INTRODUCTION

Top Abstract Introduction Procedures Results Discussion References

Human erythrocytes are highly permeable to water, urea, and glycerol (1-3). The existence of membrane proteins that facilitate water and solute movements in these cells has been postulated, but most of these proteins remain to be characterized. In 1992, Agre and co-workers identified a very abundant protein of the human red blood cell as the first water-selective channel that was named aquaporin-1 (AQP1)¹ (4). Colton-null erythrocytes, lacking AQP1, exhibit a reduced osmotic water permeability (5). However, these cells are found to be residually permeable to water, suggesting the presence of additional, non-AQP1, water channels.

The existence of a protein-mediated glycerol transport in erythrocytes has relied mostly on pharmacological evidence. In particular, inhibition of glycerol transport by sulfhydryl reagents, phloretin, and copper ions has been reported (1, 3, 6). Yet, the identity of the erythrocyte glycerol carrier remains unknown. The glycerol facilitators identified to date all belong to the Major Intrinsic Protein (MIP) family (7). They have been characterized in several organisms: GlpF, a bacterial glycerol permease facilitator (8); Fps1p, a yeast glycerol exporter (9); aquaporin-3 (AQP3), initially characterized in mammalian kidney (10-12); and AQP7, recently identified in rat testis (13). Compared with other mammalian aquaporins, which are selective mostly for water, AQP3 is moderately permeable to water, but highly permeable to glycerol and possibly to urea (11, 12, 14). AQP3 expression has been reported in several mammalian tissues, including kidney, intestine, stomach, spleen, and eye (11, 15, 16).

The aim of this study was to investigate the nature of the red cell glycerol transporter. Due to the high permeability of AQP3 to glycerol, we hypothesized that AQP3, if expressed in human erythrocytes, could account for their facilitated glycerol permeability. Since AQP1 itself was previously suggested to transport glycerol (17), in our study we used red cells from an individual who does not express AQP1, associated to the Colton group antigens (18). We found no difference in glycerol transport between the normal and Colton-null (Co(a-b-)) red cells, suggesting that AQP1 does not constitute a major pathway for glycerol. By contrast, our immunological data demonstrate that AQP3 is expressed in both normal and Colton-null red cells. Pharmacological evidence strongly supports the hypothesis that AQP3 could account for the high glycerol permeability of these cells and that, in Colton-null erythrocytes, AQP3 could constitute a residual, mercury-sensitive pathway for water.

	EXPERIMENTAL PROCEDURES

Top Abstract Introduction Procedures Results Discussion References

Materials-- Control and Co(a-b-) red cells (18) were obtained from the Institut National de Transfusion Sanguine (Paris, France). Rabbit polyclonal antibodies against purified human AQP1 were raised as described previously (19). Polyclonal antibodies against a 26-amino acid synthetic peptide corresponding to the carboxyl terminus of rat kidney AQP3 were raised in rabbits (Neosystem, France). The anti-AQP3 serum was affinity purified by chromatography (immunobilization Kit 2, Pierce). The secondary antibodies used in Western blot experiments (goat anti-rabbit-peroxidase) were from Promega (Madison, WI); those used in immunofluorescence (mouse anti-rabbit-CY3) were from Jackson Immunoresearch. ¹⁴C-Glycerol (120 mCi/mM) was synthesized by the Service des Molécules Marquées (CEA/Saclay, France). Phenylmethylsulfonyl fluoride, pepstatin, DIDS, and phloretin were from Sigma Immunochemicals. Reagents used in electrophoresis and immunoblotting were from Pharmacia LKB Biotechnology (Uppsala, Sweden) and from Amersham Corporation (Buckinghamshire, United Kingdom). All other reagents were at least analytical grade.

Red Cell Preparation-- Thawed normal and Co(a-b-) red cells were washed three times in a Carlsen buffer (3) (in mM): NaCl, 154; D-glucose, 5; KH₂PO₄, 0.25; Na₂HPO₄, 0.25; pH 7.4, 300 mosm/kg H₂O, by centrifugations at 2,000 × g for 10 min at 8 °C. The cells were then resuspended in the Carlsen buffer at a hematocrit of 1.5%, corresponding to ~10⁷ cells/ml and directly used for stopped-flow experiments.

Stopped-flow Experiments-- Kinetics of red cell volume changes were followed at 26 °C, by 90° light scattering (_exc = 600 nm) using a stopped-flow spectrophotometer (SFM3, Biologic, Claix, France). An emission wavelength > 500 nm was obtained by using a cut-on filter (Specivex, J526a). Cell osmotic water permeability was measured by mixing 100 µl of cells with an equal volume of a hyperosmotic solution of sucrose to produce a 100 mosm/kg H₂O inwardly-directed osmotic gradient. Data from at least 10 time-courses were averaged and fitted to single exponential functions (k is the rate constant, in s¹) by using the Simplex procedure of the BIOKINE software (Biologic, France). The osmotic water permeability coefficient, P_f, in cm/s, was determined using the following equation (5),

(Eq. 1)

where V(t) is the relative red cell volume as a function of time, [SAV] is the red cell surface area to initial volume ratio, MVW is the molar volume of water (18 cm³/mol) and C_in and C_out (mol/cm³) are the initial concentrations of intracellular and extracellular solute. SAV was taken equal to 13,500 cm¹ (20). Theoretical kinetics were simulated for various arbitrary values of P_f and fitted by the BIOKINE software as single exponentials. The obtained calibration curve (P_f versus k) was used to determine P_f from the experimental k values.

One-step Pink Ghost Preparation and Glycerol Efflux Measurements-- The hemoglobin-free erythrocyte ghosts were prepared by the method described by Ojcius (22). Briefly, washed normal and Colton-null red cells were lysed in a hypotonic medium, 5 mM Na₂HPO₄, pH 8.0, containing 20 µg/ml phenylmethylsulfonyl fluoride, and 2 µg/ml pepstatin, and centrifuged at 25,000 × g for 20 min at 4 °C. The membranes were resuspended in 5 mM Na₂HPO₄, 100 mM NaCl, and 100 mM glycerol, pH 8.0, at 10⁸ ghosts/ml, and sealed by incubation at 37 °C for 1 h. Glycerol effluxes were measured at 20 °C by a rapid filtration method (23). Samples of 6 × 10⁷ ghosts were loaded for 1 h with ¹⁴C-glycerol (5 µCi/ml) and ³H-mannitol (5 µCi/ml). Mannitol was used as an impermeant intracellular marker, to quantify the ghosts retained on glass-fiber filters (Whatman GF/A). The filters were rinsed over a preset period of 500 ms to 10 s at a flow rate of 1 ml/s. The cold washout solution contained 100 mM raffinose instead of glycerol. The retained radioactivity (¹⁴C/³H) was quantified by liquid scintillation. Glycerol permeability coefficient (P_gly, in cm/s) was calculated from the rate constant of the exponential time course of glycerol efflux and from ghost surface area to initial volume ratio.

Electrophoresis and Immunoblotting-- Human and rat hemoglobin-free ghosts were prepared as described above. Purified human AQP1 was obtained as described previously (17). Bloodless rat kidney outer medulla was prepared according to Ecelbarger et al. (15). Protein concentration of the membrane fractions was measured using the Pierce BCA Protein Assay reagent kit. Five µg of membrane proteins and 0.6 µg of pure AQP1 were denatured at 65 °C for 10 min in Laemmli sample buffer (24), and separated by 12.5% SDS-polyacrylamide gel electrophoresis. Proteins were transferred to PVDF membranes (NEN Life Science Products) and probed with the affinity-purified anti-AQP3 at approximately 0.25 µg/ml or the anti-AQP1 whole serum diluted 1:500. Controls were carried out by using the affinity-purified anti-AQP3 preabsorbed for 10 min with 0.2 mg/ml of immunizing peptide. Immunoreactive proteins were revealed by the ECL Western blotting technique (Enhanced ChemiLuminescence, Amersham Pharmacia Biotech).

Indirect Immunofluorescence-- Erythrocytes washed in PBS were fixed for 1 h in 4% paraformaldehyde in PBS, followed by PBS washes. After a 3-min centrifugation at 2000 × g, the cell pellet was infiltrated in PBS containing 2.3 M sucrose. The samples were frozen in liquid nitrogen, and 0.75 µm sections were cut on an ultracryomicrotome at 70 °C (Reichert Ultracut, Leica, Wien, Austria) and collected on Superfrost Plus glass slides. Adult Sprague-Dawley male rats were anesthetized (pentobarbital, 5 mg/100 g of body weight, intraperitoneally) and perfused with a fixative containing 7.1 mM Na₂HPO₄, 30.4 mM NaH₂PO₄, 75 mM lysine, 10 mM sodium periodate, and 2% paraformaldehyde (pH 7.4). Kidneys were removed, sliced, and kept in fixative overnight at 4 °C, followed by extensive PBS washes. Tissue slices were infiltrated overnight in PBS containing 30% sucrose and frozen in liquid N₂. Cryosections of 5 µm were cut on a cryostat (Leica) and collected on Superfrost Plus glass slides. The slides were incubated for 5 min in PBS containing 1% bovine serum albumin (PBS/BSA), followed by 1-h incubation in primary antibody (1:300 dilution of anti-AQP3 antiserum with or without preincubation of the serum with the peptide used for immunization, or 1:100 dilution of anti-AQP1 anti-serum) in PBS/BSA. The sections were then washed 3 × 10 min in PBS, followed by a 45-min incubation in CY3-conjugated mouse anti-rabbit antibodies (6 µg/ml) in PBS/BSA. The sections were washed 2 × 10 min in PBS and mounted for observation under a fluorescence microscope (Olympus Vanox-T).

	RESULTS

Top Abstract Introduction Procedures Results Discussion References

Immunoblot Analysis of AQP3 Expression in Various Tissues-- Fig. 1A shows results from immunoblots probed with affinity-purified antibodies raised against a rat AQP3 carboxyl-terminus peptide. In rat renal outer medulla (Fig. 1A, lane 4), a characteristic profile of three stained bands was observed: one band at ~25 kDa and two broader bands at 33-40 and >67 kDa. None of these bands was detected when the anti-AQP3 antibody was preabsorbed with the immunizing peptide (Fig. 1A, lane 1). A similar profile with three stained bands was also observed in rat red cell membranes probed with the anti-AQP3 antibody (Fig. 1A, lane 5), suggesting that, in addition to kidney, AQP3 was also expressed in rat erythrocytes. Again, no signal was detected when the anti-AQP3 was preincubated with the immunizing peptide (Fig. 1A, lane 2). Although the 26 amino acids of human AQP3 carboxyl terminus differ from those of rat by three residues, the anti-AQP3 antibody also recognized three bands in human normal ghosts (Fig. 1A, lane 7): a faint 25-kDa band and two broad 37-48-kDa and >70-kDa bands. An identical signal was observed in Co(a-b-) ghosts (Fig. 1A, lane 6). Similar to controls with rat membrane samples, no signal was detected in human red cells when the anti-AQP3 was first blocked by the immunizing peptide (Fig. 1A, lane 3). The anti-AQP3 antibody failed to recognize human AQP1 purified from red blood cells (Fig. 1A, lane 8), thus confirming the specificity of anti-AQP3 antibodies. Most importantly, our results indicate that AQP3 is present in normal and Colton-null human red cells.

View larger version (36K): [in this window] [in a new window]   Fig. 1.   Immunoblots of AQP3 (A) and AQP1 (B) in various tissues. Five µg of membrane proteins and 0.6 µg of pure AQP1 were loaded in individual lanes and were probed with either the affinity-purified anti-AQP3 at a concentration of 0.25 µg/ml (A) or the anti-AQP1 whole serum diluted 1:500 (B). A, lanes 1-3, the anti-AQP3 antiserum was preabsorbed with 0.2 mg of immunizing peptide; lanes 1 and 4, rat renal outer medulla; lanes 2 and 5, rat ghosts; lanes 3 and 7, normal human ghosts; lane 6, Colton-null ghosts; lane 8, purified human AQP1. B, lane 1, purified human AQP1; lane 2, Colton-null ghosts; lane 3, normal human ghosts.

Indirect Immunofluorescence-- In experiments on human red blood cells, the anti-AQP3 serum strongly stained the plasma membranes of both normal (Fig. 2a) and Colton-null cells (Figs. 2, c and d). No staining was observed when the antiserum was preincubated with the peptide used for immunization (Fig. 2b, normal cells). In contrast, the polyclonal antiserum against purified AQP1 stained the plasma membranes of normal human red blood cells (Fig. 2e) but not those of Colton-null cells (Fig. 2f). In rat kidney, the anti-AQP3 antibodies recognized the basolateral plasma membranes of collecting duct principal cells (Fig. 2g) but not those of adjacent intercalated cells (Fig. 2g, arrowheads). Although no staining was observed in other cell types of the rat kidney nephron, as previously reported (16), we also found staining of rat red blood cells (Fig. 2h). Altogether, these results confirm the presence of AQP3 in both human (normal and Colton-null) and rat red blood cell membranes.

View larger version (45K): [in this window] [in a new window]   Fig. 2.   Indirect immunofluorescence staining of red blood cells and rat kidney. On cryosections, both anti-AQP3 (a) and anti-AQP1 antiserum (e) stained the normal human red blood cell plasma membranes. At two different magnifications, Colton-null human red blood cells were also stained by anti-AQP3 antibodies (c and d) but not by anti-AQP1 antibodies (f). No staining was observed when the anti-AQP3 antiserum was preabsorbed with the peptide used for immunization (b, normal red blood cells shown). In rat kidney nephrons, the anti-AQP3 only stained the basolateral plasma membrane of collecting duct principal cells (g), whereas intercalated cells were unstained (arrowheads). In addition, rat red blood cells were also stained by the anti-AQP3 antibodies (h). a-f, bars = 5 µm; g and h, bar = 15 µm.

Water and Glycerol Permeabilities of Normal and Colton-null Red Cells-- Fig. 3A shows a typical time course of osmotic shrinkage induced by a 100 mosm/kg H₂O sucrose gradient, carried out with thawed normal and Colton-null red cells at 26 °C. Averaged P_f was (1.32 ± 0.09) × 10² cm/s and (2.16 ± 0.16) × 10³ cm/s (n = 6) for normal and Co(a-b-), respectively. As previously reported (5), the P_f values in Co(a-b-) cells were significantly lower than in normal cells.

View larger version (15K): [in this window] [in a new window]   Fig. 3.   Measurements of water (A) and glycerol (B) transport in human normal (thick line) and Colton-null (thin line) erythrocytes by stopped-flow light scattering. A, a suspension of erythrocytes at ~107 cells/ml was submitted to a 100 mosm/kg H2O sucrose gradient. The increase in 90° scattered light intensity, corresponding to water efflux, was recorded at 600 nm as a function of time. The curves were fitted as single exponentials and Pf were calculated from rate constants as described under "Experimental Procedures." B, the dilute suspension of erythrocytes was submitted to an inwardly directed 100 mM glycerol gradient. The biphasic curves show the osmotically induced cell shrinkage consecutive to initial water efflux, followed by concomitant glycerol influx and cell swelling. The rates of glycerol transport were estimated by single exponential fits on the second part of the curves.

1

1

6

6

View larger version (10K): [in this window] [in a new window]   Fig. 4.   Time courses of glycerol efflux from 100 mM 14C-glycerol-loaded resealed ghosts (filled squares, normal; open squares, Colton-null). Results are presented as the ratio of the apparent volume of 14C-glycerol remaining in the cells (Vt) over the initial volume (V0) as a function of time. Pgly were calculated as indicated under "Experimental Procedures."

Inhibition Studies-- The effects of transport inhibitors were tested on normal and Colton-null red cells using the stopped-flow technique. Glycerol effluxes were followed in isoosmotic conditions, to avoid initial water movements. Fig. 5 shows the effects of 0.2 mM DIDS, 0.1 or 0.5 mM phloretin, 0.1 mM CuSO₄, or 0.5 mM HgCl₂ on glycerol (Fig. 5A, normal; Fig. 5B, Co(a-b-)) and water (Fig. 5C, normal; Fig. 5D, Co(a-b-)) efflux rates. DIDS inhibited the glycerol outflow from normal and Co(a-b-) cells by approximately 45%. It also reduced the rate of water efflux of both cell types, with a greater efficiency in Co(a-b-) erythrocytes. A large inhibition of glycerol transport in normal and Co(a-b-) red cells was observed in the presence of 0.1 mM (80 and 67% inhibition, respectively) and 0.5 mM phloretin (93 and 73% inhibition, respectively). In addition, phloretin markedly inhibited water transport in Colton-null erythrocytes, at the 2 concentrations tested (~53% inhibition). CuSO₄ significantly reduced the glycerol permeability of normal (69% inhibition) and Co(a-b-) (33% inhibition) red cells, and also strongly reduced the water permeability of Colton-null erythrocytes (60% inhibition). HgCl₂ dramatically inhibited both glycerol and water transport in normal and Co(a-b-) erythrocytes. Percentages of inhibition of glycerol efflux were 75 and 72% for normal and Colton-null cells, respectively, and the water efflux was inhibited by 90% in normal and by 50% in Colton-null erythrocytes. This indicates the presence, in the latter, of a residual mercury-sensitive path for water. In conclusion, all the reagents inhibited glycerol transport in human normal and Colton-null red blood cells in a similar manner. In contrast, these molecules had differential effects on water transport in both cell genotypes. These data support the idea that the major path for water transport was disrupted in Colton-null cells while that for glycerol was not.

View larger version (26K): [in this window] [in a new window]   Fig. 5.   Inhibition of glycerol and water transport in normal (A and C) and Colton-null (B and D) erythrocytes. The cells were submitted in a stopped-flow apparatus to a 100 mM outwardly directed glycerol gradient (A and B) or to a 100 mosmol/kg H2O inwardly directed sucrose gradient (C and D). Inhibitors (0.2 mM DIDS, 0.1 and 0.5 mM phloretin, 0.1 mM CuSO4, 0.5 mM HgCl2) were added 10 min (or 5 min for HgCl2) before the stopped-flow assay. Results are expressed as percentages of the maximal (no inhibitor) glycerol and water efflux rate constants. Results are means ± S.E. of the indicated number of individual experiments (0.001 < p < 0.030).

	DISCUSSION

Top Abstract Introduction Procedures Results Discussion References

AQP1 is the major water channel of human red blood cells. However, as we showed here and as previously reported (5), a significant mercury-sensitive part of osmotic water permeability remains in AQP1-knocked out (Colton-null) red cells. The existence of additional, non-AQP1, protein pathways for water transport has been discussed by Toon and Solomon (26). Red blood cells also exhibit a high glycerol transport capacity, whose molecular bases have remained unknown. AQP3 is a mammalian aquaporin which, in addition to being moderately permeable to water, is highly permeable to glycerol and, to a lesser extent, to urea (11, 12, 14). In the present work, we provide evidence that AQP3 is expressed in normal and Colton-null red cells and likely mediates not only the residual water transport, but also most of glycerol transport in human erythrocytes.

Human AQP3 was detected by Western blotting in both normal and Colton-null red cells and exhibited a migration profile very similar to that of rat AQP3 in renal outer medulla and erythrocytes. The broad high molecular weight bands revealed in human tissues appeared slightly larger than those of rat tissues and may correspond to different glycosylation patterns. Indirect immunofluorescence confirmed the presence of AQP3 in the plasma membranes of normal and Colton-null cells. Interestingly, the presence in red blood cells of AQP1 and AQP3 constitutes a new example of aquaporin colocalization on the same plasma membrane. AQP3 and AQP4 were colocalized on the basolateral membrane of kidney collecting duct principal cells (16). AQP1 and AQP7, another aquaporin permeable to glycerol (13), are colocalized in the kidney proximal tubule brush-border membrane (27). AQP7 (13) and AQP8 (28) were both detected in rat testis.

To investigate the relevance of AQP3-mediated transport in erythrocytes, we characterized glycerol transport in both normal and Colton-null red cells. For this purpose, two distinct methods were used. The rates of red cell volume changes were estimated by light-scattering stopped-flow experiments on intact cells submitted to an osmotic glycerol gradient. This technique, which requires small amounts of material, was particularly useful for the study of Colton-null cells, which are not easily available. However, stopped-flow light scattering measurements do not allow accurate calculations of solute permeability coefficients, because of uncertainties in red cell surface area changes and refractive index effects (2, 29). Consequently, we considered in these experiments only the shrinking or swelling rate constants related to glycerol movements. In a second type of experiment, the glycerol permeability was directly measured by a rapid filtration method using one-step pink ghosts equilibrated in ¹⁴C-glycerol. In this case, the P_gly can be calculated from the rate of radioactive glycerol efflux and was 1.4 and 1.6 × 10⁶ cm/s in normal and Colton-null ghosts, respectively, i.e. in the range of already published values for human red blood cells (3, 25). These values are higher than the P_gly of bovine red cells, lacking the facilitated diffusion mechanism for glycerol transport (30). These values remain also 2 orders of magnitude lower than the P_f of Colton-null cells, indicating that, even in these cells, the rate of water flow was not limiting (29). Thus, both stopped-flow and tracer methods indicate a similar glycerol transport capacity in normal and AQP1-deficient red blood cells. This means that, in its native environment, AQP1 does not constitute a major pathway for glycerol. We previously reported that AQP1 itself had a small glycerol permeability (17). However, the glycerol transport capacity of AQP1, estimated after expression in Xenopus oocytes, is much lower than that of AQP3,² and may therefore not be detectable in the intact red cell membrane. The unitary P_gly of the bacterial GlpF is 2 × 10⁵ molecules/s (31). If we assume that the unitary P_gly of AQP3 is in the same range, the number of AQP3 copies that account for P_gly in erythrocytes would be around 15,000, i.e. at least ten times lower than the number of AQP1 in these cells (32). Conversely, AQP1 appears to have a higher unitary P_f than AQP3. This, with its higher abundance, explains why AQP1 dominates water transport in the red blood cell membrane.

A pharmacological characterization of glycerol and water transport was performed in normal and Colton-null red cells. All the reagents tested exerted similar inhibitory effects on glycerol transport in normal and Colton-null red cells, confirming the idea that a protein, distinct of AQP1, mediates most of glycerol transport in the red cell membrane. Some of the inhibitors tested can have unspecific membrane effects (33), and indeed, their use may be controversial. In particular, Ma et al. (12) did not observe any inhibitory effect of 0.2 mM DIDS on glycerol transport in human erythrocytes. In contrast, an inhibition of water transport by DIDS has been reported by Toon and Solomon (34) in human red blood cells. In our experiments, DIDS was inhibitory both on glycerol and water effluxes. Copper was found to affect the glycerol transport, as already described (3), and also the water transport in an unexplained fashion. Both mercury and phloretin have been shown to alter water transport in AQP3-injected oocytes (11, 14). HgCl₂ had the strongest inhibitory effects, suggesting that this reagent blocked the protein-mediated pathways for both water and glycerol transport. Phloretin markedly inhibited the residual water transport of the Colton-null cells, further supporting the hypothesis that AQP3 could be the phloretin-sensitive water channel of these AQP1-deficient cells.

Taken together, our results demonstrate the presence of functional AQP3 in the human red cell membrane, which can account for the glycerol permeability of these cells and the residual water permeability of AQP1-deficient erythrocytes. Our findings may help explain the lack of clinical defects in Colton-null patients (18). While AQP1 transports mostly water, the significance of AQP3 expression in red blood cells may reside in its ability to transport solutes. However, the role devoted to a glycerol transporter in erythrocytes is not elucidated. In contrast to testis or liver, where an important metabolism of glycerol has been described (13, 35), glycerol appears not to be metabolized by erythrocytes. The presence of a glycerol facilitator in red cells could make them less susceptible to osmotic stress upon local exposure to high glycerol concentrations. A similar hypothesis has been advanced by Macey (36), who suggested that the high urea permeability of human red blood cells could protect these cells when they reach the deeper regions of the renal medulla. AQP3 can also transport urea to some extent and yet to be discovered solutes that can be important for the red cell osmoregulation. Also, the possibility that a glycerol facilitator serves physiologically for another purpose should not be dismissed.