Advances in the development of novel compounds targeting cation-chloride cotransporter physiology

Abstract

For about half a century, the pharmacology of electroneutral cation-chloride cotransporters has been dominated by a few drugs that are widely used in clinical medicine. Because these diuretic drugs are so good at what they do, there has been little incentive in expanding their pharmacology. The increasing realization that cation-chloride cotransporters are involved in many other key physiological processes and the knowledge that different tissues express homologous proteins with matching transport functions have rekindled interest in drug discovery. This review summarizes the methods available to assess the function of these transporters and describe the multiple efforts that have made to identify new compounds. We describe multiple screens targeting KCC2 function and one screen designed to find compounds that discriminate between NKCC1 and NKCC2. Two of the KCC2 screens identified new inhibitors that are 3–4 orders of magnitude more potent than furosemide. Additional screens identified compounds that purportedly increase cell surface expression of the cotransporter, as well as several FDA-approved drugs that increase KCC2 transcription and expression. The technical details of each screen biased them toward specific processes in the life cycle of the transporter, making these efforts independent and complementary. In addition, each drug discovery effort contributes to our understanding of the biology of the cotransporters.

INTRODUCTION: THE TARGETS

The era of molecular biology not only provided clarity about the identity of several known transport mechanisms, but also led to the discovery of many novel ion channels and transporters. The cloning era clarified the composition of the electroneutral cation-chloride cotransporter family, with some members already known as functional units and new members discovered by homology cloning (1–6). The importance of cation-chloride cotransporters in physiology is supported by their link to human diseases and the development of genetically modified mouse models (Table 1). One key parameter that defines the functional properties of these transporters is their sensitivity to inhibitors and/or activators. The multiplicity of genes with very similar transport function, but somewhat specialized roles due to gene-specific tissue expression, highlights the need to develop new research tools and therapeutic agents. This review article was written to highlight the physiology and pathology of these transporters, describe the methodologies available to study their function, and describe recent drug discovery efforts.

Table 1.

SLC12 transporters, human diseases, mouse phenotypes, and historical inhibitors

Gene	Transporter	Disease	Mouse Phenotype	Inhibitor	References
A1	NKCC2 ↓	Bartter	Lethality, polyurea	Furosemide	(7, 8)
A2	NKCC1 ↓	Deafness, lung and intestinal obstruction, saliva production	Deafness, intestinal obstruction, saliva production, male sterility, etc.	FurosemideBumetanide	(9–14)
A3	NCC ↓NCC ↑	GitelmanGordon	NaCl wasting, hypokalemia	Thiazides	(15, 16)
A4	KCC1	None	None	Furosemide
A5	KCC2	SeizuresPossibly autism	Lethality, brain hyperexcitability, clonic/tonic seizures	Furosemide	(17–21)
A6	KCC3 ↓KCC3 ↑	HSMN/ACC	Peripheral neuropathyPeripheral neuropathy	Furosemide	(22–24)
A7	KCC4	None	Deafness, renal tubular acidosis	Furosemide	(25)

HSMN/ACC, hereditary sensorimotor neuropathy/agenesis corpus callosum; KCC1–4, K-Cl cotransporter-1–4; NCC, Na-Cl cotransporter; NKCC1, Na-K-2Cl cotransporter-1; NKCC2, Na-K-2Cl cotransporter-2.

The first cation-chloride cotransporter gene is solute carrier family 12 A1 (SLC12A1), which encodes for the Na-K-2Cl cotransporter 2, or NKCC2. This protein is almost exclusively expressed in the thick ascending limb of Henle (hTAL), where it participates in 20%–25% of Na⁺-reabsorption. Recessive loss-of-function mutations in SLC12A1 result in Bartter syndrome, a salt wasting disorder associated with hypokalemic alkalosis, low blood pressure, and hypercalciuria (7). The NKCC2 knockout mouse closely recapitulates the human phenotype with severe polyuria leading to perinatal lethality. The early death can be minimized with treatment with indomethacin (8), a nonsteroidal anti-inflammatory drug that also helps patients with Bartter syndrome (26). Cotransporter function is reduced or eliminated with furosemide, bumetanide, or analogs, which are collectively referred to as “loop” diuretics.

SLC12A2 encodes a second Na-K-2Cl cotransporter: NKCC1. Compared with NKCC2, which has a highly restricted pattern of expression, NKCC1 is expressed widely throughout the body. It is found in many blood cells, in muscle, heart, brain, intestine, lung, etc. (Fig. 1). It is therefore not surprising that the NKCC1 knockout mouse displays a plethora of phenotypes including deafness and imbalance (inner ear deficit), intestinal obstruction, lack of saliva secretion, pain perception deficit, and male sterility (9, 10, 13, 27–29). As summarized in a recent review, deleterious mutations in NKCC1 have now been found in humans, leading to disorders reminiscent of the phenotypes observed in the genetically modified mouse models (30). These disorders include inner ear deficit, muscle weakness, lung and intestinal obstruction, and abnormal mucus secretion (11, 12, 14, 31). In addition, SLC12A2 is a haplo-insufficient gene, also linked to neurodevelopmental disorders such as intellectual disability and autism (32). In fact bumetanide, which also inhibits NKCC1, is currently tested in clinical trial for autism (33–35). Because inhibition of NKCC2 in the thick ascending limb of Henle results in enhanced K⁺ secretion in the distal tubule, close attention was being paid to the children’s serum potassium levels and when needed, K⁺ supplementation was given to prevent severe hypokalemia. The development of an inhibitor discriminating NKCC1 and NKCC2 would alleviate this unwarranted complication.

Figure 1.

Message RNA abundance of the seven cation-chloride cotransporters in tissues. Representative panels showing the abundance of Na-K-2Cl cotransporter-1 (NKCC1), Na-K-2Cl cotransporter-2 (NKCC2), Na-Cl cotransporter (NCC), and K-Cl cotransporter-1–4 (KCC1-4) in 19 selected tissues. As highlighted by the arrows, NKCC2 and NCC are almost exclusively expressed in kidney, whereas KCC2 expression is abundant in the central nervous system. Values are given as Normalized eXpression (NX). Note that NX values for NKCC2 and NCC in kidney far exceed the scale, which is represented by the broken bars. The NX values are given for these two bars. Data were collected and redrawn from the Human protein atlas (https://www.proteinatlas.org/).

SLC12A3 encodes the distal convoluted tubule Na-Cl cotransporter, NCC. Absence of a key tyrosine residue coordinating K⁺ in NKCC and KCC (36, 37) makes the transport of Na⁺ and Cl⁻ through NCC independent of K⁺. Recessive loss-of-function mutations in SLC12A3 result in Gitelman syndrome, another salt wasting disorder associated with hypokalemic alkalosis (16). Several mouse models recapitulate the Gitelman phenotype: the Slc12a3 knockout mouse (15); the knockout of sterile 20-related proline-alanine-rich kinase (SPAK), a kinase that binds, phosphorylates, and activates NCC (38, 39); and the knockout of with-no-lysine kinase 4 (WNK4), a kinase that acts upstream of SPAK (40). Interestingly, gain of NCC function is observed in pseudohypoaldosteronism type II (PHAII) (41), a disorder caused by mutations in several proteins regulating NCC function. These proteins are the WNK4 and WNK1 kinases (41, 42) and culin-3 and Kelsh-like protein 3, proteins part of E3 ligase complex (43, 44). Patients with PHAII have increased blood pressure accompanied by hyperkalemia and metabolic acidosis. This disorder is also mimicked in mouse models leading to increased WNK4 expression (45) or increased SPAK activity (46). NCC is inhibited by thiazide diuretics such as hydrochlorothiazide and metolazone.

SLC12A4 encodes the first Na⁺-independent K-Cl cotransporter, KCC1 (4). KCC1 shares only 33%–36% amino acid identity with the three Na⁺-dependent cation chloride cotransporters, which have 75% amino acid identity among themselves. This clearly places this cotransporter in a subfamily apart from NCC, NKCC1, and NKCC2. This gene is expressed in many tissues with highest expression levels in smooth muscle cells and breast tissue (Fig. 1). Function of KCC1 is absent or negligible under isosmotic conditions and is activated under hypotonic, cell swelling, conditions or following N-ethylmaleimide (NEM) treatment (4). The mouse knockout exhibits no overt phenotype, unless another K-Cl cotransporter is also eliminated. Red blood cells lacking KCC1 and KCC3 (see SLC12A6) exhibit increased mean corpuscular volume leading to their increased susceptibility to osmotic lysis (47). Functional redundancy indicates the importance of KCC function in the volume control of red blood cells.

The SLC12A5 gene encodes KCC2, a K-Cl cotransporter mostly restricted to central neurons (Fig. 1). Compared with the other three K-Cl cotransporters, KCC2 exhibits basal transport activity under isosmotic conditions. The transporter serves as the “Cl⁻ pump” driving intracellular Cl⁻ well below its electrochemical potential equilibrium, thereby facilitating γ-aminobutyric acid (GABA)-mediated hyperpolarization and inhibition (48, 49). KCC2 transports Cl⁻ out of the cell using the energy of the large K⁺ gradient maintained by the Na⁺/K⁺ pump. KCC2 readily prevents Cl⁻ loading during sustained GABA release due to rapid and repetitive neuronal activity (50). Consequently, decreased KCC2 function leads to a significant increase in neuronal activity/firing (51). Although the full knockout of KCC2 in mice shows postnatal lethality due to respiratory failure (52), the knockout of the major isoform (KCC2b) survives for 2 wk after birth and demonstrates touch-evoked tonic-clonic seizure-like behavior consistent with brain hyper-excitability (21). In addition, seizures can be observed in older heterozygous animals, compared with their wild-type counterparts. Mutations in human KCC2 have been found in patients with epilepsy (18, 19). The development of the GABA inhibitory system is also critical for the proper development of brain, especially during the perinatal period. Thus, it is not surprising that mutations in KCC2 have also been reported in patients with neurodevelopmental disorders, like autism spectrum disorder (53). KCC2 expression levels are significantly reduced in patients with Rett syndrome (54, 55) and in the methyl CpG binding protein 2 (Mecp2) animal model of Rett syndrome (56). In addition, restoring KCC2 function rescues behavioral deficits in that animal model (57).

SLC12A6 expresses the K-Cl cotransporter-3 or KCC3. Once cloned and mapped to human chromosome 15 (5, 58, 59), it took only two years to link the transporter to a crippling peripheral neuropathy disorder predominant in Quebec (Canada) and to confirm the phenotype through the development of KCC3 knockout mice (22, 23). The disorder, originally known as Andermann syndrome or agenesis of corpus callosum with peripheral neuropathy, is now called hereditary sensorimotor neuropathy with agenesis of corpus callosum or hereditary sensorimotor neuropathy/agenesis corpus callosum (HSMN/ACC). Although the exact mechanism by which dysfunction in KCC3 results in HSMN/ACC, it is clear that it involves neurons (60, 61) and the maintenance of their volume (62, 63). Additional patients have been identified (64), including single copy SLC12A6 variants in patients with peripheral nerve disease (65). In addition, a patient and its mouse model with gain-of-function mutations in KCC3 also suffer from sensorimotor peripheral neuropathy (24). The mutation involves a key phosphorylation residue Thr991, which when substituted to alanine renders the transporter constitutively active.

SLC12A7 encodes for the fourth isoform of K-Cl cotransport, KCC4. The transporter is expressed in many tissues, although at relatively low levels (Fig. 1). The knockout mouse was generated in 2002 and showed early-onset deafness and renal tubular acidosis (25). Immunolocalization of the cotransporter in the inner ear revealed expression at the base of outer hair cells and in supporting Deiter’s cells. The transport of K⁺ at the base of hair cells is critical for K⁺ recycling and the maintenance of the endochoclear potential. In the kidney, the cotransporter is localized to the basolateral membrane of intercalated cells of type A, or acid producing cells. Absence of KCC4 leads to accumulation of intracellular Cl⁻ resulting in inhibition of acid secretion and an alkaline urine (25). To date, there are no human disease-associated mutations in SLC12A7.

TECHNIQUES TO MEASURE CATION-CHLORIDE COTRANSPORTER FUNCTION

Direct Measurement of Ion Transport

Radioisotopes.

Isotopes have been extensively used in biology and physiology since the 1950s. Relevant isotopes for cation-chloride cotransporters are those tracing Na⁺, K⁺, and Cl⁻ transport. Note that a limited number of ionic species can substitute for the natural substrates of the cotransporters. On the cationic side, the Na⁺ binding site is rather selective, compared with the more promiscuous K⁺ binding site. Indeed, Na⁺ can hardly be replaced by Li⁺. This is demonstrated in Fig. 2A, where K⁺ influx through NKCC1 occurs in Li⁺ at a rate that is 26% of the rate in Na⁺. The consensus in the field is that K⁺ can readily be substituted by Rb⁺ with little to no effect on transport activity. As seen in Fig. 2, B and C, Tl⁺ is also transported through NKCC1 and KCC2, although at about half the rate of K⁺. At this point, it is unclear whether Tl⁺ can also substitute for Na⁺, as the cation utilizes the K⁺ binding site. Interestingly for neuroscientists using Cs⁺ in their perforated patch pipette solution, the heavier cation moves easily through KCC2 as seen in Fig. 2D. This finding is a bit surprising, as the ionic radius of Cs⁺ is substantially greater than that of K⁺ and the coordination site in the cryo-EM structure seem to be rather tight (Fig. 3). On the anionic side, it is known that Br⁻ is slightly better transported than Cl⁻ by the sheep red blood cell K-Cl cotransporter (66). In contrast, as indicated in Fig. 2E, K⁺ flux by NKCC1 in the presence of Br⁻ is reduced, compared with Cl⁻. Relevant for those thinking about using I⁻ as strong interactors of halide sensitive dyes, this anion is poorly transported by the Na-K-2Cl cotransporter (Fig. 2E).

Figure 2.

Substitution of cations and anions as substrates for Na-K-2Cl cotransporter-1 (NKCC1) and K-Cl cotransporter-2 (KCC2). A: bumetanide-sensitive K⁺ influx in Xenopus laevis oocytes injected with mouse NKCC1 cRNA and incubated in either 96 mM NaCl, 96 mM LiCl, or 96 mM N-methyl glucamine (NMDG)Cl. All solutions had 4 mM KCl and 5 µCi/ml ⁸⁶Rb. B: Xenopus oocytes injected with mouse NKCC1 cRNA were fluxed either in 96 mM NaCl with either 2 mM K₂(SO₄) or 2 mM Tl₂(SO₄) and 5 µCi/mL ⁸⁶Rb. C: HEK293 cells overexpressing rat KCC2 were fluxed in 96 mM NMDGCl (Na⁺-free solution) with either 2 mM K₂(SO₄) or 2 mM Tl₂(SO₄) and 1 µCi/mL ⁸⁶Rb. D: Xenopus oocytes injected with rat KCC2 cRNA were fluxed either in 96 mM NaCl with 1 and 8 mM KCl or CsCl and 5 µCi/mL ⁸⁶Rb. E: X. laevis oocytes injected with mouse NKCC1 cRNA were fluxed in 96 mM NaCl, 96 mM NaBr, or 96 mM NaI. All solutions had 4 mM KNO₃ and 5 µCi/mL ⁸⁶Rb. For oocytes, bars represent means ± SEM (n = 20-25 oocytes). For HEK293 cells, bars represent means ± SEM (n = 3 dishes).

Figure 3.

Atomic sizes of monovalent cations and anions. Left: for each ion, the atomic number is located in the top left corner, while the atomic radius (Van der Waals) in picometer (pm, or 10⁻¹² m) is placed in the bottom right corner. The size of the drawn ion is proportional to its atomic radius. Lines (red) are placed around the size of the Rb⁺ ion, which is transported by Na-K-2Cl and K-Cl cotransporters. Cs⁺ is much larger, whereas K⁺ has an equivalent size. Na⁺, Tl⁺, and Li⁺ are much smaller. Note that cations are alkali metals (column 1 in periodic table), except for Thallium, which is a post-transition metal (column 13 in periodic table). The anions are generally smaller with I⁻ of a size similar to Na⁺. Br⁻ and Cl⁻ are slightly smaller and Fl⁻ is much smaller than I⁻. The atomic radius values were obtained from the Los Alamos National Laboratories (https://periodic.lanl.gov/list.shtml). Right: ribbon structure of portions of TM1, TM3, and TM6 showing the location of ions in the cryo-EM structure of hKCC1, PDB ID: 6KKT (37). K⁺ is in blue, whereas Cl⁻ ions are in green. The tyrosine residue located in TM3 which coordinate K⁺ binding is drawn as stick.

Sodium.

A short-lived sodium isotope (²⁴Na) was first used to follow Na⁺ in animal and human tissues (67–69), including urinary Na⁺ excretion (70). The isotope was also used to trace Na⁺ movement through the Na⁺/K⁺ pump (71, 72), and the Na-K-2Cl cotransporter (73). Its use was rapidly replaced by ²²Na which has a much longer half-life (Table 2). This specific isotope has been used to trace Na⁺ movement through the Na⁺/H⁺ exchanger (74), the Na-K-2Cl cotransporter (75), the Na-Cl cotransporter (76), and the Na-HCO₃ cotransporter (77).

Table 2.

Radioactive isotopes of Na⁺, K⁺ (Rb⁺), and Cl

Isotope	Atomic No.	Half-Life	Decay Mode	Gamma Intensity
24Na	11	15 h	β− (1.39 MeV)	Yes
22Na	11	2.58 yr	β+ (89%, 0.54 MeV), ε (11%)	Yes
42K	19	12.4 h	β− (82%, 3.53 MeV) (18%, 2.03 MeV)	Low
43K	19	22.4 h	β− (0.83 MeV)	Yes
83Rb	37	83 days	ε (0.9 MeV)	Yes
86Rb	37	18.77 days	β− (1.77 MeV)	Low
36Cl	17	3 × 105 yr	β− (98%, 0.71 MeV), ε (2%)	No

Atomic number or number of protons in the nucleus of the atom. Half-life is the rate of radioactive decay or time it takes for activity to be reduced by ½. Decay mode: β⁻, negative beta emission or energetic electron; β⁺ positron emission, ε, orbital electron capture or absorption of an inner atomic electron. Energy of decay is given in parenthesis. Where more than one decay exists, separate energy values and distribution are given. The last column indicates whether gamma ray energy is released.

Potassium.

Isotopes of potassium (⁴²K and ⁴³K) were originally used to label erythrocytes (78), trace the movement of the cation from blood to cerebrospinal fluid (79), measure the K⁺ permeability in cortical segments of the nephron (80), study the distribution of body K⁺ (81), or examine the stoichiometry of the sodium pump (71). Because of their short half-lives (12.4 h for ⁴²K and 22.4 h for ⁴³K), the use of these isotopes was cumbersome, as radioactive decay needed to be included in all flux calculations. Replacement of radioactive potassium isotopes with ⁸⁶Rb, which has a half-life of 18.6 days (Table 2), greatly facilitated these experiments. Although it was shown that ⁸⁶Rb had its own characteristic behavior and was not suitable for tracing potassium throughout the body (82), it is wildly used to trace the transmembrane movement of K⁺ through pumps, channels, and transporters (83–90). Unfortunately, Perkin Elmer discontinued the production and sale of ⁸⁶Rb in 2019. Since then, our laboratory has been using ⁸³Rb, an isotope with a half-life of 83 days (see Table 2). The isotope is produced by the Brookhaven National Laboratory and available through Oak Ridge National Laboratory.

Chloride.

In 1964, Goodford (91) used ³⁶Cl to confirm that smooth muscle cells of the guinea pig colon had a relatively high Cl⁻ content. The isotope was also used to trace Cl⁻ movement during intestinal fluid secretion (92), mammary epithelium milk secretion (93), and chloride transport in erythrocytes (94). As for Na⁺ and K⁺ isotopes, ³⁶Cl has been extensively used as a tracer to measure the unidirectional movement of the halide through transport proteins. For transporters that carry multiple ions, e.g., Na-K-2Cl, Na-Cl, and K-Cl cotransporters, ³⁶Cl is useful to measure stoichiometry of transport. However, because the specific activity of ³⁶Cl is so much lower than that of ⁸⁶Rb (due to the high Cl⁻ concentration vs. low K⁺ concentrations in external solutions), the use of the ⁸⁶Rb is preferred. All radioactive isotopes relevant to cation-chloride cotransporters are listed in Fig. 4.

Figure 4.

Methods to assay the transport of Na⁺, K⁺, Cl⁻ through Na-Cl cotransporter (NCC), Na-K-2Cl cotransporter (NKCC), and K-Cl cotransporter (KCC). Isotopes of Na⁺, Cl⁻, K⁺, and Rb⁺ are used to trace the transport of ions through the cotransporters. Isotopes of Rb⁺ are used to trace K⁺ movement. Nonradioactive Rb⁺ (⁸⁵Rb⁺) can also be used as a tracer as the cold isotope is not naturally present in biological fluids. It can be detected using flame or atomic absorption spectrometry. Note that tissue/cell Na⁺ and K⁺ contents can also be determined by flame or atomic absorption spectrometry. Fluorescent indicators are also widely used. Indicators sensitive to Tl⁺ (another non-biological cation mimicking K⁺), Na⁺, K⁺, and Cl⁻ are listed.

Atomic absorption spectrometry.

Adaptation of atomic absorption spectrometry to biological samples was originally made by Walsh in 1955 (95). The concentration of specific cations in a solution can be determined by first vaporizing (atomizing) the sample via a flame or an electrothermal graphite tube, then emitting light through the gas sample to measure its absorption by specific elements/atoms. Atomic absorption measures the amount of energy absorbed by a sample and this amount at discrete wavelengths (ion signature) is proportional to the concentration. The method was first used to measure amounts of calcium and magnesium in biological samples (96, 97). The movement of ions such as Na⁺ and K⁺ through biological membranes was also followed using atomic absorption spectrometry (98, 99). Nonradioactive Rb⁺ is also used as a tracer for K⁺ movement across membranes via quantitation using atomic absorption spectrometry (100, 101). The cold rubidium tracing method is preferred by those who cannot use radioactive materials. Its use will likely grow, as radioisotopes of Rb⁺ are increasingly difficult to obtain.

Fluorescent Indicators/Dyes

Sodium dyes.

The intracellular Na⁺ content of cells or tissues has been first determined using a variety of methods, including X-ray microanalysis, isotope equilibration methods, and flame absorption spectrometry. Although X-ray microanalysis was useful in comparing Na⁺ content from tissue to tissue and, for example, determine that tumor cells had a significantly higher Na⁺ content than normal cells (102), the reported values are expressed as a function of dry weight and not concentrations. Other methods like flame/atomic absorption spectrometry also require the knowledge of the water content of a cell/tissue to calculate its ion concentration. It also requires the knowledge of the extracellular fluid contribution, as the cation is abundant in the extracellular space. Determination of extracellular space contribution is typically made using membrane impermeant markers such as radiolabeled sucrose or inulin (103). Na⁺ concentrations in cells ranges from 5 mM to 15 mM (103), but local Na⁺ concentrations can transiently reach 40 mM in neuronal dendrites (104).

Multiple compounds that change fluorescence intensity with changes in intracellular Na⁺ have been developed (Fig. 4). They are SBFI (a benzofuranyl fluorophore linked to a crown ether chelator) (104, 105), CoroNa green (106), sodium green (107), and Ansante Natrium Green-2 or ANG-2 (108, 109). Although Na⁺ dyes are not completely insensitive to K⁺ and are often sensitive to changes in intracellular pH, ionic strength, and viscosity, they are useful in following changes in intracellular Na⁺. Note that unlike unidirectional ion flux measurements, a change in intracellular Na⁺ concentration due to Na⁺ transport is an indirect measurement that reflects net effect of transport in and out of the cell. Although the development of these new dyes is significant, they have not yet found their way into drug development.

Potassium dyes.

K⁺ dyes were created by modifying the chemistry of existing Na⁺ dyes. PBFI and Ansante Potassium Green (APG-2, APG4) were developed for greater sensitivity to K⁺ versus Na⁺. Although the APG4 dye has improved K⁺ selectivity over previous dyes, caution is required since the fluorescence is affected by intracellular proteins (molecular crowding) and cell volume (110).

Chloride dyes.

Both chemical dyes and genetically encoded [protein] dyes have been developed to survey Cl⁻ in cells and tissues (Fig. 4). The first chemical dyes developed for measuring changes in intracellular Cl⁻ were heterocyclic organic compounds containing a quaternary nitrogen group (111). The compounds are SPQ [6-methoxy-N-(3-sulfopropyl)quinolinium] (112), MQAE [N-(ethoxycarbonylmethyl)-6-methoxyquinolinium bromide] (113), and diH-MEQ [6-methoxy-N-ethylquinolinium iodide] (114). Dyes based on different scaffolds and with brighter fluorescence were also developed (114). One significant disadvantage of Cl⁻ indicators is the fact that they are quenching indicators, meaning that their fluorescence decreases with increased concentration of the halide. These dyes have been used extensively before the appearance of yellow fluorescent protein (YFP)-based Cl⁻ indicators (115). Note that generally, iodine produces stronger quenching and is often used as a substitute for Cl⁻ in ion transport studies involving Cl⁻ indicators. This is the case for studies of CFTR, which seem to readily transport iodine (116, 117). As indicated in the Radioisotopes section, however, the alternate halide is not as useful with cation-chloride cotransporters as iodine is a poor substrate for these transporters (Fig. 2E).

Genetically encoded Cl⁻ dyes are based on yellow fluorescent protein and the cooperative binding of halide ions near the chromophore (118). Fluorescence decreases significantly as Cl⁻ is raised from 0 mM to 150 mM. Clomeleon was engineered as the first ratiometric indicator for Cl⁻ as it is a fusion protein of YFP-cyan fluorescent protein (CFP), with YFP being sensitive and CFP insensitive to the anion (119). Its K_D for Cl⁻ is 160 mM. Using cell-free protein engineering methods, mutations were made in the protein to significantly enhance the K_D for Cl⁻ without sacrificing the Förster Resonance Energy Transfer (FRET) properties. This effort led to the development of a new Superclomeleon protein with a brighter CFP, a shorter linker (enhancing FRET), and YFP mutations (Q69T/V153A) that reduce the K_D for Cl⁻ to ∼20 mM, which is far more compatible with physiological concentrations (120). A different fusion protein made use of an enhanced T203Y mutant YFP-fused to DsRed, allowing the simultaneous measurement of chloride and pH. The K_D for Cl⁻ was determined to be 13 mM, also within the physiological range (121). The sensor was used to assess intracellular Cl⁻ in Drosophila renal epithelial cells (122) and SLC26A3 expressing Chinese hamster ovary cells (CHO) cells (123).

Thallium dyes.

(Tl⁺) is an interesting element and important in the context of high throughput screening (HTS). It is a metalloid element located with boron and aluminum in the IIIa (13th) column in the periodic table. Its atomic weight is 81 (Fig. 2). Its most common oxidation state is +1 making it “similar” to alkali metals. As indicated in Fig. 2, the atomic radius of Tl⁺ is 196 pm, which is a bit smaller than Na⁺ (227 pm), K⁺ (275 pm), and Rb⁺ (303 pm). Thus, Tl⁺ can replace both sodium and potassium ions in the glutamate transporter excitatory amino acid-1 (124); potassium in K⁺ channels (125) and Na-K-2Cl and K-Cl cotransporters (126, 127). Note that a smaller size cation does not necessarily translate into more effective transport. For NKCC1 and KCC2, the transport of Tl⁺ was reduced by ∼50% compared with K⁺, indicating that the natural cation is a much better substrate.

As it is the case for the transport of isotopes (see Radioisotopes section), the movement and detection of Tl⁺ directly report the transport of the cation across the plasma membrane. Thus, the use of Tl⁺ fluorescence indicators is quite different from the use of dyes that monitor the intracellular concentration of an ion. The first Tl⁺ indicator described was the coumarin benzothiazole-based Ca²⁺ indicator, BTC (125). The dye is conjugated to an acetoxymethyl (AM) group to facilitate its loading into cells. Because of its relatively low affinity for divalent cation, calcium-dependent changes in fluorescence are minimal when compared with the magnitude of the Tl⁺-dependent effects. In addition, Ca²⁺ binding decreases BTC fluorescence, whereas Tl⁺ binding increases it. The dye generates a strong positive signal upon interaction following Tl⁺ uptake into the cells. Other thallium-sensitive dyes are the zinc indicators fluozin-2 and 3 (127, 128), and Thallos, a fluorescein moiety coupled to an amino dicarboxylic acid metal binding unit (129, 130).

PHARMACOLOGY OF CATION-CHLORIDE COTRANSPORTER INHIBITION: A HISTORICAL VIEW

The pharmacology of cation-chloride cotransporters started with the clinical need for diuretics, and this was long before their molecular targets were identified. In conditions of decreased cardiac function such as in congestive heart failure, blood volume accumulates causing peripheral edema or abnormal accumulation of fluid in tissue spaces. This problem is typically treated first with beta-blockers (e.g., atenolol), angiotensin-converting enzyme (ACE) inhibitors (e.g., Lisinopril), or angiotensin-II receptor blockers (e.g., losartan), and second diuretics to eliminate the extra volume. The first agent used to induce diuresis was mercury (131), an element which use traces back to ancient times. In 1553, Paracelsus (Swiss physician, alchemist, and philosopher) recorded that mercuric chloride was a cure for dropsy (edema). This was followed by less toxic forms of mercury such as mersalyl (132). Possible sites of action of mercury are the distal tubule Na-K-2Cl and Na-Cl cotransporters (76), which have two highly conserved neighboring cysteine residues in the extracellular loop connecting TM7 and TM8. In the zebrafish cryo-EM structure of NKCC1, Cys496 and Cys507 form a disulfydral bridge, which likely stabilizes the cotransporter structure. Mercurial compounds could interfere with this bridge and inhibit cotransporter function.

Thiazides were the first non-mercurial compounds used in the clinic to induce diuresis (133). They were shortly followed by frusemide (134), now called furosemide, which has some chemical similarity with hydrochlorothiazide (Fig. 5). Interestingly, frusemide at its maximal dose produced more than twice the sodium excretion than any of the benzothiadiazine compounds, which at the time of discovery indicated a different mechanism of action. Bumetanide (3-(butylamino)-4-phenoxy-5-sulfamoylbenzoic acid) then came to clinical medicine in 1972 as a diuretic with higher potency than furosemide. A study published in the British Medical Journal reported its use in 27 edematous patients (135). The dosage of bumetanide required for similar effect was 1/40 the dosage of frusemide. The efficacy of furosemide as cornerstone loop diuretic is attributable in large measure to its active secretion by the proximal tubule organic anion transporters OAT1 and OAT3 (136). The resulting concentration of furosemide in the tubular fluid and the absence of albumin binding (90+% in plasma) in the tubular fluid greatly favors inhibition of TALH NKCC2 over systemic NKCC1.

Figure 5.

Two classes of diuretics (thiazide and loop) share structure similarities. Hydrochlorothiazide and furosemide have conserved phenol rings with a Cl⁻ ion and a sulfonamide group. Furosemide and bumetanide share the sulfamoylbenzoic acid but the Cl⁻ ion in furosemide is replaced with a phenoxy group. Conserved groups in structures are highlighted by the gray background.

Today, we now know that thiazide diuretics inhibits the Na-Cl cotransporter expressed in the distal convoluted tubule of the kidney, whereas furosemide and bumetanide target the Na-K-2Cl cotransporter expressed in the thick ascending limb of Henle (137, 138). As seen in Fig. 4, there are significant chemical similarities between chlorothiazide, furosemide, and bumetanide. Piretanide and torasemide are additional compounds with inhibitory effect on NKCC function. Ethacrynic acid, which was synthesized to mimic the sulfydryl reactivity of mercurial compounds, is also an inhibitor of Na-K-2Cl cotransport and is also used to treat fluid retention (139).

Furosemide was also key in the discovery of the K-Cl cotransport in sheep and human red blood cells (99, 140). Indeed, the cotransporter was originally defined as a NEM or cell swelling activatable furosemide-sensitive and Cl⁻-dependent K⁺ flux. The loop diuretic has, however, a much weaker effect on K-Cl cotransport, as it works 2–3 order of magnitude less efficiently than for inhibition of Na-K-2Cl cotransport. Interestingly, the potency of furosemides depends upon the level of extracellular K⁺. In the absence of external K⁺ (or Rb⁺), the EC₅₀ for furosemide inhibition of K⁺ efflux in LK sheep red blood cells was greater than 1 mM, but in the presence of 23 mM external Rb⁺, the EC₅₀ dropped to 50 µM (141). In addition, there was no furosemide effect in the absence of external Cl⁻, indicating that anion binding is also required for furosemide inhibition. This observation is consistent with an ordered model of binding where Cl⁻ binds first followed by K⁺ (142). When Cl⁻ is bound, furosemide can bind and affect both the V_max and the apparent affinity for K⁺/Rb⁺ (141).

In toadfish, duck, and sheep red blood cells, 2,2′diisothiocyanatostilbene-2,2′ disulfonic acid (DIDS) or its derivative H₂-DIDS strongly inhibits K-Cl cotransport (143). As with furosemide, DIDS inhibition required external Cl⁻ and was enhanced by the presence of external K⁺. Non-competitive inhibition of K⁺ influx was observed for DIDS versus external K⁺ or Cl⁻ concentrations. The effect of DIDS on red blood cell K-Cl cotransport might be species-specific, as human red blood cell K-Cl cotransporter activity was unaffected by the stilbene derivative (144).

WHAT ARE THE COMPOUNDS TRULY TARGETING?

There are no “true” potentiators or positive allosteric modulators of cation-chloride cotransporters. For K-Cl cotransport, the closest molecule to function as a potentiator is N-ethylmaleimide (NEM), which activates K-Cl cotransport-mediated K⁺ influx in LK sheep red blood cells by fourfold (101). Because Na-K-2Cl cotransport is otherwise inhibited by NEM (145, 146), it is currently believed that the thiol alkylating agent acts on a kinase inhibiting K-Cl cotransport and activating Na-K-2Cl cotransport. Work in the past 20 years has identified the WNK/SPAK kinase pathways as direct regulators of NKCC and KCC function (147–151). In 2006, we showed that NEM readily inhibits SPAK kinase activity using in vitro phosphorylation assays (152).

The NEM effect establishes that compounds can indirectly affect the function of a transport protein. The overall transport at the cell level depends upon the amount of protein expressed at the cell surface. For ion channel proteins, the current flowing through the membrane is typically defined by N × P_o, where N is the total number of functional channels expressed at the plasma membrane, while P_o is their open probability. For carriers, an equivalent parameter is V_max or the maximum rate of transport. Expression of a protein at the cell surface involves many steps, including transcription/translation, forward trafficking, membrane residence time and internalization, degradation, and/or recycling. In addition to “N,” transport activity is also affected by post-translational modification and the channels or transporters at the membrane can be activated/deactivated by cellular processes. For the purpose of drug discovery, it is important to remember that all these steps are possible targets for manipulation, and the design of an experiment can favor one step over the other.

HIGH THROUGHPUT DRUG DISCOVERY SCREENING ACTIVITIES

First, recognition should be made to the original efforts of Hoechst AG, the German chemical and life-sciences company, for creating and testing in the 1950s and 1960s countless compounds that led to the discovery of chlorothiazide, furosemide, and bumetanide, drugs that are among the most utilized drugs in clinical medicine. It then took 30–40 years for Academia to be attracted to the development of new compounds. Drug discovery efforts were facilitated and justified by the cloning and molecular identification of the transporters, the study of their more diverse biological functions, and the need to discriminate pharmacologically between these transporters.

From West Point, Pennsylvania to Paris, France

In the early 1980s, the Merck Sharp & Dohme Research Laboratories in West Point, Pennsylvania teamed with a group of neurosurgeons at Albany Medical College to develop drugs that prevent the swelling associated with brain injury. They performed chemical modification around ethacrynic acid and developed a series of (aryloxy) alkanoic acid compounds, some of which had significant beneficial effect following brain injury. Among them, (indanyloxy) alkaloid acids were devoid of diuretic activity (153, 154). In collaboration with a group in Paris, they tested a subset of these compounds on N-ethylmaleimide stimulated K⁺ efflux in red blood cells and found four [(dihydroindenyl)oxy]alkanoic acids with potent inhibitory activity. One compound, called DIOA for simplicity, demonstrated an EC₅₀ of 10 µM, without effect on NKCC1 (155). The compound was also active in inhibiting swelling-activated K⁺ efflux. It was noted that high concentrations of DIOA induced an increase in nonspecific membrane leak of K⁺, indicating some toxicity.

Drug Discovery at Vanderbilt University (Nashville, TN)

In 2009, we published the first HTS effort in developing modulators of KCC2 function (127). At the time, inhibitors were furosemide with an EC₅₀ of 0.6 mM, and DIOA with an EC₅₀ of 1 µM, and as activator was N-ethylmaleimide, which was effective at a dose of 1–2 mM. For our assay development, we used HEK293 cells transfected with the rat KCC2 cDNA (6) inserted into pIRES_puro2. Following puromycin selection, multiple individual clones were picked, expanded, and tested functionally using Tl⁺-induced fluozin-2 fluorescence increase. Expression of KCC2 was also confirmed by Western blot analysis. One of the clones was selected and used for the entire screen. The assay used an isosmotic solution containing 0.2 mM ouabain to prevent Tl⁺ uptake through the Na⁺/K⁺ pump. Compounds were preincubated for a period of 8 min before the addition of the Tl⁺-containing solution. The relatively short preincubation time did preclude any long-term effect of the drug. The fluorescence signal acquisition time that followed the addition of Tl⁺ was 5–6 min. A library of 234,560 compounds was screened yielding 4,933 primary hits (2%). Interestingly, a larger number of compounds increased (2,954) rather than decreased (1,880) the fluorescence signal. A secondary screen was done comparing KCC2-expressing HEK293 cells versus wild-type HEK293 cell and 465 inhibitory compounds showed specificity to KCC2-expressing cells. They were then re-tested in triplicate with 11-point concentration response curves and a total of 26 inhibitory compounds were selected. VU0240551 (D4) was identified as the most potent inhibitor with an IC₅₀ of 560 nM. Unfortunately, as far as the activating compounds are concerned, the follow-up screens were unsuccessful to identify specific “agonist” of KCC2 (127).

Chemical modifications of the lead inhibitory compound were instructive as they pointed out to key chemical features of VU0240551. First, the position of the methyl group on the thiazol was critical as the compound was active with the methyl at position 4 (4-methyl thiazole), whereas completely inactive at position 5 (5-methyl thiazole). Second, the secondary amide could be replaced by a tertiary N-Me amide (VU0255011, ML077) with similar potency, suggesting additional possible modifications at this site. In fact, a follow-up chemical optimization study showed that replacing the methyl group of the tertiary amide by a cyclopropyl group, increased potency by 10-fold (156). This improved compound is VU0463271. The study also examined the pyridazine group and demonstrated intolerance to its replacement by pyridines, pyrazines, pyrimidines, and thiadiazoles (156). The KCC-specific inhibitors VU0240551 and VU0463271 are now widely used to study KCC2 function. Just in the past few months, we identified five studies making use of these inhibitors (157–161).

Drug Discovery at Union Chimique Belge, Braine L’Alleud, Belgium

Human liver adenocarcinoma SK-Hep cells were transfected with rat KCC2 and a clone was selected using ⁸⁶Rb uptake (162). A screen was performed by pretreating these cells with NEM to activate the cotransporter and then exposing them to 5 mM RbCl for 10 min. Uptake of Rb⁺ was assayed using atomic absorption spectrometry. Because of the maximal activation of the cotransporter, this screen was conceptually biased to find inhibitors. Benzyl 1-acetyl-2-benzylprolinate (+)-2 was identified as an inhibitor with an EC₅₀ of 350 nM. Structure activity relationship (SAR) efforts led to the identification of more potent inhibitors, but with increased lipophilicity. Conversely, SAR also identified a compound with slightly (two- to threefold) reduced potency over the original compound, but with decreased lipophilicity and much better pharmacokinetic properties. All compounds tested had some minimal inhibitory effect on NKCC1 function (162). Thus, this family of compounds represents a separate class of KCC inhibitors.

Search for KCC2 Agonists (Montreal, QC, Canada)

In 2013, Yves Dekoninck published the first screen that used measurements of intracellular Cl⁻ levels as a readout for KCC2 activity (163). The cells utilized in this case were NG108-15 cells, a Mus musculus neuroblastoma/Rattus norvegicus glioma cell hybrid that natively expresses low levels of KCC2 expression. This low KCC2 level mimics the conditions of neuropathic pain and thus the screen was intended to identify compounds that increase KCC2 function. To assess KCC2 function, they transfected the chloride indicator: Clomeleon. They screened a library of 92,500 compounds and identified 78 positive hits. Secondary screens were done with cells that express NKCC1, KCC1, KCC2, and KCC4. They then focused on CL-058, (5Z)-5-[(2-hydroxyphenyl)methylidene]-2-piperidin-1-yl-1,3-thiazol-4-one, a three-ring compound with a middle thiazol group. After synthesis of 300 analogs, they identified a lead compound (CLP257) which had an added Fl⁻ ion to the hydroxyphenyl group and that replaced the pyridine group with a pyridazine. The substitution increased the potency of the compound from an EC₅₀ of 31.5 mM to 50 nN. Importantly, the effect seemed to be specific to KCC2 as Rb⁺ flux assays performed in Xenopus laevis oocytes expressing NKCC1, KCC1, KCC3, and KCC4 failed to demonstrate increased transport. Experiments were done at appropriate osmolarities, as KCC1, 3, and 4 require hypotonic swelling for activation. Effect of the compound on the GABA receptor was also tested using heterologous expression of receptor subunits in CHO cells.

CLP257 was shown to significantly increase the rate of Cl⁻ accumulation upon elevation of extracellular K⁺, in two models of pain hypersensitivity. In these experiments, because freshly isolated spinal neurons were used to assay Cl⁻, the permeable MQAE Cl⁻ indicator was used. Cl⁻ load was imposed onto the neurons by delivery through the recording pipette. The neurons were from spinal cords treated with brain-derived neurotrophic factor (BDNF) or isolated from rats exposed to peripheral nerve injury. In both models, CLP257 application resulted in hyperpolarized E_GABA, indicative of a higher rate of Cl⁻ extrusion (163). Cell surface expression of KCC2 was determined to be higher after application of the compound, indicating that CLP257 affects protein turnover. Additional modifications were made to CLP257 to improve its pharmacokinetic profile and its usefulness in in vivo studies (163).

Note that Cardarelli et al. (164) challenged the idea of a direct effect of CLP257 on KCC2. They argued that CLP257 potentiates the GABA_A receptor, providing an alternative explanation for the beneficial effect of the compound on pain perception following peripheral nerve injury. Additional evidence against CLP257 affecting KCC2 was the demonstration that the compound failed to interact with the protein by backscattering interferometry, while a direct interaction could be demonstrated between the inhibitor VU0463271 and the cotransporter (164). Also, several attempts to capture a CLP257-bound KCC2 structure failed (Guo, personal communication), whereas a VU0463271-bound hKCC1 cryo-EM structure was recently reported (165). It seems therefore that CLP257 is not a positive modulator of KCC2 function and its identification originated for some unknown reasons as a false-positive hit in the screen.

Search for Transcriptional Enhancers (Boston, MA)

The promoter region responsible for KCC2b expression, the forebrain and predominant isoform of KCC2, contains a typical 5′-cytosine-phosphate-guanine (CpG) island making it responsive to epigenetic modifications (Fig. 6A). In a 2012 study, Yeo et al. (166) demonstrated that bisphenol A, an environmental chemical released from polycarbonate plastics and epoxy resins, retarded the developmental upregulation in KCC2 mRNA expression in perinatal rat cortical neurons. This effect was rescued by using trichostatin A, a pan-histone deacetylase inhibitor or by reducing expression of HDAC1. Treatment of neurons with bisphenol A led to increased binding of MECP2 and decreased binding of H3K9ac to the KCC2b promoter, providing a mechanism for the decrease in KCC2b expression by the plastic toxin. In vivo experiments further confirmed the toxicity of bisphenol A on neuronal development (166).

Figure 6.

Targets and drugs associated with K-Cl cotransporter (KCC2) function. A: activity of the KCC2 promoter is affected by compounds that target histone acetylation and methylation of 5′-cytosine-phosphate-guanine (CpG) sites. The “plastic” toxin Bisphenol A inhibits the histone acetyltransferase which tend to increase KCC2 activity. Thus, the toxin reduces KCC2 expression. This phenomenon is reversed by Trichostatin A (Tri. A), which inhibits HDAC1. The screen of Jaenisch group (20) identified an inhibitor of GSK3β, a kinase that stimulates HDAC. Inhibition of GSK3β results in increased KCC2 expression. Methyl CpG binding protein 2 (Mecp2) binds to methylated CpG sites likely to stimulates CpG demethylases leading to stimulation of KCC2 expression. B: historical and new inhibitors to KCC2 function. The compounds are believed to interfere with the translocation of ions. Note that the phosphorylation events leading to inactivation of KCC2 or internalization of KCC2 are also targets for drug discovery. Cryo-EM model of KCC2 was drawn using Pymol and PDB ID: 7D8Z.

Jaenisch’s group (20) recently developed a sophisticated screen designed to identify compounds enhancing the transcription of KCC2. They introduced, via CRISPR/cas9, a luciferase reporter gene upstream of the stop codon in the SLC12A5 gene in human embryonic stem cells. They then differentiated the cells into neurons expressing KCC2. As the luciferase reporter protein was not fused to the KCC2 protein but expressed separately and in an equal amount to KCC2, the assay reports KCC2 with fidelity, but without affecting translation and function of the cotransporter. From a combination of three targeted chemical libraries, they screened 929 compounds and found 14 hits. Note that the compounds were added to the neurons for 1 wk, which seems to be an awfully long time to capture transcriptional effects and unfortunately likely to invite unspecific effects. The hit compounds included: KW-2449, an inhibitor of FLT3 (receptor-type tyrosine-protein kinase); BIO (6-bromoindirubin-3′-oxime), an inhibitor of the glycogen synthase kinase 3 β (GSK3b) pathway; and resveratrol. These compounds were shown to enhance KCC2 expression by twofold (20).

Consistent with an increase in KCC2 expression, these compounds led to a hyperpolarizing shift in E_GABA in immature neurons, increased KCC2 expression in MECP2-null RTT (Rett syndrome) human neurons, and the rescue of functional and morphological deficits in RTT neurons (20). In vivo efficacy of KW-2449 and piperine was tested in a Mecp2 mutant mouse model of RTT (Mecp2 −/y) and the compounds improved the disease-related behavioral pathologies in this mouse model (20).

Recent Chemistry on Bumetanide (Genoa, Italy)

As indicated in a comprehensive review published in 2000, bumetanide inhibits both NKCC1 and NKCC2 with EC₅₀s in the range of 10⁻⁷–10⁻⁶M (167). In addition, bumetanide is 40-fold more efficient in inducing diuresis in humans than its analog, furosemide (135). To target NKCC1 and avoid renal effects due to NKCC2 inhibition, Savardi et al. (168) synthetized new bumetanide analogs and tested them using a membrane-bound yellow fluorescent protein as Cl⁻ indicator in HEK293 cells transfected with NKCC1 or NKCC2. They tested 135,000 compounds from several chemical libraries and selected 90 compounds for secondary screens. They selected two compounds out of these hits: ARN22392 and ARN22393, with moderate inhibitory effect on NKCC1. Upon chemical modifications, they identified a compound, ARN23746, with improved solubility and potency. Although the compound still had moderate effect on NKCC1 (38% at 10 µM and 95% at 100 µM) but was substantially weaker on NKCC2 (6.9% inhibition at 10 µM). The compound reduced GABA-mediated Ca²⁺-influx in neurons, consistent with the role that NKCC1 plays in accumulating Cl⁻ and promoting GABA depolarization. ARN23746 had no significant off target activity and an improved pharmacokinetic profile, when compared with bumetanide. Finally, the compound was tested in vivo and shown to have no effect on diuresis, whereas it rescued the social deficits and repetitive behaviors in the valproic acid mouse model of autism, which is attributed to abnormal NKCC1 function (169). Thus, ARN23746 seems to be better (potentially better therapeutic) than bumetanide for in vivo studies involving NKCC1.

FUTURE PERSPECTIVES: DRUG DISCOVERY THROUGH COMPUTATIONAL DOCKING?

Over the past year, several cryo-EM structures of cation-chloride cotransporters have been published. The structure of the zebrafish NKCC1 came first (36), closely followed by the structures of the human KCC1 (37), mouse KCC4 (170), human KCC2, KCC2, and KCC4 (171). These major advances in the field now open the possibility of structure-based drug design. Structure-based or computer-aided drug design predicts the binding of small molecules to a protein, i.e., determine their position within the three-dimensional representation of the protein structure and estimate binding affinities. Computational approaches have the advantage of being relatively cheap compared with chemical screens. The cost largely consists of processor time on super computers.

Because the accuracy of a model is based on the number of simulations, it is useful to have some indication about the general vicinity of binding. This information can be gained through an iterative process with a first set of simulations identifying likely sites of binding, followed-up by additional simulations to refine the site of interaction. Note that a major limitation is the unique three-dimentional representation of a protein given by a crystal or cryo-EM structure. That representation only represents one state or configuration of a protein and that configuration might not be appropriate to model the binding of a drug. For instance, structures of membrane proteins typically favor inside versus outside configurations. This is the case for the several cation-chloride cotransporter structures just mentioned as well as for the crystal structures of related bacterial transporters (172). Based on the available structure of hKCC1, we simulated the binding of VU0463271 (and other compounds) using Autodock Vina, an open-source molecular docking software designed by the Scripps Institute (173). We started with the assumption that the drug must bind close to the pathway utilized by the ions during transport and were encouraged to find a putative binding site with very-high binding energy. Importantly, binding was coordinated by a tyrosine residue which is part of the K⁺ binding site, a key observation since the binding of VU0240551 is competitive with respect to K⁺ (127). One significant limitation of the docking experiment is the fact that the structure available was in inside-facing configuration with a closed extracellular pore. Would the binding be similar if the protein was in its outside-facing configuration with an open external pore? One way to answer this question would be to solve the structure of the transporter conjugated with the inhibitor. A paper recently uploaded to BioRxiv presents such a structure with the compound located as predicted within the ion permeation pathway, but closer to the external pore (165). In fact, the presence of the inhibitor likely stabilized the structure of the cotransporter in its outside-facing conformation. In the absence of published PDB coordinates, we cannot run molecular docking simulations and compare the positions and binding energies of the small inhibitory molecule in the two separate configurations.

Once the position of a small molecule within the structure of a protein is known, it becomes easier to test the effect of chemical modifications. This process can be automated, and the binding of thousands of chemical structures can be simulated to identify molecules with higher binding affinities. The process can also be done with multiple isoforms and small molecules can be identified based on strong binding with one isoform and much weaker binding with a separate but related isoform. These computer-based drug-screening efforts are great to generate hypotheses that can be tested experimentally.

CONCLUSIONS

The life cycle of a membrane transporter is complex and involves a multitude of steps that precede and follow the period of transport activity. A drug can affect the level of transport by tampering with any of these steps. Specificity of a drug toward the transport activity will depend upon the precise target and thus, the highest confidence for specificity is reached when the drug interferes directly with the transporter and its function. This is certainly the case for the thiazide and loop diuretics that inhibit NKCC1/NKCC2, and for the new compounds that inhibit KCC2 (Fig. 6B). That said, it is obvious that direct binding to the target alone does not guaranty specificity. Furosemide, for instance, is a potent inhibitor of NKCC1/NKCC2—but is also known to affect the activity of some GABA_A receptors (174). Also, because of gene redundancy, paralogous genes typically encode proteins with high degree of conservation/homology, and therefore these proteins are likely to be affected by the same drugs. Not only furosemide and bumetanide affect the NKCC1 and NKCC2 transporters, but with lower affinity, the compounds also inhibit the related K-Cl cotransporters. High throughput screening efforts can identify compounds that discriminate between transporter isoforms, but success can also be obtained by screening large numbers of chemical variants. Also discussed in our last section of this review, we believe that computational docking simulations have the potential to help identifying small molecules that would bind to one isoform but be excluded from another.

The proteins involved in gene regulation, transcription, translation, trafficking, post-translational modification, internalization, and recycling are specific to individual transporters. What makes the process specific to each transporter protein is the unique combinatory sequence of all the regulatory players involved. Because each of these regulatory proteins participates in the regulation of many other proteins, their targeting makes it unspecific. Thus, targeting transcription factors or protein kinases might lead to the desired effect with a transporter under study, but will likely lead to unintended consequences. For instance, targeting GSK3β will not only lead to the desired upregulation of KCC2, but will also affect all the other processes regulated by the kinase. Let us remember, however, that most drugs currently used in medicine have known (and possibly unknown) side effects, and they are used because the benefit of using them to treat disorders far outweigh their unwarranted off target activities.

The pharmacology of Slc12 transporters is still limited to a few inhibitors. To date, all these compounds seem to act as full inhibitors rather than partial inhibitors. The development of partial inhibitors might be of use for diseases with abnormally high transporter activity. This would be for instance the case for patients with gain-of-function mutations in K-Cl cotransport (24, 175). Conversely, whether small molecules can be found to bind to the transporters and stimulate their function is still unknown. Any compound that would bind and stabilize the dimer, accelerate the translocation of ions, or increase the time of residency at the plasma membrane would increase the function of the transporter. Any compound that would indirectly affect these processes would also increase the function of the transporter. Functional screens are typically unbiased toward direct versus indirect mechanisms, and once a small molecule is identified as a modulator, additional work is needed to resolve the mechanism of action. As the physiology and pathophysiology of Slc12 transporters becomes increasingly clearer, the need to find additional modulators should only get bigger.

GRANTS

This work was supported in part by National Institute of Diabetes and Digestive and Kidney Diseases Grant DK093501 and by Leducq foundation for Cardiovascular Research Grant 17CVD05.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author.

AUTHOR CONTRIBUTIONS

E.D. prepared figures; drafted manuscript; edited and revised manuscript; and approved final version of manuscript.