The Na+/H+ Exchanger NHE6 Modulates Endosomal pH to Control Processing of Amyloid Precursor Protein in a Cell Culture Model of Alzheimer Disease*

Background: NHE6 transports protons out of the endosomal lumen in exchange for cations. Results: Localization and processing of amyloid precursor protein are altered by NHE6 expression. Conclusion: NHE6 controls Aβ production and is down-regulated in Alzheimer disease brains. Significance: NHE6 and related orthologs may be risk factors for Alzheimer disease. Early intervention may be key to safe and effective therapies in patients with Alzheimer disease. Endosomal dysfunction is an early step in neurodegeneration. Endosomes are a major site of production of Aβ peptide from the processing of amyloid precursor protein (APP) by clipping enzymes (β- and γ-secretases). The β-secretase enzyme BACE1 requires acidic lumen pH for optimum function, and acid pH promotes Aβ aggregation. The Na+/H+ exchanger NHE6 provides a leak pathway for protons, limiting luminal acidification by proton pumps. Like APP, NHE6 expression was induced upon differentiation of SH-SY5Y neuroblastoma cells and localized to an endosomal compartment. Therefore, we investigated whether NHE6 expression altered APP localization and processing in a stably transfected cell culture model of human APP expression. We show that co-expression with NHE6 or treatment with the Na+/H+ ionophore monensin shifted APP away from the trans-Golgi network into early and recycling endosomes in HEK293 cells. NHE6 alkalinized the endosomal lumen, similar to monensin, and significantly attenuated APP processing and Aβ secretion. In contrast, Aβ production was elevated upon NHE6 knockdown. We show that NHE6 transcript and protein levels are lowered in Alzheimer brains relative to control. These findings, taken together with emerging genetic evidence linking endosomal Na+/H+ exchangers with Alzheimer disease, suggest that proton leak pathways may regulate Aβ generation and contribute to disease etiology.

via pH-dependent effect on the formation of b-sheets and the binding of metal ions to Ab (13). Most recently, (iv) acidic luminal pH has been shown to attenuate lysosomal degradation of Ab by disrupting SORLA-Ab interaction (14). These observations suggest that modulation of luminal pH by endosomal NHE activity may offer novel strategies to control APP trafficking and processing.
SLC9A6, the gene encoding NHE6, is one of the six most recurrently mutated loci in patients with X-linked intellectual disability and has been linked to autism comorbid with seizures, Christianson syndrome, and Angelman-like syndrome (15)(16)(17)(18)(19). Autism spectrum disorder has emerged as a major public health concern worldwide with an urgent unmet need for effective and safe interventions (1). With an aging population, Alzheimer and related neurodegenerative disorders pose a similar challenge (20). Interestingly, the A␤ and Tau pathologies linked by the amyloid cascade hypothesis have been implicated in autism, Fragile X syndrome, and 15q duplication in patients and animal models (21)(22)(23). These recent revelations about the shared pathology of altered processing of amyloid precursor protein (APP) between autism spectrum disorder and Alzheimer disease may point the way to new treatments. Indeed, histopathological examination of post-mortem brain tissue from patients with NHE6 mutations shows widespread neuronal loss, gliosis, and deposits of the hyperphosphorylated microtubule binding protein Tau in tangles and inclusions in neurons and glia (18). Research showing NHE6 down-regulation in substantia nigra in patients with Parkinson disease further signifies an emerging role for endosomal pH and NHE6 activity in the underlying mechanisms of neurodegeneration (24). Finally, closely related NHE6 paralogs have also been implicated in AD in genome-wide association studies (25,26).
In this study, we examine the relationship between NHE6 and APP in human brains, differentiated neuroblastoma cells in culture, and an AD model of stable human APP expression in HEK293 cells. We show that NHE6 expression is linked to APP and other early onset Alzheimer risk genes. In cultured cells, NHE6 colocalizes with and alters processing of APP by modulating endosomal pH. Together with reduced protein levels of NHE6 in AD brains, our findings suggest that endosomal pH modulation by Na ϩ /H ϩ exchangers may be an important factor in neurodegenerative disease.

EXPERIMENTAL PROCEDURES
Cell Culture-Human neuroblastoma SH-SY5Y cells (gift from Dr. Svetlana Lutsenko, Johns Hopkins University) were maintained in minimal essential medium/F-12 supplemented with 10% FBS (Invitrogen). Neuronal differentiation of SH-SY5Y was induced as described earlier (27). Briefly, cells were sequentially treated with 10 nM retinoic acid (Sigma) for 5 days and 50 ng/ml brain-derived neurotrophic factor (Shenandoah Biotechnology) in serum-free medium for the subsequent 3 days. Stable HEK293 cells expressing human APP695 (gift from Dr. Randy Schekman, University of California, Berkeley, CA) were cultured in DMEM/high glucose/sodium pyruvate (Invitrogen) with the addition of 10% FBS (Invitrogen), Glutamax (Invitrogen), NEAA (Quality Biological), 650 g/ml Geneticin/ G418 (Corning Cellgro) and maintained in a humidified incu-bator with 5% CO 2 at 37°C. NHE6-GFP and control vector (GFP) expression were introduced using lentiviral packaging and expression, as described earlier (28). For transferrin uptake and endosomal pH measurement experiments, to avoid spectral overlap with Alexa Fluor 488 or FITC-tagged transferrin, we expressed NHE6 (wild type and mutant) tagged to C-terminal mCherry in HEK293 cells by transient transfection using Lipofectamine 2000 reagent, as per the manufacturer's instructions. Cell viability was measured using the trypan blue exclusion method. Briefly, cells were mixed 1:1 in suspension with a 0.4% solution of trypan blue Stain. A 10-l sample was added to a disposable cell counting chamber slide and inserted into a Countess automated cell counter (Invitrogen). The cell viability and total cell numbers were recorded from six independent experiments, as described earlier (29).
Plasmids-Human NHE6-EGFP was cloned into FUGWlentiviral vector into the BamHI site. NHE6 was cloned into mCherry pcDNA3 vector using HindIII and EcoRI sites to yield NHE6-mCherry construct. A QuikChange Lightning site-directed mutagenesis kit was used to engineer ⌬WST 372 in-frame deletion into NHE6. The resulting plasmid was confirmed by sequencing. Mouse BACE1 plasmid developed by the Thinakaran laboratory was obtained from Addgene (plasmid 26736). This construct has a C-terminal FLAG tag and engineered mutations, previously shown to neither alter localization nor function of BACE1 (30). HsNHE6 short hairpin RNA (shRNA) sequence (5Ј-CCGGCGTCCTAGTGCATGTCTCTGTTCAAGAGACAGA-GACATGCACTAGGACTTTTTC-3Ј) was designed to target the 3Ј-untranslated region (3Ј-UTR). Both HsNHE6 shRNA and scrambled control construct were cloned into pLKO.1 lentiviral vector with a puromycin-selectable marker.
Western Blotting and ELISA-Human brain tissue was obtained from the Johns Hopkins Brain Bank and used in accordance with institutional guidelines. For Western blots, samples of medial frontal cortex gray matter were obtained by punch biopsy and dissolved in 1% Triton and 0.2% SDS lysis buffer. For Western blots of cell culture lysate, cells were lysed using Nonidet P-40 (1%) supplemented with protease inhibitor mixture (Roche Applied Science). Cells were sonicated, and the protein concentration was determined by a BCA assay. Equal amounts of proteins were separated by SDS-PAGE under reducing conditions and then electrophoretically transferred onto nitrocellulose membranes (Bio-Rad). After protein transfer, the membranes were treated with the blocking buffer, fol-lowed by incubations with primary antibodies and then with HRP-conjugated secondary antibodies (GE Healthcare). Super-Signal West Pico substrate was used for detection. Fujifilm LAS 3000 imaging system was used to capture images. Tubulin or glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as a loading control. For quantification of A␤40 in the culture supernatant, 24-h conditioned medium was harvested with the addition of protease inhibitor mixture (Roche Applied Science). Medium was diluted 20-fold and subjected to A␤40 ELISA analysis using a commercially available A␤40 human ELISA kit (Invitrogen) according to the manufacturer's instructions. All experiments were performed at least three times independently.
Quantitative Real-time PCR-mRNA was isolated using RNeasy minikit from Qiagen following the manufacturer's instructions. A high capacity RNA-cDNA kit (Applied Biosystems, catalog no. 4387406) was used to make cDNA from RNA, following the manufacturer's instructions. Quantitative realtime PCR analysis was performed using the 7500 real-time PCR system (Applied Biosystems) using Taqman Fast universal PCR Master Mix (Applied Biosystems, catalog no. 4352042). Taqman gene expression assay probes used were Hs00234723_m1 (SLC9A6 solute carrier family 9 (sodium/hydrogen exchanger), member 6), Hs00543518_m1 (SLC9A9 solute carrier family 9 (sodium/hydrogen exchanger), member 9), and Hs00169098_ m1 (APP). Hs02758991_g1 (GAPDH) was used as an endogenous control. Ct values were used for all manipulations and were first normalized to endogenous control levels by calculating the ⌬Ct for each sample. Values were then calculated relative to control to generate a ⌬⌬Ct value. -Fold change was calculated using the equation, expression -fold change ϭ 2 Ϫ⌬⌬Ct . Each experiment was repeated three times independently.
Steady-state Transferrin Uptake-HEK 293 cells plated on polylysine-coated coverslips were rinsed and incubated in serum-free medium for 30 min to remove any residual transferrin and then were exposed to 100 g/ml transferrin conjugated with Alexa Fluor 488 at 37°C for 60 min. Uptake was stopped by chilling on ice and washing three times with ice-cold PBS. Bound transferrin was removed by an acid wash in PBS (pH 5.0), followed by a wash with PBS at pH 7.0. Cells were fixed with a solution of 4% paraformaldehyde, DAPI-stained, and mounted onto slides to determine colocalization of transferrin with NHE6-mCherry by confocal imaging.
Endosomal pH Measurement-Measurement of pH of transferrin-positive endosomal compartments was performed as described earlier. Briefly, HEK293 cells were rinsed and incubated in serum-free medium for 30 min to remove any residual transferrin and then were treated for 60 min with a pH-sensitive fluorescence indicator (Tfn-FITC) in conjunction with pH-insensitive cargo (Tfn-Alexa Fluor 633) (Molecular Probes) to normalize for endosomal loading. The fluorescence intensity of internalized transferrin was measured for ϳ10,000 cells by flow cytometry using the FACSAria instrument (BD Biosciences), and the average intensity of the cell population was recorded. For calibration of pH-dependent fluorescence, cells were exposed to K ϩ /H ϩ ionophore nigericin (100 mM) and K ϩ -rich medium (125 mM KCl, 25 mM NaCl, 25 mM HEPES or 25 mM MES), adjusted to defined pH values (5.0, 6.0, 7.0, and 8.0).
Immunofluorescence and Image Quantification-Cultured cells on collagen-coated (SH-SY5Y) or polylysine-coated (HEK293) coverslips were preextracted with PHEM buffer containing 0.025% saponin for 2 min and then washed twice for 2 min with PHEM buffer containing 0.025% saponin and 8% sucrose. The cells were fixed with a solution of 4% paraformaldehyde (Electron Microscopy Sciences) and 8% sucrose in PBS for 30 min at room temperature. Cells were stained for primary antibodies and Alexa Fluor-conjugated secondary antibodies, DAPI-stained, and mounted onto slides using Dako fluorescent mounting medium. Cells were imaged using Zeiss AxioObserver inverted microscope with an LSM700 laser-scanning confocal module with advanced variable secondary dichroic beamsplitter that facilitates four-color imaging by eliminating fluorophore bleed-through. Slides were imaged with a ϫ63 oil immersion objective using four laser lines that excite at 405, 488, 561, and 639 nm. Expression of empty GFP vector and NHE6-GFP was detected using GFP (green) signal. Secondary antibodies without spectral overlap with DAPI and GFP fluorescence were selected. Alexa Fluor 568 (red) goat anti-mouse IgG (H ϩ L) and Alexa Fluor 633 (far red) goat anti-rabbit IgG (H ϩ L), both highly cross-adsorbed, were used for detecting subcellular markers and APP, respectively. The human eye is more sensitive to hues in the green and red color groups. Therefore, for better visualization of overlap between two images, we pseudocolored them green and red, and the colocalization was demonstrated by yellow fluorescence in the merge image. Fractional colocalization was determined from Manders' coefficient using the JACoP ImageJ plugin, which measures the direct overlap of green and red pixels in the confocal section and represents it as mean Ϯ S.E. A minimum of 20 cells/condition from at least two representative experiments were included in the analysis.
Structural Modeling and Bioinformatics-Three-dimensional homology model structure of NHE6 was developed using the crystal structure of bacterial cation proton antiporter ortholog Escherichia coli NhaA as a template using multiple state-of-the-art approaches and evolutionary conservation analysis, as described earlier (1,28). A brain RNA sequencing gene expression data set from 578 samples represented as log base 2 of RPKM (reads per kilobase of exon model per million mapped sequence reads) values across different developmental periods and different brain regions was obtained from the BrainSpan atlas (available on the World Wide Web). Hierarchical clustering with XLSTAT (Addinsoft, Paris, France) was performed under nearest neighbor methodology, and results were displayed as a dendrogram and heat map. Microarray data sets for the study included (a) GSE803 on the GPL8300 Affymetrix Human Genome U95 version 2 array analyzing gene expression in a variety of normal human tissues (n ϭ 24) and (b) GSE1297 on the GPL96 Affymetrix Human Genome U133A array analyzing gene expression in different stages of AD (incipient, moderate, and severe) as compared with normal controls (n ϭ 31). We validated our results by performing pooled analysis of gene expression profiles from independent studies of AD versus control brains, taken from anatomically and functionally distinct brain regions. To perform meta-analysis, we used normalized data obtained from Genevestigator (Nebion AG) that facil-itates integration of data from multiple experiments. The pooled estimate and confidence interval of differential expression of NHE6, NHE7, and NHE9 genes were obtained using the RevMan program (Nordic Cochrane Centre). The I 2 statistic was used to assess heterogeneity among microarray data sets. An I 2 value more than 50% indicates the presence of significant heterogeneity (where more than 50% of the variation is due to heterogeneity rather than chance). A fixed effects model was used for data pooling for NHE7 (I 2 ϭ 14%) expression. For NHE6 (I 2 ϭ 65%) and NHE9 (I 2 ϭ 87%), a random effects (Der-Simonian and Laird) model was used for data pooling due to observed substantial heterogeneity.

RESULTS
Expression of NHE6 in Brain Is Linked to APP-Although ubiquitous in tissue distribution, expression of NHE6 (SLC9A6) is highest in the spinal cord and brain (Fig. 1A), where it is particularly enriched in the hippocampus, cortex, and Purkinje cell layer of the cerebellum (1). Expression levels of NHE6 remain high through human development, including the adult (Fig. 1B). This distribution is consistent with emerging reports implicating NHE6 function in neurogenesis, spine dynamics, dendritic arborization, and synaptic strength (1,17,31). Similarly, impaired neurogenesis has been reported to occur early in AD pathology (32).
Genetic architecture and co-expression networks hold the key to understanding Alzheimer disease pathogenesis (33) and are based on the premise that genes that are expressed together are likely to function together (34). We noted that the pattern of APP expression closely parallels that of NHE6 through normal human brain development ( Fig. 1C; Pearson correlation 0.86; n ϭ 578; p ϭ 2.28 ϫ 10 Ϫ172 ) and in all areas of the brain (Fig.  1D). In an independent analysis of the human brain RNA sequencing data, NHE6 was among the top 0.5% of human brain transcriptome transcripts (of 52,376) whose expression was most correlated with the expression of APP across 524 samples. In contrast, no correlation was observed between NHE6 and cytochrome c 1 , an established housekeeping gene (29) (Pearson correlation 0.062; n ϭ 524; p ϭ 0.15). Next, we performed hierarchical clustering of brain NHE6 expression with 15 genes strongly linked to Alzheimer disease and found association of NHE6 with early onset AD genes, including APP and PSEN2, and with the gene encoding ␤-secretase (BACE1) (9,11,35) (Fig. 1E). Notably, clustering of NHE6 was distinct from that of related NHE isoforms (NHE9 and NHE7). We suggest that this clustering could provide clues to functional interactions and form the basis for new hypothesis-driven research, as evidenced by the close association of the enzyme-substrate pair (BACE1, APP) (10) and apolipoproteins (APOE, APOJ/ FIGURE 1. Expression of NHE6 in brain is linked to APP. A, NHE6 gene expression was extracted from the microarray data set of normal human tissues published by Yanai et al. (74). B, expression plots of NHE6 across different human development stages: prenatal period (embryogenesis to birth), infancy (birth to 1 year), childhood (2-10 years), adolescence (11-20 years), and adulthood (21 years and older). C, expression plot of NHE6 versus APP across a total of 578 samples obtained from different developmental periods. Note the prominent linear correlation of APP with NHE6 during normal human brain development (Pearson correlation coefficient, 0.86; p ϭ 2.28 ϫ 10 Ϫ172 ). D, plot showing high Pearson correlation of NHE6 with APP expression across different brain regions. Dashed red line, average. DFC, dorsolateral prefrontal cortex (CTX) (Brodmann area 9 (BA9), BA46); VFC, ventrolateral prefrontal CTX (BA44, BA45); MFC, medial prefrontal CTX (BA32, BA33, BA34); OFC, orbital frontal CTX (BA11); M1C, primary motor CTX (BA4); S1C, primary somatosensory CTX (BA1-BA3); IPC, posterior inferior parietal CTX (BA40); A1C, primary auditory CTX (BA41); STC, posterior superior temporal CTX (BA22); ITC, inferior temporal CTX (BA20); V1C, primary visual CTX (BA17); HIP, hippocampal formation; AMY, amygdaloid complex; STR, striatum; MDT, mediodorsal nucleus of thalamus; CBC, cerebellar CTX. E, hierarchical cluster-ing and expression heat map (from low (blue) to high (red)) of NHE6 with 15 genes strongly linked to Alzheimer disease showing clustering of NHE6 with early onset AD genes, including APP and PSEN2, and with the gene encoding ␤-secretase (BACE1). Notably, a strong association was observed for NHE6 with APP (red asterisk), similar to observations in C. The RNA sequencing gene expression data set included a total of 578 samples represented as log base 2 of reads per kilobase of exon model per million mapped sequence reads (RPKM) values across different developmental periods and different brain regions (data used for B-E were from the BrainSpan atlas Web site. pcw, postconception weeks; mos, postnatal months; yrs, age in years. Error bars, S.D. clusterin) (36). Furthermore, recent studies have validated functional links between other close associations observed in our clustering analysis: CD33 with TREM2 (37) and PICALM with PSEN1 (38). Intriguingly, we observed functional clustering of genes involved in innate immune responses implicated in AD (TREM2, CD33, and MS4A6A) (9,35). These correlative observations prompted our hypothesis that ion transport activity and/or expression of NHE6 may contribute to AD pathology by altering luminal pH, endosomal trafficking, BACE1 activation, and A␤ production.
NHE6 Colocalizes with APP to Alter Trafficking in Cultured Cells-Human SH-SY5Y neuroblastoma cells can be differentiated with retinoic acid and brain-derived neurotrophic factor into cells with morphological and biochemical characteristics of mature neurons, which show increased expression of amyloid precursor protein (27,39). We observed a ϳ3.5-fold increase in APP mRNA upon differentiation of SH-SY5Y cells that was accompanied by a significant increase in NHE6 expression (ϳ2.3-fold) ( Fig. 2A). Previously, we established that NHE6 localizes to endosomal recycling compartments and traffics in part to the plasma membrane (2). Using a rabbit polyclonal antibody raised against the C-terminal tail of NHE6, we localized endogenous NHE6 with recycling (Rab11) endosomal markers in SH-SY5Y cells (Fig. 2B), similar to previous observations in hippocampal neurons (17,31). Colocalization of NHE6 with APP in both undifferentiated cells (Fig. 2C) and in differentiated neurites (Fig. 2D) was consistent with a potential role for NHE6 in APP function.
To further dissect the role of NHE6 in APP function, we utilized a well characterized HEK293-derived cell line stably expressing human APP695, developed by Choy et al. (40) for endosomal APP trafficking studies. Elegant studies by the Schekman group (40) using these cells have led to a model in which plasma membrane APP is endocytosed and trafficked to the trans-Golgi network (TGN), where the majority of processing occurs. Using an antibody that recognizes the C-terminal domain of human APP, as well as the ␤-CTF fragment, we show that APP colocalizes in part with the endosomal markers EEA1 and Rab11 and largely with the TGN marker Golgin 97, but not with the lysosomal marker lysobisphosphatidic acid (LBPA; Fig.  3A). As has been previously established, NHE6-GFP colocalized in part with the endosomal marker EEA1 and predominantly with transferrin-positive endosomes (not shown) (1,2,17). We noted prominent colocalization of APP with NHE6-GFP (Pearson correlation coefficient, 0.83 Ϯ 0.03; Fig. 3B), similar to observations with endogenous NHE6 in differentiated SH-SY5Y cells (Fig. 2). Intriguingly, ectopic expression of NHE6-GFP resulted in striking redistribution of APP from the TGN to endosomes, resulting in increased colocalization with EEA1 and Rab11 (Fig. 3C), as quantified by Manders' coefficient ( Fig. 3D). Redistribution of APP from the TGN also resulted in endocytic vesicles of variable sizes, similar to observations by Choy et al. (40). Given the emerging links between luminal pH and retrograde cargo exit out of endosomes (41), we hypothesized that the effect of elevated NHE6 activity on endosomal pH FIGURE 2. NHE6 colocalizes with APP in differentiated SH-SY5Y cells. A, quantitative PCR analysis of NHE6 and APP mRNA, in human SH-SY5Y neuroblastoma cells normalized to reference GAPDH. SH-SY5Y cells differentiated by sequential treatment with retinoic acid and brain-derived neurotrophic factor show a significant increase in APP and NHE6 expression. Error bars, S.D. determined from triplicate measurements. B, subcellular localization of endogenous NHE6 in differentiated SH-SY5Y cells determined by immunofluorescence confocal microscopy. The merge image (left) shows overlap of NHE6 (red) and recycling endosomal marker Rab11 (green). The boxed region is magnified for better representation (right). NHE6 (red) partly localizes with recycling endosome marker, Rab11 (green), as seen in the merge. An orthogonal view of subcellular localization is shown on the extreme right of the merge image. Colocalization of endogenous NHE6 (red) with APP (green) in undifferentiated SH-SY5Y cells (C) and differentiated neurites (D) is shown. Orthogonal views of subcellular localization are shown on the extreme right of the merge images. APP was immunostained using mouse monoclonal (6E10) antibody, and NHE6 was detected using rabbit polyclonal antibody raised against the C terminus. Scale bars, 10 m. DIC, differential interference contrast.
underlies the blockade of retrograde trafficking of APP from the endosome to the trans-Golgi network.
NHE6 and Monensin Alter APP Trafficking by Modulating Endosomal Acidification-The ionophore monensin mediates Na ϩ /H ϩ exchange and mimics constitutively activated NHE6 to stimulate the endosomal recycling pathway and alter vesicular traffic (42,43). Upon treatment of APP695-expressing HEK293 cells with monensin (1 M for 8 h), APP redistributed from a perinuclear tubulo-reticular pattern into vesicles that colocalized prominently with the early endosomal marker EEA1, as quantified by Manders' coefficient ( Fig. 4, A-C). This finding was strikingly analogous to the effect of NHE6 on APP trafficking and suggested that endosomal pH modulation was a common causal mechanism.
To measure endosomal pH, we took advantage of the overlap of NHE6 with endocytosed transferrin (Tfn), as shown in Fig. 4D. We measured endosomal pH by a ratiometric method using Tfn coupled to a mixture of pH-sensitive (Tfn-FITC) and pHinsensitive (Tfn-Alexa Fluor 633) fluorescence probes to normalize for endosomal loading, as described under "Experimental Procedures." We expressed NHE6 with a C-terminal mCherry tag to avoid spectral overlap with FITC-tagged transferrin. NHE6 expression alkalinized endosomal pH from 5.85 Ϯ 0.11 to 6.34 Ϯ 0.04 (Fig. 4H). These results are consistent with endosomal alkalization observed by Xinhan et al. (44) in HeLa overexpressing NHE6 and hyperacidification seen in NHE6knockdown cells. Luminal endosomal pH in HEK293 cells treated with monensin was also elevated (to 6.48 Ϯ 0.07), similar to cells expressing NHE6-mCherry (Fig. 4H).
To rule out nonspecific overexpression effects, we examined a patient-associated mutation within the membrane-embedded transporter domain of NHE6, which was predicted to result in loss of function (1). Specifically, ⌬WST 372 is an in-frame deletion of three amino acids that was described by Garbern et  FEBRUARY 27, 2015 • VOLUME 290 • NUMBER 9 JOURNAL OF BIOLOGICAL CHEMISTRY 5317 al. (18) in patients with severe intellectual disability and autistic symptoms accompanied by neuronal loss and Tau deposition in the brain. For a structure-driven assessment of NHE6 variants, we developed a three-dimensional model structure of NHE6 on the basis of the inward-open NhaA crystal structure using evolutionary conservation-based approaches, described previously (1,28). We mapped the ⌬WST 372 mutation within the membrane-embedded transporter domain that corresponds to transmembrane helix VII in NhaA, predicted to be non-functional (Fig. 4, E-G). Quantitative PCR showed ϳ80% expression of the mutant transcript relative to wild-type NHE6. We confirmed that cell viability, measured by trypan blue exclusion, remained unaltered by manipulations in NHE6 (vector control, 95.70 Ϯ 1.23%; NHE6 expression, 94.45 Ϯ 1.57%; ⌬WST 372 expression, 95.75 Ϯ 1.54%; percentages represent the proportion of viable unstained cells relative to the total cells counted). NHE6 carrying the ⌬WST 372 mutation colocalized with transferrin ( Fig. 4D) but failed to elevate endosomal pH in HEK293 cells (Fig. 4H), confirming loss of function as predicted using a structure-function approach. Taken together, these results establish perturbations in endosomal pH as an important causal mechanism in subcellular APP redistribution.

NHE6 Regulates A␤ Production in Endosomes
NHE6 Alters APP Processing in Cultured Cells-Several lines of evidence suggest that BACE1 is predominantly localized to the trans-Golgi network, where the majority of APP processing occurs (30,40). We hypothesized that as an evolutionarily conserved mechanism for protection against excess endosomal acidification, NHE6 activity might be one of the fundamental strategies by which neurons largely restrict APP and BACE1 to separate organelles, thus limiting generation of A␤ under normal physiological conditions. To test this hypothesis in our cell culture model, we transfected the BACE1 construct into HEK293-derived cells stably expressing APP695 and confirmed prominent colocalization of APP with BACE1 (Pearson correlation coefficient, 0.89 Ϯ 0.04; Manders' overlap coefficient, 0.83 Ϯ 0.05; n ϭ 30; Fig. 5A), as reported in the literature (30). Intriguingly, unlike APP (Fig. 3B), BACE1 did not show much colocalization with NHE6-GFP in HEK293 cells (Pearson correlation coefficient, 0.39 Ϯ 0.08; Manders' overlap coefficient, 0.22 Ϯ 0.06; n ϭ 20; Fig. 5A). Upon cotransfection of NHE6-GFP and BACE1 into APP stable cells, we documented prominent colocalization of APP with NHE6-GFP and reduced overlap of APP with BACE1. Quantitative analysis using Manders' coefficient revealed the fraction of APP overlapping with BACE1 to NHE6 coexpression and concomitant increase in full-length APP. Note reductions in both ␣-CTF and ␤-CTF forms suggesting a perturbation in trafficking/ activity of multiple secretase enzymes. NHE6-GFP was detected using anti-GFP antibody. D, quantitation of A␤40 by ELISA showed a significant reduction of secreted A␤ peptide from cells expressing elevated levels of NHE6 (mean Ϯ S.D.; **, p Ͻ 0.01; n ϭ 3; two-tailed t test). Total A␤ in the conditioned medium was normalized to total cellular protein. E, expression of disease-associated ⌬WST 372 loss-of-function mutant in HEK293 cells stably expressing APP695 fails to significantly alter APP processing. Western blot analysis was performed with lysates collected from control cells with mCherry empty vector and cells transfected with NHE6⌬WST 372 -mCherry (detected using anti-mCherry rat monoclonal antibody). Holo-APP was detected using polyclonal C-terminal antibody, and tubulin was used as a loading control. decrease from 0.83 Ϯ 0.05 to 0.39 Ϯ 0.11 (p ϭ 8.27 ϫ 10 Ϫ28 ; n ϭ 30) upon NHE6-GFP expression. In previous studies, treatment of cells stably expressing APP with destruxin E, a V-ATPase inhibitor, resulted in a similar decrease in colocalization of APP with BACE1 and reduced processing of APP and A␤ generation (45). Inhibition of V-ATPase is expected to alkalinize endosomes and mimic the activity of NHE6, consistent with a critical role for endosomal pH in A␤ biogenesis.
Upon NHE6-GFP lentivirus-mediated expression in stable APP695 cells, quantitative PCR analysis showed that expression of NHE6 mRNA was enhanced by 3.2-fold (p ϭ 1.97 ϫ 10 Ϫ5 ; two-tailed t test), whereas APP mRNA levels remained unaltered (p ϭ 0.53; two-tailed t test (Fig. 5B). Choy et al. (40) have reported that depletion of retromer subunits interrupts retrograde trafficking of APP from the endosomes to TGN and reduces A␤ production in the stable APP695 cell culture model. Considering that NHE6 overexpression also blocked delivery of APP to TGN, we predicted a similar effect. NHE6 is known to regulate clathrin-dependent endocytosis and modulate endosomal pH (44); therefore, we hypothesized that NHE6 activity could potentially alter (a) ␣-site cleavage of cell surface APP by endocytosis via clathrin-coated pits and (b) ␤-site cleavage of APP by modulation of luminal pH and ␤-secretase activity. Consistent with our hypothesis, Western blot analysis showed that ectopic expression of NHE6-GFP in HEK293 cells stably expressing APP695 significantly reduced A␤ levels and concomitantly increased full-length APP (Fig. 5C). We also noted reductions in both CTF forms, suggesting a perturbation in trafficking and/or activity of multiple secretase enzymes. HEK293 cells stably expressing APP695 produce low levels of A␤42 (40); hence, we did not measure secreted A␤42. Quantitation by ELISA (Fig. 5D) confirmed a significant reduction of secreted A␤40 peptide from cells expressing elevated levels of NHE6. We also confirmed that expression of ⌬WST 372 loss-offunction NHE6 mutant failed to significantly alter APP processing (Fig. 5E). Intriguingly, our findings with NHE6 recapitulate multiple studies in the literature, conducted over the past 2 decades reporting the effects of monensin on APP processing (46 -49). Like NHE6, monensin has been shown to increase full-length APP expression, enhance intracellular accumulation of APP, inhibit ␤-secretase activity, and alter APP processing (46 -49). Therefore, our new findings present a physiological and mechanistic basis for beneficial effects of monensin and strengthen the validity of endosomal pH as a therapeutic target for Alzheimer disease.
We next investigated whether NHE6 knockdown altered the processing of amyloid precursor protein. Lentivirus-mediated knockdown of NHE6 (by ϳ70%; p value ϭ 3.07 ϫ 10 Ϫ05 ; twotailed t test; Fig. 6A) in HEK293 cells stably expressing APP did not alter transcript levels for APP. We consistently observed a compensatory increase in NHE9 levels (by ϳ3-fold; p value ϭ 0.014; two-tailed t test) upon knockdown of NHE6, similar to our previous observations in primary astrocytes (28). Quantitation of Western blot signal intensity analysis showed ϳ80% depletion (p value ϭ 0.015; two-tailed t test) of NHE6 protein by lentiviral knockdown (Fig. 6, B and C). We confirmed that cell viability, measured by trypan blue exclusion, remained unal- (error bars) determined from triplicate measurements. Note the compensatory increase in NHE9 levels upon knockdown of NHE6, similar to previous observations in primary astrocytes (28). Shown are a representative Western blot (B) and quantification of three biological replicates (C) showing the efficacy of short hairpin RNA knockdown of NHE6 (NHE6Ϫ) normalized to tubulin levels. Control cells were transfected with scrambled shRNA, and NHE6 was detected using rabbit polyclonal antibody raised against the C terminus. Western blot analysis (D) and densitometry quantification (E) of three biological replicates demonstrated a significant increase in ␤-CTF and A␤ levels with NHE6 depletion. Holo-APP was detected using monoclonal 6E10 antibody, and tubulin was used as a loading control. a.u., arbitrary units. FEBRUARY 27, 2015 • VOLUME 290 • NUMBER 9 JOURNAL OF BIOLOGICAL CHEMISTRY 5319 tered by NHE6 knockdown (scrambled control, 95.40 Ϯ 1.12%; NHE6 knockdown, 96.45 Ϯ 1.63%; percentages represent the proportion of viable unstained cells relative to the total cells counted). Although overlapping distribution of the closely related NHE9 paralog could result in functional redundancy and thus underestimate functional effect in cultured cell models, Western blot analysis showed a significant increase in levels of ␤-CTF (2.03-fold; p value ϭ 0.003; two-tailed t test) and A␤ levels (1.67-fold; p value ϭ 0.017; two-tailed t test) upon NHE6 knockdown (Fig. 6, D and E). We also observed a reduction in full-length APP (by 17.4%) and increase in ␣-CTF (1.29-fold), although the differences did not reach statistical significance. Notably, NHE6 depletion has been previously shown to cause hyperacidification of endosomes both in cell culture models and in primary neurons (17,44). Taken together, these findings indicate that NHE6 functions as a proton leak pathway in endosomes acting as a brake against excessive amyloidogenic processing of APP by controlling luminal acidification.

NHE6 Regulates A␤ Production in Endosomes
Reduced NHE6 Expression in Alzheimer Disease Brains-We investigated whether the expression of NHE6 was altered in AD brains by analyzing publically available microarray data, as described under "Experimental Procedures." We found statistically significant reduction of NHE6 gene expression in different stages of AD (incipient, moderate, and severe), as compared with normal controls, in microarray data reported by Blalock et al. (50) (n ϭ 31; p Ͻ 0.05; ANOVA); Fig. 7A). Furthermore, independent studies of gene expression profiles from anatomically and functionally distinct brain regions showed that NHE6 was down-regulated in AD brains, with a pooled average of Ϫ1.15 (p Ͻ 0.00001; Fig. 7B and Table 1) (51-53). In contrast, pooled analysis of NHE7 and NHE9 expression did not show a

transcript expression in Alzheimer disease brains
Differential NHE6, NHE7, and NHE9 expression between AD and control (Ctrl) data sets and pooled average represented as mean difference and 95% confidence interval (CI) of log base 2 expression, as depicted in Fig. 7, B-D, were obtained as described under "Experimental Procedures." The publicly available microarray data sets are from anatomically and functionally distinct brain regions taken from three independent studies (51-53) comprising a total of 96 AD and 82 control post-mortem brains. Pooled analysis from weighted data sets (indicated in columns) showed significant down-regulation of NHE6 in AD brains. In contrast, pooled analysis of expression of related intracellular Na ϩ /H ϩ exchangers, NHE7 and NHE9, did not show significant difference between AD and control groups. NS, not significant; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.00001.

TABLE 2 NHE6 correlates with synapse genes down-regulated in AD
Pearson correlation coefficient of hippocampal expression of NHE6 with 22 different synapse genes down-regulated in AD and 20 synapse genes up-regulated in AD in microarray data reported by Blalock et al. (50,57). Correlations of NHE6 with synapse genes were calculated across all 31 subjects, including control and clinical stages of AD. Note the positive correlation of NHE6 with most genes down-regulated in AD and negative NHE6 correlation with most genes up-regulated in AD, suggesting that changes in synapse genes in AD involve a cellular mechanism that includes NHE6. Ϫ0.23 0.20 significant difference between AD and control groups (Fig. 7, C and D). To evaluate the clinico-pathological correlation of NHE6 expression with AD phenotypes, we tested the association of NHE6 gene expression with ante-mortem cognitive function evaluated by means of the mini-mental state examination (MMSE; positive correlation, p ϭ 0.0035; Fig. 7E) and with post-mortem brain Tau pathology assessed via neurofibrillary tangle scores (negative correlation, p ϭ 0.0077; Fig. 7F). These correlations suggest that reduced NHE6 expression and, by extrapolation, reduced endosomal pH may underlie AD phenotypes. We then hypothesized that changes in NHE6 are associated with broader changes in gene expression, such as downregulation of synapse genes, known to occur in AD (54,55).

Down-regulated genes
Notably, synaptic loss occurs early in pathogenesis of AD and precedes neuronal loss by several decades (56). To investigate our hypothesis, we evaluated Pearson correlation of NHE6 levels with average expression of 22 synapse-associated genes found to be down-regulated in AD in the data set of Blalock et al. (50,57). As shown in Fig. 7G, NHE6 levels strongly correlated with the synapse genes (Pearson correlation coefficient, 0.88; p ϭ 2.82 ϫ 10 Ϫ11 ), suggesting that changes in synapse genes in AD, at least in part, involve a cellular mechanism that includes NHE6. Notably (Table 2), NHE6 showed a strong association with neurexin-1 (NRXN1; Pearson correlation coefficient, 0.92), GABA-B receptor 2 (GABBR2; Pearson correlation coefficient, 0.90), and synaptic vesicle glycoprotein 2B (SV2B;  ). B, differential NHE6 expression between AD and control from data described in Table 1, represented as a Forest plot of mean difference and 95% confidence interval (CI) of log base 2 expression. Black diamond, pooled average of Ϫ1.15 (95% CI: Ϫ1.56, Ϫ0.74) (Z-test, 5.47; control, n ϭ 82; AD patients, n ϭ 96; p Ͻ 0.00001). C, differential NHE7 expression between AD and control from data described in Table 1, represented as a Forest plot of mean difference and 95% confidence interval of log base 2 expression. Black diamond, pooled average of Ϫ0.03 (Ϫ0.09, 0.04) (Z-test, 0.72; control, n ϭ 82; AD patients, n ϭ 96; p ϭ 0.47). D, differential NHE9 expression between AD and control from data described in Table 1 FEBRUARY 27, 2015 • VOLUME 290 • NUMBER 9

NHE6 Regulates A␤ Production in Endosomes
Pearson correlation coefficient, 0.86). Similarly, we had previously noted clustering of NHE6 (SLC9A6) with NRXN1 and other synapse genes expressed during normal brain development (1). In contrast, NHE6 expression did not correlate with average expression of 20 synapse genes up-regulated in AD in the data set of Blalock et al. (50,57) (Fig. 7H).
To confirm and extend our observations from microarray studies, we compared expression of NHE6 protein in gray matter punch samples of medial frontal cortex from autopsies of patients with clinically and neuropathologically classified AD versus samples from age-matched controls ( Fig. 8A and Table  3). We chose this brain region because medial frontal cortex is used in both Braak and CERAD staging as representative neocortex and is commonly affected in AD. NHE6 protein levels were significantly reduced in AD brains relative to control (Fig.  8B), consistent with down-regulation of NHE6 transcript observed in a microarray of AD brains. We note that Na ϩ /H ϩ exchangers are estimated to have high transport rates of ϳ1,500 ions/s (58), so that even small changes may be functionally significant. We observed no changes in levels of loading control GAPDH protein between AD and control brains. For comparison, we examined the expression of activity-regulated cyto-skeleton-associated (Arc) protein, a neuronal AD-linked protein known to be involved in endosomal mechanisms (59). It is worth noting that both NHE6 and Arc have been previously linked to autism and Angelman syndrome (16,60). The levels of Arc protein were elevated in AD samples, consistent with previous findings (59) (Fig. 8C). Taken together with the functional link between endosomal pH and A␤ generation, these observations suggest that NHE6 may be a potential risk factor in the etiology of AD.

DISCUSSION
Emerging Links between NHE and AD-Phylogenetic clustering by Brett et al. (61) placed three intracellular NHE isoforms (NHE6, NHE7, and NHE9) in one distinct clade, separate from plasma membrane transporters (NHE1-5). Notably, all three members of the endosomal/TGN clade have since been linked to neurodegenerative disorders, consistent with overlapping yet distinct functions. We propose that intracellular NHE isoforms are non-redundant and are likely to differ in one or more of the following: (a) brain temporal and spatial expression patterns, (b) precise subcellular distribution, (c) transport function, and (d) interacting partners and regulation (1). X-linked FIGURE 8. Reduced NHE6 protein expression in Alzheimer disease brains. A, Western blot of NHE6 and activity-regulated cytoskeleton-associated (Arc) protein in samples of gray matter from the medial frontal cortex of brains from AD patients compared with age-matched controls. Patient data are shown in Table 3. Shown is quantitation of NHE6 (B) and Arc (C) relative to GAPDH, used as a loading control (control, n ϭ 7; AD patients, n ϭ 9; *, p Ͻ 0.05; two-tailed t test). Error bars, S.E. mutations in NHE6 cause Christianson syndrome, characterized by loss of neurons and glia, massive atrophy of cerebellum, and Tau pathologies (16,18,62). In a study aimed at understanding age-related cognitive changes in the human brain, the NHE6 transcript was found to be significantly down-regulated in aged neocortical areas important for cognitive functions, memory, speech production, and language understanding (association cortex, Broca and Wernicke areas) (63). Although the gene for NHE6 (SLC9A6) has not been linked to AD by genome-wide association studies, the closely related genes SLC9A7 (NHE7) and SLC9A9 (NHE9) have recently been identified as risk factors. NHE7 has been identified as a novel late onset Alzheimer disease risk gene through analyses of neuroimaging AD phenotypes and single nucleotide polymorphism (SNP) genetic variants (26). Furthermore, SNPs in NHE9 have been associated with (a) late onset Alzheimer disease and (b) response to cholinesterase inhibitor treatment in AD (25,64). More recently, significant association between NHE9 variants and Alzheimer disease was identified in meta-analysis of a large genome-wide association study data set including ϳ5000 individuals (65). Collectively, emerging genetic, molecular, and histopathological evidence points to the TGN/endosomal NHE isoforms as potential risk factors for neurodegenerative diseases.
Vesicle Trafficking Dysfunction in AD-Endosomal pathology is an understudied component of Alzheimer disease and a promising target for therapy. Well known pathological hallmarks of AD include extracellular amyloid plaques, composed of A␤ fragments, and intracellular neurofibrillary tangles of the microtubule-binding protein Tau (66). However, these lesions appear later in disease progression, and efforts are under way to identify initial, treatable signs of dysfunction in AD. APP processing occurs within endosomal/TGN compartments by the sequential action of BACE1 and ␥-secretase (Fig. 9). Enlarged endosomes are one of the earliest cellular phenotypes associated with AD, correlating with up-regulation of Rab proteins (8). Familial, AD-linked mutations in the ␥-secretase subunit PS1 led to retention of APP within TGN compartments and increased cleavage (67), whereas silencing retromer subunits blocked retrograde transport of APP to decrease cleavage (40). Furthermore, exosomal secretion of Tau occurs early in AD and is thought to contribute to disease progression (68). These and many other data have been consolidated into a unifying hypothesis centering on vesicle trafficking dysfunction in AD. In this context, our demonstration that NHE6 alters APP processing implicates endosomal pH as an important modulator of amyloidogenic A␤ production and potentially of Tau aberrations (66). Notably, NHE6 deletion in mice leads to cellular phenotypes reminiscent of AD, including endosomal-lysosomal dysfunction, accumulation of unesterified cholesterol in endosomes, and neurodegeneration (69), although amyloidogenesis remains to be evaluated. Therefore, corroboration of our cell culture findings in the mouse model of nhe6 KO is warranted. There is growing interest in associations between neurodevelopmental and neurodegenerative disorders. Based on our data, we propose that enhanced amyloidogenic processing of APP may contribute to autism phenotypes seen in patients with NHE6 mutations. Given that NHE6 is a relatively new candidate gene for autism and intellectual disability, it is imperative to determine whether a subset of autists with deregulated NHE6 activity, either by loss-of-function mutations or by down-regulated gene expression, have a higher risk of premature aging and developing neurodegenerative disorders, such as Alzheimer disease.
Endosomal pH as a Therapeutic Target for Alzheimer Disease-Targeting the endosomal microenvironment to modulate activity of ␤and ␥-secretases has the advantage of avoiding toxicity ensuing from direct inhibition of these developmentally important enzymes. Luminal pH is set by a balance of pump and leak pathways from opposing proton transport by the V-ATPase and Na ϩ /H ϩ exchanger (1) (Fig. 9). As we show here, increasing the leak pathway by NHE6 expression alkalinizes the endosomal lumen and limits A␤ production. It has long been appreciated that V-ATPase inhibitors (e.g. bafilomycin and concanamycin), alkalizing drugs (e.g. chloroquine and ammonium chloride), and ionophore drugs that mediate Na ϩ (K ϩ )/H ϩ exchange (e.g. monensin and nigericin) alter APP processing and A␤ production (46 -49, 70 -72). Alkalization of endosomes could, at least in part, underlie APP redistribution seen in retromer/VPS35-depleted cells (40), given our observations that VPS35 deletion in yeast results in alkaline luminal pH (6). Furthermore, endosomal alkalization might explain APP redistribution and reduced A␤ production reported in APP stable cells treated with destruxin E, a V-ATPase inhibitor (45). However, V-ATPase inhibitors, alkalinizing drugs, and ionophores exert potent and multiple compartmental effects and have unwanted changes in vesicle trafficking and lysosomal function and have not been exploited for AD therapy thus far. We suggest that therapeutic elevation of endosomal pH using mildly alkalinizing drugs or by enhancing NHE6 activity will significantly reduce amyloid and Tau pathologies. For example, the weak bases bepridil and amiodarone, already in clinical use as calcium antagonists, were shown to mildly alkalinize endosomes and inhibit cleavage by ␤-secretase in cultured cells, in neurons, and in an in vivo guinea pig model (73). Membrane transporters like NHE6 are eminently "druggable targets" for Alzheimer disease and other neurological disorders that share endosomal dysfunction and trafficking defects. Given the success with discovery of small molecule potentiators for membrane transporters, such as CFTR, NHE6 could be an attractive therapeutic candidate to develop drugs to exert compartment-selective effects on luminal pH and minimize unwanted changes in vesicle function and trafficking.