Water and Glycerol Permeabilities of Aquaporins 1–5 and MIP Determined Quantitatively by Expression of Epitope-tagged Constructs inXenopus Oocytes*

The goal of this study was to compare single channel water and glycerol permeabilities of mammalian aquaporins (AQP) 1–5 and the major intrinsic protein of lens fiber (MIP). Each of the six cloned cDNAs from rat was left untagged or was epitope-tagged with c-Myc or FLAG at either the N or C terminus so that results would not depend on epitope identity or location. The constructs were expressed in Xenopus oocytes for measurement of osmotic water permeability (P f ), [3H]glycerol uptake, and protein expression. Each of the 30 epitope-tagged constructs was expressed strongly at the oocyte plasma membrane. The 10-min uptake of [3H]glycerol was increased significantly (range of 4.5–8-fold over control) in oocytes expressing untagged AQP3 (GLIP) and each of the four tagged AQP3 constructs; [3H]glycerol uptake was not increased in oocytes expressing AQP1, AQP2, AQP4, AQP5, or MIP. In oocytes microinjected with 5 ng of cRNA, average P f values (in cm/s × 10−3) were 0.67 ± 0.06 (control), 19 ± 2 (AQP1), 10 ± 1 (AQP2), 8 ± 2 (AQP3), 29 ± 1 (AQP4), 10 ± 1 (AQP5), and 1.3 ± 0.2 (MIP), and they were relatively insensitive to the presence, identity, or location of the epitope tag. P f values were not affected by protein kinase A or C activation. After normalization for plasma membrane expression by immunoprecipitation of microdissected plasma membranes, single channel water permeabilities (p f , referenced to the AQP1 p f of 6 × 10−14cm3/s) were (in cm3/s × 10−14) 3.3 ± 0.2 (AQP2), 2.1 ± 0.3 (AQP3), 24 ± 0.6 (AQP4), 5.0 ± 0.4 (AQP5), and 0.25 ± 0.05 (MIP); p f values were insensitive to epitope identity and location. These results indicate very different intrinsic water permeabilities for the mammalian aquaporin homologs, with thep f value for AQP4 remarkably higher than those for the others. The p f values establish limits on aquaporin tissue densities required for physiological function and suggest significant structural and functional differences among the aquaporins.

Five proteins with homology to the major intrinsic protein of lens fiber (MIP) (1) have been cloned in mammals and given the name aquaporins. AQP1 1 (original name CHIP28) (2) is found in erythrocytes, kidney proximal tubule, and various epithelia and endothelia. AQP2 (original name WCH-CD) (3) is the vasopressin-inducible water channel expressed in kidney collecting duct. AQP3 (alternate name GLIP (glycerol-transporting intrinsic protein)) was cloned by three laboratories (4 -6) and is expressed at the basolateral membrane of kidney collecting duct and in multiple epithelia (7). AQP4 (original name MIWC (mercurial-insensitive water channel)) (8) is expressed strongly in brain and colocalizes with AQP3 in several tissues (7). AQP5 (9) is expressed in lung, salivary gland, and eye. Amino acid sequences for various proteins share several conserved sequence motifs with overall identities of 25-60% after sequence alignment. Subsequent to the initial cloning reports cited above, various isoforms of AQP1-5 were identified in different organs and/or species (see Refs. 10 -13 for review). In addition, an aquaporin (AQP2L, subsequently renamed AQP6) (14) with 52% amino acid identity to human AQP2 was cloned that was expressed only in human kidney, and several homologous water-transporting proteins have been cloned from plants (15) and amphibia (16,17).
Functional studies have been carried out on AQP1-5 and MIP by expression of cRNAs in Xenopus oocytes. Several studies indicate that expression of AQP1, AQP2, AQP4, and AQP5 significantly increases oocyte water permeability (2-6, 8, 9, 18 -20), although permeability values varied widely in different laboratories. Transfection of AQP1 and AQP4 in cultured mammalian cells also confers increased plasma membrane water permeability (21,22), and transfection of AQP2 confers vasopressin-regulated water permeability (23). However, there are conflicting reports for expression of AQP3: two groups report increased oocyte water permeability (4,6), whereas our laboratory found little increase in water permeability for AQP3 expression in oocytes (5). For MIP, studies in lens membrane vesicles and proteoliposomes (24) and in Xenopus oocytes (25,26) suggest that MIP has some intrinsic water permeability, but probably much less than the aquaporins. In the case of AQP1, where purification to homogeneity has been possible, measurement of "per channel" or "single channel" water permeability (p f ) indicates that each AQP1 monomer has a relatively low p f of ϳ6 ϫ 10 Ϫ14 cm 3 /s (27,28). This relatively low p f requires the presence of very high densities of AQP1 water channels in membranes (Ͼ10 3 /m 2 , compared with generally Ͻ1 ion channel/m 2 ) to confer significantly increased water permeability. No information is available about p f for the other aquaporins because purification and reconstitution have not been accomplished. Most studies of aquaporin function have concluded that AQP1, AQP2, AQP4, and AQP5 are selective for transport of water without passage of ions, protons, urea, and glycerol, whereas Abrami et al. (29) reported that AQP1 may transport small polar non-electrolytes in addition to water. It is difficult to compare these investigations because the levels of protein expression at the plasma membrane have not been quantified.
The principal goal of this study was to determine quantitatively the single channel water and glycerol permeabilities of mammalian AQP1-5 and MIP. A second goal was to examine whether activators of protein kinases A and C regulate the water permeabilities of these proteins, a controversial issue that has received recent attention (see "Discussion"). Our strategy was to utilize epitope tagging to detect oocyte plasma membrane expression of each protein with similar efficiencies. Single channel water permeabilities were determined from ratios of measured oocyte permeabilities to plasma membrane expression levels. The principal findings were that significant glycerol permeability was found only for AQP3 and that single channel water permeabilities differed widely for the aquaporins. Phosphorylation by protein kinase A or C did not affect MIP-or aquaporin-mediated water permeability. The considerable heterogeneity in water channel function of the aquaporin family proteins was an unexpected finding with interesting implications for water channel structure and physiology.
RNA Transcription-Complementary RNA was transcribed in vitro using SP6 polymerase (Life Technologies, Inc.) and 5 g of plasmid DNA in a 100-l volume at 37°C for 1 h in the presence of diguanosine triphosphate (1 A 250 unit; Pharmacia Biotech Inc.). Plasmid DNA was digested with RNase-free DNase (Invitrogen), extracted with phenol/ chloroform, and precipitated twice in ethanol. The cRNA was suspended in distilled water for oocyte injection.
Cell-free Translation-In vitro transcribed cRNA was added to a rabbit reticulocyte lysate mixture containing [ 35 S]methionine (32). Microsomes prepared from dog pancreas were added in some experiments at the start of translation to a final concentration of 8 A 280 units. Translation was performed at 24°C for 1 h. Samples were analyzed by SDS-polyacrylamide gel electrophoresis, EN 3 HANCE fluorography, and autoradiography. Oocyte Expression and Transport Assays-Stage V and VI oocytes from Xenopus laevis were isolated and defolliculated with collagenase (type IA, Sigma; 1 mg/ml for 2-4 h at 20°C) in Barth's buffer (200 mosM). Oocytes were microinjected with 50-nl samples of cRNA (0 -200 ng/l) and incubated at 18°C for 24 -27 h. Osmotic water permeability (P f ) was measured from the time course of oocyte swelling at 10°C in response to a 1.5-or 5-fold dilution of the extracellular Barth's buffer with distilled water (33). Oocyte P f was calculated from the initial rate of swelling (d( , V w ϭ 18 cm 3 /mol, and osM out Ϫ osM in ϭ 190 mosM. In some experiments, oocytes were incubated with forskolin or phorbol 12-myristate 13-acetate for 15-30 min at room temperature prior to water permeability measurements. Uptake of [ 3 H] glycerol (200 mCi/mmol) was carried out as described previously (5). Groups of 10 oocytes were incubated in 200 l of Barth's buffer containing 20 Ci of the radiolabeled glycerol (nonradioactive glycerol was added to give a 1 mM final concentration) at room temperature. After 0 or 10 min, oocytes were washed rapidly three times in ice-cold Barth's buffer, and individual oocytes were dissolved in 5% SDS for scintillation counting. Results are reported as percentage uptake assuming an oocyte aqueous volume of 0.25 l, after subtraction of the 0 time contribution from surface binding (generally Ͻ5%).
Immunoprecipitation-Metabolic labeling was performed by incubating groups of 10 oocytes for 24 h at 18°C in 100 l of Barth's buffer containing 50 Ci of [ 35 S]methionine (Amersham Corp.). Oocytes were gently washed three times in ice-cold Barth's buffer, and plasma membrane complexes were peeled from the surface of the oocytes with fine forceps and collected. Membranes were disrupted in 1 ml of homogenization buffer (7.5 mM Na 2 HPO 4 , pH 7.4, 1 mM EDTA, 20 g/ml phenylmethylsulfonyl fluoride, 1 g/ml pepstatin A, and 1 g/ml leupeptin) by vigorous vortexing and repeated pipetting. Cellular debris was pelleted at 750 ϫ g for 5 min, and the membranes were then pelleted at 16,000 ϫ g for 30 min. Membrane pellets were solubilized in 1 ml of phosphate-buffered saline, pH 7.4, containing 100 mM ␤-octyl glucoside for 1 h and incubated with protein A-Sepharose CL-4B beads (Pharmacia Biotech Inc.) for 1 h at 4°C with continuous rocking. The beads were pelleted at 10,000 ϫ g. The supernatant was incubated with primary antibody (1:300) for 1 h and then with a rabbit anti-mouse secondary antibody for 1 h. Protein A-Sepharose CL-4B beads were added, pelleted, and washed four times with phosphate-buffered saline, pH 7.4, containing 0.1% SDS, 1% deoxycholate, and 1% Triton X-100. The proteins were released from the beads in 30 l of SDS loading buffer and resolved on 12% SDS-polyacrylamide gel. Gels were soaked in 15% (w/v) 2,5-diphenyloxazole in Me 2 SO for 30 min, dried, and exposed to Hyperfilm (Amersham Corp.) at Ϫ70°C for 3-10 days.

RESULTS
The cDNAs encoding rat AQP1-5 and MIP were epitopetagged at the N or C terminus with the c-Myc or FLAG sequence, and the untagged and tagged constructs were subcloned into an oocyte expression vector. The constructs were confirmed by sequence analysis and characterized initially by cell-free translation. Fig. 1 shows representative cell-free translation results for the C-terminal c-Myc (A) and FLAG (B) epitope-tagged constructs. Each construct was translated efficiently. Protein sizes were as expected from prior translation and immunoblot studies after correction for the small c-Myc (1.2 kDa) or FLAG (1.0 kDa) epitope. As found previously, AQP1-3 and AQP5 were partially glycosylated, whereas AQP4 and MIP were not. Similar cell-free translation profiles were obtained for the N-terminal tagged constructs and after immunoprecipitation of each of the 24 tagged cDNAs with c-Myc or FLAG antibody (data not shown).
Functional measurements were made in Xenopus oocytes after microinjection of 5 ng of each cRNA and a 24 -27-h incubation at 18°C. All 24 epitope-tagged constructs and the six untagged constructs were tested. Fig. 2A shows representative measurements of osmotic water permeability by oocyte swelling, and Fig. 2B summarizes averaged results. For each of the constructs tested, a consistent rank order of water permeabilities was found, with P f for AQP4 Ͼ AQP1 Ͼ AQP5 Ͼ AQP2 ϳ AQP3 Ͼ Ͼ MIP. The water permeability of MIP-expressing oocytes was small, but was significantly greater than that of control (water-injected) oocytes. The water permeability values were relatively insensitive to the presence, location, and identity of the epitope tag (see below). For an equal quantity of injected cRNA, AQP4-expressing oocytes had the greatest water permeability; however, these results cannot be interpreted in terms of intrinsic water permeability per aquaporin monomer because of possible differences in expression among the various proteins. Fig. 3 shows a summary of 10-min glycerol uptake measurements in oocytes from the same groups used for the water transport measurements. Glycerol uptake was remarkably increased compared with control in oocytes expressing untagged AQP3 as well as each of the four tagged AQP3 constructs. Glycerol uptake was not increased significantly above control in oocytes expressing AQP1, AQP2, AQP4, AQP5, or MIP.
Experiments were next carried out to determine the protein expression levels of AQP1-5 and MIP in Xenopus oocytes microinjected with 5 ng of each cRNA. Several complementary approaches were used, including immunofluorescence, immunoprecipitation, and immunoblot analysis. Fig. 4 shows repre-sentative immunofluorescence data for oocytes expressing the C-terminal c-Myc epitope. Immunofluorescence in oocytes expressing each of the tagged proteins was clearly above control, with qualitatively the greatest expression of AQP1 and the least expression of AQP4. However, the staining efficiency and specificity of the monoclonal antibodies were inferior to polyclonal AQP1 and AQP4 antibodies used previously for measurement of plasma membrane aquaporin expression (34) and were not suitable for quantitative analysis.
Quantitative analysis of aquaporin plasma membrane expression was accomplished by inclusion of [ 35 S]methionine in the incubation solution followed by oocyte plasma membrane microdissection and immunoprecipitation. Fig. 5 shows autoradiograms of immunoprecipitated proteins labeled with each of the epitopes (c-Myc or FLAG) at their C or N termini. The relative expression levels of AQP1-5 and MIP were in qualitative agreement with the data in Fig. 4, with strong expression of AQP1 and the least expression of AQP4. Below each lane is given the averaged P f values measured in the same oocytes that were blotted. These values are needed to compute single channel water permeabilities (see below).
Control measurements were carried out to confirm the suitability and linearity of the immunoprecipitation method. Oocytes were microinjected with different amounts (0 -10 ng) of cRNA encoding AQP1 tagged at its N terminus with the c-Myc epitope. P f was measured by the swelling assay, and plasma membrane expression was measured by densitometry of autoradiograms of immunoprecipitated protein. The AQP1 protein was immunoprecipitated with AQP1 antibody or c-Myc antibody. Fig. 6A shows a representative autoradiogram of the immunoprecipitated proteins with different amounts of microinjected cRNA. Fig. 6B shows a comparison of oocyte P p versus the amount of immunoprecipitated protein (by densitometry of autoradiograms). There was a nearly linear relationship between plasma membrane water permeability and relative amount of 35 S-labeled AQP1 in immunoprecipitations performed with the AQP1 and c-Myc antibodies. The y intercept of the line was just above zero because of the non-zero base-line water permeability of control oocytes.
The strategy to determine single channel water permeabilities (p f , in units of cm 3 /s) involves taking ratios of oocyte P f to integrated band intensity and normalizing the ratios to that for untagged AQP1, which has a known p f(AQP1) of 6 ϫ 10 Ϫ14 cm 3 /s (27), tag y is the single channel water permeability (cm 3 /s) of AQPx with tag y (y is N-or C-terminal c-Myc or FLAG), P f(AQPx) tag y is the water permeability (cm/s) of oocytes expressing AQPx with tag y, and IB AQPx tag y (Ab z) is the integrated band intensity (arbitrary units) of immunoprecipitated proteins from oocytes expressing AQPx with tag y using antibody z. The normalization is done in two steps. The first bracketed term in Equation 1 compares water permeabilities of untagged AQP1 to those of each of the four tagged AQP1 constructs. (AQP1 antibody is used because untagged AQP1 cannot be immunoprecipitated with c-Myc or FLAG antibody.) The second bracketed term compares water permeabilities of AQPx with tag y to that of AQP1 with tag y. (Immunoprecipitations were done with antibody y). Fig. 7A shows an autoradiogram of the AQP1 proteins immunoprecipitated with AQP1 antibody along with corresponding measured P f values. Fig. 7B summarizes the deduced p f values referenced to that for untagged AQP1. The values did not differ significantly among the AQP1 constructs; the first bracketed term in Equation 1 will thus be taken as unity. Fig. 8A summarizes the single channel water permeabilities of each of the four tagged constructs (N-or C-terminal c-Myc or FLAG) of AQP1-5 and MIP. The p f values were computed using Equation 1, in which the second bracketed term was deduced from experiments as in Fig. 5. The results show a considerable variation in p f values among the aquaporins (see "Discussion").
Notably, p f for each of the proteins was relatively insensitive to the identity and location of the epitope tag.
To investigate whether the intrinsic water permeabilities of AQP1-5 and MIP are regulated by phosphorylation, effects of activators of protein kinases A and C were studied (Fig. 8).
Utilizing activators and incubation conditions that have been established to effectively phosphorylate proteins in oocytes (35), the protein kinase activators did not affect P f for any of the proteins. Incubation of oocytes with the cell-permeable protein kinase A inhibitor (R p )-cAMP-S (0.1 mM, 15 min) also did not affect P f values (data not shown). These results indicate that the single channel water permeabilities reported in Fig. 8A are insensitive to the action of the major cellular protein kinases. DISCUSSION The goal of this study was to compare single channel water and glycerol permeabilities of the six principal mammalian aquaporin homologs. Because purified functional aquaporins have not been available except for AQP1, our strategy was to engineer epitope tags so that expression of the various aquaporins could be quantified with approximately equal efficiencies. Two different tags were used at each of two different locations, as well as the untagged molecule, to ensure that the results would be independent of tag identity and location. It was found that the epitope tags had little effect on aquaporin expression and function in Xenopus oocytes. After correction of permeability data for plasma membrane expression, it was found that AQP4 had remarkably higher intrinsic water permeability than AQP1 and AQP5, with lower water permeabilities for AQP2 and AQP3. MIP had little intrinsic water permeability. Only AQP3 expression conferred increased glycerol permeability. These findings have important implications for aquaporin structure and physiology as discussed below.
There was a very small but measurable single channel water permeability for rat MIP of 2.5 ϫ 10 Ϫ15 cm 3 /s, in agreement  4. Immunofluorescence of oocytes expressing each of the constructs tagged with c-Myc at the C terminus. Oocytes were imaged using a Leitz epifluorescence microscope (10 ϫ objective) and printed under identical conditions as described previously (34). Scale bar ϭ 50 m.
with the small (Ͻ4 ϫ 10 Ϫ15 cm 3 /s) water permeability for bovine lens MIP in reconstituted proteoliposomes measured by stopped-flow light scattering (24). The relative water permeability for MIP versus AQP1 of 0.042 found here is comparable to that of 0.023 for bovine MIP, in which MIP density was estimated by electron microscopy of oocyte membranes (26). If the MIP p f of 2.5 ϫ 10 Ϫ15 cm 3 /s measured here is the same as that in the lens cell membrane in vivo, then a significant increase in membrane water permeability (P f ϭ 0.01 cm/s) would require a density of ϳ10,000 MIP monomers/m 2 of membrane. Assuming that MIP forms tetramers in membranes, this density is equivalent to a 10-nm spacing between MIP particles. Electron microscopy of lens shows an ϳ6-nm spacing between MIP particles. Thus, if MIP-mediated water permeability is important in lens, the high MIP density may be necessary because of the low intrinsic MIP water permeability. It should be noted, however, that the non-zero intrinsic water permeability of MIP in Xenopus oocytes may be related to its role as a transporter of other substances; similar small increases in water permeability have been found for oocytes expressing cRNAs encoding the sodium-independent (36) and -dependent (37) glucose transporters and the CFTR chloride channel (35).
AQP3 was found to have significant water and glycerol per-meabilities in this study. In our initial report on AQP3 cloning (5), we found high glycerol permeability, but little water permeability, similar to the finding for the homologous protein GlpF (38), the bacterial glycerol facilitator protein. The story about the water and substrate specificities of AQP3 is probably incomplete and is under investigation. Recent measurements of AQP3-mediated water and glycerol permeabilities using inhibitors have suggested distinct water and solute pathways (39). In our laboratory, closely related AQP3 cDNAs have recently been isolated from cDNA libraries and by polymerase chain reaction; all of the cDNAs have high glycerol permeability, whereas a subfraction of cDNAs (used in this study) give substantial water permeability. (For comparison, the "low P f " AQP3 cDNAs have the same intrinsic glycerol permeability as the cDNA used here, but have a low intrinsic p f of (0.037 Ϯ 0.022) ϫ 10 Ϫ14 cm 3 /s. Studies are needed to measure AQP3mediated water versus glycerol permeability in native tissues. We note that AQP3 is coexpressed with the efficient water transporter AQP4 in various fluid-transporting tissues in kidney, airways, and intestine (7) and by itself in tissues that are not involved in fluid transport, including conjunctiva, urinary bladder, and epidermis (40). Transgenic knockout mice should be useful to resolve the function and physiological significance of AQP3, as was done recently for AQP4 (41). The p f of 3.3 ϫ 10 Ϫ14 cm 3 /s determined for AQP2 provides estimates of the density of AQP2 protein at the apical membrane of the vasopressin-stimulated collecting duct and the number of AQP2-containing vesicles that fuse with the apical membrane. Using a P f of ϳ0.02 cm/s for the collecting duct apical membrane after vasopressin stimulation (42) and the single channel p f found here, a predicted density of ϳ6000 AQP2 monomers/m 2 of apical plasma membrane is computed. Kishore et al. (43) have estimated an AQP2 density of 4.3 ϫ 10 10 molecules/cm 2 of cortical collecting duct in dehydrated rats. If every AQP2 molecule is at the apical plasma membrane and functional, then apical membrane P f would be ϳ0.1 cm/s. In AQP2-containing endosomes from collecting duct of ϳ150-nm diameter, a P f of 0.03 cm/s has been measured (44). If all AQP2 molecules are active as water transporters, then it is estimated that there are ϳ650 AQP2 monomers/endosome and that ϳ1000 endosomes fuse with the apical plasma membrane of each principal cell upon vasopressin stimulation.
The considerable heterogeneity in the single channel water permeabilities of the aquaporins was an unexpected finding. Although it is likely that structural differences account for the heterogeneity, the possibility of different "open probabilities" cannot be discounted. It is difficult to test whether water channel gating occurs because there is no electrical signature for the movement of water. The most information about structure is available for AQP1. AQP1 monomers assemble into tetramers in membranes (45) in which each monomer functions independently as a water channel (34). Structure analysis by electron crystallography suggests that each monomer of human AQP1 consists of six membrane-spanning helical domains arranged around a central cavity (46,47); however, resolution is inadequate to define details about the water transport pathway. Although the other aquaporins probably also contain six membrane-spanning domains, no information is available about their assembly in membranes or the structure of their putative water pores. High resolution x-ray diffraction of three-dimensional crystals will be required to establish whether the heterogeneity in single channel water permeability of the aquaporins is due to differences in water pore length, diameter, and/or shape. There may be additional factors that influence aquaporin water permeability, such as enhanced osmotic flow for AQP4 resulting from the assembly of AQP4 into large, closely packed orthogonal arrays (22) and the physical state of membrane lipids. Finally, the data available do not rule out the possibility that in some aquaporins, the potential space between or around aquaporin monomers might permit water flow.
It has been thought that AQP1, AQP2, AQP4, and AQP5 function as water-selective channels that exclude the small solutes urea and glycerol as well as monovalent ions and protons (10 -12). It was argued that significant permeability of aquaporins to solutes other than water would be deleterious to cell function because of the high density of aquaporins required for increased water permeability (18). For example, a high proton permeability for AQP2 would produce profound intracellular acidification of principal cells of kidney collecting duct because the AQP2-containing luminal membrane is bathed in FIG. 7. Determination of single channel water permeabilities (p f ) for the untagged and tagged AQP1 proteins. A, tagged and untagged AQP1 proteins immunoprecipitated by AQP1 antibody. Oocytes were microinjected with water (control) or 5 ng of cRNA encoding each construct and incubated for 36 h. Measured oocyte water permeabilities (P f ) are given for each condition. B, computed (single channel) p f values referenced to the p f for untagged AQP1 of 6 ϫ 10 Ϫ14 cm 3 /s. FIG. 8. Single channel water permeabilities (p f for epitopetagged AQP1-5 and MIP. A, single channel water permeabilities (mean Ϯ S.E., three sets of measurements) were computed from ratios of measured oocyte water permeabilities (P f ) to integrated band intensities of immunoprecipitated proteins per Equation 1 under "Results." B, shown are the effects of protein kinase activators on the water permeabilities of AQP1-5 and MIP. Oocytes were microinjected with cRNAs (0.5 ng) encoding each of the untagged proteins. Where indicated, oocytes were incubated in 10 M forskolin or 100 nM phorbol 12-myristate 13-acetate (PMA) for 15-30 min at room temperature prior to water permeability assay. P f values (mean Ϯ S.E., n ϭ 8 -12, two separate batches of oocytes) are shown for swelling measurements done at 10°C. acidic urine. Abrami et al. (29) reported that AQP1 has significant permeability to glycerol and several small polar nonelectrolytes. Their results were based primarily on measurements of volume changes in oocytes expressing AQP1 in response to solute gradients. The experiments here do not show measurable glycerol transport by oocytes expressing AQP1, AQP2, AQP4, AQP5, and MIP. The reason for the substantial AQP1-mediated solute permeability found by Abrami et al. (29) is unclear. One possibility is that under certain conditions, possibly in the presence of osmotic gradients in oocytes, solutes may pass through or around AQP1 monomers. For example, mercurial inhibition of AQP1 was associated with increased urea transport via a pathway probably distinct from the AQP1 water pore (24). Further work is needed to identify conditions in which various aquaporins might transport substances other than water.
Water channel phosphorylation in oocytes by protein kinases A and C did not affect water permeability. Recently, it was reported that exposure of oocytes expressing AQP1 to forskolin was associated with increased water permeability and the appearance of an ion conductance (48). These results have been controversial and have not been repeatable by several laboratories. Under conditions in which ion channel phosphorylation in oocytes is strongly stimulated, we did not find an effect of protein kinase A activation on P f in oocytes expressing AQP1, and in related studies, we did not find an effect of phosphorylation agonists on AQP1-mediated water permeability and cation leak in human erythrocyte membranes (49). It is noted that the AQP1 sequence does not contain a classical consensus site for phosphorylation by protein kinase A, although AQP2 does contain such a site at its C terminus. A small, 20 -30% increase in water permeability has been reported in oocytes expressing AQP2 (50). This result has also been controversial in oocyte studies (51) as well as in an in vitro phosphorylation study using AQP2-containing endosomes from kidney collecting duct (52). The data here, with 5-10% S.E. in P f determination, do not confirm the increase in P f in AQP2-expressing oocytes. Recent data suggest that a more likely role for the phosphorylation consensus site in AQP2 is in the regulation of trafficking in mammalian cells (53,54). Effects of protein kinase A phosphorylation on AQP3-5 and MIP have not been studied previously, nor have effects of protein kinase C phosphorylation on any of the proteins. Interestingly, protein kinase A activation increases water permeability by 2-fold in oocytes expressing the plant water channel ␣-TIP (55). Taken together, the results here do not support a role for phosphorylation in regulating the intrinsic water permeabilities of rat AQP1-5 and MIP.
In summary, the aquaporins and MIP have heterogeneous intrinsic water permeabilities. Measurable glycerol permeability was found only for AQP3. Mechanistic evaluation of the heterogeneous transport properties of the aquaporins will require high resolution structure studies. The biological reason(s) for the existence of a family of related but functionally distinct proteins remains to be resolved.