Four Na+/H+ Exchanger Isoforms Are Distributed to Golgi and Post-Golgi Compartments and Are Involved in Organelle pH Regulation*

Four isoforms of the Na+/H+ exchanger (NHE6–NHE9) are distributed to intracellular compartments in human cells. They are localized to Golgi and post-Golgi endocytic compartments as follows: mid- to trans-Golgi, NHE8; trans-Golgi network, NHE7; early recycling endosomes, NHE6; and late recycling endosomes, NHE9. No significant localization of these NHEs was observed in lysosomes. The distribution of these NHEs is not discrete in the cells, and there is partial overlap with other isoforms, suggesting that the intracellular localization of the NHEs is established by the balance of transport in and out of the post-Golgi compartments as the dynamic membrane trafficking. The overexpression of NHE isoforms increased the luminal pH of the compartments in which the protein resided from the mildly acidic pH to the cytosolic pH, suggesting that their in vivo function is to regulate the pH and monovalent cation concentration in these organelles. We propose that the specific NHE isoforms contribute to the maintenance of the unique acidic pH values of the Golgi and post-Golgi compartments in the cell.

The luminal ionic composition of intracellular compartments differs from the cytoplasm, and each compartment is characterized by a unique, organelle-specific ion concentration. This specific ionic composition is thought to be an important determinant for organelle function and is maintained by the concerted action of ion transport carriers on the membrane (1,2). Organelles of the secretory and endocytic pathways exhibit differential weak acidity in their lumen with a gradient of pH values decreasing toward the trafficking destination, from ER 1 (pH ϳ7.1) to Golgi (pH ϳ6.2-7.0), trans-Golgi network (TGN) (pH ϳ6.0), and secretory granules (pH ϳ5.0) and from early and late endosomes (pH ϳ6.5) to lysosomes (pH ϳ4.5) (1,3,4). This progressive acidification is essential for compartmentalizing cellular events, such as post-translational modifications, sorting of newly synthesized proteins into the secretory path-way, and the degradation or recycling of internalized ligandreceptor complexes and fluid-phase solutes in the endocytic pathway (3,5). Even pH differences of less than 0.5 between organelles can be essential for the compartmentalizing cellular events (6).
The differential ionic milieu of the organelles is maintained by a suite of ion carriers on the membrane, including pumps, channels, and transporters. Luminal acidity is primarily generated by the vacuolar-type H ϩ -translocating ATPase (V-ATPase) (4,5). Because the V-ATPase is electrogenic, the pumping activity could be affected by membrane potential and availability of permeant counterions such as chloride and potassium. In vitro studies using isolated endosomes and Golgi suggested a critical role for Cl Ϫ in shunting the inside-positive membrane potential generated by H ϩ pumping (7)(8)(9). However, although Cl Ϫ and K ϩ serve as counterions for H ϩ pumping, their conductances on the Golgi complex and the TGN were large compared with the rate of H ϩ pumping, arguing against modulation of Cl Ϫ and K ϩ conductances as the mechanism for pH regulation and suggesting that the electrical potential across the membrane is not a determinant of steady-state pH in the Golgi and TGN (10 -14). Instead, the steady-state pH is thought to be controlled by the balance between the rate of H ϩ pumping by V-ATPase and the magnitude of H ϩ leak from the organelle lumen. The H ϩ leak mechanism is postulated based on the rapid dissipation of the proton gradient across the membrane after inhibiting the V-ATPase with bafilomycin A1, however the molecular mechanism is still unknown (10,11,(13)(14)(15). The decreasing pH values of organelles along the secretory pathway is established by gradually increasing the density of active H ϩ -pumps from the ER to the Golgi while concomitantly decreasing the H ϩ permeability from ER to Golgi to secretory granules (14). Thus, the H ϩ leak acts as a key determinant of organellar pH. These data emphasize the importance of identifying the molecular components of this system, which could conceivably involve proton channels, proton-coupled cotransporters, or proton-exchanging transporters.
Na ϩ /H ϩ exchanger (NHE) proteins are integral membrane proteins that mediate electroneutral exchange of H ϩ for Na ϩ and K ϩ across the membrane, down their concentration gradients (16). NHEs are composed of an N-terminal 10 -12 membrane-spanning domain mediating the ion exchange and a large hydrophilic C-terminal domain that serves a regulatory function in ion transport (17,18). To date, eight NHE isoforms have been identified in mammals (19). NHE1-NHE5 are localized to the plasma membrane in various cell types (20 -28), and NHE6 and NHE7 reside on organellar membranes (29,30). NHE8 is the most recently identified isoform, but its cellular localization remains unknown (31). Plasma membrane NHEs are known to participate in the maintenance of intracellular pH and cell volume and are also crucial for absorption of Na ϩ across epithelia (19,32). Recently, it was reported that NHE7 on the trans-Golgi network and the plant vacuolar homologue AtNHX1 catalyze low affinity transport of Na ϩ and K ϩ in exchange for H ϩ (29,33). However, the physiological function of the organellar membrane NHEs is still unknown.
In this study, we characterized organellar alkali-cation/proton exchanger proteins and found that four NHE isoforms are distributed to Golgi and post-Golgi compartments. From these observations we suggest a mechanism regulating the luminal pH and cation composition of the intracellular compartments by NHEs.
cDNA Cloning-GenBank TM data bases of expressed sequence tags (EST) and the human cDNA sequencing project were searched by using amino acid sequences of the mammalian NHE isoforms. A human clone KIAA0939 encoding NHE8 (accession number AB023156) 2 was provided by the Kazusa DNA Research Institute (Kisarazu, Japan). Sequencing analysis of the cDNA suggested that this clone was missing coding information of ϳ0.5 kbp at the 5Ј region (corresponding to amino acids 1-169 in the mouse homologue). Two mouse EST clones (accession numbers BF782177 and BF787440) encoding NHE8 were obtained from Invitrogen. The nucleotide sequence of BF787440 has been deposited in the DDBJ/GenBank TM (accession number AB089793). A human EST clone (accession number BI600396) encoding NHE9 was obtained from Invitrogen. This EST clone contained the entire NHE9 ORF but carried nonsense (C3 A) and frameshift (insertion of C) mutations at positions of 26 and 272 nt from the initiation ATG. We cloned cDNA corresponding to NHE9 ORF by reverse-transcription (RT)-PCR amplification of poly(A) ϩ RNA isolated from human cultured cell lines, HEp2, HEK293, HeLa, and Caco-2, using the following primers: 5Ј-CTAAGGAATC-CCAAGAAGACTGGGG-3Ј and 5Ј-TGATTACATCTGTACTCTTCAT-GCC-3Ј. The nucleotide sequence of the NHE9 cDNA was determined by sequencing three independent PCR products from each cell line and was deposited in the DDBJ/GenBank TM (accession number AB089794). A human NHE6 cDNA clone, KIAA0267, was obtained from the Kazusa DNA Research Institute (Kisarazu, Japan). Human NHE7 cDNA was cloned by PCR amplification as follows. A cDNA fragment of the ORF (160 nt from initiation ATG to the termination codon) was amplified from poly(A) ϩ RNA isolated from HEK293 cells. The 159-bp DNA fragment corresponding to the 5Ј-end of ORF was cloned by amplifying a human genome clone, accession number AL050307 (provided by Children's Hospital Oakland Research Institute, Oakland, CA). The fulllength NHE7 ORF was reconstituted by PCR, and the integrity of the construct was verified by DNA sequencing.
Spectroscopy and pH Titration of EGFP Mutants in Vitro-The recombinant 6xHis-GFP mutants were produced in E. coli BL21(DE3) and purified by Ni 2ϩ -affinity chromatography using a linear gradient of 0 -1 M imidazole in buffer containing 10 mM Tris-HCl (pH 7.5) and 0.3 M NaCl. Titrations of GFP fluorescence versus pH were performed by cuvette fluorometry. Purified GFPs (10 g/ml, 134-fold dilution of the purified proteins) were dissolved in either 0.2 M sodium phosphate (pH 8.0 to pH 5.0) or 0.1 M sodium phosphate and 0.1 M sodium acetate (pH 4.5 and pH 4.0). The filters used for excitation were 480 nm Ϯ 10 nm (EGFP), 440 nm Ϯ 10 nm (ECFP), and 505 nm Ϯ 10 nm (EYFP).
Reconstitution of NHE8 Proteins into Liposomes-For reconstitution of NHE8 proteins into liposomes, we used the protocol developed previously for reconstitution of the purified His 6 -tagged H ϩ -ATPase, AHA2, and Na ϩ /H ϩ exchanger, AtNHX1 (33,42). NHE8 (50 g) was mixed with 2.34 mg of soybean phospholipids type II-S (Sigma) in a total volume of 208 l of reconstitution buffer (20 mM BTP-MES (pH 7.5), 25 mM (NH 4 ) 2 SO 4 , 10% glycerol, 2.5 mM pyranine). The sample was solubilized by addition of 12 l of 1 M n-octyl-␤-D-glucoside and loaded onto a 2.5-ml spin column filled with Sephadex G-50 (Amersham Biosciences) preloaded with 2.5 mM pyranine in reconstitution buffer. After centrifugation for 5 min at 180 ϫ g, the eluate was incubated for 30 min at room temperature with 100 mg of wet Bio-Beads (SM-2, Bio-Rad) and again passed over a G-50 spin column.
Measurement of Na ϩ /H ϩ Exchange in Vitro-Pyranine fluorescence was recorded at 463 nm excitation wavelength and 510 nm emission wavelength. Fluorescence of the sample was adjusted to appropriate levels by diluting the sample with reconstitution buffer. The liposomes containing NHE8 were diluted in the reconstitution buffer without (NH 4 ) 2 SO 4 and pyranine in a 2-ml reaction cuvette held at 25°C. The 40-fold NH 4 ϩ dilution resulted in acid loading of the vesicles because of outward diffusion of NH 3 (44). The resulting pH inside the vesicles determined from the fluorescence of pyranine was 5.73. Thereafter, inward cation gradients were imposed by the addition of NaCl outside the vesicles. Proton efflux coupled to cation influx was monitored by the increase of pyranine fluorescence. The initial rate of fluorescence variation after the addition of cations (⌬F⅐min Ϫ1 ) was used to determine the affinity for cations and maximum initial rate of vesicle alkalinization by Hanes-Woolf plotting. For radioactive sodium uptake, the liposomes (12.5 l) were suspended in 0.5 ml of reaction buffer (20 mM BTP-MES (pH 7.5), 10% glycerol, 160 mM 22 NaCl (0.23 MBq/mmol)). At various times, ice-cold reaction buffer (1.5 ml) without 22 Na ϩ was added to stop the reaction. The sample was applied to a membrane filter (HATF, 0.45-m pore size, Millipore, Billerica, MA), washed three times with 2 ml of reaction buffer without NaCl, and then dried for quantitation of radioactivity. The background radioactivity was determined by adding 25 mM (NH 4 ) 2 SO 4 to the reaction buffer prior to starting the reaction. Radioactivity was measured with an ACSII liquid scintillation counter (Amersham Biosciences) and a Beckman LS6500 liquid scintillation counter (Beckman Coulter, Inc., Fullerton, CA).
Fluorescence Microscopy-Cells were fixed with 2% formaldehyde and then permeabilized in phosphate-buffered saline containing 1% bovine serum albumin, 0.4% saponin, and 2% normal goat serum (34). The cells were incubated with primary antibodies against NHE proteins, the epitope tags, or organelle marker proteins and then stained with fluorescently labeled secondary antibodies. After washing with phosphate-buffered saline, samples were filled with 10% glycerol and then observed under an Olympus BX51 microscope equipped with differential interference contrast optics (Olympus, Tokyo, Japan). COS7 cells were transfected with pCMV-NHE6-HA or pcDNA3.1-NHE9-mycHis and incubated for 2 days. Cells were serum-starved for 30 min at 37°C to deplete them of transferrin, chilled on ice, and allowed to bind Alexa 594-Tfn (60 g/ml in phosphate-buffered saline) for 45 min on ice. Cells were washed briefly in phosphate-buffered saline and placed in Dulbecco's modified Eagle's medium at 37°C. After incubation for the indicated times, cells were fixed, permeabilized, and stained with anti-HA or anti-Myc antibodies.
Measurement of Organelle pH in Vivo-Between 2 and 4 days after transfection, cells were imaged at 26°C with an inverted fluorescence microscope (IX70, Olympus, Tokyo, Japan) equipped with a cooled charge-coupled device camera (ORCA-ER, Hamamatsu Photonics, Hamamatsu, Japan). Cells were incubated initially for 5 min with  (15). The interference filters (Omega Optical and Chroma Technology, Rockingham, VT) used for excitation and emission were 430 -450 and 465-495 nm for ECFP, 470 -400 and 515-550 nm for EGFP, and 485-515 and 528 -568 nm for EYFP. Image acquisition was controlled by the Aquacosmos software (Hamamatsu Photonics, Hamamatsu, Japan). Regions of interest were selected manually, and pixel intensities were spatially averaged after background subtraction.
Other Materials-22 NaCl (37 MBq/ml, 13.2 GBq/mg) was obtained from PerkinElmer Life Sciences. Other chemicals were from Sigma, unless otherwise specified.

RESULTS
Novel Organellar Na ϩ /H ϩ Exchangers, NHE8 and NHE9 -Previously, seven isoforms of the NHE protein have been identified using molecular biological approaches (19). Recent progress on the human genome sequencing project provided a powerful tool to identify novel molecules showing characteristic primary structures, and this enabled us to identify all genes encoding organellar NHE members. We searched the human genome sequence, and found that six regions (chromosomes 3q24, 10q11.21, 12q23.1, 20q13.3, 22q11.21, and Xq21.2) contain sequences showing similarity to previously known NHEs, in addition to the seven loci for previously known NHEs and one NHE3 pseudogene (45). Sequence analysis revealed that four of the six chromosome regions (chromosomes 10q11.21, 12q23.1, 22q11.21, and Xq21.2) contain none of the exon-intron boundaries predicted from the genomic structures of the NHE family and carry both nonsense and frameshift mutations. We did not find the corresponding cDNAs in the mammalian EST data bases. Thus, these regions are thought to be pseudogenes.
We searched the EST databases for cDNA clones corresponding to the regions at chromosomes 20q13.3 and 3q24, and found KIAA0939 (named NHE8) and BI600394 (named NHE9) for chromosomes 20q13.3 and 3q24, respectively. NHE8/ KIAA0939 is the most recently identified NHE homologue reported by Goyal et al. (31). As reported by Goyal et al. (31), the KIAA0939 clones lacked ϳ500 bp in length at the 5Ј-end of the coding region. We searched mouse cDNA libraries and obtained two cDNA clones (BF782177 and BF787440) carrying the entire length of the ORF from mouse kidney. Sequence analysis of the cDNA (BI600396) corresponding to NHE9 revealed that FIG. 1. A, alignment of amino acid sequence of NHE isoforms. The alignment was generated using the ClustalV algorithm (73). Amino acid residues conserved over three or more isoforms, and transmembrane regions predicted by TMHMM2.0 (43) are inverted and shaded, respectively. The amino acid sequences used are as follows: NHE1 (human, P19634); NHE6 (human, Q92581); NHE7 (human, AAK54508); NHE8 (mouse, AB089793); and NHE9 (human, AB089794). B, hydropathy plots of amino acid sequences of mouse NHE8 (upper panel) and human NHE9 (lower panel). Average hydropathy indexes were calculated by the Kyte-Doolittle method using a window of 9. Positive and negative values represent hydrophilicity and hydrophobicity, respectively. Twelve predicted membrane-spanning domains are indicated by the bars below each plot. C, phylogenetic tree of mammalian NHE isoforms. The phylogenetic relationships among the NHE isoforms were calculated using the ClustalV algorithm. Amino acid sequence used are as follows: NHE2 (human, AF073299); NHE3 (human, U28043); NHE4 (rat, AAA41703); NHE5 (human, XM_007868). this clone contained a nonsense (C3 A, 26 nt from initiation ATG) and a frameshift (deletion of C at 272 nt from initiation ATG) mutation in the possible coding region when compared with the human genomic sequence. We cloned NHE9 cDNAs from several human cell lines and determined the nucleotide sequence.
From these examinations, we concluded that the human NHE family is composed of nine members. Phylogenetic analysis of the amino acid sequences revealed that the nine NHE isoforms are classified into two groups of NHE1-NHE5 and NHE6 -NHE9 (Fig. 1C). NHE8 and NHE9 are grouped with NHE6 and NHE7, which are thought to reside on recycling endosomes and the trans-Golgi network (29,30). These results suggest that NHEs can be classified into two groups, plasma membrane NHEs (NHE1-NHE5) and organelle-specific NHEs (NHE6 -NHE9) that include the novel NHE members NHE8 and NHE9.
Tissue and Cellular Distribution of NHE6 -NHE9 Isoforms-We examined the distribution of NHE8 and NHE9 transcripts in human tissues by Northern blotting analysis. Previous studies have revealed that plasma membrane NHEs show unique and isoform-specific tissue expression, but NHE6 and NHE7 are ubiquitous (21,26,28,29,46). Poly(A) ϩ RNA prepared from various human tissues was hybridized with 32 P-labeled DNA probes corresponding to the open reading frames of human NHE8 and NHE9 and then autoradiographed. NHE8 (6.4 and 4.6 kb) and NHE9 (3.7 kb) transcripts were detected in all human tissues tested ( Fig. 2A), suggesting that these isoforms are ubiquitously expressed. The highest amounts of NHE8 transcript were found in skeletal muscle and kidney. The NHE9 transcript was detected at similar levels in all the human tissues. We further examined the expression of NHE8 and NHE9 in several human cultured cell lines, HEK293 (embryonic kidney), HeLa (cervix), HEp2 (laryngeal epithelium), and Caco-2 (colon), by RT-PCR. The broad and ubiquitous expression of NHE8 and NHE9 shown by Northern blotting ( Fig. 2A) was confirmed at the cellular level. The transcripts of the NHE6 and NHE7 were also detected in all the cell lines by the RT-PCR analysis, suggesting that NHE6-NHE9 isoforms each play a role in basic cellular functioning.
Cation/H ϩ Exchange Activity of NHE8 -We examined the alkali-cation/H ϩ exchange activity of the NHE8 gene product. NHE8 protein tagged with His 6 was expressed in yeast cells, purified using a nickel-nitrilotriacetic acid column, and subjected to ion-transport assays using artificial liposomes. The NHE8 expressed in yeast cells appeared as a 55-kDa protein on SDS-PAGE (Fig. 3A), which is slightly smaller (ϳ10 kDa) than the calculated mass, as has been noted for other eukaryotic Na ϩ /H ϩ exchanger proteins (47-50). The protein was localized to intracellular compartments, primarily the ER (Fig. 3B).
Microsomal membranes were isolated from the yeast cells expressing NHE8 -6xHis, and the proteins were solubilized using n-dodecyl-␤-D-maltoside. The protein (Fig. 3C) was reconstituted in soybean phospholipid vesicles containing the pH indicator pyranine in the presence of (NH 4 ) 2 SO 4 . Dilution of the proteoliposome solution in reconstitution buffer without (NH 4 ) 2 SO 4 resulted in an immediate fluorescence diminution of vesicle-trapped pyranine, reflective of the internal acidification of the vesicles (Fig. 4A, time 30 s). When proteoliposomes were diluted under equilibrium conditions (equal concentrations of ammonium inside and outside the vesicles), the lumen of the liposome was not acidified (data not shown). The internal pH, estimated using standard pH buffers, was 5.73. When vesicles were diluted in the presence of (NH 4 ) 2 SO 4 , the internal pH as indicated from pyranine fluorescence was 7.5. The variation of pyranine fluorescence (pK a 7.2) was approximately linear between pH 5.7 and 7.5 (data not shown) and therefore is directly related to intravesicular pH changes. The addition of NaCl (300 mM) increased the pyranine fluorescence, which indicates the alkalization of the liposome lumen to pH 6.4 (Fig.  4A). No significant recovery of pyranine fluorescence was observed upon the addition of organic cation, choline chloride, or addition of NaCl to control liposomes that did not contain the NHE8 protein (Fig. 4A). The addition of (NH 4 ) 2 SO 4 fully collapsed the pH gradient in both liposomes and proteoliposomes. These observations are consistent with NHE8-mediated Na ϩ - dependent proton efflux from the liposomes. The ion exchange reaction showed saturable kinetics with increasing cation concentrations (Fig. 4B). The affinity of NHE8 for Na ϩ and K ϩ was relatively low, with apparent K m values of 130 and 75 mM, respectively. The affinities for Li ϩ , Cs ϩ , and Rb ϩ were much lower, and the K m value was difficult to determine because it is relatively high (Ͼ200 mM) (data not shown). The maximum initial rate of vesicle alkalinization (V max ) obtained from fitting the data was 3.6 and 3.8 ⌬F⅐min Ϫ1 . We examined Na ϩ uptake into the proteoliposomes. If the increase in pH in the liposomes is because of proton efflux coupled with Na ϩ influx through Na ϩ /H ϩ antiport, then the proteoliposomes should sequester Na ϩ . As expected, 22 Na ϩ uptake was observed with these proteoliposomes (Fig. 4B), whereas 22 Na ϩ uptake was negligible in liposomes without NHE8 and in proteoliposomes carrying NHE8 mutated at the glutamate 220 and asparagine 225 residues (Fig. 4B), which are important for the exchange activity in NHE1 (51,52). These results indicate that NHE8 is capable of performing alkali-cation/H ϩ exchange.
Intracellular Distribution of NHE6 -NHE9 Isoforms-The similarity of primary structures among NHE6 to NHE9 isoforms suggested that these isoforms act on the intracellular compartments. We examined the subcellular localization of NHE8 and NHE9 by immunofluorescence microscopy, and compared their distributions with those of NHE6 and NHE7. COS7 cells were fixed, permeabilized, and stained with antibodies against either NHE8, NHE9, or epitope tags fused to the C terminus of the NHE proteins. Fluorescence microscopy revealed that NHE8 proteins were found primarily in curvy tubular structures at the region juxtaposed to the nucleus (Fig.  5A). A portion of the signal was detected in punctate structures dispersed throughout the cell. Double staining of NHE8 with organelle-marker proteins suggested that the localization of NHE8 is quite similar to that of GM130, a marker for cis-Golgi compartments (Fig. 5A). However, the signals from NHE8 and GM130 are not overlapping but rather closely adjacent to one another at higher magnifications (Fig. 5A). The same result was obtained by staining NHE8-FLAG transiently expressed in COS7 cells (data not shown). We did not find colocalization with protein markers for other compartments, including ER (calnexin), early endosomes (EEA1), late endosomes and lysosomes (Lamp-2), TGN and late endosomes (cation-independent mannose 6-phosphate receptor) (data not shown). From these observations, we concluded that NHE8 is mainly localized to the mid-to trans-Golgi compartments.
We next investigated the intracellular localization of NHE9 by immunofluorescence microscopy. This protein is also localized to intracellular compartments. The signals were found as punctate structures, highly concentrated around the nucleus and dispersed throughout the cell periphery (Fig. 5B). NHE9 showed a partial colocalization with NHE6 in punctate structures between the cell periphery and the peri-nuclear region (Fig. 5C).
Since the distribution of NHE9-positive compartments is similar to that of endosomes, we performed double staining with EEA1, Lamp-2, and cation-independent mannose 6-phosphate receptors. No significant overlap of NHE9 with Lamp-2 or cation-independent mannose 6-phosphate receptor was observed, and only a small portion of NHE9 was colocalized with EEA1 (Fig. 5D). Signals from NHE9 were found in compartments labeled by rhodamine-conjugated transferrin (Fig. 5E), suggesting the localization of NHE9 to the recycling endosomes rather than EEA1-residing early endosomes. To differentiate among endosomal compartments in the recycling pathway, rhodamine-labeled transferrin was internalized and chased in the presence of nonlabeled transferrin for up to 30 min. After fixation, NHE6 and NHE9 were detected by immunostaining. Early after internalization (2-10 min), the major population of transferrin-labeled structures was found at the cell periphery and labeled with NHE6 (Fig. 6), suggesting that NHE6 is localized to an early stage of recycling pathway. Consistently, most of the NHE6-positive compartments contained EEA1, a marker for early endosomes (data not shown). The colocalization with NHE6 was transient and decreased after longer chase periods. The signals from labeled transferrin were translocated to the juxtanuclear region, and the number of transferrinlabeled endosomes containing NHE9 increased slowly from 10 to 15 to 30 min of chase (Fig. 6). These observations, together with the partial colocalization of NHE6 and NHE9 (see Fig.  5C), suggest that the proteins reside on compartments of the

FIG. 3. Expression and purification of NHE8 protein expressed in yeast.
A, lysates of yeast cells (W303-1B) expressing mouse NHE8 or NHE8-GFP were electrophoresed, transferred to a nylon membrane, and then immunoblotted using an anti-NHE8 antibody. B, microscopic images of yeast cells expressing mouse NHE8-GFP. Left and right panels show the fluorescence and differential interference contrast (DIC) images, respectively. C, affinity purification of NHE8 -6xHis expressed in yeast cells. Yeast cells (TYY3) expressing mouse NHE8 -6xHis were lysed (total), and the microsomal membranes (P100) were recovered by centrifuging at 100,000 x g. The proteins were solubilized in 0.4% n-dodecyl-␤-Dmaltoside and then subjected to Ni 2ϩ -affinity column chromatography. Proteins were eluted from the column with a linear gradient of 80 -400 mM imidazole and then processed for immunoblotting of NHE8. Numbers shown below the panels indicate the fold purification of NHE8. FT, flowthrough fraction from the Ni 2ϩaffinity resin. early and late stages of recycling pathway, respectively, with partial overlap, which are possibly early endosome and recycling endosome, respectively.
Overexpression of Na ϩ /H ϩ Exchangers Causes Alkalinization of the Compartment-The isoform-specific distribution of NHE6 -NHE9 prompted us to investigate their function in regulation of organellar pH and cation concentration. Some GFP mutants are reported to exhibit pH-sensitive fluorescence (53,54). We examined the pH sensitivity of EGFP (F64L and S65T), EYFP (S65G, V68L, S72A, and T203Y), and ECFP (F64L, S65T, Y66W, N146I, M153T, and V163A) expressed in E. coli cells. EGFP and EYFP showed an acidification-dependent decrease in the fluorescence emission at 510 and 528 nm, respectively (Fig. 7A). The apparent pK a values of EGFP and EYFP were 5.9 and 5.4 with Hill coefficients (n) of 2.1 and 2.3, respectively. The pH-dependent change in fluorescence of ECFP was smaller than that of EGFP or EYFP (pK a 6.2). The fluorescence change was reversible in the pH range 4 -8 for all three proteins (data not shown).
Because these pH-dependent changes span the pH range of most subcellular compartments, we reasoned that these GFP mutants could be used as physiological pH indicators for Golgi and endosomes. EYFP was fused with integral membrane protein markers for the mid-to trans-Golgi, GalT (55), or the recycling endosome VAMP3 (56). The N-terminal 82-aa residues of GalT fused to GFP has been shown to localize to the mid-and trans-Golgi with the GFP moiety residing in the Golgi lumen (54). COS7 cells expressing GalT-GFP and either NHE8- FIG. 4. NHE8 protein transports Na ؉ and K ؉ ion in exchange for H ؉ . A, cation/H ϩ exchange activity in reconstituted liposomes. Affinity-isolated proteins (50 g) of NHE8 -6xHis (black line) or control proteins (gray line) were reconstituted in liposomes containing ammonium. A pH gradient was created by a 20-fold dilution of proteoliposomes in ammonium-free medium at pH 7.5. After dilution, the pH inside the liposomes was 5.73. The cation/H ϩ exchange reaction was started by the addition of 300 mM NaCl (arrow). A fast fluorescence recovery corresponding to vesicle alkalinization was observed upon addition of Na ϩ to the vesicles, indicating cation/H ϩ exchange. The addition of 25 mM (NH 4 ) 2 SO 4 (arrowhead) resulted in full recovery of fluorescence. B, different concentrations of Na ϩ (left) and K ϩ (right) were added to acid-loaded vesicles. C, Na ϩ uptake into the proteoliposomes coupled with ⌬pH. Liposomes carrying either NHE8 -6xHis or NHE8(E220Q,D225N)-6xHis, and the control liposomes without the protein were diluted 40-fold with ammonium-free buffer to generate a pH gradient and 22 NaCl (160 mM) was added to initiate the subsequent Na ϩ uptake. Reactions were terminated after 2 min by adding 3 volumes of ice-cold medium without 22 Na ϩ , and then the radioactivity of 22 Na ϩ sequestrated in the liposomes was measured by using a liquid scintillation counter. FLAG or NHE8-ECFP were observed under the fluorescence microscope. NHE8-FLAG and NHE8-ECFP were colocalized with GalT-GFP and GalT-EYFP, respectively (Fig. 8). This confirms the mid-and trans-Golgi localization of NHE8 protein. Steady-state Golgi pH in COS7 cells was 6.50 Ϯ 0.05, which is consistent with values measured in HeLa and Vero cells (Fig. 9A) (15,54). Overexpression of NHE8 dissipated the acidic pH of the Golgi complex and increased the pH by about 0.78 pH unit to pH 7.28 Ϯ 0.15 (Fig. 9A), indicating that NHE8 mediates proton efflux from the Golgi lumen.
We examined the pH in recycling endosomes using VAMP3-GFP. VAMP3 is a type II integral membrane protein on the recycling endosome (56) and the chimera of VAMP3-GFP localizes to recycling endosomes with GFP residing in the endosome lumen (57). The VAMP3-EYFP was localized to punctate structures in the peri-nuclear region and cell periphery of COS7 cells (Fig. 8). Most interestingly, NHE9-ECFP was partially colocalized with VAMP3-EYFP in the peri-nuclear region (Fig.  8), but NHE6-ECFP was not significantly colocalized with VAMP3-EYFP (data not shown). The resting pH in recycling endosomes of COS7 cells was 6.73 Ϯ 0.03, slightly higher than that reported previously (pH 6.2; Refs. 58 and 59). When NHE9-ECFP was overexpressed in cells, the pH in the recycling endosomes increased 0.41 pH unit to 7.14 Ϯ 0.07 (Fig.  9B). Most interestingly, overexpression of NHE6 did not cause the alkalinization of VAMP3 compartments (pH 6.73 Ϯ 0.01, Fig. 9B), consistent with the differential localization of NHE9 and NHE6 in cells.

DISCUSSION
In this study, we showed that four types of Na ϩ /H ϩ exchanger proteins, including the two novel isoforms NHE8 and NHE9, were distributed to intracellular compartments, Golgi complex, and post-Golgi endosomal compartments in human cells. The distribution of these NHEs was isoform-specific, but there was partial overlap among the distributions. This suggests that different NHEs function in particular organelles and regulate the specific compositions of monovalent alkali cations and protons in the Golgi and post-Golgi compartments.
Four Organellar NHE Isoforms in the Human NHE Family-By searching the human genome sequence, we found that the NHE family is composed of nine members that are classified into two groups, NHE1-NHE5 and NHE6 -NHE9. The sequence analysis of the entire human genome and EST data base enabled us to identify all members of this protein family; indeed, all NHE isoforms reported previously were found in the data base. Although the overall similarity of amino acid sequences among the NHE isoforms is relatively low (17-67% identity), all the NHE isoforms possess a characteristic secondary structure composed of multiple transmembrane domains at the N terminus and a large hydrophilic domain at the C terminus. Charged amino acid residues in the predicted trans- membrane segments are conserved in all members of this family. Although little is known about the specific amino acids involved in the exchanging transport and about the mechanism of operation in mammalian NHEs, mutations in the conserved residues, glutamine 262, glutamine 391, and asparagine-267 (in NHE1), greatly reduced the exchange activity (Fig. 4) (51,52). These data suggest the importance of the charged residues in ion transport across the membrane and essentially the same molecular mechanism of ion transport mediated by the NHE isoforms. We showed that NHE8 mediates the exchange of H ϩ for Na ϩ using the recombinant protein reconstituted in liposomes. Most interestingly, the luminal acidity of the liposome was not fully collapsed by the exchange reaction but appeared to reach an equilibrium at the mildly acidic pH of 6.4 (Fig. 4). These results indicate that NHE8 acts as an ion exchanger on the membrane and also suggest that this protein modulates the acidic pH of the lumen. Further investigations of the pH sensitivity and ion selectivity are now in progress using the in vitro assay and will provide insight into the molecular mechanism of transport and the in vivo function of this protein.
In contrast to the NHE1-NHE5 isoforms, which are known to act on the plasma membrane and which exhibit distinctive patterns of tissue and cell expression in order to fulfill their tissue-and cell-specific functions, NHE6 -NHE9 isoforms are ubiquitously and concurrently expressed in single cells. In addition, the yeast endosomal NHE homologue, Nhx1p (60) falls into the NHE6 -NHE9 group by phylogenetic analysis (data not  The fluorescence intensity of purified recombinant GFP mutants as a function of pH was measured in a fluorometer as indicated under "Experimental Procedures." Symbols used are as follows: circles, EGFP; triangles, ECFP; squares, EYFP. B, schematic drawing of constructs for Golgi-and recycling endosome-specific expression of GFP mutants. The GFP mutants were fused to the C terminus of either the human GalT (␤-1,4-galactosyltransferase) N-terminal 82amino acid residues or the entire rat VAMP3 sequence. NHE isoforms were fused with ECFP to examine expression and localization.
shown). From these observations, we hypothesized that NHE6 -NHE9 are members of organellar NHEs concurrently acting in the cell and involved in fundamental cellular events. Of the four isoforms, NHE8 is phylogenetically distinct from other organelle NHE isoforms, whereas NHE6, NHE7, and NHE9 show more closely related primary and secondary structures across their entire length. Noteworthy, the C-terminal hydrophilic domain of NHE8 (ϳ100 amino acid residues) is shorter than those of other NHE isoforms (150 -190 amino acid residues) and shows no significant amino acid similarity, suggesting that the function and the regulation of NHE8 are different from other organellar NHE isoforms. The transcript of NHE8 was relatively high in skeletal muscle and kidney, where it is abundant in proximal tubules in the outer medulla and cortex of the kidney, as reported by Goyal et al. (31). This may suggest a specialized function of NHE8 in Golgi apparatus and these tissues.
Isoform-specific Localization of NHEs-Microscopic observations revealed that NHE6 -NHE9 isoforms were localized to distinct intracellular compartments. NHE6 was found in early recycling endosomes, NHE7 in the trans-Golgi network, NHE8 in the mid-and trans-Golgi, and NHE9 localized to late recycling endosomes. Most interestingly, the distributions of these isoforms are not discrete in these compartments but partially overlap one another. Post-Golgi compartments are connected by dynamic membrane flow of exocytic and endocytic mem-brane trafficking (61,62). The overlapping distribution of the NHE isoforms suggests that they are circulating in the post-Golgi membrane traffic, and the distribution of the isoforms would be maintained by the balance of export from and retention in each compartment. Most interestingly, the amino acid sequence of four regions, the N terminus (aa 1-29 in NHE9), second loop (aa 70 -127), and two segments in the C-terminal hydrophilic domain (aa 495-517 and 570 -645), are diverged in the NHE6, NHE7, and NHE9 isoforms. Recent topological analyses of plant vacuolar and yeast endosomal Na ϩ /H ϩ exchangers suggested that the N termini and second loop are exposed to the cytoplasm and that the C-terminal domain resides in the organelle lumen (63,64). This topological feature implies that the N terminus and the second loop are involved in the localization of the NHEs through recognition by cytoplasmic machinery for vesicular membrane trafficking and that the C-terminal region may act as a regulatory domain for ion transport by sensing the intra-compartment ionic concentration. Further examination will uncover how these isoforms are distinguished and localized to specific compartments.
Role of NHE6 -NHE9 Isoforms in Maintaining Organelle Ion Homeostasis and Their Unique Acidity-Previous studies revealed that intracellular compartments are mildly acidified to specific pH values and exhibit a gradient of decreasing pH from ER (pH ϳ7.1) to Golgi (pH ϳ6.2-7.0), trans-Golgi network (pH ϳ6.0), and secretory granules (pH ϳ5.0) and from early and late endosomes (pH ϳ6.5) to lysosomes (pH ϳ4.5) (1, 3, 4). The low pH environment is crucial for a number of well defined processes, such as the activation of endoproteases (65), the maturation and modification of secretory proteins (66), and the dissociation of receptor-ligand complexes (67,68). Steady-state pH values of ER, Golgi, and secretory vesicles appeared to be controlled by rates of H ϩ pumping and by the gradual decrease in organelle H ϩ permeability from ER to Golgi to secretory vesicles in the secretory pathway, whereas the membrane potential in Golgi and secretory vesicles is small and not perturbed by large changes in Cl Ϫ and K ϩ conductances, indicating that membrane potential is not a determinant of steadystate pH values (10,11,(13)(14)(15).
Based on these findings, we hypothesized the role of NHEs in controlling luminal pH through the cation/H ϩ exchange activity, and examined the organellar pH using pH-dependent GFP mutants. Overexpression of NHE8 and NHE9 caused luminal alkalinization to near cytosolic pH of the compartments in which they reside. This observation suggests that these NHEs effuse luminal H ϩ in exchange for cytosolic cations. Electroneutral one for one ion exchange driven by the concentration gradients across the membrane equalizes the ratios of inside and outside ion concentrations, i.e.
The Golgi luminal [K ϩ ] is ϳ107 mM, which is only slightly lower than the cytosolic concentration (12). Because of the much higher concentration of K ϩ than H ϩ , the electroneutral exchange by overexpressed NHEs would dissipate the transmembrane ⌬pH of ϳ 0.7 without changing both Golgi and cytosolic k ϩ concentrations and neutralize the compartment to near the cytosolic pH. This assumption is validated by incubating cells in the presence of nigericin, a K ϩ /H ϩ ionophore, which caused rapid neutralization of the Golgi pH (data not shown). The protein levels of NHE8 and NHE9 proteins in overexpressing cells were increased by 51 Ϯ 33-fold (55 overexpressing cells, total 150 cells) and 99 Ϯ 83-fold (67 overexpressing cells, total 138 cells), respectively, compared with the endogenous levels under the fluorescence microscope. Further examination such as knock-down experiments would be required to confirm that endogenous NHE8 and NHE9 are involved in the organel- lar pH regulation. A detailed analysis of the relation between protein levels and luminal pH is currently in progress to provide further insight into the action of NHEs to control the organellar pH.
Recently, a plant vacuolar Na ϩ /H ϩ exchanger has been suggested to play a role in the regulation of vacuolar pH (69). Vacuolar pH is known to be a main determinant for flower coloration, and a shift from reddish purple buds to blue open flowers correlates with an increase in the vacuolar pH. The vacuolar Na ϩ /H ϩ exchanger proteins utilize the proton electrochemical gradient generated by the V-ATPase and H ϩ -translocating pyrophosphatase to couple the movement of H ϩ down its concentration gradient in exchange for cytosolic alkali cation (70). Mutations in the gene encoding the vacuolar Na ϩ /H ϩ exchanger protein NHX1 causes excessive acidification of vacuolar lumen and, consequently, empurpling of flowers of the Japanese morning glory. The fact that NHX1 confers salt tolerance to plants by sequestrating excess Na ϩ in the vacuole (71) is consistent with the idea that the vacuolar Na ϩ /H ϩ exchanger transports luminal H ϩ out of the vacuole, exchanging it for cytosolic Na ϩ . These observations support our idea that K ϩ /H ϩ exchange mediates the H ϩ leak that has been proposed to regulate the luminal pH of intracellular compartments and that specific NHE isoforms present on Golgi and post-Golgi compartments are involved in controlling the unique acidic pH of these compartments. We found no apparent localization of NHE isoforms to lysosomes, which is consistent with the previous observation that rat liver lysosomes have no detectable Na ϩ /H ϩ exchange activity (72). Lysosomes are the most acidic compartment of the cell, pH ϳ4.5, and it may be that the H ϩ leak mechanism is omitted to accomplish such high acidification, or an alternative mechanism may exist in the lysosomal system.
Further analyses of the organelle NHE isoforms by examining the ion transporting activity, including the ion selectivity and pH dependence in vitro and by interchanging isoforms residing on the compartments in vivo, will uncover the role of NHEs in organelle ion homeostasis and the mechanism of organelle pH regulation in the cell.