Chansporter complexes in cell signalling

Abstract

Ion channels facilitate diffusion of ions across cell membranes for such diverse purposes as neuronal signalling, muscular contraction, and fluid homeostasis. Solute transporters often utilize ionic gradients to move aqueous solutes up their concentration gradient, also fulfilling a wide variety of tasks. Recently, an increasing number of ion channel-transporter (“chansporter”) complexes has been discovered. Chansporter complex formation may overcome what could otherwise be considerable spatial barriers to rapid signal integration and feedback between channels and transporters, the ions and other substrates they transport, and environmental factors to which they must respond. Here, current knowledge in this field is summarized, covering both heterologous expression structure/function findings and potential mechanisms by which chansporter complexes fulfil contrasting roles in cell signalling in vivo.

1. Introduction

Cells in biological systems require constant and highly regulated movement of ions across cell membranes, to perform or regulate virtually all biological processes. Rapidly gating (opening and closing) voltage-gated sodium channels and potassium channels generate and propagate the action potentials required for nervous signalling, the heartbeat and other processes in excitable cells [1–3]. An array of calcium channels, including voltage-gated and store-operated channels on the plasma membrane, and channels in the sarcoplasmic/endoplasmic reticulum membrane that are opened by a variety of stimuli including Ca²⁺ and coupling to plasma membrane Ca²⁺ channels, increase cytosolic Ca²⁺ levels to trigger processes as diverse as wound healing, fertilization, gene transcription, immune responses, and long-term potentiation [4–9]. Other classes of ion channels with various grades of ion selectivity are required for ion and fluid homeostasis and secretion, regulation of resting membrane potential, and hormone production and secretion, often in non-excitable cells such as polarized epithelial cells [10–14].

Ion channels do not actively transport ions, but instead they facilitate ion diffusion down an electrochemical gradient, by creating an aqueous pathway through otherwise inhospitable cell membranes. These pathways are typically selective either for specific ions or classes of ions, and diffusion is highly regulated. This regulation often takes the form of gates intrinsic to the channel protein, which open and close to permit or stem ion flow; other forms of regulation can involve extrinsic moieties, e.g., inward rectification of some ion channels via block by Mg²⁺ and other molecules at depolarized voltages. Most ion channels, in addition to opening (activation) and closing (deactivation) also have an inactivation or desensitization gate that can close to stem the flow of ions even when the activating stimulus (e.g., membrane depolarization, ligand binding) is still present [15–18]. Another type of protein termed a uniporter, distinct from channels, also mediates movement of solutes across the cell membrane down their concentration gradient, but instead works by facilitative diffusion. Thus, solutes bind to e.g., the extracellular side of the protein, which then changes conformation to expose the bound solute to the intracellular side, permitting diffusion into the cell. Examples include the facilitated glucose transporters GLUT1-4 [19, 20].

In contrast, many other classes of transporters move solutes across membranes up their concentration gradient. This is achieved via one of various mechanisms. Secondary active transporters utilize a downhill ionic gradient, to move a solute up its concentration gradient concomitant with downhill transport of the ion [21–23]. For example, SGLT1 utilizes a downhill Na⁺ gradient to move glucose into cells, up its concentration gradient. Sodium-coupled transporters represent a major class of secondary active transporters for a range of ions and solutes, and several will be covered in this review. They are termed “secondary” active transporters because energy from ATP hydrolysis is required to establish the Na⁺ gradient they require for function, but the hydrolysis is not performed by the sodium-coupled transporter itself [24, 25]. Instead, the Na⁺ gradient is established by another protein, such as the Na⁺/K⁺-ATPase, which is often described as pump. The Na⁺/K⁺-ATPase is a primary active transporter because it hydrolyzes ATP to provide energy to pump K⁺ into the cell and Na⁺ out of the cell, both against their electrochemical gradients [26].

For many cellular processes, ion channels must work in concert with ion and solute transporters. The Na⁺/K⁺-ATPase is a prime example. Aside from establishing the Na⁺ gradient for Na⁺-coupled solute transport through other transporter proteins, the Na⁺ and K⁺ gradients established by the Na⁺/K⁺-ATPase are essential for maintaining resting membrane potential and setting the stage for the action potential [27–30], a process orchestrated by rapidly-gating ion channels [2, 31–35]. The importance of the Na⁺/K⁺-ATPase for nervous signalling is amply illustrated by the calculation that this one type of pump protein accounts for as much as two-thirds of neuronal energy expenditure [36].

By the same token, solute transporters, including pumps, may be reliant upon ion channels. As an example, gastric acidification in mammals requires the gastric H⁺/K⁺-ATPase to secrete protons into the stomach. The H⁺/K⁺-ATPase is positioned intracellularly in vesicles, until stimulated to traffic to gastric (oxyntic) pits on the apical surface of parietal cells, and then pump protons into the stomach. As protons move out, K⁺ ions move through the H⁺/K⁺-ATPase into the cell, where they would accumulate in the absence of a suitable conduit to release them back into the stomach lumen (where they are required for further cycles of the H⁺/K⁺-ATPase) [37, 38]. Several apically located ion channels are thought to serve the function of the K⁺ recycling conduit in parietal cells. KCNQ1-KCNE2 is a constitutively active potassium channel apically localized in parietal cells, which is potentiated by low extracellular pH, making it an ideal candidate for this process. Accordingly, genetic deletion from mice of either Kcnq1 or Kcne2 results in achlorhydria (loss of gastric acid secretion), indicating an essential role for KCNQ1-KCNE2 in gastric acid secretion, by supporting H⁺/K⁺-ATPase function [38–45]. Inward rectifying K⁺ channels including Kir1.1, Kir2.1, Kir4.1 and Kir4.2 are also thought to be important for gastric acid secretion [46–50].

Beyond functional cooperation between ion channels and transporters, some channels have been found to form physical complexes with solute transporters, a phenomenon that has not yet been described for the H⁺/K⁺-ATPase or Na⁺/K⁺-ATPase. Novel channel-transporter (“chansporter”) complexes are being discovered more frequently in the last 5 years than ever before, suggesting they may represent a previously overlooked, yet important mode of cellular signalling. In this review, the author describes examples of channel-transporter complexes reported in the literature, together with their known or suspected physiological roles and the types of signal integration they facilitate. The complexes covered in this review are summarized in Table 1.

Table 1.

Chansporter complexes covered in this review.

Transporter class	Transporter (other names)	Interacting channels (other names)	Channel class	Native expression	References
Sodium-coupled sugar/polyol transporters	SLC5A3 (SMIT1)	KCNQ1, 2, 2/3 (Kv7.1, 2, 2/3)	KCNQ family voltage-gated potassium channels	Nervous system	[110, 130]
	SLC5A11 (SMIT2)	KCNQ1, 2, 2/3		Nervous system	[110, 130]
	dSLC5A11 (cupcake)	dKCNQ		Nervous system	[137]
Chloride and bicarbonate exchangers	SLC26A5 (prestin)	KCNQ2, 3, 4 (Kv7.2, 3, 4)		Putative complex formation in outer hair cells	[150]
	SLC26A2, 3, 4 (pendrin), 5, 6, 9	CFTR	Chloride channel/ABC transporter-like	Epithelia, e.g., pancreas, salivary glands	[151]
ABC-family non-transporters	Sulfonylurea receptors (Sur1, Sur2)	Kir6.1, 6.2	Inward rectifier K+ channels	Pancreas, heart	[166]
		TRPM4	Calcium-activated, monovalent cation-nonselective channel	Post-injury nervous system	[168]
Calcium ATPase	secretory pathway Ca2+-ATPase isoform 2 (SPCA2; ATP2C2)	Orai1	calcium channel	Mammary epithelium	[170]
Amino acid transporters	SLC6A11 (GAT3)	MaxiK (BK; Slo1)	calcium-activated potassium channel	Nervous system	[176]
	SLC61A2 (EAAT2; GLT-1)	AQP4	water channel	Astrocytes	[178]
Sodium chloride co-transporter	SLC12A3 (NCC)	ENaC	epithelial sodium channel	Distal convoluted tubules	[185]

d, Drosophila melanogaster. All other complexes listed involve mammalian proteins.

2. SLC5A sodium-coupled solute transporter interactions with KCNQ potassium channels

Voltage-gated potassium (Kv) channels are formed by pore-forming α subunits that co-assemble into tetramers around an aqueous ion-conducting pore. They are in general highly selective (>10:1) for K⁺ over Na⁺, a property important for their role in repolarizing excitable cells to end each action potential. Kv channel α subunits are encoded by 40 different genes in the human genome, each of which fall into one of ~12 (depending on which nomenclature is used) subfamilies. Most Kv α subunits are functional as homomers, but many form heteromers with other α subunits (typically but not always from the same subfamily) that display distinct functional characteristics and may be prerequisite for some biological processes [2].

Most if not all Kv α subunits also co-assemble with some form of ancillary or β subunit to facilitate their surface trafficking or targeting, and/or modify their function for specific biological activities [51–54]. Still others, termed “silent” α subunits, either silence other α subunits upon heterotetramerization, or require heteromerization with other α subunit types for ion conduction [55, 56]. Kv α subunits, including silent α subunits, each possess six transmembrane domains split into a voltage sensing domain (VSD; transmembrane helices S1–S4) and a pore module (S5 and S6) [57–59] (Figure 1A). Kv channel ancillary subunits can be single transmembrane-spanning (1TM) such as the KCNEs (Figure 1A) or entirely cytosolic and their importance to normal channel function in vivo is emphasized by the pathogenic effects of their disruption [60–65].

Figure 1. KCNQ1-KCNE channel structure and function.

A. Representative electrophysiological recordings [65] from Xenopus oocytes expressing KCNQ1 alone (centre), or with KCNE1 (left) or KCNE3 (right). Voltage protocol inset. Subunit topologies shown beneath corresponding traces. VSD, voltage sensing domain.

B. High-resolution structures obtained by cryo-EM of Xenopus KCNQ1 [186] (red) bound to calmodulin (blue); left, one subunit; right, tetramer. View from inside membrane perspective.

C. A tetramer of KCNQ1 [186] viewed from inside the membrane (left) or outside the cell (right) perspectives. Individual subunits are coloured orange, green, light green and gold (KCNQ1) versus red and blue (calmodulin). Putative position of KCNEs [187] (black circles) shown in right-hand image. For all figures, structures were obtained from the Rutgers/UC San Diego/SDSC (RCSB) protein data bank and viewed using Jsmol (Javascript) (this panel) or NGL Viewer [188, 189] (all other structure images).

Kv α subunits in the KCNQ (Kv7) subfamily are represented by 5 isoforms in the human genome, KCNQ1-5. KCNQ channels are notable for their slow gating kinetics, small and flickery unitary conductance, and relatively little inactivation. KCNQ1 is a ubiquitous channel which is both important for human cardiomyocyte repolarization, and prominent in the essential functions of many different types of non-excitable, polarized epithelial cells. These contrasting roles are made possible by interaction of KCNQ1 with a variety of β subunits, most notably the KCNE 1TM subunits [66].

Each of the five KCNE isoforms co-assembles with and regulates KCNQ1 with divergent functional outcomes. For example, KCNE1 slows KCNQ1 activation 5–10-fold, positive-shifts KCNQ1 voltage-dependence of activation such that more comprehensive membrane depolarization is required for channel activation (Figure 1A), removes inactivation [67–70], and increases sensitivity 100-fold to current potentiation by phosphatidylinositol 4,5-biphosphate (PIP₂) [71]. PIP₂ is a minor membrane phospholipid constituent that plays a major role in regulating membrane excitability by binding to and regulating the gating of many different ion channels [72–74], and features heavily in the discussions below. KCNE1 slows KCNQ1 activation by a mechanism involving probably both the transmembrane domain and C-terminal domain of KCNE1. One school of thought is that KCNE1 slows down movement of the charged, voltage-sensing S4 segment of KCNQ1 to slow its activation, possibly also involving regulation of how S4 interacts with S5, and formation of a stable pre-open state [75–78]; another school of thought is that in addition to S4 slowing, KCNQ1-KCNE1 channels require all 4 voltage sensors to activate before channel opening, while KCNQ1 activation without KCNE1 requires only 1 voltage sensor to move [79, 80]. KCNQ1-KCNE1 channels generate the slowly activating potassium current (I_Ks) that is important for cardiomyocyte repolarization, and also from a channel required for endolymph secretion in the inner ear. Hence, individuals with severe loss-of-function mutations, typically in both alleles, suffer from the cardioauditory Jervell and Lange-Nielsen syndrome, which presents as Long QT Syndrome (delayed ventricular repolarization) and congenital profound bilateral sensorineural hearing loss [39, 81–83]. KCNQ1 requires co-assembly with calmodulin for surface expression and normal function (Figure 1B). KCNE1 is thought to fit into a pocket between the VSD of one KCNQ1 subunit and the pore module of another, within I_Ks channel complexes (Figure 1C). It is likely that other KCNEs are similarly localized. Two KCNE1 subunits are thought to exist in each functional I_Ks complex with four KCNQ1 α subunits (Figure 1C) although some suggest this can be variable and rise to 4:4 stoichiometry [84–88].

In contrast, as mentioned above, KCNQ1-KCNE2 (and KCNQ1-KCNE3; Figure 1A) channels are constitutively active, and also non-inactivating, meaning that they can provide a constant repolarizing force when cells are positive to the potassium equilibrium potential (E_K), which is typically around −80 mV under physiological conditions. Thus, in polarized epithelial cells, which often remain at potentials between −40 mV and −20 mV, KCNQ1-KCNE2 can facilitate steady K⁺ efflux unless closed by other forms or regulation [40, 42, 43, 89]. KCNQ1-KCNE2 channels are required for gastric acid secretion (as explained above) and also for optimal thyroid hormone secretion; genetic deletion of either protein causes hypothyroidism [90–92]. KCNQ1 and KCNE2 are also expressed in pancreatic β cells, where they may also form a heteromeric channel complex. Germline deletion of Kcnq1 increases glucose tolerance in mice [93], whereas deletion of Kcne2 decreases glucose tolerance, decreases insulin sensitivity, and impairs insulin secretion in response to glucose [94, 95]. This hints at the complex effects of KCNE2 on KCNQ1: the channel is converted to a constitutively active one, but channel outward current is decreased (possibly by reduced surface expression and/or other mechanisms). Thus, KCNE2 co-assembly has opposite effects on KCNQ1 current magnitude depending on the membrane potential at which is it measured [89]. KCNE2 appears to be most highly expressed in the choroid plexus epithelium, where it regulates KCNQ1 and also the Kv1.3 (KCNA3) potassium channel on the apical side [96].

KCNQ1-KCNE3 channels are also constitutively activated and have been mechanistically studied more than KCNQ1-KCNE2, largely because they generate more robust currents, making it easier to study effects of mutagenesis, etc. We and others have shown that KCNE3 locks open the KCNQ1 VSD, via residues including V72 in the transmembrane segment and D54 and D55 in the KCNE3 N-terminal segment [97–104]. KCNQ1-KCNE3 generates a K⁺ current important for regulating cAMP-stimulated chloride secretion in the intestine [105]. KCNE4 and KCNE5, in contrast, each inhibit KCNQ1 activity; in the case of KCNE4 this effect involves calmodulin but probably not impaired trafficking [106–108], while the KCNE5 effect involves a positive-shift in the voltage dependence of activation [109]. For detailed reviews of KCNE family regulation of KCNQ1 and other α subunits, see [63–66].

KCNQ1 interaction with sodium-coupled myo-inositol transporters

The choroid plexus epithelium is the primary site for cerebrospinal fluid (CSF) production and secretion; deletion of KCNE2 depolarizes choroid epithelial cells, alters their cellular Kv currents, increases seizure susceptibility, and decreases immobile time in the tail suspension test – both of the latter suggesting increased neuronal excitability. Metabolomics analysis revealed that KCNE2 deletion in mice decreases CSF levels of the major osmolyte, myo-inositol, by ~30%. Mega-dosing with myo-inositol by intraperitoneal injections restored normal seizure susceptibility and tail suspension behaviour in Kcne2^−/− mice without affecting wild-type mice, suggesting a causative role for altered myo-inositol handling in the neural defects in Kcne2^−/− mice [96, 110]. Accordingly, KCNQ1-KCNE2 channels were found to form physical complexes with the sodium-coupled myo-inositol transporter, SMIT1 (encoded by SLC5A3), in mouse choroid plexus and when heterologously co-expressed in Chinese Hamster ovary (CHO) cells [110]. SMIT1 is predicted to be a ~14-transmembrane-segment protein (Figure 2A), closely related to the sodium-coupled glucose transporter, SGLT1 [25, 111–113], but favouring myo-inositol versus glucose as a substrate.

Figure 2. KCNQ-transporter interactions.

A. Transmembrane topologies of vSGLT (a representative of the SLC5A transporters) [190] and SLC26Dg (a representative of the SLC26 transporters) [191]. Substrates (magenta) shown in binding sites.

B. In this and subsequent panels, high-resolution structures obtained by X-ray crystallography or cryo-EM are shown for proteins known to participate in chansporter complexes, or structures of their closest relatives for which high resolution structural information is available. Unless otherwise stated, colouring progresses from (red to blue) N-terminal to C-terminal highlighting individual secondary structural elements such as individual transmembrane helices. This panel: Xenopus KCNQ1 with calmodulin bound [186] represents KCNQ1, 2, 3 or 4; vSGLT [190] is in place of SMIT1 or 2; Deinococcus geothermalis SLC26 [191] is in place of prestin. Red arrows indicate known or putative (“?”) direct physical interaction. Green lines approximate membrane boundaries. Also shown is the myo-inositol (MI) pathway, indicating how SMIT-transported myo-inositol can generate PIP₂ which then regulates KCNQ gating. Abbreviations: PI4K, phosphatidylinositol 4-kinase; PI5K, phosphatidylinositol 5-kinase; PLC, phospholipase C; DAG, diacylglycerol; IP₃, inositol(1,4,5)-triposphate, IP₂, inositol(1,4)-bisphosphate; IP, inositol monophosphate; IPP, inositol-1,4 bisphosphate 1-phosphatase; IMPase, inositol-1(or 4)-monophosphatase; G6P, glucose 6-phosphate; Ino-1, inositol synthase.

C. Hypothetical configuration of individual SLC26 and SLC5A transporters with a KCNQ1 tetramer viewed from outside the cell.

D. Hypothetical 4:4 stoichiometry (KCNQ1:SLC5A) viewed from outside the cell.

Myo-inositol performs a variety of crucial cellular functions aside from its osmoregulatory properties. Myo-inositol is converted to phosphatidylinositol (PI) and phosphatidylinositol 4,5-biphosphate (PIP₂) by phosphatidylinositol 4-kinase (PI4K) and phosphatidylinositol 5-kinase (PI5K). PIP₂ is a prerequisite for normal functioning of many ion channels (e.g., KCNQs, inward rectifier K⁺ channels, transient receptor potential (TRP) channels and ryanodine receptors), transporters (e.g., sodium/hydrogen exchangers), and pumps (e.g., plasmalemmal calcium pumps) [72–74]. In addition, Gq-coupled receptor stimulation activates phospholipase C (PLC), cleaving PIP₂ into inositol(1,4,5)-triposphate (IP₃), and diacylglycerol (DAG) (Figure 2B), each of which contributes to protein kinase C (PKC) activation and subsequent downstream signalling cascades. IP₃ also activates IP₃ receptors, releasing Ca²⁺ from intracellular stores such as the endoplasmic reticulum.

While KCNQ1 co-expression enhanced SMIT1 myo-inositol uptake, KCNQ1-KCNE2 inhibited it, as did co-expression with R231A-KCNQ1, a constitutively active mutant with the VSD locked in the active state [110]. This suggests a more complicated channel-transporter relationship than merely Kv channel cellular hyperpolarization increasing the electrical gradient to enhance Na⁺ coupled myo-inositol influx, as R231A-KCNQ1 would be predicted to serve this role very efficiently. R231A-KCNQ1 also inhibited activity of SMIT2, a SMIT1-related myo-inositol transporter encoded by SLC5A11, while SMIT2 inhibited KCNQ1 activity, when the two were co-expressed in Xenopus oocytes [110].

Later work showed that SMIT2 regulates KCNQ channels in mammalian and Drosophila neurons (see below). Further, the closely related KCNQ4 channel did not augment or otherwise alter SMIT1 activity when co-expressed in oocytes, also suggesting against solely the hyperpolarizing effects of KCNQ1 as being important for SMIT1 regulation. However, it appears that KCNQ1 pore activity is important for this role, as inhibiting with KCNQ channel blocker XE991 also inhibited co-expressed SMIT1 activity [110]. The author’s lab is currently investigating how SMIT1 co-expression alters KCNQ1 channel functional properties and pharmacology. Initial findings suggest intimate association of SMIT1 with the KCNQ1 channel pore module. Thus, SMIT1 co-expression, in the absence of extracellular myo-inositol or other SMIT1 substrates, changes ion selectivity, voltage-dependence and pharmacology of KCNQ1 in a KCNE1-dependent manner, consistent with a model in which SMIT1 physically interacts with the KCNQ1 pore or close enough to it to alter its fundamental attributes. This association may include altering the way in which KCNE1 modulates KCNQ1, or even competition between SMIT1 and KCNE1 for interacting with similar sites in the KCNQ1 pore or gating apparatus [114].

One could speculate that transporters such as SMIT1 might interact one per KCNQ1 tetramer, potentially leaving room for other types of transporter to co-assemble as well (Figure 2C), or 4:4 giving an octameric complexes with the tetramer of KCNQ1 α subunits (Figure 2D).

KCNQ1-KCNE2 channels also functionally interact with the sodium/iodide symporter (NIS; encoded by SLC5A5), a close relative of SMIT1. Germline deletion of either Kcnq1 or Kcne2 in mice causes hypothyroidism, and KCNQ1-KCNE2 channels are required for optimal thyroid cell iodide uptake by NIS, for the purpose of thyroid hormone production [90, 92]. The mechanisms by which KCNQ1-KCNE2 regulates NIS have yet to be fully established.

KCNQ2/3 interaction with myo-inositol transporters

While KCNQ1 is synonymous with human ventricular repolarization and roles in various secretory epithelia, the remainder of the KCNQ family members (2–5) are better known for generating the muscarinic-inhibited (M) current in mammalian neurons [115–119]. KCNQ2 and KCNQ3 in particular are expressed widely in the brain, where they from heteromeric Kv channels in axon initial segments (AIS) of neurons. In addition, KCNQ2 and KCNQ3 form channels alone or with one another in nodes of Ranvier in nerves including the sciatic nerve [120–123]. KCNQ2/3 channels act as gatekeepers, controlling neuronal excitability by raising the threshold for action potential firing. Upon activation of the muscarinic acetylcholine receptor, KCNQ2/3 channels are inhibited, depolarizing the cell sufficiently to initiate an action potential [121].

As was also found for KCNQ1 channels, KCNQ2/3 channels are regulated by PIP₂, which negative-shifts their voltage dependence such that they activate at more hyperpolarized membrane potentials, further quelling membrane excitability [73, 124–128]. Aside from being an important osmolyte, myo-inositol is a substrate for PIP₂, and therefore SMIT1 can increase cellular PIP₂ content by transporting myo-inositol into the cell. Dai and colleagues showed that this is one mechanism by which neuronal electrical activity can be linked to extracellular osmotic changes: increased extracellular myo-inositol is transported into the cell by SMIT1, raising PIP₂ levels, negative-shifting the voltage dependence of KCNQ2/3 channel activation, hyperpolarizing the cell and reducing cellular excitability [129].

We found that not only is SMIT1-transported myo-inositol able to generate sufficient PIP₂ to modulate KCNQ2/3 activity, as reported by Dai et al., but also that SMIT1 forms physical complexes with KCNQ2, and that KCNQ2 and KCNQ3 each co-localize with SMIT1 and/or SMIT2 in axon initial segments and sciatic nerve nodes of Ranvier [130]. Co-localization/co-assembly of KCNQ2/3 channels with SMIT1/2 may facilitate the channel being able exploit locally high PIP₂ concentrations generated from SMIT1/2-transported myo-inositol, an hypothesis supported by previous studies showing that PIP₂ diffusion is quite slow in the presence of intact cytoskeleton [131], and our finding that disruption of the cytoskeleton impairs the ability of SMIT1 to potentiate KCNQ2/3 current after 3 h incubation in 1 mM myo-inositol [130]. Reciprocally, KCNQ2/3 channels protect SMIT1 activity from the otherwise inhibitory effect of cellular depolarization via high extracellular K⁺ [130]. In tight confines such as axons, where diffusion might be slower for some molecules than in the soma, close physical proximity between KCNQ2/3 and SMITs could be particular advantageous, but what other advantages might KCNQ2/3-SMIT1/2 co-assembly confer? Ongoing work in the author’s laboratory suggests that SMIT1 binds directly to the KCNQ2 pore module, and in the absence of myo-inositol, SMIT1 binding itself alters KCNQ2/3 ion selectivity, gating and pharmacology, indicative of direct influence on pore architecture and fundamental channel properties [114]. Thus, SMIT-KCNQ co-regulation probably involves both direct physical contact and regulation by myo-inositol-derived PIP₂ (Figure 2B).

The physiological significance in vivo of the regulatory modes described above remains to be determined. However, there is a striking similarity between the phenotypes of mice with either Kcnq2 or Slc5a3 (SMIT1) germline-deleted: disruption of either gene causes death within hours after birth from hypoventilation [132, 133]. Likewise, some human KCNQ2 sequence variants also cause respiratory dysfunction [134]. In Slc5a3−/− mice, this phenotype can be avoided by supplementing the drinking-water of the dam during gestation and lactation with 1 mM myo-inositol. Thus, the myo-inositol transported into phrenic nerves by SMIT1 is required for normal development and function of peripheral nerves, and for innervation of the diaphragm and ventilation [135]. An interesting experiment would be to similarly try to rescue Kcnq2−/− mice using myo-inositol, but this may not work, as just having myo-inositol, but not the channel that the PIP₂ it generates presumably regulates in phrenic nerves (KCNQ2), may not suffice. Nevertheless, the phenotypic similarity of the two mouse lines, and the KCNQ2/3-SMIT1/2 co-localization we observed in sciatic nerve, are highly suggestive of requisite cooperation and chansporter complex formation in vivo. Unlike other family members, KCNQ4 and KCNQ5 have not yet been found to interact with SLC5A family transporters, but KCNQ4 is suggested to interact with SLC26A5 (Prestin) (see below).

dKCNQ interaction with cupcake

Vinegar flies (Drosophila melanogaster) express an orthologue of mammalian SMIT2 (gene name SLC5A11 in either case), termed cupcake [136]. Deletion of SLC5A11 renders Drosophila insensitive to the nutritional value of sugars. Cupcake is required for the increased excitability of ellipsoid body R4 neurons observed during fly starvation, which normally drives feeding and other hunger-driven behaviours. Cupcake expression in the brain, accordingly, is increased during starvation and decreased shortly after feeding of starved flies. It appears, however, that cupcake is not a sugar transporter, but rather it may act as a sugar sensor. Thus, unlike related mammalian transporters such as SGLT1, cupcake lacks a consensus sugar binding site. Neither does cupcake generate glucose-dependent currents when it is studied in Xenopus oocytes, although it does pass a large constitutive current that requires sodium and a conserved sodium-binding site [136, 137]. However, as described for mammalian SMITs above, cupcake co-assembles with dKCNQ, a Drosophila Kv channel orthologous to mammalian KCNQs. Cupcake co-assembly inhibits dKCNQ (as we previously found for the mammalian cupcake orthologue, SMIT2, with KCNQ1 [110]), and dKCNQ is required for normal feeding behaviour and food selection by controlling the excitability of neurons expressing cupcake [137].

3. SLC26A family anion transporter interactions with ion channels

Prestin interaction with KCNQ channels

KCNQ4 is a Kv channel α subunit structurally and functionally closely related to KCNQ2, 3 and 5. KCNQ4 is distinguished from these other KCNQ family members primarily by its expression in mammalian cochlea and auditory neurons, and its association when disrupted with human and mouse hearing loss. KCNQ4 is required for normal cochlear ion homeostasis and inherited sequence variants in KCNQ4 cause DFNA2, which presents as progressive sensorineural hearing loss at all frequencies [138]. Many of the KCNQ4 DFNA-linked mutations impair KCNQ4 trafficking to the plasma membrane and/or impair channel conductance when at the membrane [139, 140].

Prestin (encoded by SLC26A5) is part of the SLC26 family of anion transporters that also includes pendrin (encoded by SLC26A4) (Figure 2A). While pendrin is a more conventional, sodium-independent, electroneutral chloride/iodide transporter (anion exchanger), prestin is a motor protein in addition to being an anion transporter. Mechanical elongation and contraction of prestin underlies electromotility of outer hair cells of the inner ear, important for the amplification of sound waves that is required for high sensitivity hearing [141, 142]. Prestin is thought to be intrinsically voltage-sensitive by virtue of the ability to change shape in response to binding of intracellular chloride ions. Independent of this voltage sensing, prestin is also thought to have the capacity to transport anions, although was unable to do so when heterologously expressed in Xenopus oocytes, suggesting the possibility that oocytes lack an essential component required for the transport function of prestin [143–147].

Like KCNQ4, inherited sequence variants in SLC26A5, the gene encoding prestin, cause non-syndromic hearing loss [148, 149]. It is therefore very interesting that prestin expressed in CHO cells left-shifted the voltage dependence of activation of co-expressed KCNQ4, by about −15 mV; pendrin co-expression, in contrast, did not affect KCNQ4 activation. A similar −15 mV shift in the voltage dependence of activation was found when KCNQ2/3 channels were co-expressed with prestin, but a lesser effect (−5 mV) was observed when prestin was co-expressed with Kv1.1 (KCNA1) [150]. It is not yet known whether the functional effects of prestin on KCNQ channels arise from direct physical interaction (Figure 2B, C), nor whether the interaction occurs in native outer hair cells.

SLC26A chloride and bicarbonate exchanger interaction with CFTR

Prestin, pendrin, and also related SLC26 chloride and bicarbonate exchanger family members A2 and A3, form complexes with another ion channel - the cystic fibrosis transmembrane conductance regulator (CFTR) [151]. CFTR is a highly unusual protein that displays some functional properties of a chloride channel but also those of an ATP-binding cassette ABC transporter; while most ABC transporters operate by active transport, CFTR is a Cl⁻ channel through which Cl⁻ passes by diffusion [152, 153]. SLC26Ax-CFTR complexes are thought to form in, e.g., the pancreas and salivary glands, where the native epithelial chloride absorption and bicarbonate secretion characteristics are more similar to those of the complexes than of CFTR alone; dysfunction of the complexes may contribute to diseases including cystic fibrosis and congenital chloride diarrhoea [154, 155]. Thus, CFTR mutations cause cystic fibrosis [156–158]; SLC26A2, A3 and A4 mutations cause diastrophic dysplasia [159], congenital chloride-losing diarrhoea [160], and Pendred syndrome (deafness, goitre and organification defects) [161], respectively.

The molecular basis of SLC26A-CFTR interaction has been well studied. The STAS domain of some SLC26A transporters has been found to mediate interaction with CFTR via the phosphorylation site-rich R domain of the latter, assisted by PDZ ligand-mediated binding of both proteins to scaffolding proteins such as CAP70 (Figure 3A). When non-phosphorylated, the CFTR R domain binds to CFTR NBD1 to stop it interacting with NBD2, which would result in CFTR Cl- channel activity. The SLC26A STAS domain stabilizes this auto-inhibition. When phosphorylated by PKA, R domain-NBD1 interaction is altered and R domain-STAS domain binding is enhanced – this activates both CFTR and SLC26A anion transport leading to fluid and electrolyte secretion [152, 154, 162, 163]. The isolated STAS domain from SLC26A3 potentiates CFTR open probability (P_o) >threefold, and the number of active channels (n) × channel open probability (nP_o) sixfold, while leaving total surface CFTR protein levels unaffected. Similarly, CFTR co-assembly potentiates SLC26A6 activity [154, 155]. There, is, however, further complexity to STAS-CFTR interactions – for SLC26A3 or SLC26A6 this interaction results in reciprocal potentiation, but in the case of SLC26A9, CFTR is activated but SLC26A9 is inhibited by the interaction [164, 165].

Figure 3. ABC and ATPase chansporter complex interactions.

A. High-resolution structures obtained by X-ray crystallography or cryo-EM are shown for proteins known to participate in chansporter complexes, or structures of their closest relatives for which high resolution structural information is available. Red arrows indicate direct physical interaction. Green or pink lines approximate membrane boundaries.

A. Deinococcus geothermalis SLC26 [191] in place of mammalian SLC26A transporters (e.g., prestin, pendrin), the rat prestin STAS domain [192, 193] known to interact with CFTR, and human CFTR [153]. CAP70 (a PDZ-domain containing scaffolding protein) is shown to mediate the interaction. TMD, transmembrane domain; NBD, nucleotide binding domain.

B. Mouse/Syrian hamster Kir6.2-SUR1, for which the complex structure has been solved by cryo-EM [194]; and polycystin-2 (TRPP2) [195] in place of TRPM4, which interacts with SUR1 (probably not while it is in complexes with Kir6.2).

C. Known 4:4 stoichiometry (Kir6.2:SUR1) viewed from outside the cell, coloured (left) from N-terminal (red) to C-terminal (blue) or (center) by individual subunits; right, hypothetical 4:4 stoichiometry (TRP:SUR1) viewed from outside the cell, coloured by individual subunits.

D. Drosophila melanogaster Orai1 [196] and rabbit SERCA (Sarco/Endoplasmic Reticulum Ca²⁺-ATPase) [197] in place of SPCA2 (Secretory Pathway Ca²⁺-ATPase 2).

ABC and ATPase interactions with ion channels

Sulfonylurea receptor interactions with ion channels

Sulfonylurea receptors (SURs) are integral membrane proteins that hydrolyse ATP. SURs are members of the ABC superfamily of active transporters, but unlike the majority of family members, SURs are not active transporters, nor do they even pass ions themselves (unlike CFTR). Instead, SURs are best known for acting as regulatory subunits for K_ATP channels, which comprise an octamer of four SURs (Sur1 or 2) and four Kir6.1 or Kir6.2 α subunits (Figure 3B, C). K_ATP channels act as metabolic sensors in tissues including the pancreas and heart; ATP binding to the Kir6.x subunits closes the channel, MgATP binding to SURs counteracts this by inducing a conformational change in the channel [166]. It is interesting, therefore, to view K_ATP channels as another example of channel-transporter complexes, albeit the “transporter” belongs to a transporter family without actually transporting anything. The theme of the “transporter” in a chansporter complex not actually transporting anything when in the complex appears to also apply to the Drosophila cupcake (encoded by SLC5A11) protein related to mammalian SMIT2 (also encoded by SLC5A11) (see above).

SUR1 also participates in another type of channel complex, with TRPM4, a calcium-activated monovalent cation-nonselective channel (Figure 3B). SUR1/TRPM4 channels are not expressed under normal physiological conditions, but are upregulated in neurons, astrocytes and capillary endothelial cells following various types of brain or spinal cord injury, and are thus important mediators of cerebral oedema formation. Co-assembly with SUR1 renders TRPM4 channel activity sensitive to drugs that bind to SUR1, including inhibitors glibenclamide and tolbutamaide, and activator diazoxide [167]. TRPM4/SUR1 channels only activate after ATP depletion [168]. As is observed for K_ATP channels, one could speculate a 4:4 octameric stoichiometry for SUR1/TRPM4 channels (Figure 3C).

Ca²⁺-ATPase interaction with Orai1

The secretory pathway Ca²⁺-ATPase isoform 2 (SPCA2, encoded by ATP2C2) is a Golgi-localized P-type calcium ion transporter that has the capacity, as with other ATPases, to catalyse the hydrolysis of ATP into ADP and a free phosphate ion. In the case of SPCA2, this energy is used to power uphill transport of Ca²⁺ to prime the secretory pathway (Golgi and post-Golgi vesicles). SPCA2 was found to form physical complexes with Orai1, a Ca²⁺ channel better known for its role in the plasma membrane mediating store-operated Ca²⁺ influx by communicating with the ER-localized STIM1 protein, which senses ER Ca²⁺ depletion [169]. In contrast, Orai1-SPCA2 complexes mediate Ca²⁺ signalling without STIM1 binding, and are thought to operate store-independently.

SPCA2-Orai1 complexes influence Ca²⁺ regulation and tumorigenicity in human mammary epithelium [170]. SPCA2 acts in this context as an ion channel chaperone and store-independent activator of Orai1, via a C-terminal domain on the transporter [170]. This control of Orai1 activity by the SPCA2 C-terminus is regulated by the SPCA2 N-terminus. In addition, the SPCA2 N-terminal is also important for SPCA2 Ca²⁺ transport activity [170–173]. Interestingly, a prior study had shown that a 20 kDa C-terminal fragment of SPCA2 is expressed by a short transcript, under the control of the MIST1 transcription factor, in pancreatic acinar cells and various glandular tissues. Germline deletion of MIST1 in mice eliminated the short transcript and also disrupted Ca²⁺ signalling [174], shown subsequently to be because of the ability of the C-terminal SPCA2 fragment to regulate Orai1. There is debate about exactly how SPCA2 and Orai1 work together to mediate store-independent Ca²⁺ entry, but one hypothesis is that SPCA2 positioned in membranes of the secretory pathway physically and functionally interacts with plasma membrane Orai1 to promote Ca²⁺ influx through both proteins, from the extracellular side into the secretory vesicles and/or Golgi [173] (Figure 3D).

4. Amino acid transporter (SLC6A11 and SLC1A2) interactions with ion and water channels

MaxiK-GAT3

Although no functional effects have yet been reported for this interaction, a recent proteomic study uncovered complex formation between MaxiK (a.k.a. BK or Slo1, a Ca²⁺-activated K⁺ channel), and GAT3, an electrogenic sodium-dependent gamma-aminobutyric acid (GABA) transporter encoded by SLC6A11 in mouse brain. GAT3 probably transports 2 Na⁺, one GABA and one Cl⁻ per transport cycle, although the Cl- transport is under debate [175] (Figure 4A). Subsequent co-immunoprecipitation and co-localization studies using heterologous co-expression in human embryonic kidney (HEK) 293 cells supported the proteomic findings [176]. It will be fascinating to evaluate whether there are functional consequences of this interaction, which potentially has implications for regulation of neuronal excitability.

Figure 4. Amino acid transporter interactions with ion and water channels.

High-resolution structures obtained by X-ray crystallography or cryo-EM are shown for proteins known to participate in chansporter complexes, or structures of their closest relatives for which high resolution structural information is available. Red arrows indicate known or putative (“?”) direct physical interaction. Green lines approximate membrane boundaries.

A. Aplysia Slo1 channel [198, 199] and the SERT human serotonin receptor (SLC6A4) [200] in place of GAT3 (SLC6A11).

B. Human AQP4 [201], human EAAT1 [202] in place of EAAT2 (EAAT1 does not bind to AQP4), and mouse μ-opioid receptor [203].

AQP4-EAAT2

Deficiency in the water channel aquaporin 4 (AQP4) has long been known to downregulate glutamate uptake and expression of the sodium-dependent glutamate transporter, EAAT2 (Excitatory Amino Acid Transporter 2; also known as GLT-1, encoded by SLC1A2). This relationship is relatively specific as EAAT1 (GLAST, the sodium-dependent glutamate aspartate transporter, encoded by SLC1A3) was reportedly not affected by AQP4 deficiency [177]. Very recently, AQP4 was found to form a tripartite physical complex with EAAT2 and the μ-opioid receptor in astrocytes (Figure 4B). The co-assembly between AQP4 and EAAT2 is mediated by a region within amino acids 252–323 of the AQP4 C-terminus. Interestingly, morphine was found to regulate expression of the complex via activation of protein kinase C, and it is suggested that the complex might play a role in morphine dependence [178]. In another recent study, the astrocytic Fc receptor was found to be required for autoantibody-induced internalization of AQP4 and EAAT2, a finding which may aid development of therapies to treat neuromyelitis optica, an autoimmune disease in which AQP4 autoantibodies cause CNS immunopathology and secondary CNS demyelination [179]. The findings of these two studies together with others described above underline the potential importance of chansporter complexes and their disruption to human disease processes.

5. Sodium-chloride symporter (SLC12A3) interaction with ENaC

The sodium-chloride symporter, NCC (encoded by SLC12A3) is another secondary active transporter, which uses the sodium ion gradient across the apical membrane of distal convoluted tubule cells to electroneutrally transport sodium and chloride ions into the cells from tubular fluid [180]. NCC forms physical complexes in distal convoluted tubule cells with ENaC, an epithelial sodium channel that also brings Na⁺ into distal convoluted tubule cells, by diffusion down an electrochemical gradient [181]. NCC and ENaC reciprocally regulate one another, and blocking NCC using thiazide also impairs co-expressed ENaC function, by reducing its open probability [182].

6. Conclusions and Perspectives

New ion channel-transporter (chansporter) complexes are being discovered at an increasing rate. Chansporter complexes appear to represent a novel paradigm in cell signalling, facilitating crosstalk between channels and transporters and their responses to voltage, membrane lipids, osmolarity, and various ions and other solutes. In some cases, the “transporter” lacks the capacity to transport, (e.g., SUR1); in other cases small fragments of the transporter can achieve regulatory effects upon the channel in the absence of the rest of the transporter (e.g., SPCA2). Outstanding questions include the stoichiometry and specific interaction interfaces of chansporter complexes, which are mostly unknown, and how widely these complexes are in utilized in mammalian cell signalling. For instance, do the majority of channels or transporters participate in these types of complexes? Also, what were the primary evolutionary driving forces that favoured selection of the ability to form specific chansporter complexes? The answers will differ depending on the specific complex but there may be patterns or family-specific generalities regarding which specific evolutionary advantages might be conferred by formation of each of the various chansporter complexes. Several hypotheses are given below.

1. Nanodomains. Chansporter complexes may act in nanodomains that exist to reduce spatiotemporal barriers between different signalling mechanisms. Diffusion of substrates imported by the transporters may be important in some cases for regulating co-assembled channel function. In these instances, it may be advantageous to position specific channel-transporter pairs close to one another. In the case of myo-inositol transporters such as SMIT1, for example, the myo-inositol they bring in is converted to PIP₂, which happens to be a potent regulator of channels that form complexes with SMIT1. Because PIP₂ diffusion within the membrane is quite slow, KCNQ-SMIT proximity could help ensure locally high PIP₂ concentrations for activation of KCNQ channels. This may be especially important in constrained spaces such as axons [130].

2. Biosensor sharing. Channels and transporters might co-assemble because it is advantageous to exploit one another’s biosensory capabilities. This is particularly obvious in the cases where one of the partners has lost its transport capabilities. In the case of K_ATP channels, although SURs do not possess transport machinery, they sense the metabolic state of the cell and report this to the Kir6 channels to regulate their function [166]. Likewise, cupcake may not possess sugar transport capabilities, but it is suggested to sense sugar levels [136, 137] and it is possible that via co-assembly, it can communicate these to dKCNQ to regulate feeding behaviour.

   KCNQ1 is adept at sensing its immediate environment, and it is possible that by responding to this in the form of altered channel activity and/or conformation, it may be able to regulate co-assembled SMIT activity and thus act as a biosensor for its transporter partner. KCNQ1 requires bound calmodulin for channel trafficking and function. One possibility is that Ca²⁺ binding to KCNQ1-bound calmodulin exerts conformational effects that can be transmitted to co-assembled SMITs to Ca²⁺-dependently regulate myo-inositol intake. KCNQs, and all other voltage-gated ion channels, sense membrane potential via their VSDs. Thus, KCNQs can communicate information about the membrane potential of the cell to SMIT1 via either S4 position or channel activity, or perhaps both. Thus, we found that KCNQ2/3 co-expression enabled SMIT1 to maintain its transport activity during cellular depolarization, whereas without KCNQ2/3, cellular depolarization lowered SMIT1 transport activity [130]. In contrast, locking the KCNQ1 S4 open (to mimic aspects of cell depolarization without actually depolarizing the cell) inhibited activity of co-expressed SMIT1 [110]. In addition, the sensitivity of KCNQ1 to extracellular pH (pH_o) is modulated by co-expressed KCNEs. For example, homomeric KCNQ1 is inhibited by low pH_o whereas KCNQ1-KCNE2 activity is potentiated by low pH_o [43]. It is possible, then, that depending on the co-assembled KCNE subunit, KCNQ1 could transmit information about extracellular pH to SMITs or other co-assembled transporters to regulate their activity.

3. Feedback. As described above, KCNQ1 activity is sensitive to PIP₂ concentration, and may therefore communicate information about myo-inositol-derived PIP₂ levels to bound SMITs to regulate their myo-insoitol uptake activity. This could be important in certain circumstances, e.g., if a strong downhill Na⁺ gradient was favouring excessive myo-inositol uptake but local PIP₂ levels were already inappropriately high. Co-assembly with different KCNEs can change the PIP₂ sensitivity of KCNQ1, the established example being that KCNE1 increases PIP₂ sensitivity 100-fold compared to KCNQ1 alone [71]. Thus, cells may be able to tune how KCNQ1 controls SMIT activity by differentially expressing the various KCNEs.

4. Tuning of channel properties. We recently found that SMIT1 can alter the ion selectivity, gating kinetics and pharmacology of co-expressed KCNQs, via direct interaction with the pore module [114]. This is good evidence of close physical interaction, but some of these modifications may also serve a physiological role. We found that SMIT1 increased KCNQ Na⁺ permeability, which could alter action potential morphology or tune the resting membrane potential in the cells where these complexes occur. Interestingly, SMIT1 co-assembly increases KCNQ2/3 Cs⁺ permeability, potentially resolving a long-standing discrepancy between the relative permeability for Cs⁺ of cloned KCNQ2/3 versus native M current generated by this channel [114]. In addition, modification of KCNQ gating kinetics by SMIT1, which in the case of KCNQ1 was KCNE1-dependent, could also modify action potential morphology. We do not yet know whether addition of substrate changes the effects of SMIT1 on KCNQ channel properties, but this could add another dynamic mode of regulation or feedback.

5. Location and surface stability. Finally, membrane proteins each contain within their primary sequences information that determines how, when and where they traffic to the cell surface, how long they remain there, and what is their fate once they are recycled from the plasma membrane. When one or more different membrane proteins co-assemble, they can thus influence one another’s fate. In K_ATP channels, for example, Kir6 and SUR subunits contain ER retention motifs that dictate they can only reach the surface if co-assembled [183, 184]. For other chansporter complexes, the effects may be more subtle, but if complexes form early in biogenesis, could ensure that a specific channel only traffics to a particular axonal region or side of a polarized cell, when a specific transporter partner co-assembles with it. KCNEs can also dictate Kv α subunit subcellular localization, subunit composition, and endocytosis rate, and therefore could also dictate these same properties for specific chansporter complexes if the channel were the dominant partner in this respect. In this model, expression by the cell of specific channel subunit combinations could ensure that a partnered transporter reached the appropriate part of a cell to fulfil a specific task, in a specific manner dictated by the channel, and could also regulate how long the transporter remained in that location. Of course, the reciprocal is equally likely to be the case.

High-resolution structures are now available for either the proteins, or their close surrogates, within many of the known chansporter complexes (Figures 1–4). Thus, future in silico modelling guided by site-directed mutagenesis and functional assays, and vice-versa, may help elucidate how chansporter complexes fit together, how the partners regulate one another, and the reasons why complex formation offers an advantage over more conventional, long-range signalling.

Nonstandard abbreviations

ABC

ATP-binding cassette

CFTR

cystic fibrosis transmembrane conductance regulator

CSF

cerebrospinal fluid

ENaC

epithelial sodium channel

NCC

sodium chloride cotransporter

NIS

sodium iodide symporter

PI

phosphatidyl inositol

PIP₂

phosphatidylinositol 4,5-bisphosphate

SGLT1

sodium-dependent glucose transporter 1

SMIT1

sodium-dependent myo-inositol transporter 1

SMIT2

sodium-dependent myo-inositol transporter 2

SPCA2

secretory pathway Ca²⁺-ATPase isoform 2

STIM1

stromal interaction molecule 1

SUR

sulfonylurea receptor