Relative CO₂/NH₃ selectivities of AQP1, AQP4, AQP5, AmtB, and RhAG

Abstract

The water channel aquaporin 1 (AQP1) and certain Rh-family members are permeable to CO₂ and NH₃. Here, we use changes in surface pH (pH_S) to assess relative CO₂ vs. NH₃ permeability of Xenopus oocytes expressing members of the AQP or Rh family. Exposed to CO₂ or NH₃, AQP1 oocytes exhibit a greater maximal magnitude of pH_S change (ΔpH_S) compared with day-matched controls injected with H₂O or with RNA encoding SGLT1, NKCC2, or PepT1. With CO₂, AQP1 oocytes also have faster time constants for pH_S relaxation (τ_pHs). Thus, AQP1, but not the other proteins, conduct CO₂ and NH₃. Oocytes expressing rat AQP4, rat AQP5, human RhAG, or the bacterial Rh homolog AmtB also exhibit greater ΔpH_S(CO₂) and faster τ_pHs compared with controls. Oocytes expressing AmtB and RhAG, but not AQP4 or AQP5, exhibit greater ΔpH_S(NH₃) values. Only AQPs exhibited significant osmotic water permeability (P_f). We computed channel-dependent (*) ΔpH_S or P_f by subtracting values for H₂O oocytes from those of channel-expressing oocytes. For the ratio ΔpH_S(CO₂)*/P_f*, the sequence was AQP5 > AQP1 ≅ AQP4. For ΔpH_S(CO₂)*/ΔpH_S(NH₃)*, the sequence was AQP4 ≅ AQP5 > AQP1 > AmtB > RhAG. Thus, each channel exhibits a characteristic ratio for indices of CO₂ vs. NH₃ permeability, demonstrating that, like ion channels, gas channels can exhibit selectivity.

Gas transport through membranes is of fundamental importance for nutritive transport, photosynthesis, oxidative metabolism, and signaling. For most of the past century, we assumed that gas molecules cross biological membranes merely by diffusing through the lipid phase. This dogma was challenged by 2 observations: (i) Apical membranes of gastric-gland cells have no demonstrable permeability to CO₂ or NH₃ (1). (ii) Heterologous expression of the water channel aquaporin 1 (AQP1) increases the CO₂ permeability of Xenopus oocytes (2). Cooper and Boron (3) and Prasad et al. (4) confirmed and extended this observation. Uehlein (5) showed that an AQP plays a physiological role by enhancing CO₂ uptake by plants. Endeward et al. (6) demonstrated that AQP1 accounts for ≈60% of the CO₂ permeability of human red blood cells (RBCs). Molecular dynamics simulations suggest that CO₂ can pass through the 4 aquapores of an AQP1 tetramer (7) and especially through the central pore between the 4 monomers (7). Additional data indicate that AQP1 is permeable to nitric oxide (8), and that—when expressed in Xenopus oocytes (9, 10) or when reconstituted into planar lipid bilayers (11)—AQP1, AQP3, AQP8, AQP9, and the plant aquaporin TIP2;1 are all permeable to NH₃.

The AmtB/MEP/Rh proteins represent a second family of gas channels (12–15). Early work showed that AmtB and MEP transport NH₃ or NH₄⁺, thereby playing a nutritive role in archaea, bacteria, and fungi (16, 17). The crystal structures of the bacterial AmtB (18–20) and Rh50 (21) and the fungal Amt-1 (22) are consistent with the idea that NH₃ passes through a pore in each monomer of the homotrimer. Indeed, reconstituted AmtB conducts NH₃ (14), and RhAG is necessary for NH₃ transport in mammalian RBCs (23). Soupene et al. found that Rh1 deficiency impairs the growth of the green alga C. reinhardtii (24) and suggested that Rh1 plays a role in CO₂ transport. In RBCs, RhAG accounts for ≈50% of CO₂ transport (25).

In 2006, we introduced an approach (6) to assess CO₂ transport by pushing a blunt microelectrode against the surface of an oocyte, while monitoring surface pH (pH_S). Introducing extracellular CO₂ causes a transient pH_S increase, the maximum magnitude of which (ΔpH_S) is an index of maximal CO₂ influx. Earlier, De Hemptinne and Huguenin (26) had observed such a CO₂-induced transient while monitoring extracellular pH (pH_o) of rat soleus muscle. Moreover, Chesler (27) had found that exposing lamprey neurons to NH₃ causes a transient decreases in pH_o. Here, we exploit CO₂- and NH₃-induced pH_S transients to study the CO₂ vs. NH₃ permeability of 4 channels abundantly expressed in cells that mediate high rates of gas transport: human AQP1 (RBCs; ref. 28), the M23 variant of rat AQP4 (astrocytic endfeet at the blood–brain barrier, ref. 29), rat AQP5 (alveolar type I pneumocytes; ref. 30), and human RhAG (RBCs, ref. 31). We also studied bacterial AmtB. Our results show that all 5 channels are permeable to CO₂, and all but AQP4 and AQP5 are permeable to NH₃. A relative index of CO₂/NH₃ permeability varied widely: AQP4 ≅ AQP5 > > AQP1 > AmtB > RhAG. Thus, as is true for ion channels, gas channels exhibit substantial solute selectivity, which could play an important physiological role in controlling gas fluxes.

Results

pH_S Transients Caused by Applying CO₂ vs. NH₃.

Fig. 1A illustrates schematically how the influx of CO₂ leads to a fall in [CO₂] near the extracellular surface of the membrane ([CO₂]_S), which in turn leads to a rise in pH_S. Fig. 1B shows how the influx of NH₃ leads to a fall in pH_S. As described in ref. 6, exposing an AQP1-expressing oocyte to a solution containing 5% CO₂/33 mM HCO₃⁻ at a constant pH of 7.50 causes a transient rise in pH_S, followed by an exponential decay (Fig. 1C Left, green record). After the washout of CO₂ (see SI Text and Fig. S3), exposing the same oocyte to 0.5 mM NH₃/NH₄⁺ causes a transient fall in pH_S (Fig. 1C Right, green record), as noted elsewhere (32). Additional data are consistent with the hypothesis that Xenopus oocytes handle NH₃ in an unusual way, sequestering most incoming NH₃ in an intracellular compartment as NH₄⁺ (32).

Fig. 1.

Basis of surface pH changes. (A) CO₂ influx. At the outer surface of the membrane, the CO₂ influx creates a CO₂ deficit. The reaction HCO₃⁻ + H⁺ → CO₂ + H₂O, in part, replenishes the CO₂, raising pH_S. (B) NH₃ influx. The reaction NH₄⁺ → NH₃ + H⁺, in part, replenishes NH₃, lowering pH_S. (C–E) Representative pH_S transients from oocytes injected with H₂O or expressing AQP1 (record repeated in the 3 images), SGLT1, NKCC2, or PepT1. All data in C–E were obtained on the same day, from the same batch of oocytes, exposed first to CO₂/HCO₃⁻ and then (after CO₂ removal) to NH₃/NH₄⁺. Also shown in C are records with no oocyte present. Before and after solution changes, we retracted the pH electrode to the bulk extracellular solution (pH 7.50) for calibration. (F and G) Summary of extreme excursions of pH_S (ΔpH_S) for CO₂ and NH₃ data. (H and I) Representative pH_S transients from day-matched H₂O or AQP1 oocytes exposed to CO₂/HCO₃⁻ or 30 mM butyrate. (J and K) Summary of ΔpH_S for experiments like those in H or I. Values are means ± SE, with numbers of oocytes in parentheses. For F and G, statistical comparison between H₂O-injected controls and other oocytes (separately for CO₂ and NH₃ data) were made using a 1-way ANOVA for 5 groups, followed by Dunnett's multiple comparison. ΔpH_S values for H₂O, NKCC2, and PepT1 do not differ from one other. For J and K, statistical comparisons were made using unpaired 2-tailed t tests.

The maximal pH_S transients are much smaller in day-matched oocytes injected with H₂O (orange) or cRNA encoding the Na/glucose cotransporter SGLT1 (black), and are totally lacking in the absence of an oocyte (gray). Moreover, oocytes expressing the Na/K/Cl cotransporter NKCC2 or the H/oligopeptide cotransporter PepT1 have ΔpH_S values similar to those of H₂O-injected oocytes (Fig. 1 D and E).

Fig. 1 F and G summarize ΔpH_S data for a larger number of experiments like those in Fig. 1 C–E and show that ΔpH_S for AQP1 is significantly greater than for all other oocyte groups. As expected, the time constant (τ_pHs) for the decay of pH_S from its peak, an index of the time required for CO₂ to equilibrate across the membrane, has a pattern that is the inverse of that for ΔpH_S (Fig. S1 A and B). Because of the oocyte's unusual NH₃ handling, the pH_S relaxation during NH₃ exposures is prolonged, precluding the calculation of a τ_pHs for NH₃. In the CO₂ protocol, the smaller ΔpH_S for SGLT1 vs. H₂O oocytes could reflect a decrease in the expression of other proteins or an increase in membrane–protein/lipid ratio.

Given our observations with CO₂, one might ask whether AQP1 also would enhance the flux of a permeant weak acid like butyric acid (33). While confirming that AQP1 increases the ΔpH_S in oocytes exposed to CO₂ (Fig. 1 H and J), we found that the channel has no effect either on ΔpH_S (Fig. 1 I and K) or τ_pHs (Fig. S1 C and D) in oocytes exposed to butyric acid.

In principle, AQP1 could enhance the CO₂-induced pH_S spike in Fig. 1 C or H, not because AQP1 is a CO₂ channel, but because it has unanticipated carbonic anhydrase (CA) activity and thus catalyzes the extracellular reaction HCO₃⁻ + H⁺ → CO₂ + H₂O (see Fig. 1A). To test the CA hypothesis, we injected oocytes with cRNA encoding CA IV, coupled via a GPI linkage to the outer surface of the membrane. Increased CA-IV expression causes a graded increase in ΔpH_S (Fig. S2), and we determined the dose of CA-IV cRNA that produces the same ΔpH_S as a typical AQP1 oocyte. Fig. 2 shows that membrane preparations of oocytes injected with this dose of CA-IV cRNA, compared with those from H₂O oocytes, require a much shorter time to achieve a pH endpoint in a CA assay. However, membrane preparations of AQP1 oocytes are indistinguishable from those of H₂O, allowing us to rule out the CA hypothesis. Thus, on the basis of ΔpH_S and τ_pHs measurements, AQP1, but not SGLT1, NKCC1, and PepT1, acts as a channel for CO₂ and NH₃.

Fig. 2.

Carbonic anhydrase activities of Xenopus oocytes. CA activity was determined from membrane preparations of 100 oocytes injected with H₂O, 0.25 ng of cRNA encoding hCA IV, or 25 ng of cRNA encoding hAQP1. We divided each membrane preparation into aliquots containing 20 μg of total protein and performed a colorimetric CA assay on each aliquot (see SI Text). The sample mixtures containing CA-IV were run ± 100 μM methazolamide (MTZ), a CA inhibitor. Each N refers to a CA assay on 1 aliquot.

Comparison of AQPs with AmtB and RhAG.

Using the same protocol shown in Fig. 1 C–E, we systematically examined the effects of sequential exposures to CO₂/HCO₃⁻ and then NH₃/NH₄⁺ on oocytes expressing AQP1, AQP4, AQP5, AmtB, or RhAG. Fig. 3A shows again that CO₂ and NH₃ elicit larger pH_S spikes in AQP1 (green) than in day-matched H₂O (orange) oocytes. Both AQP4 (Fig. 3B) and AQP5 (Fig. 3C) enhance the pH_S spike with CO₂ but not with NH₃. Both AmtB (Fig. 3D) and RhAG (Fig. 3E) enhance the pH_S spike with CO₂ but are especially effective with NH₃. Asp¹⁶⁰ is vital for AmtB activity (19, 34), and the homologous Asp¹⁶⁷ is required by RhAG (35). We found that the inactive D160A mutant of AmtB (34, 36) and the inactive D167N mutant of RhAG (35) are inactive as either CO₂ or NH₃ channels, presumably because the mutations cause major structural changes (35). Figs. S3 and S4 A and D show full-length experiments similar to those in Fig. 3 A–E.

Fig. 3.

Surface pH changes caused by CO₂ and NH₃ influx in oocytes expressing gas-channel proteins. (A–E) Typical pH_S transients from oocytes injected with H₂O or expressing AQP1, AQP4, AQP5, WT AmtB or its inactive D160A mutant, or WT RhAG or its inactive D167A mutant. The protocol was the same as in Fig. 1. (F–J) Summary of extreme excursions of pH_S (ΔpH_S) caused by CO₂ influx. Each image represents mean values for day-matched oocytes. (K–O) Summary of ΔpH_S caused by NH₃ influx. Each image (F–O) represents mean values for day-matched oocytes. Some H₂O oocytes served as controls in more than 1 panel (total number of H₂O oocytes: 54 for CO₂, 61 for NH₃). Values are means ± SE, with numbers of oocytes in parentheses. For F–H and K–M, statistical comparisons were made using unpaired 2-tailed t tests. For I and J and N and O, statistical comparisons were made using 1-way ANOVAs for 3 groups, followed by Student–Newman–Keuls analyses.

In the above experiments, all AmtB and RhAG constructs were C-terminally tagged with EGFP (enhanced GFP), and fluorescence measurements confirmed trafficking to near the oocyte surface. The tagged and untagged constructs yielded identical results in pH_S assays (Fig. S4).

Fig. 3 F–J are analogous to Fig. 1F, except that in Fig. 3 F–J, we pair each oocyte expressing a WT or mutant channel with its day-matched H₂O-injected control. Each WT protein yields a ΔpH_S that is significantly greater than the H₂O control or (as applicable) the mutant protein. Moreover, the ΔpH_S values of the mutants are not different from the corresponding H₂O oocytes. Fig. 3 K–O is a summary of the NH₃ data. The results are comparable to the CO₂ data, except that the magnitudes of ΔpH_S for AQP4 and AQP5 in the NH₃ protocol are not different from those of their corresponding H₂O-injected controls. Fig. S5 shows that the relationship for the τ_pHs values is the inverse of that for the ΔpH_S values in Fig. 3. Thus, each of the proteins, AQP1, AQP4, AQP5, AmtB, and RhAG, is permeable to CO₂. Moreover, AQP1, AmtB, and RhAG, but not AQP4 or AQP5, are permeable to NH₃.

Cleavage of AmtB Signal Peptide by Oocytes.

Our AmtB cDNA encodes a signal peptide that Escherichia coli naturally cleaves (37), so that the new N terminus is extracellular. The A22K point mutation in AmtB prevents the cleavage in E. coli, although the mutant AmtB still forms trimers and is active (37). To verify that Xenopus oocytes also cleave the signal peptide, we added a C-terminal His tag to WT AmtB, A22K-AmtB, and an AmtB variant with the signal peptide truncated. Fig. 4 shows Western blots of plasma-membrane preparations from oocytes expressing the 3 constructs. The molecular mass of the major band of WT AmtB is appropriately less than that of A22K-AmtB, but the same as truncated AmtB. Thus, oocytes do indeed cleave the signal peptide of WT AmtB. Densitometry indicates that >90% of the AmtB in oocytes is appropriately cleaved. Additional data reveal that AmtB-His is active as both a CO₂ and an NH₃ channel.

Fig. 4.

Western blots testing cleavage of the AmtB signal peptide in Xenopus oocytes. (A) Wild-type AmtB vs. uncleavable A22K mutant. (B) Wild-type AmtB vs. AmtB with truncated signal peptide. Data are representative of 4 similar experiments. All constructs were His tagged at the C terminus and detected with an anti-His antibody.

P_f in Oocytes Expressing Different Channels.

So that we could relate our CO₂ and NH₃ data to the wealth of information on the osmotic water permeability (P_f) of AQP-expressing oocytes, we determined P_f for each AQP oocyte and its day-matched control from the dataset in Fig. 3. As summarized by the 3 pairs of bars on the left side of Fig. 5, the mean P_f value for each AQP was significantly and substantially greater than the matched controls. We separately assessed P_f for matched oocytes expressing AQP1, AmtB, or RhAG vs. day-matched H₂O oocytes. The right side of Fig. 5 shows that only for AQP1 oocytes was the mean P_f value significantly greater than that for H₂O oocytes; the P_f values for H₂O, AmtB, and RhAG oocytes were not significantly different from one another. Thus, despite the hypothesized presence of H₂O in the NH₃ pore of AmtB (20, 22), the 2 Rh proteins do not function as water channels.

Fig. 5.

Osmotic water permeabilities of Xenopus oocytes. P_f (cm/s) of oocytes injected with H₂O or cRNA encoding AQP1, AQP4, AQP5, AmtB, or RhAG. Values are means ± SE, with numbers of oocytes in parentheses. Statistical comparison between H₂O vs. AQP oocytes were made using unpaired 2-tailed t tests. Statistical comparisons among the 4 groups were made using a 1-way ANOVA, followed by Student–Newman–Keuls analyses.

Discussion

Channel-Dependent ΔpH_S and P_f Values.

Given our experimental design, the magnitude of ΔpH_S is a semiquantitative index of both the flux of, and membrane permeability to, CO₂ or NH₃. For CO₂, the same is true of τ_pHs. Note that the quantitative relationship between ΔpH_S on the one hand and absolute flux or permeability on the other is likely to be different for CO₂ vs. NH₃. The portion of the CO₂-induced ΔpH_S signal that we can ascribe to a particular channel is the difference between the ΔpH_S of each channel-expressing oocyte (e.g., green record in Fig. 3A) and the ΔpH_S of its day-matched H₂O-injected control (e.g., orange record in Fig. 3A). Fig. 6A summarizes these differences, the channel-dependent signal (ΔpH_S*)_CO2, for the CO₂ data, computed oocyte by oocyte. Similarly, Fig. 6B summarizes the analogous differences, the channel-specific signal (ΔpH_S*)_NH3, for the NH₃ data. Note that the mean (ΔpH_S*)_NH3 values for AQP4 and AQP5 are not significantly different from zero.

Fig. 6.

Comparison of channel-dependent properties. (A) Indices of channel-dependent CO₂ permeability. For each ΔpH_S from a channel-expressing oocyte, we subtracted the mean, day-matched ΔpH_S for H₂O oocytes. Bars represent mean subtracted values, the channel-dependent ΔpH_S for CO₂, or (ΔpHS*)_CO2. Note: Oocytes in A are the same as those in B and C. (B) Indices of channel-dependent NH₃ permeability. We computed (ΔpH_S*)_NH3 using the same approach as in A. (C) Channel-dependent water permeabilities. For each P_f from a channel-expressing oocyte, we subtracted the mean, day-matched P_f for H₂O oocytes. Bars represent mean subtracted values, the channel-dependent P_f, or P_f*. (D) Indices of channel-dependent CO₂ and NH₃ permeability, normalized to P_f*. For each oocyte, we divided (ΔpH_S*)_CO2 and (ΔpH_S*)_NH3 by its P_f*. (E) Indices of gas selectivity. For each oocyte expressing AQP1, AmtB, or RhAG, we divided (ΔpH_S*)_CO2 by −(ΔpH_S*)_NH3. Because (ΔpH_S*)_NH3 was not significantly different from zero for oocytes expressing AQP4 or AQP5, we represent these ratios as “infinity.” Statistical comparisons were made using 1-way ANOVAs for 3 groups, followed by Student–Newman–Keuls analyses.

Ratios of Indices of Permeability.

Because we do not know the number of AQP molecules at the plasma membrane in each oocyte, it is impossible to normalize our (ΔpH_S*)_CO2 or (ΔpH_S*)_NH3 data to protein abundance. However, for each AQP oocyte, we also have a channel-dependent P_f (P_f*), summarized in Fig. 6C. For each oocyte, we divided (ΔpH_S*)_CO2 or (ΔpH_S*)_NH3 by P_f*. Fig. 6D summarizes these mean values, which represent semiquantitative indices of the CO₂/H₂O or NH₃/H₂O permeability ratios. By a factor of 2, AQP5 has the highest (ΔpH_S*)_CO2/P_f*, and the values for AQP1 and AQP4 are indistinguishable.

Because we have both (ΔpH_S*)_CO2 and (ΔpH_S*)_NH3 for each of a large number of oocytes, it is also possible to compute the ratio (ΔpH_S*)_CO2/(ΔpH_S*)_NH3, a relative index of the CO₂/NH₃ permeability ratio, for AQP1, AmtB, and RhAG. Fig. 6E summarizes these values. Because (ΔpH_S*)_NH3 for AQP4 and AQP5 do not differ from zero, the ratios for these proteins are theoretically infinite. Thus, among the channels tested, AQP4 and AQP5 have the highest CO₂/NH₃ permeability ratios, followed by AQP1, AmtB, and RhAG. Conversely, RhAG has the highest NH₃/CO₂ permeability ratio (see Fig. S6).

Thus, compared with the AQPs tested, the Rh-like proteins tested are relatively more selective for NH₃, whereas the AQPs are relatively more selective for CO₂.

Significance.

Our data demonstrate that channel proteins can exhibit gas selectivity by channel proteins. The basis of the selectivity is probably not the size of the transiting molecules—H₂O, CO₂, and NH₃ have similar minimum diameters—but rather their chemistries and the chemistries of the monomeric pores vs. the pore at the center of the multimers. The electronic configuration of NH₃ is identical to that of H₂O, which moves exclusively through the monomeric aquapores of AQP1. Thus, the hydrophilic NH₃ probably also moves exclusively through the monomeric aquapores of AQP1 and through the monomeric ammonia pores of AmtB and RhAG. The less hydrophilic CO₂, however, could move through the hydrophobic central pores of all 5 channels. NO is also known to move through AQP1 (8), and indirect evidence is consistent with the idea that O₂ moves through AQP1 (38). We suggest that the hydrophobic NO and O₂ move through the central pores. Crystallographic data show that xenon can enter the central pore of the bacterial Rh50 (21). Moreover, in the case of AQP1, molecular-dynamics simulations show that CO₂ could penetrate the 4 aquapores or, with greater ease, the central pore (7). We hypothesize that the CO₂/NH₃ selectivities that we observe reflect the relative permeabilities of the 2 gases through the monomeric vs. the central pores of the 5 channels we studied.

In a cell like the human RBC, whose plasma-membrane lipid has an intrinsically low gas permeability (6), the gas selectivity of AQP1 and the Rh complex would provide control over dissolved gases crossing the membrane. The NH₃ permeability of AQP1 and RhAG could enhance the ability of RBCs to pick up NH₃ in various tissues (where the NH₃ gradient would favor NH₃ uptake) and then to off-load it in the liver (where the gradient would favor NH₃ efflux from RBCs and uptake by hepatocytes) for NH₃ detoxification. In the hypertonic renal medulla, this NH₃ permeability could reduce the reflection coefficient for NH₃ and thereby reduce cell-volume changes. However, the low NH₃ permeability of AQP4 could protect the brain from rising blood levels of NH₃, while still allowing CO₂, and perhaps NO and O₂, to pass.

Materials and Methods

Molecular Biology.

AQPs.

Human AQP1 cDNA (GenBank accession no. NM_198098), cDNA encoding the rat AQP4/M23 splice variant (GenBank accession no. NM_012825), and human AQP5 cDNA (GenBank accession no. NM_012779) were gifts of Peter Agre (Johns Hopkins University, Baltimore).

AmtB.

We cloned E. coli AmtB cDNA (GenBank accession no. ECU40429) by PCR from genomic DNA and subcloned the ≈1.3-kb PCR product into the Xenopus expression vector pGH19 (39). Using PCR, we created an additional construct in which we replaced the nucleotides encoding the signal sequence (i.e., first 22 residues) with ATG. At the 3′ end of some constructs, we added in-frame cDNA encoding either EGFP (Clontech, ref. 40) or a His tag.

RhAG.

Human RhAG cDNA in pT7TS (GenBank accession no. NM_000324, a gift of Baya Chérif-Zahar, Université René Descartes, INSERM, Paris) (41) was subcloned into pGH19. We tagged RhAG at its 3′ end with EGFP.

Other cDNAs.

Rabbit NKCC2 (42) was a gift of Biff Forbush (Yale University, New Haven, CT). SGLT1 (43) and PepT1 (44) were gifts of Matthias Hediger (Brigham and Women's Hospital and Harvard Medical School, Boston).

Site-Directed Mutagenesis.

We used the QuikChange Site-Directed Mutagenesis Kit (Stratagene), following the manufacturer's instructions.

cRNA Preparation.

We generated cRNA using the Message Machine kit (Ambion) and, unless otherwise stated, injected oocytes with 50 nL of 0.5 ng/nL of cRNA or 25 nL of 1 ng/nL of cRNA.

Western Blot Analysis.

Plasma-membrane proteins were prepared from oocytes (45), separated on a 13% SDS polyacrylamide gel, blotted on a PVDF membrane, probed with a monoclonal anti-His antibody (Catalog no. 70796–3, Novagen), and detected using ECL plus Western Blotting Detection Reagents (Amersham Biosciences).

Solutions for Physiological Assays.

The ND96 solution contained: 96 mM NaCl, 2 mM KCl, 1 mM MgCl₂, 1.8 mM CaCl₂, and 5 mM Hepes, pH 7.50, osmolality 195 mOsm. For P_f assays, we used a hypotonic ND96 variant (100 mOsm) that contained only 43 mM NaCl. The CO₂/HCO₃⁻ solution was identical to ND96 except that 33 mM NaHCO₃ replaced 33 mM NaCl, and the solution was bubbled with 5% CO₂/balance O₂. The NH₃/NH₄⁺ solution was a variant of ND96 in which we replaced 0.5 mM NaCl with 0.5 mM NH₄Cl. The butyrate solution was a variant of ND96 in which we replace 30 mM Na-butyrate with 30 mM NaCl.

Carbonic-Anhydrase Assay.

Carbonic-anhydrase activity was assessed in 20 μg of membrane preparation of CA-IV or AQP1 oocytes using a colorimetric technique (46). The assay measures the rate at which the pH of a weakly buffered alkaline solution (imidazole-Tris, 50% CO₂, with ρ-nitrophenol as indicator at 0 °C), falls in the presence or absence of CA, noted by a color change from yellow to clear, due to the reaction CO₂ + H₂O → HCO₃⁻ + H⁺.

Measurement of Oocyte Water Permeability.

We used a volumetric assay (47, 48) to measure osmotic water permeability (P_f). Briefly, after dropping oocytes into a Petri dish containing the hypotonic solution, we acquired video images every 1–2 s, obtaining the time course of the projection area of the oocyte. Assuming the oocyte to be a sphere, and the true surface area (S) to be 8-fold greater than the idealized area (49), we computed P_f as:

where V_o is initial oocyte volume, d(V/V_o)/dt is the maximal fractional rate of volume increase, ΔOsm is the osmotic gradient across the membrane, and V_w is the molar volume of water.

Measurement of Surface pH.

We used microelectrodes to measure pH_S (6, 50). Briefly, the pH electrode had a tip diameter of 15 μm, was filled at its tip with H⁺ ionophore mixture B (Catalog no. 95293, Fluka), and was connected to a FD223 electrometer (World Precision Instruments). The extracellular reference electrode was a glass micropipette filled with 3 M KCl and connected via a calomel half cell to a 750 electrometer (World Precision Instruments). The extracellular solution flowed at 3 mL/min, and the sampling rate was 1 per 500 ms. Using an MPC-200 system micromanipulator (Sutter Instrument), we positioned the pH_S electrode tip either in the bulk extracellular fluid or dimpling ≈40 μm onto the oocyte surface, in the “shadow” of the oocyte. Although not displayed in the figures, membrane potential (V_m) and intracellular pH were also monitored (see SI Text). All oocytes had initial V_m values at least as negative as −40 mV.

We verified delivery of EGFP-tagged proteins to a region near the plasma membrane by using a 96-well plate reader (BMG Labtechnologies) to assess fluorescence (40).

Data Analysis.

Before applying CO₂/HCO₃⁻ or NH₃/NH₄⁺, we computed pH_S from the preceding calibration, with the electrode tip in the bulk phase of the ND96 solution (pH 7.50). After applying CO₂/HCO₃⁻ or NH₃/NH₄⁺, we computed pH_S from a second calibration in the bulk phase of the new solution (also pH 7.50). The maximum pH_S excursion (ΔpH_S) was taken as the maximum pH_S after the application of CO₂/HCO₃⁻ (or the minimum pH_S after application of NH₃/NH₄⁺) minus the pH_S prevailing just before the solution change from ND96.

Statistics.

Data are presented as mean ± SEM. To compare the difference between 2 means, we performed Student's t tests (two tails). To compare more than 2 means, we performed a 1-way ANOVA followed by a Dunnett's or a Student–Newman–Keuls posthoc analysis, using KaleidaGraph (Version 4, Synergy Software). P < 0.05 was considered significant.