Gating of O2-sensitive K + Channels of Arterial Chemoreceptor Cells and Kinetic Modifications Induced by Low PO2

Gating of O2-sensitive K + Channels of Arterial Chemoreceptor Cells and Kinetic Modifications Induced by Low PO2 MARIA DOLORES GANFORNINA and Jos~ LOPEZoBARNEO From the Departamento de Fisiologia y Biofisica, Facultad de Medicina, Universidad de Sevilla, 41009 Sevilla, Spain A B STRACT We have studied the kinetic properties of the O2-sensitive K + channels (Ko~ channels) of dissociated glomus cells from rabbit carotid bodies exposed to variable 02 tension (P02). Experiments were done using single-channel and wholecell recording techniques. The major gating properties of Ko 2 channels in excised membrane patches can be explained by a minimal kinetic scheme that includes several closed states (Co to C4), an open state (O), and two inactivated states (Io and It). At negative membrane potentials most channels are distributed between the left-most closed states (Co and Cl), but membrane depolarization displaces the equilibrium toward the open state. After opening, channels undergo reversible transitions to a short-living closed state (C4). These transitions configure a burst, which terminates by channels either returning to a closed state in the activation pathway (Ca) or entering a reversible inactivated conformation (Io). Burst duration increases with membrane depolarization. During a maintained depolarization, Ko 2 channels make several bursts before ending at a nonreversible, absorbing, inactivated state (I1). On moderate depolarizations, Ko 2 channels inactivate very often from a closed state. Exposure to low P02 reversibly induces an increase in the first latency, a decrease in the number of bursts per trace, and a higher occurrence of closed-state inactivation. The open state and the transitions to adjacent closed or inactivated states seem to be unaltered by hypoxia. Thus, at low P02 the number of channels that open in response to a depolarization decreases, and those channels that follow the activation pathway open more slowly and inactivate faster. At the macroscopic level, these changes are paralleled by a reduction in the peak current amplitude, slowing down of the activation kinetics, and acceleration of the inactivation time course. The effects of low P02 can be explained by assuming that under this condition the closed state Co is stabilized and the transitions to the absorbing inactivated state I i are favored. The fact that hypoxia modifies kinetically defined conforrnational states of the channels suggests that 02 levels determine the structure of specific domains of the Ko 2 channel molecule. These results help to understand the molecular mechanisms underlying the enhancement of the excitability of glomus cells in response to hypoxia. Address reprint requests to Dr. Jos~ L6pez-Barneo, Departamento de Fisiologia y Biofisica, Facultad de Medicina, Avenida S~nchez Pizju:in, 4, 41009 Sevilla, Spain. j. GEN. PHYSIOL. © The Rockefeller University Press • 0022-1295/92/09/0427/29 $2.00 Volume I00 September 1992 427-455 427 on September 22, 2014jgp.rupress.orgDownloaded from Published September 1, 1992 428 THE JOURNAL OF GENERAL PHYSIOLOGY • VOLUME 100 • 1992 INTRODUCTION Recent electrophysiological studies on type I (or glomus) cells of the mammalian carotid body have shown that they can generate action potentials due to the activity of voltage-gated Na +, Ca 2+, and K + channels of their plasma membrane (Duchen, Caddy, Kirby, Patterson, Ponte, and Biscoe, 1988; L6pez-Barneo, L6pez-L6pez, Urefia, and Gonz~ilez, 1988; Hescheler, Delpiano, Acker, and Pietruschka, 1989; Urefia, L6pez-L6pez, Gonz~ilez, and L6pez-Barneo, 1989), and that low environmental Oz tension (Poz) produces a selective and reversible attenuation of the K + current (L6pez-Barneo et al., 1988; Delpiano and Hescheler, 1989; L6pez-L6pez, Gonz~ilez, Urefia, and L6pez-Barneo, 1989; Stea and Nurse, 1991). These electrophysiological findings have opened new perspectives on the physiology of the carotid body and provided a deeper understanding of the mechanisms responsible for the chemosensory properties of type I cells. Inhibition of the K + current in response to low PO 2 produces an increase in the firing frequency of the cells which may lead to Ca 2+ entry and enhanced release of the transmitters that activate the afferent fibers of the sinus nerve (L6pez-L6pez et al., 1989). The present experiments were designed to address two major questions: (a) Can 02 interact with a specific type of K + channel in a manner that would account for the modulation of the macroscopic K ÷ current? If so, (b) what modifications in channel gating are induced by altering 02? In the preceding article (Ganfornina and L6pez-Barneo, 1992) we have reported the existence of three major classes of voltage-activated K + channels in glomus cells and shown that changes in PO 2 modulate the activity of only one type, referred to as the Ko 2 channel. The experimental data indicate that low PO2 decreases single-channel open probability and suggest that 02 may act through a sensor intrinsic to the membrane closely associated with the channel protein. Thus, the unique characteristics of Ko 2 channels explain the effect of altering Poz on the electrical properties of glomus cells. In this article we have aimed at a more detailed and quantitative study of the gating kinetics of Ko~ channels by the combined analysis of single-channel and macroscopic currents. Voltage-gated K + channels are extraordinarily diverse and appear to be expressed in many different cell types (for review, see Rudy, 1988), but detailed, nonstationary kinetic analysis has been done in only a few preparations (see Hoshi and Aldrich, 1988; Cooper and Shrier, 1989; Solc and Aldrich, 1990; Zagotta and Aldrich, 1990). These channels, although primarily regulated by changes in the transmembrane electric field, can be modulated by a number of ions, metabolites, and enzymes (for review, see Levitan, 1988). The interaction of modulators with K + channels and the precise changes induced in their gating properties are, however, poorly studied and far from being understood (see Perozo, Jong, and Bezanilla, 1991). Here we propose a kinetic scheme including the minimal conformational states required to explain the basic gating properties of Ko 2 channels. The model is based on sequential state diagrams of inactivating Na + and K ÷ channels (see, for example, Armstrong and Bezanilla, 1977; Armstrong and Gilly, 1979; Aldrich, Corey, and Stevens, 1983; Zagotta and Aldrich, 1990) and it provides the conceptual framework necessary for organizing and understanding the effect of Oz on channel function. Our results suggest that the effects of low PO2 can be explained by on September 22, 2014jgp.rupress.orgDownloaded from Published September 1, 1992 GANrORNINA AND LrPEZ-B~a~NEO Gating of Single Ko 2 Channels 429 stabilization of kinetically defined conformational states of the Ko 2 channels, which may indicate that changes in Po2 induce structural modifications in specific domains of the channel molecule. METHODS Culture and Recording Techniques Details of the methodology used appear in the preceding paper (Ganfornina and Lrpez- Barneo, 1992). Enzymatic dispersion of the cells and recording were performed using techniques described previously (see Urefia et al., 1989). For whole-cell and single-channel recording we used the different variants of the patch clamp technique (Hamill, Marty, Neher, Sakmann, and Sigworth, 1981). Digitized records were corrected for linear leakage and uncompensated capacity currents before analysis. Solutions, indicated as external//internal, as well as the protocol followed to test the effect of low Po2 on the properties of single-channel and whole-cell currents were the same as in the preceding paper. All experiments were conducted at room temperature (22-25°C). Analysis Macroscopic whole-cell currents and the ensemble averages obtained from original singlechannel current traces were analyzed following standard procedures. The final frequency response of the cascade of Bessel filters used (patch clamp and filter) was characterized, at different cutoff frequencies, by the rise time (Tr) of the output signal when a square wave was applied at the input. Tr was used afterwards to impose a resolution to the measurements of duration and amplitude of single-channel events. Amplitude of unitary currents was measured only for openings of duration > 2"Tr. We used a 50% threshold-crossing method to detect opening and closing transitions and a resolution of 1.3'Tr was subsequently imposed on the interval duration as indicated by Colquhoun and Sigworth (1983; see also Colquhoun, 1988). The effective cutoff frequency, the resolution imposed on the data, and the digitizing sample interval used in each experiment are indicated in the figure legends. Unless otherwise noted, no correction for unresolved events was performed. All the data used for the calculations of this paper were obtained during depolarizing pulses. In all experiments the interval between test pulses was at least 30 s to avoid cumulative inactivation. We have measured the duration of open and closed events, first latency, and burst length, as well as the number of bursts per trace. Because even with small pipettes the probability of finding a patch with only one or two Ko~ channels is very low, and, in addition, these channels recover from inactivation very slowly (which prevents the acquisition of a large number of stable recordings in a reasonable period of time), the number of events was relatively small in some of the time interval distributions studied. To minimize this limitation, open time, closed time, and burst length distributions were fitted with probability density functions (pdt) with one or the sum of several exponential terms by the method of maximum likelihood. This method of fitting a variable takes into account each one of the individual measurements and thus gives a more sensible estimate than other methods where the data are grouped into bins (Colquhoun and Sigworth, 1983). We used the conditional pdf, f(t), given that it is restricted to a range between tmin and tma~ (see Colquhoun and Hawkes, 1983; Colquhoun and Sigworth, 1983). For the graphic representation of the data, closed and open time as well as burst duration were displayed as cumulative histograms and the scaled function g(t) = 1 - F(t) (where F(t) is a distribution function integral of pdf) was superimposed on them. The number of events in each distribution is given in the figure legends. First latency distributions were displayed as a cumulative histogram and a distribution function was fitted to the data by a on September 22, 2014jgp.rupress.orgDownloaded from Published September 1, 1992 430 THE JOURNAL OF GENERAL PHYSIOLOGY • VOLUME 100 • 1992 least-squares method. The number of bursts per trace was fitted to the sum of one or more geometric terms by the maximum likelihood method. Distributions were fitted by several pdf or distribution functions with a different number of exponential terms or unknown parameters. In all instances, the extent to which a model with more parameters improved the fit was evaluated with a likelihood ratio test as described by Horn (1987). Most kinetic parameters were estimated in patches with only one channel, although occasionally first latency and open time distributions were also estimated in patches containing two channels (the number of channels in a patch was estimated as indicated in the preceding paper). In these cases, the true first latency distribution was obtained after correction for the number of channels as indicated by Patlak and Horn (1982). When the opening of two channels overlapped, we assumed that the channel that opened first closed first 50% of the times (Aldrich et al., 1983; Cooper and Shrier, 1989). In patches with more than one channel the true probability of the occurrence of records that had no openings (blank records) was taken as the Nth root of the apparent probability (Aldrich et al., 1983). When appropriate, significance of differences between mean values obtained in different experimental conditions was determined with a Student's t test for paired samples. The nonparametric Mann-Whitney U test was used to compare data distributions. The Spearman coefficient (rs) was used to test the existence of correlation between random variables. Unless indicated otherwise, the significance level (o0 of the tests was set at 0.05. RESULTS Gating of Ko 2 Channels Channels traverse several closed states during activation. Activation of the Ko 2 channels follows a sigmoidal time course with an initial delay, which suggests that they must traverse more than one closed state before opening. Fig. 1 shows the time course of the macroscopic O2-sensitive K + current (top), the ensemble average current from an outside-out patch with at least four Ko 2 channels (middle), and the cumulative first latency distribution in an inside-out patch with one Ko 2 channel (bottom). This distribution illustrates that upon step depolarization the channel opens with a variable lag (depicted in the inset to the figure) and specifies the probability that the first channel opening (ordinate) occurred at time _< t (abscissa) since the onset of the pulse. The traces reveal that the time courses of the whole-cell and the ensemble average currents are almost identical and comparable to the shape of the first latency histogram. Macroscopic and ensemble average currents reach half-maximal amplitude (ill2) in 1.1 and 1.4 ms, respectively, which are values quite similar to the time at which the first latency distribution crosses 50% of its maximum value (median first latency = 1.2 ms). In the cumulative first latency distribution probability does not reach unity because some pulses had no channel openings (blank traces). Since blank sweeps were observed even during pulses lasting up to 3 s (a time period clearly larger than the longest latency observed) it can be hypothesized, in accordance with previous work in other channel types (Horn, Patlak, and Stevens, 1981; Hoshi and Aldrich, 1988; Cooper and Shrier, 1989), that Ko 2 channels undergo direct conformational transitions from closed to inactivated states without traversing the open state, or that they are inactivated with ~ 20% probability before the pulse. Transitions to the open state are voltage dependent. The number of closed states leading to the open state and the value of the rate constants characterizing transitions among states were estimated in two steps. First, we evaluated the number on September 22, 2014jgp.rupress.orgDownloaded from Published September 1, 1992 GANFORNINA AND LOPEZ-BARNEO Gating of Single Ko 2 Channels 431 of states in the activation pathway by fitting a distribution function with several exponential terms to cumulative first latency histograms obtained at various membrane potentials. For simplicity the fitting was done on the basis of several assumptions summarized in Scheme 1. k 3= 2= = Co ' "0 SCHEME 1 It is proposed that transitions from state C1 to the open state (O) come about by the independent movement of three identical charges. This constrains rate constants 0.8 _Q o 0.4 n o 0.. 0 whole-cell ~~~I0 pA first latency J i ~ i i i i 0 1 2 3 4 5 t (ms) FIGURE 1. Activation time course of the macroscopic O~-sensitive K + current and Ko 2 channels. Macroscopic K ÷ current (top), ensemble average current (20 pulses) from an outsideout patch with at least four K% channels (m/dd/e), and cumulative first latency distribution from an inside-out patch with N = 1 K% channel (bottom). In all cases pulses were applied from -80 to +20 mV and time was measured from the onset of the pulse (arrows). Solutions: standard Na, qTX//standard K, 10 EGTA (top and middle) and standard Na, TFX//130 K, 10 EGTA (bottom). for transitions to be integer multiples of the voltage-dependent first order transition rate et, and therefore reduces the number of parameters that must be determined. We also supposed that at positive membrane potentials the deactivation rate constant (13) can be neglected and that at resting all the channels are in state Co. Thus, the Co to Cl transition is characterized by an unconstrained first order rate constant (k) which was assumed to be independent of voltage. This model fitted the data significantly better than the ones with fewer states, as well as models where the rate constant k was made voltage dependent and a multiple of ~t (4a) or all the rate constants were unconstrained. Moreover, it provided a simpler explanation for the effects of low Po2. As shown in another section, hypoxia slows down the activation kinetics of Ko 2 channels without altering the transitions adjacent to the open state. Hence, it was conceived as likely that at negative membrane potentials Ko 2 channels are distributed on September 22, 2014jgp.rupress.orgDownloaded from Published September 1, 1992 432 THE JOURNAL OF GENERAL PHYSIOLOGY - VOLUME 100 • 1992 between states Co and C] and that transitions between these states are regulated by Po 2 . A final estimate of the rate constants in the opening pathway was done by fitting to cumulative first latency histograms the numerical solutions of the state diagram represented in Scheme 2. k 1 3,', 2= " Co-- '-C1., "C2- k ,8 2.8 3,8 -i SCHEME 2 The voltage-independent rate constants hi and k_ 1 were chosen such as to determine that at resting and under normoxic conditions 97.5% of the channels are in state C1, and the value of the rate constant 0t calculated with the constraint that 13, estimated from the fitting to the voltage dependence of the burst duration (see Fig. 4 C, below), was set to 32 s -t at 0 mV. Fig. 2 A shows first latency histograms obtained during depolarizations to various membrane potentials. The fittings obtained following Scheme 2 scaled by the number of blanks observed at each membrane potential are superimposed. The data illustrate the shortening of the median first latency with depolarization as a result of the opposite voltage dependence of the rate constants ot and 13. The values of et (filled symbols) and 13 (open symbols) estimated at different membrane potentials are plotted in Fig. 2 B. The solid lines follow the exponential equations: ee(V) = A,'e(V/V~) (1) 13(V) = Ap'e-(V/V~) (2) where Aa andAp are the values of~t and 13 at 0 mV (201 and 32 s -l, respectively), and Va (22 mV) and V~ (21 mV) are slope constants specifying the membrane potential displacement required for an e-fold increase or reduction of a and 13, respectively. Matching of the model to experimental findings is further shown in Fig. 2 C, where it is illustrated that ensemble averages of single K% channel currents recorded at various membrane potentials have a rising phase comparable to the theoretical fits plotted in Fig. 2 A (compare also traces in Fig. 1, B and C). In these records half-time to peak current also varies exponentially with the membrane potential with a slope constant of - 12 mV (Fig. 2 D). Closed interval distribution. During a step depolarization, Ko 2 channels exhibit, after the first opening, several open/closed transitions, with distinct closed intervals separating consecutive openings. These intervals, tentatively grouped into three major categories, are illustrated in Fig. 3 A by representative recordings obtained from a patch with one channel. Open periods were interrupted by fast transitions to a nonconducting state with a lifetime of only a few hundred microseconds (single asterisks). The occurrence of these brief closed events was surely underestimated because most transitions could have been partially, or totally, filtered due to the limited recording bandwidth. Fast open/closed transitions were defined as within a "burst" following a burst criterion duration (Magleby and Pallota, 1983). We set the on September 22, 2014jgp.rupress.orgDownloaded from Published September 1, 1992 GANFORNINA AND L6PEZ-BARNEO Gating of Single Koz Channels 433 burst criterion at 2.5 ms (five times the fastest time constant of the closed interval distribution; Colquhoun, 1988), which means that longer closed durations were considered as interburst intervals. Interburst closed intervals of a lifetime between ~ 3 and 10 ms (double asterisks) could be distinguished from intervals lasting several A B 1.0 ~ = 1400 0.8 FE_ 0.6 O 0.4 0_ 0.2 0.0 C 4 3 < ~ 2 v m 1050 / - I ,5 7o0 350 f T ~ I I 0 0 40 50 50 D i i i 10 20 30 t (ms) i i i i i 10 20 30 40 50 t (ms) E v _i i 0 50 0 i -25 0 25 Vm (mV) 2O 15 I0 5 0 -25 0 25 Vm (mV) 90 60 7 30 ~" FIGURE 2. Voltage dependence of Ko 2 channel activation. (A) Cumulative first latency histograms obtained at -20 (~7), 0 (V), +20 (O), and +40 (Q) mV in an inside-out patch. Correction for the number of channels in the patch (N = 2) was performed as indicated in Methods. Solid lines drawn over the data points are the fitted functions following the model of Scheme 2 scaled to the number of blanks at each membrane potential. (B) Estimated values of a (filled circles) and [3 (open circles) obtained from fittings to first latency at various membrane potentials. The solid lines are the solutions of Eqs. 1 and 2 in the text. (C) Ensemble average currents obtained in an outside-out patch with at least three Ko~ channels in response to pulses from -80 to -20, -10, 0, +20, and +40 mV (from bottom to top). (D) Half-time to peak (hi2) from averages in C vs. membrane voltage. Data are fitted to an exponential function with a slope factor of -12 mV. Effective cutoff frequency = 0.95 kHz and sampling interval = 250 I~s. Solutions: standard Na, "ITX//130 K, 10 EGTA. tens to hundreds of milliseconds (triple asterisks) or even longer (see also Fig. 5 A). The intraburst closed events and the short-lasting interburst intervals have durations smaller than the first latency and are, in principle, compatible with transitions between the open and closed states of the activation pathway. However, the duration on September 22, 2014jgp.rupress.orgDownloaded from Published September 1, 1992 434 A THE JOURNAL OF GENERAL PHYSIOLOGY • VOLUME 100 • 1992 0 C - 0 - 0 - 12pA J B 1.0 20 ms C 1.0 0.8 ".---- 0.6 U 0.4 0.2 0.8 = 0.6 JO 0 0.4 o_ 0.2 0 5 0 0 25 50 75 1 O0 close time (ms) i i i 0 2 3 4 close time (ms) FIGURE 3. Closed interval distribution. (A) Representative single-channel recordings from an inside-out patch with N = 1 Ko~ channel during depolarizations from -80 to +20 inV. Examples of intraburst (*) and short- (**) and long- (***) lasting interburst intervals are illustrated. (B) Cumulative histogram of closed intervals during 400-ms pulses to +20 inV. A distribution function with three exponential components is superimposed on the data. The estimated time constants and weighting factors are: ~ = 506 p~s (al = 0.61), ~ = 5.4 ms (a~ = 0.1), and ~3 = 58.4 ms (a3 = 0.28). Number of events = 164. (C) Cumulative distribution of closed events shorter than or equal to 3 ms. A single exponential function with "r = 0.46 ms is superimposed on the histogram. Number of events = 82. In the cumulative distributions presented in this and following figures the fitted function specifies the probability that an interval duration was longer than the time on the abscissa. Sampling interval = 250 I~s and imposed time resolution = 468 ~s. Solutions: standard Na, TTX//130 K, 10 EGTA. of the long-lasting interburst intervals is clearly longer than any of the first latency values observed at this membrane potential; hence, it can be assumed that they represent transitions to a reversible inactivated state (Iverson and Rudy, 1990; Zagotta and Aldrich, 1990). As indicated in Methods, a precise quantitative analysis of closed intervals was hampered by the difficulty in obtaining patches with one channel and by the fact that on September 22, 2014jgp.rupress.orgDownloaded from Published September 1, 1992 GANFORNINA AND LOPEZ-BARNEO Gating of Single Ko 2 Channels 435 at -80 mV Ko~ channels require a resting period of at least 30--40 s to recover completely from inactivation, which prevents rapid pulsing. Due to these limitations the parameters estimated must be taken only as approximate although they have the value that they come from measurements done in patches (n = 3) containing one Ko~ channel where we could compare in each case the distributions at normal and low Po 2. Under normoxic conditions, closed interval distributions were consistently best fitted by a three-exponential function with average time constants of 0.5, 4.3, and 180 ms at +20 mV (Fig. 3 B and Table I). The weight of each component of the distribution was also consistent in the three patches studied. These data further TABLE I Kinetic Parameters of Ko 2 Channels in Control and Hypoxic Solutions Control Hypoxia Median first latency (ms) (6) 2.15 ± 1.31 4.18 ± 1.42" Closed time distribution (3) "rl (ms) 0.5 ± 0.13 0.47 ± 0.13 Amplitude 0.68 -+ 0.16 0.77 ± 0.1 "r2 (ms) 4.32 -+ 2.7 3.84 ± 2.4 Amplitude 0.17 ± 0.! 0.13 ± 0.02 "r3 (ms) 180 ± 194 217 ± 289 Amplitude 0.15 ± 0.1 0.1 ± 0.09 Open time distribution To (ms) (3) 35.4 ± 21 34 ± 18 "to (ms) (1) 8.93 8.98 Burst length distribution (3) % (ms) 46.5 ± 6 47.6 ± 10.7 Number of bursts per trace distribution (p) (3) 0.41 ± 0.07 0.55 +- 0.06** Probability of blanks (5) 0.18 ± 0.02 0.47 ± 0.08* Values are given by mean - SD and the number of patches is given in parentheses. "to values are from an experiment after correction for unresolved events. In the number of bursts per trace distribution, p is the probability for the transition from state I0 to I 1 (see Scheme 4). All measurements were done during pulses to +20 inV. *Statistically significant difference with respect to control values (paired t test, ~x < 0.01). **Significance level et = 0.116 (see text for further details). suggest that there are at least three kinetically different closed states from which Koe channels can reopen. The fast intraburst closed intervals can be explained by adding to Scheme 2 a short-lived closed state (C4) from where channels can flicker to the open conformation. In agreement with this idea, the distribution of the closed events < 3 ms, shown expanded in Fig. 3 C, was best fitted by an exponential function with a time constant of 0.46 ms. Intraburst open time distributions were fitted too by a single exponential function that after correction for missed events yielded a mean open time value of 8.9 ms at +20 mV (see below). The three possible kinetic diagrams compatible with the on September 22, 2014jgp.rupress.orgDownloaded from Published September 1, 1992 442 THE JOURNAL OF GENERAL PHYSIOLOGY • VOLUME 100 • 1992 TABLE III Inactivation Time Course of Ko2 Channels Control Hypoxia Fast component "c (ms) 32.6 -+ 13 47.4 -+ 12 Amplitude (normalized) 0.47 -+ 0.01 0.9 -+ 0.05* Slow component (ms) 267 -+ 39 351 -+ 139 Amplitude (normalized) 0.52 -+ 0.01 0.09 -+ 0.05* Values are given by the mean - SD (n = 3 patches). Data were obtained by doubleexponential fit to ensemble averages. Pulse duration = 3 s. Membrane potential = +20 mV. *Statistically significant differences with respect to the control values. effect, low Po2 slows down the activation time course of the K + current, as illustrated by the scaled traces in Fig. 7 B. Activation kinetics of the O2-sensitive K + current are quantified in Fig. 8 A by the time to reach half-maximal current amplitude (tl/2) in control and under hypoxic conditions at various membrane potentials. At + 20 mV, for example, low Po2 reduces by almost 50% the speed of activation. These effects of hypoxia were very reproducible and qualitatively similar in the 79 type I cells tested so far. The results of Fig. 7 predict, on the basis of a sequential model of gating, that low Po2 would retard Ko 2 channel first opening, as is in fact shown in Fig. 8 B, where cumulative first latency distributions at +20 mV obtained in normoxic and hypoxic solutions are compared. The average median first latency values calculated in six patches are given in Table I. In good agreement with the effect on the macroscopic current, at +20 mV hypoxia increases by almost a factor of two the median first latency of Ko 2 channels. The increase of first latency induced by low Po2 could be due to stabilization of any, or several, of the closed states existing in the activation A <;2S ............. H +40 H +30 N +2O B 0.5 nA 5 ms FIGURE 7. Effect of low Po2 on the rising phase of the macroscopic O2-sensitive K + current. (A) Current generated upon step depolarizations from -80 mV to the indicated membrane potentials in control (C: Po2 = 150 mmHg) and hypoxic (H: Po2 ~ 5-10 mmHg) solutions. Recovery (R) after returning to control solution is shown at +40 mV. (B) The same traces as in A are shown scaled to compare their time course. Effective cutoff frequency = 10 kHz and sampling interval = 100 Ixs. Solutions: 80 K, TI'X//standard K, 10 EGTA. on September 22, 2014jgp.rupress.orgDownloaded from Published September 1, 1992 GANFORNINA AND LOPEZ-BARNEO Gating of Single Ko 2 Channels 443 pathway. The simpler explanation is that hypoxia specifically influences the transition between states Co and Cz. The solid lines in Fig. 8 B are the theoretical fits to first latency histograms obtained from a patch exposed to normoxic and hypoxic solutions following the kinetic diagram of Scheme 2. In this and in two more patches good fittings were obtained in the two experimental conditions without altering the values of~x and 13 and assuming that the modifications in the rate constants ki and k_ 1 upon exposure to extreme hypoxia (Po 2 --~ 5-10 mmHg) determine that the percentage of channels in state Co changes from 2.5 to 73.3% (see Fig. 13). Thus, at -80 mV (a potential at which most of the channels are in state C1) hypoxia shifts the equilibrium further to the left and favors the residence of the channels in the closed A B 5 E 4 3 2 6 1.0 0.8 .~ 0.6 O o 0.4 [:1_ 0.2 i I I 0 0 30 50 70 0 Vm (mV) v n u ////n I 5 10 15 180200 time (ms) FIGURE 8. Effect of hypoxia on the activation time course of the macroscopic O2-sensitive K ÷ current and Ko~ channels. (A) Half-time to peak of the macroscopic K + current (tn/2, ordinate) plotted vs. step voltage (Vm, abscissa) in control (filled circles) and hypoxic (open circles) conditions. Data were obtained from the experiment shown in Fig. 7. The open triangle indicates the recovery after returning to the control solution. (B) Cumulative first latency distributions obtained in an inside-out patch with N = 1 Ko 2 channel in control (filled circles, Po2 = 150 mmHg) and hypoxic (open circles, Po~ ~ 5-10 mmHg) conditions. Pulses from -80 to +20 mV were delivered every 30 s while alternating exposure to control and hypoxic solutions was applied. Data are fitted by the model of Scheme 2 using the rate constant values indicated in Fig. 13A. Effective cutoff frequency = 0.95 kHz; sampling interval = 250 Ixs. Solutions: standard Na, TIX//130 K, 10 EGTA. state Co. This could explain why low Poe increases the activation latency without affecting cx and its more pronounced effect at moderate depolarizations, since it can be expected that the relative change of kl and k-n will have a larger influence on activation at potentials at which cx is not too large. Transitions near the open state are unaffected by hypoxia. It was shown before that after the first opening Ko 2 channels undergo reversible transitions to adjacent nonconducting states (Schemes 3 and 4). Fig. 9A shows the cumulative closed interval histogram obtained from a channel exposed to hypoxia and superimposed (solid line) a three-exponential distribution function obtained by fitting the experimental values. The fit obtained in the same patch with normal Po 2 and at the same on September 22, 2014jgp.rupress.orgDownloaded from Published September 1, 1992 444 THE JOURNAL OF GENERAL PHYSIOLOGY • VOLUME 100 • 1992 A 1.0 B 1.0 0.8 .=-E 0.6 .a o o.4 iX. 0.2 C 1.0 0.8 .~ 0.6 J3 J3 2 0.4 13- 0.2 0 0 25 50 75 100 125 close time (ms) t 0 50 100 open time (ms) I 150 D 0.8 .'--E 0.6 o 0.4 &.. 0.2 X3 P fl_ 1.0 0.8 0.6 0.4 0.2 0 2 3 4 close time (ms) ! 0 50 100 150 200 burst length (ms) FIGURE 9. Closed-time, open-time, and burst length distributions at low Po2. (A) Cumulative histogram of closed-time intervals from records of the same experiment shown in Fig. 3 obtained while the patch was bathed in a hypoxic solution (Po2 ~ 5-10 mmHg). The continuous line is the distribution function fitted to the data (% -- 0.44 ms, al = 0.81; "2 = 4.3 ms, a~ = 0.14; % = 46 ms, a3 = 0.05; number of events = 81). The fitting obtained under normoxic conditions is also shown (dashed line) for comparison (see Fig. 3 B). These distributions were not statistically different (U test, tx > 0.05). (B) Cumulative histogram of closed events < 3 ms (see Fig. 3 C). The single exponential functions obtained in hypoxic (continuous line, "r = 0.55 ms) and control (dashed line, "r = 0.46 ms) conditions are superimposed on the histogram. Number of events = 66. Data were obtained from an inside-out patch with N = 1 K% channel. Sampling interval = 250 Ws; imposed time resolution = 468 V.s. Solutions: standard Na, TTX//130 K, 10 EGTA. (C and D) Open-time (C) and burst length (D) cumulative distributions in hypoxia during 400-ms pulses from -80 to +20 mV. Number of events = 61 and 38, respectively. The scaled single-exponential distribution functions are shown superimposed to the histograms. The fittings obtained under normoxic conditions (dashed lines) are also shown for comparison. Data were obtained from an inside-out patch with N = 1 K% channel. Sampling interval = 250 p.s; imposed time resolution = 468 V.s. Solutions: standard Na, TIX//130 K, 10 EGTA. membrane potential (+20 mV; see Fig. 3B) is also shown (discontinuous line). Average values of the estimated parameters (time constant and relative amplitude of each component) obtained from the fits to closed interval distributions from several patches (Table I) indicate that they are not significantly modified by low Po2 (paired on September 22, 2014jgp.rupress.orgDownloaded from Published September 1, 1992 GANFORNINA AND L6PEz-BARNEO Gating of Single Ko 2 Channels 445 t test). In agreement with these results, the cumulative histogram of intraburst closed intervals measured during exposure to low Poe, plotted separately in Fig. 9 B, is fitted by a single exponential function (solid line; "r = 0.51 ms) that does not appreciably differ from the one estimated under normoxic conditions (discontinuous line, ~" = 0.46 ms; see also Fig. 3 C). Thus, it can be concluded that the rate constants that determine the exit of the channels from the reversible nonconducting states (either a closed state or the reversible inactivated state) are unaffected by hypoxia. Fig. 9 also illustrates the open time (C) and burst duration (D) distributions of Ko 2 channels in low Po2. In both cases the solid lines are the single exponential functions that fitted the data and, to facilitate comparison, the discontinuous lines are the functions obtained at normal Po2. Average values of the mean open time (%) at +20 mV from five patches studied in the two experimental conditions are given in Table I, which also includes estimates from an experiment after correction for missed events 0") as indicated by Colquhoun (1988). Since no statistically significant difference was found (U test) between the distributions obtained in the two experimental conditions, it should be concluded that low Po2 does not alter the mean open time of Ko 2 channels. Further support for this idea is provided by the fact that the burst length distribution, a variable that can be measured without the restriction imposed by the time resolution, has a similar single exponential shape in control and under hypoxic conditions (Fig. 9 D; Table I). These distributions were not significantly different (U test), indicating that the rate constants hi and 313, which determine burst length (i.e., the exit of the channels from flickering between states C4 and O; see Scheme 3), are not modified by low Po2. The experiments described in the previous paragraphs show that at +20 mV transitions between the open state and any of the adjacent states are not under the influence of 02 tension. We also tested whether at negative potentials closing of Ko 2 channels is similar in control and hypoxic conditions. These experiments were performed by recording macroscopic K + tail currents in cells bathed in high K + solutions. Channels were opened by 20-ms step depolarizations to +20 mV and at the instant of repolarization to -80 mV (indicated by the arrows in Fig. 10) we recorded tail currents whose time course reflect the kinetics of the transition to nonconducting states of the channels open at the end of the pulse. At -80 mV, the tails probably represent deactivation of the channels (returning to closed states in the activation pathway), although they may also have a small component due to channel inactivation. In all cells studied (n = 4) tail currents were fitted by the sum of two exponential components, drawn superimposed on the current traces in the example of Fig. 10, A and B. Low Po~ produced a perfectly reversible attenuation in the amplitude of the two components of the tails, reflecting the decrease in the number of channels open during the pulse, but their respective time constant values were unchanged (Table II). Low Po2 favors closed-state inactivation. Besides the lengthening of the activation latency, one of the most striking effects of hypoxia is a marked increase in the appearance of blank records. The average values of blank record probability at +20 mV in five patches exposed to normoxic and low Po2 solutions are given in Table I. Hypoxia increased blank probability by a factor of almost 2.5, a relative change comparable to that induced by low Po~ on the activation kinetics of the macroscopic on September 22, 2014jgp.rupress.orgDownloaded from Published September 1, 1992 446 THE JOURNAL OF GENERAL PHYSIOLOGY " VOLUME 100 • 1992 O2-sensitive K + current and the median first latency of Ko~ channels at the same membrane potential. These results suggest that low Po2 highly favors the direct transition of Ko 2 channels from closed to inactivated states. Since in our model closed to inactivated transitions occur from state C1 and the voltage-dependent forward rate constant in the activation pathway (~t) is not affected by hypoxia, the higher occurrence of blank traces leads us to suggest that low Po 2 increases the rate constant 8, perhaps due to a further stabilization of state I1 or to a decrease in the activation energy of the transitions leading to the absorbing inactivated state. Low Po2 favors the transitions to the absorbing inactivated state. The effect of low Po2 on the inactivation time course of Ko 2 channels is illustrated in Fig. 11 by ensemble average currents obtained from an outside-out patch containing two channels activated by 3-s pulses to +20 mV. The control record (the darker trace in Fig. 11 A) shows that in ~ 0.5-1 s Ko 2 channels inactivate almost completely (see also Fig. 5, A A Control B Hypoxia 130 K+// 130 K + I O0 ms 1 nA FIGURE 10. Effects of hypoxia on the deactivation kinetics of the O2-sensitive K ÷ current. Tail currents elicited upon repolarization to -80 mV after a 20-ms pulse to +20 mV with a Po 2 of 150 mmHg (A, control and recovery) and a Po2 < 5 mmHg (B). The instant of repolarization is indicated by the arrows. Note the smaller amplitude of the current at the end of the pulse in hypoxia compared with control. Tail currents are fitted by a double-exponential function whose estimated parameters are shown in Table II. The first two sample points of the tails were not included in the fitting, Effective cutoff frequency = 10 kHz; sampling interval = 500 V,s. Solutions: 130 K, EGTA//130 K, 10 EGTA. and B). Exposure to hypoxia produces a decrease in the peak amplitude of the average current and accelerates inactivation, which in this condition is nearly complete in < 500 ms (Fig. 11 B). This effect is reversible, as shown by the recovery trace (the lighter trace in Fig. 11 A), and was consistently observed in all patches studied (n = 29). In the two experimental conditions, the inactivation time course was fitted by the sum of two (fast and slow) exponential functions and the mean values of the time constants as well as the weight of each component estimated in several patches are given in Table III. Low Po2 produces a drastic reduction of the slow component of inactivation, leaving a fast component with an almost unaltered time course. Since it was shown before that hypoxia does not modify the rate constants that characterize the transitions between the open and the reversible inactivated states (hi and h-l in Scheme 4), we interpret these results as indicating that low Po2 favors the transition from state I0 to Ii. This interpretation also predicts on September 22, 2014jgp.rupress.orgDownloaded from Published September 1, 1992 GANFORNINA AND LOI'Ez-BARNEO Gating of Single Ko z Channels 447 that low Po~ should disfavor reopening of Ko 2 channels from state I0 and therefore decrease the number of bursts per trace. This discontinuous variable is expected to follow a geometric distribution (Colquhoun and Sigworth, 1983), with the unknown parameter (p) being the probability of entering the absorbing inactivated state I1. The number of bursts per trace distribution was well fitted to a single geometric distribution function in three patches with one Ko~ channel. The estimated average values of p were 0.41 in control and 0.55 in hypoxic solutions (Fable I). The difference between these two values, although varying in the direction of the predictions of the model, was not statistically significant (et = 0.116). Nonetheless, it must be noticed that the number of bursts per trace and the difference ofp in the two experimental conditions have surely been underestimated, since they were calculated from data obtained during pulses of 200 or 400 ms, a duration too short for allowing complete inactivation of the channels. A B l~ Contro v "~,~ A~,~. ~.~ Hypoxio I pA 500 ms FIGURE 11. Modification of the inactivation time course of Ko~ channels by hypoxia. (A) Ensemble average of records obtained during 3-s pulses to +20 in an outside-out patch with N = 2 Ko~ channels exposed to a normoxic solution. The lighter trace shows the average obtained after recovery from a 4-min exposure to low Po2. (B) Ensemble average of singlechannel records obtained during exposure to hypoxia. Smooth lines in A and B are the fittings to double-expo- nential functions whose estimated parameters are shown in Table III. Effective cutoff frequency = 1 kHz; sampling interval = 3.9 ms. Pulses were delivered with a minimum interval of 45 s. Solutions: standard Na, TTX//standard K, 10 EGTA. The modifications induced by hypoxia on the inactivation time course of Ko 2 channels are summarized in Fig. 12. The two panels at the right of the figure show representative single-channel currents from an inside-out patch with one Ko 2 channel exposed to control (B) and hypoxic (D) solutions. The respective ensemble average currents are plotted in A and C, illustrating that the probability of channel opening decreases in low Po2. Superimposed on the ensemble averages we have plotted the convolutions of the open time and first latency distributions (continuous smooth lines) and of burst duration and first latency distributions (discontinuous smooth lines) estimated in the two experimental conditions. These theoretical functions predict the time course of ensemble averages of inactivating channels that either open once (continuous line) or make a single burst (discontinuous line) per pulse on September 22, 2014jgp.rupress.orgDownloaded from Published September 1, 1992 448 THE JOURNAL OF GENERAL PHYSIOLOGY • VOLUME 100 • 1992 (Aldrich et al., 1983; Kirsch and Brown, 1989). Fig. 12A illustrates that the time course of Ko 2 channel open probability clearly deviates from any of these predictions because they normally make several bursts during a pulse, which slows down their inactivation time course. The deviation from the theoretical functions is, however, markedly reduced by low Po2, indicating that under this experimental condition A B 0.8 0.6 °_ 0.4 o .o o o_ 0.2 Control 0 50 100 150 200 t C 0.8 0.6 -~ 0.4 o o o_ 0.2 Hypoxia 0 50 100 150 200 time (ms) D 50 ms J2 pA FIGURE 12. Effect of low PO2 on Ko 2 channel inactivation. (B and D) Representative recordings illustrating the activity of a Ko~ channel exposed to normoxic and hypoxic solutions, respectively. (A and C) Ensemble averages from a set of single-channel records obtained from the same inside-out patch (N = 1 Ko~ channel) exposed to control and hypoxic solutions. Ensemble averages are normalized by the single-channel amplitude. In both panels the continuous line is the convolution of the open-time and first latency distributions and the dashed line is the convolution of the burst length and first latency distributions calculated from the probability density functions estimated in the two experimental conditions. Effective cutoff frequency = 0.95 kHz and sampling interval = 500 p.s. All measurements were done in traces obtained by 200-ms depolarizations to +20 mV delivered every 30 s. Solutions: standard Na, TrX//130 K, 10 EGTA. reopening of Ko 2 channels is less likely and, as a consequence, their inactivation time course follows more closely the predictions of a model with a single burst per trace (discontinuous line). A model for the effect of hypoxia on Ko2 channel gating. Low Po2 modifies specific transitions and states in Ko 2 channels which can be simply summarized by assuming on September 22, 2014jgp.rupress.orgDownloaded from Published September 1, 1992 GANFORNINA AND I.~PEZ-BARNEO Gating of Single Ko2 Channels 449 that state Co is stabilized and transitions to state I1 favored. The state diagram of Ko 2 channel gating proposed in Scheme 4 is shown in Fig. 13 A with the values (in seconds -1) estimated for the different rate constants at +20 mV under normoxic conditions (Po~ = 150 mmHg). The values at low Po~ (~5-10 mmHg) of the rate constants altered upon exposure to hypoxia are given in parentheses. As explained in the previous sections, the different variables were estimated from the fits to the A (2000) (g) 3900 1950 1300 650 111 200 2.8 Co ;--"C ~ =----~- C 2 ~ " C 3-:-~ O~C ,~-"I o ---~I 1 1 O0 14 28 42 2200 5 (5500) I 3es 0ooo) B C 0.8 0.8 0.6 °~ o 0.4 t~ 8 13_ 0.2 0.0 D 0.8 >, 0.6 J3 o 0.4 t'~ 8 13- 0.2 0.0 0.6 0.4 ,.0 8 Q- 0.2 0.0 200 400 600 800 lOOO t (ms) E 0.8 .'''''--- ~ 0.6 o 0.4 .O 8 a. 0.2 0.0 0 2 4 6 8 10 i i i t t 0 200 400 600 800 1000 t (ms) i I I i I 0 200 400 600 800 1000 t (~) t (~) FIGURE 13. Kinetic model of Ko~ channel gating and modifications induced by low Po~. (A) State diagram of Ko~ channel gating (Scheme 4 in the text) with the numerical values of the rate constants at +20 mV in normoxic (Po2 = 150 mmHg) and hypoxic (Po~ = 5 mmHg) conditions. In parentheses are the values of the rate constants altered upon exposure to low Po 2. (B and C) Comparison of the time course of the ensemble average current of K% channels (control record in Fig. 11 A ) and the solution of the model under normoxic conditions and at +20 mV. (D and E) Predictions of the model under control (C) and hypoxic (H) conditions. In D, trace H has been scaled (discontinuous line) to the same amplitude as trace C to illustrate the deceleration of the activation time course at low Po 2. on September 22, 2014jgp.rupress.orgDownloaded from Published September 1, 1992 450 THE JOURNAL OF GENERAL PHYSIOLOGY • VOLUME 100 • 1992 first latency and to the open and closed interval distributions. The rate constants h2 and 8, specifying the irreversible transitions to the absorbing inactivated state, were calculated from: and pI0-I1 = h2/(h2 + h-l) (8) Pb = 8/(3a + k-1 + 8) (9) pI0 - Ii (Eq. 8) is the probability of entering the absorbing state estimated from fittings of the number of burst per trace distributions. This variable was increased by 15-25% to correct for the underestimation of the number of bursts since, as indicated above, it was calculated from pulses lasting less than the time required for complete inactivation of Ko 2 channels. We preferred to assume that closed-state inactivation begins at state C1, rather than at state C2 or C3, because if this mode of inactivation occurs from a state near the open state, the increase of 8 on exposure to low PO 2 produces by mass action an acceleration of the activation kinetics incompatible with the experimental findings. Given this assumption, the value of 8 was estimated from Eq. 9, where Pb is the probability of blanks under normoxic conditions. At low Po2, we chose a value for 8 such as to determine the same probability of blank traces as experimentally observed. Our model quantitatively reproduces the activation and inactivation kinetics of the macroscopic O2-sensitive K + current, as well as the peak open probability of Ko 2 channels at various membrane potentials. In Fig. 13, B and C, we show, for comparison, the ensemble average current from a patch with one Ko 2 channel at +20 mV (B) and the computer solution of the model at the same membrane potential (C). Adequate matching of the model to the experimental data can be further evaluated by its comparison with recordings of whole-cell K + currents (see, for example, the trace at +20 mV in Fig. 5 C). In Fig. 13, D and E, computer solutions of the model are shown under control (C) and low Po2 (H) conditions, illustrating that its predictions (decrease in channel open probability, retardation of activation kinetics, and acceleration of inactivation time course) are in good agreement with the experimental results. DISCUSSION In this article we show that the major gating properties of K% channels can be explained on the basis of a minimal kinetic scheme and that low PO2, rather than a general alteration of this scheme, reversibly modifies specific states and transitions. These findings explain the decrease in Ko~ channel open probability under exposure to low Po2 and help to understand the role of Ko 2 channels in regulating the excitability of glomus cells under normoxic and hypoxic conditions. Kinetic Properties of K% Channels The kinetic model proposed for Ko~ channel gating is based on sequential schemes of other channels (see, for example, Armstrong and Bezanilla, 1977) and resembles in several ways the model developed by Zagotta and Aldrich (1990) for Shaker K + on September 22, 2014jgp.rupress.orgDownloaded from Published September 1, 1992 GANFORNINA AND I~PEZ-BARNEO Gating of Single Ko 2 Channels 451 channels of Drosophila myotubes, albeit there are some qualitative and quantitative differences. The major features of the gating scheme are as follows: (a) Resting channels must traverse several closed states before opening. (b) Ko~ channels exhibit a single open state from which they flicker to a short-lived closed state. These fast transitions configurate a burst whose duration increases with membrane depolarization. (c) Termination of a burst occurs either by returning to a dosed state in the activation pathway or by entering a reversible inactivated state. (d) During a maintained depolarization, channels normally make several bursts before they inactivate. Inactivation usually proceeds through the reversible inactivated state before entering an absorbing state. (e) Channels can directly inactivate, before opening, from a closed state of the activation pathway. Fits to first latency histograms indicate that Ko 2 channels traverse at least three closed states (Cl to C3) before opening. In our model, distribution of the channels between the left-most closed state (Co) and state CI is assumed to be dependent on 02 tension. Once open, Ko 2 channels display several fast transitions to a nonconducting state of short lifetime. This indicates that, as in other K ÷ channels (Magleby and Pallotta, 1983; Zagotta and Aldrich, 1990), the open state is relatively unstable and that in the process of ion conduction channels intermittently reside in a short-lived closed state (C4 in Scheme 3). In a number of voltage-gated K ÷ channels (Zagotta and Aldrich, 1990; Perozo et al., 1991) it has been shown that the mean open time is independent of the membrane potential. However, it has also been reported that in some mammalian K + channels this parameter increases with depolarization (Hoshi and Aldrich, 1988; Cooper and Shrier, 1989), a result also observed in our analysis of Ko~ channel gating. If channels open intermittently, in the form of bursts, the open time is a variable whose estimation is, in many cases, subjected to a high degree of uncertainty, because many fast closed events are missed due to the limited recording bandwidth necessary to resolve single-channel currents. We preferred to study the voltage dependence of the burst duration, a parameter that can be measured with more accuracy, and found that in Ko~ channels it increases with depolarization. This is a distinct property that differentiates K + channels of glomus cells from native Shaker K + channels (Zagotta and Aldrich, 1990), and strongly suggests that after the first opening Ko 2 channels can return to a closed state in the activation pathway. Thus, bursts terminate by channels entering either an inactivated state or a closed state, and their change in duration with voltage comes from the voltage dependence of the deactivation rate constant lB. In our kinetic scheme the short-lived intraburst closed state is placed after the open state, though several alternatives are possible (Scheme 3), because this provides a better quantitative fit to the data and also gives a straightforward explanation for the voltage dependence of channel closing on repolarization in voltage-dependent K + channels. Another characteristic feature of Ko~ channel gating is that they normally make several bursts during a pulse, which is interpreted to mean that the closed or inactivated states occupied after termination of the first burst are very reversible. In this respect, Ko 2 channels differ also from native Skaker K + channels that exhibit one or two bursts per trace (Zagotta and Aldrich, 1990). Reopening, in the sense that it reflects returning to the open state from either a closed or an inactivated state, may on September 22, 2014jgp.rupress.orgDownloaded from Published September 1, 1992