Structure and function of water channels
To secure ion channel functions, the water channel AQP must exclude ion permeation when water molecules permeate under osmotic pressure. Thirteen water channels have been identified in the human body, AQP0 to AQP12. These water channels are involved in numerous physiological processes. For example, AQP1 molecules are expressed in many cells and tissues and maintain homeostasis.30) Three billion water molecules can permeate a single channel of AQP1 in a second while permeation of ions and even protons is excluded. The pH regulation for biological cells is crucially important, because pH is strongly related to signal transduction, proliferation, and survival of cells. Hydronium ions, however, do not need to move from one side to the other to change the pH conditions. Protons easily pass through membranes by exchanging hydrogen bonding partners of a water wire, like toppling dominoes (Fig. 4A). Therefore, proton exclusion requires the separation of hydrogen bonds that cause a large potential barrier. To prevent proton permeation, therefore, water cannot pass through with the high speed of 3 billion water molecules per second in one channel. Therefore, how do water channels accomplish such difficult functional roles? To answer these puzzling questions, we proposed the following hydrogen-bond isolation mechanism based on the structure of AQP1 analyzed by electron crystallography.7)
|Fig. 4. |
Proton permeation mechanism and structure of the water channel AQP1. A: Protons can easily permeate through the so-called water wire by changing hydrogen bonding partner like toppling dominos. B: Ribbon representation of the AQP1 structure, which is formed by a right handed six-helical bundle and two short helices. C: Cross sectional view parallel to the membrane surfaces at the NPA motifs. Loops B and E fold back into the membrane from the cytoplasmic and extracellular membrane sides, respectively. After the NPA motifs, which are almost perfectly conserved in water channels, the loops form short helices. The asparagine back bonded by its carbonyl group with NH group of main chain and, as a result, the amide groups of the asparagines come into the channel almost parallel to the channel axis. The constriction of 3 Å is formed by hydrophobic amino acid residues except the asparagines as all related residues are indicated in stick representations.
The monomer structure in the AQP1 tetramer has a right-handed helical bundle and two short helices, HB and HE, whose N-terminal sides face each other, especially the proline residues of the two asparagine-proline-alanine (NPA) sequences, which are almost perfectly conserved in AQP family proteins, called the NPA motif (Fig. 4B, C). Figure 4C shows a sectional view of the atomic model of AQP1 at the NPA sequences, where the two short helices begin to form helical structures by the back-bonded carbonyl group of asparagine with the NH-group of the main chain. The arrangement of the amino acid residues at the NPA motifs forms a narrow hydrophobic constriction except the two asparagine residues (Fig. 4C). This constriction formed at the NPA sequences with 3 Å in diameter is measured by van der Waals distance and therefore blocks the passage of any solutes and hydrated ions whose sizes are ∼8 Å or larger. Only water molecules whose diameter is ∼2.8 Å can pass through the constriction at which the two amide groups of the asparagine residues of the NPA motifs come into the channel almost parallel to the channel axis. The narrowest constriction, which is far narrower than the water diameter of 2.8 Å, is observed at the extracellular side from the NPA motifs and named the ar/R part, indicating conserved aromatic residues and arginine (R). Therefore, the ar/R pore must be enlarged when water passes through the constriction, whereas the conformation of the other parts might not change.
The electrostatic field formed by two short helices, whose helical dipole is indicated by the blue and red colors in schematic Fig. 5A, forces the oxygen of the water molecule to orient to the NPA side, coming close to the motifs to allow the water oxygen to smoothly form hydrogen bonds with amide groups of the two asparagine residues of the NPA motifs at the pore constriction (Fig. 5A, F). The two hydrogen atoms of the water molecule orient perpendicular to the channel axis because of the molecular orbitals of water and also the arrangement of the two amide groups of the asparagine residues. The orientation might prevent the water molecule from forming hydrogen bonds with the other water molecules that are close to the NPA motifs because the perpendicular arrangement of the water molecule to the channel axis renders the two hydrogen atoms too far away from the adjacent water molecules. We named this the hydrogen-bond (H-bond) isolation mechanism. We were unable to observe the water arrangement in the channel of AQP1, because the resolution was quite limited at 3.8 Å.7) Nevertheless, high-quality 2D-crystals of AQP0 were prepared by Walz’s group, and we could analyze the structure at a high resolution, up to 1.9 Å.23) The water molecules were clearly discriminated in the channel, and the distance between these water molecules was importantly ∼4 Å, preventing the formation of H-bonds between adjacent water molecules (Fig. 5B). The true structure of AQP0 thus shows an arrangement of water molecules very similar to that in the modeled AQP1 shown in Fig. 5A, and the H-bond isolation mechanism was supported by high-resolution analysis of this water channel. The analyzed structure of AQP0 was in a closed state, however, and actually only three water molecules were observed in the channel (Fig. 5B). Therefore, high-resolution structure analysis of a fast water permeation channel is still needed.
|Fig. 5. |
Hydrogen bond isolation mechanism based on structural analysis of 2D-crystals. A: Schematic emphasizing the helical dipole of two short helices in the water channel. The helical dipole moment is greater when the water channel is in the lipid bilayer, which has a characteristic distribution of dielectric constants. B: Densities of water molecules in AQP0 channel analyzed at 1.9 Å resolution by electron crystallography. This structure shows a closed state of AQP0 and only three water molecule densities are observed. C: Densities of eight water molecules in the AQP4 channel analyzed by electron crystallography. Water molecules are numbered from the extracellular side to the cytoplasmic side. Highly conserved parts, NPA and ar/R motifs, in the water channels are indicated. D: Water densities are blurred in the AQP4 structure, which was analyzed in a detergent-solubilized condition by X-ray crystallography. E: Schematic figure of a water channel in a native condition. Electrostatic fields formed by strong helical dipoles force water molecules approaching the center to orient facing the NPA site and, in cooperation with the orientation of the water molecules, polypeptide carbonyl groups sticking out from the hydrophobic channel surface. Therefore, the eight water molecule positions in the channel are stabilized by the carbonyl and amide groups, which are hydrogen-bonding partners for the hydrogen and oxygen atoms, respectively, of the permeating water molecules. These partners of the water molecules form a guide rail on the hydrophobic wall of the channel fitting to the water orientation produced by the dipole moments of the two half pore helices HB and HE. F: Structure of AQP4 including water molecules in the channel indicating the number of water molecules. The hydrogen bonds among the water molecules in the channel are separated close to the NPA motif, because the distances between W4 and W5 and W5 and W6 are 3.9 Å and 3.8 Å, respectively.
The water permeation speed of AQP4, which is the predominant water channel in the brain expressed in the endfeet of astrocytes, ependymal cells, and glial lamellae of the hypothalamus, is identical to that of AQP1, a fast water permeation channel. We analyzed the AQP4 structure at 2.8 Å resolution and could clearly discriminate eight water molecules in the channel (Fig. 5C). The densities of the water molecules in the channel at the higher resolution of 1.8 Å by X-ray, however, were blurred (Fig. 5D).31) This counterintuitive notion could be attributed to the different surrounding conditions of the membrane protein with and without a lipid bilayer of the analyzed structures by electron and X-ray, respectively (Fig. 5C, D). We could actually analyze the structures of lipid molecules together with AQP4 molecules by electron crystallography, but the X-ray structure was analyzed without lipid molecules in the detergent-solubilized condition. The lipid bilayer has a characteristic distribution of dielectric constants, as shown in the schematic figures (Fig. 5A and E). The conspicuous distribution of dielectric constants in the membrane produces a large dipole moment of the half membrane-penetrating short helices, whereas the dipole moment without a lipid bilayer is small.32) In cooperation with the electrostatic field formed by the two short helices having a large dipole moment, the arrangement of carbonyl groups in the channel acts as a binding site in highly hydrophobic surfaces and lowers the energy barrier for water molecules to enter the water channel at specific locations in the narrow hydrophobic channel. Together with the strong helical dipole, these carbonyl groups might be essential for the very fast water permeation through an otherwise perfectly hydrophobic narrow channel (Fig. 5E). These water stable positions guided by the polypeptide carbonyl groups and amide groups of NPA motifs may allow for 3 billion or more water molecules pass through in 1 second. Distances between water positions close to the NPA center are 3.9 Å (W4–W5) and 3.8 Å (W5–W6), by which no hydrogen bonds could be formed and proton permeation is inhibited (Fig. 5F). The importance of the helical dipole moment can be observed in many examples of water and ion channels. Negative ions are stabilized by the positive electrostatic field of the short helices (Fig. 6A) and vice versa, as shown in the schematic representations of ion channels in Fig. 6B–D. These examples strongly suggest the importance of the helix dipole of short helices for ion channel functions. In gap junction channels, which are discussed later, a plug is formed by six short helices at the N-termini of connexin molecules, but their arrangement is not symmetrical and their helical charge may help to stabilize the up-down configuration of these helices (Fig. 6E).
|Fig. 6. |
Schematic figures showing arrangements of short helices with helical dipole moments. A: Chloride channel. B: Potassium channel. C: Sodium channel. D: Calcium channels of TRPV1 and TRPVA1. Interestingly, TRPV1 has one short helix in one subunit, but TRPVA1 has two short helices in one subunit. E: The plug of the gap junction channel is formed by six short helices formed by the N-termini of connexin molecules that are arranged, not symmetrically, but rather in an up-down configuration, as observed in the 3D-structure analyzed by electron crystallography. For clarity, the schematic figure shows only three of the six helices of the plug.
When we speak a language, as a typical example of a human action, our brains consume oxygen and glucose, and the temperature rises at an active local area. Different languages may activate different local areas in the brain. Therefore, regulation of these various conditions in the brain is important for the neural activities and functions, and AQP4 molecules coincidentally show characteristic expression patterns in the brain, especially in the endfeet of astrocytes, ependymal cells, and glial lamellae of the hypothalamus. At the endfeet of astrocytes, crystalline arrays called “orthogonal arrays” were observed long ago, which were later confirmed to be AQP4.33) AQP4 has splicing variants and the short isoform, which starts from the 23rd methionine, AQP4M23, produces large orthogonal arrays, whereas the long isoform starting from AQP4M1 disrupts the array and never yields large orthogonal arrays. Structural analysis of the 2D-crystal by electron crystallography revealed the same orthogonal array lattice constant of 69 Å as observed in intact cells. Our structure provides a basis for explaining the orthogonal array forming intermolecular interactions (Fig. 7A, B). The mechanism underlying the disruption of the arrays by 22 amino acid residues at the N-terminus (Fig. 7C), however, was unclear, even when the structure was carefully observed. The array formation and disruption were then observed in Chinese hamster ovary (CHO) cell membranes by deleting the amino acids one by one from the N-terminus end using the freeze-fracture method. Sodium-dodecyl sulfate-digested freeze fracture replica labeling was performed as described previously34) with small modifications. When the 17th amino acid, a cysteine residue, was deleted, large arrays were suddenly observed. The cysteine residues at positions 13 and 17 of AQP4M1 are conserved in homologs from different species. To verify the possible involvement of Cys13 and/or Cys17 in interfering with the formation of square arrays, we generated AQP4M1 mutants, in which either one or both cysteine residues were mutated to alanine residues. CHO cells transfected with these mutant constructs were analyzed by sodium-dodecyl sulfate-freeze fracture replica labeling. Electron microscopy showed that the single cysteine mutant AQP4M1/C13A formed no square arrays, while the other single cysteine mutant, AQP4M1/C17A, formed a few small arrays. By contrast, the double cysteine mutant, AQP4M1/C13,17A, formed many and large square arrays (Fig. 7D, E). These findings indicate that the presence of even one of the two cysteine residues is sufficient to disrupt the formation of square arrays. Cysteine residues can serve as sites for posttranslational modifications by fatty acids. Our chemical analyses revealed that both Cys13 and Cys17 of AQP4M1 expressed in CHO cells are palmitoylated.35) Palmitoylation of these cysteine residues was confirmed and the lipid modification of even one cysteine was confirmed to disrupt the formation of arrays in AQP4. Without any lipid modification, interactions of tetramers form the orthogonal arrays but a lipid modified amino terminus hinders the array-forming interaction (Fig. 7D, E). Therefore, a longer isoform with cysteine residues cannot form orthogonal arrays, while a shorter isoform without cysteine residues forms large orthogonal arrays.
|Fig. 7. |
Array formation and disruption of AQP4 molecules. A: The AQP4 molecules form an array structure that was analyzed by electron crystallography to reveal a molecular distance in the crystal similar to that of the array. B: Enlarged structure of the square region in A showing intermolecular interactions by which the array might be stabilized. C: Amino acid sequence of the N-terminal side of AQP4 and two methionine residues are observed at the positions 1 and 23. There are long (M1AQP4) and short (M23AQP4) isoforms in this type of water channel. Two cysteine residues are also observed at positions 13 and 17. D: M1AQP4 molecules expressed in CHO cells form no array structure as shown in the freeze-fracture image. E: M1AQP4 mutant molecules without C13 and C17 form large array structures as shown in the freeze-fracture image, suggesting the importance of both cysteine residues.
Our double-layered crystal structure showed an adhesive homophilic interaction of AQP4 molecules (Fig. 8A). Although this may represent the cell adhesive function of AQP4 molecules in vivo, it is possible that this adhesive interaction of AQP4 artificially occurs in the crystal. We therefore confirmed the adhesive function of AQP4 in L-cells expressing no endogenous cell adhesive molecules. Whereas AQP1 showed no cell adhesion, AQP4 was confirmed to have a weak cell adhesive function (Fig. 8B). Interestingly, the arrangement of the channel’s upper and lower layers is not straight but shifted, i.e., staggered (Fig. 8C), and the adhesive interaction between 310 helices is also rather weak. Therefore, water permeation through AQP4 channels could separate the adhesive membranes. The movement could act as an osmotic pressure sensor (Fig. 8D, E). Actually, large numbers of AQP4 molecules are expressed in the glial lamellae of the hypothalamus, which has important osmo-, thermo-, and glucose-sensory brain functions. AQP4 is involved in cerebral edema, multiple sclerosis, neuromyelitis optica, and bipolar disorder.30) Large amounts of AQP4 molecules are expressed in the brains of patients with bipolar disorder.36) How this multifunctional water channel is involved in such higher order cerebral dysfunctions, however, remains unclear. To better understand the protein functions, we attempted to analyze knockout mice. Simple AQP4 knockout mice, however, show no significant phenotype other than reduced edema. As a typical example, the structure of the glial lamellae of simple knockout mice is similar to that in wild-type mice. Therefore, based on the structural analysis of AQP4, we produced genetically modified mice expressing mutant AQP4 without the 310 helix to elucidate the physiological adhesive functions of the water channel. The phenotypes of these mice should prove to be very interesting once they are fully studied.
|Fig. 8. |
Cell adhesive function of AQP4. A: Structure of adjoining molecules of AQP4 monomers in 2D-crystals. The interaction between the 310 helices of loop C is suggested to be at the homophilic binding part in AQP4. B: Cell adhesive function of AQP4 was confirmed in L-cells, which have no endogenous cell adhesive molecules, whereas AQP1 has no such function. C: AQP4 molecules in upper (brown) and lower (blue) layers are not aligned straight, but are half shifted. D: Schematic figure of glial lamellae of the hypothalamus. Green indicates the AQP4 tetramer. E: Water permeation and its pressure separate the adhesive membranes of the glial lamellae.