Entry to “\batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\mathrm{HCO_{3}^{-}}\) \end{document} Tunnel” Revealed by SLC4A4 Human Mutation and Structural Model*

Glaucoma, cataracts, and proximal renal tubular acidosis are diseases caused by point mutations in the human electrogenic Na+ bicarbonate cotransporter (NBCe1/SLC4A4) (1, 2). One such mutation, R298S, is located in the cytoplasmic N-terminal domain of NBCe1 and has only moderate (75%) function. As SLC transporters have high similarity in their membrane and N-terminal primary sequences, we homology-modeled NBCe1 onto the crystal structure coordinates of Band 3(AE1) (3). Arg-298 is predicted to be located in a solvent-inaccessible subsurface pocket and to associate with Glu-91 or Glu-295 via H-bonding and charge-charge interactions. We perturbed these putative interactions between Glu-91 and Arg-298 by site-directed mutagenesis and used expression in Xenopus oocyte to test our structural model. Mutagenesis of either residue resulted in reduced transport function. Function was “repaired” by charge reversal (E91R/R298E), implying that these two residues are interchangeable and interdependent. These results contrast the current understanding of the AE1 N terminus as protein-binding sites and propose that hkNBCe1 (and other SLC4) cytoplasmic N termini play roles in controlling \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\mathrm{HCO_{3}^{-}}\) \end{document} permeation.

Regulating and maintaining acid-base homeostasis is critical for normal cell, tissue, and systemic function. Transporters in several transporter families are involved in this multilevel pH regulation: Slc4 (bicarbonate transporters), Slc9 4 (Na ϩ -H ϩ exchangers), Slc16 (H ϩ /monocarboxylate cotransporters), and Slc26 (anion and bicarbonate transporters). Na ϩ -H ϩ exchangers (NHEs, 5 Slc9 proteins) play important pH regulatory roles in many cells and tissues. Nevertheless, in many cells, HCO 3 Ϫ transporters carry even more acid-base equivalents than NHEs and are often more active in CO 2 /HCO 3 Ϫ environments (normal cellular and tissue buffering system).
In the renal proximal tubule, the major apical step of bicarbonate absorption is acid secretion to the forming urine by the NHE3, Na ϩ -H ϩ exchanger (6). The basolateral step of proximal tubule bicarbonate absorption appears to rely exclusively on NBCe1 function (Na ϩ /nHCO 3 Ϫ cotransport). For example, the NHE3 knock-out mice have only a slight metabolic acidosis (blood pH ϳ7.27 and [HCO 3 Ϫ ] ϭ 21 mM) (7,8), indicating that NHE3 is not the rate-limiting step in transepithelial bicarbonate absorption. However, both humans with NBCe1 mutations (1, 9 -13) and the NBCe1 knock-out mice (14) have significant metabolic acidosis (humans, see above; NBCe1 Ϫ/Ϫ , blood pH 6.86, [HCO 3 Ϫ ] ϭ 5.3 mM). Taken together, these data indicate that the basolateral exit of HCO 3 Ϫ via NBCe1 rather than apical H ϩ secretion via NHE3 is the dominant and rate-limiting step in kidney bicarbonate absorption. This loss-of-function/reduced-function phenotype also indicates that NBCe1 is the only HCO 3 Ϫ absorption pathway in the renal proximal tubule and that NBCe1 plays a key role for maintaining ocular pressure and corneal clarity.
Loss of NBCe1 function may result from (a) aberrant protein processing or folding (12,13,15), (b) protein truncation (10), or (c) misfunction of the NBCe1 protein (1,9,16). For example, S427L in transmembrane span 1 is a biophysical (functional) mutation resulting in unidirectional transport at 10% of wildtype (9), whereas L522P in transmembrane span 5 is a protein trafficking problem (12). R298S-NBCe1 was originally reported as having ϳ50% wild-type function (1), i.e. also a biophysical mutation, but was more recently reported as a protein traffick-ing problem (1,9,16). A transmembrane topography of the human NBCe1-B has been proposed based on glycosylation studies (17). None of the proposed structural models dispute the Arg-298 location, i.e. residing in the center of the cytoplasmic N terminus of NBCe1 and not obviously associated with the transmembrane domain. How then does this placement translate to malfunction of the R298S-NBCe1 protein? Does this imply that HCO 3 Ϫ permeation and/or affinity is associated with the cytoplasmic N terminus? Transmembrane domains of membrane proteins are generally thought to control ion permeability across membranes. However, knowing the sequence and predicted structural location (based on linear sequence) is not always the best predictor of structure.
NBCe1 is a member of the HCO 3 Ϫ transporter gene family that includes Band 3 (AE1/SLC4A1). All SLC4 members have Ͼ35% sequence identity, particularly in predicted membrane spans. Although crystals were recently obtained for this region of NBCe1, only gross topology rather than amino acid assignment was reported (18). We hypothesized that we could gain insights into NBCe1 structure and function by mapping its amino acid sequence onto the AE1 N-terminal structure (3). This structural prediction indicated that Arg-298, a conserved residue in SLC4 proteins, is located in a solvent-inaccessible pocket. The model further predicted that Arg-298 has charge interactions with Glu-295 and Glu-91, both of which are also conserved in SLC4 HCO 3 Ϫ transporter proteins. Are these sequence alignments coincidence, or is the Band 3 N-terminal structure a good predictor of NBCe1 N-terminal structure? In this study, we use point mutations to perturb the charge interaction between Glu-91 and Arg-298. Our results indicate that Arg-298, Glu-91, and their interaction are crucial for the NBCe1 N-terminal structure as well as the normal physiological function of NBCe1. Thus, this structural model and the following experiments elucidate, on the molecular level, "why" R298S causes a proximal RTA with bilateral cataracts and glaucoma. These results suggest that the NBCe1 cytoplasmic N terminus dictates/controls HCO 3 Ϫ permeation or affinity. These results also challenge the general belief that the N terminus of membrane proteins, like NBCe1, is not directly involved with substrate transport but rather is merely serving as a protein anchor, as is believed for the AE1 N terminus (3).

EXPERIMENTAL PROCEDURES
NBCe1 N Terminus Structure Modeling-A pair-wise alignment of sequences of human kidney NBCe1 (residues 62-371) and AE1 (residues 55-356) accession codes M27819 was prepared in the Swiss-PdbViewer (SPDBV (19)) or externally with SIM and the Blosum62 algorithm (20). These nearly identical alignments, when submitted together with the coordinates for AE1 (Protein Data Bank accession code 1HYN) to the Swissmodel server (21), did not yield an initial model due to failure of identifying appropriate loops. Despite a 36.5% overall sequence identity between these proteins, there is a region among NBCe1 (residues 113-174) and AE1 (residues 165-218) that does not show much sequence similarity, reducing the threading reliability in this area. Thus, multiple sequence alignments with the ClustalW algorithms (DNASTAR) were made to identify boundaries of conserved and variable sequence regions within and across homologous domains of the Slc4 protein families. The resulting manually optimized binary alignment between NBCe1 and AE1 served as input for Swiss model. The returned initial model, based on the four individual copies of the domain in HYN, was briefly minimized. Side chain conformations were subsequently optimized with SCWRL3 (22) and minimized to yield the present model structure (Fig. 2B). The NBCe1 model was then analyzed with the programs VADAR, Procheck, and WHAT IF (23-25) (Fig. 2,C-E).
Cloning and Mutations-The human kidney NBCe1 (hkNBCe1) clone in a Xenopus expression plasmid was previously described (9). hkNBCe1 mutations were generated using QuikChange (Stratagene, La Jolla, CA) and sequenced for verification (W. M. Keck Biotechnology Resource Laboratory, New Haven, CT). Linearized cDNA was used to make capped cRNA with the SP6 mMessage mMachine kit (Ambion, Austin, TX) as described previously (26).
A hemagglutinin affinity tag "HA tag" was engineered into the extracellular loop of hkNBCe1 at the Ser-596 3 Ser-610 region with a linker (SNDTTLAP-DYPYDVPDYAG-EYLPTMS) as that described in McAlear et al. (27). This HA tag insertion does not affect NBCe1 activity or sensitivity to stilbenes (27). Using an anti-HA 1 o antibody and horseradish peroxidase-conjugated 2 o antibody with a chemiluminescent substrate, we were able to quantify surface expression of hkNBCe1 clones in a luminometer. The single-oocyte chemiluminescence technique utilizes enzyme amplification of chemiluminescence substrate to quantify a HA-tagged protein expressed at the cell surface (28,29). This technique has a linear relationship between surface expression detected by single-oocyte chemiluminescence and functional activity of the K ϩ channel, ROMK (Kir 1.1), as reported by Yoo et al. (30).
Oocyte Electrophysiology-50 nL of water (control) or RNA solution (25 ng of hkNBCe1 or mutant cRNA) was injected into stage V/VI Xenopus oocytes. Voltage electrodes, made from fiber-capillary borosilicate and filled with 3 M KCl, had resistances of 1-10 megaohms (31). Ion-selective electrodes were pulled similarly and silanized with bis-(dimethylamino)-dimethylsilane (Fluka Chemical Corp., Ronkonkoma, NY). pH electrode tips were filled with hydrogen ionophore 1 mixture B (Fluka) and back-filled with phosphate buffer (pH 7.0). Electrodes were connected to a high-impedance electrometer (WPI-FD223 for intracellular pH (pH i ) and V m experiments), and digitized output data (filtered at 10Hz) were acquired by PCLAMP software sampling at 0.5 Hz. All ion-selective microelectrodes had slopes of Ϫ54 to Ϫ57 mV/decade ion concentration (or activity). pH electrodes were calibrated at pH 6.0 and 8.0. For voltage-clamp experiments, electrodes were filled with 3 M KCl/agar and 3 M KCl and had resistances of 0.2-0.5 megaohms. Oocytes were clamped at Ϫ60 mV, and current was constantly monitored and recorded at 10 Hz (Warner Inst. Co., Oocyte Clamp OC-725C). Voltage steps pulses (75 ms) were executed from Ϫ160 to ϩ60 mV in 20 mV steps; the resulting I-V traces were filtered at 2 kHz (8 pole Bessel filter) and sampled at 10 kHz. Data were acquired and analyzed using Pulse and PulseFit (HEKA Instruments, Germany).
Oocyte Surface Protein Expression-Oocyte labeling was performed at 4°C. Oocytes were fixed with 4% paraformaldehyde in ND96 for 15 min, washed and incubated in 1% bovine serum albumin-ND96 blocking solution for 30 min. Oocytes were labeled with a 1 o antibody (1:200 dilution, monoclonal rat-␣-HA 1 o antibody (Roche Applied Science)) for 60 min, and then with a 2 o antibody (1:2000 dilution, horseradish peroxidase-conjugated goat-␣-rat IgG (Jackson ImmunoResearch Laboratories)) for 30 min in 1% bovine serum albumin-ND96 blocking solution. Labeled oocytes were washed several times and incubated in ND96 for 10 min before exposure to 50 l of the premixed SuperSignal ELISA Femto substrate solution (Pierce) at room temperature. Chemiluminescent was measured from single oocytes in a microcentrifuge tube using a TD-20/20n luminometer (Turner BioSystems). Measurements were taken at 30 s after initiation of the luminescent reaction.
Statistical Analysis-Values, quantity of ion activities, or currents are indicated as the mean Ϯ S.E. The total apparent buffering power (␤ T , see Table 1) is defined as the change in pH i before and after application of CO 2 /HCO 3 Ϫ (once steady state is reached) divided by the pH i change elicited from the same solution changes, i.e. ␤ T ϭ ⌬[HCO 3 Ϫ ] steady state /⌬pH i (9). Statistical analysis was performed with a one-tailed Student's t test to have a significant difference at p Ͻ 0.05 or less.

RESULTS
The addition of CO 2 to a solution, results in H ϩ formation because Ϫ . When CO 2 crosses to the inside of a Xenopus oocyte, this H ϩ formation results in a cellular acidification (decrease of pH i ). The addition of 5% CO 2 , 33 mM HCO 3 Ϫ , pH 7.5 (CO 2 /HCO 3 Ϫ ) to the bath solution resulted in a fast acidification in a water-injected control oocyte (Fig. 1A). When hkNBCe1 was expressed in Xenopus oocytes, this CO 2 -induced acidification in oocytes was much less. This decrease reflects the transport of extracellular HCO 3 Ϫ into the cell (via NBCe1), counteracting acidification by CO 2 hydration (Fig. 1B).
The bicarbonate transport is accompanied by a large positive (outward) current ( Fig. 1, B and E, squares) due to the 1 Na ϩ : nHCO 3 Ϫ stoichiometry of NBCe1 and the negative charge movement (32)(33)(34). No similar current can be observed in the water-injected oocytes (Fig. 1, A and E, circles). Sodium replacement with choline (hereafter referred as Na ϩ removal or "0Na ϩ ") in a HCO 3 Ϫ solution reverses the HCO 3 Ϫ transport direction (Fig. 1, B and D, 0Na ϩ ). That is, HCO 3 Ϫ is now moving out of the cell resulting in a fast acidification and an inward current for a hkNBCe1-expressing oocyte (9) (Fig. 1, B and D, 0Na ϩ , and Fig. 1F), equivalent to renal NaHCO 3 absorption. Na ϩ removal in HCO 3 Ϫ creates no detectable pH i change or current in the water-injected oocytes. One can use these parameters (current magnitude, I max , and the rate of acidifica-  Table 1. D, expanded time scale of Na ϩ removal from panel (gray box) over 6.6 min before and after Na ϩ was removed from the HCO 3 Ϫ solution. NBCe1 transport HCO 3 Ϫ into the cell working against acidification due to the CO 2 exposure resulting in alkalized pH i before the Na ϩ removal was only observed in WT-hkNBCe1. After Na ϩ removal, pH i change (pH units/s) of WT-hkNBCe1 (Ϫ130 ϫ 10 Ϫ5 pH units/s) decreased fastest among the three (R298S ϭ Ϫ71, ϫ 10 Ϫ5 ; water ϭ Ϫ10 ϫ 10 Ϫ5 ). E and F, currentvoltage relationship of hkNBCe1 mutants. Oocytes were voltage-clamped at Ϫ60 mV and stepped as indicated under "Experimental Procedures." Data shown are corrected I-V traces by subtracting initial measurements in ND-96 at corresponding voltage-steps; sample sizes are in parentheses. E, averaged data collected in CO 2 / HCO 3 Ϫ when the elicited outward currents reached a steady state. The reversal potentials are Ϫ80 mV. F, averaged traces when the inward current reached a steady state in the 0Na ϩ -CO 2 /HCO 3 Ϫ solution. The WT and R298S currents all rectified at positive holding voltages with no obvious reversal potential. hkNBCe1 Cytoplasmic N Terminus Controlling HCO 3 ؊ Permeation tion, dpH i /dt) with or without CO 2 /HCO 3 Ϫ solution ϮNa ϩ to depict the transporter function ( Table 1). The overall NBCe1transporter contribution can also be portrayed by the buffering power (␤ T ) in the unit of mM/pH unit. The voltage dependence of these hkNBCe1 currents (I-V curves; Fig. 1, E and F) illustrate that the membrane potential (electrical driving force) can overcome the chemical gradients for both Na ϩ and HCO 3 Ϫ . To understand the R298S pathophysiology, we recreated this point mutation in wild-type hkNBCe1 (WT) and expressed it in Xenopus oocytes. Results show that R298S is a moderately functioning mutation with decreased HCO 3 Ϫ affinity/capacity ( Fig.  1, C-F, and Table 1). The CO 2 /HCO 3 Ϫ -elicited outward current and the R298S-Na ϩ -dependent currents are significantly lower than those observed in WT. The R298S acidification rate is 78% faster than that of WT in CO 2 /HCO 3 Ϫ (indicating slower transport), and the acidification due to Na ϩ removal (renal absorption) is significantly slower than that of WT ( Fig. 1D; Ϫ71 versus Ϫ130 ϫ 10 Ϫ5 pH units/s). Comparing the rate before (NaHCO 3 influx) and after (NaHCO 3 efflux) Na ϩ removal in HCO 3 Ϫ solution, the dpH i /dt of R298S is decreased rather than increased as in WT, showing that HCO 3 Ϫ transport function is deficient in this human mutation. The ␤ T of R298S is also much less than that of WT ( Table 1). The differences in the current magnitude between WT and R298S are voltage-dependent, yet the reversal potential (at 0 current) remains unchanged (Fig.  1E). The NBCe1 I-V responses further illustrate that R298S is a moderate mutation with lower apparent HCO 3 Ϫ transport capacity and Na ϩ affinity than WT.
To elucidate the role of R298S in NBCe1 transport, we initiated structure-function studies. Our rationale was that sequence alignments of highly conserved residues of well characterized Slc4 protein sequences from divergent species could reveal candidate residues of transport importance (Fig. 2).
Members of the SLC4 family share 48.7% sequence identity through predicted membrane-spanning regions, although the animal Slc4 family includes functionally distinct transporters (35): (a) anion exchangers, (b) Na ϩ /HCO 3 Ϫ cotransporters, and (c) Na ϩdriven Cl Ϫ -HCO 3 Ϫ exchangers. Interestingly, even higher identities among Slc4 members are found within their N termini, particularly within spans of the folded N-terminal cytoplasmic domains (57.2% on average; NBCe1 has 67% identity among orthologs). Indeed, several 5-10-amino-acid stretches have 100% identity (36) ( Fig. 2A), including the absolutely conserved Arg-298 (hkNBCe1 numbering, equivalent to Arg-283 in AE1) and its sequential neighbors. Since mutation of Arg-298 results in renal and ocular disease and appears important for proper transporter function, we reasoned that functional insight into NBCe1 would be gained by homology modeling. The human NBCe1 N-terminal amino acid model (Fig. 2, B-E) is based on the only known Slc4 structure, the structure for the cytoplasmic N terminus Band 3 (AE1/SLC4A1) solved by x-ray crystallography (3).
High sequence similarity of these two transporters and the other family members facilitated a straightforward structure prediction (Fig. 2, B-E). Corresponding sequence and secondary structural elements are found for AE1 and NBCe1 in this domain, thus confirming the identity of the entire fold. The lowest identity is observed for one larger loop region (NBCe1 residues 114 -170; Fig. 2, B and D, light blue region), suggesting structural differences around the hairpin loop that binds ankyrin in Band 3. These differences are likely responsible for the lack of ankyrin binding by NBCe1. This domain fold is comprised of a central sheet surrounded by multiple helices. The spatially and somewhat separate helix-loop-helix motif represents a dimerization domain (Fig. 2B,  right). Similar to Band 3, this domain in NBCe1 appears to dimerize (18). We represent NBCe1 as a monomer since we lack ultimate proof of the dimeric nature of NBCe1.
The structural model indicates that Arg-298 is located in a largely solvent-inaccessible, polar subsurface pocket (Fig. 2C) and that Arg-298 is surrounded by multiple other charged and polar residues. Foremost, Arg-298 likely forms H-bonds with either Glu-295 or Glu-91 (approximate residue distance 3-3.5 Å) (Fig. 2, C and E). NBCe1 residues Glu-295 and Glu-91 are equivalent to human AE1 residues Gln-283 and Glu-85, respectively. Other residues of this "pocket" are Thr-108 at the top

transport and elicited currents from WT-and mutant NBCe1
Calculations are as indicated under "Experimental Procedures" and as previously described (9). The value for ⌬␤ hkNBCe1 is ␤ T (hkNBCe1) Ϫ ␤ T (water). These data were collected using the three-electrode experiments (see "Experimental Procedures") to voltage-clamped oocytes at Ϫ60 mV while also measuring pH i . I m is membrane current. For pH i and ⌬pH i values, there are actually four significant digits, although three are shown for readability. Italicized columns are the average value for each parameter for each clone or control. % values for each clone are normalized to the wild-type hkNBCe1 (taken as 100%). Note that values should be compared with both WT-NBCe1 as well as the water-control because the water-control versus the WT-NBCe1 percentage increases and decreases depending on the parameter. (not shown) and Thr-302 at the bottom (Fig. 2D). The pocket is flanked by two other potentially charged residues: His-105 and His-294. The residues that form this pocket are particularly well conserved among the Slc4 HCO 3 Ϫ transporter sequences ( Fig. 2A), indicating that this pocket is likely a general feature of all Slc4 HCO 3 Ϫ transporters. To test our putative structure for the NBCe1 N terminus, we created point mutations in hkNBCe1 to perturb the putative charge interactions among three residues (Arg-298, Glu-91, and Glu-295) (Fig. 3). Our NBCe1 model (Fig. 2, B and E) predicted that the mutations would cause charge repulsion, thereby opening the N-terminal structure and altering NBCe1 function. We began by evaluating the effect of R298E on the hkNBCe1 transport function by altering the charge of the disease-mutation from a neutrally charged (Ser) to a negatively charged residue (37). The HCO 3 Ϫ -evoked currents and Na ϩdependent currents in R298E are ϳ71% of WT (Fig. 3, A and E, and Table 1). CO 2 acidifies R298E oocytes 72% faster than WT, whereas Na ϩ removal in HCO 3 Ϫ acidifies WT 74% faster than R298E (Fig. 3, A and D). These results represent an impaired NBCe1 transport function resulting from the R298E mutation. All of these results are consistent with R298E-expressing oocytes having a buffering power (␤ T ) three times smaller than that of WT. Nevertheless, the I-V relationships for R298E (Fig. 3, E and F) are similar to the R298S disease mutation (Fig. 1, E and F). The Interestingly, E91R-hkNBCe1 exhibits very severe defects in ion transport function (Fig. 3B and Fig.  S1B). The CO 2 /HCO 3 Ϫ -evoked current in E91R appears gradually and reaches a plateau only slowly, instead of maximizing quickly fol-  Residues marked as green are residues within 4 Å from Glu-91, Arg-295, and Arg-298 (highlighted in red) and highly likely to have charge interactions with these three residues. B, ribbon diagram structure of hkNBCe1 amino acid sequence 62-313 mapped onto the corresponding region of the Band 3 (human AE1) crystal structure (PDB number 1HYN) of Low and co-workers (3). The blue to red progressive color scheme denotes secondary structure succession from N terminus toward C terminus. C, human missense mutation R298S is located in the solvent-inaccessible pocket. Highlighted are residues Glu-91 and Glu-295 about 3.5 Å from the Arg-298 side chains putatively forming H-bond (green) with Arg-298. Side-chain charges of the three residues are color-coded to negative (red) and positive (blue). D, chain of polar residues creating a polar channel in the domain core. E, close up view on the hydrogen bond network of polar residues. The NCBI/GenBank TM accession numbers for these sequences are hkNBCe1 (human kidney form electrogenic Na ϩ bicarbonate cotransporter 1; AF007216), hAE1-hAE3 (human anion exchanger 1-3; M27819, U62531, and U05596), drNDAE1 (Drosophila sodium-dependent anion exchanger 1; AF047468), ceNBC (Caenorhabditis elegans Na ϩ bicarbonate cotransporter; AF004926), aNBCe1 (Ambystoma electrogenic Na ϩ bicarbonate cotransporter 1; AF001958), rkNBCe1 (rat kidney electrogenic Na ϩ bicarbonate cotransporter 1; NM_053424), mNBCe1 (murine electrogenic Na ϩ bicarbonate cotransporter 1; AF141934), rb1NBCe1 and -2 (rat brain electrogenic Na ϩ bicarbonate cotransporter 1 and 2; AF124441 and AF254802), hpNBCe1 (human pancreas form electrogenic Na ϩ bicarbonate cotransporter 1; AF053753), NBCe1-dace (Osorezan dace electrogenic Na ϩ bicarbonate cotransporter 1; AB055467), NBCe1-trout (rainbow trout Na ϩ bicarbonate cotransporter 1; AF434166), NBC3 (human Na ϩ bicarbonate cotransporter 3; AF069512), NBC4c (human Na ϩ bicarbonate cotransporter 4; AF293337), NBCn1-D (rat electroneutral Na ϩ bicarbonate cotransporter 1-D, NM_058211), NCBE (Na ϩ -Cl Ϫ /bicarbonate exchanger; AB040457), NDCBE(Na ϩ -driven chloride/bicarbonate exchanger; AY151155). 3 ؊ Permeation lowed by a slow decay as seen in the WT (Fig. 1A and Fig. S1A). The current magnitude is 20 times smaller than in WT (Fig.  3, E and F, and Table 1). Accordingly, the CO 2 -induced acidification is much faster for E91R (Fig. 3B), i.e. greatly reduced HCO 3 Ϫ transport. Na ϩ -dependent currents (Fig. 3, E and F, circles) and acidification (Fig. 3D) are also much less and slower in E91R. These properties translate into significantly less HCO 3 Ϫ transport (lower ␤ T ) of E91R than that of WT. The E91R I-V relationship has no clear reversal potential in CO 2 /HCO 3 Ϫ , but the small HCO 3 Ϫ -elicited currents are still voltage-dependent (Fig. 3, E and F, circles). E91R is a much more impaired mutation than R298S, as indicated by the E91R I-V relationships resembling that of water-injected controls. However, these HCO 3 Ϫ -elicited and Na ϩ -dependent currents of E91R are significantly higher than those of water-injected controls ( Table 1). The dpH i /dt (HCO 3 Ϫ transport) due to solution change for E91R (Fig. 3, B and D) is also significantly different from waterinjected controls (Fig. 1, A and D).

hkNBCe1 Cytoplasmic N Terminus Controlling HCO
The surface expression of the NBCe1 transporter protein in the oocytes was quantified by detecting an HA tag engineered into the extracellular loop of hkNBCe1 and the NBCe1 mutants. Basal luminescence of oocyte surfaces was determined using oocytes expressing the untagged WT-hkNBCe1 transporter. The luminescence values of the HA-tagged mutants were not significantly different from that of the HA-tagged WT-hkNBCe1, demonstrating that there was no difference in surface expression of NBCe1 protein between the WT and mutants. These data confirm that the E91R-hkNBCe1 protein is appropriately trafficked and expressed at the oocyte plasma membrane (Fig.  4A). These results also verified that the extracellular HA tag did not alter the HCO 3 Ϫ -elicited current (Fig. 4B) or the 0Na ϩ /HCO 3 Ϫelicited currents (Fig. 4C) when compared with untagged NBCe1 proteins (Figs. 1, E and F, and 3, E and F).

DISCUSSION
We hypothesized that we might gain additional insights of NBCe1 structure and function by mapping the hkNBCe1, SLC4A4, amino acid sequence onto the AE1 crystal structure (Protein Data Bank (PDB) number 1HYN) (3). From this NBCe1 structural model, Arg-298 of NBCe1 is predicted to hydrogen-bond with Glu-91 and Glu-295. R298S is a human NBCe1 mutation resulting in renal and eye disease (see "Results"). Interestingly, Arg-298 is a conserved residue in the animal Slc4 protein family and is predicted to lie in a "solventinaccessible pocket." To test whether our structural model of NBCe1 was valid, we mutated these putative interacting resi-  Table 1. The R298E mutant has decreased currents and less substantial pH i changes in response to the addition of CO 2 /HCO 3 Ϫ and/or Na ϩ removal. The E91R mutant transport function is almost abolished in terms of current magnitude and pH i changes. Double mutant E91R/R298E operates approximately like a wild-type hkNBCe1. D, expanded time scale of Na ϩ removal from panel (gray box) over 6.6 min before and after Na ϩ was removed from the HCO 3 Ϫ solution. NBCe1 transports HCO 3 Ϫ into the cell, working against acidification due to CO 2 exposure, resulting in increased pH i of different magnitudes prior to 0Na ϩ . After Na ϩ removal, the pH i changes (ϫ10 Ϫ5 pH units/s) are E91R/R298E Ϸ WT Ͼ R298E (ϭ R298S) Ͼ E91R. E and F, current-voltage relationship of hkNBCe1 mutants. Data shown are corrected I-V traces by subtracting initial measurements in ND-96 at corresponding voltage-steps; numbers are in parentheses. E, averaged data collected in CO 2 /HCO 3 Ϫ . The reversal potentials are approximately Ϫ80 mV. F, averaged steady state inward current with 0Na ϩ -CO 2 /HCO 3 Ϫ solution. The wild-type and mutants currents all rectified at positive holding voltages with no obvious reversal potential. hkNBCe1 Cytoplasmic N Terminus Controlling HCO 3 ؊ Permeation dues. The modeling and functional data of the E91R-NBCe1 mutant imply that charge alterations (particularly the introduction of additional cationic side chains) are disruptive to the pocket structure around Arg-298. To further probe the role of residue charge at this putative solvent-inaccessible pocket, we examined the effect of charge reversal through a double mutation, E91R/R298E-hkNBCe1 (Fig. 3, C-F and Fig. S1C). If structure and charge interaction rather than exact sequence are important, we predicted that a clone that simultaneously reverses the charge of Glu-91 and Arg-298 (E91R/R298E) would have transporter function that resembles those of wildtype NBCe1. The data presented here show that the double mutant (E91R/ R298E) can restore E91R to full "WT-like" function (Fig. 5, A  and B). Thus, either the double mutant reverses the structural effect of E91R on wild-type function or R298E quenches the effects of E91R on the transport function (Figs. 1, 3, and 4 and Fig. S1) by restoring the native structure (Fig. 2, B-E). Although all tested mutants imply a role for this domain and particularly this pocket in the transmembrane ion transport, this rescue mutation presents the strongest evidence and validation for the close proximity between Glu-91 and Arg-298, i.e. correctness of this model, and for a direct residue-and location-specific correlation of the structural and functional changes. A similar res-idue charge interaction was reported recently between a pair of arginine and glutamate residues in a Torpedo acetylcholine receptor structure model (38). Acetylcholine-evoked singlechannel currents were measured from receptors with R209E or E45R mutation and the corresponding double mutation. The charge-reversal double mutation yielded surface expression and rescued wild-type acetylcholine-function abolished by single mutations.
Glu-91 and its sequential neighbor Glu-92 are part of the highly conserved motif (WRETARWIKFEE; amino acids 336 -347 in hkNBCe1) ( Fig. 2A) proposed to determine pH sensitivity of murine anion exchanger AE2/SLC4A2 (39 -41). Intriguingly, Glu-92 similarly is suggested by the model to be part of a second unusual feature of this fold. Glu-92 is located on the opposite side of the ␤-sheet from Glu-91. It is involved in a network of interacting residues as it hydrogen-bonds to arginine 86 (Arg-86) and potentially Lys-227 (Fig. 2E). Arg-86 in turn interacts with Glu-272. Residues Arg-86, Glu-92, Lys-227, and Glu-272 are highly conserved in the Slc4 bicarbonate transporters ( Fig. 2A), and these functionally important residues share number, charge, and residue type, nearly duplicating the properties of the pocket as for Glu-91, Glu-295, and Arg-298.

NBCe1
Na + nHCO 3 -efflux FIGURE 5. Summary of structural effects on hkNBCe1 function. A, averaged peak outward currents collected in CO 2 /HCO 3 Ϫ (pH 7.5) solution, i.e. inward NaHCO 3 uptake (right). B, average peak inward current in 0Na ϩ -CO 2 / HCO 3 Ϫ (pH 7.5) solution, i.e. outward renal transport (right). Oocytes expressing different hkNBC1 mutants and water-injected oocytes were voltageclamped at Ϫ60 mV. The current changes were recorded in various experimental solutions, and peak current changes were calculated by subtracting the elicited currents before and after the HCO 3 Ϫ or 0Na ϩ HCO 3 Ϫ solutions were introduced. The sample size for each clone is in Table 1. R298S and R298E are mutants with less pronounced peak current changes, which indicates altered transport functions. E91R is a severe mutation with nearly abolished transport function having significantly less current responses to solution changes. R298S, E91R, and R298E currents are stoically different from WT and E91R/R298E. E91R/R298E double mutant has peak current changes identical to WT-hkNBCe1, indicating that E91R/R298E restores E91R to a fully WT-hkNBCe1-like function (not statistically different from WT).

hkNBCe1 Cytoplasmic N Terminus Controlling HCO 3 ؊ Permeation
Together, these residues create an unusual continuous chain of interconnected polar residues and a steady path of high polarity through the core of this domain from the membrane oriented C-terminal side (Fig. 2, D and E) to the interior. It is intriguing to speculate on the function of this feature. This pathway may attenuate ion transport (or even serve as an ion transport pathway). Interestingly, when the same group of residues is mutated in AE2 (R341A, W342A, E346A, and E347A), pH sensitivity of wild-type anion transport is abolished (39 -41). A mutagenesis study of murine AE2 residues identified a histidine residue, corresponding to His-105 in NBCe1 amino acid sequence, important for regulation of Cl Ϫ transport (40). Histidine and lysine are hydrophilic, positively charged basic amino acids highly likely to be a potential pH sensor(s) for NBCe1. These residues have been characterized as pH sensors in many studies: acid-sensing ion channels (42), tandem pore domain acid-sensitive K ϩ channel (TASK-3) (43), Na ϩ /H ϩ antiporter (44), and ROMK1 channels (45). Glutamate, negatively charged and acidic amino acid, was also identified as the pH sensor in other investigations of uncoupling protein (46), TRPV5 channel (47), and ClC-2G Cl Ϫ channel (48).
Finally, it is noteworthy that the Slc4 gene family spans eukaryotes from humans to yeast to plants. In plants and yeast, Slc4 proteins have not been shown to transport HCO 3 Ϫ but rather borate (49). The Arabidopsis and mammalian boron transporter (BOR1/Slc4a11) (50) members lack ϳ391 cytoplasmic N-terminal sequence found in mammalian NBCe1 or other animal Slc4 members, and these boron transporters do not transport HCO 3 Ϫ . In addition, human pancreatic NBCe1 isoform (pNBC1/NBCe1-B) has an N-terminal variation with a lower bicarbonate transport capacity (17), which is disinhibited by an inositol 1,4,5-trisphosphate receptor binding protein (IRBIT) (51). These results corroborate the suggestion of a critical role of NBCe1 N terminus in HCO 3 Ϫ transport. The structure modeling points us to candidate residues for mutation analysis that eventually gave rise to a severe functional mutation, E91R (Fig. 2).
The effect of the R298S-hkNBCe1 mutation is unclear in the literature. R298S has been reported reducing wild-type function (1) and as a protein trafficking problem (1,9,16). This latter report uses Xenopus oocytes as we have in this study. Horita et al. (16) implied oocyte surface expression by coincident fluorescence of a NBCe1-A N-terminal antibody (intracellular epitope) and wheat germ agglutinin as a general marker of plasma membrane (extracellular). The data presented in Fig. 4A use an extracellular tag of the hkNBCe1 molecule, i.e. a direct assessment of the NBCe1 proteins at the plasma membrane. Contrary to the previous Xenopus oocyte report (1,9,16), the data in Fig. 4 also explicitly show that R298S-hkNBCe1 affects NBCe1 function and not NBCe1 protein processing.
This report provides a structure model and biophysical role for the NBCe1 N terminus based in part on a human NBCe1 disease mutation (R298S), summarized in Fig. 5. R298S-hkNBCe1 affects NBCe1 function and not NBCe1 protein processing (Fig. 4). Further, we detect the very unusual polarity of multiple core residues in the N-terminal domain, suggesting that this chain of connected residues may create and ion transport pathway, thus providing a possible explanation for its ion transport role and putative pH sensitivity. This solvent-inaccessible pocket appears conserved in all HCO 3 Ϫ -transporting Slc4 proteins. Thus, this work brings to light a new structural domain critical for HCO 3 Ϫ transport in the Slc4 proteins.