Cryo-EM structures of DrNKCC1 and hKCC1: a new milestone in the physiology of cation-chloride cotransporters

Abstract

New milestones have been reached in the field of cation-Cl⁻ cotransporters with the recently released cryo-electron microscopy (EM) structures of the Danio rerio (zebrafish) Na⁺-K⁺-2Cl⁻ cotransporter (DrNKCC1) and the human K⁺-Cl⁻ cotransporter (hKCC1). In this review we provide a brief timeline that identifies the multiple breakthroughs in the field of solute carrier 12 transporters that led to the structure resolution of two of its key members. While cation-Cl⁻ cotransporters share the overall architecture of carriers belonging to the amino acid-polyamine-organocation (APC) superfamily and some of their substrate binding sites, several new insights are gained from the two individual structures. A first major feature relates to the largest extracellular domain between transmembrane domain (TMD) 5 and TMD6 of KCC1, which stabilizes the dimer and forms a cap that likely participates in extracellular gating. A second feature is the conservation of the K⁺ and Cl⁻ binding sites in both structures and evidence of an unexpected second Cl⁻ coordination site in the KCC1 structure. Structural data are discussed in the context of previously published studies that examined the basic and kinetics properties of these cotransport mechanisms. A third characteristic is the evidence of an extracellular gate formed by conserved salt bridges between charged residues located toward the end of TMD3 and TMD4 in both transporters and the existence of an additional neighboring bridge in the hKCC1 structure. A fourth feature of these newly solved structures relates to the multiple points of contacts between the monomer forming the cotransporter homodimer units. These involve the TMDs, the COOH-terminal domains, and the large extracellular loop for hKCC1.

INTRODUCTION

The discovery of the Na⁺-K⁺-2Cl⁻ cotransporter (NKCC) and K⁺-Cl⁻ cotransporter (KCC) as functional transport units came 40 years ago this year. This was the product of seminal work by Peter Geck (26), Peter Lauf (53), and Philip Dunham and Clive Ellory (22) and their colleagues. Further breakthroughs were achieved some 10–15 years later with the molecular identification/cloning of these cotransporters (16, 25, 27, 33, 60, 63, 64, 77), the creation of global knockout mice (15, 20, 23, 34, 35, 62, 73), and the identification of the main regulatory SPS1-related proline/alanine-rich kinase (SPAK)/Lys-deficient protein kinase (WNK) signaling cascade (24, 59, 66, 67, 75). Milestones were again crossed during these past few months with the determination of the cryo-electron microscopy (EM) structure of the zebrafish (Dano rerio) NKCC (DrNKCC) (7) and three cryo-EM structures of the human KCC (hKCC) (54). A timeline for the transporter is provided in Fig. 1. It should be also mentioned that an X-ray structure of the COOH-terminal tail of a bacterial cation-Cl⁻ cotransporter (CCC) was resolved and published in 2009 (76).

Fig. 1.

Time line highlighting milestones related to Na⁺-K⁺-2Cl⁻ and K⁺-Cl⁻ cotransport. The first publications mentioning furosemide (frusemide) and bumetanide appeared in 1964 and 1972, respectively. Na⁺-K⁺-2Cl⁻ cotransporter 1 (NKCC1) and K⁺-Cl⁻ cotransporters (KCCs) were discovered as functional units in 1980. For NKCC1, additional milestones were the discovery of NKCC1 as a phosphoprotein (55), the cloning in 1994, the generation of knockout (KO) mice in 1999, the discovery of regulatory kinases in 2001–2002, and the cryo-EM structure in 2019. For KCC, additional milestones were the 2-state phosphorylation model in1990 (40), the asymmetry of binding in 1991, the molecular cloning in 1996, the generation of knockout mice in 2002, the discovery of potent inhibitors in 2009 (10) and 2012 (9), and the cryo-EM structure in 2019. The surface-filled models of Danio rerio NKCC1 (DrNKCC1; top) and human KCC (hKCC; bottom) show dimers with individual cores shown in blue and light purple. For NKCC1, the COOH termini are shown in dark blue and purple. The extracellular domain of the blue monomer is shown in green, whereas the extracellular domain of the purple monomer is shown in orange. pNKCC1 and pKCC, phosphorylated NKCC1 and KCC, respectively; NEM, N-ethylmaleimide; SPAK, SPS1-related Pro/Ala-rich kinase; WNK, Lys-deficient protein kinase.

Now that the structures of two major CCCs have been solved, we thought it would be important to highlight their major features and useful to revisit some older functional data. For both cotransporters, the structure of the cytosolic NH₂ tail was, unfortunately, not resolved due to the highly flexible nature of the whole domain or parts thereof. Access to the structure of the NH₂ terminus of NKCC1 and NKCC2 would have been informative, as it contains a series of Thr residues that are keys to the activation of the cotransporters. It is worth noting that the authors of the NKCC1 structure paper could not detect an interaction between the NH₂ terminus and the core or the COOH termini, which is unfortunate in light of its importance in transport activation (7).

GENERAL ARCHITECTURE

The cryo-EM structures reveal dimeric assembly of both DrNKCC1 and hKCC1. Each monomer is composed of a short NH₂-terminal and a longer COOH-terminal cytosolic domain flanking a lipophilic core of 12 transmembrane (TM) domains (TMDs), with TM1–TM5 and TM6–TM10 being inverted structural repeats. This overall structure is conserved within the superfamily of amino acid-polyamine-organocation (APC) transporters (78). While this general description fits both transporters well, a major difference exists within the extracellular domains. In the KCC, a large 120-amino acid linker exists between TM5 and TM6. This large domain (highlighted in green and orange in Fig. 1) is unique to the KCC. A much shorter 30-residue loop links these two TMDs in the NKCC and NCC, the K⁺-independent Na⁺-Cl⁻ cotransporter. The KCC extracellular domain is composed of two pairs of antiparallel β-strands and four short α-helices. The architecture is stabilized by two disulfide bonds, Cys³⁰⁸–Cys³²³ and Cys³⁴³–Cys³⁵³. These bridges seem to be critical for function, since mutation of each of these four highly conserved Cys residues abolishes KCC2 activity (32).

Whether the disulfide bridges are always formed and are required for the stability of the dimer and/or for the capping function is unknown. It should be noted that KCC activity is activated severalfold by Cys-modifying reagents. In fact, K⁺-Cl⁻ cotransport was codiscovered as an N-ethylmaleimide (NEM)-activated K⁺ transport mechanism (53). The cotransporter is also activated by diamide, a reagent that promotes the oxidation of thiol groups in proteins to disulfides (48). Interestingly, the transporter is inhibited by iodoacetamide, another sulfhydryl (SH) reagent acting with slower kinetics than NEM. Activation and inhibition of the KCC by SH reagents with different kinetics indicated the presence of multiple SH targets (51). While the precise sites of action of NEM, diamide, and iodoacetamide are unknown, some have been postulated to be cytosolic (46). Kinases, for instance, are likely to be a site of action for NEM, diamide, and reactive oxygen species. In 2006, we demonstrated that SPAK autophosphorylation was diminished by 30% by 100 μM NEM, 40% by 100 μM diamide, and 70% by 0.03% H₂O₂. These observations, however, do not exclude the possibility that the Cys residues in the extracellular domain of the KCC may be also sensitive to thiol-modifying reagents. For this to be the case, the bridge between the two Cys pairs would have to be accessible to these reagents and able to form and dissolve reversibly.

In the two cotransporter structures, a salt bridge exists between an Arg residue located toward the end of TM1b and a Glu residue located toward the end of TM3. The residues are Arg²²⁹, Glu³¹¹ in DrNKCC1 and Arg¹⁴⁰, Glu²²² in hKCC1 (Fig. 2). Distances between the amino group of Arg and the hydroxyl group of Glu are 3.1 and 3.0 Å in DrNKCC1 and hKCC1, respectively. These residues are hypothesized to be part of a closed extracellular gate, as both transporter structures were captured in their “inside” conformation. As in the case of the Cys bridges mentioned above, whether the formation and breakage of these salt bridges are part of the conformational changes that occur during transport cycles also remains to be determined. Disruption of the salt bridge by mutation of conserved Glu²⁸⁹ in mouse KCC3 (mKCC3) to glycine completely abrogated transport (19). Interestingly, mutation of the residue only in one monomer was sufficient to affect the function of the dimer, suggesting that activation/deactivation of KCC occurs simultaneously within the two units of a dimer. An additional salt bridge (2.5 Å) exists in hKCC1 between Lys⁴⁸⁵ and Asp⁵⁷⁵. These residues are located in the extracellular loop between TM7 and TM8 and at the beginning of TM10. These residues are not conserved in NKCC1.

Fig. 2.

The extracellular gate in Danio rerio Na⁺-K⁺-2Cl⁻ cotransporter (DrNKCC1) and human K⁺-Cl⁻ cotransporter (hKCC1) structures. The gate is formed by salt bridges between conserved Glu and Arg residues in NKCC1 (blue) and KCC1 (green). An additional interaction exists in KCC1 between Lys⁴⁸⁵ and Asp⁵⁷⁵. TM, transmembrane.

DIMER FORMATION

The dimeric assembly in DrNKCC1 and hKCC1 is well in line with biochemical and functional studies of the CCCs. Visualization of proteins at ~160 kDa and at >250 kDa (depending of the strength of reducing agents) is quite standard in the field (2). Using cross-linking reagents such as 3,3′-dithiobis(sulfosuccinimidylpropionate) (DTSSP), Moore-Hoon and Turner were able to show that the structural unit of NKCC1 from the rat parotid gland was a homodimer (58). The DTSSP dimer was resistant to Triton X-100 and CHAPS treatments but sensitive to SDS. The complex could be broken by SDS concentrations >0.01%. Similar evidence of homodimerization was shown for the NCC (8) and KCCs (71). Because KCCs exist in four isoforms, KCC1–KCC4, and cells and tissues often express multiple isoforms (65, 69), the possibility that KCCs form heterodimers was tested in biochemical and functional studies. These studies provide solid evidence that different KCC isoforms can create structural and functional dimers (6, 18, 71). While the significance of this observation has yet to be established in vivo, it potentially provides a new dimension to the physiological roles of the transporters.

There are many points of contact between the two monomers in a cotransporter dimer. The structure of KCC1 reveals interaction between TM9 and TM12 of adjacent monomers. The interaction seals the dimer interface and produces a hydrophobic cavity that is filled with two putative glyco-diosgenin (GDN) detergent molecules. In addition to the TM interaction, which is loose in comparison with the Leu transporter (LeuT) and Na⁺-dependent glucose (SGLT) cotransporter of Vibrio parahaemolyticus (vSGLT) structures, the dimer is strengthened by hydrogen bonds between the carboxyl oxygen of Ser⁴⁰³ and the amine group of Lys⁴⁰⁵ of the corresponding monomer and additional hydrophobic interactions between extracellular domain residues of the two monomers (54). The situation seems to be a bit different for the NKCC, where the monomers interact through an interface consisting of TM11, TM12, and the COOH-terminal end of TM10 (7). No subunit interaction involving NKCC1 extracellular domains was reported. However, the NKCC1 structure reveals points of interaction between the COOH termini of both monomers in a dimer. Helix α₀ and 10 flanking residue linkers form a scissor-like structure allowing two COOH termini to interdigitate with a domain-swapped architecture. Points of interactions also involve the additional α-helices and β-strands, namely, α₃ and β₃. Interestingly, truncation of the COOH terminus of hNKCC1 and mouse NKCC1 (mNKCC1) allows for a truncated monomer to interact with another truncated monomer or with a full-length monomer, as long as the truncation occurs past the first structured domain that is involved in this interaction (17, 44).

ION BINDING SITES

The conservation of the ion binding sites in the KCC and NKCC structures is quite remarkable. The K⁺ binding site (S_K) in both cotransporters is formed by the side chains of adjacent Asp and Ile residues located at the end of TM1a, Pro and Thr residues at the end of TM6a, and a Tyr residue in the middle of TM3 (Fig. 3). K⁺ also forms a bond with a Cl⁻ located 3.8 Å above. The fact that the S_K is formed by five identical residues in the KCC and NKCC structures is puzzling in light of the knowledge that the affinity for external K⁺ is 5- to 10- fold higher in NKCC1 than KCC. Table 1 summarizes the K_m values for external K⁺ for stimulated KCCs and NKCCs from different cell types and species. Figure 4 shows the overlap between the side chains of residues involved in K⁺ coordination in both structures. With such similar coordinating sites, how can the binding affinities be so different? One possible explanation is that the average coordination distance of K⁺ is 2.94 Å in DrNKCC1, a bit shorter than that in hKCC1, which is 3.09 Å. Interestingly, binding affinity studies using chimeric proteins have highlighted individual residues in other TMDs affecting ion binding. Indeed, in their 1997–1998 papers, Paul Isenring and Biff Forbush and their colleagues created chimeric proteins using pieces of human and shark NKCC1 cDNAs. They identified key residues in TM2, TM4, and TM7 as involved in ion binding affinities (36–38). Again, as shown in Fig. 5A, there is 100% conservation between human and shark in the residues involved in coordinating the four ions (1 Na⁺, 1 K⁺, and 2 Cl⁻), while residues underlying the affinity differences are located in TM2, TM4, and TM7. Figure 5 also highlights some TM2 residues involved in modifying ion affinities. As seen in the structure (Fig. 5B), those TM2 residues highlighted in red are relatively far from the ions. The distance of the side chain of the four residues to K⁺ is >15 Å. TM2 and TM7 are still part of the core domain (54), and some portions of these participate in the creation of cavities. Thus, as these cavities provide access to the binding sites, the residues lining these cavities might be as important in determining binding affinities as are the ion coordination sites. It is important to note that the authors of the chimera studies understood the limitation of the approach, as they stated, “it is possible that the two pairs of residues [in TM2, i.e., Ser/Val or Met/Met (Fig. 5A)] are not directly involved in ion binding but that they confer affinity differences through conformational interactions.”

Fig. 3.

Ion binding sites in Danio rerio Na⁺-K⁺-2Cl⁻ cotransporter (DrNKCC1) and human K⁺-Cl⁻ cotransporter (hKCC1) structures. A: schematic representation of 5 conserved transmembrane (TM) domains of DrNKCC1 and hKCC1, with TM1 and TM6 showing broken helices that form K⁺ and Cl⁻ binding sites. Identical residues between DrNKCC1 and hKCC1 are shown in red over yellow background. Conserved residues are shown with green background. Otherwise, the DrNKCC1 sequence is shown in blue font and the hKCC1 sequence in black font. Dashed lines indicate interaction between residues and ions. For clarity, the valence of the ions is not drawn. B: location of the residues coordinating the ions in transmembrane domains. Numbers preceding amino acids show TM domain. S_K, K⁺ binding site; S_Cl1, Cl⁻ binding site 1; S_Cl2, Cl⁻ binding site 2; S_Na, Na⁺ binding site.

Table 1.

Affinity of KCC and NKCC1 for external K⁺

K+ Km, mM	Cell Type	Activation	References
	KCC
25	LK sheep RBC	Swelling	12
34	LK sheep RBC	Swelling	21
27	LK sheep RBC	NEM	49
19	LK sheep RBC	Diamide	47
58	LK sheep RBC	Swelling	1
17–30	Human RBC	Basal, NEM	42
27	Human RBC	Ghosts	70
21	Human RBC	Ghosts	61
115–130	Human RBC	Swelling	41
	NKCC
4.7	Human RBC		22
2	HEK-293 cells*	Low C1−	36
3.5–6	Human RBC		4
12	LLC-PK1 cells	Apical vesicles	3
	NKCC1
8	Xenopus laevis oocytes	Shrinkage	11

Affinity (K_m) is reported for the measurement of K⁺ influx in the presence of external Na⁺ [for Na⁺-K⁺-2Cl⁻ cotransporter 1 (NCC1)] and Cl⁻ at physiological concentrations. KCC, K⁺-Cl⁻ cotransporter; LK, low-K⁺; NEM, N-ethylmaleimide.

^*Transfected cells.

Fig. 4.

Superimposition of residues coordinating K⁺ in Danio rerio Na⁺-K⁺-2Cl⁻ cotransporter (DrNKCC1) and human K⁺-Cl⁻ cotransporter (hKCC1) structures. Backbones of DrNKCC1 residues are drawn in blue and those of hKCC1 residues in green. Bonds are indicated by dashed lines. Amino acid position of drawn residues is indicated by numbers. Position of K⁺ ions and Cl⁻ ion at Cl⁻ binding site 1 is indicated by spheres.

Fig. 5.

Comparison of human and shark Na⁺-K⁺-2Cl⁻ cotransporter (NKCC1) sequence and ion binding affinities. A: sequence of residues forming the K⁺ binding site (S_K), Cl⁻ binding site 1 (S_Cl1), and Cl⁻ binding site 2 (S_Cl2) and surrounding residues is identical between human and shark proteins. As shown in the box at left, affinities (K_m) for ions differ remarkably between the cotransporters from both species. Pairs of different residues in transmembrane (TM) 2 (Ser/Ala, Val/Leu, Met/Gly, and Met/Thr) were identified as affecting cation, but not anion, binding. As shown in the box at right, when the 4 human residues were introduced into the shark transporter, the cation affinities were intermediate between human and shark. B: cartoon showing that the pairs of residues in TM2 (labeled in red) are relatively far from the ion binding sites. K⁺ is shown as a purple sphere; the 2 Cl⁻ ions are shown as red spheres. TM1a and TM1b are drawn in blue and TM6 and TM2 in yellow and green, respectively.

The position of the first Cl⁻ binding site (S_Cl1) is coordinated by three consecutive residues located between TM1a and TM1b (Gly, Val, and Ile) above K⁺ at S_K (Fig. 3). Note that these binding site-forming residues are highly conserved within KCC and NKCC isoforms from Caenorhabditis elegans to human. We also noted that these residues were conserved in the orphan “transporter” encoded by the SLC12A8 gene, but not in the orphan protein encoded by the SLC12A9 gene (see alignment in Fig. 6). Interestingly, the NCC, which does not transport K⁺, lacks the key Tyr residue located in TM3 (this residue is, instead, replaced by a His).

Fig. 6.

Transmembrane (TM) domain alignments of 7 solute carrier family 12 (SLC12A) transporters and 2 SLC12A orphan proteins. Residues involved in K⁺ coordination [K⁺ binding site (S_K)], Cl⁻ coordination [Cl⁻ binding sites 1 (S_Cl1) and 2 (S_Cl2)], and Na⁺ coordination [Na⁺ binding site (S_Na)] are shown as colored symbols. Dashed lines between boxed residues represent interactions involved in closing the extracellular gate. The line between the Lys residue after EL6 and the Asp residue in TM10 is only for K⁺-Cl⁻ cotransporters (KCCs). Identical residues across all proteins are highlighted in red font over yellow background, conserved residues in black or blue over green or blue backgrounds, and nonconserved residues in black over white background. Alignments were done using Vector Nti 6.0. NCC, Na⁺-Cl⁻ cotransporter; NKCC, Na⁺-K⁺-2Cl⁻ cotransporter; CCC, cation-Cl⁻ cotransporter.

The second Cl⁻ binding site (S_Cl2) of NKCC1 is located below the S_K (Fig. 5), out of reach for ionic interaction with the cation. It is coordinated by the side chains of Gly, Ile, and Leu residues located at the beginning of TM6b and a Tyr residue positioned in the middle of TM10. Interestingly, this site also exists in the KCC structure (S_Cl2 in Fig. 6). Because it is understood that transport through KCC is electroneutral (1 K⁺ and 1 Cl⁻) (5, 39), one of the Cl⁻ sites must not be a transport site. As we suggest elsewhere (54), S_Cl1 is likely the transport site, since 1) S_Cl1 sits (from an inside perspective) deeply in the transporter, while S_Cl2 is more exposed to cytosolic solvent; 2) S_Cl1 and S_K align well with substrates in LeuT and Geobacillus kaustophilus APC transporter (GkApcT); and 3) the two ions can coordinate with each other, while S_Cl2 is 7.8 Å away from S_K. Note that the residues forming S_Cl2 are also conserved in the NCC, which couples the transport of 1 Na⁺ and 1 Cl⁻.

In a transport kinetic study performed in sheep red blood cells, we showed that the binding of DIDS to the cotransporter required external Cl⁻. However, the binding of DIDS was noncompetitive with respect to both K⁺ and Cl⁻ (13). This observation suggested the presence of a Cl⁻ “allosteric site” distinct from the Cl⁻ transport site. As indicated in the KCC1 structure paper (54), Cl⁻ binding at S_Cl2 stabilizes not only the S_K, but also the interactions between TM6a, TM6b, and TM10. Thus, one could speculate that Cl⁻ binding at S_Cl2 is required for DIDS binding to the transporter, but then the stilbene binding is noncompetitive with respect to the K⁺ and Cl⁻ transport sites. In a later transport kinetic study performed in HEK-293 cells overexpressing KCC2, we established that binding of ML077 to the transporter resulted in competitive inhibition with respect to K⁺ and noncompetitive inhibition with respect to Cl⁻ (10). These observations meant that ML077 binding to the free transporter somehow prevented K⁺ binding but did not interfere with Cl⁻ binding. These data were also consistent with those obtained with furosemide inhibition of sheep red blood cell K⁺-Cl⁻ cotransport (50). Together, these binding studies indicate complex interaction of drugs with the cotransporter. The fact that ML077 and furosemide interaction with the cotransporter prevents K⁺ binding, while DIDS interaction has no effect on K⁺, indicates distinct binding modalities of the two types of drugs to the KCC (see below).

If Cl⁻ (at S_Cl1) and K⁺ coordinate with each other, do both ions require the presence of their counterion to bind to their binding site? This is highly unlikely based on previous transport kinetic studies. Instead, one ion must be able to bind to its binding site, while the other site is unoccupied, and this initial binding facilitates the interaction of the counterion with its own coordination site. The electrostatic interaction between cation and anion then stabilizes the complex. In measuring K⁺ efflux in sheep red blood cells, we demonstrated that either K⁺ or Cl⁻ in the inward conformation could initiate the first binding step (Fig. 7). The random nature of K⁺ versus Cl⁻ binding was established through transport kinetics of K⁺ efflux from red blood cells loaded with different internal K⁺ and Cl⁻ concentrations ([K⁺] and [Cl⁻]). The kinetics data, determined using a model of rapid equilibrium, were consistent with random binding (12). This was later confirmed by the ability of internal Cl⁻ to trans-inhibit K⁺ influx in the absence of internal K⁺ and, conversely, the ability of internal K⁺ to trans-inhibit K⁺ influx in the absence of internal Cl⁻ (14). For the trans effect to be possible, each ion had to be able to bind to the transporter in the absence of its counterion.

Fig. 7.

Rapid-equilibrium kinetics of K⁺-Cl⁻ cotransporter. Model shows ordered binding with Cl⁻ interacting first on the outside, followed by K⁺. On the inside configuration, either ion can bind first (random binding). Gray bar represents the plasma membrane; arrows in and out represent conformational changes, which we used to term translocation steps. [Model was redrawn from Delpire and Lauf (12).]

Interestingly, the situation was different when the transporter was sitting in the outside configuration. The kinetics data of K⁺ influx at different external [K⁺] and [Cl⁻] were only consistent with a model where Cl⁻ binds first, followed by K⁺ (12). In agreement with this ordered-binding model, it was also demonstrated that K⁺ efflux could be trans-inhibited by external Cl⁻ in the absence of external K⁺, whereas K⁺ was unable to affect K⁺ efflux in the absence of Cl⁻. These functional data demonstrated asymmetry of binding on the outside and inside aspects of the membrane and clearly indicated different states of the ion binding sites when they were exposed to outside versus inside environments.

The ordered binding of ions from the outside in the KCC with Cl⁻ first, followed by K⁺, argues against ion access to the binding sites in single file, as K⁺ sits deeper than Cl⁻ in the structure. For Cl⁻ to bind first, followed by K⁺, Cl⁻ binding must not hinder the movement of K⁺ to its coordination site. Single-file access was proposed for the NKCC (37), as a strict order of binding on one side is followed by a strict reverse order of binding on the other (56). This assumption, however, was based on kinetic data obtained from β-adrenergic-stimulated duck red blood cells, and a different binding order was observed in a study that used mNKCC1 cRNA injected into Xenopus laevis oocytes (11). Also, the presence of a narrow channel in the upper part of the hKCC1 structure would not explain why K⁺ in the absence of Cl⁻ does not produce a trans effect. If the site is available for binding, then K⁺ should trap transporters in this configuration and reduce the pool of available transporter for efflux. Thus it is more likely that coordination at the K⁺ site requires, first, the binding of Cl⁻.

Molecular dynamic simulations of the hKCC1 structure hinted that binding at S_Cl2 was also critical. The simulation showed that, in the absence of Cl⁻ at S_Cl2, the loop between TM6a and TM6b relaxes, leading to the disruption of K⁺ at S_K and the weakening and loss of Cl⁻ at S_Cl1. Mutation of key residues coordinating the second Cl⁻ ion also abolished transport. Could S_Cl2 be the transport site, and S_Cl1, instead, be the “allosteric” site? If that was the case, why would the random binding of K⁺ and Cl⁻ on the inside be allowed? It is more likely that occupancy of Cl⁻ at the “allosteric” S_Cl2 is a prerequisite for K⁺ and Cl⁻ binding at their respective S_K and S_Cl1 transport sites.

The rapid-equilibrium kinetics of K⁺-Cl⁻ cotransport are consistent with one Cl⁻ binding site and Cl⁻ being released alongside K⁺. One could argue that the rapid-equilibrium model would fit the flux data if Cl⁻ at S_Cl2 was not released. In 2006, Javier Corzo highlighted the importance of time in the formation of a protein-ligand complex (7a). If Cl⁻ interaction at S_Cl2 was long-lasting, i.e., of an order of magnitude different from ion binding constants at S_Cl1 and S_K (K_off, the rate constant for the dissociation of the protein-ligand complex being extremely small), the rapid-equilibrium kinetics would only reveal the binding sites involved in K⁺-Cl⁻ transport. At this point, we cannot explain why S_Cl2 is used as a transport site in the NKCC, but not the KCC or NCC. As mentioned above, the requirement of Cl⁻ for DIDS binding to the sheep KCC suggested the presence of a Cl⁻ allosteric site, and S_Cl2 could represent such a site. In addition, kinetics analyses of K⁺ transport in duck red blood cells demonstrated that Cl⁻ reduces bumetanide inhibition (29) and [³H]bumetanide binding (28). These observations were interpreted as bumetanide and Cl⁻ competing for a common binding site. Additional evidence for the competitive nature of furosemide and Cl⁻ is presented in the 2000 comprehensive review by John Russell (68).

The substitution of Cl⁻ with different anions has also been used to characterize the properties of the KCC. In human and sheep red blood cells, Br⁻ and Cl⁻ are similarly carried by KCC, whereas $\text{N}{\text{O}}_3^{-}$, I⁻, and SCN⁻ are not transported (21, 53). An interesting behavior was observed, however, when cells were preincubated with SCN⁻, I⁻, or $\text{N}{\text{O}}_3^{-}$ at 37°C and returned to Cl⁻ when K⁺ flux was measured. In cells preincubated with the foreign anions, the KCC-mediated flux was activated by severalfold over cells that were preincubated with Cl⁻ and fluxed in Cl⁻. The anion series for the preincubation was SCN⁻ > I⁻ > $\text{N}{\text{O}}_3^{-}$ > Cl⁻ ≥ Br⁻ (45). In light of the presence of a second Cl⁻ binding site that is not directly involved in transport, we wonder if the foreign anions do not preferentially bind to (and “fail” to be easily released from) S_Cl2. A somewhat similar situation might exist with NKCC1. Indeed, while studies have shown that, in the absence of external Cl⁻, $\text{N}{\text{O}}_3^{-}$ does not allow NKCC1-mediated Na⁺ or K⁺ transport, in the presence of low (e.g., 10 mM) Cl⁻, $\text{N}{\text{O}}_3^{-}$ or SCN⁻ was readily transported through NKCC1 (43, 57).

DOCKING OF INHIBITORS TO THE hKCC1 STRUCTURE

To identify the putative drug binding sites on the hKCC1 structure, we hypothesized that the inhibitors were likely to utilize the ion permeation pathway and overlap with some of the ion binding sites. The idea was based on transport kinetic studies that demonstrated competitive inhibition with respect to K⁺ (10, 50). By using the ligand docking program AutoDock Vina from the Scripps Research Institute (74a), we first modeled the binding of ML077 starting from the position of the three ions: S_Cl1, S_K, and S_Cl2. ML077 is a potent inhibitor of KCC function: it binds rat KCC2 with an EC₅₀ of 560 nM (10). The xyz coordinates of the three ions were extracted from the cryo-EM structure and used as centers to dock the KCC2 inhibitor (Table 2). The top binding energies for the best solutions from these three positions were: −7.6, −7.6, and −8.5 kcal/mol. Clearly, the compound did fit better in the cytosolic portion of the permeation pathway (Fig. 8A). The ring structures of the inhibitory compound nicely fit within the hydrophobic groove or cavity, which is located at the bottom of the transporter core. Furthermore, the most distal methyl group is conveniently positioned 3.6 Å from the side chain of Tyr²¹⁶ (Fig. 8B). Since this residue also coordinates the placement of K⁺, it makes sense that the binding of ML077 would be competitive with the binding of K⁺ (10).

Table 2.

Predicted binding affinities of ligands to different positions in hKCC1

Center Coordinates	Ligand	Energy, kcal/mol	RMSDlow, Å	RMSDup, Å	Y216, Å
SCl1	ML077	−7.6	Best fit	Best fit	No
SK		−7.6	Best fit	Best fit	No
SCl2		−8.5	Best fit	Best fit	3.6
SCl2	D4.1 (25067400)	−8.2	Best fit	Best fit	3.5
	VU0463271	−8.5	Best fit	Best fit	3.4
	R4 (70694655)	−8.1	Best fit	Best fit	No
		−7.9	3.405	5.519	6.6
	R5 (70696692)	−8.6	Best fit	Best fit	No
	D4.6 (3228019)	−8.3	Best fit	Best fit	No
SCI2	DIDS	−7.7	Best fit	Best fit	No
		−7.5	3.361	8.975	No
		−7.5	3.450	4.980	No
	H2DIDS	−7.6	Best fit	Best fit	No
SCI2	Furosemide	−7.5	Best fit	Best fit	1.8

Coordinates were as follows: x = 109.058, y = 89.73, z = 119.906 for Cl⁻ binding site 1 (S_Cl1); x = 108.077, y = 89.328, z = 116.223 for the K⁺ binding site (S_K); x = 105.852, y = 91.041, z = 109.121 for Cl⁻ binding site 2 (S_Cl2). These parameters were used as center coordinates in the docking process. Size of the surrounding box was set to 30 Å, and exhaustiveness was set at 300. Root-mean-square deviation (RMSD) values are compared with best fit; both values (low and up) are 0.000 Å for the best fit. Values in Y261 column show distance between the inhibitor and the hydroxyl group of Tyr²¹⁶. Numbers in parentheses in Ligand column are PubChem Compound ID numbers. hKCC1, human K⁺-Cl⁻ cotransporter.

Fig. 8.

Docking of ML077 and some derivatives to the human K⁺-Cl⁻ cotransporter (hKCC1) structure. A: surface rendering of a portion of the hKCC1 structure, a large central cavity, where ML077 sits. Note positions of the 2 rings at the center of the cavity and position of the thiazol ring. B: some transmembrane (TM) domains, ions, and Tyr²¹⁶ residue are shown to better represent the position of the inhibitor. Note the 3.6-Å distance between the thiazol methyl group of ML077 and the hydroxyl group of Tyr²¹⁶. C: the inhibitor VU0463271 binds in the same configuration. Note position of the N-cyclopropyl group, replacing the methyl group of ML077 facing toward Cl⁻ binding site 2 (S_Cl2). D: addition of 1 carbon (to N-cyclobutyl) increases the bulkiness of this group and prevents binding of the compound to its original position. Docking now places the compound ∼6.6 Å from Tyr²¹⁶.

Also note the position of the side methyl group that faces the Cl⁻ at S_Cl2. When the methyl group was moved from position 4 to position 5 on the thiazol group (D4.1), the energy decreased to −8.2 kcal/mol, and the side methyl group moved away from the Cl⁻ position. D4.1 is an inactive compound (10). If the side methyl group is replaced by an N-cyclopropyl (3-carbon ring) moiety, KCC inhibition is increased. The affinity of this compound (VU0463271) is increased to 61 nM (9). The docking binding energy of this compound, versus ML077, was still −8.5 kcal/mol, but the distance of the methyl group was decreased to 3.4 Å (Fig. 8C). When the steric bulk of the tertiary amide was increased to an N-cyclobutyl (4-carbon ring) or an N-cyclopentyl (5-carbon ring) moiety (9), the compound could not be positioned in the proper location but moved toward the lower portion of the cytosolic cavity with no possibility of interaction with the side chain of Tyr²¹⁶ (Fig. 8D). While the N-cyclobutyl-containing compound retained some inhibitory properties, the larger N-cyclopentyl-containing compound failed to inhibit the cotransporter (9).

When we docked the stilbene derivative DIDS to the hKCC1 structure, the binding energy was weaker (−7.7 kcal/mol) and the molecule moved lower in the groove, with one sulfur group oriented toward S_Cl2. Note that binding modalities with slightly weaker binding energy (−7.5 kcal/mol) (Table 2) exist and place the inhibitor a bit higher in the groove, with the nitrogen atom or the sulfur atom of one of the isothiocyanate groups located 3.3 or 3.5 Å from Tyr²¹⁶ and one sulfonic acid group oriented either toward S_Cl2 or toward Gln⁵²¹ in TM8 (Fig. 9A). In these two latest configurations, the second sulfonic acid group orients toward the outside of the cavity, whereas the second isothiocyanate group fits nicely within the groove (Fig. 9B). These two positions are preferred over the first position, as docking of the related H₂DIDS compound, another potent inhibitor of sheep red blood cell K⁺-Cl⁻ cotransport (13), preferentially adopts this conformation (Table 2). Note that these latest conformations do not involve Tyr²¹⁶ and do not seem to overlap with K⁺ binding; therefore, these docking data are consistent with noncompetitive DIDS inhibition with regard to K⁺ (13). The placement of DIDS within the cavity, however, seems to trap Cl⁻ binding at S_Cl2, which, again, is consistent with DIDS binding requiring the presence of Cl⁻ without competing for its binding site (13).

Fig. 9.

Docking of DIDS, H₂DIDS, and furosemide to human K⁺-Cl⁻ cotransporter (hKCC1) structure. A: both DIDS (orange) and H₂DIDS (blue) acquire the same conformation within the hKCC1 structure. B: sphere representation of H₂DIDS showing how close the inhibitors fit within the cavity. One N = C = S group (top) faces toward exterior of the cavity, whereas the other N = C = S group (bottom) enters deeper into a niche within the cavity. C: position of furosemide with close-range (1.8-Å) hydrogen bond between furosemide and Tyr²¹⁶. D: sphere representation shows also how well furosemide fits in the transporter cavity. S, C, N, and H, sulfur, carbon, nitrogen, and hydrogen atoms.

Finally, docking of furosemide to the hKCC1 structure also preferentially placed the loop diuretic within the cavity with a possible hydrogen bond between the amine group of the ligand and the hydroxyl group of Tyr²¹⁶ (distance 1.8 Å) (Table 2, Fig. 9, C and D). Involvement of the Tyr residue in coordinating furosemide binding is, once again, consistent with competitive inhibition of furosemide binding with respect to K⁺ (50).

Thus, it seems that ML077 and its derivative DIDS and furosemide binding can be modeled to locations that are consistent with older functional data. However, single-site mutagenesis and functional experiments should be done to confirm these predictions. Because some of the residues will likely be involved in K⁺ or Cl⁻ binding or be essential to cotransport function, they will be resistant to characterization. Tyr²¹⁶, for instance, coordinates K⁺ binding, and its disruption leads to absence of function. Thus, no information about inhibitor binding is likely to come from mutating this specific residue. The substitution of other residues, however, might affect inhibitor binding without affecting the overall function of the transporter. Such studies should teach us about ligand-protein interaction. Alternatively, solving the structure of transporters occupied by the inhibitor will also confirm the docking studies. Also note that the docking simulations were done with the only structure that exists today, i.e., in an inside configuration. The binding of the drugs within the cotransporter internal cavity does not argue for them entering from the cytosolic side. It is far more likely that the drugs enter the cavity when the transporter is in the outward-facing configuration. Solving the structure in the outward-facing configuration should verify the validity of our binding predictions.

THE COOH TERMINUS AND ITS INTERACTION WITH THE CORE DOMAIN

As mentioned in the introduction, the structure of the COOH-terminal tail of a bacterial cation-Cl⁻ transporter was crystalized in 2009 (76). The structure revealed the presence of two highly structured domains, each containing five β-strands and three or four α-helices. The two domains are still present in the DrNKCC1 structure, and the structure alignment between the domains of the two transporters is reasonably good, despite low sequence similarity. Figure 10 shows the alignments of domains I and II. While the β-sheet cores are highly conserved, there are additional α-helices in both domains of DrNKCC1. This observation indicates conservation and divergence in structure of the COOH terminus in CCCs.

Fig. 10.

Alignment of the 2 structured domains of the COOH terminus of human K⁺-Cl⁻ cotransporter (hKCC1) and Danio rerio Na⁺-K⁺-2Cl⁻ cotransporter (DrNKCC1). Chains drawn in blue are from the Methanosarcina acetivorans cation-Cl⁻ cotransporter structure; chains drawn in purple are from the DrNKCC1 structure.

The structure of DrNKCC1 also reveals proximity between the α-helix located between TM2 and TM3, the intracellular loop between TM10 and TM11, and the NH₂-terminal end of the first (scissor) helix of the COOH terminus. Whether this proximity allows for domain interaction is speculative.

CRYO-EM STRUCTURES AND FUTURE WORK

As seen in Fig. 7, the model of K⁺-Cl⁻ cotransport (if correct) consists of seven reaction states: three transporter states on the outside and four on the inside. As four ions bind to NKCC1, the total number of states for Na⁺-K⁺-2Cl⁻ cotransport is 10. Because carrier-mediated transport can be described by rapid-equilibrium kinetics (12, 72), we know that, at any given time, transporters are not equally divided per state (i.e., 1 of 7 for KCC and 1 of 10 for NKCC1). Since the rapid-equilibrium method assumes that binding steps are fast and translocation (transition) steps are slow, in the presence of ions, most transporters should exist in the loaded configuration. Thus, for both KCC and NKCC, the majority of transporters should be loaded carriers in the outside and inside configurations. The number of transporters on each side is dictated by the ion gradients (52). If the product [K⁺] × [Cl⁻] on one side exceeds the same product on the other side, then more transporters should be loaded on that particular side. Note that this might only be true if [K⁺] and [Cl⁻] are higher than the K_m values for each ion. In the absence of ions, the number of transporters on each side is possibly determined by some intrinsic carrier properties that favor one orientation. Kinetics properties of volume-stimulated KCC in sheep red blood cells were indicative of transporters favoring the inside configuration. This makes intuitive sense for a transporter that typically functions as an efflux transport mechanism.

The fact that the KCC structure reveals an inward configuration with a closed extracellular gate with strong ionic bonding raises the following question: What factors are involved in disrupting the extracellular gate? One could easily assume that once the transporter is fully loaded on the inside, the transporter transitions to the outside configuration, allowing the extracellular gate to open for ion release. The rapid-equilibrium kinetics, however, are only valid if the empty transporter is able to also transition from inside to outside configurations, and vice versa, and the extracellular gate is open to allow ions to bind on the outside. Trans-inhibition of K⁺ efflux by outside Cl⁻ in the absence of external K⁺ is also only possible if translocation (or transition) is allowed for the empty carrier. Thus, the extracellular gate should also be able to form and dissolve in the absence of ions.

It is fascinating that the hKCC1 and DrNKCC1 structures are captured in the inside configuration. As NKCC1 is an inward transport mechanism, one would assume that it possesses intrinsic asymmetry that would favor the outside configuration. Why is it that, in the presence of all ions on both sides, half of the cryo-EM particles do not capture the cotransporters in the inside configuration, while the other half of the particles depict the cotransporters in the outside configuration? The single-particle cryo-EM technique only determines high-resolution structures from rigid and homogeneous particles. It is possible that the outside configuration is unstable or, rather, heterogeneous in detergent environments, and those particles are discarded during the data processing. Clearly, for both hKCC1 and DrNKCC1 in detergent micelles, it seems that only the inside states are stable states.

Future work should focus on finding ways to resolve the structure of the cotransporters in other configurations. Is it possible to isolate the transporters with unequal concentrations of ions on each side of the membrane? One way would be to reconstitute the transporters into liposomes, where inside and outside solutions can be differentially adjusted. Tonggu and Wang recently reported a 3.5-Å cryo-EM structure of the large-conductance Ca²⁺-activated K⁺ (BK) channel reconstituted in liposome membranes (74). If the cryo-EM particles of hKCC1 were to be analyzed in the absence of external K⁺, for instance, but in the presence of Cl⁻, would the transporter be seen as partially loaded in its outside configuration? Also, for KCCs that seem to have a Cl⁻ allosteric site that allows the binding of foreign anions, what would a structure of hKCC1 reveal if different amounts of SCN⁻ or I⁻ versus Cl⁻ are used? Because of the lipid solubility of these lyotropic anions, it is also possible that they also affect the protein-lipid interface. This could likely be visualized with additional cryo-EM experiments. Finally, can the structure be solved in the presence of furosemide, DIDS, or ML077 binding? Much more information can be gained by manipulating the conditions for cryo-EM particle preparation.

Future work should also address the possibility that CCCs directly mediate the cotransport of water and ions. This notion is counter to the classical view of ion transport, which is that water is supposed to “follow” the movement of ions across membranes through other pathways, such as water channels. Experiments performed in mudpuppy choroid plexus epithelium have shown that the movement of water is directly linked to the movement of ions through the KCC (79). The movement of water was inhibited by furosemide, and as a hallmark cotransport, it could be elicited against a water (osmotic) gradient. A fixed stoichiometry of 500 molecules of water per 1 K⁺ and 1 Cl⁻ per transport cycle was determined for KCC. Similar water-transporting properties were also associated with the NKCC1 (30) with a measured stoichiometry of 570:1:1:2 for water, Na⁺, K⁺, and Cl⁻ (31). Because of the water transport properties of the Na⁺-coupled glucose transporter SGLT1, one could assume that this phenomenon is a property of all cotransport mechanisms. However, while NKCC1 was able to cotransport water with ions in X. laevis oocytes, NKCC2 in the same system was shown to transport ions but not water (80). Many questions can now be addressed with the help of structural data. What are the structural determinants of water transport through NKCC1 and KCC? Why would water flow along the path of the ions through the protein, rather than having ions shedding their water sphere on one side as they bind and gaining a new sphere from available water on the other side when binding is released? Why would NKCC1, but not NKCC2, transport water?

Additional work should also be guided by the structural information. For instance, before the structures were known, the conserved Tyr was easily identified as a residue involved in K⁺ binding, as it exists in all K⁺-dependent cotransporters but is missing in the NCC. In fact, we had attempted to substitute the corresponding His in the NCC back to Tyr to force the NCC to transport K⁺. This effort was unsuccessful, indicating that the approach was too simplistic. The knowledge of the structure gives us opportunities to design better experiments.

In conclusion, a wealth of new information came with the resolution of the structure of NKCC1 and KCC. This information not only can help explain biochemical and functional data gathered in the past, but it also provides an opportunity to devise experiments to answer new questions. This new trove of data should keep us busy for the foreseeable future, until the field of CCCs is driven anew by another breakthrough.

GRANTS

This work was supported by National Institute of Diabetes and Digestive and Kidney Diseases Grants DK-93501 and DK-110375 and by Transatlantic Network of Excellence Grant 17CVD05 from the Leducq Foundation.

DISCLOSURES

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

AUTHOR CONTRIBUTIONS

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