key: cord-0038676-bhjpvwbg
authors: Engel, Andreas; Walz, Thomas; Agre, Peter
title: The aquaporin family of membrane water channels
date: 2004-02-24
journal: Curr Opin Struct Biol
DOI: 10.1016/s0959-440x(94)90217-8
sha: 5b7f659e286c6342c626a0bf43edcf2c06286fb5
doc_id: 38676
cord_uid: bhjpvwbg

The rapid movement of water across the plasma membranes of certain cells has been a long-standing puzzle to membrane biophysicists and physiologists; the discovery of the red cell channel-forming integral protein has provided a molecular solution to this puzzle. The identification of this protein has led to the recognition of a family of related water-selective channels, the aquaporins, that are found in animals, plants and microbial organisms. In order to provide insight into the remarkable but simple function of these membrane proteins, their structures are being elucidated.

All biological membranes exhibit some water permeability as a result of diffusion across the lipid bilayer; however, the degree of water permeability differs considerably between tissues and cell types. Mammalian red blood cells and renal proximal tubules are extremely permeable to water molecules. This remarkable feature led biophysicists to propose the hypothesis that water-selective channels must exist. The physiological importance of red cell water channels is still unclear; however, the kidney has a major role in body water balance, and permits survival despite severe water deprivation. The plasma membranes of certain other tissues are also highly permeable to water molecules, and transmembrane water movements are involved in diverse physiological processes including the secretion of cerebrospinal fluid, aqueous humor, sweat, tears, saliva, bile and amniotic fluid. Several observations have provided clues about how water channels function (reviewed in [1]). For example, it is thought that water traverses water channels as a singlefile column of molecules, because the ratio of osmoticdriven water permeability (Pf) to diffusional water permeability (Pd) is much greater than one. The selectivity for water is apparently explained by a narrow aqueous pore since such a channel would be permeable to water but not to protons, other ions, or small molecules. The Arrhenius activation energy measured for red cell Pf is similar to the diffusion of water in bulk solution (<5 kcalmol-1), which is considerably below the activation energy of the diffusional water permeability of simple phospholipid bilayers. Red cell and renal tubule water channels are selectively inhibited by HgC12 and certain organomercurials, indicating the presence of a critical sulflaydryl group in the pore [2] . Also, the wa-Biology 1994, 4:545-553 ter channels of red cells and renal proximal tubules seem to be constitutively activated, whereas the renal collecting duct is only permeable to water in the presence of vasopressin, which leads to the redistribution of collecting duct water channels from intracellular vesicles to the apical surface (reviewed in [3] ). Radiation target analysis initially showed that the size of the molecular water channel is 30kDa [4] , but attempts to determine its structure by biochemical approaches [5] and expression cloning [6] were unsuccessful. This was in part due to the inability to label the water channel with its substrate (as water is ubiquitous), the lack of a highly specific inhibitor (HgC12 labels all free sulfllydryls), and the diffusional permeability of cell membranes. Recently, however, significant progress has been made in identifying and characterizing the proteins responsible for water permeability, and the first steps towards a determination of their structure have been made.

The first molecular explanation for transmembrane water movements emerged from the study of a novel red cell membrane protein identified by the serendipitous development of an antibody. This 28 kDa channelforming integral protein (CHIP) has not been noticed by traditional SDS-PAGE as it fails to bind Coomassie stain [7] . CHIP resides within the lipid bilayer, as it is quantitatively included in the membrane vesicle fraction even after extraction with 1 M KI. CHIP is very abundant in red blood cells, -200 000 copies per cell [8] , and in apical brush border cells of renal tubules where it constitutes 4 % of the total protein [9] . ence of Mg 2+ ions. STEM mass measurements (Fig. 3b ) yielded a mass-per-area of single-layered CHIP square arrays of 4.1+0.27kDanm-2 (n=2160). The mass of one unit cell is therefore 378 kDa, accommodating two CHIP tetramers of 268kDa and phospholipids. Correlation averages calculated from electron micrographs of negatively stained CHIP lattices exhibited a prominent four-fold rotational symmetry [34"'] . As shown in Fig. 3d each CHIP tetramer has four stain-excluding elongated domains of approximately 2.7 nm length and 1.6 nm width (bright shades) surrounding a central stain filled depression (dark shades). The tetramers are separated by rhomboid stained areas of approximately 7.3 nm length and 4.7 nm width that represent the liPid bilayer.

Vesicles folded from CHIP lattices have been shown to retain full biological activity [34" ]. The 3 ~tm diameter vesicles exhibited a high degree of water permeability and were shrunken within 10msec. The osmotic water permeability, Pf, was calculated as 0.472 cmsec-1, allowing the unit water permeability, pf=Pfxarea per subunit=5.43 × 10 -14 cm 3sec -1 subunit -1 to be calculated from the packing density of CHIP within the two-dimensional crystals. This value agrees with that measured from smaller proteoliposomes, pf=4.6 x 10 -14 cm 3 sec-I subunit--1 [16"]).

Projections of tilted negatively-stained two-dimensional CHIP crystals allowed the three-dimensional density distribution of the CHIP tetramer to be determined at a resolution of better than 2 nm [14" ]. The three-dimensional map showed two tetramers of membrane-spanning CHIP monomers per unit cell that are incorporated in the bilayer with opposite orientation. This results in different surface topographies of adjacent tetramers as shown by the perspective view in Fig. 4 . Each tetrainer exhibits two stain penetrated indentations about the four-fold axis. One face corresponds to the extracellular side of the tetramer and has subunits that project approximately 0.7 nm from the bilayer and surround a deep central cavity o f -3 nm in diameter. The second face corresponds to the cytoplasmic side of the tetramer and has subunits that project approximately 1 nm from the bilayer and surround a narrower central cavity. The wide and narrow cavities must be separated by an unresolved, thin barrier about the four-fold axis.

The initial structural predictions for CHIP were derived from the primary sequence by the Kyte and Doolittle hydropathy analysis (Fig. 5a ) [17] . This plot strongly resembles that first deduced for MIP [10] and is shared by the aquaporins and all members of the MIP family, whose function is still unknown (reviewed in [35] ). The predicted topology features a small cytoplasmic amino terminus, a 35. residue cytoplasmic carboxyl terminus, six bilayer-spanning domains of 20-25 residues (which may be helical), and five connecting loops of which A, C, and E are extracellular and B and D are cytoplasmic (Fig. 5b) . . Perspective view of the surface-rendered three-dimensional density map. As a result of the p422~ symmetry of the two-dimensional crystals, both faces of the CHIP tetramer are visible. The central CHIP tetramer extends by approximately 1 nm from the bilayer surface and surrounds a narrow pore (short arrow). The four adjacent tetramers are incorporated in the opposite orientation, and protrude by less than 0.7 nm from the bilayer. The four subunits surround a deep cavity (long arrow) that has a diameter of 3 nm. Reproduced with permission from [14" ].

As noted in the initial report [17] , the structures of loops B and E are unusual, as loop B is significantly hydrophobic and loop E is slightly hydrophobic. A second feature of the protein is the presence of an internal repeat which is present in both the amino-terminal half of the molecule and the carboxy-terminal half [36, 37] . This structure is shared by all animal and plant homologs, suggesting the duplication of an ancestral gene [35] . About 50% of the residues in the amino-and carboxyterminal halves of CHIP are conservative replacements, and 20% of the residues are identical. The most highly conserved domains are loops B and E, which both contain a sequence motif Asn-Pro-Ala (Fig. 5a,b) . These motifs are preserved in all members of the MIP gene family from diverse species [35] , suggesting an essential structural or functional role.

The topology of CHIP has been established by single site substitutions and insertional mutagenesis of CHIP expressed in Xenopus oocytes [38" ]. There are consensus glycosylation sites containing asparagine in both loop A and loop E; however, only the loop A site is occupied by glycan, and therefore only this site is known to be extracellular. Additional analyses of different multiple site-directed CHIP mutants suggest that the glycosylation is kinetically controlled, since mutants that were retained in the endoplasmic reticulum had much higher proportions ofglycosylated subunits than the normal 1:3. The 20-amino acid E1 viral epitope was introduced into the amino and carboxyl termini and into the connecting loops B, C, D, and E. The water permeabilites ofoocytes expressing each of the recombinant proteins were determined; only recombinants with inserts in loops B and E were not active, indicating that these loops are important for function. Use of antiserum specific for the E1 and epitope selective proteolytic degradation of intact oocytes expressing recombinants established that loop C is extracellular. Similar analyses of microsomes isolated from oocytes established that the amino-and carboxyl-termini and loops B and D are cytoplasmic. These findings provide experimental evidence that the tandem repeats span the bilayer in opposite orientation.

Red cell membrane water channels are known to be reversibly inhibitable by HgC12, presumably because the HgC12 modifies sulphydryl group near the outer face of the water channel aperture [2] . When the four cysteines in CHIP were individually replaced by serine and expressed in oocytes, each recombinant exhibited wild type biological activity, but the Cys189--)Ser mutation -in loop E was no longer inhibitable by HgC12, indicating that Cys!89is the mercury-sensitive site [39] . When residue 189 was replaced by a smaller residue, alanine, similar behavior was noted. When larger residues such as methionine were introduced, the Pf was greatly reduced, suggesting that residue 189 is critical for the function of the CHIP water pore. The glycosylation pattern was high mannose, indicating that the non-functional mutants at position 189 are not properly targeted to the Golgi, presumably as result of aberrant folding. These findings were subsequently confirmed by others [40] .

The identification of the mercury-sensitive residue at position 189 in loop E raised the question of whether residue Ala73, which is found at the corresponding position in loop B with respect to the Asn-Pro-Ala motif, might behave similarly. The double mutant Ala73--~Cys/Cys189--)Ser was constructed, and the Pf indicated full biological activity. When treated with HgC12, this double mutant was reversibly inhibited, although the sensitivity to the reagent was approximately one-third that of the wild-type CHIP [41°']. Although located in opposite halves of the molecule, both residues 73 and 189 appear to reside near critical narrowings of the water channel through CHIE Other studies in which similar conservative mutations were introduced at corresponding sites in loop B or loop E confirmed that these locations were structurally similar. A possible explanation of this behavior is that loop 13 may fold into and out of the cytoplasmic leaflet of the bilayer, and loop E may do the same at the extracellular side (similar to the H5 loop of potassium channels). If the two halves of the molecule are juxtaposed, loops B and E could form a single central sleeve surrounded by the six bilayer spanning domains (Fig. 5b) . This is referred to as the 'hourglass' model.

Although CHIP exists exclusively as a tetramer, numerous experiments suggest that individual subunits have independently functioning water pores. Coex- The cross-section obtained by negative stain electron microscopy and digital image processing suggests a thin barrier about the four-fold axis of the tetramer. (d) Provided that the putative a-helical membrane spans form the peripheral wall of the CHIP tetramer, loops B and E must assemble into a thin barrier that carries four water pores and joins the CHIP monomers into a tetrameric complex. Reproduced with permission from [17] and [41 " ] .

pression of certain in vitro transcribed R.NAs encoding mercury-insensitive CHIP mutants and wild-type CHIP resulted in membranes with Pfs that were partially inhibitable by HgC12 [39] and expression of dimeric CHIP constructs of varying compositions produced similar results [41" ]. Expression of certain CHIP mutants (such as Cys189--qMet and a truncation mutant, Asp237--gSTOP) in oocytes did not result in increased water permeability; however, when these two mutants were coinjected, the subunits complemented each other, and the oocytes exhibited a Pf characteristic of wild-type CHIP [41" ]. These studies confirm that multiple subunits are needed to form active CHIP water channels.

CHIP has been studied by biochemical, biophysical, cell biological, physiological, ultrastructural, crystallographic, and molecular biological techniques. These studies combined have demonstrated that CHIP is an oligomer which resides primarily between the leaflets of the lipid bilayer. It is distributed in many but not all water-permeable epithelia and has a remarkably high permeability for water, yet it is not permeable to ions or other small molecules. The protein assembles as a tetramer with a central cavity, which extends from the outer surface deep into the molecule, and with a smaller central cav-ity at the inner surface. Primary sequence and topology studies indicate that the two halves of the molecule that originate from a primordial gene duplication event span the bilayer in opposite orientations and comprise loops B and E, which both contain the highly conserved motif Asn-Pro-Ala.

This information is not easily explained by a single structural model. Our current hypothesis is that individual subunits exist as highly asymmetric, hourglass-like structures with water-conducting pores assembled from overlying loops B and E, which form a shelf adjacent to six helical bilayer-spanning domains. The inability of these subunits to exist as free monomers probably reflects the need for transmembrane proteins to have vertical hydrophobic faces surrounded by the phospholipid bilayer. When four CHIP subunits oligomerize, the shelves formed by loops B and E assemble into a thin barrier which separates the deep central cavity at the outer surface from the narrower cavity at the inner surface of the tetramer. This barrier is penetrated by four narrow pores, each contributed by an individual subunit (Fig. 5c,d) .

Since our efforts began, cDNAs encoding several homologous proteins have been isolated from various mammalian and plant tissues. Some of these proteins cause membranes to become permeable to water as assessed by the oocyte swelling assay and are termed aquaporins. The first aquaporin identified was CHIP (gene symbol AQP1). Newly identified mammalian homologs include: AQP-CD (gene symbol AQP2); the probable vasopressin-regulated water channel of collecting duct [42] , AQP3, the product of which forms the baso- Aquaporins are also likely to be involved in certain diseases of animals and plants. The presence of CHIP in alveolar capillary endothelium suggests that it may be important in the pathogenesis of fresh water drownings, and its presence in ciliary epithelium indicates a possible involvement in glaucoma. A patient with nephrogenic diabetes insipidus was found to be a compound homozygote for two mutations in AQP2 [47°] . In addition a plant homotog is known to be pathologically induced in roots during nematode infestations, a major scourge in agriculture [46• ].

The structural model presented here is based on a low resolution three-dimensional map obtained by negative Aquaporins Engel, Walz and Agre 551 stain electron microscopy, molecular biological analyses, and on the putative presence of six helical membranespanning regions suggested by structural prediction. The first aim of high-resolution electron microscopy will be the identification of these membrane spanning domains, and this goal may be reached by cryo-microscopy of the two-dimensional CHIP crystals that are currently available. The particular properties of loops B and E indicate that a full understanding of the water permeability requires the elucidation of the atomic structure.

Papers of particular interest, published within the annual period of review

Water Movement through Lipid Bilayers, Pores, and Plasma Membranes, Theory and Reality

Farmer REh Inhibition of Water and Solute Permeability in Human Red Cells

Structural-Functional Features of Vasopressin-lnduced Water Flow in the Kidney Collecting Duct

Functional Unit of 30kDa for Proximal Tubule Water Channels as Revealed by Radiation Inactivation

Current Understanding of the Cell Biology and Molecular Structure of the Antidiuretic Hormone-Stimulated Water Transport Pathway

Water Channels in Cell Membranes

Agre P: Identification, Purification, and Partial Characterization of a Novel M r 28,000 Integral Membrane Protein from Erythrocytes and Renal Tubules

Erythrocyte M r 28,000 Transmembrane Protein Exists as a Multisubunit Oligomer Similar to Channel Proteins

Agre P: CHIP28 Water Channels are Localized in Constitutively Water-Permeable Segments of the Nephron

The Major Intrinsic Protein (MIP) of the Bovine Lens Fiber Membrane: Characterization and Structure Based on cDNA Cloning

Conformational Properties of the Main Intrinsic Polypeptide (MIP26) Isolated from Lens Plasma Membranes

Secondary Structure Analysis of Purified Functional CHIP28 Water Channels by CD and FTIR Spectroscopy

Nonpolar Environment of Tryptophans in Erythrocyte Water Channel CHIP28 Determined by Fluorescence Quenching

13:in press. The first low-resolution three-dimensional map of negatively stained two-dimensional CHIP crystals

Tetrameric Assembly of CHIP28 Water Channels in Liposomes and Cell Membranes: a Freeze-Fracture Study

The size and morphology of CHIP water channels were analyzed in CHIP proteoliposomes and native membranes by electron microscopy

This paper describes the detailed ultrastrucutral and functional analysis of CHIP proteoliposomes

Isolation of the cDNA for Erythrocyte Integral Membrane Protein of 28 Kilodaltons: Member of an Ancient Channel Family

Growth Factor-Induced Delayed Early Response Genes

Isolation of a cDNA for Rat CHIP28 Water Channel: High mRNA Expression in Kidney Cortex and Inner Medulla

The Aquaporin CHIP Gene: Structure, Organization, and Chromosomal Localization

Localization of the CHIP28 Water Channel in Rat Kidney

Sabolic h Localizalization of the CHIP28 Water Channel in Reabsorptive Segments of the Rat Male Reproductive Tract

Agre P: Distribution of the Aquaporin CHIP in Secretory and Resorptive Epithelia and Capillary Endothelia

Developmental Gene Expression and Tissue Distribution of the CHIP28 Water-Channel Protein

Concurrent Expression of Erylhroid and Renal Aquaporin CHIP and Appearance of Water Channel Activity in Perinatal Rats

Aquaporin CHIP: the Archetypal Molecular Water Channel

Appearance of Water Channels in Xenopus Oocyles Expressing Red Cell CHIP28 Protein. Science

Cloning, Functional Analysis and Cell Localization of a Kidney Proximal Tubule Water Transporter Homologous to CHIP28

Expression of Multiple Water Channel Activities in Xenopus Oocytes Injected with mRNA from Rat Kidney

Cultured Bovine Corneal Endothelial Cells Express CHIP28 Water Channels

Reconstitution of Functional Water Channels in Liposomes Containing Purified Red Cell CHIP28 Protein

Functional Reconstitution of the Isolated Erythrocyte Water Channel CHIP28

2D Crystallization: from Art to Science

Biologically • , Active Two-Dimensional Crystals of Aquaporin CHIP

This is the first report on two-dimensional crystals of Aquaporin CHIP and illustrates that isolated membrane proteins can recover their full biological activity when reconstituted in two-dimensional crystals in the presence of phospholipids

The MIP Family of Integral Membrane Channel Proteins --Sequence Comparisons, Evolutionary Relationships, Reconstructed Pathway of Evolution, and Proposed Functional Differentiation of the 2 Repeated Halves of the Proteins

Evolution of the MIP Family of Integral Membrane Transport Proteins

Tandem Sequence Repeats in Transmembrane Channel Proteins

Membrane Topoi-* ogy of Aquaporin CHIP: Analysis of Functional Epitope-Scanning Mutants by Vectorial Proteolysis

A coronavirus E1 protein epitope was used to establish the external localization of loop C, and the cytoplasmic localization of loop D, and the amino and carboxy termini

The Mercury-Sensitive Residue at Cysteine 189 in the CHIP28 Water Channel

A Point Mutation at Cysteine 189 Blocks the Water permeability of Rat Kidney Water Channel CHIP28

This paper reports the construction of the double mutant Ala73--~Cys/Cys189--~Ser, which was used to demonstrate the functional symmetry of loops B and E suggested by their sequence homology. Complementation assays were used to show

Cloning and Expression of Apical Membrane Water Channel of Rat Kidney Collecting Tubule

Molecular Cloning and Expression of a New Member of the Aquaporin Family (AQP3) with Permeability to Glycerol and Urea in Addition to Water Expressed at the Basolateral Membrane of Kidney Collecting Duct Cells

This is the first aquaporin known to have glycerol and urea permeability in addition to water permeability

Verkman AS: • Molecular Cloning of a Mercurial-Insensitive Water Channel Expressed in Selected Water-Transporting Tissues

This is the first report of a mercurial-insensitive Aquaporin

The Vacular ® Membrane Protein y-TIP Creates Water Specific Channels in Xenopus Oocyles

This was the first plant Aquaporin discovered

Root-Knot Nematode-• Directed Expression of a Plant Root-Specific Gene

This paper shows that promoter elements required for nematode ToRB7 are different from those that control normal root specific expression

Human Kidney Water Channel Aquaporin-2 is Involved in Vasopressin Dependent Concentration of Urine

We are grateful to U Aebi for his continuous support and critical discussions, to G Preston for his help with the manuscript, and to D Brown for providing the micrograph shown in Fig  lb. The work was supported by the ME Miiller Foundation of Switzerland, by the Swiss National Foundation for Scientific Research grant 31-32536.91 (to AE), and by National Institutes of Health grants HL33991and HL48268 (to PA).

This is the first report on mutations in AQP2 responsible for non-X-linked nephrogenic diabetes insipidus.A Engel and T Walz, ME Mfiller-lnstitut for Microsocpic Structural Biology, Biozentrum, Klingelbergstrasse 70, CH-4056, Basel, Switzerland.