Introduction

Evidence for an electrically neutral cotransport of Na+, K+, and Cl− was first described in 1980 in Ehrlich ascites tumor cells [50]. This coupled transport of 1Na+, 1K+, and 2Cl− was shown to be inhibited by furosemide, a diuretic used clinically to inhibit renal NaCl absorption in a specific part of the kidney tubule: the thick ascending limb of Henle (TAL) [56]. The molecular identification of two Na+–K+–2Cl− cotransporters, NKCC1 and NKCC2, was made in 1994 [28, 47, 62, 132], and the genes and exon structures were soon thereafter partially characterized [63, 105]. Expression patterns were examined by Northern blot analyses, in situ hybridization, and immunofluorescence studies once specific antibodies were generated against the cotransporters [61, 69, 70, 102, 135]. Effort was also made in generating NKCC1 [27, 37, 94] and NKCC2 knockout mice [117] to assess the consequence of eliminating protein expression on the physiology of these mice. While NKCC1 is ubiquitously expressed in cells and plays a major role in cell volume regulation and Cl− secretion, NKCC2 is mainly expressed in the TAL where it mediates NaCl reabsorption. This review will examine the structure/function relationship, the physiological functions, and the regulation of the two Na–K–2Cl cotransporters.

Structure/function of SLC12A1/2 cotransporters

In humans, NKCC1 is encoded by the SLC12A2 gene which is present on chromosome 5q23, whereas NKCC2 is encoded by the SLC12A1 gene located on chromosome 15q15-q21. NKCC1 and NKCC2 form dimers, although each monomer has been demonstrated to be fully functional [87, 115]. NKCC1 and NKCC2 monomers share overall 60 % sequence homology at the protein level (Table 1). NKCC2 is slightly smaller than NKCC1 with 1,099 amino acids yielding a core molecular size of 121.3 kDa, compared to 1,212 amino acids giving a size of 131.4 kDa [28, 47]. The main size difference results from a larger cytoplasmic N-terminal tail, which is some 102 amino acids longer in NKCC1. Consequently, the N-terminal tails are the least conserved portions of the two cotransporters (22 % amino acid identity). The regulatory cytoplasmic C-terminal tail and the core are well conserved between both isoforms with 56 and 79 % amino acid identity, respectively (see Table 1).

Table 1 Comparison between human NKCC1 and NKCC2
Full size table

The cytosolic N-terminal tail

The N-terminal tail of NKCC1–2 is critical to cotransporter function. Deletion of the tail results in inactive cotransporters. However, as mentioned above, the amino terminal tail of NKCC shows the highest sequence diversity. This diversity is not only between NKCC1 and NKCC2, but also among the many NKCC1 orthologs whose sequences are found in GenBank (Fig. 1a and [45]). How can a domain with such diversity be so integral to transport function? The answer comes from two particular subdomains that are highly conserved within the Na+-dependent cotransporters (NKCC and NCC). The first subdomain is a binding site for kinases, and the second subdomain consists of phosphorylation sites located near the plasma membrane. Sequences upstream of the kinase binding site and between the kinase binding site and the phosphorylation sites seem to be of less importance. We will discuss in a later section the regulation of NKCC1–2 by Ste20 (Sterile20) kinases and other regulatory proteins. In this section, however, we will briefly discuss the properties of these two highly conserved subdomains of the N termini. Using a yeast-two-hybrid analysis screen, we found that the N-terminus of cation–chloride cotransporters interacts with two specific Ste20 kinases: SPAK and OSR1. The site of interaction was reduced to nine amino acid residues containing a core RFXV (or RFXV/I) motif [101]. In fact, the tail of NKCC1 contains two such motifs, whereas the tail of NKCC2 only contains the first one. It is of interest to note that the second motif in NKCC1 overlaps with a PP1 binding motif, raising the possibility that the two proteins compete at that particular binding site [45]. Deletion mutants and site-directed mutagenesis studies demonstrated that only one motif was necessary for cotransporter activation, while the loss of the two motifs resulted in the absence of cotransporter function [45]. The presence of a PP1 binding motif in NKCC1 but not NKCC2 is puzzling. Studies have shown that NKCC1 activity is modulated by PP1 and mutation of this residue affects dephosphorylation of the cotransporter [22]. Results from our laboratory have also shown that binding of PP1 to the N-terminal tail of NKCC1 not only results in dephosphorylation of the cotransporter, but also to the dephosphorylation and inactivation of the Ste20 kinases [41, 46]. Once NKCC2 is activated by phosphorylation, it also needs to be dephosphorylated for inhibition. At this stage, it is not clear which phosphatase is involved and how it interacts with the N-terminal tail of NKCC2 and its phospho-sites. The second highly conserved subdomain is a peptide of 25–30 amino acids that contain multiple threonines and/or serine residues, some of which are targets of the kinases that bind to the upstream RFXV sites. Figure 1b highlights the information obtained from mass spectrometry/Edman sequencing analysis as well as site-directed mutagenesis studies. While several residues are involved, the most consistent residues are a pair of threonine residues separated by four amino acids: R/QT 206FGY/HNT 211M/I (numbers taken from mouse NKCC1). There are some interesting functional differences between the cotransporters at these two sites that are not yet explained. Indeed, mutation of T206 in NKCC1 completely abrogates cotransporter function [45], while mutation of the corresponding residue in NKCC2 reduced function by one third [55], and mutation in the Na–Cl cotransporter (NCC) has no functional effect [95]. Similarly, while mutation of T211 in NKCC1 [23, 45] and the corresponding residue in NCC [95] abolish cotransporter function, mutation of the corresponding residue in NKCC2 fails to do so [55]. Finally, note that no amino acid substitution of single or multiple threonine residues into either aspartic acid or glutamic acid residues, to mimic phosphorylation, confers constitutive activity to NKCC1. Instead, the substitutions disrupt cotransporter function [23, 45].

Fig. 1
figure 1

Sequence homology and domains of the NH2-terminal tails of NKCC1 and NKCC2. a Schematic comparison of NKCC1 and NKCC2 amino terminal tails from vertebrates: mammal (Homo sapiens), bird (Gallus gallus), reptile (Anolis carolinensis), amphibian (Xenopus laevis), bone fish (Oreochromis niloticus), and cartilage fish (Squalus acanthias). Identical residues are highlighted in red, conserved residues are highlighted in green, blocks of identical residues are shown in blue, and nonconserved residues are indicated in white (gaps). Note the presence of two to three regions of high degree of conservation, corresponding to the SPAK/OSR1 binding motifs, and the phospho-threonines and phospho-serines. Bars indicate the RFxV motifs or SPAK/OSR1 binding site (filled bar), and the VxFxD motif or PP1 binding site (open bar). b Amino acid alignment of a portion of the N-terminal tail of human and mouse NCC, NKCC1, and NKCC2. Protein sequences were aligned using VectorNti (Invitrogen) and a portion of the alignment is displayed. Threonine and serine residues that are targets of phosphorylation are highlighted with references. The position of the R5 peptide used to generate the R5 antibody [38] is indicated. Residues highlighted in yellow are identical residues, whereas amino acids highlighted in green/blue represent conserved residues

Full size image

The transmembrane core

The core of the NKCC proteins consists of 12 hydrophobic transmembrane (TM) α-helices and a large extracellular loop between TM7 and TM8. The cation–chloride cotransporters belong to a large superfamily of transporters: the amino acid–polyamine–organocation superfamily, whose basic structure consists of an inverted repeat of five transmembrane domains followed by two additional transmembrane α-helical segments at their N- and/or C-termini [129]. The Phyre2 protein fold recognition server (www.sbg.bio.ic.ac.uk/phyre2/) drew a model (>90 % confidence) of the first 10 transmembrane domains of NKCC1, which is represented in Fig. 2a. The model is based on structure similarity with several resolved structures (Protein Data Bank IDs: 4dji, 3gia, 3lrc, 2jln, and 3dh4). A characteristic of amino acid–polyamine–organocation transporters is the duplication or repeat of five transmembrane domains. This 5 + 5 symmetry can be seen with pairs of TM1–TM6, TM2–TM7, TM3–TM8, and TM4–TM9 being in close proximity of each other, whereas TM5 and TM10 being separate (Fig. 2a, b). To further demonstrate the 5 + 5 symmetry, we aligned the five first TMs with the next five and significant amino acid conservation is observed (Fig. 2c). Note the conservation of two pairs of residues between TM2 and TM7, the SV and MM residues in TM2 having been shown to participate in cation binding (see below). While the hydropathy plots of NKCC reported in multiple studies [28, 47, 52, 132] diverges somewhat from that of LeuTAa, a bacterial homolog of neurotransmitter sodium symporters belonging to the same amino acid–polyamine–organocation superfamily, there are some striking high spots of conservation. For instance, four of the six residues in TM9 and 7 out of 10 residues in TM12 that have been shown to promote cotransporter dimerization [134] are conserved in NKCCs. Glycosylation of the large TM7–TM8 loop increases the molecular size of NKCC1 and NKCC2 by some 30–40 kDa [90] and is essential for membrane expression, transport activity, and affinity for loop diuretics [97]. Indeed, inhibition of glycosylation by tunicamycin drastically reduces the activity of both cotransporters. Site-directed mutagenesis of the two N-linked glycosylated sites (N442Q, N452Q) of NKCC2 reduces its membrane expression and activity when single or double mutants are expressed in the heterologous Xenopus laevis oocyte system. While cation affinity was unchanged in mutants versus wild-type NKCC2, Cl− affinity was increased and binding of the loop diuretic bumetanide was decreased in the single N442Q and double N442Q/N452Q mutants. Interestingly, the N442 glycosylated site is conserved in NKCC1, and its mutation also confers an increase in Cl− and a decrease in bumetanide binding [97]. This observation suggests that a reduction in the net negative charge at the extracellular surface of both cotransporters, induced by the absence of glycosylation, increases Cl− affinity but decreases inhibition by loop diuretics. Moreover, the fact that bumetanide and Cl− affinity are significantly modified when glycosylation is prevented, could imply that the loop diuretic and the Cl− binding sites are located near the N422 residue. Taking advantage of differences between the living environments of men and sharks, and consequently of a several fold difference in ion binding affinities between human NKCC1 and shark NKCC1, Biff Forbush’s group created chimeras to map the ion binding domains on the cotransporters [66]. They demonstrated that the Na+ affinity of NKCC1 was 15 mM in human versus 109 mM in shark. The sequence of shark NKCC1 TM2 differs from human NKCC1 TM2 at the two pairs of residues highlighted in Fig. 2c (AL residues instead of SV and GT residues instead of MM). Substitution of these pairs of residues in the shark sequence by their human counterparts significantly decreased the affinity for both Na+ and K+ ions [66]. It is worth noting that similar to TM2, TM7 is also involved in cation binding [65], which is consistent with the symmetry model (Fig. 2a or b). As NKCC1 is inhibited by bumetanide, the chimeric approach was also used to map residues affecting bumetanide binding. It was shown that multiple transmembrane domains (TM2, 11, and M12) participate in bumetanide binding [65]. Recently, cysteine- and tryptophan-scanning mutagenesis experiments have shown that TM3 also plays a major role in ion translocation and binding of loop diuretics [114]. Additional studies are required to specify the role of each transmembrane domain and identify amino acids involved in ion and loop diuretics binding in both cotransporters.

Fig. 2
figure 2

Structure of the core of human NKCC1. a Structure of the transmembrane domains and other alpha helices was modeled using the Phyre2 protein fold recognition server [72]. The PDB file was viewed using the Visual Molecular Dynamics software (University of Illinois) and the image was rendered after highlighting the α-helices. The first five TMs are colored in blue, whereas the next five TMs are colored in red. Note the proximity and parallelism of TM1–6, TM2–7, TM3–8, and TM4–9. b Predicted topology of the first 10 transmembrane domains based on similarity to crystal structure of amino acid–polyamine–organocation transporters. The last two transmembrane domains are not drawn. c Alignment of two segments of the human NKCC1 core showing conservation of transmembrane domains. The SV and MM/VV amino acids in human NKCC1 TMD2–7 are replaced by AL and GT in shark NKCC1. The alignment was done with VectorNti 6 (Invitrogen)

Full size image

The chimeric approach was also used to decipher the molecular basis for the high osmolarity/volume sensitivity of NKCC1 versus low osmolarity/volume response of NKCC2. The Na–K–2Cl cotransporter has been known for many decades to be activated by hypertonicity or cell shrinkage (for an early review, see [49]). We noticed, however, a significant difference in the level of activation from 3- to 4-fold increase in K+ influx for NKCC1 versus 1.5-fold increase for NKCC2 [40]. While the activation requires phosphorylation of the N-terminal tail (see above), we demonstrated that swapping the tails of the cotransporters did not affect their level of activation. In contrast, we showed that portions of the protein that includes the second transmembrane domain (TM2) along with the first intracellular loop (IL1) and the second extracellular loop (EL2) “encode” for the high sensitivity of NKCC1. It is of significance that not only the transmembrane domain but also the loops are involved in the fold activation [40].

With the exception of the squid axon NKCC1, whose stoichiometry of transport was shown to be 2Na+:1K+:3Cl− [109], the typical stoichiometry for both NKCC1 and NKCC2 is 1Na+:1K+:2Cl− [50]. However, when we measured unidirectional K+ and Cl− influxes in NKCC1-injected oocytes exposed to hypertonicity, we observed identical ion fluxes [40]. This was interpreted as NKCC1 functioning partially in K+/K+ exchange mode. Two kinetics models have been described for NKCC1 [26, 81]. The first complete model of ion binding to NKCC1 came from studies performed in duck red cells [81, 85]. It was shown that the free transporter binds Na+ first, followed by Cl−, then K+ and the second Cl− (Fig. 3a). Interestingly, it was postulated that the order of ion release on the other side is reversed so the first ion attached to the cotransporter would come off first. This postulate gave rise to a gliding model, conceptually different than the symmetrical model [81]. Note that K+/K+ exchange in this model can be explained if, close to the end of the cycle, the partially unloaded transporter picks a K+ ion and reverses its orientation (Fig. 3a, red arrows). This model has remained unchallenged until we observed kinetic behavior better compatible with a model where the transporter binds Cl− first, followed by Na+, the second Cl− and then K+ (Fig. 3b). In this model, K+/K+ exchange can still occur when partially loaded carriers reverse their orientation and re-load K+ ions. Note that an unusual NKCC1 stoichiometry has also been measured in neonatal neurons [9].

Fig. 3
figure 3

Two kinetic models of Na–K–2Cl cotransport. a Glide symmetry model where a Na+ ion binding first on the cotransporter in the outside facing configuration is the first ion coming off in the inward facing configuration. b Alternative model with more conventional kinetics where Cl− binds first followed by Na+, the second Cl−, and K+. Partial reactions or slippages which are electroneutral in nature (e.g., NaCl movement) are allowed. See text for details and references

Full size image

Both NKCC1 and NKCC2 are capable of transporting alternative ions as substitute for K+ and Cl−. For instance, rubidium (Rb+) is often used in its radioactive form (86Rb) as a tracer for K+ because the isotope half-life is far more convenient than isotopes of K+, and the cotransporter affinity for Rb+ is similar to that of K+ (discussed in [26]). Thallium (Tl+) has been recently used in combination with intracellular fluorescent dyes to assess the function of NKCC1 and NKCC2 [17, 25, 51]. To date, how Tl+ affinity compared to that of K+ is unknown. NH4 + can also substitute for K+ and be inwardly transported by both cotransporters. Fast accumulation of NH4 + within the cell results in acidification of the intracellular pH, which in turn, inhibits the activity of the cotransporters [7]. Note that, as it will be described below, NKCC2 in the kidney and NKCC1 in the intestine regulate acid–base balance by promoting NH4 + transport. As far as the anions are concerned, Br− substitutes quite well for Cl− on NKCC1, whereas I− is a poor substitute for Cl− (see Fig. 4). Note, however, that the cotransporter is able to mediate some K+ movement in the presence of iodide, as we were able to measure a small but significant bumetanide-sensitive K+ influx in the presence of this anion.

Fig. 4
figure 4

NKCC1 transport of Cl−, Br−, and I−. K+ influx was measured in NKCC1-injected Xenopus laevis oocytes in a hyperosmotic solution containing 96 mM NaCl, 4 mM KNO3, 2 mM Ca(NO3)2, 1 mM MgSO4, 60 mM sucrose, and 1 mM ouabain in the presence or absence of 20 μM bumetanide (pH 7.4, osmolality 260 mOsM). For halide substitution, NaCl was replaced with NaBr, or NaI. Fluxes are expressed in nanomoles K+ per oocyte per hour. Bars represent means ± SEM (n = 20–25 oocytes). Values in inset represent sizes (radius) of atoms and their ions in pm

Full size image

The cytosolic C-terminal tail

The C-terminal tails of NKCC1 and NKCC2 are important for maturation, dimerization, and protein trafficking to the plasma membrane. While no splicing events involving the C-terminal tail of NKCC2 have been reported, there is one isoform of NKCC1 that has a different carboxyl-terminus. The alternatively spliced isoform involves a cassette exon (exon 21 which is 48-bp-long and encodes 16 amino acids) which is omitted in some tissues [105, 122]. Although the 3′ end of the exon creates a PKA consensus site when present, there is no evidence that protein kinase A modulates NKCC1 function at that locus. In contrast, the 16 amino acids fragment was shown to contain a di-leucine motif as a basolateral sorting motif, therefore targeting NKCC1 to the basolateral membrane of epithelia [13]. This sequence is absent in both the NKCC1 isoform lacking exon 21 and NKCC2. Absence of the basolateral targeting sequence explains the apical localization of NKCC2 in the thick ascending limb of Henle. When we identified the alternatively spliced isoform of NKCC1 in 1997, we found it to be abundant in the brain. Whether or not the apical localization of NKCC1 in choroid plexus [102, 131] is due to the absence of exon 21 is unknown. The C-terminal tail of NKCC1 has been shown to facilitate di- and tetramerization of the cotransporter. This was shown using a yeast-two-hybrid method [111] and fluorescence resonance energy transfer (FRET) analysis [86].

Physiological functions of the SLC12A1/2 cotransporters and relation to human diseases

NKCC2

NKCC2 is expressed on the apical membrane of the epithelial cells of the thick ascending limb of Henle (TAL) which reabsorbs about 20–30 % of the NaCl filtered by the glomerulus ([4], Fig. 5a, Table 2). Note that the main function of the TAL is to reabsorb large amount of NaCl while reabsorbing no water, which leads to dilution of the forming urine in the tubule lumen [6, 82]. Loss-of-function mutations in SLC12A1 gene result in Bartter’s syndrome, an autosomal recessive disease characterized by plasma volume reduction, polyuria, hyponatremia, hypotension, hypochloremia, hypokalemia, magnesuria, metabolic alkalosis, and hypercalciuria [11, 112]. These multiple characteristics can be explained as follows: a decrease in Na+ reabsorption in the TAL results in an increased delivery of Na+ in the collecting duct, where its reabsorption is mediated by the epithelial Na+ channel, ENaC. Increased ENaC activity results in an increased K+ secretion by the channel ROMK [78], leading to hypokalemia and gives rise to an increased H+ secretion by intercalated cells H+-ATPase, leading to metabolic alkalosis [124]. Furthermore, the decrease in plasma volume associated with the defect in TAL Na+ reabsorption stimulates the renin–angiotensin–aldosterone system leading to an increased production of angiotensin II and aldosterone which respectively stimulates proximal Na+/H+ exchanger and distal ENaC, thereby leading to a further secretion of H+ and K+ [74]. This phenotype has been mostly reproduced in a mouse model [117]. The homozygous NKCC2 knockout mice exhibit a severe decrease in plasma volume, high plasma renin, hypokalemia, and metabolic alkalosis, while heterozygous knockout mice did not show any physiological difference when compared with wild-type mice. Reversely, enhanced activity of NKCC2 has been linked to hypertension and hypertensive disorders in rodents and in humans [12, 15, 118].

Fig. 5
figure 5

Na–K–2Cl cotransport in two types of epithelial cells. a Thick ascending limb epithelial cell model showing apical NKCC2 localization. The driving force for NaCl reabsorption is provided by the basolateral Na+/K+ pump. Apical K+ channel (ROMK) delivers K+ in the lumen for NKCC2 function and creates an electropositive lumen. ClCK(A and B) on the basolateral membrane creates a path for Cl− movement and participate in the electronegative blood side. The electrical field generated by the epithelial cells favors the paracellular reabsorption of divalent cations. b Model of a Cl− secreting epithelial cell with localization of NKCC1 on the basolateral membrane. The cotransporter replenishes Cl− as the anion is transported across the apical through CFTR or other Cl− channels. K+, which enters through the pump and NKCC1, leaves the cell through apical and basolateral K+ channels (not shown). Note that this model can be used for stria vascularis marginal cells (K+ secreting cell) with CFTR substituted for KCNQ1

Full size image
Table 2 Functions of NKCC1 and NKCC2
Full size table

NKCC2 is also present in the apical membrane of the macula densa, which is located in the juxtaglomerular apparatus and is strategically in contact with the distal tubule and the extraglomerular mesangium close to the afferent arteriole [70]. The macula densa cells are NaCl sensors able to adjust the glomerular filtration by vasodilatation or vasoconstriction of the afferent arteriole. While a decrease in tubular NaCl concentration results in vasodilation of the afferent arteriole and in renin release by the granular cells, an increase in NaCl concentration induces a vasoconstriction of the afferent arteriole and thereby a decrease in glomerular filtration. This latter mechanism is known as the tubuloglomerular feedback and NKCC2 has been shown to play a pivotal role in the sensing of this high tubular NaCl concentration ([100, 110], Table 2). Note that NKCC1 is also highly expressed in the afferent arteriole renin-positive cells and in juxtaglomerular mesangium [69], possibly also related to tubuloglomerular feedback.

Alternative splicing of the exon 4 of the SLC12A1 gene gives rise to three NKCC2 variants [61, 98]. These variants have different transport characteristics and different localizations along the TAL and the macula densa (Fig. 6). NKCC2F, which is the most abundant variant, is exclusively expressed in the medullary TAL, while NKCC2A is detected in both medullary and cortical TAL and NKCC2B is present in the cortical TAL. Macula densa express both NKCC2A and NKCC2B. Note that in human, NKCC2A is the predominant variant, possibly a reflection of a slightly different diet and physiology [18]. Affinity for Cl− differs between variants, and although there is some variability between species, the sequence order is NKCC2B > NKCC2A > NKCC2F for mouse and rabbit cotransporters [19]. The presence of both NKCC2A and B variants in the macula densa facilitates efficient NaCl sensing over a wide range of fluctuating NaCl concentrations [18, 19]. Specific NKCC2 variant knockouts did not show severe salt-wasting phenotype probably because of compensation by the other NKCC2 variants [93].

Fig. 6
figure 6

Diagram of the mouse medullary and cortical thick ascending limb of Henle expressing NKCC2. The low affinity isoform, NKCC2F, is expressed in the lower portion of the outer medullary TAL where the urine Na+ concentration is high. The high affinity isoform, NKCC2B, is expressed in cortical TAL where the urine Na+ concentration is low. The intermediate affinity isoform, NKCC2A, is expressed along the entire TAL. In the outer medulla, the major role of NKCC2 is to transport Na+ without water, thus raising the osmolarity of the interstitium. Osmolarity of the urine and interstitium and the urea and Na+ concentrations in the urine are provided. Glom glomerulus

Full size image

NKCC2 is also a major player in NH4 + reabsorption and acid–base homeostasis (Table 2). NH3 is produced by glutamine metabolism in renal epithelial cells and exists predominantly in its acidic form, NH4 +, at physiological pH. The proximal tubule is primarily responsible for luminal NH4 + secretion, which is substantially increased by metabolic acidosis. Thereafter, NH4 + is reabsorbed in the TAL by NKCC2 and, subsequently, accumulated in renal interstitium where it contributes to the high medullary osmolarity. It is finally secreted back into the urine by Rhesus glycoproteins in the collecting duct [73, 126].

NKCC1

In contrast to NKCC2, which exhibits a highly restricted pattern of expression, NKCC1 is widely expressed in human, mouse, and other vertebrates. To date, there are no mutations that have been found in the human SCL12A2 gene, which suggests that loss of NKCC1 function might be embryonic lethal in human. Considering the many tissues and functions supported by NKCC1 (see Table 2), it is surprising that the NKCC1 knockout mouse is viable [27, 30, 37, 94]. The most striking deficit of the NKCC1 knockout mouse is sensorineural deafness and imbalance, which originates in cochlea and semicircular canals and vestibule, respectively. In the cochlea, the stria vascularis stratified epithelium composed of marginal cells expresses the Na+/K+-ATPase and NKCC1 on the basolateral membrane. As in many epithelia, the Na+/K+ pump provides the force for K+ and Cl− influx through NKCC1. While Na+ and Cl− are subsequently recycled respectively through the Na+/K+ pump and the ClCK Cl− channel, K+ is secreted into the scala media by the complex of K+ channels: KCNQ1/KCNE1 [75, 89, 121]. The high potassium endolymph generates a positive (+85 to +100 mV) endocochlear potential with a large driving force for the movement of K+ into sensory hair cells, which transduce mechanical into electrical signals [138].

NKCC1 is also highly expressed in inner ear spiral (cochlear) and vestibular ganglia [27]. While the role of the cotransporter in these specific neurons is unknown, this observation is interesting in light of the role that NKCC1 plays in sensory afferent neurons. In primary afferent neurons of dorsal root ganglia, NKCC1 participates in Cl− accumulation in both the neuronal cell body and axon terminals. High Cl− in the terminals leads to GABA depolarization, presynaptic inhibition, and filtering of sensory noise [2, 116, 128]. As GABAA receptor subunits are also very abundant in rat spiral ganglion [133], it is tempting to speculate that the cotransporter in inner ear ganglion neurons might be participating in the filtering of auditory noise.

Regulation of intracellular Cl− seems to be an important function of NKCC1 in adult neurons located outside the central nervous system (CNS). NKCC1 is abundant not only in primary afferent dorsal root ganglion and inner ear neurons, but also in olfactory neurons [107] and gonadotropin releasing hormone neurons [24]. In contrast in the CNS, NKCC1 expression is only elevated in immature neurons but then decreases during neuronal maturation [32, 103]. The role of the cotransporter in immature neurons is, however, not very well understood. First, there is currently a debate in the field on whether the cotransporter truly accumulates Cl− in immature neurons. In reality, only minimal changes in the GABA-mediated Cl− reversal potential are measured in wild-type neurons exposed to bumetanide [32, 140] or in neurons lacking NKCC1 expression versus wild-type neurons [5, 139]. Second, there is conflicting information regarding the pro- versus anticonvulsant nature of the cotransporter as studies have demonstrated that bumetanide decreases spontaneous network activity in the immature hippocampus [32], whereas absence of NKCC1 significantly increased spontaneous activity in immature CA3 pyramidal neurons [140]. Bumetanide treatment has been recently effectively used in treating adult patients with temporal lobe epilepsy [33]. However, it is difficult to know if the effect of the loop diuretic is a direct effect on neurons or an indirect effect due to volume depletion. Indeed, the diuretic in the circulation is bound to albumin and does not permeate the blood–brain barrier. The diuretic is currently in clinical trial for intractable neonatal seizures (http://ClinicalTrials.gov). In these newborns, the blood–brain barrier might not be completely tight and the diuretic might be crossing to the brain.

NKCC1 also plays an important role in epithelial function (Fig. 5b). In salivary gland, NKCC1 is highly expressed at the basolateral membrane of acinar cells where it participates in the secretion of fluid that accompanies secreted α-amylase and mucine. Inhibition or deletion of NKCC1 gives rise to a reduction in the volume of saliva secreted in response to muscarinic agonists [34]. In the intestine, the cotransporter is expressed on the basolateral membrane of epithelial cells where it is also involved in fluid secretion [37, 67, 91]. Interestingly, the cotransporter is also expressed in cells involved in peristalsis or contraction and relaxation of smooth muscles that helps the movement of the food along the gastrointestinal tract. Indeed, the cotransporter is found in the interstitial cells of Cajal which surround the myenteric plexus and serve as a pacemaker. These cells generate slow waves that synchronize smooth muscle cell contraction, thus promoting an efficient intestinal transit. NKCC1 is responsible for maintaining intracellular Cl− concentration above equilibrium and its absence in NKCC1 knockout mice induces a depolarization of the smooth muscle cell membranes, leading to a decrease in the frequency and amplitude of the slow waves. In wild-type cells, bumetanide reversibly affects the shape, frequency, and amplitude of slow waves in the smooth muscle cells of the jejunum [130]. Deficit in peristalsis in the NKCC1 knockout mouse might explain the intestinal obstructions and high lethality observed in animals prior to weaning [37]. In humans, impaired myenteric cells of Cajal have been reported in chronic intestinal pseudo-obstruction, infantile pyloric stenosis, Hirschsprung’s disease, slow transit constipation, and diabetic gastroparesis [35, 119, 120, 125].

Two additional striking phenotypes of the NKCC1 knockout mouse are hypotension and male infertility. The hypotension is due to decreased vascular tone, as studies performed in isolated rat aorta have shown that NKCC1 is activated by vasoconstrictors, inhibited by nitrovasodilators, and is upregulated in hypertension [37, 48]. Furthermore, intravenous infusion of bumetanide resulted in a rapid drop in blood pressure in wild-type mice, but not in NKCC1 knockout mice [48]. NKCC1 knockout males show smaller testis size than their wild-type counterpart. Histological analysis demonstrates a smaller diameter of the seminiferous tubules and absence of mature spermatids near the lumen of the tubule [94]. This deficiency might be due to a deficit in the hypothalamus/pituitary/testis axis, as reduced circulating levels of testosterone and luteinizing hormone are observed in the knockout mice [43].

Regulation of SLC12A1/2

Regulation of the gene level

In contrast to other transport mechanisms, there is very little information about the regulation of the SLC12A1/2 genes at the transcriptional level. The SLC12A2 gene (NKCC1) is encoded by 28 exons, and as discussed above, exon 21 is alternatively spliced. Analysis of the genomic DNA sequence upstream of the putative transcriptional initiator site revealed a promoter region characteristic of a housekeeping gene: a TATA-less promoter and several Specificity Protein 1 (Sp1) sites [105]. Activator protein-2 (AP-2) binding sites are also located ~30 and ~500 bp upstream of the transcription initiator site. In plants, AP-2 transcription factors have been implicated in response to environmental signals in stress acclimation [29]. AP-2 s are also involved in the regulation of cell proliferation, differentiation, apoptosis, and carcinogenesis [99]. Several studies have examined changes in NKCC1 mRNA/protein expression upon changes in physiological conditions. Experiments performed in rats dehydrated for 2 days or fed with NH4Cl for 6–7 days to induce metabolic acidosis have shown a significant increase in NKCC1 abundance in both outer and inner medullary collecting ducts [64]. In another study performed in cell culture with T-84 cells, it was shown that hypoxia leads to a significant decrease in NKCC1 transcription [60]. This decrease was due to HIF-1α, a transcription factor mediating transcriptional activation of genes in response to decreased oxygen concentration [92]. Colonic mucosal scrapings from HIF-1α knockout animals showed a ~10-fold increase in NKCC1 expression [60]. Finally, NKCC1 expression was shown to be significantly downregulated in a skin-specific knockout of the forkhead transcription factor, FoxA1 [21]. NKCC1 expression is also epigenetically regulated during the development of hypertension in rat models [20]. While the NKCC1 promoter is methylated with age in normotensive rats, the promoter is hypomethylated in spontaneous hypertensive (SHR) rats. Hypomethylation of the NKCC1 promoter leads to an increase expression of the cotransporter whose function affects vascular tone. It was demonstrated that activity of the DNA methyltransferase is attenuated in hypertensive rats, while upregulated in normotensive rats [20].

The SLC12A1 (NKCC2) gene is encoded by 27 exons with alternative splicing of exon 5 which exists in three versions encoding the A, B, and F variants. Nuclear run-off assays experiments demonstrated that the kidney-specific expression of NKCC2 is due to interactions between kidney-enriched transcription factors and regulatory elements in the 5′-flanked region of the SLC12A1 gene [63]. Deletion analysis revealed the presence of positive regulatory elements controlling the NKCC2 promoter in TAL cells in an ~280-bp fragment not too far upstream from the TATA box. Those transcriptional elements include a putative Hepatocyte Nuclear Factor 1 (HNF-1) site, a GA microsatellite, two consensus CCAAT Enhancer-Binding Protein (C/EBP) binding sites, and one Enhancer Box (E-box) site. In addition, negative regulatory elements were shown to be present in a 725-bp fragment located about 1 kb upstream of the previous region. Both transcription and mRNA stability of NKCC2 are increased upon chronic metabolic alkalosis [71]. This is not unexpected as the cotransporter is a major pathway for NH4 + reabsorption in the medullary thick ascending limb.

Posttranslational modification

Both NKCC1 and NKCC2 are activated by an increase in the external osmolarity and a decrease in the intracellular Cl− concentration. Activation by low Cl− is not related to a gradient effect, but due to a Cl−-sensitive mechanism [80]. Interestingly, the two stimuli are interrelated and working in opposite directions. Indeed, an increase in external osmolarity results in cell shrinkage, which is accompanied by an increase in intracellular Cl−. Thus, there is an activation signal due to a decrease in cell volume and an inhibition signal due to an increase in intracellular Cl−. Whether under these conditions there is overall activation of Na–K–2Cl cotransport is a matter of debate. Two pieces of evidence indicate that it might not be the case. First, as we discussed above, cell shrinkage induces an increase in K+/K+ exchange, which is neutral for net transport [26]. Second, most cells fail to exhibit a regulatory volume increase (RVI) response when exposed to a hypertonicity solution, but demonstrate RVI only when they are returned to isosmotic conditions following a hypo-osmotic challenge (for review, see [59]). Stimulation of NKCC in this case is due to the loss of Cl− that the cells experience during volume restoration under hypotonicity. Similarly, stimulation of NKCC is sometimes observed upon mild hypotonic shock. This can also be explained by a decrease in intracellular Cl− that occurs when low Cl− solutions are used to decrease the external osmolarity.

There are many additional factors, e.g., physical, hormones, cytokines, and signaling cascades, that affect the activity of NKCC1 and NKCC2 [4, 14, 53, 68]. It is worth noting that not all of them affect the cotransporters directly but rather indirectly through other transport mechanisms. For instance, adenylcyclase activators (e.g., forskolin), phosphodiesterase inhibitors (IBMX), and cAMP have a profound effect on NKCC1 function in Cl− secreting epithelia (from shark rectal gland [79] to human airway [58]). It is not at all clear, however, that the cyclic nucleotide and/or protein kinase A have a direct stimulatory effect on the cotransporter. Indeed, in this case, NKCC1 activation is likely due to stimulated Cl− secretion on the apical membrane (i.e., through CFTR), resulting in a reduction in intracellular Cl− and secondary activation of the cotransporter. This mode of activation likely occurs in other situations where K+ channels and Cl− channels activities are affected and the level of intracellular Cl− is decreased.

Evidence that NKCC is activated by phosphorylation comes from the demonstration by John Russell that ATP depletion caused a significant decrease in Na–K–Cl transport activity in squid giant axon [1]. In addition, orthovanadate and fluoride, two inhibitors of protein phosphatases, decreased this rate of inhibition supporting the hypothesis that the cotransporter in squid axon is regulated by phosphorylation–dephosphorylation. This observation was later pursued by Bliss Forbush’s group who showed that NKCC in shark rectal gland acquires phosphate at serine and threonine residues in response to forskolin and cell shrinkage stimuli [79]. The identity of the kinases remained unknown until the discovery of Ste20-kinase association with K–Cl and Na–K–2Cl cotransporters [31, 101].

The regulation of SPAK/OSR1 and of NKCC1/NKCC2 is complex and has been addressed thoroughly in a recent review [42]. Here, we will briefly summarize some of the major findings. The current working model involves WNK kinases and PKC kinase isotypes acting upstream of SPAK and OSR1 [44, 76, 113] with the best characterized activation cascade involving WNK4–SPAK. In this cascade, WNK4 and SPAK physically interact through an RFxV motif in WNK4 and the CCT/PF2 domain in SPAK. WNK4 then phosphorylates SPAK at residues T243 in the catalytic domain and S383 in the regulatory domain, resulting in the activation of the Ste20 kinase. SPAK is then able to activate and phosphorylate the Na–K–2Cl cotransporters [44, 123]. An additional protein, which seems to play a critical role in cotransporter activation, is the calcium binding protein-39 (Cab39), also called mouse protein-25 (MO25). The adaptor protein comes in two isoforms Cab39 and Cab39-like, which are the products of two related genes. Both Cab39/MO25 isoforms serve as scaffolds, stabilizing Ste20-related kinases in their close or active confirmation [8, 36]. Knockdown of Cab39/MO25 suppresses the SPAK/OSR1-mediated phosphorylation and activation of NKCC1 [36]. We recently demonstrated that Cab39 can facilitate the trans phosphorylation and activation of SPAK monomers, in the absence of upstream activation by WNK kinases [104]. This finding argues that the Ste20 and WNK kinases can also act independently of each other, making the interpretation of the in vivo whole animal model data much more complicated.

Important questions in the field involve the multiplicity and possible redundancy of kinases and adaptor proteins. Why do we need multiple Ste20 kinases, WNK kinases, and Cab39 proteins? One possible answer is that the homologous proteins might be expressed in different cell types, might have slightly different properties, and might differently modulate the activity of specific transport proteins. Another conceivable answer is the need for redundancy that would minimize the prospect of affecting important physiological functions. For the Ste20 kinases, we have considered the need for two Ste20 kinases and the answer is mixed. On one hand, we found SPAK and OSR1 both expressed in sensory neurons and each participating in the activation of NKCC1 [51]. On the other hand, SPAK knockout mice mainly display a distal convoluted tubule (DCT) phenotype [136], whereas OSR1 knockout mice predominantly exhibit a TAL phenotype [77]. There are also significant differences between the two kinases as OSR1 is mostly found as a full-length product, whereas SPAK in kidney medulla is also observed as smaller fragments [57, 84]. Similarly, there are some major differences between WNK kinases, both in expression patterns, as well as molecular properties [83].

Another regulator of NKCC2 function in the thick ascending limb of Henle is a protein called SORLA (Sortilin-related receptor, LDLR class A repeats-containing) and which is mostly studied in the nervous system [96]. SORLA has been shown to regulate the subcellular localization of SPAK and its ability to phosphorylate NKCC2. In SORLA knockout mice, a smaller fraction of SPAK colocalizes with NKCC2 and phosphorylation of NKCC2 is decreased, whereas the apical expression is unchanged [106].

Vasopressin is the main hormone that affects NaCl reabsorption in the TAL [88, 127]. Its infusion in live mice increases apical expression and phosphorylation of N-terminal Thr96 and Thr101 residues of NKCC2 [54]. Vasopressin increases intracellular cAMP which phosphorylates and activates protein kinase A (PKA), this latter inducing phosphorylation (N-terminal Ser126 and C-terminal Ser874) of rat NKCC2, leading to increase trafficking of the cotransporter to the plasma membrane [10]. While vasopressin, calcitonin, glucagon, and β-adrenergic agonists stimulate NKCC2 trafficking by a cAMP-dependent PKA pathway, atrial natriuretic peptides, endothelin, α-adrenergic agonists, and nitric oxide induce an increase in intracellular cGMP, leading to a stimulation of the phosphodiesterase 2 (PDE2), resulting in a decrease in the apical expression of NKCC2 [3]. The AMP activated kinase (AMPK) seems to also regulate NKCC2 activity by phosphorylation at Ser126 (as shown in a murine macula densa cell line) and at Ser130 in HEK cells [39, 108]. However, it is not known whether AMPK regulates NKCC2 activity in the TAL. Yeast-two-hybrid assay experiments have shown that the glycolytic enzyme, aldolase B, and the vesicle scaffolding protein, SCAMP2, both interact with the C-terminal tail of NKCC2. While an increased in aldolase B or in SCAMP2 expression decreases membrane expression of NKCC2 in co-transfected opossum kidney epithelial cells, their possible role in TAL is unknown [137]. Finally, the lipid raft-associated protein MAL/VIP17 which is expressed in the TAL interacts with the C-terminal tail to possibly decrease the endocytosis of NKCC2 [16].

Final remarks

It is clear that much progress has been made over the past 30 years in understanding the physiology of the two Na–K–2Cl cotransporters. The availability of furosemide and bumetanide as potent and rather specific inhibitors at low concentrations has greatly facilitated the identification and characterization of cotransporter function in many tissues. Likely, it is also the efficacy of furosemide and bumetanide as potent and safe inhibitors of Na+ reabsorption in the thick ascending limb of Henle that has prevented the need to discover new inhibitory compounds. Considering the multiple roles that the NKCC1 plays outside the kidney, it would be useful to have compounds that modulate NKCC1 function without affecting NKCC2. This is particularly true when bumetanide is used in clinical settings to target NKCC1 function, with the knowledge that significant confounding effects are likely to result from blood volume reduction.

Aside from new pharmaceutical agents, crystal structures of the entire cotransporter proteins or protein domains are also in need. These structures should allow us to better understand how phosphorylation of residues in the N-terminal tail of the cotransporters results in cotransporter activation or how the structure of the tail affects the transmembrane core and the movement of the ions. The crystal structure will also give precious information on specific residues that interact with transported ions and inhibitors. These structures might also help in understanding the slippage modes of the transporter and the basis for the K+/K+ exchange.

As we have alluded to in this review, we have very limited information on regulation of both cotransporters at the transcriptional level. It is very unlikely that cotransporter regulation occurs mostly at the posttranslational level, as synthesis and stability of mRNA and protein must also be modulated by physiological processes. It is also clear that more and more proteins associating with the cotransporters are being identified. Whether or not all these proteins will have a significant impact on our understanding of cotransporter function is an unknown. It is clear that the discovery of Ste20 and WNK kinases has had a profound impact on our understanding of NKCC1, NKCC2, and even NCC activation, even if many of the details still need to be worked out. It is increasingly clear that these proteins are working in context-specific environments and that there are likely major differences in cotransporter regulation that are cell-type specific. The challenge will be to decipher these unique features in situations as close as physiologically possible.