Aquaporins, a large family of integral membrane proteins that share a common and novel fold, transport water across cell membranes at a rate so fast that it is effectively limited only by the rate of diffusion. A large number of family members (around 200 at the last count) have been found in a wide variety of organisms, including humans, bacteria, and plants. Their prevalence may seem unusual at first; it was long assumed that simple diffusion accounted for all water movement through biological membranes. However, the ability to control water uptake and release is fundamental to life, so once the existence of aquaporins had been confirmed it was not so very surprising to find them present - and their fold to be conserved - between many different organisms. It is also unsurprising, then, to find that aquaporin malfunction may be involved in many physiological disorders, including diabetes inspidus (DI) and cerebral oedema. However, it is only in the past year or so that atomic or near-atomic resolution structures have begun to emerge for these proteins, bringing with them the promise of a detailed model for the action of these proteins; how, for example, is such a high specificity for water maintained?

The Discovery Of Aquaporins

In the beginning, nobody was looking for aquaporins. Their discovery was a matter of chance; one day in 1987, during isolation of the 32kD transmembrane (TM) component of the red cell Rh blood group antigen, a 23kD polypeptide was also found. The new protein stained poorly with coomassie, and was initially assumed to represent a proteolytic fragment of Rh. This was made unlikely when initial studies showed that the protein had a hydrophobic composition and also existed in an N-glycosylated form; it was disproved completely when the N-terminal sequence of the new protein was found to be unrelated to that of Rh. This was a new protein. Because of its abundance in erythrocyte and renal membranes, it was suggested as the long sought-after water channel, and designated AQP1; supporting evidence included the fact that water permeation was observed to be inhibited by the addition of mercury, consistent with a proteinaceous channel being involved.

AQP1 Structure And Function

The advancement of understanding in this field has been rapid, not least as a result of the ready availability of milligram quantities of the protein, from erythrocytes. It has also helped that there is a simple functional assay for the protein, involving expression in Xenopus oocytes that are placed in hypotonic buffer. The rate at which they explode increases proportionally with the amount of AQP1 present.

Within the last six months, the structure of AQP1 has been obtained by electron diffraction, using 2D crystals, at a resolution of 3.8Å and an R factor of 0.399. This, together with the phase residual of 40.3°, would seem to indicate that the quality of the data is relatively poor; however, it corresponds to a figure of merit of 0.76, which is similar to or better than the initial phase normally determined by X-ray crystallography studies. The data is also in good agreement with previous structure determination at lower resolutions.

The AQP1 monomer contains 269 amino acid residues, within which there are two tandem repeats. Each repeat folds in the same manner, with three TM a-helices and one short connecting loop helix; in addition, there is a 20% sequence conservation between the two halves of the protein. All of this suggests that there was a gene duplication event early in the protein's evolution. The organisation of a folded monomer is shown in figure one; note that both ends of the polypeptide chain are to be found on the cytoplasmic side of the membrane.

Overall, there are six tilted TM helices, plus two half-helices that dip into the membrane from opposite sides and meet in the middle and together form a right-handed coiled bundle. A single aqueous pathway is formed through the centre of the protein. In the 3.8Å structure, protrusions corresponding to amino acid side chains can be clearly resolved and permit unambiguous assignment of the 1° structure data. There is a site at asparagine 41, within loop A (extracellular, connects M1 and M2), for attachment of a polylactosaminoglycan moiety; loop D, the symmetric counterpart of loop A, does not have such a site because it is a cytoplasmic loop.

Loop B folds back into the membrane and contains an asparagine-proline-alanine (NPA) motif that is repeated in an equivalent position in loop E. As a result, the two sequences are juxtaposed within the membrane, and are a key component of the aqueous pore structure. They hold the loops together by interactions between pro77 and pro193, whilst the asparagine residues cap the N-terminal ends of HB and HE by hydrogen-bonding with the main-chain NH groups at val79 and arg195. The positions of HB and HE are further stabilised an ion pair (his74, in HB, with glu17, in M1), a salt bridge (arg195, in HE, with glu142, in M4), and more hydrogen bonds.

In cellular membranes, four AQP1 monomers associate into a tetramer (figure two). The stability of this complex is thought to arise from accommodation of each monomer as a tight-fitting wedge. Amino acid residues interacting with the acyl chains of the lipids in the tetramer are found in M3, M6, HB, and HE, all of which are at the exteriors of the monomers. In addition, each monomer interacts with two neighbours via its TM a-helices. M1 and M2 form left-handed coiled-coil interactions with M4 and M5 of the neighbouring monomer, respectively, whilst M1 and M5 interact at the intracellular surface, and M2 and M4 interact at the cytoplasmic surface of the membrane. This last is thought to involve a stabilising network of hydrogen bonds involving such residues as ser59, glu62, and gln65 (all in M2), and gln148, cys152, and thr156 (all in M4). Finally, there may be interactions between loops that contribute to tetramer stability; loops A surround the four-fold axis on the extracellular surface, whereas loops D do so on the intracellular surface.

The Mechanism Of Water Transport

The NPA motifs interact with each other and form part of the surface of the aqueous pore (figure three); M2, M5, and the C-terminal halves of M1 and M4 make up the remaining surface.

At its narrowest point, the pore has a 3Å diameter (measured from Van der Waal's surfaces); a single water molecule has a diameter of 2.8Å, and so even though the span of this constriction is only one amino acid residue, it is sufficient to exclude larger ions and molecules. AQP1 does not contain a structure to liberate ions from their hydration shells (unlike, say, the potassium channel), which also helps maintain selectivity. Lastly, there is a hydrophobic surface formed lining the pore adjacent to asn76 and asn192 of the NPA motifs.

Key residues involved in water permeation through the pore include cys189, which projects into the pore and is the site of inhibition by mercury; and his180, which is apparently crucial to selectivity. In the structurally related aquaglyceroporins (permeable, as the name suggests, to glycerol), his180 is notably replaced by a glycine residue.

The exact mechanism of water transport is still not known with certainty. The most likely model makes use of the partial charges of helix dipoles to restrict the orientation of water molecules passing through the pore constriction such that their oxygen molecules face the NPA motifs (figure four). The two asparagine residues mentioned previously are particularly important in this model, as they are believed to be the residues that interrupt the continuous chain of hydrogen bonds that a single file of water molecules would otherwise form - and thus H3O+ are prevented from moving through the pore. This mechanism involves the breaking and formation of hydrogen bonds, and the activation energy of AQP1 has been shown to be around 3 kcal, or about the same as the energy in one hydrogen bond.

The Aquaporin Family

There are many members of the aquaporin family, as mentioned above; currently, somewhere in the region of two hundred examples are known. They come from a wide variety of organisms, from E.coli to plants.

At this point, it should be made clear that there are two main branches of the AQP family; there are the 'true' aquaporins (AQPs), which transport water as described above, and the aquaglyceroporins (AQGPs), which transport glycerol (and other related compounds). There are no significant differences between the folds of the two families (as exemplified by the structures of AQP1 and GlpF, the E.coli glycerol transporter), so for a time it was unclear how the two groups could be so specific for different molecules. However, the structure of GlpF has now been determined at 2.2Å resolution, from crystals grown in the presence of 2M glycerol. As a result, the crystal structure also shows the locations of three glycerol molecules, as well as an intervening water molecule, trapped as they passed through the pore.

It seems that the pore dimensions of AQP1 and GlpF are broadly similar; however, the central cavity of GlpF is somewhat the wider of the two (although still not wide enough to accommodate a solvated ion), with a constriction on the extracellular side caused by trp48 (termed a 'hydrophobic corner'), in the region designated as the selectivity filter. As this name suggests, it is considered to be very important in determining the selectivity of the channel, and this is supported by the observation that in AQP1 trp48 is replaced by ile60, which is a significantly smaller residue. The GlpF pore has an amphipathic lining that allows glycerol molecules to be oriented so that their hydroxyl groups from hydrogen bonds to the polar half of the pore lining, whilst their carbon atoms interact with the apolar half. The pathway through the channel can be viewed as consisting of a series of activation barriers that separate intermediate binding sites through the channel.

The Physiological Roles Of Aquaporins

AQP1, the first aquaporin to be identified, has been found to be constitutively present in apical and basolateral membranes in kidney nephrons, at levels sufficient to account for the observed level of water reabsorption by that tissue; in a variety of eye tissues; and in a variety of capillary endothelia. Partly on the basis of this quite extensive distribution, AQP1 was at first thought to be an essential gene product. However, subsequent investigation has revealed that there are - rarely - humans who have no AQP1 at all. In fact, all such people so far identified have been women who developed anti-AQP1 antibody during pregnancy. Surprisingly, though, none have showed an obvious clinical phenotype; it is unclear whether this is due to the presence of natural backup systems or of compensating mutations. Mouse studies suggest that it may be the latter; AQP1 is critical for renal water reabsorption in that animal, although they tolerate the lack of the protein well until they are deprived of water.

To date, nine other mammalian aquaporins have been found, named AQP0 to AQP9 (AQP0 was first known as the major intrinsic protein in lens, MIP, but has subsequently been renamed). This group includes both AQPs and AQGPs; six of the former (AQP0, AQP2, AQP4, AQP5, AQP6, and AQP8), and three of the latter (AQP3, AQP7, and AQP9). All show high structural homology to AQP1, but each has some variations related to differences in their function and regulation. Their localisations and properties are summarised in table one.

AQP2 deserves examination in more detail. The first hints of its existence came from studies on AQP1- mutants, where the observation was made that renal ducts retain vasopressin-sensitive water permeability. The protein itself was found by a homology cloning approach and, as indicated in the table, thus far its expression has only been observed in the renal collecting duct. What is interesting about it is that it is not constitutive; it is regulated by vasopressin, which has long been known to cause the redistribution of intracellular vesicles to the apical surfaces of cells by the so-called 'membrane shuttle mechanism'. The exact mechanism is thought to involve a basolateral membrane V2 receptor couple to activation of adenylyl cyclase and subsequent phosphorylation of the C-terminus of the AQP2 protein. Participation of a heterotrimeric member of the Gi family is required. There has also been a cAMP regulatory element found in the 5' flanking DNA of the AQP2 gene, which supports a role for transcriptional regulation in the story of this pore.

However, the chief interest in AQP2 lies not with the protein itself, but with its clinical importance; it is implicated in many imbalances of water metabolism. For example, diabetes insipidus (DI) results from inadequate levels of vasopressin, and leads to secretion of large volumes of dilute urine, even as the affected individual becomes dehydrated. Nephrogenic DI occurs when the kidney fails to respond to vasopressin that is present, and has been associated with functionally disruptive mutations in the AQP2 gene. In addition, it is thought that many disorders may produce secondary decreases or increases in AQP2 expression. Bipolar disorder is commonly treated with lithium, for example, and polyuria is a problematic side effect; lithium has been observed to cause a 90% decrease in AQP2 expression. Patients suffering congestive heart failure often succumb to refractory pulmonary oedema caused by water retention; this has been observed to occur concurrently with increased AQP2 expression. AQP2 has thus attracted considerable attention as a target for pharmacologic intervention.

Table One: Mammalian Aquaporins

AQUAPORIN

EXPRESSED IN

ROLE

CLINICAL IMPORTANCE?

NOTES

AQP0

Lens

Confers low permeability

Mouse mutants suffer congenital cataracts

Not inhibited by Hg

AQP1

Kidney nephrons; eye; capillary endothelia

Primarily water reabsorption

No apparent change in phenotype in deficient humans

The first AQP found

AQP2

Principle cells of the renal collecting duct

Not constitutive; regulated by vasopressin through short-term exocytosis to the PM and long-term biosynthetic mechanisms

Nephrogenic DI has been associated with AQP2 mutants; may often be an indicator of other disease states

Vasopressin action involves adenylyl cyclase and phosphorylation of the C-terminus of AQP2

AQ(G)P3

Kidney

Renal water reabsorption?

?

Shows homology to GlpF; is somewhat permeable to urea; may be vasopressin-regulated

AQP4

Brain; optic nerve; fast-twitch muscle in rats

As an exit port for excess brain water; to restore osmotic balance in other tissues

Excess brain water can be lethal in cerebral oedema; expression is decreased in fast-twitch muscle in a mouse model of muscular dystrophy

Has two possible translation initiation sites, so exists in two possible forms; is not affected by Hg; may be actively regulated by PKC

AQP5

Apical membranes of respiratory pathways

Involved in airway humidification, and the release of saliva and tears

Possible gene therapy agent; its gene has been engineered into an adenoviral vector

-

AQP6

Kidney

?

?

Shows low permeability to water (like AQP0)

AQ(G)P7

Spermatids and seminiferous tubules

A port for water and glycerol (carbon source) in mature sperm

?

The human homologue may be expressed in adipose tissue and involved in lipolysis

AQP8

Isolated from testis; pancreas; liver; colon

Unclear. May be permeable to water and urea

?

-

AQ(G)P9

Leucocytes

Appears to transport urea but not glycerol

?

-

Aquaporin Expression In Complex Tissues

A large number of important tissues, including lungs, airways, and secretory glands have complex expression patterns involving multiple aquaporins. The precise subcellular localisations of different aquaporins during these expression patterns are being confirmed by electron microscopy studies; two examples are shown in figure five. In general, it is found that unique aquaporin homologs are expressed at unique intracellular sites and function together to provide transcellular water flow. For example, in kidneys AQP2 is found in the apical membranes, whereas AQP3 and AQP4 reside at the basolateral membranes in the outer and inner medulla.

Nonmammalian Aquaporins

Other groups of organisms show special requirements for aquaporins when compared to mammals. The reasons for this can be as simple as 'the organism lives in a very wet environment' (e.g. amphibians; so far, homology cloning has found one constitutively expressed member of the family, in toad bladder). Sometimes the reasons are unclear; a Drosophila homologue named Big Brain has been found, with its expression being essential for correct tissue development, but does not appear to transport water. In addition, as might be expected, bacterial homologues for both branches of the family have been found although - with the exception of E.coli, which has GlpF and AQPZ - never both in the same organism. AQPZ is a tetramer with structural homology to AQP1 that is important under conditions of maximum growth rate.

However, the group with the most aquaporins known is - mammals included - plants. Plants are entirely dependent on their local environment for water, and it shows; for example, Arabidopsis thaliana has recently been found to have at least 23 different aquaporin homologs. There are two basic subgroups; those that are expressed in the plasma membrane (PM Intrinsic Proteins, PIPs), and those that are expressed in the tonoplast (TIPs). A wide variety of physiological proteins have been ascribed to plant aquaporins; for example, g-TIP (the first plant homologue evaluated in the Xenopus oocytes system) behaves similarly to AQP1, and is found in the stem. Mutant studies have led to the conclusion that mod in Brassica encodes an AQP involved in water transfer from stigma to pollen. The sunflower aquaporins SunTIP17 and SunTIP20 are markedly upregulated in stomatal closure.

Conclusions

The aquaporin family is widespread; there are undoubtedly many more members that remain to be discovered. The structures of family members are, in general, highly conserved when compared with one another, both within species and between species. It seems that these are proteins of such fundamental importance that little can (or needs) to be changed, although it is possible that there are other water transporters waiting to be discovered.

Progress in understanding the structure and function of these proteins has been rapid; however, there is undoubtedly still much to learn, both at a basic mechanistic level and at the physiological level. One of the most intriguing findings to come out of the recent structural studies of GlpF is that the tetrameric complex has a potential ion channel down its centre; electron density near the periplasmic surface suggests coordination of two cations. The pathway is notable because many multimeric ion channels form their channels on their symmetry axis - for example, the tetrameric KcsA potassium channel. However, in GlpF it is unlikely to be constitutively accessible, and probably requires some form of regulation if it exists at all.

In any case, further studies are needed; but this field promises to be exciting for some time to come.