Histone deacetylase–mediated regulation of endolysosomal pH

The pH of the endolysosomal system is tightly regulated by a balance of proton pump and leak mechanisms that are critical for storage, recycling, turnover, and signaling functions in the cell. Dysregulation of endolysosomal pH has been linked to aging, amyloidogenesis, synaptic dysfunction, and various neurodegenerative disorders, including Alzheimer's disease. Therefore, understanding the mechanisms that regulate luminal pH may be key to identifying new targets for managing these disorders. Meta-analysis of yeast microarray databases revealed that nutrient-limiting conditions inhibited the histone deacetylase (HDAC) Rpd3 and thereby up-regulated transcription of the endosomal Na+/H+ exchanger Nhx1, resulting in vacuolar alkalinization. Consistent with these findings, Rpd3 inhibition by the HDAC inhibitor and antifungal drug trichostatin A induced Nhx1 expression and vacuolar alkalinization. Bioinformatics analysis of Drosophila and mouse databases revealed that caloric control of the Nhx1 orthologs DmNHE3 and NHE6, respectively, is also mediated by HDACs. We show that NHE6 is a target of the transcription factor cAMP-response element–binding protein (CREB), a known regulator of cellular responses to low-nutrient conditions, providing a molecular mechanism for nutrient- and HDAC-dependent regulation of endosomal pH. Of note, pharmacological targeting of the CREB pathway to increase NHE6 expression helped regulate endosomal pH and correct defective clearance of amyloid Aβ in an apoE4 astrocyte model of Alzheimer's disease. These observations from yeast, fly, mouse, and cell culture models point to an evolutionarily conserved mechanism for HDAC-mediated regulation of endosomal NHE expression. Our insights offer new therapeutic strategies for modulation of endolysosomal pH in fungal infection and human disease.

The endosome is a busy way station that handles cargo traffic at the cross-roads of degradation and recycling. An important hallmark of the endolysosomal system is a pH gradient that is increasingly acidic through the transition from early, recycling, and late endosomes to the lysosome (1). Precise tempo-spatial pH lumen regulation within each of these specialized compartments is critical for a range of functions, including controlling cargo sorting, vesicular budding and formation of intraluminal vesicles, dissociation of internalized ligands, receptor recycling and turnover, modulating enzyme activity, antigen processing, and cellular signaling (2).
Endosomes are acidified by the electrogenic pumping of protons by the V-type H ϩ -ATPase in conjunction with vesicular chloride transporters of the CLC family that shunt the membrane potential, effectively allowing a build-up of luminal protons (3,4). Proton leak mechanisms consisting of proton conduction channels or proton-coupled antiporters limit the acidification, and the balance of pump and leak pathways defines the pH set point. Although the molecular identity of vesicular proton channels remains poorly defined, more recently, endosomal Na ϩ /H ϩ exchangers (eNHE) 3 have emerged as the dominant leak pathway for luminal protons. Because exchanger mechanisms have high transport rates of ϳ1,500 ions/s (5), they overwhelm the capacity of the proton pump so that even small changes in eNHE activity or expression result in significant pH changes within the endolysosomal system.
The fundamental importance of endosomal pH is highlighted by the fact that a growing number of human diseases are associated with mutations in V-ATPase subunits (e.g. osteopetrosis, renal tubular acidosis, cutis laxa), chloride transporters (e.g. Dent's disease), and eNHE (Christianson syndrome, autism, and attention deficit hyperactivity disorder) (3,4,6). Furthermore, defective endolysosomal pH regulation is being increasingly linked to cellular aging, amyloidogenesis, synaptic dysfunction, and neurodegenerative disorders, including Alzheimer's disease (7,8). Therefore, the discovery of novel regulators of endolysosomal pH may be crucial to identifying new diagnostic and therapeutic targets for these disorders.
As a starting point, we turned to the yeast Saccharomyces cerevisiae, which has powerfully informed our understanding of fundamental mechanisms of endocytosis and vesicular traffic that are functionally conserved with higher organisms (9). By leveraging data from yeast to fly and mammalian models, we discovered an evolutionarily conserved mechanism for epigenetic control of eNHE by histone deacetylases (HDACs) in response to nutrient availability. In mammalian cells, HDACmediated control of endolysosomal pH occurs by transcriptional regulation of the endosomal Na ϩ /H ϩ exchanger NHE6 (SLC9A6) via the cAMP-response element-binding protein (CREB) pathway. We show that elevation of cAMP levels by the FDA-approved drug Rolipram results in CREB activation, increased NHE6 expression, and restoration of defective A␤ clearance in the apoE4 astrocyte model of late onset Alzheimer's disease.
Histone acetylation has been widely recognized as a key modulator of global chromatin structure that dynamically couples extracellular signals to gene transcription. Whereas lysine acetylation on histone tails interferes with generation of higherorder chromatin structures and promotes access and binding of transcription factors, histone deacetylation favors chromatin packing and represses gene transcription (10 -12). HDAC inhibitors have been approved for the treatment of hematologic cancers and are being considered as a promising therapy to reverse disease-associated epigenetic states in a range of cardiovascular, neurodegenerative, and inflammatory diseases (11)(12)(13)(14). Given their broad application, our findings offer potential therapeutic options to treat endolysosomal pH dysfunction in autism, AD, and other neurodegenerative disorders.

Nutrient control of endolysosomal pH in yeast
Yeast microarray data sets are a valuable resource for discovery-driven mining efforts. A meta-analysis of 45 microarray experiments that included 937 samples revealed increasing expression of Nhx1, with growth phase stationary Ͼ mid-log Ͼ Figure 1. Nutrient control of endolysosomal pH in yeast. A, microarray data analysis revealed increasing expression of Nhx1 with growth, sampled as indicated in the growth-phase representation. Note that maximal expression was in stationary phase, which is associated with glucose/nitrogen limitation). B, qPCR validation of Nhx1 expression showed 3.2-fold up-regulation of Nhx1 transcript in stationary-phase yeast, compared with mid-log phase (****, p Ͻ 0.0001; n ϭ 3; Student's t test). No change in V-ATPase subunit Vma1 levels was observed (p ϭ 0.092; Student's t test). C, glucose removal from mid-log-phase yeast cultures showed a ϳ2-fold increase in Nhx1 mRNA within 60 min of depletion (***, p Ͻ 0.001; n ϭ 3; Student's t test). Note significant down-regulation of V-ATPase subunits with acute glucose depletion (***, p Ͻ 0.001, Vma1; ***, p Ͻ 0.001, Vma7). D, conversely, the addition of glucose (2%) to stationary-phase cultures showed significant decrease in Nhx1 mRNA within 30 min (***, p Ͻ 0.001; n ϭ 3; Student's t test). Note significant up-regulation of V-ATPase subunits with glucose addition (*, p Ͻ 0.05, Vma1; ***, p Ͻ 0.001, Vma7). C and D, in contrast to Vma1 and Vma7, secretory pathway isoform Stv1 expression showed modest or no changes in response to glucose, suggesting that the effect of glucose regulation on the V-ATPase is largely localized to endosomal/vacuolar compartments. E, schematic to show that Nhx1 functions as a leak pathway for protons within the endosomal/prevacuolar compartment, to balance the proton pump activity of the V-ATPase (Vma). F and G, representative micrographs (F) and quantification (G) showing loss of quinacrine fluorescence with acute glucose depletion, similar to V-ATPase-deficient vma7⌬ mutants, pointing to increased vacuolar alkalization (****, p Ͻ 0.0001; n ϭ 100; Student's t test). NS, not significant. Error bars, S.D.

Histone deacetylases control endolysosomal pH
early log (Fig. 1A), which we independently confirmed; Nhx1 transcript increased by 3.2-fold in stationary-phase cells, compared with mid-log phase (Fig. 1B). Stationary-phase growth in yeast is associated with depletion of nutrients. To determine whether growth-phase regulation of Nhx1 was linked to glucose availability, we tested the effect of glucose removal from mid-log phase yeast cultures. Nhx1 mRNA was increased by nearly 2-fold within 60 min of glucose removal (p Ͻ 0.001; Fig.  1C). In contrast, the addition of glucose (2%) to stationaryphase cultures elicited significant decrease in Nhx1 mRNA within 30 min (p Ͻ 0.001; Fig. 1D).
Nhx1 functions as a leak pathway for protons within the endosome/prevacuolar compartment, in opposition to the proton pump activity of the V-ATPase (15) (Fig. 1E). The balance of proton pump and leak sets the pH within the endolysosomal compartments. Therefore, we also analyzed expression of V-ATPase subunits under the same experimental conditions. Although we did not observe a growth phase-dependent transcriptional change in the V-ATPase catalytic subunit Vma1 (Fig. 1B), V-ATPase subunits were significantly down-regulated with acute glucose depletion (p Ͻ 0.001, Vma1; p Ͻ 0.001, Vma7; Fig. 1C) and up-regulated upon glucose addition (p Ͻ 0.05, Vma1; p Ͻ 0.001, Vma7; Fig. 1D) in contrast to Nhx1. Previous studies have shown that yeast V-ATPase subunit Stv1 is selectively targeted to the Golgi/secretory pathway and is not required for vacuolar acidification and function (16,17). In contrast to Vma1 and Vma7, Stv1 expression showed modest or no changes in response to glucose (Fig. 1, C and D), suggesting that the effect of glucose regulation on the V-ATPase is largely localized to late endosomal/vacuolar compartments.
We predicted that reciprocal transcriptional changes in Nhx1 and V-ATPase subunits would alter the pH of the yeast vacuole. Vacuolar pH can be probed in situ with the pH-sensitive fluorescent dye quinacrine, which accumulates in acidic environments (18). Following glucose depletion, yeast vacuoles showed little or no quinacrine staining, similar to V-ATPase-deficient vma7⌬ mutants (Fig. 1, F and G; p Ͻ 0.0001, n ϭ 100). Consistent with the potential role for Nhx1 in regulating nutrient control of endolysosomal pH in yeast, nhx1⌬ mutants showed severely reduced survival during the stationary-phase growth (19). Post-translational regulation of V-ATPase disassembly and reassembly in response to glucose concentration is well established (20,21). Here, we conclude that glucose also elicits reciprocal transcriptional changes in proton pump and leak mechanisms to regulate the pH gradient across yeast vacuoles.

Regulation of endolysosomal pH in yeast is mediated by a histone deacetylase
We extended our unbiased bioinformatics approach to analyze data from 284 microarray studies that comprised a wide range of experimental conditions, including gene deletion, overexpression, and mutations; drug/toxin treatment; nutrient limitation or excess; metabolic/environmental stress; and more. We identified experimental conditions eliciting a minimum of Ϯ2-fold change in NHX1 gene expression ( Fig. 2A and Fig. S1A). The highest up-regulation of Nhx1 was observed under conditions of (i) genetic deletion of Rpd3, (ii) glucose depletion, and (iii) nitrogen depletion. Pathway analysis revealed that loss or inhibition of HDAC activity of Rpd3 was common to all three of these conditions (22,23) (Fig. 2B). Thus, nutrient (glucose or nitrogen) depletion preserves histone acetylation by replacement and inactivation of the HDAC-Rpd3 complex from the promoters of specific genes (22). Consistent with this finding, Rpd3 deletion phenocopies calorie restriction by enhancing lifespan of yeast cells (24,25). Conversely, all five conditions leading to maximal down-regulation of Nhx1 were linked to depletion of Abf1, an essential transcription factor that is negatively regulated by Rpd3 (26 -28) ( Fig. 2A and Fig. S1A). Recent evidence suggests that Abf1 is activated during nutritional stress (29). Furthermore, Abf1 is predicted to bind to the NHX1 promoter (Fig. S1B), and chromatin immu-

Histone deacetylases control endolysosomal pH
noprecipitation with DNA microarray (ChIP-on-chip) data from the Saccharomyces Genome Database (SGD) provided experimental evidence for Abf1 binding to the NHX1 promoter (30). For comparison, depletion of another general regulatory transcription factor, Rap1, had no effect on Nhx1 expression ( Fig. 2A and Fig. S1A). Although more direct studies are needed to confirm Rpd3-Abf1 occupancy of the NHX1 promoter, bioinformatics analysis of all top hits in NHX1 gene expression (both up-and down-regulation) funneled into a single pathway: the Rpd3-Abf1-Nhx1 axis (Fig. 2C).
To confirm and extend these in silico observations, we independently demonstrated that Nhx1 transcript was elevated by 1.9-fold in the rpd3⌬ mutant, relative to isogenic WT yeast (Fig.  3A). Other histone deacetylase gene deletions conferred more modest induction (1.35-fold in hda1⌬; Fig. 3A) or repression (0.6-fold in sir2⌬; Fig. 3A). All three yeast deletion strains showed similar growth (Fig. S2A). Restoration of Rpd3 by ectopic expression under a galactose-inducible promoter in rpd3⌬ not only abolished Nhx1 up-regulation in stationary phase, but also caused further repression (68% lower) in midlogarithmic cells and inhibited yeast growth (Fig. S2, B and C).
Yeast Rpd3 is a target of the histone deacetylase inhibitor and fungal antibiotic trichostatin A (TSA), and the transcriptional profile of TSA-treated WT yeast is similar to that of rpd3⌬ yeast (23). We show that that TSA treatment of WT yeast elicited significant 2.9-fold up-regulation of Nhx1 transcript (Fig. 3B), mimicking the effect of Rpd3 deletion. In contrast, we observed significant down-regulation of V-ATPase subunits Vma1 (ϳ30% lower; Fig. 3C) and Vma7 (ϳ37% lower; Fig. 3D) in TSAtreated yeast, compared with vehicle treatment. In contrast to Vma1 and Vma7, the levels of Stv1 did not change significantly with TSA treatment (Fig. S2D), suggesting that the effect of TSA is specific to V-ATPase localized to the vacuolar compartment. Figure 3. Regulation of endolysosomal pH in yeast is mediated by a histone deacetylase. A, deletion of Rpd3 (rpd3⌬ strain) increased NHX1 transcript by 1.9-fold relative to isogenic WT yeast (**, p ϭ 0.002; n ϭ 3; Student's t test). Deletion of two other histone deacetylase genes led to lesser NHX1 induction (1.35-fold in hda1⌬; ***, p ϭ 0.0008; n ϭ 3; Student's t test) or repression (0.6-fold in sir2⌬; ***, p ϭ 0.0003; n ϭ 3; Student's t test). B, TSA treatment of WT yeast elicited significant 2.88-fold up-regulation of Nhx1 transcript (**, p ϭ 0.001; n ϭ 3; Student's t test) mimicking the effect of Rpd3 deletion. C and D, qPCR analysis of V-ATPase subunits showed significant down-regulation of Vma1 (ϳ30% lower; ***, p ϭ 0.0003; n ϭ 3; Student's t test) (C) and Vma7 (ϳ37% lower; **, p ϭ 0.001; n ϭ 3; Student's t test) (D) in TSA-treated yeast, compared with vehicle treatment. E and F, representative micrographs (E) and quantification (F) showing loss of quinacrine fluorescence with TSA treatment, similar to treatment with a V-ATPase inhibitor concanamycin A (ConA), pointing to increased vacuolar alkalization (***, p Ͻ 0.0001; n ϭ 100; Student's t test). G, alkaline pH sensitivity is a defining phenotype of increased vacuolar pH. TSA-treated cells show dose-dependent reduction of growth at medium pH of 7.0 as compared with pH of 4.3 (***, p Ͻ 0.001; n ϭ 4; Student's t test). H, a spotting assay on YPD agar plates was performed to demonstrate hypersensitivity to TSA in the vma7⌬ yeast, lacking V-ATPase activity, indicative of synthetic lethality. See related Fig. S2. Error bars, S.D.

Histone deacetylases control endolysosomal pH
The transcriptional up-regulation of Nhx1 and concomitant down-regulation of vacuolar V-ATPase subunits by TSA is predicted to alkalinize the yeast vacuole (3). Compared with robust quinacrine fluorescence in WT yeast vacuoles (18), TSA treatment drastically reduced vacuolar accumulation of quinacrine, similar to the effect of the well-characterized V-ATPase inhibitor concanamycin A, pointing to increased vacuolar pH (Fig. 3, E and F). Alkaline pH sensitivity is a defining phenotype of V-ATPase mutants, indicative of the reduced ability to acidify vacuoles (31,32). We found that TSA-treated cells show dosedependent reduction of growth at alkaline medium pH (Fig.  3G). Hypersensitivity to TSA in the vma7⌬ yeast, lacking V-ATPase activity, is evidenced by synthetic lethality (Fig. 3H and Fig. S2E). TSA synergizes with the antifungal action of an azole drug fluconazole that we previously showed to act via inhibition of V-ATPase activity (31) (Fig. S2F). Together, these data provide strong evidence for regulation of yeast vacuolar pH by HDAC inhibition.

Evolutionary conservation of eNHE regulation by nutrient availability and histone deacetylases
To investigate whether the mechanism of up-regulation of eNHE in response to nutrient limitation and HDAC activity is evolutionarily conserved, we evaluated publicly available expression data in the fruit fly Drosophila melanogaster. Fruit fly has one eNHE isoform, DmNHE3, which, like its mamma-lian cousin NHE6, is highly expressed in the brain (33) (Fig. 4A). Analysis of a microarray data set studying the effects of starvation in Drosophila larvae (GSE14531) revealed significant down-regulation of class I HDAC DmRpd3 (Ϫ2.14-fold; We also evaluated the effect of nutrient depletion on transcript levels of Vha68 and Vha14, which are orthologs of Vma1 and Vma7, respectively. Unlike our observations in yeast, no change in expression of these V-ATPase subunits was noted in Drosophila, suggesting that the transcriptional effect of starvation might be limited to endosomal NHE in higher eukaryotes. Chromatin-modifying complexes are conserved between yeast and mammals. Eighteen mammalian HDACs have been described, grouped into four functional classes based on similarity to yeast orthologs (10). Independent studies have shown that insulin, the principal hormone orchestrating response to glucose availability, alters HDAC nuclear translocation in mammalian cells (35). Calorie restriction has been shown to improve insulin sensitivity and mimic the effects caused by

Histone deacetylases control endolysosomal pH
HDAC inhibitors, including on histone acetylation and activation of CREB (36 -39). Intriguingly, the mammalian eNHE ortholog NHE6 is a reported insulin-responsive protein, and its subcellular trafficking is regulated by insulin (40). Furthermore, in neurons, loss of NHE6 has been reported to attenuate brainderived neurotrophic factor (BDNF) signaling that is downstream of insulin and the CREB cascade (41,42). Given that Na ϩ /H ϩ exchangers are estimated to have very high transport rates of ϳ1,500 ions/s (5), even small changes in expression may be functionally significant. Therefore, to test whether calorie restriction or HDAC inhibitors enhance mammalian eNHE expression, we first analyzed publicly available expression data on caloric restriction (ϳ10% lower than ad lib intake) in the mouse (Fig. 5A). These data revealed significant up-regulation of NHE6 in the neocortex (Fig. 5B) but not of the closely related endosomal NHE9 or V-ATPase subunit Vma1/ATP6V1A (Fig.  S4, A and B) (43). Next, we analyzed two independent microar-ray experiments studying the effect of TSA in cell culture at shorter (2 h; GSE8488) and longer (18 h; GSE9247) time points (44,45). Expression data from both of these experiments revealed significant up-regulation of NHE6 with TSA treatment (Fig. 5, C and D), but not NHE9 (Fig. S5, A-D) or NHE1 (Fig. S5, B-E). Activation of transcription factor CREB is a common response to both calorie restriction and pharmacological HDAC inhibition (12,39). Thus, for comparison, we evaluated a prototypic CREB target BDNF and documented significant up-regulation with calorie restriction in mice (Fig. S4C) and HDAC inhibition in cell culture (Fig. S5, C-F), as predicted. These observations from yeast, fly, mouse, and cell culture models support our hypothesis of an evolutionarily conserved mechanism for regulation of eNHE expression by histone deacetylases.

NHE6 is a transcriptional target of CREB
Next, we attempted to identify the HDAC-regulated transcription factor equivalent to yeast Abf1 in higher eukaryotes. In the absence of direct Abf1 orthologs in the Drosophila or human genome, we looked for parallel starvation response pathways between human and yeast. Various studies suggest that CREB transcription factor regulates the response to nutrient deprivation and glucose availability in mammalian cells, analogous to the Abf1 transcription factor in yeast (39,46). Both class I (HDAC1) and class II (HDAC4) HDACs directly interact with CREB and negatively regulate its function (47,48). Furthermore, similar to the Abf1 transcription factor in yeast, the phosphorylation state of CREB is tightly regulated in response to nutrient and glucose availability (39,49). Based on all of these observations, we hypothesized NHE6 as a potential CREB target (Fig. 6A).
Analysis of a publicly available data set of CREB target genes, from ChIP with DNA microarray (ChIP-on-chip) (50), using the recommended cutoff for ChIP-positive genes (binding ratio Ն1.5 and log 10 p value Ն3) revealed NHE6 as a potential CREB target. The ChIP binding data for NHE6 were similar to those of DUSP1, a known CREB target, whereas no CREB occupancy was detected for the related NHE7 isoform (Fig. 6, B and C). Consistent with our hypothesis in Fig. 6A, TSA treatment resulted in a ϳ5.2-fold increase in NHE6 expression in HEK293 cells (Fig. 6D), whereas, modest increases were observed in NHE7 (ϳ1.3-fold; Fig. S6A) and NHE9 (ϳ1.4-fold; Fig. 6D), and no changes in expression of V-ATPase subunits V0a1 (lysosomal) and V0a2 (putative endosomal) were observed (Fig. S6,  B and C). We used the firefly and Renilla luciferase reporter gene system to demonstrate significant activation of cAMPresponse element (CRE) with TSA treatment (10 M, 12 h), relative to vehicle control (p ϭ 0.0064; Fig. 6, D and E). To establish a causal link between CREB activation and NHE6 expression, we treated HEK293 cells with forskolin, an adenylyl cyclase activator, and confirmed a significant increase in NHE6 transcript levels that was blocked by expression of a constitutively active/nuclear HDAC4 mutant (S246A/S467A/S632A) (Fig. 6G).
To directly test whether NHE6 is a CREB target, we expressed FLAG-tagged CREB1 plasmid in HEK293 cells and assessed (i) functional expression of NHE6 and (ii) binding of

Histone deacetylases control endolysosomal pH
CREB to the NHE6 promoter by ChIP-qPCR (Fig. 7A). Expression of CREB1 in HEK293 cells significantly elevated NHE6 transcript (Fig. 7B), resulting in alkalization of endosomal pH from 5.9 Ϯ 0.03 to 6.12 Ϯ 0.05 (Fig. 7C). In contrast, no changes in NHE6 transcript or endosomal pH were observed with the nonphosphorylatable mutant of CREB1 (S133A; Fig. 7, B and C). ChIP was performed with a ChIP grade anti-FLAG mouse antibody (Fig. 7D). Using real-time qPCR, we amplified three predicted CREB binding sites in the promoter of NHE6 (SLC9A6) at positions Ϫ12, ϩ1, and ϩ2 kb relative to the transcription start site (TSS) (Fig. 7D and Fig. S6D). For comparison, we used primers targeted to the NHE9 (SLC9A9) promoter at position Ϫ1 kb relative to TSS (Fig. 7E). We confirmed CREB binding to NHE6 promoter at all three predicted sites (Fig. 7,  F-H). In contrast, we observed no significant binding of CREB to the NHE9 promoter (Fig. 7I). Specific pulldown of CREB1 was confirmed by probing the ChIP samples with an antibody against CREB raised in rabbits (Fig. 7J). Consistent with our ChIP-qPCR findings, we note that the pattern of NHE6 expression closely parallels that of CREB1 through normal human brain development (Fig. S7K; Pearson correlation, 0.62) and in all areas of the brain (Fig. 7L). In contrast, NHE9 expression showed no correlation with CREB ( Fig. 7M; Pearson correlation, Ϫ0.018). Taken together, our findings reveal a novel mechanism of CRE-mediated transcriptional regulation of NHE6 and provide a molecular mechanism for nutrient-and HDAC-dependent regulation of endosomal pH.

Control of NHE6 expression by pharmacological targeting of CREB pathway
Christianson syndrome patients, with loss of function mutations in NHE6, clinically present with neurodevelopmental disorders of intellectual disability and severe autism and show agedependent hallmarks of neurodegeneration, including loss of Figure 6. Regulation of NHE6 expression is mediated by a histone deacetylase. A, schematic to draw parallels between regulation of eNHE in yeast (top) and mammalian cells (bottom). CREB transcription factor regulates response to nutrient deprivation and glucose availability in mammalian cells, analogous to the Abf1 transcription factor in yeast. Both class I (HDAC1) and class II (HDAC4) HDACs directly interact with CREB and negatively regulate its function. Based on this predictive model, we hypothesized NHE6 as a potential CREB target. B and C, ChIP-on-chip (ChIP with DNA microarray) data analyzed using the recommended cutoff for ChIP-positive genes (binding ratio Ն1.5 and log 10 p value Ն3) revealed NHE6 as a potential CREB target. Note that the ChIP binding data for NHE6 were similar to those of DUSP1, a known CREB target, whereas no CREB occupancy was detected for the related NHE7 isoform. D, qPCR analysis demonstrating prominent ϳ5.2-fold increase in NHE6 expression (****, p Ͻ 0.0001; n ϭ 3; Student's t test) and modest increases in NHE9 (ϳ1.4-fold; **, p ϭ 0.0011; n ϭ 3; Student's t test) with TSA treatment of HEK293 cells. E and F, luciferase assay to confirm TSA-induced activation of CREB using a CRE that drives a firefly luciferase reporter gene and is measured luminometrically (E). Renilla luciferase driven by a constitutively active SV40 promoter was used to normalize for variation of cell number and transfection efficiency. Significant activation of the CRE was seen with TSA treatment (10 M; 12 h; F) in HEK293 cells, relative to vehicle control (**, p ϭ 0.0064; n ϭ 3; Student's t test). G, qPCR analysis to determine the potential of forskolin to augment the expression of NHE6 in HEK293 cells. Note the significant increase in NHE6 transcript levels with forskolin treatment (***, p ϭ 0.0005; n ϭ 3; Student's t test), which was blocked by expression of a constitutively active/nuclear HDAC4 mutant (S246A/S467A/S632A) (p ϭ 0.116; n ϭ 3; Student's t test). NS, not significant. See related Fig. S6. Error bars, S.D.

Histone deacetylases control endolysosomal pH
cerebellar and cortical mass, prominent glial pathology, and hyperphosphorylated Tau deposition (3,(51)(52)(53). Recent revelations on the shared pathology of defective A␤ clearance and altered processing of amyloid precursor protein between autism and Alzheimer's disease suggest that endosomal pH regulation may be one critical mechanistic link between neurodevelopmental and neurodegenerative disorders (54). NHE6 gene expression is depressed in both AD and autism (55,56). CREmediated transcriptional regulation of NHE6 expression offers opportunities to expand the repertoire of drugs that could potentially correct human pathologies resulting from NHE6 down-regulation and aberrant endosomal hyperacidification (Fig. 8A). For example, rolipram, a PDE4 inhibitor that elevates cAMP levels, resulting in CREB activation, has shown efficacy as memory enhancer and for amelioration of A␤-related neuropathology in mouse models (57).
We used mouse astrocytes with knock-in of human ApoE alleles that alter genetic risk for Alzheimer's disease (58). Recently, we showed that NHE6 expression is severely reduced in astrocytes carrying the pathological apoE4 variant, relative to the isogenic nonrisk ApoE3 allele (8). Consistent with this observation, CREB is down-regulated in apoE4 brains (59). First, we confirmed CREB-mediated NHE6 regulation in apoE3 astrocytes using the CREB inhibitor (KG501); we observed dose-dependent down-regulation of NHE6 transcript levels with KG501 treatment (Fig. 8B). Rolipram was effective in dosedependent activation of CREB targets in apoE4 astrocytes, including the prototypic CREB target BDNF (Fig. S7A), and astrocyte-specific CREB targets, ALDH1A1 and FBXO2 (60) (Fig. S7, B and C). We observed significant induction of CREB mRNA by rolipram as well (Fig. 8C), consistent with similar observations in rat hippocampus (61). Here, we demonstrate a dose-dependent increase in NHE6 transcript levels with rolipram treatment (1.25-10 M) in apoE4 astrocytes (Fig. 8D). To evaluate the efficacy of rolipram treatment on NHE6-mediated function, we monitored A␤ clearance by endocytosis in cultured apoE4 astrocytes. Cell-associated A␤ was monitored using flow cytometry and confocal microscopy following a short (1-h) incubation with fluorescently tagged peptide (58, 62) (Fig. 8E). We show that rolipram elicits a significant increase in A␤ clearance, evidenced by an increase in cell-associated fluorescence by flow cytometry of apoE4 astrocytes (Fig.  8F). Similarly, endocytosed A␤, monitored by confocal microscopy, exhibited an average of 2.66-fold increase following treat-ment with rolipram (Fig. 8, G and H). Although the increase in cell-associated A␤ could be, in part, due to a reduction in A␤ degradation, especially at longer time points, our experimental conditions are consistent with increased A␤ uptake. Taken together, our findings suggest that CREB activation of NHE6 expression is translatable into correction of A␤ clearance deficits in apoE4 astrocytes.

Nutrient control of endolysosomal pH
The discovery of eNHE in the mid-1990s and subsequent work in a range of model organisms and mammalian systems have established the fundamental importance of these ion transporters in human health and disease (3,63). Although the contribution of the V-ATPase in nutrient control of endolysosomal pH is well established in yeast (20,21), the role of eNHE in this mechanism was previously unknown. Here, we leveraged public data from yeast, fly, and mammalian studies to make an unbiased in silico prediction leading to the discovery of an evolutionarily conserved mechanism for regulation of eNHE expression by nutrients and HDACs. Previous studies have established that glucose triggers rapid, post-translational regulation of disassembly and reassembly of V-ATPase to regulate the pH gradient across yeast vacuoles (20,21). In this study, we show that glucose also elicits reciprocal transcriptional changes in Nhx1 and V-ATPase subunits to control vacuolar pH. This finding is in agreement with previous work showing severely reduced survival of Nhx1 mutants during nutrient-depleted stationary phase (19).
The regulation of endolysosomal pH by Nhx1 and V-ATPase (or their orthologs in higher eukaryotes) might promote cell survival during nutrient depletion or starvation by at least three potential mechanisms, as follows. (i) Vacuole acidity facilitates proton-dependent import of amino acids into the lumen through Avt1, a neutral amino acid transporter important for starvation resistance (18,64). In nutrient-depleted cells, a decrease in vacuolar acidification would reduce vacuolar import of amino acids and help maintain cytoplasmic amino acid homeostasis (Fig. S8, A and B). (ii) Transcriptional control of V-ATPase subunits could supplement post-translational mechanisms of V-ATPase disassembly as an energy-saving mechanism to conserve ATP under nutrient-limiting conditions. (iii) We have previously shown that Nhx1 regulates vesi- Figure 7. CREB-mediated transcriptional regulation of NHE6 expression. A, experimental framework to directly test whether NHE6 is a CREB target. FLAG-tagged CREB1 plasmid was expressed in HEK293 cells and (i) NHE6 expression and its functional outcome (elevation of endosomal pH) and (ii) binding of CREB to NHE6 promoter by ChIP-qPCR were assessed. B and C, expression of CREB1 in HEK293 cells significantly elevated NHE6 transcript (****, p Ͻ 0.0001; n ϭ 3; Student's t test; B), resulting in alkalization of endosomal pH from 5.9 Ϯ 0.03 to 6.12 Ϯ 0.05 (**, p ϭ 0.0075; n ϭ 3; Student's t test; C). No changes in NHE6 transcript (p ϭ 0.636; n ϭ 3; Student's t test; B) or endosomal pH (p ϭ 0.916; n ϭ 3; Student's t test; C) were observed with the nonphosphorylatable mutant of CREB1 (S133A). D and E, schematic of NHE6 (top; D) and NHE9 (bottom; E) promoter element showing transcription start site, CREB-binding sites, and locations of ChIP-qPCR assays. F-H, ChIP-qPCR amplification of three predicted CREB-binding sites in the promoter of NHE6 at positions Ϫ12, ϩ1, and ϩ2 kb relative to the TSS in HEK293 cells. Note significant CREB binding to the NHE6 promoter at all three predicted sites. I, no significant binding of CREB to NHE9 promoter was identified using a primer targeting position Ϫ1 kb relative to the TSS. J, Western blotting to confirm specific pulldown of CREB1 by probing the ChIP samples from HEK293 cells with an antibody against CREB raised in rabbits. ChIP was performed using an antibody raised in mice. Note the high-molecular weight smear due to formaldehyde cross-linking. K and L, the pattern of NHE6 expression closely parallels that of CREB1 through normal human brain development (K) (Pearson correlation, 0.62; n ϭ 244; ****, p Ͻ 0.0001) and in all areas of the brain (L). 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. M, NHE9 expression showed no correlation with CREB1 (Pearson correlation, Ϫ0.018; n ϭ 244; p ϭ 0.782). Data were obtained from the Allen Brain atlas. NS, not significant. Error bars, S.D. See related Fig. S6. cle trafficking and lysosomal biogenesis to control internalization of plasma membrane proteins and degradation in vacuoles (15,63). Increased Nhx1 expression in glucose-depleted cells might therefore help promote endocytosis and provide resources for survival during glucose starvation (65). Intriguingly, eNHE mutants in Caenorhabditis elegans (Nhx5) and fly (DmNHE3) were found to be resistant to metformin, a biguanide drug commonly used to treat type-2 diabetes, potentially through dysregulation of autophagy (66). Given that metformin is known to induce a caloric restriction-like state (67), our findings suggest that up-regulation of eNHE and alkalinization of endosomal pH may underlie response to metformin therapy.

Antifungal drugs target endolysosomal pH
Pathogenic fungi have become a leading cause of morbidity and mortality in the world (68). Fungi need V-ATPase function to infect their hosts, and genetic inactivation of vacuole-specific V o subunit Vph1, but not secretory pathway-specific subunit Stv1, abolished virulence and invasive growth of C. albicans, underscoring the importance of endolysosomal pH as a drug target (16,17). Consistent with this finding, the antifungal activity of azoles, which inhibit biosynthesis of fungus-specific sterol, ergosterol, is mediated by reduced V-ATPase activity in ergosterol-deficient vacuoles (31). Similarly, decreased V-AT-Pase activity may underlie the antifungal activity of amphotericin B, which is known to directly bind and sequester ergosterol in yeast cells (69). Furthermore, many naturally occurring antifungal compounds, such as terpenoid phenols, function at least in part by disrupting V-ATPase function (70). Recent studies have documented the potential benefits of repurposing FDAapproved amphipathic drugs, such as amiodarone, already in clinical use as calcium antagonists, that partition into acidic compartments, alkalinize endolysosomes, and potentiate antifungal effects and reverse resistance to azole drugs (71). Here we show that an antifungal HDAC inhibitor, TSA, increases vacuole pH and enhances the antifungal effects of fluconazole. Intriguingly, both amiodarone and HDAC inhibitors confer significant neuroprotection in cell culture and animal models (11,12,72). Together, these studies and our current findings makes a strong case for (i) repurposing antifungal drugs as potential drugs to correct human pathologies arising from dysregulation of endolysosomal pH and (ii) repurposing FDA-approved drugs having off-target effects on pH homeostasis as candidates for the development of new therapeutic strategies for fungal disease.

Targeting endolysosomal pH in human diseases
Dysregulation of endolysosomal pH is an emerging theme in pathogenesis of a wide range of human diseases, ranging from neurodegeneration to cancer (7,73). There is clearly potential for intervention to exploit the disease-modifying effects of endosomal pH, and this should be investigated further. From lessons learned with bakers' yeast, we know that gene disruption of NHX1 leads to cellular phenotypes reminiscent of Alzheimer's disease (AD) that include an enlarged endosome/prevacuolar compartment, hyperacidic luminal pH, enhanced proteolysis of the chaperone protein Vps10 (a homolog of the AD susceptibility factor SORL1), mistrafficking of vacuolar hydrolases, and profound deficiencies in lysosomal cargo delivery (15,74). Several studies have identified links between eNHE and a host of neurological disorders, including autism, intellectual disability, epilepsy, Parkinson's disease, multiple sclerosis, and more recently late-onset AD (3,8,55,75), although mechanisms underlying eNHE regulation have remained obscure.
Here we show that expression of the Christianson syndrome protein NHE6 is under the control of CREB transcription factor. A translatable finding from our studies is that pharmacological HDAC inhibition (by TSA) or CREB activation (by forskolin or rolipram) elevates endosomal pH and has the potential to mitigate human pathologies, such as AD and autism, associated with aberrant endosomal hyperacidification. Furthermore, regulation of endolysosomal pH may at least partially explain the working mechanism of HDAC inhibitors in cancer therapy. Tumor cells in the periphery have better blood supply, whereas cells inside tumor masses are under starved conditions, because nutrients are not fully supplied to these cells. As we have shown earlier, a subset of brain cancer glioblastoma multiforme show significant up-regulation of eNHE isoform NHE9 that might help in the survival of tumor cells in nutrient-limited conditions (76). Together, these data suggest that, by governing fundamentally critical endosomal pH homeostasis, eNHE plays an important role in mediating nutrient control of endolysosomal pH. This importance makes them an attractive target as antifungal drugs and in several human diseases. Exploration of the conserved mechanism for regulation of endolysosomal pH by histone deacetylases provides a foundation for understanding the pathways regulating pH homeostasis and a platform for the development of drugs that could potentially correct human pathologies resulting from dysregulated endolysosomal acidification. Figure 8. Control of NHE6 expression by pharmacological targeting of CREB pathway. A, extracellular signals induce adenylyl cyclase, which elevates cAMP levels and activates the cAMP-dependent protein kinase A (PKA), which in turn induces CREB phosphorylation at serine 133. Active phospho-CREB translocates to the nucleus and activates transcription of NHE6, which contains CRE in its promoter. At least three different pharmacological approaches could be used to activate CREB and enhance NHE6 transcription. First, HDAC inhibitors (HDACi) activate CREB by attenuating HDAC-mediated repression. Second, forskolin (FOR), an agonist of adenylyl cyclase, stimulates production of cAMP and activates CREB. Third, rolipram (ROL), a PDE4 inhibitor, inhibits hydrolysis of cAMP and activates CREB. B, qPCR analysis to confirm CREB-mediated NHE6 regulation in apoE3 astrocytes using the CREB inhibitor (KG501). Note the dose-dependent down-regulation of NHE6 transcript levels with KG501 treatment (****, p Ͻ 0.0001; n ϭ 3; Student's t test). C, qPCR analysis demonstrating dose-dependent increase in CREB1 transcript levels with rolipram treatment in apoE4 astrocytes. D, qPCR analysis to demonstrate dose-dependent increase in NHE6 transcript levels with rolipram treatment (1.25-10 M) in apoE4 astrocytes (****, p Ͻ 0.0001; n ϭ 3; Student's t test). E, fluorescence-based assay to monitor clearance of A␤ peptides by apoE4 astrocytes. F, quantitation of A␤ clearance from FACS analysis of 10,000 cells in biological triplicates confirmed restoration of A␤ clearance in apoE4 astrocytes with rolipram treatment (****, p Ͻ 0.0001; n ϭ 3; Student's t test). G and H, representative micrographs (G) and quantification (H) showing a ϳ2.66-fold increase in cell-associated A␤ in apoE4 astrocytes with rolipram treatment (****, p Ͻ 0.0001; n ϭ 160/condition; Student's t test). Error bars, S.D. See related Fig. S7.

Yeast strains and growth
All S. cerevisiae strains used were derivatives of BY4741 (a gift from Dr. Susan Michaelis (The Johns Hopkins University) or purchased from transOMIC Technologies). Yeast cells were grown in YPD (yeast extract, peptone, and dextrose) medium at 30°C with shaking at 250 rpm. Acute glucose depletion was performed by growing yeast in in YP (yeast extract and peptone) medium. Yeast growth was determined by A 600 . A spotting assay on YPD agar plates was performed to detect druginduced growth inhibition phenotypes.

Alkaline pH sensitivity and quinacrine staining
Increased luminal pH in the yeast vacuole was detected using two defining phenotypes: alkaline pH sensitivity and loss of quinacrine fluorescence. Sensitivity to alkaline medium pH was assayed in YPD medium, which was adjusted to pH 7 or pH 4.3. For quinacrine staining, cells were suspended in YPD medium containing 100 mM HEPES, pH 7.6, 200 M quinacrine and stained for 10 min at 30°C. Stained cells were washed three times with 100 mM HEPES, pH 7.6, 0.2% glucose and examined by fluorescence microscopy immediately. Quantification of images was done using ImageJ software.

Cell culture
Human ApoE isoform-expressing (apoE3 and apoE4) immortalized astrocytes (gift from Dr. David M Holtzman, Washington University, St. Louis) were maintained in DMEM/ F-12 (Invitrogen) supplemented with 10% fetal bovine serum (Invitrogen) and 200 g/ml Geneticin/G418 (Corning Cellgro). HEK293T cells were obtained from ATCC and grown in DMEM/high glucose/sodium pyruvate (Invitrogen) containing 10% fetal bovine serum. Culture conditions were in a 5% CO 2 incubator at 37°C. Cell viability was measured using the trypan blue exclusion method.

A␤ clearance assays
To quantify A␤ clearance, apoE4 astrocytes were washed with serum-free medium followed by incubation with 100 nM fluorescently labeled HiLyte Fluor 647-A␤40 (AS-64161, AnaSpec). Following several washes with PBS, cells were fixed for confocal imaging using the LSM 700 confocal microscope (Zeiss) or trypsinized for flow cytometry analysis of ϳ10,000 cells in biological triplicates using the FACSCalibur instrument (BD Biosciences). Unstained cells without any exposure to fluorescently labeled A␤ were used as a control for background fluorescence. Quantification of confocal images was done using ImageJ software.

Endosomal pH measurement
Endosomal pH was measured using flow cytometry, as we described previously (55,75). Briefly, cells were rinsed and incubated in serum-free medium for 30 min to remove residual transferrin and then incubated with pH-sensitive FITC-transferrin (T2871, Thermo Fisher Scientific) (75 g/ml) together with pH-nonsensitive Alexa Fluor 633-transferrin (T23362, Thermo Fisher Scientific) (25 g/ml) at 37°C for 1 h. Cells were placed on ice to stop transferrin uptake, and excess transferrin was removed by washing twice with ice-cold serum-free DMEM and PBS. Bound transferrin was removed by washing with ice-cold PBS, pH 5.0, and PBS, pH 7.0. Cells were trypsinized, and pH was determined by flow cytometry analysis of ϳ10,000 cells in biological triplicates using the FACSCalibur instrument (BD Biosciences). A four-point calibration curve with different pH values (4.5, 5.5, 6.5, and 7.5) was generated using the Intracellular pH Calibration Buffer Kit (P35379,

Histone deacetylases control endolysosomal pH
Thermo Fisher Scientific) in the presence of 10 M K ϩ /H ϩ ionophore nigericin and 10 M K ϩ ionophore valinomycin.

Bioinformatics
An unbiased in silico approach was used to identify experimental conditions that significantly altered Nhx1 expression, by analyzing data from 284 Saccharomyces cerevisiae microarray studies, which comprised a wide range of studies, including gene deletion, overexpression and mutations, drug/toxin treatment, nutrient limitation or excess, metabolic/environmental stress, and many others. Experimental conditions with a minimum gene expression -fold change of Ϯ2 were selected for further thorough pathway analysis. A predictive model for regulation of Nhx1 expression was created based on the overlap of pathways from experimental conditions that resulted in the highest degree of changes in gene expression (both up-and down-regulation) and selected for in vitro validation. Meta-analysis of 45 microarray experiments, including 937 samples, determined the effect of growth phase on Nhx1 expression. Normalized yeast expression data were obtained from Genevestigator (Nebion AG). Drosophila microarray experiment GSE14531 was analyzed to study the starvation effect on fly larvae. Mammalian gene expression data sets included in the study were GSE11291, GSE8488, and GSE9247. Human brain and Drosophila gene expression data were obtained from the Allen Brain and FlyBase atlas, respectively.

ChIP-qPCR
To assess physical binding of the FLAG-tagged CREB to the NHE6 gene promoter, ChIP-qPCR was carried out. ChIP was performed using the EpiQuik ChIP kit (P-2002, Epigentek), according to the manufacturer's instructions. In brief, cells were treated with 1% formaldehyde for 10 min to cross-link proteins to DNA. To generate DNA fragments ϳ200 -1,000 base pairs long, the cross-linked chromatin was sonicated on ice. ChIP assays were performed using a ChIP grade anti-FLAG antibody. Specific pulldown of CREB1 was confirmed by probing the ChIP samples with an antibody against CREB raised in rabbits using Western blotting. After reversal of the cross-links, qPCR was performed using the EpiTect ChIP-qPCR primer assay for NHE6/SLC9A6 (GPH1013950(Ϫ)12A, Ϫ12 kb from the transcription start site; GPH1013950(ϩ)01A, ϩ1 kb from the transcription start site; and GPH1013950(ϩ)02A, ϩ2 kb from the transcription start site). The ChIP-qPCR primer assay for the NHE9/SLC9A9 promoter (GPH1023313(Ϫ)01A, Ϫ1 kb from the transcription start site) was used as a control (all Qiagen). The resulting amplification and melt curves were analyzed to ensure specific PCR product. Ct values were used to calculate the -fold enhancement.

Luciferase assay
pCRE-Luc (encoding for CRE and firefly luciferase) and pSV40-RL (encoding for an SV40 promoter and Renilla luciferase) plasmids (gift of Dr. Jennifer L. Pluznick, The Johns Hopkins University) were transiently transfected into HEK293T cells using Lipofectamine 2000 reagent, as per the manufacturer's instructions. The ratio of firefly luciferase and Renilla luciferase was measured using the Dual-Luciferase assay system (E1910, Promega) with data collected using a FLUOstar Omega automated plate reader (BMG LabTech) to evaluate the CREB activation following treatment with an HDAC inhibitor.
Author contributions-H. P. designed, conducted, and analyzed experiments and wrote the paper. R. R. designed and interpreted experiments and wrote the paper.