Structural and functional characterization of transmembrane segment IV of the NHE1 isoform of the Na+/H+ exchanger.

The Na(+)/H(+) exchanger isoform 1 is a ubiquitously expressed integral membrane protein that regulates intracellular pH in mammals. We characterized the structural and functional aspects of the critical transmembrane (TM) segment IV. Each residue was mutated to cysteine in cysteine-less NHE1. TM IV was exquisitely sensitive to mutation with 10 of 23 mutations causing greatly reduced expression and/or activity. The Phe(161) --> Cys mutant was inhibited by treatment with the water-soluble sulfhydryl-reactive compounds [2-(trimethylammonium)ethyl]methanethiosulfonate and [2-sulfonatoethyl]methanethiosulfonate, suggesting it is a pore-lining residue. The structure of purified TM IV peptide was determined using high resolution NMR in a CD(3)OH:CDCl(3):H(2)O mixture and in Me(2)SO. In CD(3)OH: CDCl(3):H(2)O, TM IV was structured but not as a canonical alpha-helix. Residues Asp(159)-Leu(162) were a series of beta-turns; residues Leu(165)-Pro(168) showed an extended structure, and residues Ile(169)-Phe(176) were helical in character. These three structured regions rotated quite freely with respect to the others. In Me(2)SO, the structure was much less defined. Our results demonstrate that TM IV is an unusually structured transmembrane segment that is exquisitely sensitive to mutagenesis and that Phe(161) is a pore-lining residue.

The mammalian Na ϩ /H ϩ exchanger isoform 1 (NHE1) 1 is a ubiquitously expressed integral membrane protein that mediates the removal of one intracellular proton in exchange for one extracellular sodium ion (1). NHE1 thereby protects cells from intracellular acidification (2,3); and stimulation of its activity promotes cell growth and differentiation (2) and regulates sodium fluxes and cell volume after osmotic shrinkage (2,3). The Na ϩ /H ϩ exchanger also plays an important role in the damage that occurs to the human myocardium during ischemia and reperfusion, and it has been shown that inhibition of the exchanger has beneficial effects on the myocardium under these conditions (4). Amiloride and its derivatives are inhibitors of the NHE1 isoform of the Na ϩ /H ϩ exchanger, and a new generation of Na ϩ /H ϩ exchanger inhibitors is being developed for clinical treatment of heart disease (5).
Although the activity of NHE1 has been extensively examined in many tissues, only recently is information starting to be elucidated on how this antiporter actually binds and transports Na ϩ ions and protons. NHE1 is composed of two domains as follows: an N-terminal membrane domain of ϳ500 amino acids and a C-terminal regulatory domain of about 315 amino acids (1,4) (Fig. 1). The N-terminal membrane domain is responsible for ion movement, and it is reported to have 12 transmembrane (TM) segments and 3 membrane-associated segments (6). Transmembrane segment four (TM IV; residues 155-177) has been implicated in the ion transport and inhibitor binding properties of NHE1 (7)(8)(9). The sequence of human TM IV of NHE1 is 155 FLQSDVFFLFLLPPIILDAGYFL 177 . The underlined residues have been shown to affect Na ϩ affinity or the inhibitor resistance of mammalian NHE1 (7)(8)(9). Recently, we have shown that prolines 167 and 168 are critical forNHE1 function, targeting, and expression (10). These data provide a strong case for the importance of many amino acid residues of TM IV in the ion binding, structure, and transport properties of NHE1.
In this communication, we examine both structural and functional aspects of TM IV of the NHE1 isoform of the Na ϩ /H ϩ exchanger. We use cysteine-scanning mutagenesis to characterize which amino acids are important in function and are * This work was supported in part by funds from the Canadian Institutes of Health Research (to L. F.) and from the Canadian Institutes of Health Research and Protein Engineering Network of Centers of Excellence (to B. D. S.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The atomic coordinates and structure factors ( likely pore-lining residues of TM IV. To study the structure of TM IV, the segment was expressed and purified, and we compared its structure in two different membrane mimetic solvent conditions. This provides insight into structured regions, in our case, without fixing the segment as a whole into a single conformation. The use of solution conditions having a low dielectric constant to mimic a membrane environment has become quite common for structural studies of transmembrane proteins or peptides and has provided structures of isolated protein segments consistent with their structures in the full protein (11)(12)(13). Our study demonstrates that whereas TM IV is structured, only a 4 -6-residue stretch of the segment is helical. In addition, we identify Phe 161 as a pore-lining residue in NHE1.

EXPERIMENTAL PROCEDURES
Materials-15 N-Labeled ammonium sulfate and all deuterated NMR solvents were from Cambridge Isotope Laboratories (Andover, MA). Pwo DNA polymerase was from Roche Applied Science. Lipo-fectamine™ 2000 Reagent was from Invitrogen. MTSET and MTSES were from Toronto Research Chemicals (Toronto, Ontario, Canada).
Site-directed Mutagenesis-Mutations in TM IV were made to an expression plasmid containing a hemagglutinin (HA)-tagged human NHE1 isoform of the Na ϩ /H ϩ exchanger. The plasmid pYN4-C contains the cDNA of the entire coding of NHE1 with all 10 native cysteine residues mutated to serine as described earlier (10). Each residue of TM IV was individually mutated to cysteine using the cysteine-less pYN4-C as a template. Site-directed mutagenesis was performed using amplification with Pwo DNA polymerase followed by use of the Stratagene (La Jolla, CA) QuikChange TM site-directed mutagenesis kit as recommended by the manufacturer. Mutations were designed to create a new restriction enzyme site for use in screening transformants. DNA sequencing confirmed the accuracy of the mutations.
Cell Culture and Stable Transfection-AP-1 cells were used to examine Na ϩ /H ϩ exchanger expression and activity. These cells lack an endogenous Na ϩ /H ϩ exchanger. Stable cell lines were made of all mutants by transfection with Lipofectamine TM 2000 Reagent according to the manufacturer's protocol as described earlier (10). Transfected cells were selected using 800 g/ml geneticin (G418), and stable cell lines for experiments were regularly re-established from frozen stocks at passage numbers between 5 and 15.
SDS-PAGE and Immunoblotting-To confirm NHE1 expression immunoblot, analysis was used on samples from total cell lysates of AP-1 cells. Cell lysates were made as described earlier (10). For Western blot analysis, equal amounts of each sample (100 g total protein) were resolved on 10% SDS-polyacrylamide gels. The gel was transferred onto a nitrocellulose membrane and immunostained using anti-HA monoclonal antibody (Roche Applied Science) and peroxidase-conjugated goat anti-mouse antibody (Bio/Can, Mississauga, Ontario, Canada). The enhanced chemiluminescence Western blotting and detection system (Amersham Biosciences) was used to visualize immunoreactive proteins. Densitometric analysis of x-ray films was carried out using NIH Image 1.63 software (National Institutes of Health, Bethesda).
Cell Surface Expression-Cell surface expression was measured essentially as described earlier (10). Briefly, cells were labeled with Sulfo-NHS-SS-Biotin (Pierce), and immobilized streptavidin resin was used to remove cell surface-labeled Na ϩ /H ϩ exchanger. Equivalent amounts of the total and unbound proteins were analyzed by SDS-PAGE, and Western blotting was as described above. The relative amount of NHE1 on the cell surface was calculated by comparing both the 110-and the 95-kDa species of NHE1 in Western blots of the total and unbound fractions. Results are shown as mean Ϯ S.E. with statistical significance determined by a Mann-Whitney U test.
Na ϩ /H ϩ Exchange Activity and Sulfhydryl-reactive Reagent Inhibition-Na ϩ /H ϩ exchange activity was measured using a PTI Deltascan spectrofluorometer or a Schimadzu RF 5000 spectrofluorometer. The initial rate of Na ϩ -induced recovery of cytosolic pH (pH i ) was measured after acute acid load using 2Ј,7-bis(2-carboxyethyl)-5 (6) carboxyfluorescein-AM (Molecular Probes Inc., Eugene, OR). Ammonium chloride (50 mM for 3 min) was used to transiently induce an acid load, and the recovery in the presence of 135 mM NaCl was measured as described previously (10). There were no differences in buffering capacities of stable cell lines as indicated by the degree of acidification induced by ammonium chloride. Results are shown as mean Ϯ S.E., and statistical significance was determined using a Mann-Whitney U test.
To test the effect of MTSET and MTSES on activity of the NHE1 mutants, we used the standard Na ϩ /H ϩ exchanger assay with ammonium chloride-induced acidification of the cells. In this case cells were acidified two times as described above. After a first control acidification and recovery, either MTSET or MTSES was added to a final concentration of 10 mM for 10 min in Na ϩ -free buffer. The cells were subsequently washed three times in Na ϩ -free buffer prior to the second ammonium chloride-induced acidification and recovery. To calculate residual activity, Equation 1 was used, % residual activity ϭ pH change after (reagent) pH change without (reagent) ϫ 100% (Eq. 1) Production and Purification of TM IV-We produced TM IV as a fusion protein with the immunoglobulin binding domain of streptococcal protein G (GB1 domain). DNA encoding the residues of human NHE1 amino acids 155-180 ( 155 FLQSDVFFLFLLPPIILDAGYFLPLR-180 ) was synthesized synthetically in the core facility of the Department of Biochemistry. Two self-annealing primers were designed to flank the protein sequence with Met residues and to change the codon preference to that favored in Escherichia coli. The primers were 5Ј-CTAGCATGTT TCTGCAGAGCGATGTGTTTTTTCTGTTTCTGCTGCCGCCGATTAT-TCTGGATGCGGGCTATTTTCTGCCGCTGCGCATGCCCGGGC-3Ј and 5Ј-TCGAGCCCGGGCATGCGCAGCGGCAGAAAATAGCCCGCA-TCCAGAATAATCGGCGGCAGCAGAAACAGAAAAAACACATCGCT-CTGCAGAAACATG-3Ј. The primers were designed to have overhanging "sticky" NheI and XhoI ends so as to insert into the GEV-1 vector (14,15). The annealed primers were ligated into the cut vector and transformed in E. coli XL1Blue and then later transformed into E. coli (BL21 DE3 pLysS) for protein production. The correct clones were verified by restriction enzyme digests and DNA sequencing. Cultures were induced with isopropyl 1-thio-␤-D-galactopyranoside and treated to release the soluble protein as described earlier (14). For routine production of the protein, cultures were started from a single colony and grown in logarithmic stages until 2.0 liters. At A 600 of 0.8, cultures were induced with 1 mM isopropyl 1-thio-␤-D-galactopyranoside for 5 h at 37°C. Bacterial pellets were resuspended in 40 ml of 1ϫ phosphatebuffered saline with 1% Triton X-100. The fusion protein was constructed such that a His 6 tag flanked the C-terminal Met. This was used for purification of the protein via nickel-nitrilotriacetic acid affinity chromatography as described by the manufacturer (Qiagen). The fusion protein was eluted from the column using 1 M imidazole. After purification, the eluted proteins concentrated using an Amicon concentrator and resuspended in 150 mM NaCl, 50 mM Na 2 HPO 4 . They were then desalted by using a Sephadex G-25 column in 10 mM NH 4 HCO 3 , pH 8.0, and then were lyophilized to dryness. To cleave the transmembrane segment IV free of GB1, standard techniques were employed. Briefly, a given purified batch of fusion peptide was lyophilized twice out of 0.05% (v/v) trifluoroacetic acid (trifluoroacetic acid; 99%, Aldrich) in water, dissolved to ϳ2 mg/ml in 50% (v/v) trifluoroacetic acid in water with excess cyanogen bromide (97%, Aldrich), and allowed to react in darkness for ϳ16 h at room temperature. The reaction mixture was quenched with an equal volume of deionized water and subsequently lyophilized. Reverse-phase HPLC was used to separate the various peptide products of the cleavage, and fractions containing the TM IV segment were pooled and lyophilized at least twice prior to use. The correct fragment identity was determined by MALDI mass spectrometry.
Preliminary experiments produced unlabeled TM IV. Later experiments produced labeled TM IV for more detailed characterization of the peptide structure. For these experiments, E. coli cells were grown in a minimal medium containing 87 mM NaH 2  Samples for extended analysis were prepared in Me 2 SO and in CD 3 OH:CDCl 3 :H 2 O (4:4:1 v/v/v) at ϳ2 mM peptide and 1.0 mM DSS. One-dimensional 1 H, natural abundance 15 N HSQC, TOCSY (60-ms mix; DIPSI spin lock), and NOESY (150-ms mix) experiments were acquired on the NANUC Varian INOVA 800 MHz spectrometer for each sample. Double-quantum filtered correlation spectroscopy and NOESY spectra at various mixing times were acquired at 600 MHz on either Varian UNITY or INOVA spectrometers. By using uniformly 15 N-labeled peptide, the three-dimensional 15 N-edited NOESY-HSQC (150-ms mixture), TOCSY-HSQC (57-ms mixture, DIPSI spin lock), and HNHA experiments were acquired at 500 MHz on a Varian Inova spectrometer. Spectra were processed either using VNMR (Varian Inc., Palo Alto, CA) or NMRPipe (17); spectral analysis was facilitated with Sparky 3 (T. D. Goddard and D. G. Kneller, University of California, San Francisco).
Structure calculation was carried out in CNS version 1.1 (18) using NOE contacts derived from the 150-ms mixing time experiments at 800 MHz and 3 J-coupling constants derived from the HNHA experiments at 500 MHz as described by Vuister and Bax (19). Homonuclear NOESY and HNHA peaks were manually picked in Sparky, and volumes were calculated using Gaussian fits, with motion of the peak center generally allowed; in some cases (Ͻ2%) with the NOESY spectra, Sparky's gaussian fit algorithm did not find a convergent solution, and a summed signal intensity was used instead over a manually specified region. Initial NOE calibration was carried out empirically from peak volumes to provide a value in the range of 1.8 -5.0 Å. These estimates were used to bin each contact into one of strong (1.8 -2.8 Å), medium (1.8 -3.6 Å), or weak (1.8 -5.0 Å). Peaks with an ambiguous assignment between two possible contacts were assigned to the more conservative contact (where possible) as a starting point.
An iterative procedure was employed to refine and prune the NOE contacts using a philosophy similar to that of Wang et al. (20). Families of 500 -1000 structures were generated by simulated annealing using the default CNS heating and cooling cycles and NOE energetics. NOE violations were analyzed over each ensemble, and any violations observed in more than ϳ10% of the ensemble structures were examined in detail. If unambiguous, these were typically lengthened, and if ambiguous, they were reassigned or discarded. During this process, restraints were extended to a fourth "very weak" category (1.8 -6.0 Å) as necessary (20). After all major NOE violations were resolved, the three-bond J-coupling values were incorporated directly into the simulated annealing protocol (21) using Karplus coefficients as specified by Vuister and Bax (19) for the HNHA experiment for the more structured CD 3 OH: CDCl 3 :H 2 O condition. A further pair of iterations was carried out to optimize the compatibility between 3 J-coupling values and NOE restraints. The homoserine lactone residue was incorporated into structure calculations using a protonated residue structure attached to an N-terminal methyl ketone group calculated by MMϩ molecular mechanics with bond dipole electrostatics with atomic charges obtained by Restricted Hartree-Fock calculation in HyperChem 3 (Hypercube, Gainesville, FL). XPLO2D 3.2.2 (22) was used to produce CNS topology and parameter files. Structurally convergent segments of the peptide were determined by examining the lowest energy 600 structures out of the final ensemble of 1000 with NMRCORE 1.0 (23); structural features of the 600-member ensemble were analyzed using PROMOTIF-NMR 3.0 (24). The final sets of restraints have been deposited in the Protein Data Bank (entry 1Y4E) along with the 100 lowest energy structures for the CD 3 OH:CDCl 3 :H 2 O solvent condition. Fig. 1 illustrates a general model of the Na ϩ /H ϩ exchanger (Fig. 1A) and a schematic model illustrating TM IV (Fig. 1B). To examine which amino acids of TM IV were critical for the activity of the Na ϩ /H ϩ exchanger and which amino acids were pore-lining, we used the cysteine-less Na ϩ /H ϩ exchanger (cNHE). Each residue in TM IV of cNHE1 was individually mutated to a cysteine residue. Initial experiments examined whether these mutant forms of the Na ϩ /H ϩ exchanger had activity. Most surprisingly, TM IV was exquisitely sensitive to mutation. Fig. 2 shows the effects of mutation of amino acids of TM IV to cysteine on NHE activity. Substitution of any of the amino acids with cysteine in TM IV resulted in significant reductions of the measurable activity of the protein in all mutants (p Ͻ 0.05). In particular, residues Phe 155 , Leu 156 , Ser 158 , Asp 159 , Phe 162 , Phe 164 , Pro 167 , Pro 168 , Asp 172 , Tyr 175 , and Phe 176 retained less than 20% of the control activity. These mutants were not used for further characterization of activity. We have demonstrated previously that Pro 178 is not critical for function of NHE1 (10). Fig. 3 illustrates a Western blot of total cell extracts from AP1 cells stably expressing the single cysteine mutants. Both the mutant and wild-type exchangers displayed the same pattern of immunoreactive bands, with a larger band at ϳ110 kDa that represents the glycosylated form of the mature Na ϩ /H ϩ exchanger and a smaller band at ϳ95 kDa that represents an immature form of the exchanger that is not fully glycosylated (10). Both the native wild-type NHE and the cysteine-less NHE1 showed strong immunoreactive bands of 110 kDa size and a weaker 95-kDa, unglycosylated NHE1. The amount of mature 110-kDa NHE is quantified below each lane in We examined the sensitivity to MTSET or MTSES of the mutant Na ϩ /H ϩ exchangers that had greater than 20% residual activity of the cysteine-less NHE1 (Fig. 4). Of the active Na ϩ /H ϩ exchangers, only F161C was affected by treatment with either MTSET or MTSES. This resulted in decreases in Na ϩ /H ϩ exchanger activity of ϳ60 and 80%, respectively. To The orientation of TM segments I-XII of the NHE1 isoform of the Na ϩ /H ϩ exchanger is illustrated (6). B, crude model of amino acids present in TM IV. confirm that Phe 161 was a critical residue, we created the mutant F161K, which introduced a positive charge into this location in the wild-type NHE1 background. This mutant retained only ϳ23% of wild-type NHE activity (Fig. 2), had reduced expression (not shown), and reduced surface processing (Table I) relative to wild-type NHE1 supporting the importance of Phe 161 in NHE1 function.

Activity and Expression of NHE1 Mutants-
Subcellular Localization of the Mutant and Wild-type Na ϩ /H ϩ Exchangers-We have earlier found that mutation of amino acids of the Na ϩ /H ϩ exchanger sometimes causes the protein to be targeted to an intracellular location (10,25). We therefore used quantitative measurement of the intracellular localization of NHE1 within AP-1 cells. Cells were treated with sulfo-NHS-SS-biotin and then lysed and solubilized, and labeled proteins were bound to streptavidin-agarose beads. An equal amount of the total cell lysate and unbound lysate was separated by size using SDS-PAGE followed by Western blotting with anti-HA antibody to identify tagged NHE1 protein. Fig. 5 illustrates examples of the results, and Table I summarizes the results quantitatively. In Fig. 5, the total lanes (T) and intracellular lanes (I) illustrate the NHE1 protein present in these samples. The intracellular lane I was that fraction of the Na ϩ /H ϩ exchanger that did not bind to the streptavidinagarose beads. Cysteine-less NHE1 and most of the mutant exchanger proteins were predominantly present on the plasma membrane in similar amounts. Na ϩ /H ϩ exchanger mutants D159C and D172C were significantly reduced in expression on the plasma membrane of the cell. Correcting the activity of these mutants for changes in targeting (Fig. 2) demonstrated that a significant amount of the apparent reduction in activity was because of mistargeting of these proteins. Nevertheless, both still retained less than half of the activity of the cysteineless NHE1 protein. Correcting for targeting did not greatly change the values for activity of other mutants with the next largest changes being increases in activity of about 20% for L166C and I170C proteins.
Production and Characterization of TM IV-To examine the structure of TM IV, we produced it as a fusion protein with GB1. On SDS-PAGE, the fusion protein appeared as a band of ϳ8 kDa in size (not shown). This was slightly less than the predicted size of 11 kDa. We confirmed the identity of the protein by mass spectrometry, suggesting that the protein ran with a somewhat anomalous weight, typical of many membrane proteins. The yield of the GB1-TM IV fusion protein was typically ϳ25 mg/liter of cells. After CNBr treatment to free TM IV, it was then purified by reverse phase HPLC. We confirmed the identity of TM IV by mass spectroscopy. For the unlabeled TM4, the expected mass (with a C-terminal homo- Na ϩ /H ϩ exchanger activity was measured after transient induction with an acid load as described under "Experimental Procedures." The activity of cNHE1 stably transfected with NHE1 was 3.2 pH units/min, and this value was set to 100%. All mutations to cysteine were done in the background of the cysteine-less NHE1, and activities are percent of those of cNHE. The F161K mutation was done in the wild-type NHE1 background, and its activity is expressed as a percent of the wild-type NHE1 activity (4.25 pH units/min). All results are means Ϯ S.E. of at least 10 determinations from two independently made stable cell lines. Results are shown for mean activity both uncorrected (black) and normalized for surface processing (hatched). * indicates mutants with unnormalized activity that is less than 20% of cNHE1. serine lactone) is 3138.8 atomic mass units, and the observed mass in retained preparative HPLC fractions ranged over 3138.4 -3134.0 atomic mass units (parent peak mass differed slightly from fraction to fraction; ionic adducts also observed). Impurities at 4405-4411 and 9910 -9915 atomic mass units were detected in less than 25% of the total pooled fractions each, at intensities ranging from 0.5 to 2% relative to the parent peak plus ionic adducts. For the uniformly 15 N-labeled peptide, the expected mass was 3165.8, and observed mass range was 3165.6 -3171.6 atomic mass units. Impurities at mass 3085-3086, 4428 -4430, and 4816 -4820 atomic mass units were detected in less than 10% of the total pooled fractions each, at intensities ranging from 1 to 4% relative to the parent peak plus ionic adducts. Purity was therefore estimated to be greater than 95% for each of the peptides after pooling of the HPLC fractions.
Choice of NMR Conditions-To determine the structure of TM IV, we tested several solvent mixtures. The trifluoroethanol:H 2 O mixtures and methanol appeared more promising than either the CD 3 OH:CDCl 3 :H 2 O mixture or Me 2 SO, in that wider dispersions of chemical shifts were observed in the backbone H N (7-9 ppm) and H ␣ (3.5-5 ppm) regions of one-dimensional 1 H spectra (data not shown). Unfortunately, the peptide precipitated out of each of these solutions within days, making multidimensional NMR data acquisition impractical. In the range of 5-40°C, 30°C gave rise to well resolved, minimally broadened 1 H spectra (data not shown). Therefore, structural studies were carried out in CD 3 OH:CDCl 3 :H 2 O and Me 2 SO at 30°C. These ϳ2 mM peptide samples with 0.5 mM DSS as the internal chemical shift reference are stable at room temperature, providing reproducible spectra over the course of several months.
Resonance Assignment-Standard sequential assignment methods were applied (26), using TOCSY and double-quantum filtered correlation spectroscopy experiments along with H N values observed by natural abundance 15 N-HSQC experiments for spin-system assignment and NOESY connectivity data to walk through the sequence. This provided assignment of all backbone H N and H ␣ , except for the N-terminal H N , all side chain H ␤ resonances, and most other side chain proton resonances. The C-terminal residue was identified as a homoserine lactone rather than free homoserine, because its H ␣ and side chain 1 H chemical shifts correspond to ␣-amino-␥-butyrolactone rather than to L-homoserine (27). This was upheld by the molecular weight indicated by MALDI-MS of the purified peptide when considering the fact that homoserine lactone residues are not subsequently undergoing conversion to the free homoserine moiety, because this reaction is reported to be very slow at the peptide C terminus even with catalytic basic conditions (28). Due to spectral overlap even at 800 MHz, caused primarily by repetition in the primary sequence, a few side chain protons could not be resolved. In

TABLE I Summary of subcellular localization in AP-1 cells of wild type (wNHE1), cysteine-less NHE1 (cNHE1), and
TM IV Na ϩ /H ϩ exchanger mutants Analysis of localization was as described under "Experimental Procedures." The percent of the total NHE1 protein that was found within the plasma membrane is indicated. The results are mean Ϯ S.E. of at least three determinations.  Structure Calculation-Homonuclear NOESY experiments were used exclusively to define NOE contacts between protons. A variety of mixing times was examined under each solvent condition (100 -450 ms; data not shown). In each of the solvents, the 150-ms mixing time provided peaks of sufficient intensity to analyze without apparently extensive spin diffusion. Therefore, this mixing time was employed for quantitative peak volume analysis. A large number of intra-and interresidue contacts were observed in each solvent condition. In Me 2 SO, 302 intra-molecular, 244 sequential, and 62 medium range contacts were unambiguously assigned. In CD 3 OH: CDCl 3 :H 2 O, there were 222 intra-residue, 140 sequential, and 96 medium range unambiguous contacts assigned. More contacts were able to be assigned in the Me 2 SO conditions due to the additional Phe and Tyr ring proton assignments. In CD 3 OH:CDCl 3 :H 2 O, the TM IV sample showed two regions with significant i,i ϩ 3 contacts, often indicative of ␣-helical character, which were not apparent in Me 2 SO (Fig. 6). The first of these regions is coincident with a large number of i,i ϩ 2 contacts, however, which together indicate potential ␤-turns. Chemical shifts have not at present been incorporated into the structural calculations because a quantitative comparison to random coil or BioMagResBank average chemical shifts observed in aqueous solution may be altogether incorrect. 3 Jcoupling constants extracted from 15 N-edited HNHA experiments were used where possible to aid in specifying -dihedral angles during the simulated annealing protocol. Because the peptide was less structured in Me 2 SO than in CD 3 OH:CDCl 3 : H 2 O, as evinced by the greatly reduced number of medium range NOE contacts (Fig. 6), we analyzed in detail only the CD 3 OH:CDCl 3 :H 2 O structural features present. The characteristics of the final ensemble of retained structures are given in Table II. DISCUSSION Functional Analysis of TM IV-Transmembrane segment four (TM IV; residues 155-177) of the NHE1 isoform of the Na ϩ /H ϩ exchanger has been implicated in the ion transport and inhibitor binding properties of the protein (7-9). We have recently shown (10) that the double proline pair Pro 167 and Pro 168 is critical for the function of NHE1. They were suggested to be critical in the maintenance of an appropriate structure of TM IV and necessary for normal NHE1 transport function. One method to determine the functional role of individual amino acids of a transmembrane segment of a membrane protein is cysteine-scanning mutagenesis in combination with reaction with sulfhydryl reagents. Cysteine-scanning mutagenesis takes advantage of the fact that the sulfhydryl moiety is the most reactive functional group in a protein (29). It has been used to determine pore-lining residues in numerous membrane proteins (30 -32), including the lactose permease of E. coli (33), the mouse acetylcholine receptor (34), the human glucose transporter Glut1, and the human anion exchanger isoform 1 (AE1) (35). Cysteine-scanning mutagenesis uses the highly reactive sulfhydryl moiety to determine the accessibility of side chains of amino acids. Two sulfhydryl-reactive reagents (36) that are often used in these studies are MTSET and MTSES, which are membrane-impermeant (37,38), react with porelining residues surrounded by water, and cannot reach residues within the hydrophobic bilayer.
Initial experiments substituted each of the amino acids in TM IV of cysteine-less NHE1 with cysteine residues. We found that TM IV was exquisitely sensitive to mutation. Twelve of the cysteine mutants had less than 20% of the activity of the control cNHE1. Western blotting revealed that in some cases the reduction in activity was due to a lower level of expression of the protein (Fig. 3); however, in most cases correction for the amount of plasma membrane protein (Table I and Fig. 2) did not explain the loss of activity. Several of the mutants were expressed mainly as unglycosylated protein, and many but not all of these had greatly reduced activity. We have found previously that mutation of some amino acids can greatly affect the glycosylation levels of the proteins. We examined if these mutations affected the surface targeting of the protein. Our findings that many of the mutations of TM IV affected targeting, expression, or activity are similar to the results found with the human anion exchanger (35) and with TM XI of lactose permease (33). In contrast, for the tetracycline-resistant transporter of E. coli (39) and in P-glycoprotein (40), it was possible   to mutate all the amino acids of a TM segment and always retain appreciable activity. This susceptibility to mutation appears to vary not only between proteins but also within different transmembrane segments of the same protein. TM XI of lactose permease had several sensitive amino acids (33), whereas TM XII only had one (41). This property could be reflective of both the importance of the residues in the particular segment, and of the importance of the segment itself to the structure and function of the protein. In our case, it was clear that there was a relatively strict requirement for many of the amino acids of TM IV, and this could be indicative of precise structural and functional roles. Treatment with both positively charged MTSET and negatively charged MTSES revealed that Phe 161 is accessible to and reacts with these sulfhydryl-reactive reagents (Fig. 4). The most likely explanation for this reactivity is an interaction at a site that lines the ion translocation pore. The reagents then at least partially block the pore and inhibit ion transport (35). Thus, Phe 161 in TM IV is the first residue to be unambiguously identified as lining the ion transport pore of NHE1. Mutation of Phe 161 to Lys resulted in loss of most of NHE1 activity (Fig. 2), confirming that this is a critical amino acid.
Structural Analysis of TM IV-We examined the structure of TM IV in CD 3 OH:CDCl 3 :H 2 O and in Me 2 SO at 30°C. As with many peptides studied by NMR (20,42) the TM IV segment does not, as a whole, assume a single conformation. Rather, sections within the peptide converge structurally. Consideration of these sections relative to each other then provides insight into the TM IV segment overall. In a given state of the intact NHE1 protein, a single conformation would presumably be preferred by the TM IV segment. However, the structures that we present here represent the most energetically favorable forms of this TM segment in a low dielectric environment mimicking both a lipid bilayer and a protein interior. Therefore, it is reasonable to presume that they would also be favorable within the setting of the NHE1 protein. Indeed, a number of studies have demonstrated strong correspondence between structures of peptides or protein segments obtained in membrane mimetic solvents to structures obtained by solution state NMR in micelles or to entire membrane proteins determined by x-ray crystallography (11)(12)(13). Most interestingly, we find that CD 3 OH:CDCl 3 :H 2 O provides a well defined peptide structure, whereas the peptide in Me 2 SO appears much less structured. This implies that there is a great deal of value in determining membrane peptide structures in multiple solvent conditions, because a single solvent may not induce sufficient sampling of the structured state.
Three sections of four to nine residues of the TM IV segment converge structurally and were defined as core regions by NMRCORE (23) and confirmed by analysis of r.m.s.d. values (Table III). Namely, the stretches of residues Asp 159 -Leu 163 , Leu 165 -Pro 168 , and Ile 169 -Phe 176 each converge. These are illustrated in Fig. 7 (a-c), and their relationship is demonstrated schematically in Fig. 7d. Each section superposes extremely well, with all convergent residues identified in Fig. 7d for the 600-member ensemble having a C ␣ r.m.s.d. relative to the average structure in the range of 0.25-0.85 Å. The heavy backbone and side chain atom r.m.s.d. are Ͻ1.5 Å over Asp 159 -Leu 163 and Ile 169 -Gly 174 as well as for Leu 165 and Pro 167 (Table  III), which implies good structural convergence at the side chain level for these segments. The boundary points of converged stretches are also very clear upon examination of the r.m.s.d. values in Table III Fig. 7e as a representative structure. Given that the segment must span a membrane with the width on the order of 25-35 Å, a relatively extended ensemble member such as that pictured in Fig.  7e is more representative of the expected biological configuration than one that curls back upon itself. This particular structure must be considered as only demonstrative of the relation between the convergent segments of the TM IV segment, rather than as representative of any global structuring.
The first feature of note is that the TM IV segment certainly does not resemble a canonical transmembrane ␣-helix. This is despite the fact that there are helix-capping sequences inherent within the TM IV over the segment studied. The N-terminal sequence FLQSDV (155-160) fits well with a type Ib cap, whereas the sequence LDAGYF (171-176) fits a type Va Cterminal cap, with capping nomenclature as defined by Aurora and Rose (43). These cap sequences would imply that a helix should form between Leu 156 and Asp 173 . However, the NMR data readily demonstrate a lack of helicity, even at relatively long NOESY mixing times. Contiguous i,i ϩ 3 and i,i ϩ 4 contacts indicative of an ␣-helix (26) are not observed. In fact, as implied in Fig. 7d, only a very short stretch of the segment even approaches a helical structure. This should not be taken to imply a lack of structure; in fact, eight type IV ␤-turns occur in a major proportion of ensemble members, with four found in more than 90% of the structures. Note that this means a frequent observance of "disallowed" / angles would actually be expected (44), and correspondingly, a PROCHECK-NMR (45) analysis places 12.2% of the non-Pro and Gly residues in the disallowed region of the Ramachandran plot.
Of the three core regions defined by convergent backbone atoms, residues Asp 159 -Leu 163 are the most affected by these ␤-turns. Asp 159 is the i ϩ 2 residue in one type IV turn, and the first residue in a second, Val 160 , is the i ϩ 3 and i ϩ 1 residue for these turns, as well as the first residue in a third turn; following along the sequence, Phe 161 and Phe 162 both participate in the turns initiated at Asp 159 and Val 160 , and Leu 163 participates as the i ϩ 3 residue for Val 160 . Therefore, this series of turns, observed in 81-94% of the ensemble, serves to converge this five-residue stretch. Most interestingly, Leu 165 -Pro 168 are nowhere near as strongly influenced by ␤-turns. Rather, these residues appear to be quite extended (Fig. 7b), with Pro 168 initiating type I or IV ␤-turns in ϳ85% of the ensemble. Finally, residues Ile 169 -Phe 176 converge strongly, providing the only appreciable ␣-helical character over the entire transmembrane segment. The helical stretch extends over 4 -6 residues for ϳ77% of the ensemble members, starting between Ile 169 and Leu 171 and ending at Gly 174 , with each one of these members of the ensemble showing helical character over the four-residue segment of Leu 170 -Gly 174 . The C terminus of this convergent stretch is defined by Gly 174 and Tyr 175 , each of which act as the first residue for a sequential pair of type IV ␤-turns in 92% or more of the ensemble members. Although similar experimental resolution in NMR studies has been achieved using micelles in aqueous solution (46), such an environment would surround the TM IV peptide with hydrophobic detergent tails. In its natural setting, this segment of the NHE1 protein would contact lipid tail groups, surrounding polypeptide segments, and a pore region with a relatively high dielectric constant. In the case where a transmembrane segment has a very clearly defined structure, such as a canonical membrane-spanning ␣-helix, the micellar environment serves as an excellent conformational stabilizer. In the micellar state, we would anticipate that the allowed rotations about Phe 164 and Pro 168 /Ile 169 would be greatly reduced. However, this may actually be an artificial constraint upon the peptide structure relative to the solvent setting, which allows increased conformational sampling. We feel that the present structure in a mixture of solvents representing a variety of dielectrics provides a reliable representation of the best structured regions of the TM IV segment. This should be useful in interpretation of future experiments using more complete sections of NHE1, in particular for solid state NMR experiments that are aided by knowledge of the most likely polypeptide backbone configurations (47).
Correlating Structural and Functional Data-In light of the structural data, plausible explanations for the loss in activity observed for some of the most severe cysteine-scanning mutants can be advanced. Mutations to P167C and P168C practically abrogate all exchanger activity, and mutations of these prolines to other residues did not allow for a return to a normally functioning Na ϩ /H ϩ exchanger (10). Mutation from an imino to a much freer amino acid moiety would readily modify the structural motif observed at this pair of prolines (Fig. 7b). Given the loss of activity, it is likely that the extended structural nature of this motif is crucial in NHE1 folding into the functional form. The Y175C mutation also causes almost complete loss in activity. Most interestingly, Tyr 175 is observed to participate as the second residue in an extremely well converged double (i,i ϩ 1) type IV ␤-turn, just N-terminal to the aromatic residue Phe 176 . Note that F176C also causes significant loss of activity. Analysis of the torsion angles defined between the C ␣ -C ␤ , -C ␥ , and -C vectors of Tyr 175 and Phe 176 for the ensemble of 600 structures shows that the aromatic side chain moieties of Tyr 175 and Phe 176 have a strong tendency to fall within ϳ50°of each other relative to the polypeptide backbone over this region of the TM IV segment. Because these are large, hydrophobic side chains, it seems likely that they would be at the interface with neighboring transmembrane segments or protruding into the lipid bilayer rather than directly into the channel environment. Substitution of these aromatic groups with the significantly less bulky, polar cysteine side chain would disrupt these interactions, thereby disrupting the NHE1 structure as a whole, leading to the observed loss in activity. However, the mutated F161C side chain does appear to be pore-exposed, so this is not a certainty.
The mutation F161C allows for reactivity with the sulfhydryl-reactive reagent MTSET that inhibits Na ϩ /H ϩ exchanger activity. This implies that residue 161 is facing the NHE1 translocation pore. With this in mind, examination of the structure of the TM IV segment provides a very interesting revelation. Phe 161 side chain is actually fully exposed when put in the context of residues Phe 155 -Ile 169 in every member of the ensemble (data not shown). Because rotation of the structurally convergent sections is observed about Phe 164 (Fig. 7d), any further analysis of the two C-terminal domains relative to Phe 161 would be purely speculative, and it was therefore not possible to determine at this stage which other C-terminal amino acids face the translocation pore. However, the exposed nature of the Phe 161 side chain at the extracellular end of the segment is demonstrated in all 600 retained ensemble members satisfying the experimental constraints, meaning that the isolated TM IV segment in membrane mimetic solvent assumes a structure consistent with the reactivity observed for the exchanger as a whole. Over the convergent stretch of Asp 159 -Leu 163 (Fig. 7a), the Asp 159 side chain is uniformly observed on the same face of the peptide. Therefore, it seems likely that the anionic Asp 159 side chain would also protrude into the pore of the exchanger, allowing it to act in cation coordination during translocation. Mutation of this amino acid eliminated activity and reduced expression of the protein, consistent with a critical role for this amino acid. We have previously postulated that polar residues within transmembrane segments of the Na ϩ /H ϩ exchanger are important in cation coordination and transport (48,49), and the structure of TM IV supports such a role for this polar residue.
Our study has given the first detailed structural and functional information on TM IV of the NHE1 isoform of the Na ϩ /H ϩ exchanger. TM IV appears to be very sensitive to alterations in its amino acid sequence. We show that although it is a well structured transmembrane segment, its structure is uniquely different from a typical ␣-helix. Possibly this reflects its pivotal role in cation binding and transport. Phe 161 and possibly Asp 159 are amino acid residues that protrude into the pore of the channel.