Molecular Characterization of a Broad Selectivity Neutral Solute Channel*

In all living cells, coordination of solute and water movement across cell membranes is of critical importance for osmotic balance. The current concept is that these processes are of distinct biophysical nature. Here we report the expression cloning of a liver cDNA encoding a unique promiscuous solute channel (AQP9) that confers high permeability for both solutes and water. AQP9 mediates passage of a wide variety of non-charged solutes including carbamides, polyols, purines, and pyrimidines in a phloretin- and mercury-sensitive manner, whereas amino acids, cyclic sugars, Na+, K+, Cl−, and deprotonated monocarboxylates are excluded. The properties of AQP9 define a new evolutionary branch of the major intrinsic protein family of aquaporin proteins and describe a previously unknown mechanism by which a large variety of solutes and water can pass through a single pore, enabling rapid cellular uptake or exit of metabolites with minimal osmotic perturbation.

* This work was supported in part by National Institutes of Health Grants DK46289 (to M. A. H.) and DK32753 (to W. B. G.). 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.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM  Iso-osmotic swelling assays were performed as described (11). Briefly, the swelling rate of the oocytes was measured by a video microscope every 10 s for 2 min, after transferring the oocytes from standard Barth's solution to modified Barth's solution in which 88 mM NaCl was replaced by solutes (adjusted to final osmolality of 200 Ϯ 20 mosm) or water. Osmolality was measured by freezing point depression.
Electrophysiological Studies-A two-microelectrode voltage clamp (Dagan Clampator-1B) was used to measure currents in water-injected oocytes and oocytes expressing AQP1 or AQP9. Microelectrodes (1-5 M⍀) were filled with 3 M KCl. Oocytes were superfused at 23 Ϯ 1°C in Barth's medium and clamped at a holding potential of Ϫ50 mV.
Step changes in membrane potential (between ϩ50 and Ϫ150 mV, in 20-mV increments) were applied for a duration of 100 ms. Current was filtered at 500 Hz, digitized at 5 kHz, and acquired using the Digidata1200 interface and pCLAMP6 software (Axon Instruments). The steady-state current-voltage relationships were determined with the help of software written by Donald D. F. Loo, UCLA School of Medicine. Functional expression of AQP9 or AQP1 in individual oocytes used for voltageclamp experiments was subsequently confirmed by radiotracer uptakes or osmotic swelling (lysis) measurements.

RESULTS AND DISCUSSION
Expression Cloning of AQP9 cDNA-We first determined 1 mM [ 14 C]urea uptake in Xenopus oocytes injected with rat liver mRNA. RNA-injected oocytes exhibited a 2-fold increase in urea uptake compared with water-injected controls (Fig. 1A).
When rats were fed a high protein diet, liver mRNA injection induced an 8-fold increase in urea uptake. Liver mRNA from rats fed a high protein diet was fractionated by preparative agarose gel electrophoresis (4), and fractions were analyzed by oocyte expression. A fraction of size 1.5-2.0 kb maximally stimulated urea uptake (Ϸ12-fold above background, data not shown). This fraction was used to construct a cDNA library from which we isolated a 1.5-kb cDNA that encodes a 295amino acid residue hydrophobic protein, AQP9. AQP9 has lowto-moderate homology with the mammalian aquaporins AQP3 (48% identity) (11,13,14), AQP7 (46% identity) (15), AQP1 (30% identity) (1)(2)(3), and with the bacterial glycerol facilitator GlpF (37% identity) (16) (Fig. 1, B and C). A homologue of AQP9 was recently reported from human leucocytes (17), but its unique properties have not yet been investigated. There is no significant homology with members of the urea transporter family. Kyte-Doolittle hydropathy analysis predicted six transmembrane domains separated by five connecting loops (Fig.  1D). AQP9 has two Asn-Pro-Ala (NPA) consensus motifs, characteristic of members of the major intrinsic protein (MIP) family (1-3) (Fig. 1B).
Tissue Localization of AQP9 -High stringency Northern ]urea in Xenopus oocytes injected with 50 ng of poly(A) ϩ RNA from rats fed normal and high protein (50% w/w) diets. B, amino acid sequence alignment of AQP9, rat AQP1 (GenBank TM P29975), rat AQP3 (GenBank TM L35108), rat AQP7 (GenBank TM AB000507), and Escherichia coli GlpF (GenBank TM M55990) was performed using the PILEUP program (Genetics Computer Group). Putative membrane-spanning regions are underlined and numbered 1-6. Conserved residues are indicated by shading. Asn-Pro-Ala (NPA) consensus motifs are highlighted by boxes. The potential sites for N-linked glycosylation (Asn-142) (double line), protein kinase C phosphorylation site (Ser-222), and casein kinase II phosphorylation site (Ser-271) (asterisks) are depicted. No consensus motifs of protein kinase A phosphorylation were found. C, phylogenic analysis of AQP9. Deduced amino acid sequences were analyzed using PILEUP. The percent amino acid identity between AQP9 and others is indicated. D, structural model of AQP9. Topology was determined according to the Kyte-Doolittle algorithm. The numbering of putative membrane-spanning regions corresponds to those of B.
analysis revealed a strong 1.5-kb transcript in liver ( Fig. 2A). Signals of 2.2 kb were detected in testis and brain, suggesting alternative splicing or use of different polyadenylation sites. No hybridization signals were detected in kidney, colon, heart, and skeletal muscle. Nonradioactive in situ hybridization in rat liver showed that AQP9 is expressed evenly among hepatocytes (Fig. 2B). In testis, AQP9 mRNA was detected in the inner surface of seminiferous tubules and in the interstitial Leydig cells. In seminiferous tubules, signals were obtained in spermatocytes at early developmental stages but not at later stages and not in Sertoli cells. AQP9 mRNA was detected in astrocytes throughout the brain (data not shown).
To investigate the effect of electric charge on solute permeability, we tested uptake of the monocarboxylates lactate and ␤-hydroxybutyrate, each of physiological importance in hepatocytes. Their P s values were significantly increased at physiological pH, but an additional 4-fold increase was observed at pH 5.5 (Fig. 3B), whereas mannitol uptake was independent of pH. In voltage-clamped oocytes expressing AQP9, monocarboxylates at concentrations up to 10 mM evoked no significant currents (less than 5 nA), regardless of pH o . Together with pH dependence of radiotracer uptakes, these results indicate that monocarboxylates permeate AQP9 only in their protonated form. The low permeabilities of the purine analogues xanthine and uric acid at pH 7.4 suggest that, likewise, these compounds permeate AQP9 in their protonated form (Fig. 3A).
We measured radiotracer uptake in oocytes expressing the aquaporins AQP1 and AQP3 to compare their solute permeabilities with those of AQP9 (Fig. 3C). AQP1 did not mediate significant uptake of any solute tested, indicating that AQP1 is a selective water channel (1-3). AQP3 facilitated glycerol uptake at levels similar to that for AQP9, but a weak urea permeability was detected in AQP3 (Ϸ25% of the P s value for AQP9). In addition, AQP3 was impermeable to the urea analogue thiourea, to polyols larger than glycerol, and to purines and pyrimidines. Our results suggest that the pore selectivity in the aquaporins decreases in the order AQP1 Ͼ AQP3 Ͼ AQP9. The osmotic permeability coefficient P f was increased Ϸ30-fold (to 0.014 Ϯ 0.0013 cm/s) in oocytes expressing AQP9 (cf. 0.0005 Ϯ 0.00002 cm/s in water-injected oocytes) (Fig. 3D). AQP9 showed a P f similar to that for AQP1 (0.016 Ϯ 0.0014 cm/s), but AQP3 exhibited a lower P f (0.0073 Ϯ 0.0008 cm/s). The Arrhenius activation energy for AQP9-mediated P f was 3.6 Ϯ 1.3 kcal/mol, again consistent with movement through a pore (1-3).
Notably, unlike other known aquaporins, water permeability was sensitive to 0.1 mM phloretin (86% inhibition). In AQP9, P f was reduced 61% in the presence of 0.3 mM HgCl 2 . A cysteine residue thought to be involved in mercurial binding lies 3 amino acid residues N-terminal to the second NPA box (1-3). Thus, AQP9 possesses general features of water channels in AQP9 mRNA was found in hepatocytes (arrows). Hybridization of sense AQP9 cRNA probe did not reveal any signal (data not shown). Lower panel, cross-section of testis. Asterisks indicate the lumen of the seminiferous tubules, containing mature spermatocytes with long tails. AQP9 mRNA was detected selectively in immature spermatocytes (large arrows) and also in the interstitial Leydig cells (small arrows). Bar ϭ 100 m .  FIG. 3. Functional properties of AQP9 expressed in Xenopus oocytes. A, solute permeability in oocytes expressing AQP9. Uptake of 1 mM 14 C-or 3 H-labeled carbamides, polyols, purines, pyrimidines, nucleosides, and monocarboxylates was measured over 90 s in oocytes injected with water (control) or AQP9 cRNA. Measurements were performed at 22°C except for lactate uptakes which were performed at 4°C to minimize the addition to the distinctive phloretin-inhibitable permeability for a large variety of non-charged solutes.
Ion Conductance of AQP9 -It has been generally considered that none of the aquaporins mediates any type of ion conductance under normal conditions (1)(2)(3). A recent report (19) suggested that forskolin induced a nonspecific cation conductance via AQP1, despite the absence of a typical consensus protein kinase A phosphorylation site. This phenomenon has been disputed by others (20) applying the same protocols. By using voltage-clamped oocytes expressing either AQP1 or AQP9, we determined the macroscopic conductance from the slope of the linear current-voltage relationship (between ϩ50 mV and Ϫ150 mV) before and after superfusion with 10 M forskolin in Barth's solution for 15 min (Fig. 3E). Only low ionic conductance was observed in the presence and absence of forskolin, in oocytes expressing AQP1 or AQP9, and in water-injected oocytes. The reversal potential was unchanged in all groups following forskolin treatment. Forskolin administered by direct microinjection into the oocytes was also ineffective (not shown). Therefore, neither AQP1 nor AQP9 mediates any ionic conductance under basal or forskolin-treated conditions.
Inhibition Profiles of AQP9 -A fundamental question is whether or not solutes and water share a single pore in AQP9. To address this issue, we first examined the inhibition profiles of solute permeabilities by phloretin and mercurial compounds. Phloretin (0.1 mM) and HgCl 2 (0.3 mM) effectively inhibited the water permeability as well as the permeabilities for mannitol, uracil, adenine, and ␤-hydroxybutyrate (75-90%) (Fig. 3F). Glycerol and urea permeabilities were inhibited 35-45% by 0.1 mM phloretin. Glycerol permeability was inhibited Ϸ50% by 0. contribution from endogenous lactate transport (29). 5-FU, 5-fluorouracil; ␤-HB, ␤-hydroxybutyrate. Data are mean Ϯ S.E. from 6 to 8 oocytes. B, pH-dependent monocarboxylate permeation. Uptake of 1 mM 14 C-labeled lactate and ␤-hydroxybutyrate was measured at physiological pH (Barth's solution pH 7.3-7.5) and at pH 5.5. The increase in permeability was obtained as the ratio of the uptake at pH 5.5 to the uptake at 7.3-7.5. C, solute permeability in oocytes expressing AQP1 and AQP3. Uptake of 1 mM radiolabeled solute was measured under identical experimental conditions as in A with oocytes expressing AQP1 or AQP3. D, the osmotic water permeability (P f ) was measured in oocytes expressing AQP1, AQP3, or AQP9. Where indicated, oocytes were preincubated in the presence of either 0.1 mM phloretin (10 min) or 0.3 mM HgCl 2 (5 min) and during uptake. The reversal of mercurial inhibition was studied by incubation with 5 mM ␤-mercaptoethanol for 15 min following 10 min of treatment with 0.3 mM HgCl 2 . Data represent the mean Ϯ S.E. from 6 to 8 oocytes. E, macroscopic ionic conductance in oocytes expressing AQP9 and AQP1 under initial conditions and after 15 min superfusing with 10 M forskolin. Data are mean Ϯ S.E. from 5 oocytes in each group. F, inhibition of water and solute permeability by phloretin and HgCl 2 . Percent inhibition of water and solute permeability was examined by incubation with 0.1 mM (or 0.5 mM) phloretin or 0.3 mM HgCl 2 , for 10 and 5 min, respectively. Asterisks indicate no significant difference from zero (t test). mM HgCl 2 , but urea permeability was not significantly affected by HgCl 2 . In oocytes expressing AQP3, water and urea permeabilities were partially inhibited by 0.1 mM phloretin and 0.3 mM HgCl 2 (15-20%), whereas the glycerol permeability was unchanged. The similar inhibition profiles of solute and water permeabilites in AQP9 indicate that solutes and water share a common transmembrane pathway.
Determination of Reflection Coefficients-This common pathway was further investigated by volumetric assays under isoosmotic conditions as well as by mathematical simulation of water and solute flows. Oocytes expressing AQP9 were transferred from standard Barth's solution into Barth's solution containing test solute (176 mosm). In AQP9 expressing oocytes, rapid swelling was observed for mannitol and glycerol (Fig. 4A), whereas in AQP3 expressing oocytes, rapid swelling was observed only for glycerol (Fig. 4B). When the test solute was replaced by water (hypo-osmotic solution), oocytes expressing AQP9 exhibited the same rapid swelling as for test solutes, suggesting that there is no discrimination between permeable solutes and water.
The Staverman's reflection coefficient () which is an indicator of the efficacy of a solute to induce osmosis was determined by simulation of solute and water movement (Fig. 4, A and B) using the Kedem and Katchalsky equations (Equations 1 and 2) for volume flow (J v ) and solute flux (J s ) (21,22),