structure and function of K2P-channels - Uni Marburg

Pflugers Arch - Eur J Physiol (2015) 467:867–894 DOI 10.1007/s00424-015-1703-7

INVITED REVIEW

Much more than a leak: structure and function of K2P-channels Vijay Renigunta 1 & Günter Schlichthörl 1 & Jürgen Daut 1

Received: 6 March 2015 / Accepted: 9 March 2015 / Published online: 21 March 2015 # Springer-Verlag Berlin Heidelberg 2015

Abstract Over the last decade, we have seen an enormous increase in the number of experimental studies on two-poredomain potassium channels (K2P-channels). The collection of reviews and original articles compiled for this special issue of Pflügers Archiv aims to give an up-to-date summary of what is known about the physiology and pathophysiology of K2Pchannels. This introductory overview briefly describes the structure of K2P-channels and their function in different organs. Its main aim is to provide some background information for the 19 reviews and original articles of this special issue of Pflügers Archiv. It is not intended to be a comprehensive review; instead, this introductory overview focuses on some unresolved questions and controversial issues, such as: Do K2P-channels display voltage-dependent gating? Do K2Pchannels contribute to the generation of action potentials? What is the functional role of alternative translation initiation? Do K2P-channels have one or two or more gates? We come to the conclusion that we are just beginning to understand the extremely complex regulation of these fascinating channels, which are often inadequately described as ‘leak channels’.

Keywords K2P-channels . Action potential . Voltage dependence

This article is published as part of the special issue on K2P-channels. * Jürgen Daut [email protected] 1

Institute of Physiology and Pathophysiology, Marburg University, 35037 Marburg, Germany

Introduction Our knowledge about the structure and function of twopore-domain potassium channels (K2P-channels) has increased enormously over the last decade. This special issue of Pflügers Archiv aims to summarise what we know today about the physiology and pathophysiology of K2Pchannels. It contains 19 reviews and original articles written by experts in the field. This introductory overview has two aims: first is to provide some background information for the papers compiled in the special issue and second is to discuss in detail some open or controversial questions regarding structure and function of K2P-channels. K2P-channels are found throughout the animal kingdom and in plants [84, 102, 103, 273]. Functional K2P-channels assemble as dimers, in contrast to all other K+ channels, which are tetramers. Each subunit has four transmembrane domains (M1–M4), two-pore domains (P1 and P2) and two extracellular cap helices (C1 and C2). While K2P-channels were initially regarded merely as passive ‘leak channels’, it is now clear that they are regulated in a very complex way and are functionally important in the central nervous system, the heart, blood vessels, the kidneys, endocrine and exocrine glands and many other organs. There are a number of expert reviews on K2Pchannels [22, 69, 76, 84, 86, 88, 121, 153]. On the basis of sequence homology, eukaryotic K2Pchannels have been subdivided into six major subgroups or ‘clades’: 1. The TWIK clade (weakly inwardly rectifying K2P-channels) TWIK-1 (K2P 1.1; KCNK1) [154], TWIK-2 (K2P 6.1; KCNK6) [53, 207], TWIK-3 (K2P 7.1; KCNK7) [231].

868

2. The TREK clade (lipid and mechanosensitive K2P-channels) TREK-1 (K2P 2.1; KCNK2) [90], TREK-2 (K2P 10.1; KCNK10) [15], TRAAK (K2P 4.1; KCNK4) [91]. 3. The TASK clade (acid pH-sensitive K2P-channels) TASK-1 (K2P 3.1; KCNK3) [77], TASK-3 (K2P 9.1; KCNK9) [138, 220], TASK-5 (K2P 15.1; KCNK15) [8, 128, 134]. 4. The TALK clade (alkaline pH-activated K2P-channels) TASK-2 (K2P 5.1; KCNK5) [226], TALK-1 (K2P 16.1; KCNK16) [97], TALK-2 (K2P 17.1; KCNK17) [65, 97]. 5. The THIK clade (halothane-inhibited K2P-channels) THIK-1 (K2P 13.1; KCNK13) [219], THIK-2 (K2P 12.1; KCNK12) [219]. 6. The TRESK clade (spinal cord K2P-channel) TRESK (K2P 18.1; KCNK18 [239]. The phylogeny of K2P-channels in plants and animals is reviewed in this special issue of Pflügers Archiv [103].

A brief history of the six eukaryotic K2P-channel clades Many years before K2P-channels were identified as such, potassium currents that we now attribute to K2P-channels had been recorded. Eric Kandel and co-workers discovered the so-called S-current in Aplysia neurons [142, 143], a steadystate outward current that can be inhibited by serotonin; this current is carried by a TREK-1-like channel. Franks and Lieb detected a potassium current in a particular cell in the parietal ganglion of the pond snail Lymnaea stagnalis that can be activated by the volatile anaesthetic halothane [93]; this current is carried by TASK-1-like channels [2]. The first single-channel recordings of a K2P-channel were probably performed in 1989 by Donghee Kim [130, 133]. He found ‘bursty’ arachidonic acid-sensitive K+ channels in cardiac muscle cells. These channels could be activated by applying negative pressure to the pipette; they were most likely TREK-1 channels. K+ channels that were open at the resting potential and showed ‘flickery bursts’ were also recorded by Vogel and co-workers in myelinated nerve [145]. These channels (probably also TREK-1) were found to be relatively insensitive to block by tetraethylammonium. The first K2P-channel genes were detected in the worm Caenorhabditis elegans [233, 268], whose genome contains 46 different K2P-channels [232, 234]. TWIK-1 The first mammalian K2P-channel, TWIK-1 (Tandem of pore domains in a Weak Inward rectifier K+ channel; KCNK1),

Pflugers Arch - Eur J Physiol (2015) 467:867–894

was cloned in 1996 by Florian Lesage and co-workers [154]. When expressed in Xenopus oocytes, TWIK-1 was found to produce only small currents with weak inward rectification at physiological K+ concentrations [154]; later, it was shown that in mammalian cells, TWIK-1 currents show a nearly linear current-voltage relation at physiological K+ concentrations [68, 89]. However, in some laboratories, no measurable currents were observed upon heterologous expression of TWIK-1 in Xenopus oocytes [101, 197, 215], and it was proposed that the failure to record any TWIK-1 current was due to ‘silencing’ of the channel by sumoylation [214, 217]. The sumoylation hypothesis was later disproved [87]. Recently, it was found that TWIK-1 can convert from a K+-selective channel to a nonselective cation channel [26, 51]. The mechanisms governing the expression of TWIK-1 at the cell surface and the selectivity of TWIK-1 are reviewed in this special issue of Pflügers Archiv [26]. TREK-1 The second K2P-channel, TREK-1 (TWIK-1-related K+ channel; KCNK2), was also cloned in 1996 [90]. It was found to be a K+-selective channel that shares many properties with the Scurrent previously described in neurons of the marine snail Aplysia, including sensitivity to membrane stretch, activation by polyunsaturated fatty acids and by the volatile anaesthetic chloroform and inhibition via cAMP [206]. The roles of TREK-1 and TASK-1 channels in the thalamus [28], in immune cells [79], in the heart [37, 64, 141], in the adrenal cortex [13] and in anaesthesia and sleep [252] are reviewed in this special issue of Pflügers Archiv. TASK-1 The third K 2P -channel, TASK-1 (TWIK-related AcidSensitive K+ channel; KCNK3), was cloned in 1997 [77]. It was found to be highly sensitive to changes in extracellular pH in the range 7.5 to 6.8. Relatively large changes in extracellular pH can take place, for example, during hypoxia or ischemia, during epileptic seizures, during the activation of phagocytic white blood cells or during the bone reabsorbing activity of osteoclasts [70] in synaptic clefts and in T-tubules of cardiac or skeletal muscle. TASK-2 Another pH-sensitive K+ channel, TASK-2 (KCNK5), was cloned in 1998 from the human kidney [226]. It is inhibited by extracellular acidification, but its pK value is more alkaline than that of TASK-1 and TASK-3. Later, it was assigned to the TALK clade, together with TALK-1 and TALK-2, on the basis of sequence similarity. TASK-2, TALK-1 and TALK-2 are all strongly expressed in the pancreas and may be involved in the


exocrine secretion of bicarbonate. The regulation of TASK-2 channels and their functional role in the kidney are reviewed in this special issue of Pflügers Archiv [162]. THIK-1 THIK-1 channels (Tandem pore domain Halothane-Inhibited K+ channels; KCNK13) were cloned and characterised in 2001 [219]. The hallmark of THIK-1 channels is that they are inhibited by halothane; they are ubiquitously expressed and are particularly abundant in the kidney [219, 260]. In cerebellar Purkinje neurons, an outward current showing many characteristics of THIK-1 has been detected [49]. THIK-2 is strongly expressed in the central nervous system. The properties and the intracellular traffic of THIK-2 channels are reviewed in this special issue of Pflügers Archiv [26]. TRESK The last mammalian K2P-channel, TRESK (TWIK-related spinal cord potassium channel; KCNK18), was discovered in 2003 [239]. This channel is highly expressed in sensory neurons of the dorsal root ganglia and of the trigeminal ganglion, where it may play a role in nociception. TRESK has a unique long intracellular loop between M2 and M3 and is the only K2P-channel that is activated by a rise in intracellular Ca2+ ions [83, 85, 157, 272]. The regulation of TRESK channels [85] and the roles of TRESK and other K2P-channels in the sensation of pain [173] are reviewed in this special issue of Pflügers Archiv.

Are K2P-channels leak channels? K2P-channels are often described as ‘leak channels’ [99, 131, 171], ‘background channels’ [88, 121], ‘open rectifier’ channels [99] or ‘K+-selective holes’ [77]. Thus, it may be worthwhile to have a closer look at what is implied by these terms. In their quantitative analysis of the mechanisms underlying the action potential in squid axons, Hodgkin and Huxley introduced a nonselective leak conductance to make the total ionic current zero at the resting potential [119]. They assumed this leakage current to consist of chloride, sodium, potassium and other ions and wrote that B…so many components contribute toward a leakage current that measurement of its properties is unlikely to give useful information on the nature of the charged particles on which it depends^ [118]. Non-selective leak currents with a linear current-voltage relation and a reversal potential near 0 mV are also often observed in patchclamp experiments, especially in cells with a very high input resistance or in cell-free patches. This is usually an artefact related to an insufficient seal between the patch pipette and the cell membrane. However, similar leak channels are sometimes

869

observed in native cells expressing nonselective cation channels. What is an ‘ion-selective leak channel’? Potassiumselective leak channels have been pragmatically defined as constitutively open pores whose current-voltage relation at different external K+ concentrations can be described by the Goldman-Hodgkin-Katz (GHK) equation [84, 99, 121]. The GHK equation had been derived long before it was known that ions permeate membranes via pores formed by integral membrane proteins. It is based on the constant field theory, which was developed to analyse the determinants of ionic permeability changes of the membrane of nerve cells. It assumes that the ions are driven by a constant electrical field, i.e. the voltage gradient across the membrane is assumed to be linear and the concentration of the ions within the membrane is assumed to vary accordingly. In addition, ions are assumed to cross the membrane independently, i.e. without interacting with each other, and to partition into the membrane instantaneously. The GHK equation predicts that the conductance depends on the concentration of ions at either side of the membrane. When ions flow from the less concentrated side to the more concentrated side, the conductance is lower than in the reverse case. As a consequence, the current-voltage relation of potassium channels is expected to show strong outward rectification under physiological conditions, and it should be linear when the K+ concentration is symmetrical, i.e. equal on both sides, as illustrated, for example, in chapter 14 of Bertil Hille’s book [117] or in Fig. S1 of Eric Honoré’s excellent review on ‘background channels’ [121]. The assumptions underlying the GHK equation turned out to be incorrect: ions do not move independently but interact within the selectivity filter, and the electrical field across the membrane is not constant; the largest potential drop probably occurs at the selectivity filter [123]. Furthermore, the open probability of most ion channels, including K2P-channels, is voltage and time dependent. Despite these shortcomings, the GHK equation was often used successfully as a theoretical starting point when trying to analyse current-voltage relations, and in many cases, the behaviour of K+ channels is described reasonably well by the GHK equation [117]. When TASK-1 channels were first expressed in Xenopus oocytes, the resulting currents were found to be potassium selective and outwardly rectifying [77]. The current-voltage relation at various external K+ concentrations could be fitted by the GHK equation. Hence, the TASK-1 channels were described as ‘K+-selective holes’ in the cell membrane that are open at all membrane potentials [77]. Another K2P-channel, dORK1, was found in neuromuscular tissues of Drosophila and could also be described by the GHK equation and was therefore denoted an ‘open rectifier’ [100]. Subsequently, K2P-channels were usually denoted ‘potassium-selective leak channels’ or ‘open rectifiers’ [99, 217].

870


However, at least in the case of TASK-1 channels, the reasonable agreement of the measured current-voltage relation and the predictions of the GHK equation was rather coincidental. When the paper of Duprat et al. [77] was written, it was not yet known that the single-channel currentvoltage relation of TASK-1 is inwardly rectifying with symmetrical K+ concentrations (Fig. 1c) and that the open probability increases with depolarisation (see Fig. 1c of this paper and Fig. 3 of Kim et al. 1999). Both of these effects contribute to the macroscopic current-voltage relation. Furthermore, the open probability of TASK-1 channels under physiological conditions is probably very low (Fig. 1c). These properties are not compatible with the concept of an ‘open rectifier’ [99] or a ‘K+-selective hole’ [77]. The current-voltage relation of TASK-3 can only be very roughly approximated by the GHK equation [84]. The typical Fig. 1 Single-channel recordings of TREK-1 and TASK-1. a, b TREK-1 channels in rat cardiac muscle. The channels were recorded in an outside-out patch of an isolated rat ventricular cardiomyocyte with symmetrical high-potassium solution. Panel a shows a low-conductance TREK1 channel (41 pS); panel b shows a high-conductance TREK-1 channel (132 pS). The two different conductances are caused by alternative translation initiation. From ref. [156], with permission. c TASK-1 channels in rat cardiac muscle. The channels were recorded in an outside-out patch of an isolated rat ventricular cardiomyocyte with symmetrical high-potassium solution. It can be clearly seen that the single-channel conductance was larger and that the mean open time was smaller at −80 mV compared to +80 mV. The mean open time at −80 mV was 0.33 ms. From ref. [216], with permission

a

c

outward rectification of TASK-3 channels recorded under physiological conditions is the result of inward rectification of the single-channel current-voltage relation and a relatively strong increase in open probability upon depolarisation. Other K2P-channels, too, deviate substantially from the predictions of the GHK equation (see below). Similar to voltage-activated K+ channels, K2P-channels also show time-dependent gating [9, 167]. For example, when TREK-1 channels are activated by a depolarising voltage step, an instantaneous component is followed by a time-dependent component [167], and upon repolarisation, a pronounced deactivation tail is observed [167], as illustrated in Fig. 2a. Thus, the behaviour of K2P-channels is inadequately described by the term ‘leak’: They are not permanently open, their open probability increases markedly with depolarisation and they show time- and voltage-dependent gating.

b

mV +80 +60 +40 +20 0 -20 -40 40 -60 -80

2 pA 50 ms


871

a TREK-1a

b TASK-3

TASK-3

c Kv3.1 Kv3.1

Fig. 2 Comparison of outward currents produced by rat TREK-1a (a), human TASK-3 (b) and human Kv3.1 (c) in Xenopus oocytes. The heterologous expression in Xenopus oocytes, the two-electrode voltage clamp technique and the solutions channels were exactly as described previously [222, 279]. To achieve sufficient time resolution, optimal capacity compensation with the Tec10CX amplifier (npi electronics, Tamm, Germany) was required. The applied voltage protocol is described in the text. The black curves represent original recordings, the red curves represent exponential fits

The most striking property of K2P-channels is their very complex regulation by many different physical and chemical stimuli. In view of their modulatory role in many cellular functions, K2P-channels have been denoted signal integrators [121]. In conclusion, K2P-channels have moved out of the background in recent years; the characterisation of K2P-channels as ‘leak channels’ or ‘background channels’ is no longer applicable (or even misleading). If any such functional attribution is required, the characteristics of K2Pchannels may be better summarised by referring to them as ‘regulatory K+ channels’.

Do K2P-channels display voltage-dependent gating? In contrast to Kv, Nav and Cav channels, K2P-channels do not possess a specialised voltage sensing domain and, apart from TWIK-2, do not show any inactivation during depolarising

Fig. 3 Action potential clamp in Xenopus oocytes expressing TREK-1a (a), human TASK-3 (b) and human Kv3.1 (c). The voltage command was a simulated action potential of 5 ms duration at half-maximal amplitude. These measurements give a rough idea about the kinetics of the three channels during neuronal action potentials

voltage steps. Therefore, K2P-channels are sometimes described as voltage insensitive, and in studies of the function of K 2P -channels in native cells, their activation by depolarisation is often not appreciated. We therefore give a brief summary of what is known about the voltage dependence of K2P-channels. The potassium conductance (gK) of a membrane can be calculated from the equation gK =γ N Po, where γ is the single-channel conductance, N is the copy number of channels in the membrane and Po is the open probability. Both γ and Po can be dependent on membrane potential and contribute to the shape of the current-voltage relation. Increases in Po elicited by depolarisation give rise to outward rectification in the steady-state current-voltage relation and shape the time course of currents following stepwise changes of membrane potential. Upon depolarisation from negative potentials, all K2Pchannels show an instantaneous current component followed by a time-dependent component of outward current. The instantaneous component is attributable to the fact that all K2P-

872

channels have a certain open probability at negative potentials, although this is usually quite low (100 ms [75, 97, 125]. The single-channel current-voltage relations show inward rectification; thus, the outward rectification is most likely caused by an increase in open probability with depolarisation. THIK-1 and THIK-2 With symmetrical K+ concentrations, both THIK-1 and THIK-2 currents show slight inward rectification [52, 219, 225]; the currents are activated by depolarisation with a time constant of ∼1 ms [219]. TRESK The steady-state current-voltage relation of human TRESK showed outward rectification with symmetrical K+; upon depolarisation, the delayed component of TRESK activated with a time constant in the order of 30–50 ms [239]. The single-channel conductance also showed slight inward rectification, and the open probability of TRESK channels approximately doubled upon depolarisation from −60 to +60 mV [61, 127]. Interestingly, the temporal pattern of the openings also depended on membrane potential. At negative potentials, burst-like openings were much more prominent than at positive potentials [61, 127]. In conclusion, the steady-state open probability of K2Pchannels usually increases with depolarisation, albeit to a variable degree in different subtypes. In addition, most of the K2P-channels display a substantial time-dependent component of outward current after a depolarising voltage step; in the cases where this has been investigated in detail [160, 167, 207, 220, 226], the time-dependent component usually comprised 50 % or more of the total current at positive potentials. These findings suggest that there must be some intrinsic voltage-dependent gating of K2P-channels. The mechanisms underlying the voltage sensitivity of K2P-channels remain to be investigated.

Do K2P-channels contribute to the generation of action potentials? In most K2P-channels, the steady-state current-voltage relation under physiological conditions (with 4–5 mM extracellular K+) shows marked outward rectification. Therefore, the ‘standing outward current’ [30, 177] observed at depolarised potentials is much larger than the current flowing through

873

K2P-channels at the resting potential. This raises the possibility that K2P-channels may contribute to the configuration of the action potential. As neurons usually express several different potassium channels, the precise role of individual ion channel species in the repolarisation phase of the action potential is difficult to assess. Action potentials measured in the cell soma are caused by synaptic inputs and are normally triggered in the initial segment of the axon. The changes in the open probability of various ion channels during the action potential are controlled by voltage, but the current flowing through the channels, in turn, causes a change in voltage. Even in voltage-independent channels, the change in membrane potential will alter the electrochemical driving force for the permeant ions and thus the transmembrane current, which again changes the potential. Thus, the action potential configuration is shaped by the dynamic interplay of current and voltage for several ion channels, and the time course of the contribution of any type of individual channels can only be determined by using an action potential clamp and specific ion channel blockers or by computer simulation [24]. Moreover, there is a considerable diversity of action potential waveforms in different types of neurons [24]. The voltage-activated potassium channels (Kv channels) relevant for repolarisation usually do not activate completely during the upstroke of the action potential and deactivate or inactivate rapidly during the repolarisation phase, so that only a fraction of the current that could be produced by longer rectangular voltage steps actually flows through the channels. During repetitive firing, things are even more complicated, because the time course of recovery from inactivation comes into play and cumulative activation and inactivation of potassium channels during a train of action potentials may occur. Despite these uncertainties, it is possible to ascribe typical functions to certain types of ion channels in neurons. In trying to understand the possible function of K2P-channels in shaping neuronal action potentials, it may therefore be useful to compare the voltage-dependent properties of K2P-channels with those of the voltage-activated potassium channels of the Kv3 family which activate and deactivate very fast and thus show some similarities to K2P-channels [230]. Voltage-gated potassium channels of the Kv3 type have been identified as major determinants of the repolarisation phase in ‘fast-spiking’ neurons, for example, in the hippocampus, in the basal ganglia, in the thalamus, in auditory nuclei, in the cerebellum and in the neocortex [24, 230]. These neurons often have relatively short action potentials and are capable of firing at high frequencies with little decrease during prolonged stimulation [24]. Kv3 channels are characterised by a fast rate of deactivation, which makes them suitable for supporting high action potential frequencies with little adaptation. Figure 2c shows the time course of activation and deactivation of Kv3.1 channels expressed in Xenopus oocytes. The membrane potential was stepped from a holding potential of −80 to

874

+40 mV for 10 ms, back to -20 mV for 55 ms (to get an impression of the rate of deactivation during the repolarisation phase) and finally back to -80 mV. To facilitate comparison with K2P-channels, the current amplitude was normalised to the steady-state current measured during prolonged depolarisation. The final phase of the activation could be fitted by an exponential function with a time constant of 0.73 ms. It can be seen that the Kv3.1 reached a steady state within the 10-ms depolarising voltage step, but only a small fraction of the current was activated during the first ms (i.e. during the upstroke of a normal action potential). The repolarising voltage step to −20 mV was associated with an instantaneous current jump, followed by complete deactivation with a time constant of 3.4 ms. Figure 2a shows the time course of activation and deactivation of TREK-1 channels expressed in Xenopus oocytes. The voltage protocol was the same as for Kv3.1, and again the current amplitude was normalised to the steady-state current measured during prolonged depolarisation. It can be seen that the depolarising voltage step from −80 to +40 mV gave rise to an instantaneous current jump, followed by a voltageactivated component; the time constant of activation was 2.0 ms. The repolarising voltage step to −20 mV gave rise to a current jump, followed by partial deactivation with a time constant of 1.2 ms. Thus, the deactivation of TREK-1 was approximately threefold faster than that of Kv3.1. As expected, there was a steady-state component of outward current at 20 mV. This component decayed upon repolarisation to 80 mV with a time constant of 1.0 ms. For comparison, the time course of activation and deactivation of TASK-3 channels expressed in Xenopus oocytes is shown in Fig. 2b. Again, the current amplitude was normalised to the steady-state current measured during prolonged depolarisation. The instantaneous current jumps of TASK-3 measured during the depolarising voltage step from -80 to + 40 mV and during the repolarising voltage step to -20 mV were somewhat larger than in the case of TREK-1; the time constants of activation and inactivation were similar to those for TREK-1. The steady-state component of outward current at -20 mV was also a little larger for TASK-3 than that for TREK-1. In both TREK-1 and TASK-1, the rate of rise of the current was slow enough to avoid too much overlap with the activation of Nav channels, which would prevent excessive shunting of the depolarising current. The results presented in Fig. 2 illustrate that despite some quantitative differences, the kinetics of activation and inactivation of the K2P-channels TREK-1 and TASK-3 on the one hand and of the voltage-activated potassium channel Kv3.1 are surprisingly similar. The major difference is that in the two K2P-channels, there is a residual current at -20 mV. None of the three currents discussed here would be fully activated during a normal neuronal action potential, and the relative contribution of these channels to the repolarisation


phase would depend on the density of the respective channels. The time course of the Kv3.1, TREK-1 and TASK-3 currents during a neuronal action potential cannot be reliably predicted from these measurements because it depends on the dynamic time- and voltage-dependent activation and deactivation of the channels. To get an estimate of the time course of Kv3.1, TREK-1 and TASK-3 currents during an action potential, we expressed these channels (individually) in Xenopus oocytes and applied an action potential clamp, as illustrated in Fig. 3. The simulated action potential had a duration of 5 ms at halfmaximal amplitude. The currents flowing during the action potential clamp were preceded by a capacitive artefact and differed in their temporal relation to the imposed action potential. The peak of the Kv3.1 current (Fig. 3c) was reached later than that of TREK-1 (Fig. 3a) and TASK-3 (Fig. 3b). This may be attributable to the fact that the threshold for activation of Kv3.1 is around -20 mV, whereas the K2P have an instantaneous component flowing at all voltages during the upstroke of the action potential. The decay of the Kv3.1 current was also steeper than that of the two K2P-channels. This can also be explained by the steeper voltage dependence of Kv3.1. The decay of TREK-1 current was a little faster than that of TASK-3, in agreement with the steeper voltage dependence of TREK-1. This data suggests that the fast voltage-activated channel Kv3.1 and the K2P-channels TREK-1 and TASK-3 are available for reopening immediately after the end of the action potential. They show virtually no inactivation, which implies that the time course of recovery from inactivation is irrelevant and cumulative inactivation is not possible, and they are thus capable of supporting the repolarisation of rapidly firing neurons. This may also apply for other neuronal K2P-channels. In conclusion, our measurements suggest that Kv3.1 currents on the one hand and K2P-channel currents on the other hand display a surprisingly similar time course during an action potential. Since the shape of neuronal action potentials is quite variable [24], it would depend on the biophysical characteristics and on the relation between synaptic input and axonal output whether Kv channels or K2P-channels, or indeed a combination of Kv and K2P-channels, would be more suitable for a specific type of neurons. A study of the electrical activity of cerebellar granule neurons in TASK-3 knockout mice [40] showed a very interesting phenotype consistent with the above considerations: While the neurons of wild-type mice showed sustained repetitive firing during depolarising current pulse, the neurons of the TASK-3 knockout mice had lost the capability of firing continuously at high frequencies. The same depolarising current pulse that produced high-frequency spikes in wild-type neurons gave rise to rapidly adapting action potentials with reduced overshoot and subsequent complete action potential failure [40]. In addition, the action potentials became longer and the after hyperpolarisation was reduced in TASK-3 knockout mice. These findings support the idea that TASK-3


channels may contribute to the fast repolarisation of rapidly firing cerebellar granule neurons; this is expected to reduce the activation (and subsequent inactivation) of Kv channels and to promote the recovery from inactivation of voltage-activated sodium (Nav) channels. Collectively, these changes would facilitate continuous rapid firing of the neurons. Thus, K2Pchannels such as TASK-3 may in fact play an essential role in determining the action potential pattern and frequency elicited by an excitatory synaptic input. The idea that K2P-channels may contribute to the action potential is reinforced by a recent study published in this issue of Pflügers Archiv [165]. The authors combined theoretical calculations with dynamic patch-clamp measurements in cells expressing TREK-1 or TASK-3 channels. The computer simulation showed that introducing a time-independent K+ conductance (described by the Goldman-Hodgkin-Katz equation) in addition to a voltage-activated sodium current (described by Hodgkin-Huxley-type parameters) was sufficient to generate action potentials. Heterologous expression of TREK-1 or TASK-3 channels in HEK-293 cells confirmed this conclusion: even after blockage of endogenous voltage-activated K+ channels by application of TEA, these cells were capable of generating action potentials when a voltage activated sodium conductance was mimicked by a dynamic patch-clamp system [165]. In the computer simulations mentioned above [165], the time dependence of TREK-1 and TASK-3 currents was ignored; perhaps the overshoot of the simulated action potentials would have been even larger if the kinetics of activation were taken into consideration, because less of the initial sodium current would have been shunted by the potassium conductance.

How is the number of K2P-channels at the cell surface regulated? The density of K2P-channels in the cell membrane can be regulated at the transcriptional level [183, 211, 275, 276] and at the posttranslational level [195, 218, 237]. In the past, the regulation of the intracellular transport to and from the surface membrane has often been ignored. Over the last decade, it has become clear that the posttranslational regulation of the number of channels in the cell membrane is functionally very important because it modulates the density (and, thus, the functional role) of many ion channels. Generally, there is a continuous turnover of membrane proteins at the cell surface, and the surface expression depends on the balance between forward transport along the secretory pathway and endocytosis. Thus, the abundance of ion channels at the cell surface can also be regulated via modulation of the rate of endocytosis. The role of endocytosis in the regulation of the functional expression of K2P-channels is reviewed in this special issue of Pflügers Archiv [194].

875

One of the first proteins found to interact specifically with K2P-channels was microtubule-associated protein 2 (Mtap2) [237]. Co-expression of Mtap2 increased the surface expression of TREK-1 and TREK-2. This effect may be mediated by the simultaneous docking of Mtap2 to microtubules and to the channels [237]. The intracellular traffic of TASK-1 and TASK-3 channels has been studied in some detail [195, 218, 222, 223, 279]. The forward transport of these channels strongly depends on the binding of 14-3-3 proteins to their extreme C-terminus. The interaction between 14-3-3 proteins and TASK-1 and TASK-3 depends on phosphorylation of a serine residue at the penultimate position of the C-terminus of the channels. The abundance of TASK-1 and TASK-3 channels at the cell surface can be regulated via activation of protein kinase A [170], which phosphorylates a serine residue in the binding domain for 143-3 proteins. Other proteins that interact with TASK-1 channels and influence their abundance at the cell membrane are p11 (also known as S100A10) [223] and syntaxin-8 [222]. The role of protein-protein interactions in the intracellular traffic of TASK-1 and TASK-3 is reviewed in this special issue of Pflügers Archiv [129]. Human THIK-1 and THIK-2 channels are closely related; they show 64 % amino acid identity [219]. THIK-1 can be readily expressed in mammalian cell lines and in Xenopus oocytes, whereas for many years, all attempts to express THIK-2 channels failed. Recently, it has been shown that THIK-2 carries an arginine-based retention signal in its Nterminus that drastically reduces its surface expression (but does not abolish it) [52, 224]. Mutation of this signal gave rise to measurable THIK-2 currents. In situ hybridisation experiments have shown that THIK-2 channels are highly expressed in many nuclei of the brain [219]. Like THIK-2, TWIK-1 also shows a very low expression at the cell surface under normal conditions. One of the mechanisms underlying the predominantly intracellular localisation of this channel was elucidated when Lesage and co-workers detected an unconventional endocytosis signal, two isoleucine residues at positions 293 and 294, that provides for constitutive, rapid endocytosis of TWIK-1 [89]. Mutation of the diisoleucine motif gave rise to a substantial increase in current amplitude, which facilitated the study of the mechanisms regulating gating, ion selectivity and intracellular traffic of TWIK-1 channels [51]. The mechanisms underlying the ‘silence’ of TWIK-1 and THIK-2 are reviewed in this special issue of Pflügers Archiv [26].

Alternative splicing and alternative translation initiation K2P-channels show a very unusual pattern of alternative splicing and, in addition, alternative translation initiation.

876

Furthermore, different splice variants are subject to different degrees of alternative translation initiation, so the two phenomena are interdependent. However, the functional implications are still far from clear. In the following, we summarise and discuss some of the findings published so far.

Alternative splicing The open reading frame of K2P-channels is encoded by 2 to 7 exons; human TASK-1 has only two exons, whereas the TREK clade has seven exons in all species studied. One particularly interesting splice site, which is conserved across the entire K2P-channel family, is in the middle of the pore signature domain GY//G. The striking conservation of gene structure suggests that this ion channel family evolved by several gene duplication events. In TREK-1, TREK-2 and TRAAK, several spice variants of exon 1 have been described, resulting in N-terminal cytosolic domains of different length [105, 198, 207]. The pattern of these N-terminal splice variants, too, has been conserved across species. For example, in one splice variant of human TREK-1 (Variant 2, isoform b, accession number NM_014217.3; denoted TREK-1a in ref. [228]), the first exon contributes only one nucleotide to the code of the first amino acid, methionine, and orthologous splice variants have been found in mouse, zebrafish, chicken and cow. Interestingly, the N-terminal splice variants are differentially expressed in various tissues, which suggests that the N-terminus or the N-terminal noncoding region of the mRNA may be relevant for cellular function. However, no difference in the biophysical properties of the different N-terminal splice variants has been found so far. A splice variant of TREK-1 lacking exon 5 has been described recently [228]. The skipping of exon 5 leads to a frame shift in exon 6, which is associated with a novel C-terminal amino acid sequence and a premature termination of translation. This splice variant, denoted TREK-1e, lacks the transmembrane domains M3 and M4 and carries a retention signal at its novel C-terminal end. When expressed alone, TREK-1e is localised to the endoplasmic reticulum; when it is coexpressed with wild-type TREK-1 (splice variants TREK1a-d), it reduces the measured surface expression and current density. Another splice variant of TREK-1 lacking exon 4 has been detected in rat and mouse brain [264]. The alternative splicing causes a frame shift leading to a premature stop codon. As a result, this alternatively spliced isoform comprises only the first transmembrane domain and a unique extracellular domain. This splice variant produces no functional channel but has a dominant negative effect on TREK-1. Alternative splicing was also found in TALK-1 [112] and TASK-1 [139].


Alternative translation initiation In most, but not all, cases, translation initiation sites on the mRNA (AUG, coding for methionine) are recognised by the ribosome. The reliability with which the translation initiation site (start codon) is detected depends on the nucleotides flanking the AUG codon. If the nucleotide context is unfavourable, the start codon may be skipped and the next downstream start codon is used (‘leaky scanning’), resulting in synthesis of a truncated protein [146]. This process has been denoted alternative translation initiation (ATI). The first ion channel transcript in which ATI was shown to be relevant was TREK-1. The mRNA of TREK-1 has two potential translation initiation sites, and both result in functional channels: a ‘long isoform’ form and a ‘short isoform’ (with a truncated cytosolic N-terminus). Interestingly, the splice variant of TREK-1 lacking exon 4 (see above) had a dominant negative effect only on the long isoform, whereas the short isoform was unaffected [264]. Surprisingly, the short and the long isoforms were found to have different biophysical characteristics. The short isoform was reported to have a 30 % larger single-channel conductance and a lower selectivity for K+-ions [261]. The mechanism by which the truncation of the N-terminus may influence single-channel conductance is unknown. It may be related to the influence of the cytoplasmic domain on the permeation pathway [20]. The inference that the short isoform of TREK-1 has a relatively high permeability to sodium ions (PNA/PK ∼0.18) was based mainly on the observation that the reversal potential (measured in Xenopus oocytes) was shifted in the positive direction. However, it is difficult to be confident about these conclusions because the short isoform gave much smaller currents than the long isoform. With such small currents, the reversal potential measured in Xenopus oocytes can be misleading due to the presence of endogenous nonselective cation channels or chloride channels. ATI has also been found in the closely related channel TREK-2 [248]. In TREK-2, ATI also gives rise to cytosolic N-termini of different lengths and to potassium channels with different conductances. TREK-2 has three potential translation initiation sites at its N-terminus; systematic mutagenesis of these sites was employed to produce channels with a short, an intermediate or a long N-terminus [248]. The long isoform displayed a conductance of 52 pS, whereas the intermediate and the short TREK-2 isoforms predominantly produced a channel with a conductance of ∼200 pS. No change in ion selectivity was found between the different ATI isoforms of TREK-2 [248]. As noted above, TREK-2 channels have at least three Nterminal splice variants, denoted TREK-2a, b and c, which differ in their tissue distribution [105]. These splice variants all have three potential translation initiation sites in their Ntermini. Since the first translation initiation sites show different efficiencies, the three isoforms produce different relative


877

amounts of truncated TREK-2 channels due to leaky scanning of the mRNA [251]. Thus, the combination of alternative splicing at the C-terminus and ATI can produce a large variety of protein isoforms. Since the relative amount of ATI appears to vary between different tissues, any alteration of the biophysical properties of truncated TREK-1 channels may be functionally relevant. The first studies of the biophysical properties of TREK-1 channels in heterologous expression systems reported a single-channel conductance of 95–130 pS at positive potentials [111, 122, 206]. However, a later study showed that both in native rat cardiomyocytes and in heterologous expression systems TREK-1 channels with two different conductances were observed: one with ∼40 pS and one with ∼130 pS [156], as illustrated in Fig. 1a, b. The low-conductance channel was found to be more abundant, and in some cases, sudden transitions between the two conductances were found. For lack of an alternative explanation, the two different conductances were interpreted as two gating modes of the same channel [156]. With hindsight, it appears likely that the two different conductances were attributable to ATI; the observed sudden transitions between the two conductances [156] may have been coincidental. Thus, ATI of TREK-1 may indeed occur in the heart. In magnocellular neurosecretory cells of the hypothalamus, too, two different channels with TREK-1-like properties were found, and again the low-conductance channel (‘a novel TREK-like channel’) was found to be more abundant [111]. This might also be attributable to ATI. In conclusion, ATI can produce TREK channels with different properties both in native cells and in heterologous expression systems. The reason for the differences in singlechannel conductance and the physiological relevance of having channels with different conductances in the same cells (or in different cells) remains to be established.

The structure of K2P-channels Beginning in 2012 [43, 181], there has been enormous progress in our understanding of the structure of K2P-channels. Since this is relevant for analysing the many functions of K2P-channels described in this special issue of Pflügers Archiv, we here provide a very simple overview of the main structural features of K2P-channels (Fig. 4). Basic properties

Fig. 4 The structure of K2P-channels. a Topology of K2P-channels. b Sketch of the structure of the N-terminal part of the two subunits (including the M1, C1, C2, P1 and M2 domains). c Sketch of the structure of the C-terminal part of the two subunits (including the M3, P2 and M4 domains). The helices are not drawn to scale. For clarity, the pore helices are relatively small

The crystal structure of two K2P-channels, TWIK-1 and TRAAK, has been solved by Miller and Long [181] and Brohawn et al. [43]. Most of the structural characteristics of TWIK-1 and TRAAK were found to be similar: The channels have an extracellular cap (consisting of two cap helices, C1 and C2) that is unique among ion channel structures (Fig. 4a).

The cap extends ∼35 Å above the extracellular membrane and covers an extracellular vestibule that has two lateral portals for K+ ions. The pore-lining helices, M2 and M4, run obliquely through the cell membrane, whereas the outer helices, M1 and M3, are more vertically oriented. These basic properties may apply to all K 2P-channels. Both groups found that the

878

transmembrane segments of the channel have lateral openings (between M3 and M4) that face the lipid environment. The selectivity filter and the pore helices showed approximately fourfold symmetry, and the inner mouth of the channel was wide open. The extracellular cap, the central cavity and the inner mouth of the channel showed roughly twofold symmetry instead of the fourfold symmetry found in other potassium channels. Basically, the two subunits were found to be assembled in a simple antiparallel manner.


a

M3 M4

M1

TREK-1

M2 M2 M4

M1

M3

Swapped domains However, in a subsequent paper, Brohawn et al. [41] used an antibody fragment to crystallise TRAAK channels. Working with a higher resolution (2.75 Å), they refined their analysis and found that the C1 and the M1 helices were swapped to the opposite side of the channel (Fig. 4b). Brohawn et al. [41] state that at the resolution of the first structure (3.8 Å), it was in fact not possible to distinguish between domain-swapped and non-domainswapped models. Since the structures of TREK-1 and TREK-2, which have been deposited in the PDB database by Carpenter and co-workers, also show swapped domains (PDB codes 4TWK and 4BW5), it is likely that this is the generally applicable model for K2P-channels. This is illustrated in the simplified sketch of Fig. 4b which depicts the first pore domain (including M1, C1, C2, P1 and M2) of both subunits (orange and cyan), which are opposite to one another. The C1 and C2 helices of each subunit form a ‘tent’ with an angle of about 45° (whereas in the original papers, C1 and C2 were assumed to be adjacent and run parallel [43, 181]). The C1–C2 linkers are connected via a disulfide bridge. The two opposite pore helices and the corresponding pore loops contribute to forming the selectivity filter. The M1 and C1 helices are flipped by 180° to the other side. Thus, the M1 helix of one subunit is adjacent to the end M2 helix of the other subunit, whereas in all other potassium channels, the outer and inner helix of the same subunit are adjacent. The second pore domain of both subunits (including M3, P2 and M4) is depicted in Fig. 4c. Here, the view is rotated by 90°. It can be seen that in this case, the inner and outer helix of the same subunit are adjacent. The two opposite pore helices and pore loops form the other half of the selectivity filter. The M2 and M3 helices are connected by dotted lines to illustrate the overall spatial arrangement of the four transmembrane helices of each subunit. It is obvious from Fig. 4 that the helical cap and the transmembrane helices show only two-fold symmetry. This may be relevant for the mechanism of gating, which in tetrameric K+ channels involves fourfold

b

M1

M4

M3

TWIK-1

M2

M2 M4 M3

M1

Fig. 5 Space-filling model of TREK-1 (a) and TWIK-1 (b) viewed from the cytosolic side. The four transmembrane helices are shown in different colours; the helices belonging to one subunit are labelled with bold fonts, the helices belonging to the other subunit are labelled with normal fonts. The structure of TREK-1 has been deposited in the PDB database by Carpenter and co-workers (code 4TWK). The structure of TWIK-1 has been published by Miller and Long [181]. Although it is not known whether domain swapping also takes place in TWIK-1, the M1 domain of TWIK-1 has been swapped here, in analogy to TREK-1 and TRAAK

symmetric movement of the inner helices [123]. There are also some interesting differences between the crystal structures obtained for TRAAK and TWIK-1. The surface of TREK-1 channels viewed from the cytoplasmic side is presented in Fig. 5a; the transmembrane domains are colour coded. It can be seen that the ‘footprint’ of TREK-1 channels is rhomboid in shape and shows a roughly twofold symmetry instead of the fourfold symmetry found in other potassium channels. In contrast, the footprint of TWIK-1 is almost square-shaped (Fig. 5b) [181].

Charged residues The M4 helix of TWIK-1 shows a bend of ∼80° at L271, and the amphipathic distal part of M4, termed C-helix, runs parallel to the membrane surface [181]. The C-helix is amphipathic with charged residues facing the cytoplasmic side. TRAAK also has an amphipathic helix at the end of M4, although it is not parallel to the membrane but at an angle of about 30°. In most of the human K2P-channels (except for the TALK clade),


the cytosolic end of M4 carries two to four charged residues and appears to form an amphipathic helix [43, 181]. Since the distal end of the M4 helix in TRAAK [43] and TREK-2 as well as the C-helix in TWIK-1 [181] make contact with the cytosolic ends of both M1 and M2, it is tempting to speculate that they may play a role in gating of the channels [181]. Similar horizontally oriented amphipathic helices have been found in voltage-activated K+ channels (the S4–S5 helices) [257], in mammalian inward rectifier channels [20, 159, 256], in the bacterial K+ channel KcsA [58] and in the bacterial inward rectifier channel KirBac1.1 [147]. The role of these amphipathic helices (also denoted ‘slide helix’ or ‘interfacial helix’) in the function of potassium channels is not entirely clear; it has been proposed that they may serve as an anchor for the ‘cytosolic pore’ of the channel [66] and that they may couple the pore opening to the movement of the cytosolic domains [66, 209]

879

a

Cytosolic pore In addition to the transmembrane pore, most potassium channels also possess a cytosolic pore that extends relatively far (∼60 Å) into the cytosol [144, 147, 192, 210, 249] and may contribute to the gating and permeation properties of the channels [106, 191, 210, 250]. So far, we know nothing about the cytosolic pore of K2P-channels, although, in view of the long C-terminal domains, the M2-M3-linker and the (usually shorter) N-terminal cytoplasmic domains, it is likely to exist and may well contribute to the regulation and the permeation properties of the channels. Interestingly, truncation of the cytosolic N-terminus changed the conductance of TREK-1 and TREK-2 channels [248, 261].

b L151 V152

M2

F272

P155

I159

L276 I279

M4

L283

Lateral fenestrations The K2P-channels that have been crystallised so far have a lateral opening at the interface between the M2 helix of one subunit and the M4 helix of the other subunit [43, 181]. This opening connects the inner cavity with the hydrophobic core of the cell membrane. Figure 6 shows a lateral view of TRAA K; the domains belonging to the two subunits are coloured. The hydrophobic residues lining the lateral fenestration are highlighted in green. The lateral cleft is likely to be filled either with the hydrophobic tails of membrane phospholipids (their headgroups being located in the same plane as the headgroups of the phospholipids in the bulk lipid bilayer) [151, 152]. The interaction of such ‘buried lipids’ with the M2 and M4 domains of K2P-channels could be relatively specific and might also influence the gating the channels [151, 190, 242, 243]. If the alkyl chains of membrane lipids protrude into the cavity through the lateral opening, they might

Fig. 6 Space-filling model of TRAAK [41] viewed from hydrophobic core of the cell membrane. Part of panel a is shown at higher magnification in panel b. The two subunits are colour coded. The hydrophobic residues of the lateral fenestration (in the down state) are highlighted in green.

influence the conductance and the pharmacological properties and of K2P-channels [42]. Glycine hinge Interestingly, the glycine residue which serves as a gating hinge in most potassium channels [123] is conserved in both the M2 and the M4 helices of all K2P-channels. In hTWIK-1, for example, the two glycine residues are at positions 141 and 256; in hTRAAK, the two glycines are at positions 153 and

880

268 [43]. Recently, it has been found that both the M2 and the M4 domain can indeed bend around the hinge of G268 [42, 158]; this bending moves the M4 helix from a ‘down’ to an ‘up’ configuration. It has been suggested that stretch of the membrane may facilitate the transition to the up configuration, which seals the fenestration and thus prevents the entry of lipids into the inner cavity of the channel [42]. The hydrophobic inner cavity The central cavity of many potassium channels, including K2P-channels, is lined with hydrophobic residues [5]. It has been shown that ‘nanopores’ with a diameter of