Human Rhesus-associated glycoprotein mediates facilitated transport of NH₃ into red blood cells

Abstract

Rhesus (Rh) antigens are carried by a membrane complex that includes Rh proteins (D and CcEe), Rh-associated glycoproteins (RhAG), and accessory chains (LW and CD47) associated by noncovalent bonds. In heterologous expression systems, RhAG and its kidney orthologs function as ammonium transporters. In red blood cells (RBCs), it is generally accepted that NH₃ permeates by membrane lipid diffusion. We have revisited these issues by studying RBC and ghosts from human and mouse genetic variants with defects of proteins that comprise the Rh complex. In both normal and mutant cells, stopped-flow analyses of intracellular pH changes in the presence of inwardly directed methylammonium (CH₃NH⁺₃+CH₃NH₂) or ammonium (NH⁺₄+NH₃) gradients showed a rapid alkalinization phase. Cells from human and mouse variants exhibited a decrease in their kinetic rate constants that was strictly correlated to the degree of reduction of their RhAG/Rhag expression level. Rate constants were not affected by a reduction of Rh, CD47, or LW. CH₃NH₂/NH₃ transport was characterized by (i) a sensitivity to mercurials that is reversible by 2-mercaptoethanol and (ii) a reduction of alkalinization rate constants after bromelain digestion, which cleaves RhAG. The results show that RhAG facilitates CH₃NH₂/NH₃ movement across the RBC membrane and represents a potential example of a gas channel in mammalian cells. In RBCs, RhAG may transport NH₃ to detoxifying organs, like kidney and liver, and together with nonerythroid tissue orthologs may contribute to the regulation of the systemic acid–base balance.

Rhesus (Rh) is a clinically significant blood group system in transfusion medicine (1). Rh blood group antigens are defined by a complex association of membrane polypeptides that includes the nonglycosylated Rh proteins (carriers of RhD and RhCcEe antigens) and Rh-associated glycoprotein (RhAG), which is strictly required for cell-surface expression of Rh antigens (2, 3). In red blood cells (RBCs), the core of the Rh complex may be a tetramer composed of two Rh and two RhAG subunits, to which accessory chains (CD47, LW, and GPB) are associated by noncovalent linkages (2, 4).

The Rh_null phenotype is an inherited condition in which various Rh antigen deficiencies result in a clinical syndrome characterized by a hemolytic anemia of varying severity, with abnormalities of the red cell shape (stomato-spherocytosis), cation transport, and membrane phospholipid organization (2, 5). Rh_null disease is caused by several different mutations that occur in either the RHAG or RH loci on chromosomes 6p12-p21 and 1p34-p36, respectively. The Rh complex is missing or severely reduced in RBCs from Rh_null individuals without alteration of the genes encoding the accessory chains (2, 5). Because of a variable expressivity, some mutations of the RHAG gene result in the total lack of RhAG (and Rh protein), defining the Rh_null of the regulator type, but others result in weak RhAG (and Rh protein) levels, defining the Rh_mod phenotype (5). Mutations of the RH gene resulting in the total lack of Rh protein and only a reduced expression of RhAG (20% of normal) define Rh_null as the amorph type. Accordingly, primary defects of either RhAG or Rh protein result in defective cell surface expression and/or transport of the whole Rh complex.

The murine mutations normoblastosis (Ank1^nb) and spherocytosis (Spna1^sph) disrupt the RBC spectrin-based cytoskeleton through deficiencies of ankyrin and α-spectrin, respectively. In addition to the well described band 3-ankyrin–protein 4.2 and glycophorin C–protein 4.1–p55 complexes (6), the Rh complex represents a major interaction site between the membrane lipid bilayer and the spectrin-based cytoskeleton. Indeed, recent studies of the erythroid ankyrin-deficient normoblastosis mice and analysis with the yeast two-hybrid system have shown that Ank1 may interact directly with the C-terminal cytoplasmic domains of Rh protein and RhAG (7). Interestingly, in humans this interaction was abolished by a mutation found in an Rh_null patient (7).

The first indication of the biological role of Rh proteins came from the discovery of a significant homology, particularly to RhAG, with the methylammonium (MA) permease/ammonium transporter (Mep/Amt) superfamily of ammonium transporters present from bacteria to yeast but absent in vertebrates (8). This discovery was followed by the characterization of two nonerythroid orthologs, RhBG and RhCG (9, 10), and the identification of paralogs in a variety of living species, thereby suggesting the conservation of some critical function (reviewed in ref. 11). Moreover, when expressed in Mep/Amt-deficient yeast, RhAG and RhCG function as bidirectional ammonium transporters (12). Recently, functional studies of Rh glycoproteins in Xenopus laevis oocytes were performed but yielded discordant results. Some studies suggested that RhAG-mediated uptake of MA in oocytes was an electroneutral NH₄⁺/H⁺ exchange (13), whereas electrophysiological studies in voltage-clamped RhBG- and RhCG-expressing oocytes suggested that transport is (14, 15) or is not (16) electrogenic. It was also speculated that Rh and Mep/Amt proteins might be facilitators for NH₃ gas transport (17).

In physiological systems, such as RBCs, it is generally accepted that NH₃ permeates through membrane lipid diffusion (18–20). The purpose of the present study is to revisit these findings by studying the ammonium transport in RBCs from human and mouse genetic variants with various defects of proteins that comprise or interact with the Rh complex. Accordingly, stopped-flow analysis of kinetic rate constants of intracellular pH (pH_i) changes of resealed ghosts in the presence of MA or ammonium gradients as well as mercurial inhibition and protease cleavage studies were performed. The results indicate that RhAG facilitates CH₃NH₂/NH₃ movement across the red cell membrane.

Materials and Methods

Blood Samples. Four Rh_null patients of the regulator type [P1(TB), P2(YT), P3(AL), and P4(AC)], one Rh_mod patient [P5(CB)], and one Rh_null patient of the amorph type [P6(DR)] have been described (2, 5). Control RBCs from RhD-positive, RhD-negative, and LW_null individuals [P7(Big) and P8(Mil)] were provided by the Centre de Référence pour les Groupes Sanguins (Paris). RBCs from two individuals [P9(Sar) and P10(Fer)] of the AQP1_null phenotype lacking AQP1 (also called CO_null because AQP1 carries Colton antigens) have been described (21, 22). A blood sample from a 4.2-deficient patient [P11(Rem)] was provided by O. Agulles (Etablissement Français du Sang, Nancy, France). RBCs from mouse models of hereditary hemolytic anemia, normoblastosis (nb/nb) (ankyrin-deficient) and spherocytosis (sph/sph) (spectrin-deficient), and from wild-type mice (WBB6F1, C57BL/6J) were from The Jackson Laboratory. Before analysis, cryopreserved or fresh RBCs were washed twice with PBS.

Flow Cytometry Analysis. Indirect immunofluorescence of human and mouse RBCs for cell-surface or intracellular epitope analysis was performed with a FACScalibur flow cytometer (Pharmingen) by using the following monoclonal antibodies: mouse anti-human Rh protein BRIC69 (D. Anstee, International Blood Group Reference Laboratory, Bristol, U.K.), mouse anti-human CD47 (clone 3E12, Bioatlantique, Nantes, France), rat antimouse CD47 (clone miap 301, Pharmingen), mouse anti-LW BS46 (H. H. Sonneborn, Offenbach, Germany), human anti-Band 3 HIRO-58/Di^b (M. Uchikawa, Japanese Red Cross Central Blood Center, Tokyo); mouse anti-AQP1 (clone 1/A5F6, Serotec, Cergy, France), and rabbit polyclonal antibodies to human Rh protein (MPC8) and mouse Rhag (mCTRhAG). The antibodies were used at saturating conditions as described in refs. 3, 7, and 23. In control studies, a human anti-Colton a antigen (Centre de Référence pour les Groupes Sanguins, Paris) and a rabbit polyclonal antibody to AQP3 protein were used with intact and fixed permeabilized RBCs as described in refs. 24 and 25.

RBC Ghost Preparation. All preparation steps, except resealing (37°C), were carried out at 4°C, and assays were performed the same day (26). One volume of blood was washed three times in resealing buffer (140 mM NaCl/2.5 mM KCl/10 mM Hepes/NaOH, pH 7.4) and resuspended in 80 volumes of hypotonic lysis buffer (7 mM NaCl/10 mM Hepes/NaOH, pH 7.4) for 40 min on ice folllowed by resealing in resealing buffer containing 1 mM MgSO₄ and 0.15 mM pyranine, the fluorescent pH-sensitive dye (1-hydroxypyrene-3,6,8-trisulfonic acid, Sigma–Aldrich) at 37°C for 1 hr. After three washes in resealing buffer, ghosts were equilibrated in buffer A (130 mM NaCl/5 mM KCl/10 mM Hepes/NaOH, pH 7.0) and kept on ice before assays.

pH_i Determination by Stopped-Flow Analysis. Ammonium or MA (ammonium and MA refer to the sum of protonated and unprotonated forms) transport was measured in isoosmotic conditions. Membrane ghosts equilibrated in buffer A containing 5 mM p-xylene-bis(N-pyridinium bromide) (Sigma), a fluorescence quencher of external pyranine, were mixed with an equal volume of buffer B (65 mM NaCl/5 mM KCl/65 mM NH₄Cl or MA chloride/10 mM Hepes/NaOH, pH 7.0), generating an inwardly directed 32.5 milliequivalent gradient. In preliminary experiments, pH_i changes were measured with a stopped-flow instrument (SFM3, Biologic, Grenoble, France) that was modified by addition of the titrator device (Biologic), which consists of replacing the 30-μl cuvette (see below) with a thermostated (15°C) and vortex-mixed cuvette (4 ml), allowing the use of the equipment as an ordinary spectrofluorometer. The excitation wavelength was 460 nm, and the emitted light was filtered with a 520-nm cut-on filter. The pH-dependent fluorescence changes were followed and analyzed, because a fluorescence increase corresponded to a pH elevation and a fluorescence decrease corresponded to a pH reduction (27). Over the pH range used (7.0–7.8), the relative fluorescence of the dye was proportional to pH, as determined by titration on ghosts incubated in 2 ml of buffer A containing 5mM p-xylene-bis(N-pyridinium bromide) and 5 μM nigericin (Sigma) and submitted to pH change by step additions of 2 μl of KOH (1 M). Subsequently, ghost alkalinization and acidification kinetics were performed in the stopped-flow mode (30-μl FC-15 cuvette; dead time, 7.8 ms). Preliminary experiments showed that ammonium transport was similar for ghosts prepared from fresh and cryopreserved RBCs. Data from five to eight time courses were averaged, and the alkalinization phase was fitted to a single exponential function by using the Simplex procedure of the biokine software (Biologic) to calculate kinetic rate constants, k, ins^–1. In some experiments, apparent permeabilities (P′) were determined without unstirred layers correction according to Priver et al. (28) by using the following equation:

where ΔpH_max is the maximum pH change value, BC is the buffer capacity, V/SA is the volume-to-surface-area ratio (in cm) of ghosts, ΔC is the MA/ammonium gradient (mM), and t = 1/k (s), which is the reciprocal of the exponential kinetic rate constant k.

Membrane ghosts are crenated spheres in isotonic solution (29). Determination of ghost diameter by light and confocal microscopy confirmed their spheric shape and that ghosts are smaller than intact RBCs. When ghosts from normal and human variant RBCs were compared, we found no significant difference (4.95 ± 0.2 versus 5.07 ± 0.38 μm). The volume calculated from the measured mean diameter corresponds to a sphere of ≈7 × 10^–14 liter, as expected (30). The buffer capacity of ghosts in buffer A containing 5 mM p-xylene-bis(N-pyridinium bromide) was measured by a single addition of 20 mM acetate as in ref. 27, but with the stopped-flow mode. The mean value was within the same range for normal and variant ghosts (82.57 ± 9.83 versus 82.18 ± 3.71 mM per pH unit). In some experiments, ghosts were incubated for 5 min at 15°C in the presence of 0.1 mM HgCl₂ before the application of the MA gradient. In other experiments, ghosts were prepared from protease-treated RBCs that were obtained by incubating washed RBCs (10% hematocrit) for 30 min at 37°C in the presence or absence of 3 mg/ml bromelain (product no. B4882, Sigma) in PBS. Then, cysteine proteinase inhibitor transepoxysuccinyl-l-leucylamido-(-4-guanidino)butane (E-64, Calbiochem) was added to a 10 μM final concentration, and RBCs were washed three times at 4°C in PBS containing 10 μM E64.

Water and Glycerol Permeabilities and Western Blot Analysis. Water and glycerol osmotic permeabilities were determined by RBC volume change analysis in the stopped-flow spectrophotometer by measuring 90° scattered-light intensity with 600-nm excitation light, as described in ref. 31. Western blot analyses of native and protease-treated ghosts were performed with a murine monoclonal anti-RhAG (clone LA 18-18, gift from A. van dem Borne, University of Amsterdam, Amsterdam), the reactivity of which depends on the presence of the carbohydrate moiety, and with rabbit antisera against the C-ter of human RhAG (anti-C-ter) and MPC8 (12, 23).

Results and Discussion

CH₃NH₂/NH₃ Transport in Normal RBC Ghosts. Studies of ammonium permeability by ¹⁴N and ¹⁵N saturation transfer NMR spectroscopy have demonstrated that NH₃ very rapidly enters RBCs by passive lipid diffusion (20). To further investigate this question, we have studied RBC and ghosts from human and mouse genetic variants with defects of proteins that comprise the Rh complex. Because ammonium is a weak base (pK_a = 9.56 at 15°C), cell uptake results in pH_i changes, either alkalinization or acidification, depending on whether the neutral form (NH₃) or charged form (NH⁺₄) is taken up. Because of the high buffer capacity of RBCs, pH_i changes were determined on membrane ghosts resealed on a 10 mM Hepes buffer (pH 7.0) containing pyranine as a fluorescent pH-sensitive dye. Because all ghosts preparations, including those from controls from human variant RBCs analyzed below, exhibited a similar size and buffer capacity (see Materials and Methods), the kinetic rate constants of CH₃NH₂/NH₃ influx measured in transport studies strictly reflected differences in permeabilities to CH₃NH₂/NH₃.

By using ghosts from control RhD-positive RBCs with normal Rh, RhAG, LW, and CD47 expression (Table 1) submitted to a 32.5 mM inwardly directed MA or ammonium gradient (pH 7.0) in isoosmotic conditions, we found that the fluorescence intensity curve exhibited a biphasic response, consisting first of a rapid phase of alkalinization (ΔpH ≈ 0.5 unit), followed by a slow acidification phase to return to external pH (pH₀) of the incubation medium (Fig. 1A). The alkalinization can be accounted for by the uptake of CH₃NH₂ or NH₃, thereby trapping protons within the cells until the equilibrium values of these neutral forms were reached (20). The kinetic rate constant of alkalinization of ghosts from RhD-positive cells (Fig. 1B) was 0.95 ± 0.1 s^–1 for CH₃NH₂ and 4.95 ± 0.28 s^–1 for NH₃, and there was no significant difference between ghosts from RhD-positive and RhD-negative RBCs. RhD-positive and RhD-negative RBCs exhibiting normal levels of RhAG, LW, CD47, and RhCE proteins but differing by the presence or absence of the RhD protein, suggests that, by itself, the RhD protein has no significant effect on CH₃NH₂/NH₃ transport. From these data, the calculated apparent permeabilities (P′) of RhD-positive ghosts for CH₃NH₂ and NH₃ at 15°C are 0.88 ± 0.09 10^–4 and 4.46 ± 0.25 10^–4 cm·s^–1, respectively, without the correction of the unstirred layers (Table 2). A value in the same range has been published for NH₃ permeability of RBCs at 21°C (20). Although most experiments were done with a 32.5 mM gradient, some were also performed with 2.5–40 mM MA (pH 7.0) gradients, and we found a linear relationship with fluorescence variation (data not shown).

Table 1. Antigen and protein expression of human RBCs.

	Control	Variants
Membrane protein	RhD-positive (n = 5)	Rhnull regulator of P1-P4	Rhmod of P5	Rhnull amorph of P6	LWnull of P7/P8	4.2null* of P11	AQP1null of P9/P10
Rh	126 ± 10	0	7.6	0.1	90/100	112	131/136
RhAG	81 ± 7	<0.2	8.9	17.6	65/75	108	85/95
CD47	17 ± 1	1.9 ± 0.2	3	2.3	13/15	4.1	17/16
LW	3 ± 1	0	0	0	0/0	4	n.t.
Band 3	2,129 ± 60	2,076 ± 46	1,987	2,139	1,927/1,762	1,853	2,029
AQP1	298 ± 15	270 ± 24	243	318	350/380	297	<0.2
AQP3	172 ± 10	151 ± 26	152	169	133/157	199	119

Values indicate the copy number of membrane protein per red cell (×10^-3), except for Band 3, AQP1 (anti-Colton a antigen), and AQP3, which were measured as mean fluorescence intensity. n.t., not tested.

^*Only a single patient was tested.

Fig. 1.

Time course of fluorescence and pH_i changes in ghosts. (A) Ghosts from RhD-positive (Rh+) and Rh_null regulator (P1) were submitted to a 32.5 mM MA inwardly directed gradient (pH 7.0) at 15°C, and pH_i changes were followed by fluorescence change of a pH-sensitive dye in the stopped-flow spectrofluorometer by using the titrator device (see Materials and Methods). The response was biphasic. Experiments were repeated three times with different Rh_null of the regulator type (P2, P3, and P4) and yielded the same results. (Insert) Over the pH range used (7.0–7.8), the relative fluorescence of the dye was proportional to pH. (B) Expanded scale of the rapid alkalinization phase.

Table 2. Rate constants and permeability to CH₃NH₂/NH₃ of human RBC ghosts.

	Controls		Variants
Substrate	RhD-positive (n = 3)	RhD-negative (n = 3)	Rhnull regulator of P1-P4	Rhmod of P5	Rhnull amorph of P6	LWnull of P7/P8	4.2null of P11
CH3NH2
k, s-1	0.95 ± 0.1	0.81 ± 0.03	0.14 ± 0.03*	0.26	0.38	0.8/0.8	0.74
P′, cm·s-1	0.88 ± 0.09	0.71 ± 0.03	0.12 ± 0.03*	0.23	0.36	0.8/0.7	0.68
NH3
k, s-1	4.95 ± 0.28	5.4 ± 0.28	2.48 ± 0.09*	2.6	2.9	3.9/4.1	4.56
P′, cm·s-1	4.46 ± 0.25	3.91 ± 0.21	2.54 ± 0.08*	2.63	2.53	4.1/4.3	4.11

Figures indicate alkalinization rate constants (k) and permeability (P′ × 10^-4) to CH₃NH₂/NH₃ of ghosts submitted to 32.5 mM MA or ammonium gradient.

^*, P < 0.001 versus RhD-positive and RhD-negative.

The phase of slow acidification after the alkalinization step corresponds to a return to pH₀, which is a complex process, because the ghost membrane carries several transporters, such as Band 3 and specific or nonspecific ammonium/MA transporter(s), including the Na⁺/H⁺ antiport, that could participate in pH regulation. This process was not investigated further. Because the acidification phase is 25 or 45 times slower than the alkalinization phase (as estimated by t_1/2 values for CH₃NH₂ and NH₃, respectively), we assume that it does not significantly interfere with the calculation of first-phase rate constants.

CH₃NH₂/NH₃ Transport of Ghosts from Human Variants. We next determined pH_i changes of ghosts from Rh_null cells of the regulatory type (no Rh/RhAG/LW; low CD47) (Table 1), and we found also a biphasic response (Fig. 1). However, the rate constants of alkalinization were much slower (0.14 ± 0.03 s^–1) compared with control RBCs, because of a 6-fold reduction of apparent permeability to CH₃NH₂ (Table 2). Rh_mod and Rh_null (amorph) ghosts that express only 10% and 20% of RhAG, respectively (Table 1), exhibited intermediate values (Fig. 2A and Table 2), suggesting that CH₃NH₂ transport was correlated to RhAG expression level. pH_i changes were faster in the presence of the NH₄Cl gradient at pH 7.0 (Fig. 2B), but there was still an ≈2-fold reduction of alkalinization rate constants and P′ values for Rh_null regulators compared with controls, again with Rh_mod and Rh_null (amorph) exhibiting intermediate values (Table 2). The fact that the rates change in the same-sized ghosts with or without Rh proteins indicates that the unstirred layers do not play a significant role. Therefore, although a “basal” CH₃NH₂/NH₃ entry occurs by lipid diffusion in Rh_null cells, our data strongly suggest that RhAG may accelerate this process. As shown in Table 2, we also found a normal alkalinization rate constant of ghosts from LW_null RBCs (normal Rh/RhAG/CD47; no LW) (Table 1), indicating no role of LW in CH₃NH₂/NH₃ transport. In addition, the alkalinization rate constant of 4.2_null ghosts, which have a severe reduction of CD47 but normal levels of Rh, RhAG, and LW (Table 1) (23, 32), was only slightly lower in the single patient (P11) investigated (0.74 s^–1 for CH₃NH₂), indicating that CD47 does not play a significant role. We found, however, that AQP1_null (CO_null) ghosts with a normal content of RhAG (Table 1) showed a small but significant (P < 0.01) reduction of CH₃NH₂ transport (k = 0.51 s^–1 for P9 and 0.59 s^–1 for P10 versus k = 0.95 ± 0.1 s^–1 for control RBCs), suggesting that AQP1 might partly contribute to NH₃ transport, as suggested in ref. 33.

Fig. 2.

CH₃NH₂/NH₃ transport in ghosts from human and mouse variants. (A–G) Time course of fluorescence change of human (A, B, and E–G) and murine (C and D) ghosts subjected to a 32.5 mM MA (A, C, and E–G) or ammonium (B and D) inwardly directed gradient (pH 7.0) at 15°C were followed by stopped-flow analysis (see Materials and Methods). Only the rapid initial alkalinization phase is shown. (A and B) Human samples: RhD-positive (Rh+), Rh_null regulator (P1), Rh_mod (P5), Rh_null amorph (P6), 4.2_null (P11), and AQP1_null (P9). (C and D) Mouse samples: wild type (wt), nb/nb, and sph/sph.(E and F) Thiol sensitivity. Ghosts from RBCs were preincubated with 0.1 mM HgCl₂, followed by a treatment with or without 5 mM 2-mercaptoethanol (β-ME). (E) Effect on RhD-negative (Rh–) and Rh_null (P1) ghosts and reversibility by 2-mercaptoethanol. (F) Effect on AQP1_null (P9) ghosts. (G and H) Effect of proteolytic degradation. (G) MA transport was reduced in Rh-negative but not Rh_null (P1) ghosts prepared from RBCs digested with bromelain (Br). (H) Western blot analysis of ghosts prepared from untreated (U) or bromelain-digested (Br) Rh-negative RBCs. Blots were immunostained with a rabbit serum against human RhAG (anti-C-ter), the mouse monoclonal LA18-18 (anti-human RhAG), and MPC8 (anti-human Rh protein). The arrowhead indicates digested RhAG product.

The differences of alkalinization rate constants and permeabilities seen between CH₃NH₂ and NH₃ with Rh control samples (≈5-fold) might reflect the different molecular structure of these substrates (Table 2). These differences are even larger with Rh_null samples (≈20-fold), in which the transport is most likely limited to lipid diffusion. Interestingly, 4.2_null ghosts, which derived from RBCs with the same morphological characteristics as Rh_null cells (34), have a normal RhAG protein content (Table 1) and exhibit a transport activity virtually identical to that of controls (Fig. 2 A and Table 2). Moreover, Rh_null of the regulator and amorph types also have the same morphological characteristics (2), but they exhibit different rate constant values (Table 2) that follow their difference in RhAG content (Table 1). As a further test of whether cell volume variation affects the comparative analysis of rate constants, RBC permeabilities to water and glycerol under hyperosmotic conditions (100 and 150 mM sucrose and glycerol gradients, respectively) were determined. No significant differences were observed between control and Rh-variant cells (Table 3A), except for the water permeability of AQP1_null cells that, as expected (21), was severely reduced (data not shown).

Table 3. Volume change rate constants of human and mouse RBCs.

	Human RBCs					Mouse RBCs
Solute	RhD-positive (n = 3)	RhD-negative (n = 1)	Rhnull regulator	Rhmod	Rhnull amorph	WBB6F1 (n = 6)	C57BL/6J (n = 6)	nb/nb (n = 5)	sph/sph (n = 7)
Sucrose	3.29 ± 0.2	3.14	3.67 ± 0.46 (P1-P4)	3.29 (P5)	2.73 (P6)	7.73 ± 0.27	6.67 ± 0.27	5.15 ± 0.23*	3.51 ± 0.09*
Glycerol	0.14 ± 0.016	n.t.	0.14 ± 0.01 (P1 and P2)	n.t.	n.t.	0.091 ± 0.01	0.062 ± 0.002	0.1 ± 0.007	0.07 ± 0.003

Values indicate volume change rate constants of RBCs submitted to 100 mM sucrose or 150 mM glycerol gradients. Rate constants (in s^-1) were calculated from shrinkage for water movement and swelling for glycerol uptake. n.t., not tested. *, P < 0.001.

Our results show that Rh_null cells exhibit reduced permeabilities to CH₃NH₂ and NH₃ that may be related to their protein defects. However, it is known that the membrane lipid composition and the leaflet bilayer asymmetry may affect NH₃ permeability (35, 36). The composition and content of red cell lipids and phospholipids in Rh_null RBCs has been examined in only a few cases and was found to be normal (37, 38). However, there is an abnormal membrane distribution of phosphatidylethanolamine and an enhanced transbilayer movement of phosphatidylcholine (39). The membrane fluidity of Rh_null cells is poorly documented but has been found to be either normal (40, 41) or slightly increased (42), suggesting that such small variations cannot account for the large reduction of alkalinization rate constants of Rh_null ghosts (Table 2).

Our results showing that RhAG accelerates NH₃ import contradict the conclusion of a recent report based on a method inappropriate to measure fast kinetics (43).

CH₃NH₂/NH₃ Transport of Ghosts from Mutant Mice. Further studies were carried out by using mouse models of hereditary hemolytic anemia. These mice have primary defects in Ank1 (nb/nb) (44) or α-spectrin (sph/sph) (45). RBCs from nb/nb mice exhibit a sharp reduction of Rh and Rhag proteins but normal levels of CD47 (7), which is also true for RBCs from sph/sph mice (Table 4). Accordingly, we hypothesized that, similarly to Rh_null cells, these cells would exhibit reduced CH₃NH₂ or NH₃ transport activity. Indeed, the alkalinization rate constants in the presence of a MA gradient (pH 7.0) were high in wild type (1.55 s^–1) and much lower, although correlated to their respective level of Rhag protein, in nb/nb (0.38 s^–1) and sph/sph (0.66 s^–1) mice (Fig. 2C and Table 4). The rate constants were much faster with an ammonium gradient at pH 7.0 (Fig. 2D), but there was still a difference between wild type (12.8 s^–1) and mutant mice (7.6 and 5.5 s^–1). These results support the view that the Rhag protein plays a role in CH₃NH₂/NH₃ transport of mouse RBCs. In control experiments, although mouse RBCs lack AQP3 (25, 46), we found no significant differences in the passive lipid diffusion of glycerol. However, the water permeability of mutant mice RBCs was reduced compared with wild type (Table 3) but was correlated to an apparent lower level of AQP1 (Table 4).

Table 4. Antigen and protein expression of mouse RBCs.

	Controls		Variants
Membrane component	WBB6F1 (n = 3)	C57BL/6J (n = 3)	nb/nb (n = 3)	sph/sph (n = 3)
Rh	505 ± 15	n.t.	31 ± 1	181 ± 4
RhAG	272 ± 4	264 ± 18	7 ± 2	117 ± 5
CD47	104 ± 2	103 ± 2	104 ± 5	111 ± 2
Band 3	2,200 ± 30	2,139 ± 33	773 ± 40	1,088 ± 8
AQP1*	504 ± 31	519 ± 17	356 ± 20	321 ± 12
AQP3	24 ± 1	30 ± 1	25 ± 1	30 ± 2

Values indicate expression level of membrane components measured as mean fluorescence intensity. n.t., not tested.

^*Monoclonal antibody 1/A5F6.

CH₃NH₂/NH₃ Transport Function Is Mediated by a Protein-Dependent Pathway. Altogether, the kinetic studies with human and mouse variants indicate either more rapid diffusion of CH₃NH₂/NH₃ across membrane lipids in normal ghosts or accelerated movement mediated by the presence of RhAG. To provide further evidence for protein-mediated transport, we examined the effect of HgCl₂, which is known to interact with thiol groups of proteins on the alkalinization rate constants of ghosts from control RBCs (either RhD-positive or RhD-negative) and from RBCs of the Rh_null regulator type. Preincubation of RBCs with 1 mM HgCl₂ severely reduced the rate constants of RhD-negative ghosts (and RhD-positive; data not shown) to the value typical of untreated Rh_null cells, and this effect was reversible with 5 mM 2-mercaptoethanol (Fig. 2E). Moreover, pretreatment of Rh_null cells with HgCl₂ did not further reduce their transport activity (data not shown), whereas AQP1_null ghosts (no AQP1; normal RhAG) (Table 1) behaved as control cells before and after incubation with HgCl₂ (and 2-mercaptoethanol treatment) (Fig. 2F). As additional proof that permeability to CH₃NH₂/NH₃ is facilitated by a membrane protein, transport experiments were performed with ghosts prepared from protease-digested RBCs. Because bromelain cleaves RhAG and RhD, but not the RhCE protein (47), the experiments were carried out with RhD-negative cells. Ghosts from bromelain-digested RBCs exhibited a significant reduction of the alkalinization rate constant compared with the untreated control (Fig. 2G), whereas protease digestion of Rh_null (regulator) cells had no effect. Western blot analysis confirmed that RhAG but not the Rh protein of ghosts prepared from enzyme-treated RBCs was cleaved by bromelain (Fig. 2H).

Altogether, our results favor a critical role of the RhAG protein in CH₃NH₂/NH₃ transport. Moreover, the results showing that RhAG facilitates CH₃NH₂/NH₃ movement across the red-cell membrane point to RhAG as a potential example of a gas channel in mammalian cells. There are ≈2 × 10⁵ copies of RhAG per RBC and presumably 2 RhAG molecules per Rh complex (2), which leaves us with ≈10⁵ copies of functional Rh complexes. From these data and transport studies (Table 2), we tentatively calculate that for a 1 M ammonium gradient (pH 7.0), there is an approximate conductance of 2 × 10⁶ NH₃ molecules per second per Rh complex.

Our results support previous speculation (17) that Rh proteins may facilitate diffusion of NH₃. However, it was also proposed recently that the Rh1 protein paralog of the green alga Chlamydomonas reinhardtii might be regarded as a gas channel for CO₂, because mutants lacking this protein grow only very slowly when cultured in high CO₂ conditions (48). That Rh proteins might transport NH₃ and/or CO₂ is a possibility that should be further explored (32). Although still a matter of debate (49, 50), current studies suggest that the water channel AQP1 may represent another example of protein permeable to gases, because it also substantially increases the permeability of the membrane to CO₂ (50, 51) and NH₃ (33). Moreover, NtAQP1-related CO₂ permeability in plants plays a physiological role in photosynthesis and stomatal opening, particularly under low CO₂ gradient conditions (52).

Most interestingly, after this article was submitted, Khademi et al. (53,53) resolved the crystallographic structure of the bacterial NH₃ transport channel AmtB and showed by reconstitution into vesicles that AmtB conducts uncharged NH₃, which is fully consistent with our present studies on the RhAG protein.

In RBCs, RhAG may act to carry NH₃ to detoxifying organs, like kidney and liver (12). However, the transport function with regard to the substrate transported may be different according to which cells or tissues are examined and which Rh protein homologue (RhAG, RhBG, or RhCG) is present. These findings should stimulate further investigation of the physiological role of RhAG orthologs in nonerythroid tissues to understand their role in the regulation of systemic acid–base balance. These proteins might also play a role as ammonium/NH₃ sensors in the regulation of various cellular functions, such as ion transports, in response to variations in extracellular ammonium level (54–56).