Localization and Functional Analysis of CHIP28k Water Channels in Stably Transfected Chinese Hamster Ovary Cells*

CHIP28 is a major water transporting protein in erythrocytes and plasma membranes in kidney proxi- mal tubule and thin descending limb of Henle. Chinese hamster ovary cells were stably transfected with the coding sequence of cloned rat kidney CHIP28k using expression vectors containing cytomegalovirus or Rous sarcoma virus promoters. Clonal cell populations expressed a 1.3-kilobase mRNA on Northern blot probed by CHIP28k cDNA and a 28-kDa protein on immunoblot probed by a polyclonal CHIP28 antibody. The clone with greatest expression produced -8 X 10’ copies of CHIP28k protein/cell. Plasma membrane os- motic water permeability (Pf), measured by stopped-flow light scattering, was 0.004 cm/s in control (vec- tor-transfected) cells (10 “C) and 0.014 cm/s in the CHIP28k-transfected cells. P f in CHIP28k-transfected cells had an activation energy of 4.9 kcal/mol and was reversibly inhibited by HgC12. CHIP28k expression did not affect the transport of protons and the small polar non-electrolytes urea and formamide. CHIP28k im-munoreactivity and function was then determined in subcellular fractions. P f in 6-carboxyfluorescein-la-beled endocytic vesicles, measured by as described above. Osmotic water transport was measured in vesicle fractions suspended at -0.5 mg protein/ml in 50 mM mannitol, 5 mM sodium phosphate, pH 7.4. Vesicles were subjected to a 250 mM inwardly directed sucrose gradient and vesicle shrinkage was followed by stopped-flow light scattering. Average vesicle radii were 218 nm (endoplasmic reticulum), 147 nm (Golgi), and 162 nm (surface membranes) as determined by quasi-elastic light scattering.

CHIP28 is a major water transporting protein in erythrocytes and plasma membranes in kidney proximal tubule and thin descending limb of Henle. Chinese hamster ovary cells were stably transfected with the coding sequence of cloned rat kidney CHIP28k using expression vectors containing cytomegalovirus or Rous sarcoma virus promoters. Clonal cell populations expressed a 1.3-kilobase mRNA on Northern blot probed by CHIP28k cDNA and a 28-kDa protein on immunoblot probed by a polyclonal CHIP28 antibody. The clone with greatest expression produced -8 X 10' copies of CHIP28k protein/cell. Plasma membrane osmotic water permeability ( P f ) , measured by stoppedflow light scattering, was 0.004 cm/s in control (vector-transfected) cells (10 "C) and 0.014 cm/s in the CHIP28k-transfected cells. P f in CHIP28k-transfected cells had an activation energy of 4.9 kcal/mol and was reversibly inhibited by HgC12. CHIP28k expression did not affect the transport of protons and the small polar non-electrolytes urea and formamide. CHIP28k immunoreactivity and function was then determined in subcellular fractions. P f in 6-carboxyfluorescein-labeled endocytic vesicles, measured by a stopped-flow fluorescence quenching assay, was 0.002 cm/s (control cells) and 0.01 1 cm/s (CHIP28k-transfected cells); Pt in transfected cells was inhibited by HgC12. Immunoblotting of fractionated endoplasmic reticulum, Golgi, and plasma membranes revealed high densities of CHIP28k (-5000 monomers/pm2 in plasma membrane) with different glycosylation patterns; functional water transport activity was present only in Golgi and plasma membrane vesicles. Antibody detection of CHIP28k by confocal fluorescence microscopy and immunogold electron microscopy revealed localization to plasma membrane and intracellular vesicles. These studies establish a stably transfected somatic cell line that strongly expresses functional CHIP28k water channels. As in the original proximal tubule cells, the expressed CHIP28k protein is a selective water channel that is functional in endocytic vesicles and the cell plasma membrane.
An abundant hydrophobic 28-kDa protein from human * This work was supported by Grants DK35124 and HL42368 from the National Institutes of Health and a grant-in-aid from the American Heart Association. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "Oduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
5 Established investigator of the American Heart Association. To whom correspondence should he addressed 1065 Health Sciences East Tower, University of California, San Francisco, CA 94143-0532. Tel.: 415-476-2172;Fax: 415-476-3381. erythrocytes (CHIP28) was recently isolated and cloned (1, 2). The amino acid sequence indicates homology to a class of ancient channel-like proteins which include the major intrinsic protein of lens, related proteins from bacteria, yeast, and plants (3), and a recently cloned protein from kidney collecting duct (4). CHIP28 functions as a selective water channel in the erythrocyte and selected tubule segments of the mammalian nephron. Expression of mRNA encoding CHIP28 in Xenopus oocytes increased water but not ion permeability (5,6), and proteoliposomes reconstituted with purified CHIP28 protein had high water but not proton and urea permeabilities (6-8). Antibody staining localized CHIP28 to plasma and intracellular membranes of constitutively water-permeable tissues in rat kidney, including proximal tubule and thin descending limb of Henle (9,10). In situ hybridization confirmed this localization and indicated a highly selective tissue distribution of mRNA encoding CHIP28 in selected epithelial and/or endothelial cells in lung, intestine, brain, eye, and other tissues (6,11).
In epithelial cells in kidney proximal tubule, high plasma membrane water permeability is important for the near isosmotic reabsorption of water filtered by the glomerulus. Functional studies indicated the presence of a water transporting protein on apical and basolateral plasma membranes (12,13), as well as on apically derived endocytic vesicles (14) and clathrin-coated vesicles (15). Intact tubule and cell-free vesicle measurements showed that water permeability in these membranes is high, weakly temperature dependent, and inhibited by mercurial sulfhydryl-reactive compounds (for reviews, see Refs. 16,17). An antisense CHIP28 oligonucleotide blocked the increase in water permeability in oocytes expressing heterologous mRNA from kidney cortex (6), suggesting that CHIP28 (or homologous proteins) is a major water transporter in kidney proximal tubule.
The purpose of this study was to establish a stably transfected somatic cell line expressing functional CHIP28 water channels and to investigate the stage of biogenesis in which CHIP28 protein attains functional maturity. In plasma membranes, it is believed that individually functional CHIP28 monomers (18-20) are assembled in tetramers (1,21) in which -50% of the monomeric subunits are glycosylated. Because of the small single channel water permeability of CHIP28 cm3/s, Ref. 5) and the relatively high endogenous water permeability of biological membranes cm/s), a high membrane density of functional water channels (>loo/ mm') was required in the somatic cell to increase water permeability significantly. In our studies, high expression of CHIP28 was accomplished in CHO-K1 (wild type) cells. The transporting characteristics of the expressed CHIP28 protein were similar to those in the original kidney cells. Interestingly, as observed in the native kidney proximal tubule, CHIP28 water channels were functional in both endocytic vesicles and Stable Transfection of CHIP28k Water Channels 22757 the cell plasma membrane. CHIP28 water channels were also localized in endoplasmic reticulum and Golgi, but functional only in Golgi. The results establish the first cultured somatic cell line for study of water channel biology.

EXPERIMENTAL PROCEDURES
Plasmid Constructions-The cDNA encoding rat kidney water channel CHIP28k in plasmid pSP64.CHIP28k (6) was used as template to amplify the 807 base pair coding region and introduce HindIII (5') and XbaI (3') restriction sites. The polymerase chain reaction fragment was filled-in with Klenow DNA polymerase I. The HindIII and XbaI cut fragment was ligated into mammalian expression vectors pRc/CMV' and pRc/RSV (Invitrogen) at the HindIII and XbaI sites and propagated in Escherichia coli. The CHIP28 sequence in the plasmid constructs were confirmed by DNA double-stranded sequencing following denaturation by the dideoxy chain-termination method using a Sequenase sequencing kit (U. S. Biochemical Corp.).
Cell Culture and Transfection-CHO-K1 (wild type) cells (obtained from University of California, San Francisco Cell Culture Facility) were grown in Ham's Nutrient Mix supplemented with 10% fetal calf serum at 37 "C in 5% COP. Transfection was carried out by use of Lipofectin (22). Cells were plated at a density of 5 X lo6 cells/60-mm diameter dish 12 h before transfection. The cells were washed three times just before transfection with Opti-MEM1 media (Life Technologies, Inc.). For each transfection, 20 pg of LipofectinTM reagent (Life Technologies, Inc.) was diluted into 1 ml of serum-free Opti-MEMl media, then combined with 1 ml of serum-free Opti-MEM1 media containing 10 pg of each recombinant plasmid (or vector only) and incubated at room temperature for 15 min. The mixture was then added to the washed CHO-K1 cells. The cells were incubated at 37 "C in a humidified environment with 5% COP for 12 h, washed twice with Ham's Nutrient Mix, and incubated for 24 h in 4 ml of Ham's Nutrient Mix supplemented with 10% fetal calf serum. Cells grown on each 60-mm dish were trypsinized and transferred to three 100mm diameter dishes. Selection of cells containing stably integrated copies of transfected DNA was accomplished by adding Geneticin (G418, Life Technologies, Inc.) to the growth media at a concentration of 500 pg/ml. After selection for 10-14 days, G418-resistant cell clones were isolated and transferred to separate culture dishes for expansion and analysis.
Northern Blot Analysis-Total RNA was isolated from -10' cells of each G418-resistant clone (and mock-transfected CHO-K1 cells) by SDS-EDTA lysis followed by phenol-chloroform extraction. 50 pg of each total RNA was resolved on a formaldehyde-agarose gel and blotted overnight onto a nylon membrane. The RNA was cross-linked to the membrane by UV light, prehybridized for 2 h in 5 X SSC solution, 5 X Danhart's solution, 50% formamide, 2% SDS, and 100 pg of denatured salmon sperm DNA at 50 "C, and hybridized overnight with lo6 counts/min/ml of CHIP28 cDNA probe (corresponding to the 807-base pair coding sequence) labeled with [ o I -~' P ]~C T P (Amersham Corp.) by random priming (Bethesda Research Laboratories). After hybridization, the membrane was washed twice in 2 X SSC, 0.1% SDS at room temperature, twice at 50 "C for 15 min each, and once with 0.1 X SSC, 0.1% SDS at 50 "C for 30 min. Hybridization was visualized by autoradiography. The same membrane was washed twice with boiling water and hybridized with a rat a-tubulin cDNA probe by the same procedure.
Immunocytochemistry-For antibody staining, CHO-K1 cells were cultured on glass coverslips overnight, fixed with 4% formaldehyde in PBS for 20 min, and permeabilized with 0.2% Triton X-100 in PBS for 10 min at room temperature. The immune serum containing polyclonal CHIP28 antibody (and preimmune serum control) was incubated in PBS containing 3% bovine serum albumin and 0.2% Triton X-100 for 1 h at room temperature and then washed four times with PBS containing 1% Triton X-100. Cells were then incubated with the secondary FITC-conjugated goat anti-rabbit antibody (1:20 dilution) for 1 h and washed four times in PBS containing 1% Triton X-100. Slides were viewed in a Leitz-Technical Instruments epifluorescence confocal microscope with a cooled CCD camera detector. Confocal images were obtained with a X 60 objective (Nikon, oil immersion, numerical aperture 1.4) in which the measured z-axis resolution of the confocal optics was 1.2 pm. In separate experiments, fluorescently stained endocytic vesicles were labeled and examined as described previously (23).
For immunogold electron microscopy, cells cultured on plastic were fixed in 4% paraformaldehyde, 0.1% gluteraldehyde and washed three times in PBS (0.9% NaCl in 10 mM sodium phosphate, pH 7.4). The cells were scraped from the wells, and the pellet was infiltrated overnight in 2.3 M sucrose. A drop of the pellet was frozen in liquid N, and ultrathin (60 nm thickness) sections were cut on a Reichert FC4D ultracryomicrotome at -70 "C. Sections were collected on carbon/Parlodion-coated nickel grids. The grids were washed with PBS, preincubated with PBS containing 1% albumin for 20 min, and then incubated with the a 1:400 dilution of immune serum for 90 min. Grids were washed in albumin-PBS and incubated with a 1:50 dilution of 15-nm protein A-gold in albumin-PBS. Grids were washed and fixed for 20 min in 1% glutaraldehyde in PBS. The grids were then washed with water, stained, and embedded in 0.2% uranyl acetate and 0.5% methylcellulose in water for 10 min, and dried. Sections were observed and photographed on a Philips CMlO electron microscope.
Assays for Plasma Membrane Water, Solute, and Proton Transport-Osmotic water permeability (Pf, cm/s) was measured in freshly suspended CHO-K1 cells by a light scattering method (12,15) using a Hi-Tech SF51 stopped-flow apparatus. The instrument dead time was 1.6 ms, and sample temperature was controlled by a circulating water bath. Cells from several 100-mm diameter plastic dishes were suspended in PBS by incubation in Ca2+/Mg2+-free PBS containing 50 mM EDTA for 2 min at 23 "C. Cells were washed three times in PBS and suspended at a concentration of -2 X lo6 cells/ml. 0.1 ml of the cell suspension was mixed in 1 ms with an equal volume of PBS containing 500 mM sucrose to give a 250 mM inwardly directed osmotic gradient. The osmotic gradient caused water efflux, cell shrinkage, and an increase in light scattering. The time course of 90 ' scattered light intensity at 530 nm was recorded. Measurements were performed five to eight times in each sample for signal averaging. Data were fitted to a biexponential function; water transport rates (kt, in s-') were calculated from the initial curve slope normalized to the total signal amplitude. P/ values were calculated from kf and the cell surface-to-volume ratio (5450 cm", calculated from average cell diameter of 11 gm measured by phase-contrast microscopy) as described previously (12,15). Cell permeability (Pa, cm/s) for the small polar solutes urea and formamide was determined from the time course of scattered light intensity in response to a 250 mM inwardly directed gradient of each solute. The solute gradient caused initial cell shrinkage due to osmotic water efflux, followed by slower cell swelling due to solute and water influx.
Passive proton permeability was measured from the kinetics of intracellular acidification in response to a sudden decrease in solution pH from 7.4 to 6.0. Freshly suspended CHO-K1 cells were loaded with the fluorescent pH indicator 2,7-bis-(2-carboxyethyl)-5-(and 6-)carboxyfluorescein (BCECF) by a 20-min incubation with 2 p~ BCECF-acetoxymethylester at 37 "C. Extracellular fluorophore was removed by two washes. The BCECF-loaded cells were suspended in a buffer consisting of 50 mM NaCl, 100 mM KCl, 5 mM potassium phosphate, pH 7.4, containing 2 p~ valinomycin. The membrane potential was depolarized and the K permeability was increased so that the rate-limiting step for dissipation of an imposed pH gradient was the passive proton permeability (24). The cell suspension was mixed in the stopped-flow apparatus with an equal volume of the same buffer (titrated to pH -2.7) so that pH after mixing was 6.0. The time course of BCECF fluorescence (excitation 490 nm, emission

Stable Transfection of CHIP28k Water Channels
>515 nm) was measured and used to calculate the initial rate of dissipation of the pH gradient (pH units/s). Fluorescent Labeling of Endosomes and Water Transport Assay-Endosomes were labeled with 6-carboxyfluorescein (6CF) in intact cells by fluid-phase endocytosis as described previously (25). Cells were grown on 150-mm diameter plastic dishes and pulse-labeled by incubation for 20 min a t 37 "C in PBS containing 15 mM 6CF. After loading, cells were washed five times in PBS at 4 "C. All subsequent procedures were carried out a t 4 "C. Cells were scraped, homogenized in 50 mM mannitol, 5 mM sodium phosphate, pH 8.5 using a Potter-Elvehjem homogenizer, and centrifuged at 2500 X g for 10 min to remove heavy cellular debris. The supernatant was centrifuged at 100,000 X g for 45 min to obtain a microsomal pellet containing the 6CF-labeled endosomes. The microsomal pellet was washed in the mannitol buffer, homogenized with a 25-gauge needle, and suspended at a concentration of -0.5 mg protein/ml. Endosome PI was measured by a stopped-flow fluorescence quenching assay (26). The microsome suspension was subjected to a 100 mM inwardly directed osmotic gradient in the stopped-flow apparatus. The osmotic gradient caused endosome shrinkage, an increase in the concentration of entrapped 6CF, and instantaneous 6CF fluorescence self-quenching. The time course of 6CF fluorescence (excitation 465 f 5 nm, emission >515 nm) was recorded and used to calculate endosome P I as described previously. The average endosome diameter was 195 nm (surface-to-volume ratio, 3 X lo5 cm") as determined by transmission electron microscopy using horseradish peroxidase as a fluid-phase marker and diaminobenzidine to produce an electron dense deposit (27). Subcellular Fractionation-Fractionation was performed by a modification of the method of Balch et al. (28). Cells from ten 25-cm diameter plates were grown to confluence (1-3 X 10' cells/plate), washed three times with PBS a t 37 "C, and incubated in PBS containing 8 mM EGTA for 2 min. The solution was aspirated and the cells were released by agitation and suspended in 10 ml of PBS. Cells were washed twice by centrifugation (100 X g, 10 min) at 4 "C in homogenizing buffer (HB, 250 mM sucrose, 10 mM Tris-HC1, pH 7.4). The pellet was resuspended in HB containing antipain (1 pglml), pepstatin (1 pglml), and benzamidine (15 pg/ml), and homogenized by 20 strokes of a glass Dounce homogenizer at 4 "C. The homogenate was centrifuged at 500 X g for 10 min at 4 "C, and the supernatant was adjusted to 1.4 M sucrose by addition of an equal volume of 2.3 M ice-cold sucrose in 10 mM Tris-HC1, 0.2 mM EDTA, pH 7.4. The gradient consisted of layers of 2 M sucrose (1 ml), 1.6 M sucrose (2 ml), the homogenate (4 ml in 1.4 M sucrose), 1.2 M sucrose (4 ml), and 0.8 M sucrose (1 ml). The gradient was centrifuged for 2.5 h at 25,000 revolutions/min in an SW27 rotor, and 1-ml fractions were collected for enzyme and functional assays and Western analysis.
Protein concentration was determined by the Bradford method. The following enzyme activities were determined by standard methods: a-glucosidase (ER marker, Ref. 29), a-mannosidase (Golgi marker, Ref. 30), and alkaline phosphodiesterase I (plasma membrane marker, Ref. 31). For immunoblotting, 1.5 pg of each fraction was resolved on 13% SDS-polyacrylamide gel electrophoresis and blotted to nitrocellulose as described above. Osmotic water transport was measured in vesicle fractions suspended at -0.5 mg protein/ml in 50 mM mannitol, 5 mM sodium phosphate, pH 7.4. Vesicles were subjected to a 250 mM inwardly directed sucrose gradient and vesicle shrinkage was followed by stopped-flow light scattering. Average vesicle radii were 218 nm (endoplasmic reticulum), 147 nm (Golgi), and 162 nm (surface membranes) as determined by quasi-elastic light scattering.

RESULTS
Expression of CHIP2812 in Transfected Cells-Clonal cell lines derived from stably transfected CHO-K1 cells were expanded and examined for CHIP28k mRNA and protein. Fig. L4 shows a Northern blot of RNA isolated from a series of clonal cell lines probed with 32P-labeled CHIP28k and atubulin cDNAs. There was no expression of CHIP28k mRNA in cells transfected with vector alone ("mock-transfected") and variable levels of expression in cells transfected with the coding region of CHIP28k. Expression was variable with the CMV vector, but consistently high with the RSV vector. Expression of a-tubulin (control) was observed in every cell line.  These studies indicate the expression of CHIP28k mRNA and immunoreactive protein only in the stably transfected CHO-K1 cells.
The morphology of the CMV-4 stably transfected cells (more than five passages after cloning) was not different from that of the mock-transfected cells as observed by phasecontrast light microscopy and transmission electron microscopy (not shown). Freshly suspended mock-and stably transfected cells were spherical with an average diameter of 11 pm. The growth characteristics of the mock-and stably transfected cells were not different as assayed by [3H]thymidine incorporation (1 pCi/well of 24-well plate; relative 8 h incorporation: mock, 1.0; CHIP28k, 1.2) and cell density (in cells/ cm2) at 24 h (mock 5.7 X lo5; CHIP28k, 6.2 X lo5) and 48 h

Stable Transfection
of CHIP28k Water Channels 22759 (mock, 1.7 X lo6; CHIP28k, 1.9 X lo6) after plating at a density of 2.5 X IO5 cells/cm*. Morphological examination of patterns of endocytic uptake of 6CF and FITC-transferrin in the mock-and stably transfected cells showed no obvious differences in the rate of marker uptake (relative uptake after 5 min of pulse labeling with FITC-transferrin: mock, 1.0; CHIP28k, 1.1 +-0.2) and in the number, size and cellular location of early and late endosomes. It was noted, however, that the morphology and growth of the CMV-4 transfected cells changed after five to eight passages, whereas the transfected cells having lower levels of CHIP28k expression could be passed >15 times without notable change.
Cell Osmotic Water Permeability-A stopped-flow light scattering method was used to examine CHIP28k function in the plasma membrane of transfected cells. The adherent cells were freshly suspended by agitation in an EDTA-containing solution. Fig. 2, A and B, show the time course of scattered light intensity in response to a sudden increase in extracellular osmolality from 300 to 550 mOsm. The osmotic gradient caused water efflux, cell shrinkage, and an increase in scattered light intensity. The data are shown using three times scales to follow the full kinetics of cell shrinkage. Pf in the mock-transfected cells was 0.0039 & 0.0002 cm/s (S.E., n = 5 separate sets of measurements, each consisting of the average of five to eight individual curves) at 10 "C. Pf in the mocktransfected cells was not affected by HgC12, P-mercaptoethanol, or dimethyl sulfoxide ( Fig. 2A and Table I). The cells transfected with CHIP28k (Fig. 2B) had a remarkably increased Pf as shown by the increase in initial slope. At 10 "C, the cells used in Fig. 2B, which had the highest CHIP28k mRNA expression by Northern analysis (clone CMV-4), had a Pf of 0.014 f 0.001 cm/s ( n = 5 ) . Although a clonal cell population was used for this experiment, biexponential analysis indicated that 68% of the signal arose from cells having high Pr; the remaining cells with lower apparent Pf may have been damaged in the suspension procedure. (For this reason, Pf was measured in isolated surface membrane vesicles, see below.) The level of CHIP28k expression for several clonal cell populations (mock, CMV-1, CMV-3, RSV-1, and CMV-4; relative mRNA expression by densitometry: 0, 0.04, 0.22, 0.46, l ) , correlated reasonably well with measured plasma membrane Pf (in cm/s x at 10 "C: 3.9, 4.5, 6.9, 10, and 14, respectively).
Two additional characteristics of the CHIP28k water channel in native tissues are inhibition by mercurials and a weak temperature dependence (16, 17). Fig. 2B shows that incubation of the CHIP28k-transfected cells with 20 mM HgC12 for 5 min strongly inhibited the increase in Pf associated with CHIP28k (54 f 3% inhibition, n = 3). The inhibition by HgClz was partially reversed (to 21 f 5% inhibition) by posttreatment with ,&mercaptoethanol. Fig. 2B also shows that dimethyl sulfoxide inhibited the CHIP28k-associated increase in Pf, but had no effect on Pf in mock-transfected cells. The cAMP agonists forskolin (25 p~) and chlorophenylthio-cAMP (0.5 mM) did not affect Pf in mock-or CHIP28ktransfected cells (not shown). Fig. 2C shows an Arrhenius plot for the temperature dependence of P,. The activation energy (EJ, given by the slope of the Arrhenius plot (Fig. XC), was 4.9 f 0.6 kcal/mol for cells expressing CHIP28k and 9.2 f 0.8 kcal/mol for the mock-transfected cells. Taken together, the high Pf, the inhibition by HgClz and dimethyl sulfoxide, and the low E,, indicate that the expressed CHIP28k protein in CHO-K1 cells is functionally similar to that in the original tissue.
To investigate whether the expressed CHIP28k protein transported small solutes, measurements of urea, formamide, were averaged and fitted to a biexponential function. Increased scattered light intensity corresponds to cell shrinkage. Where indicated, cells were incubated with 20 p~ HgC1, for 5 min, followed by 5 mM j3-mercaptoethanol ( M E ) for 5 min, before stopped-flow measurements. Where indicated, cells were incubated with 500 mM dimethyl sulfoxide for 5 min before measurements. Note that the data are shown using three contiguous time scales. C, Arrhenius plot for temperature dependence of water permeability in the mock-and CHIP28k-transfected (clone CMV-4) CHO-K1 cells. Each data point is the average of 5-10 measurements; mean S.E. for In k, was 0.12. Fitted line slopes gave activation energies of 9.2 k 0.8 kcal/mol (mocktransfected) and 4.9 k 0.6 kcal/mol (CHZP28k-transfected). and proton permeability were carried out in the mock-and CHIP28k-transfected cells. Table I summarizes the results. Although water permeability differed in the mock-and stably transfected cell lines, there was no systematic difference in the permeabilities of the small solutes urea and formamide. The passive proton permeability, measured from the rate of dissipation of a preformed pH gradient, was not different in the mock-and CHIP28k-transfected cells. These results indicate that CHIP28k is a selective water transporter that excludes small solutes.  CHO-Kl cells Permeabilities (mean f S.E., n = 3-5 separate sets of measurements, each consisting of the average of five to eight individual curves) were measured as described under "Experimental Procedures." The CHIPZEk-transfected cell line was clone CMV-4 and the mocktransfected cell line was established by transfection with the empty CMV vector. Water transport experiments were carried out at 10 "C as described in the legend to Fig. 2 Endosome Water Permeability-Osmotic water permeability was measured in endosomes derived from the CHO-K1 cells. Endosomes were labeled with the membrane-impermeant fluorophore 6CF by fluid-phase endocytosis. Water permeability was assayed in a crude microsomal fraction containing the fluorescently labeled endosomes using a stopped-flow fluorescence quenching technique. The osmolality of buffer bathing the microsomes was suddenly increased from 60 to 160 mOsm, causing osmotic water efflux, microsome shrinkage, and 6CF fluorescence self-quenching. The fluorescence signal originated selectively from the fluorescently labeled endosomes, and not from the majority of nonfluorescent microsomes. Fig. 3 shows that the rate of fluorescence decrease was remarkably more rapid in endosomes derived from CHIP28k-transfected CHO-K1 cells (Fig. 3B, upper curue) than from mock-transfected cells (Fig. 3A, upper curue). Pr values were 0.0021 f 0.0001 cm/s ( n = 4, mocktransfected cells) and 0.011 ? 0,001 cm/s (CHIP28k-transfected cells). Analysis of pre-exponential factors of the biexponential fit indicated that >60% of the fluorescently labeled vesicles contained functional water channels. In control studies in which endocytosis was inhibited (33), the CHIP28ktransfected cells were labeled with 6CF at 4 "C (instead of 37 "C), or at 37 "C in the presence of NaN3. Fig. 3B (bottom curues) shows that the signal amplitude was nearly zero, suggesting that generation of the fluorescence signal required endocytosis.

Stable Transfection
The inhibition properties and temperature dependence of endosome Pr was measured. The increase in Pf in endosomes derived from CHIP28k-transfected cells was inhibited by 92 6% ( n = 2) by 0.6 mM HgC12, whereas Pf in endosomes derived from the mock-transfected cells was unaffected (inhibition 8 f 9%, Fig. 3). Similar inhibition of P, in CHIP28ktransfected cells, but not mock-transfected cells, was observed with dimethyl sulfoxide. When measured at 21 "C, endosome Pf values were 0.014 cm/s (CHIP28k-transfected) and 0.0039 cm/s (mock-transfected). Assuming a linear Arrhenius relation, calculated E, values were 4.2 kcal/mol (CHIP28k-transfected) and 9.6 kcal/mol (mock-transfected). These results indicate the presence of functional water channels in endosomes derived from the CHIP28k-transfected, but not from the mock-transfected CHO-K1 cells.
CHIP28k Localization and Function in Isolated Membrane Fractions-To examine the functional maturation of CHIP28k, cell homogenates were fractionated by sucrose gradient centrifugation for immunoblot and functional analysis. A cell-free microsomal suspension containing 6CF-labeled endocytic vesicles was prepared as described under "EXperimental Procedures." Osmotic water permeability was measured by a stopped-flow fluorescence quenching assay in which the microsome suspension was subjected to a 100 mOsm inwardly directed osmotic gradient. Decreasing fluorescence corresponds to endosome shrinkage. Each curve is the average of five to eight measurements shown with biexponential fit. Where indicated, 0.6 mM HgC12 or 200 mM dimethyl sulfoxide was added to the microsome suspension 2 min before measurements. Where indicated, cells were incubated with 6CF at 4 "C instead of 37 "C to inhibit endocytosis at the time of labeling. Where indicated, 0.05% NaN3 was added for 5 rnin before and during the 37 "C incubation with 6CF to inhibit endocytosis. Note that the data are shown using two contiguous time scales.
CHIP28k-transfected cells and is similar to previous reports in CHO cells (28). Note that a-mannosidase is not a highly specific Golgi marker; Golgi vesicles are primarily present in fractions 9-11 and should be absent in fractions 1-8. The alkaline phosphodiesterase activity in fractions 1 and 2 is due to the presence of lysosomes. Fig. 4 8 is an immunoblot probed with the anti-CHIP28 antibody. The endoplasmic reticulum vesicles showed mainly the 28 kDa band (arrow a) with some intermediate form (arrow b), whereas the Golgi and plasma membranes show a glycosylation pattern (arrow c) similar to that observed in native erythrocyte and kidney membranes. -20 kDa transfected cells had a very high Pf (0.026 cm/s) that was inhibited by >95% by HgC12 to levels observed for plasma membranes isolated from the mock-transfected cells. The analysis indicated that >95% of plasma membrane vesicles were highly water permeable. Water transport in plasma membranes from mock-transfected cells was not inhibited by HgClz (not shown). Quantitative immunoblotting showed the presence of 78 pg of CHIP28k (glycosylated + non-glycosylated) per mg total plasma membrane protein. Water transport was also high (Pf 0.015 cm/s) and HgClz inhibitable in the Golgi fraction from CHIP28k-transfected cells; Pf was low (0.0013 cm/s) and HgC12 insensitive in the Golgi fraction from the mock-transfected cells. In contrast, immunoreactive CHIP28k protein was present in the endoplasmic reticulum fraction, but not functional as shown by the low Pf that was similar to Pf in the endoplasmic reticulum fraction from mocktransfected cells, and the lack of inhibition by HgC12. These studies indicate that CHIP28k attains functional competence in the Golgi.
Zmmunolocalization of CHZP28k"Fluorescent antibody staining experiments were carried out to determine the cellular location of expressed CHIP28k. In non-permeabilized fixed cells, there was little cell staining, consistent with immunoelectron microscopy studies suggesting that the antibody binds predominantly to a cytoplasmic epitope of CHIP28k (9). Permeabilized CHIP28k-transfected cells were strongly stained (Fig. 5 B ) compared to the mock-transfected cells (Fig.  5A). The staining of the mock-transfected cells was similar to that observed for the CHIP28k-transfected cells incubated with preimmune serum (not shown).
To determine the subcellular location of CHIP28k, cells were viewed with high magnification confocal optics. Fig. 5, C and D, are confocal micrographs of permeabilized adherent cells where the focal plane was near the base of the cells (Fig.  5C) and half-way between the cell base and apex (Fig. 50).
Note that the same cell field is shown in Fig. 5, C and D. In the linear fluorescence at the cell base. In Fig. 50, staining of intracellular vesicles was observed in the cytosolic, but not the nuclear compartment. Staining of endocytic vesicles was confirmed by a double label experiment in which many FITCstained vesicles colocalized with vesicles that had previously internalized tetramethylrhodamine-dextran by fluid-phase endocytosis (not shown). No staining of intracellular vesicles was observed in the CHIP28k-transfected cells using preimmune serum or in the mock-transfected cells.
CHIP28k cellular localization was studied at higher resolution by immunogold labeling. The polyclonal anti-CHIP28 antibody did not label mock-transfected cells (Fig. 6A). In contrast, the plasma membrane of CHIP28k-transfected cells was labeled heavily (Fig.  6B). Although most of the gold labeling was located on the plasma membrane, anti-CHIP28 antibodies consistently labeled Golgi (Fig. 6C) and a population of cytoplasmic vesicles (Fig. 6D).

DISCUSSION
The purpose of this study was to establish a somatic cell line that stably expressed functional CHIP28k water channels. High levels of CHIP28k mRNA and protein were expressed in CHO-K1 cells that were stably transfected with CHIP28k cDNA in plasmids having CMV or RSV viral promoters. The functional characteristics of the expressed water channel were similar to those in the original tissue. The functional studies supported the conclusion that CHIP28k is a selective water transporting protein. The stably transfected cells expressing CHIP28k should be useful for studies of the functioning, pharmacology, and trafficking of mutant and homologous CHIP28 proteins. In addition, the abundant quantities of expressed functional protein make the transfected CHO-K1 cell system suitable for generation of mutant and homologous proteins for which there is no suitable native tissue source.
There were a number of special concerns for the stable expression of water channels in somatic cells. First, whereas

K1 cells. A and B, permeabilized fixed CHO-K1 cells (A, CMV
mock-transfected; B, CHIP28k-transfected) grown on glass coverslips were incubated with a rabbit anti-CHIP28 polyclonal antibody and fluorescently stained with an FITC-conjugated goat anti-rabbit IgG (H+L) secondary antibody. Cells were viewed by wide-field epifluorescence microscopy with FITC filter set and imaged by a cooled CCD camera. Illumination intensities and gains were identical in A and B. C and D, confocal epifluorescence micrographs of CHIP28k-transfected cells prepared as in B. Images were obtained at the cell base adjacent to the coverglass (C) and approximately half-way between the cell base and apex (D). The same field is shown in C and D. Magnifications: A and B, X 5000; C and D, X 1000. the expression of tens or hundreds of ion channels is sufficient to give a large increase in cell ion conductance, the physical factors that limit the volume of water moving through a single water channel (17) require the expression of large quantities of water channels to give a measurable increase in water permeability. In erythrocytes, kidney cells, and the "high expression" CHO-K1 cells studied here, 2-8% of membrane protein is CHIP28. Second, it was not known a priori whether the increase in water permeability in cell plasma membranes and intracellular vesicles would affect cell viability. Although water movement across membranes is always secondary to osmotic gradients and hydrostatic forces, some cell processes, such as fusion and sorting of vesicles, are associated with volume changes. It has been proposed that the kinetics of volume change may be important for vesicle fusion and other intracellular events (34). We found that the transfected CHO cells expressing large quantities of CHIP28k protein had similar morphology, growth characteristics, and vesicular trafficking to the mock-transfected (vector alone) cells.
Another concern in these studies was the functional assay for water channels. There are no good methods to measure rapid volume changes in adherent cells. A light scattering microscopy method has been applied to measure Pi in rela-tively large round cells (5774 macrophages,Ref. 35), but was unsuccessful for the CHO-K1 cells. Our approach here was to adapt the stopped-flow light scattering method to measure the rate of osmotically induced cell shrinkage of freshly suspended cells. Initial studies in suspended cells obtained by trypsinization of the adherent cultures showed no evidence of functional water channels, probably because of partial digestion of the water channel protein. High water permeability in stably transfected cells was successfully observed when cells were released from the plastic support by EDTA and mechanical agitation. The light scattering method was also used to measure water permeability in isolated subcellular vesicles. The measurement of water transport in small vesicles is not subject to effects of unstirred layers or loss of cell viability during the suspension procedure.
Although a high level of functional CHIP28k protein was expressed in the stably transfected CHO-K1 cells, -10-fold lower levels were obtained in stably transfected MDCK epithelial cells using the same transfection procedures (data not presented). Further, preliminary attempts to express high levels of CHIP28 by transient transfection of COS cells using vector pCDM8 (Invitrogen) were unsuccessful. The generation of transfected polar epithelial cells lines would be particularly useful for studies of CHIP28k targeting because CHIP28k may be one of the few transporting proteins that is present on both apical and basolateral membranes (9).
The characteristic features of the water transporting pathway in erythrocytes and kidney proximal tubule are a high and selective membrane permeability to water, inhibition of water permeability by mercurials, a weakly temperature-dependent water permeability, and a high ratio of osmotic-todiffusional water permeability (16, 17). P, in the stably transfected cells was 0.018 cm/s at 20 "C, similar to that of 0.02 cm/s in human erythrocytes (36) and 0.01-0.04 cm/s in proximal tubule apical and basolateral plasma membranes (12). It is estimated that each CHO-K1 cell contains 1.6 x lo6 copies of CHIP28k in the plasma membrane based on the density of CHIP28k measured in the surface membrane fraction (43001 pm2, see below) and cell surface area (380 pm'). Compared to the total number of immunoreactive CHIP28k molecules/cell of 8 x lo6 estimated by quantitative immunoblot, it is concluded that significant quantities of CHIP28k are intracellular. This conclusion is supported by the findings oE ( a ) immunoreactive and functional water channels in Golgi (Fig.  4), ( b ) functional water channels in the majority of endocytic vesicles (Fig. 3), and (c) immunolocalization of CHIP28k to intracellular vesicles by confocal (Fig. 5) and electron (Fig. 6) microscopy.
The expressed CHIP28k water transporter was functionally similar to that in the original kidney tissue in terms of single channel water permeability, water transport specificity, inhibition pharmacology, and Arrhenius activation energy. The single channel water permeability (Pi) of the CHIP28 water channel was -5 X cm3/s at 10 "C in reconstituted proteoliposomes containing purified CHIP28 (7,8). In the surface membrane fraction isolated from CHIP28k-transfected cells, Pf of the expressed CHIP28k protein was similar (6 X cm3/s) as calculated from a CHIP28k (monomer) channel density of 4300/pm2 (determined by quantitative immunoblot, assuming 50% of membrane weight is protein) and the measured vesicle Pf of 0.026 cm/s at 10 "C. There was no difference in proton, urea, and formamide permeabilities in the mock-and CHIP28k-transfected cells. Water permeability in the CHIP28k-transfected CHO-K1 cells was inhibited by HgClz and dimethyl sulfoxide; the remaining water permeability after inhibition was 0.0045 cm/s, similar to that of 0.004 cm/s in the mock-transfected CHO-K1 cells. The Arrhenius activation energy for the expressed water channel in CHO-K1 cells was 4.9 kcal/mol, within the range of 2-5 kcal/mol for the erythrocyte and kidney proximal tubule (16,36). The results suggest that the functional properties of expressed CHIP28 water channels are similar to those in the native erythrocyte and kidney. It was not possible to measure the ratio of osmotic-to-diffusional'water permeability in these cells because the diffusional water permeability would be seriously underestimated (and not easily correctable) because of unstirred layer effects. Immunoblot analysis indicated CHIP28k water channels were present on plasma membrane, endoplasmic reticulum, and Golgi fractions isolated from stably transfected CHO-K1 cells. Whereas little glycosylation was found in the endoplasmic reticulum fraction, the glycosylation pattern of surface membranes was similar to that in CHIP28 from native tissues in which -50% of monomers are glycosylated. The glycosylation pattern of Golgi vesicles was similar to that in plasma membranes. Functional studies were performed in the plasma membrane, endoplasmic reticulum, and Golgi vesicles by a stopped-flow light scattering technique. Plasma membrane and Golgi fractions had high water permeability that was inhibited by HgC12; however, CHIP28k was present but not functional in endoplasmic reticulum. These studies indicate that functional maturity of CHIP28k is attained in Golgi. Although site-directed mutagenesis studies indicate that CHIP28k glycosylation is not necessary for its water transporting function (ZO), there may be other topological or biochemical modifications achieved in Golgi that confer water transporting function.
Functional CHIP28k water channels were demonstrated in endosomes by a fluorescence quenching assay performed on crude microsomes containing fluorescently labeled endosomes. The assay for water transport in endosomes was used previously to demonstrate functional water channels in apically derived endosomes from kidney proximal tubule (14) and from vasopressin-stimulated kidney collecting tubule (26) and toad urinary bladder (27,33). Water permeability in endosomes isolated from these native cells was high (Pf, 0.03-0.1 cm/s), weakly temperature-dependent (E,,, 2-5 kcal/mol), and inhibited by HgC12. Using the same methods, water channels were absent in endosomes derived from a series of established epithelial cell lines derived from renal tubules and amphibian urinary bladder (25). It was found here that water permeability in endosomes derived from mock-transfected CHO-K1 cells was low and not inhibited by HgC12 and dimethyl sulfoxide, whereas endosomes from CHIP28k-transfected cells had high Pf that was weakly temperature-dependent (E,,, 4.2 kcal/mol) and strongly inhibited by HgC12 and dimethyl sulfoxide. Using a single channel CHIP28k water permeability of 5 X cm3/s, endosome Pf of 0.011 cm/s, and endosome diameter of 195 nm, it is estimated that each endosome contained 260 functional CHIP28k water channels. The estimated membrane density of water channels in the endosomal membrane (2000/pm2) was less than that of 43001 pm2 calculated for the cell plasma membrane. Analysis of the fluorescence curve shape in Fig. 3B indicated that the majority of fluorescently labeled endosomes in the CHIP28k-transfected cells were water-permeable. These results suggest that CHIP28k is present on endocytic vesicles in the stably transfected cells at a density of -50% of that found in the cell plasma membrane. Functional CHIP28k water channels are present on endocytic vesicles in kidney proximal tubule (14). The role(s) of these intracellular water channels is not known, nor is there evidence that water permeability in proximal tubule is regulated by physiological factors. Apical membrane turnover in proximal tubule is very rapid and assumed to be constitutive (37), whereas in collecting duct, water channel trafficking between the cell apical membrane and an intracellular vesicular compartment is regulated by the hormone vasopressin (38). In proximal tubule, the high density of water channels present in intracellular vesicles raises the possibility that Stable Transfection of CHIP2812 Water Channels plasma membrane water permeability may, under some conditions, be regulated by physiological factors. Studies of water transport regulation in isolated perfused proximal tubules are required to address this issue.
The stably transfected cell line developed here should be useful in studies of the biochemical and intracellular processing of CHIP28k water channels. Stably transfected cells lines have provided important data about transporting mechanisms and biochemical processing for a number of integral membrane transport proteins, including the cystic fibrosis gene product (CFTR) (39), sodium channel (40), and y-aminobutyric acid transporter (41). Recently, Lukacs et al. (42) examined the distribution and function of CFTR using CHO-K1 cells that were stably transfected with CFTR cDNA. They obtained functional evidence that CFTR was functional in endocytic vesicles. In C2C12 cells transfected with glucose transporters, the GLUT1 and GLUT4 proteins were also targeted to intracellular vesicles, however insulin did not cause glucose transporter exocytosis (43). The stably transfected cells described here should be useful for identification of the intracellular signals involved in water channel processing and plasma membrane targeting.