Selectivity of the renal collecting duct water channel aquaporin-3.

Aquaporin-3 (AQP3) is a water channel found in the basolateral cell membrane of principal cells of the renal collecting tubule as well as in other epithelia. To examine the selectivity of AQP3, the permeability to water (Pf), urea (Pur), and glycerol (Pgly) of Xenopus oocytes injected with cRNA encoding AQP3 was measured. Oocytes injected with cRNA encoding either human or rat aquaporin-1 (AQP1) were used as controls. Although both aquaporins permit water flow across the cell membrane, only AQP3 was permeable to glycerol and urea (Pgly > Pur). The uptake of glycerol into oocytes expressing AQP3 was linear up to 165 mM. For AQP3 the Arrhenius energy of activation for Pf was 3 kcal/mol, whereas for Pgly and Pur it was >12 kcal/mol. The sulfhydryl reagent p-chloromercuriphenylsulfonate (1 mM) abolished Pf of AQP3, whereas it did not affect Pgly. In addition, phloretin (0.1 mM) inhibited Pf of AQP3 by 35%, whereas it did not alter Pgly or Pur. We conclude that water does not share the same pathway with glycerol or urea in AQP3 and that this aquaporin, therefore, forms a water-selective channel.

Aquaporin-3 (AQP3) 1 is a protein located in the kidney exclusively in the basolateral cell membrane of the principal cells of the collecting tubule, and it is also found in cells of other epithelia like the stomach and colon (1)(2)(3). AQP3 is a member of the MIP family of proteins. This family includes two groups of proteins: the aquaporins, which constitute cell membrane channels selective for water, and a group of homologs that do not transport water but that may serve as transporters for small solutes or have functions that are not yet well defined (4,5). AQP3 differs from other aquaporins in that it has features found in both of these groups because, in addition to being a water channel, it transports glycerol and to some extent urea (1,2). From studies of the uptake of water, glycerol, and urea into Xenopus oocytes expressing AQP3, it has been concluded that water and these solutes most likely share the same pore as they move across the protein (1). This view implies that the selectivity of the channel is based not only on the molecular size but also on other physical properties of the transported molecules. However, it could be hypothesized that water and glyc-erol do not share the same pore in AQP3. In support of this proposal it could be mentioned that, within the MIP family, AQP3 has the largest homology with the glycerol facilitator (GlpF) of Escherichia coli (1,2). GlpF transports glycerol and other small solutes across the inner membrane of the bacteria by a pore-type mechanism (6), but it excludes water (7). Thus, it is possible that some MIP homologs (the aquaporins) form one type of pore that is a water-selective channel, whereas other homologs form a pore that is permeable only to some small uncharged solutes. AQP3 could have components of each of these two groups and form in the cell membrane two different permeation pathways, one for water and another for glycerol and possibly other solutes. To test this hypothesis, we have examined in Xenopus oocytes injected with AQP3 cRNA the effects of inhibitors of the permeability to water and to some small solutes, on the assumption that if water and solutes share the same pore the inhibitors should affect the permeability to both. As controls we have used oocytes injected with cRNA encoding rat or human aquaporin-1 (AQP1, also called AQP-CHIP) because this protein forms a channel selective for water when incorporated in liposomes (8).

Expression of Water Channels in Xenopus
Oocytes-The rat cDNAs for AQP3 and AQP1 and human AQP1 cDNA were ligated into the expression vectors pSPORT 1 (2), pBluescript SK Ϫ (9), and the Xenopus expression construct pX␤G in pBluescript II KS Ϫ (10), respectively. Capped cRNAs encoding rat AQP3 and rat or human AQP1 were synthesized using the mCAP mRNA kit (Stratagene). Xenopus laevis oocytes were isolated, microinjected with either 50 nl of water or 50 nl of water containing 5 ng of aquaporin cRNA, and incubated as described (11) during 3-4 days.
Oocyte Permeability Measurements-The oocyte osmotic water permeability coefficient (P f ) was calculated from the rate of volume increase, measured with videomicroscopy every 10 s for 2 min, upon exposure to a hyposmotic Barth's solution (11). Glycerol or urea permeabilities (P gly and P ur , respectively) were calculated from the initial rate of uptake of these solutes into the oocytes. Two methods were used to measure the uptake. In the first one (volumetric method), we calculated the uptake from the rate of oocyte swelling, measured with videomicroscopy, upon exposure for 2 min to an isosmotic modified Barth's solution in which 165 mM of either glycerol or urea substituted for 88 mM NaCl. In the second method (isotopic method), we measured the 14 C uptake per oocyte during incubation for 2 min, or in a few cases 5 min, in Barth's solution containing either 7.5 Ci/ml of [ 14 C]glycerol and 1.05 mM glycerol or 16.6 Ci/ml of [ 14 C]urea and 1.3 mM urea. Subsequently, the oocytes were rapidly rinsed twice in ice-cold modified Barth's solution and each one transferred to a vial containing 0.2 ml of 0.1 g/ml SDS for liquid scintillation counting. Oocytes exposed to label-containing medium for only 2 s had negligible 14 C uptake, indicating that the washing procedure had adequately removed the extracellular contamination with incubation medium. No attempt was made to compare in the same batch of oocytes the permeability values obtained with both methods. In water-injected oocytes, the values were quite comparable with both methods. However, in the cRNA-injected oocytes, the permeability values varied somewhat from batch to batch, depending in part on the efficiency of expression of the protein encoded by the injected cRNA. In addition, the isotopic method gave generally lower values, probably due to partial loss of the labeled solute during the oocyte * This work was supported by National Institutes of Health Grant RO1 DK11489. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ A Fellow of the Latin American Program of The Pew Charitable Trust during part of this project.
washing procedure required to reduce extracellular contamination.
Oocytes were preincubated in Barth's solution containing either 1 mM sodium p-chloromercuriphenylsulfonate (pCMBS), 0.3 mM HgCl 2 , 0.1 mM phloretin or 0.2 mM DIDS for 30, 5, 15, or 5 min, respectively, to measure the effect of these inhibitors on permeabilities. The inhibitor was also included in the medium during the permeability measurements. The uptake of water or solutes was measured at 10, 20, and 30°C in water-or AQP3 cRNA-injected oocytes to estimate the Arrhenius energy of activation (E a ) of the permeabilities (11).

RESULTS
Oocytes injected with cRNA encoding AQP3 had a 10-fold larger P f than that of water-injected oocytes (Fig. 1A), confirming results previously reported (1, 2). As expected, oocytes injected with either rat or human AQP1 cRNA also showed a larger P f than controls (Fig. 1A). In contrast, only the oocytes expressing AQP3 showed increased P gly and P ur (measured with the volumetric method) relative to the values found in water-injected oocytes (Fig. 1B). Therefore, AQP3 is permeable to water, glycerol, and slightly to urea, although these solute permeabilities are approximately 3 orders of magnitude lower than that to water.
The sulfhydryl group reagent pCMBS blocks the P f rendered to oocytes by the expression of AQP3 (2) without altering the cell membrane conductance to the major ions of intra-or extracellular fluids (11). We tested, therefore, whether exposure to 1 mM pCMBS for 30 min alters the permeability of AQP3 to water and glycerol. No change in P f of water-injected oocytes was found, whereas AQP3 water permeability was fully inhibited ( Fig. 2A). In contrast, pCMBS practically abolished P gly in water-injected oocytes, whereas it had only a small inhibitory effect in oocytes expressing AQP3 (Fig. 2B). Identical results were found after incubation of the oocytes in 0.3 mM HgCl 2 for 5 min (data not shown). These results indicate that, first, the oocytes possess an endogenous transport path for glycerol that is inhibited by the mercurial compounds. Second, the pCMBSinduced decrement in P gly observed in the oocytes expressing AQP3, about 3 ϫ 10 Ϫ6 cm/s, was of the same magnitude as that measured in water-injected oocytes, and thus the mercurial compound inhibited only the endogenous transporter and did not affect P gly imparted by the expression of AQP3. Third, the endogenous glycerol transporter is not AQP3 but a different protein. Therefore, the different effects of pCMBS on P f and P gly strongly suggest that water and glycerol molecules move through AQP3 by separate pathways.
It has been reported (1) that phloretin blocks the transport of water and urea across AQP3. We also studied the effect of exposing the oocytes to 100 M phloretin for 15 min. Phloretin did not alter P f or P gly of water-injected oocytes, whereas in oocytes that expressed AQP3 it inhibited P f by 35% while P gly was unchanged (Fig. 3, A and B). These results, therefore, are also consistent with the view that water and glycerol move through AQP3 by separate pathways.
Phloretin caused a small increase in P ur in water-injected oocytes and had no significant effect on P ur in AQP3 cRNAinjected oocytes (Fig. 3C). These results might indicate that phloretin partially inhibits P ur of AQP3. However, the unexpected increase in P ur without changes in P gly or P f in waterinjected oocytes produced by phloretin is similar to the increase in oocyte basal P ur elicited by HgCl 2 (1). This indicates that the endogenous urea transport is unspecifically altered by these compounds and, therefore, given the small magnitude of the AQP3-mediated P ur , it is difficult to assess whether phloretin inhibits P ur in oocytes injected with AQP3 cRNA.
Glycerol intrinsic protein (GLIP), a cloned MIP family member with a deduced amino acid sequence very similar to that of AQP3, is permeable to glycerol but not to urea or water (12). GLIP-mediated glycerol transport was inhibited by 88% by 0.2 mM DIDS (12). Hence, with the aim of finding an inhibitor of solute transport via AQP3, we examined the effect of exposure to 0.2 mM DIDS on oocytes injected with AQP3 cRNA. P gly was measured isotopically in groups of 10 oocytes. Control P gly in water-injected and in AQP3 cRNA-injected oocytes was 2.4 (Ϯ0.3) ϫ 10 Ϫ6 and 8.1 (Ϯ0.3) ϫ 10 Ϫ6 cm/s, respectively, and the corresponding values for the DIDS-treated oocytes were 2.6 (Ϯ0.5) ϫ 10 Ϫ6 and 8.5 (Ϯ0.8) ϫ 10 Ϫ6 cm/s. Thus, no change in P gly of AQP3 was observed with DIDS.
To better characterize the mode of transport of urea and glycerol by AQP3, we measured the values of E a for P f , P gly , and P ur . E a for P f was 3 kcal/mol, a low value, therefore consistent with AQP3 being a water channel. For P ur and P gly the E a values were much higher, 19.7 and 12.2 kcal/mol, respectively, and thus, indicative of a transport mechanism that involves a strong interaction between AQP3 and urea or glycerol.
To examine whether glycerol transport by AQP3 expressed in oocytes is saturable, we measured uptake as a function of glycerol concentration. Barth's solutions containing various glycerol concentrations were prepared by replacing NaCl with glycerol maintaining osmolality constant. A higher than 165 mM glycerol concentration was not used because it would have exposed the oocytes to a hyperosmotic medium, introducing conditions that were not comparable to the rest of the measurements. As shown in Fig. 4, glycerol uptake was linear in AQP3 cRNA-and in water-injected oocytes. Thus, in the concentration range examined, AQP3-mediated glycerol transport was not saturable. DISCUSSION The results of this study confirm reports of others that human AQP1 (13) and rat AQP1 (12) expressed in oocytes are not permeable to glycerol. In addition, rat AQP1 was not permeable to urea, and therefore, it constitutes a water-selective channel (8,15). Among the known aquaporins, ␥-tonoplast intrinsic protein (␥-TIP) from Arabidopsis thaliana (13), AQP4 (12), and AqpZ from E. coli (14) have not been found to promote increased P gly when they were expressed in the oocyte. Thus, in the MIP family, only GlpF, AQP3, and GLIP (if this homolog is indeed different from AQP3) are permeated by glycerol.
The present results and those of others (7,(12)(13)(14) show that the cell membrane of control Xenopus oocytes is slightly permeable to glycerol. Not previously reported is the observation that this permeability is practically completely inhibited by pCMBS (Fig. 2B). This suggests that glycerol crosses the oocyte membrane by facilitated diffusion via a transporter that may resemble that of E. coli because GlpF is inhibited by sulfhydrylmodifying reagents as well (6, 7). However, since pCMBS does not affect the transport of glycerol mediated by AQP3 (Fig. 2B), the endogenous transport system differs from AQP3. To examine whether glycerol shares with water the same pathway through AQP3, we studied the effect of inhibitors on the permeability to water and to this solute. The hypothesis examined in these experiments was based on a simple model of a water channel: a constantly open cylindrical pore through which water and perhaps solutes, selected by their molecular size, cross the cell membrane. Based on this model we would expect that an inhibitor that blocks the passage of water will also affect the permeability of a solute that shares the same pathway. Both pCMBS and phloretin inhibited P f , but they had no effect on the P gly attributable to the expression of AQP3. In addition, phloretin did not appear to affect P ur of AQP3. Therefore, water does not share the same pathway with glycerol or urea.
The results of the measurements of E a for the permeabilities to water and solutes can also help to understand the pathways used by these substances to cross the cell membrane through AQP3. The low value of E a for P f (3.0 kcal/mol) indicates that water moves through a water-filled pore. In contrast, the values of E a for urea (19.7 kcal/mol) and glycerol (12.2 kcal/mol) permeabilities are most consistent with the thesis that urea and glycerol do not cross by diffusion through an aqueous pore. This view differs from that of Ishibashi et al. (1) who proposed that water, urea, and glycerol share the aqueous pore of AQP3.
Comparison of the amino acid sequence of AQP3 to that of rat AQP1 and of E. coli GlpF shows differences that may help locate segments potentially involved in the formation of the glycerol transport site in AQP3. An alignment of AQP1, AQP3, and GlpF, based on that proposed by Reizer et al. (4) for the entire MIP family before the cloning of AQP3, is shown in Fig.  5. White letters in a black background highlight those residues that are most highly conserved in the family (4). Identity of AQP3 residues with those of either AQP1 or GlpF is indicated by rectangles. The larger degree of homology of AQP3 with GlpF than with AQP1 is shown. AQP3 and GlpF contain two rather homologous segments, corresponding to the residues 130 -150 and 227-239 of AQP3, which in Fig. 5 are underlined as segments I and II, respectively. Segment II, in particular, is largely identical (62%) with the corresponding one in GlpF. These segments have no homology in AQP1, AQP2, AQP4, or AQP5. Both segments are mostly polar and are probably located in extracellular loops of AQP3. It is, therefore, possible that one or both of these segments contribute to the formation of the glycerol transport pathway. Since at present it is unknown even whether AQP3 exists in the cell membrane as a homotetramer like AQP1 (16) or in some other configuration, a more specific mode of transport for glycerol cannot be proposed at this time. Certainly the absence of saturation of the rate of glycerol transport in the concentration range studied (1-165 mM) is more consistent with passage through a pore than via a transporter. However, the exact nature of the transport system remains to be elucidated.