Assorted dysfunctions of endosomal alkali cation/proton exchanger SLC9A6 variants linked to Christianson syndrome

Genetic screening has identified numerous variants of the endosomal solute carrier family 9 member A6 (SLC9A6)/(Na+,K+)/H+ exchanger 6 (NHE6) gene that cause Christianson syndrome, a debilitating X-linked developmental disorder associated with a range of neurological, somatic, and behavioral symptoms. Many of these variants cause complete loss of NHE6 expression, but how subtler missense substitutions or nonsense mutations that partially truncate its C-terminal cytoplasmic regulatory domain impair NHE6 activity and endosomal function are poorly understood. Here, we describe the molecular and cellular consequences of six unique mutations located in the N-terminal cytoplasmic segment (A9S), the membrane ion translocation domain (L188P and G383D), and the C-terminal regulatory domain (E547*, R568Q, and W570*) of human NHE6 that purportedly cause disease. Using a heterologous NHE6-deficient cell expression system, we show that the biochemical, catalytic, and cellular properties of the A9S and R568Q variants were largely indistinguishable from those of the WT transporter, which obscured their disease significance. By contrast, the L188P, G383D, E547*, and W570* mutants exhibited variable deficiencies in biosynthetic post-translational maturation, membrane sorting, pH homeostasis in recycling endosomes, and cargo trafficking, and they also triggered apoptosis. These findings broaden our understanding of the molecular dysfunctions of distinct NHE6 variants associated with Christianson syndrome.

The precise mechanisms by which altered NHE6 activity cause CS are not fully understood. The NHE6 gene is widely transcribed, with highest expression in the central nervous system (CNS), which likely explains the marked neurological morbidity of CS (36,37). Within cells, NHE6 resides in vesicles along the recycling endosomal pathway (37)(38)(39) where it operates as an electroneutral H ϩ efflux pathway to moderate intraluminal acidification driven by electrogenic vacuolar H ϩ -ATPase (V-ATPase) pumps and 2Cl Ϫ /H ϩ exchangers, the latter providing a counterion shunt to accumulate endosomal chloride that reduces the luminal positive charge and facilitates acidification. Collectively, these ion carriers attain an intraluminal pH that ranges between ϳ6.0 and 6.6 along this pathway (39 -42). This tempered acidification enables efficient dissociation and dephosphorylation of many activated ligandreceptor complexes internalized from the cell surface (43). The unbound dephosphorylated receptors, in turn, can be salvaged and returned to the plasma membrane for reuse while the freed ligands are transferred to more acidic late endosomes/multivesicular bodies (pH ϳ5.0 -5.5) and lysosomes (pH ϳ4.6 -5.0) for degradation by acid hydrolases (43)(44)(45)(46)(47). Proper acidification of endosomes is also important for their efficient movement within cells (41,48), although the mechanisms linking intraluminal acidification to the vesicular trafficking machinery are not fully understood (49,50). In the case of early/recycling endosomes, suppressing NHE6 expression in cervical cancer HeLa cells by small interfering RNA (siRNA) caused excessive endosomal acidification that correlated with impaired trafficking of a subset of clathrin-dependent recycling endosomal cargo (51). In polarized hepatoma HepG2 cells, siRNA knockdown of NHE6 selectively decreased the intraluminal pH of subapical recycling endosomes accompanied by gradual loss of apical surface bile canalicular proteins and lipids (52). Thus in CS, it has been postulated that defects in NHE6 will compromise endosomal pH homeostasis and cargo trafficking in many tissues, but especially the CNS, leading to pleiotropic pathophysiological disturbances. This is supported by studies of mouse hippocampal neurons where disruption of NHE6 expression resulted in overacidification of early and recycling endosomes and diminished signaling from the tropomyosin or tyrosine receptor kinase B (TrkB) (33) and the ␣-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (53,54) that correlated with reduced neurite outgrowth and branching, synapse density and maturation, and circuit strength.
More than 80 genetic variants of NHE6 have been identified thus far  (also see ClinVar (RRID:SCR_006169) and DECIPHER (RRID:SCR_006552)), many of which are frameshift and nonsense mutations or mRNA-splicing defects in the ion translocation domain (amino acids  that are predicted to result in total loss of NHE6 protein. However, the consequences on endosomal and cellular function of other CSlinked NHE6 mutations that may not lead to a complete loss of protein expression or function have been less well-documented (53,(55)(56)(57). These include missense substitutions, small inframe deletions, or nonsense mutations in the C-terminal regulatory domain that may generate protein products with altered functional properties (i.e. dominant-negative or constitutively-active) which, alongside other genetic factors, could contribute to reported variations in disease severity of CS patients (15).
In this report, we describe the consequences of previously uncharacterized CS-associated variants of NHE6 on various molecular and cellular parameters in transfected cells. Our results show that whereas some variants appear benign, others impair NHE6 protein biosynthetic maturation and membrane sorting to different extents, but ultimately they adversely affect recycling endosomal function and cell viability. Collectively, our findings provide greater insight into the dynamics of NHE6 dysfunction in CS and should prove valuable in the future design and development of potential therapeutic strategies for treating this neurodevelopmental disorder.

Results
To broaden our understanding of the molecular and cellular mechanisms underlying this disorder, we examined six reported inherited and de novo mutations located in the short N-terminal cytoplasmically-located segment (amino acids 1-24) (c.25G3 T:pAla 9 3 Ser, A9S) (3), the transmembrane ion translocation domain (amino acids 25-533) (c.563T3 C: Leu 188 3 Pro, L188P; c.1148G3 A:Gly 383 3 Asp, G383D) (5,15), and the longer C-terminal cytoplasmic regulatory domain (amino acids 534 -701) (c.1639G3 T: Glu 547 3 Ter, E547*; c.1703G3 A:Arg 568 3 Gln, R568Q; c.1710G3 A:Trp 570 3 Ter, W570*) (4,12,15,19,25,58) of the longest splice-variant of human NHE6 (NM_001042537.1, also called NHE6v1 but hereafter simply referred to as NHE6). The positions of these amino acids are indicated in the predicted planar transmembrane organization of NHE6 (Fig. 1A) based on the proposed topology of NHE1 (59,60). These mutations occur at highly-conserved sites in NHE6 orthologs from mammals to fruit flies, suggestive of functional importance (Fig. 1B). Transcripts containing the C-terminal premature stop codon mutations (E547* and W570*) remove the bulk of the cytoplasmic regulatory domain (i.e. the last 155 and 132 amino acids, respectively) and may be subject to nonsense-mediated mRNA decay (61). However, if translated, such truncations might remove putative signal sequences important for the membrane trafficking and/or regulation of the transporter and therefore merit further investigation as little is known about this region of the transporter.

Structural modeling of the L188P and G383D variants
Of the aforementioned six mutations, two (L188P and G383D) were localized within the transmembrane ion translocation domain, which afforded an opportunity to assess the theoretical structural and functional implications of these substitutions by structure homology modeling. In this regard, we previously described a putative structural model of the human NHE6 N-terminal transmembrane region (encompassing residues 74 -540) in its monomeric state using the spatial atomic coordinates of the outward-facing conformation of the Thermus thermophilus Na ϩ /H ϩ antiporter NapA (57,62). However, mammalian NHEs, like their bacterial homologs, normally form homodimers (63). A homodimer ribbon model of NHE6 is now displayed in its frontal and top configurations in Fig. 2A. In the wildtype (WT) exchanger, Leu 188 is adjacent to two consecutive proline residues (Pro 189 and Pro 190 ) in the middle of membrane-spanning helix 4 (M4). The latter di-proline motif would typically be expected to interrupt and break a helix (64). However, subjecting the helix to molecular dynamics (MD) simulations (for description, see "Experimental procedures") stretched out the segment containing the proline residues due to the absence of H-bonds in this location but did not break the helical formation completely (illustrated in monomer, Fig. 2B, left panel). Although very rare, it is possible to have two consecutive prolines without disrupting a helix (65,66). However, replacement of Leu 188 with proline results in three consecutive prolines, which is predicted by MD simulations to unravel the helix, resulting in a much shorter helical segment (Fig. 2B, right panel). Because the homologous M4 segment in the plasmalemmal NHE1 isoform is thought to form a critical part of the ion translocation pathway (60,67,68), it is very likely that the L188P substitution affects the functionality of the transporter.

Characterization of Christianson syndrome SLC9A6 variants
The consequences of the G383D mutation are also predicted to significantly perturb protein structure and function. In the modeled WT structure, Gly 383 is located close to the bend of an intramembranous re-entrant loop (R-loop) between helices M8 and M9 and packed tightly against amino acids Phe 373 , Ala 376 , and Glu 377 (Fig. 2C), a segment strongly implicated in ion permeation (69,70). Replacing Gly 383 with an aspartic acid residue would result in steric clashes between these residues and interfere with the packing between these segments of the R-loop (Fig. 2C). Furthermore, this substitution would also disrupt interactions of R-loop residues Trp 379 , Phe 381 , and especially Thr 382 with residues Asp 292 and Ile 296 in helix M7 and residues Ile 330 , Phe 331 , and Ser 334 in helix M8 (Fig. 2D). Notably, amino acid Asp 292 is invariant in all mammalian NHEs, and its homologous residue in human NHE1 (i.e. Asp 267 ) is critical for catalytic activity (71). Thus, it is probable that the G383D mutation would also disrupt the conformation and function of NHE6.

Biosynthetic maturation and stability of the NHE6 variants
To empirically visualize and evaluate the functional significance of the individual variants, we engineered them into the WT human NHE6 cDNA that was fused at its C terminus to either enhanced green fluorescent protein (GFP), monomeric Cherry fluorescent protein (ChFP), or influenza hemagglutinin (HA)-epitope (NHE6 GFP , NHE6 ChFP , or NHE6 HA , respectively). We previously showed that the molecular and cellular properties of these chimeric constructs were indistinguishable from unmodified NHE6 when examined in AP-1 cells, a Chinese hamster ovary-derived cell line lacking detectable levels of endogenous NHE6, and therefore serve as a useful cell model system to study this transporter (53).
To assess their biosynthetic maturation, mammalian expression plasmids containing the NHE6 GFP constructs were transiently transfected in AP-1 cells, and their total cellular expression was monitored over a 48-h period by Western blotting. Consistent with previous results (56), the homodimeric WT transporter fractionated as multiple bands that correspond to its dissociated, newly-synthesized core-glycosylated (ϳ85-90 kDa), and fully-glycosylated (ϳ110 -130 kDa) monomeric species, and its fully-glycosylated dimeric form (Ͼ250 kDa) that does not fully dissociate under standard SDS-PAGE conditions ( Fig. 3A) (53). Dimeric assembly was observed for all constructs, suggesting that formation of their quaternary structures was not grossly perturbed. However, whereas post-translational oli-

Characterization of Christianson syndrome SLC9A6 variants
gosaccharide maturation proceeded normally for A9S and R568Q, the ratio of fully-glycosylated (monomer and dimer) to total NHE6 protein expression was noticeably impaired for the other four mutants compared with WT ( Fig. 3A), showing significant reductions (ϳ70 -90%; two-way ANOVA, p Ͻ 0.05) at peak expression (between 24 and 32 h post-transfection) as quantified by densitometry (see "Experimental proce-dures" for details) (Fig. 3B). Total protein expression of these latter mutants was also noticeably diminished at 48 h posttransfection compared with WT, indicative of increased protein clearance.
The stability of the NHE6 constructs was further examined by pulse-chase experiments. After 24 h of transient transfection, cells were treated with cycloheximide for an additional and top (extracellular aspect, right panel) views of a 3D structure homology model of dimeric human NHE6 based on the crystal structure of the bacterial T. thermophilus Na ϩ /H ϩ antiporter NapA (TtNapA) (Protein Data Bank accession code 5bz3; 2.30 Å, 15% identity, 27% similarity), which provided the broadest coverage, highest resolution, and best spatial fit compared with other crystallized bacterial Na ϩ /H ϩ antiporters. The proposed structure includes only the membrane-spanning helices (M2-M12; amino acids 74 -540) that aligned with homologous segments of TtNapA. The top view includes the locations of the transmembrane-localized residues Leu-188 and Gly-383 mutated in CS. B, molecular dynamics simulation of structural changes predicted to occur in TM4 (highlighted in cyan) upon substitution of Leu-188 with Pro (L188P). The monomeric forms of NHE6 WT and L188P are illustrated. C and D, structural perturbations predicted to occur in the intramembranous re-entrant (R) loop between helices M8 and M9 upon substitution of Gly-383 with Asp (G383D). C, upper and lower panels show front and top views, respectively, of Gly-383, which is packed tightly against amino acids Phe-373, Ala-376, and Glu-377 (top left). Mutation of Gly-383 to Asp would result in steric clashes between these residues and interfere with the packing between these segments of the R-loop. D, G383D substitution would also disrupt interactions of R-loop residues Trp-379, Phe-381, and especially Thr-382 with residues Asp-92 and Ile-296 in helix M7 and residues Ile-330, Phe-331, and Ser-334 in helix M8.
4 -24 h to inhibit de novo protein synthesis while monitoring levels of the previously synthesized transporters as a function of time. As shown in Fig. 4, levels of the core-glycosylated forms of WT, A9S, and R568Q at 24 h post-transfection matured to their fully-glycosylated states within 4 h after cycloheximide treatment and then remained relatively stable over the ensuing 20-h period. By contrast, the L188P, G383D, E547*, and W570* mutants showed reduced oligosaccharide maturation and abundance over the course of the pulse-chase period. As a control for protein loading, we measured the expression of the glycolytic enzyme glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and found it to be constant over the treatment period. These results are consistent with reduced stability of these variants.  (20 g) were subjected to SDS-PAGE and immunoblotting with a polyclonal anti-GFP antibody. NHE6 migrates as multiple bands: higher molecular weight bands represent the fully-glycosylated (fg) and core-glycosylated (cg) dimeric (d) forms of the exchanger that do not fully dissociate under SDS-PAGE conditions, whereas lower molecular weight bands represent fully-glycosylated and core-glycosylated forms of the dissociated monomeric (m) protein. The blots were stripped and reprobed with a mouse monoclonal anti-GAPDH antibody to control for protein loading. B, ratios of fully-glycosylated protein (monomer and dimer)/total NHE6 protein (fg/total) were quantified by densitometry of X-ray films exposed in the linear range and analyzed using ImageJ software. Values represent the mean Ϯ S.D. of three different experiments. Significance was determined by two-way ANOVA with a post hoc Tukey test. The NHE6 variants clustered into two groups: 1) WT, A9S, and R568Q and 2) L188P, G383D, E547*, and W570*, with variants within each cluster yielding similar statistical values. Population means of the NHE6 variants are significantly different (F value ϭ 26.4, p value ϭ 1.5 ϫ 10 Ϫ16 ). Population means as a function of time are significantly different (F value ϭ 48.6, p value ϭ 1.9 ϫ 10 Ϫ19 ). § indicates significance (p Ͻ 0.01) of the means of NHE6 variants within a cluster relative to the 12-h time point. Asterisks indicate significance (ૺ, p Ͻ 0.05, and ૺૺ, p Ͻ 0.01) of the means between clusters of NHE6 variants at the indicated time points.

Characterization of Christianson syndrome SLC9A6 variants
The delayed processing and enhanced turnover of the L188P, G383D, E547*, and W570* mutants could arise by redirecting the defective transporters to proteasomes and/or lysosomes. To assess this possibility, another pulse-chase experiment was performed with transiently-transfected cells (24 h) expressing the various NHE6 constructs incubated with cycloheximide and inhibitors of proteasomal (MG132) (72) or lysosomal (leupeptin plus pepstatin, LeuP) (73, 74) proteolysis over an 8-h period. As expected, both types of inhibitors had no appreciable effect on WT, A9S, and R568Q levels compared with vehicletreated controls (Fig. 5). By comparison, suppression of proteasomal, but not lysosomal, proteolysis partially reduced the cellular loss of L188P, G383D, E547*, and W570* and resulted in the appearance of a unique high-molecular-weight NHE6 band above the fully-glycosylated dimeric band (indicated by an asterisks) for each of these variants. Although the nature of this band was not examined empirically, it likely represents polyubiquitinated or possibly aggregated forms of these variants that accumulated upon proteasomal inhibition. The absence of an effect of the lysosomal protease inhibitors leupeptin and pepstatin, which are only moderately cell-permeant, prompted us to test the effect of the lysosomotropic agent chloroquine, a weak base that readily permeates and neutralizes acidic organelles such as lysosomes and, as such, nonselectively represses the actions of resident acid hydrolases (73,75). Under these conditions, chloroquine mildly impeded their clearance at 4 h, but this was not sustained at the 8-h time point relative to diluent-treated controls (Fig. 6). Collectively, these data implicate a preferential role for the endoplasmic reticulum-associated degradation (ERAD) pathway (76,77), and to a lesser extent lysosomes, in clearance of these variants.
To biochemically assess the ability of the NHE6 variants to traffic to the plasma membrane, plasmalemmal proteins were extracted using a cell-surface biotinylation procedure and ana- lyzed by Western blotting. We previously showed that ϳ5% of the total cellular pool of NHE6 is located at the cell surface of transfected AP-1 cells (57). Because of the low levels of expression of some of the NHE6 variants, we loaded varying amounts of protein in each lane (as indicated) so that their signal intensities would be comparable and easily visualized within a single X-ray film exposure of the immunoblots. As presented in Fig.  9A, all mutants were detected at the plasma membrane, although the fractional surface abundances of L188P, G383D, E547*, and W570* were noticeably reduced compared with WT, A9S, and R568Q after taking into account differences in gel protein loading. Notably, despite their lower levels, the L188P, G383D, E547*, and W570* mutants at the cell surface were largely fully-glycosylated (monomer and dimer). This indicates that these variants have the potential to fold properly and undergo oligosaccharide maturation along the exocytic pathway, although at greatly reduced efficiency. The absence of the core-glycosylated forms (monomer and dimer) in the biotinylated fractions is consistent with their intracellular location (i.e. mainly ER). To compare their plasmalemmal abundances more quantitatively, we used a cell-based enzyme-linked immunosorbent assay (ELISA). For these experiments, AP-1 cells were transiently transfected with NHE6 constructs containing a triple-Flag epitope-tag inserted in the first exofacial loop as described previously (53). Consistent with cell-surface biotinylation, the WT and A9S variants showed equivalent levels of surface expression. The plasmalemmal abundance of R568Q was detectably higher, consistent with image analyses, but this increase did not reach statistical significance (Fig. 9B). By contrast, surface abundances of the other mutants were markedly reduced by ϳ55-70%, especially for G383D, E547*, and W570* (*, p Ͻ 0.05, one-way repeated measures ANOVA followed by a post hoc Dunnett's test).
Based on the sensitivity of the ELISA, we proceeded to measure endocytosis of the different NHE6 variants. For these experiments, cells transiently expressing the various constructs were placed on ice and incubated with a primary mouse monoclonal anti-Flag antibody and secondary goat anti-mouse antibody to label cell surface 3F NHE6 HA , and then endocytosis of the labeled pool was initiated by incubating the cells at 37°C over a 15-min period. As shown in Fig. 9C, the levels of internalization of surface-labeled A9S and R568Q were equivalent to WT, whereas L188P, G383D, E547*, and W570* were translocated more slowly (p Ͻ 0.05), indicative of their differential handling by the endocytic trafficking machinery.

Effect of NHE6 variants on recycling endosomal cargo trafficking and pH
Previous studies have shown that overexpression and knockdown of NHE6 promoted up-and down-regulation, respectively, of clathrin-mediated endocytosis of certain integral plasma membrane proteins, such as the ligand-activated transferrin receptor (Tf-TfR) but not the epidermal growth factor

Characterization of Christianson syndrome SLC9A6 variants
receptor, in a manner that depended on its ability to modulate intravesicular pH (51,53).
To determine the effect of the NHE6 variants on internalization of Tf, we measured the initial rate of uptake (5-min period) of Alexa Fluor TM 633-conjugated Tf (Tf-AF 633 ) by flow cytometry in live HeLa cells transiently expressing with GFP alone (control) or GFP-tagged constructs of NHE6 as described previously (53). For these experiments, HeLa cells were used because their TfR levels are severalfold higher than in AP-1 cells and therefore provided a stronger signal-to-noise ratio for detecting the initial uptake of Tf-AF 633 (53). The patterns of protein biosynthesis and oligosaccharide maturation for the various NHE6 constructs in transfected HeLa cells were equivalent to those observed in AP-1 cells (data not shown). Equivalent numbers of live GFP-positive HeLa cells (10 4 GFP ϩ cells) were analyzed for each construct; live cells were distinguished from nonviable cells by their ability to exclude the membraneimpermeant fluorescent dye 7-amino-actinomycin D. HeLa cells overexpressing WT, A9S, and R568Q exhibited significant increases (ϳ25-35%) in Tf-AF 633 uptake compared with control GFP-only expressing cells (p Ͻ 0.001, one-way repeated measures ANOVA followed by a post hoc Dunnett's test) (Fig.  9D). However, this up-regulation was greatly diminished or absent in cells expressing the L188P, G383D, E547*, and W570* mutants (p Ն 0.05).
The regulation of endosomal trafficking by NHE6 is dependent on its ability to regulate endosomal pH e , although the underlying mechanism linking transporter function to vesicle trafficking is unknown. NHE6, like other NHE family members, is a secondary active transporter that moves monovalent cations down their respective concentration gradients, and thus at steady-state it should operate as a "H ϩ leak" pathway (i.e. alkalinizing mechanism) to counter acidification driven by the vacuolar H ϩ -ATPase. To examine whether this property is affected, we measured the pH e of recycling endosomes in situ by fluorescence ratiometric image analysis (FRIA) (83) using the pH-sensitive fluorophore fluorescein isothiocyanate (FITC) conjugated to transferrin (Tf-FITC) as a probe. Briefly, AP-1

Characterization of Christianson syndrome SLC9A6 variants
cells were transiently transfected (48 h) with ChFP alone or the different NHE6 ChFP constructs, followed by a 60-min incubation with Tf-FITC at 37°C to allow for sufficient accumulation of the probe throughout the recycling endosomal pathway. The steady-state pH e spectrum of vesicles or puncta under each condition was determined by analyzing fluorescence signals pooled from a total of 25-36 cells (ϳ2700 -5300 vesicles collected from 8 to 12 cells per transfectant per experiment, repeated three times). In each case, the pH e spectrum of the vesicles exhibited a unimodal distribution (see Fig. S1). When the data were expressed as the mean recycling endosomal pH e per cell, the level for AP-1 cells lacking NHE6 was very acidic at pH e 5.79 Ϯ 0.11 (mean Ϯ S.D., n ϭ 27 cells) (Fig. 9E). By contrast, endosomal pH e values for WT-, A9S-, and R568Q-expressing cells were significantly more alkaline and equivalent to each other relative to parental AP-1 cells (WT: 6.23 Ϯ 0. 16 Because the L188P, G383D, E547*, and W570* variants showed significant accumulation in the ER, we next evaluated whether they influenced the luminal pH e of this compartment. To this end, AP-1 cells were co-transfected with the various mCherry-tagged NHE6 (NHE6 ChFP ) constructs and a ratiometric pH-sensitive GFP (i.e. pHluorin2 or pH2) (84) containing an ER signal sequence from calreticulin at its N terminus and an

Effect of NHE6 variants on cell viability
Previous studies have revealed that genetic disruption of NHE6 expression causes degeneration of a subset of neurons within the cortex, hippocampus, and especially the cerebellum, which exhibits extensive Purkinje cell loss with age (32,34,35).
Neurons in other brain regions, despite the loss of NHE6 expression, do not appear as severely affected. The reason for these differences in cell viability is unclear. AP-1 cells, which also lack NHE6, also show little evidence of cell death under normal culture conditions (53). However, ectopic expression of a CS-linked in-frame deletion variant of NHE6 (p.E287-S288del, ⌬ 287 ES 288 ) in AP-1 cells and in primary mouse hippocampal neurons could induce programmed cell death (53). Thus, AP-1 cells serve as a useful proxy for assessing the potential in vitro consequences of NHE6 variants on cell survival. To this end, AP-1 cells were transiently transfected with GFP alone or GFP-tagged NHE6 variants for 48 h. These cells then were incubated in the presence of the fluorescent annexin Vallophycocyanin conjugate (annexin V-APC) and propidium iodide (PI), and GFP-positive cells (10 4 cells per experiment) were analyzed by flow cytometry to determine the fraction of apoptotic cells as described previously (53). Annexin V-APCpositive and PI-negative cells represent the early stages of apoptosis, whereas annexin V-APC/PI-double-positive cells are late apoptotic or necrotic, and PI-only-positive cells are con-
At this time point, there were no significant increases in late apoptotic or necrotic cells among the CS-linked variants relative to WT. Curiously, expression of WT, A9S, and R568Q also significantly increased the percentage of apoptotic cells relative

Characterization of Christianson syndrome SLC9A6 variants
to AP-1 cells (p Ͻ 0.001), but not as greatly as the other four defective variants. The reason for this is unclear and will require further study. Nevertheless, these findings suggest that L188P, G383D, E547*, and W570* have the potential to reduce cell viability relative to WT, although the underlying mechanisms remain obscure.

Discussion
CS patients display a wide range of neurological and somatic symptoms (1,15), with their acuteness ostensibly governed by the nature of the mutation as well as other intrinsic compensatory or aggravating influences. The findings reported herein reveal differences in the biochemical properties and cellular handling of several SLC9A6/NHE6 variants associated with CS. Of the six variants examined, four (L188P, G383D, E547*, and W570*) showed varying degrees of impairment in post-translational oligosaccharide maturation, protein stability, and membrane targeting. L188P and G383D partially accumulated in the ER, but measurable fractions of L188P and to a lesser extent G383D were fully-glycosylated and sorted to a pericentriolar endosomal compartment. By contrast, both E547* and W570* were largely retained in the ER, although minor fractions did exit the ER and became fully-glycosylated. However, in each case their trafficking to the cell surface and internalization into recycling endosomes was compromised, albeit to different extents (WT Ͼ L188P Ͼ G383D Ն E547* Ϸ W570*). Each variant exhibited correspondingly diminished abilities to stimulate the internalization of endosomal cargo (i.e. ligand-activated transferrin receptor), to regulate acidification of recycling endosomes, and to maintain cell viability. Remarkably, two of the purported CS variants, A9S (3,13,25) and R568Q (4,25), exhibited properties that were essentially indistinguishable from the WT transporter that obscures their disease significance, at least when examined in AP-1 cells. Although this experimental cell system has proven informative in deciphering NHE6 function, we cannot exclude the possibility that any deleterious consequences of the A9S and R568Q variants may be more subtle and manifested only in certain neuronal cell types. However, during the preparation of this manuscript, Ouyang and co-workers (88) reported that mice genetically manipulated by CRISPR/Cas9 technology to harbor the homologous human NHE6 A9S variant (i.e. mouse A11S) appeared normal in terms of brain morphology, NHE6 subcellular localization, and intra-endosomal pH of hippocampal neurons. Although a behavioral assessment of these mice was not described, we envision that human A9S and possibly the R568Q variants are likely benign and possibly misattributed as the cause of the

Characterization of Christianson syndrome SLC9A6 variants
patients' neurological symptoms. This suggestion is further supported by the common occurrence of both variants in the general human population (genome aggregation database, gno-mAD; RRID:SCR_014964).

Preferential degradation of defective NHE6 variants by ERAD
A biochemical feature of mature WT NHE6 is the covalent attachment of N-linked oligosaccharides (53). In unpublished studies, 3 we have identified two neighboring asparagine residues located in canonical N-glycosylation motifs (Asn-Xaa-Ser/Thr) in the second extracellular loop, Asn 128 -Val-Thr and Asn 145 -Val-Ser, that serve as the sole targets for N-glycan addition. Post-translational addition of N-glycans (as well as O-glycans) in the ER and Golgi is known to facilitate intracellular processes such as the folding, multisubunit assembly, membrane trafficking, and function of many glycoproteins (89 -91). As described herein, the L188P, G383D, E547*, and W570* variants showed marked reductions in their N-oligosaccharide maturation and membrane targeting, especially the latter three, and correlated with increased retention in the ER and cellular clearance. These hypo-glycosylated variants assembled correctly as homodimers, indicating that their quaternary structures were not grossly perturbed. Instead, their glycosylation impairments are more consistent with localized misfolding, as was predicted for the L188P and G383D variants by structural modeling analyses. Aberrant conformations of N-glycoproteins, which can include exposed hydrophobic segments, improper disulfide bond formation, and immature glycans, are recognized by the ER protein quality control machinery that encompasses multiple systems capable of detecting discrete structural malformations situated in the ER lumen, within the membrane (i.e. L188P and G383D), or facing the cytoplasm (i.e. E547* and W570*) (92)(93)(94). Misshapen proteins are then subjected to repeated cycles of deglucosylationreglucosylation of the initial core N-glycan structure (Glc3-Man9 -GlcNAc2) and binding to the lectin chaperones calnexin and calreticulin, which assist in protein refolding until the proper conformation is attained (95,96). However, deformed proteins refractory to the refolding process after a prolonged ER residency are targeted for ERAD by proteasomes (93). This enhanced proteasomal proteolysis was evident for the L188P, G383D, E547*, and W570* variants. In the case of the E547* and W570* truncation variants, aside from possible misfolding, their pronounced accumulation in the ER might also result from the loss of one or more specific ER-export signal motifs (97, 98). Absence of such signals could delay their exit from the ER and favor retrotranslocation to the proteasome (99). The fate of the minor fraction of these variants that did escape ERAD, possibly in part due to overexpression, could be degradation in lysosomes. However, blocking lysosomal proteolysis with leupeptin/pepstatin or chloroquine did not substantially prevent their clearance, suggesting that other pathways such as cellular expulsion in exosomes might be an alternate route (100). Such a mechanism has been proposed for another CS-linked variant p.Gly 218 3 Arg (57) and remains to be tested for these variants.

Defective NHE6 variants disrupt recycling endosomal pH homeostasis and trafficking
The severity of sorting defects for L188P, G383D, E547*, and W570* correlated with corresponding deficits in recycling endosomal function, as revealed by excessive luminal acidification and reduced uptake of recycling endosomal cargo (e.g. ligand-activated transferrin receptor) relative to WT transfectants. Optimal acidification of organelles is required not only for efficient biochemical processing of their contents, but also for effective sorting and trafficking of proteins and lipids along the biosynthetic and endocytic pathways (48,(101)(102)(103)). Yet how intravesicular pH influences vesicle movement is not fully understood. Recent studies have indicated that the vacuolar H ϩ -ATPase pump functions not only in organellar acidification but also as a luminal pH-sensor and binding platform for members of the ADP-ribosylation factor family of small GTPases involved in vesicular trafficking (49,50,104). Curiously, despite the broad organellar distribution of the vacuolar H ϩ -ATPase, inhibiting its activity with pharmacological agents disrupted only vesicular movement between the early and late endosome/lysosomal compartments, while trafficking along the recycling endosomal pathway was relatively unaffected (49). This raises the possibility that other endosomal pH-regulatory transporters may compensate and play more prominent multifunctional roles in vesicular trafficking along other endomembrane pathways. In keeping with this notion, the yeast Saccharomyces cerevisiae endomembrane (Na ϩ ,K ϩ )/H ϩ exchanger Nhx1/Vps44 was found to form a complex with the GTPaseactivating protein Gyp6 and one of its effectors Ypt6, the homolog of the mammalian Rab6 GTPase, and served to coordinate endosomal shuttling between the Golgi and prevacuolar (late endosomes) compartments (105). Loss of Nhx1 caused hyperacidification and disruption of endosomal protein trafficking and fusion to the vacuole (106 -109). In an analogous manner, mammalian NHE6 might fulfill a similar role in pHsensing and modulation of vesicular trafficking along the recycling endosomal pathway, a process that is diminished by loss of NHE6 function (33,(51)(52)(53). Indeed, we found that excessive acidification and impaired trafficking of recycling endosomes were most pronounced for E547*-and W570*-expressing cells at levels comparable with those of NHE6-deficient AP-1 cells. This would be expected as these variants are not properly sorted to recycling endosomes. By contrast, values for luminal pH e and trafficking of L188P-containing endosomes were intermediate between WT transfectants and AP-1 cells, indicating that this variant retained some activity to counter H ϩ influx driven by the vacuolar H ϩ -ATPase. Leu 188 is a highlyconserved amino acid in transmembrane helix M4, a segment previously identified as a critical part of the ion permeation pore of the plasmalemmal NHE1 isoform (60,67,68). Thus, insertion of an additional proline residue in this region is likely to cause steric or rotamer constraints on the geometry of M4 that compromises catalytic activity. G383D also exhibited some nominal ability to elevate recycling endosomal pH e above that measured in AP-1 cells, but this level was insufficient to significantly stimulate uptake of transferrin. The crippling effect of the G383D mutation was predicted from the structural homo-3 A. Ilie and J. Orlowski, unpublished data.

Characterization of Christianson syndrome SLC9A6 variants
logy modeling analyses and consistent with earlier findings (69) showing that nonpolar and polar amino acid substitutions of the homologous Gly residue in NHE1 (i.e. human Gly 352 and rat Gly 356 ) greatly reduced its catalytic activity (ϳ10% of WT) as well as the binding of amiloride-and benzoylguanidine-based antagonists. Thus, all four variants elicit deleterious effects on recycling endosomal function.

Expression of defective NHE6 variants reduces cell viability
Progressive cerebellar atrophy, loss of Purkinje neurons, and motor regression are common features of a majority (ϳ60%) of CS patients examined thus far (1,2,6,7,10,14). Less severe neuronal loss and gliosis have been documented in other brain regions, such as the cerebral cortex and hippocampus (6). Similar degeneration of the cerebellum and mild undergrowth and atrophy of the hippocampus, striatum, and cortex were also observed in Slc9a6 knockout mice (32,34,35). The variable penetrance and seemingly selective deterioration of certain neuronal populations upon loss of NHE6 function, despite its broad expression throughout the CNS, is not well-understood but may reflect the nature of the mutation as well as other ill-defined genetic modifier effects or compensatory mechanisms. The latter could include expression of its close paralog NHE9/SLC9A9, which is also important for neurological function, as defects in NHE9 expression have been linked to attention-deficit hyperactivity and autism spectrum disorders in humans (26,(110)(111)(112)(113)(114) and autism-like behavior in mice (115,116). Using the AP-1 cell line as a surrogate model for assessing cell viability, we found that AP-1 cells lacking NHE6 showed only nominal cell death (Ͻ10%). Surprisingly, expressing WT, A9S, and R568Q raised this percentage (ϳ23%), but it was further elevated in the presence of the L188P, G383D, E547*, and W570* variants (33-40%). The mechanisms by which NHE6 modulates cell viability have yet to be elucidated. However, insights can be gleaned from recent studies showing that NHE6 and the neurotrophin TrkB receptor co-localize in recycling endosomes of cultured mouse hippocampal neurons (33). The Trk family of neurotrophin receptors are known positive regulators of nervous system development, survival, and plasticity (117)(118)(119). Notably, loss of NHE6 results in overacidification of recycling endosomes and diminished TrkB receptor signaling, neurite branching, synapse number, and circuit connectivity (33). Remarkably, these morphological and functional deficits could be rescued by activation of residual TrkB receptors with high levels of exogenous brain-derived neurotrophic factor (33). Thus, we posit that NHE6 impairment of neurotrophin Trk receptor-mediated prosurvival signals may shift the equilibrium toward neurotrophin p75 NTR receptor-mediated proapoptotic signals (120 -122) leading to degeneration and death of certain neuronal populations. Likewise, imbalances of endosomal trafficking and signaling of prosurvival growth factor receptor and proapoptotic death-or dependence-receptor signaling pathways in non-neuronal cells (123)(124)(125)(126) may also explain the observed changes in survival of AP-1 cells expressing the various NHE6 constructs.
Aside from imbalances in prosurvival versus proapoptotic signaling pathways, ER stress (127) may also be another factor that decreases the function and viability of certain cells possess-ing the CS-linked variants. Accumulation of malformed proteins in the ER initiates the unfolded protein response and ERAD in order to restore cellular homeostasis (128,129). However, prolonged ER stress can activate several protein kinases, including glycogen synthase kinase-3␤ (GSK-3␤) (130) and inositol-requiring kinase 1 (IRE1)/apoptosis signal-regulating kinase 1 (ASK1)/c-Jun N-terminal kinase signaling cascade (127,131), ultimately stimulating caspase-mediated cell death. Post-mitotic neurons are particularly vulnerable to ER stress, as they are unable to dissipate any accumulated misfolded proteins through cell division. This inability to undergo mitosis may be an additional contributing factor in the progression of CS as well as many other neurodegenerative diseases such as Alzheimer's disease, Parkinson's disease, and Huntington's disease (132).
In summary, our results provide additional insight into the molecular mechanisms by which different NHE6 variants linked to CS impair recycling endosomal pH homeostasis and cargo trafficking, and ultimately cell viability. These analyses will guide future investigations to delineate the molecular network of proteins that interact with NHE6 and influence receptor trafficking, synaptic plasticity, and learning deficits. Such knowledge has the potential to identify novel diagnostic and treatment options for CS that may also be applicable to other disorders such as autism and attention deficit hyperactivity disorders.

Antibodies and reagents
Mouse monoclonal anti-hemagglutinin (HA) antibody was purchased from Covance Inc. (Berkeley, CA); rabbit polyclonal anti-HA and mouse monoclonal anti-GAPDH were obtained from Abcam Inc. (Cambridge, MA); mouse monoclonal anti-Flag M2 and anti-␤-tubulin antibodies were from Sigma; rabbit polyclonal anti-GFP antibody was from Life Technologies, Inc.; and rabbit polyclonal anti-calnexin was from Enzo Life Sciences, Inc. Horseradish peroxidase-conjugated secondary IgG antibodies were purchased from Jackson ImmunoResearch (West Grove, PA). All Alexa Fluor-conjugated secondary antibodies, Alexa Fluor 488-and 633-conjugated transferrin (Tf-AF 488 and Tf-AF 633 , respectively), and Tf-FITC were purchased from Molecular Probes (Eugene, OR). ␣-Minimum essential medium (␣-MEM), fetal bovine serum, penicillin/ streptomycin, and trypsin-EDTA were purchased from Wisent (Saint-Bruno, Quebec, Canada). All other chemical and reagents were obtained from BioShop Canada (Burlington, Ontario, Canada), Sigma, or Thermo Fisher Scientific and were of the highest grade available.

Recombinant DNA constructs and mutagenesis
The long transcript splice-variant of human NHE6 (NHE6v1; NCBI NM_001042537) was engineered to contain the influenza virus HA (YPYDVPDYAS) epitope at its extreme C terminus (NHE6 HA ), as described previously (56). NHE6 HA was then used as a template to introduce a triple-Flag epitope (AAADY-KDDDDKGDYKDDDDKGDYKDDDDKAAA) in the first extracellular loop immediately after residue Met 53 ( 3F NHE6 HA ) (56). Enhanced GFP and monomeric ChFP were also fused to

Characterization of Christianson syndrome SLC9A6 variants
the C terminus of WT NHE6 (NHE6 GFP and NHE6 ChFP ). The various tagged NHE6 constructs were then used as templates to engineer the various CS-linked mutations by PCR mutagenesis. Insertion of the different epitope tags in the various positions did not alter the biochemical properties or cellular distribution of exogenous NHE6 compared with the endogenous protein (56). All constructs were sequenced to ensure that no additional mutations were introduced during PCR.

Western blotting
For Western blotting analyses, AP-1 and HeLa cells were grown in 10-cm dishes and transiently transfected with 5 g of plasmid DNA encoding NHE6 GFP or NHE6 HA WT or mutant constructs using Lipofectamine2000 TM (Invitrogen) according to the manufacturer's recommended procedure. Cell lysates were prepared following 6 -48 h post-transfection (as indicated in the figure legends) by washing cells twice on ice with ice-cold PBS, followed by scraping in 0.5 ml of lysis buffer (1% NP-40, 0.25% sodium deoxycholate, PBS supplemented with protease inhibitor mixture (Roche Diagnostics)). Lysates were incubated for 30 min on a rocker at 4°C and then centrifuged at 16,000 ϫ g for 20 min at 4°C to pellet the nuclei and cellular debris. Ten to 50 g of protein (as indicated in the figure legends) from the resulting supernatants were diluted in SDS-sample buffer (50 mM Tris-HCl, pH 6.8, 1% SDS, 50 mM DTT, 10% glycerol, 1% bromphenol blue), subjected to 9% SDS-PAGE, and then transferred to polyvinylidene fluoride membranes (Millipore, Nepean, Ontario, Canada) for immunoblotting. The membranes were blocked with 5% nonfat skim milk for 1 h and then incubated with the specified primary antibodies (rabbit polyclonal GFP 1:8000, mouse monoclonal HA 1:5000, ␤-tubulin 1:10,000, or GAPDH 1:50,000) in PBS containing 0.1% Tween 20, followed by extensive washes and incubation with goat antimouse (1:5000) or goat anti-rabbit (1:8000) horseradish peroxidase (HRP)-conjugated secondary antibodies for 1 h. Immunoreactive bands were detected using Western Lightning TM Plus-ECL blotting detection reagents (PerkinElmer Life Sciences). Quantification of immunoreactive signals was determined by densitometry of X-ray films exposed in the linear range and analyzed using ImageJ software.

Cell-surface biotinylation
AP-1 cells expressing NHE6 GFP WT or mutant constructs were cultured in 10-cm dishes to subconfluence, placed on ice, and washed three times with ice-cold PBS containing 1 mM MgCl 2 and 0.1 mM CaCl 2 , pH 8.0 (PBS-CM). Next, cells were incubated at 4°C for 30 min with the membrane-impermeable reagent N-hydroxysulfosuccinimidyl-SS-biotin (0.5 mg/ml) (Thermo Fisher Scientific, Rockford, IL). Cells were washed and incubated twice in quenching buffer (50 mM glycine in PBS-CM) for 7 min each on ice to remove unreacted biotin. After two more washes in PBS-CM, the cells were lysed for 30 min on ice and then centrifuged at 16,000 ϫ g for 20 min at 4°C to remove insoluble cellular debris. A fraction of the resulting supernatant was removed, and this represents the total fraction. The remaining supernatant was incubated with 100 l of 50% NeutrAvidin-agarose resin slurry (Thermo Fisher Scientific, Whitby, Ontario, Canada) in lysis buffer overnight at 4°C to extract biotinylated membrane proteins. The proteins were then resolved by SDS-PAGE and analyzed by Western blotting.

Immunofluorescence confocal microscopy
AP-1 cells were cultured on fibronectin-coated 18-mm glass coverslips, transfected with the NHE6 constructs, and fixed post-transfection (at indicated times) with 4% paraformaldehyde for 20 min at room temperature. To examine the subcellular distribution of the NHE6 constructs, AP-1 cells were cotransfected with the various NHE6 ChFP constructs and different organellar markers. To label transferrin-containing recycling endosomes, AP-1 cells were transfected with NHE6 ChFP constructs for 48 h and then incubated with Alexa Fluor 488conjugated transferrin (Tf-AF 488 ) for 45 min in serum-free ␣-MEM, washed, fixed, and mounted. To visualize the ER, AP-1 cells expressing NHE6 ChFP constructs were fixed 48 h after transfection and labeled with rabbit polyclonal anti-CANX followed by goat anti-rabbit Alexa Fluor 488-conjugated secondary antibodies, washed, and mounted. Cells were examined by laser-scanning confocal microscopy using the ZEN software of a Zeiss LSM 780 microscope equipped with a photomultiplier tube (PMT) detector, with images acquired using a ϫ63/1.4 NA oil immersion objective lens.

Fluorescence-based endocytosis assay
Cell-surface expression and internalization of the 3F NHE6 HA constructs was determined using an ELISA, as described previously (53). Briefly, AP-1 cells were grown in 10-cm dishes and transiently transfected with 6 g of empty plasmid vector (pcDNA3) or vectors containing 3F NHE6 HA WT or mutant DNA constructs. Twenty four h post-transfection, cells were transferred to 12-well plates and grown for an additional 12-24 h. Cells were chilled on ice, washed with ice-cold PBS-CM, pH 7.4, blocked in 10% goat serum/PBS-CM, and then incubated with a mouse monoclonal anti-Flag antibody (1:3000) (Sigma) on ice. Internalization of the bound antibody was initiated by incubating the cells with warm (37°C) ␣-MEM for the indicated time points and terminated by placing the plates on ice. Cell were washed and labeled with goat anti-mouse HRP-conjugated secondary antibody (1:1000) (GE Healthcare). After extensive washes with PBS-CM, cells were treated on ice with Amplex Red reagent (Invitrogen). Aliquots were transferred to 96-well plates, and fluorescence readings were taken with a POLARstar OPTIMA (BMG Labtech. Inc, Offenburg, Germany) plate reader using 544-nm excitation and 585-nm emission wavelengths. All experiments were performed in triplicate and repeated at least three times. Results were expressed as a percentage of the fluorescence recorded prior to internalization, after subtraction of the value measured with empty vectortransfected cells.

Measurements of organellar pH
Recycling endosomes-AP-1 cells were grown in 6-well plates and transfected with 1 g of ChFP alone (control) or the NHE6 ChFP constructs using Lipofectamine2000 according to the manufacturer's instructions. Twenty four hours after transfection, cells were transferred to fibronectin-coated (2 g/ml in PBS, pH 7.4, overnight at 4°C) FluoroDishes TM (World Precision Instruments, Inc.) and grown overnight. Before loading with transferrin, cells were serum-starved in serum-free ␣-MEM for 1 h at 37°C and 5% CO 2 . Next, the cells were incubated with FITC-conjugated human transferrin (Tf-FITC) (Molecular Probes) at 6.6 g/ml in serum-free ␣-MEM for 1 h at 37°C and 5% CO 2 . Cells were subsequently washed twice with warm ␣-MEM supplemented with 10% fetal bovine serum, keeping the second wash for 1 min to allow for internalization of remaining plasma-membrane bound Tf-FITC. Next, cells were washed three times in ice-cold PBS-CM (PBS containing 1 mM MgCl 2 and 0.1 mM CaCl 2 , pH 7.4) and imaged in the same solution at room temperature. Endosomal pH was measured by single-cell FRIA using a Carl Zeiss MicroImaging AxioObserver Z1 inverted microscope with a Plan-Apochromat ϫ63/ 1.4 numerical aperture oil immersion objective and equipped with a X-Cite 120Q fluorescence illumination system (Lumen Dynamics Group Inc.) and an Evolve 512 electron-multiplying charged-coupled device camera (Photometrics Technology). MetaFluor software (Molecular Devices) was used for image acquisition and vesicle selection. Images were acquired at 490and 440-nm excitation wavelengths. Average intensities of fluorescent puncta (0.3-3 m in diameter) (570-nm emission using a long-pass 525-nm emission filter) were obtained for both the 440 and 490 images, and the 490/440 ratios were calculated.
Calibration curves of fluorescence as a function of vesicular pH values were performed in situ in AP-1 cells containing Tf-FITC by clamping the vesicular pH between 5 and 7.8 in K ϩ -rich medium ( Endoplasmic reticulum-To measure pH in the ER, an ERtargeted pH-sensitive probe was engineered by replacing Dendra 2 in Dendra2_ER-5 (a gift from Michael Davidson, Addgene plasmid no. 57716) by the pH-sensitive GFP pHluorin2 using the AgeI and BspEI restriction sites. This modified construct contains an ER signal sequence from calreticulin at its N terminus and an ER retention sequence (KDEL) at its C terminus (named ERpH2). AP-1 cells were grown in 6-well plates and transfected with 1 g of ERpH2 alone or together with 1 g of ChFP NHE6 ChFP WT or mutants using Lipofectamine2000 according to the manufacturer's instructions. Twenty four hours after transfections, cells were transferred to fibronectincoated FluoroDishes TM (World Precision Instruments, Inc.) and grown overnight. ER pH was measured by single-cell FRIA at 37°C, using a Zeiss LSM 780 confocal microscope, equipped with a PMT detector. Images were acquired with a ϫ63/1.4 NA oil immersion objective lens by sequential line scanning at 405 and 488 excitation wavelengths, with emission set at 500 -550 for both channels. All cells were imaged in a heated chamber (37°C, 5% CO 2 ) at ϫ3 zoom, and the laser power, resolution, speed of scanning, digital gain and offset, pinhole opening, and line averaging were identical for both channels. Average intensities of fluorescent puncta or regions of interest (0.3 to 3 m in diameter) were obtained for both the 405 and 488 channels using the MetaXpress software (Molecular Devices, Downingtown, PA), and 488/405 ratios were calculated. Calibration curves of fluorescence as a function of pH e were performed in situ in AP-1 cells expressing ERpH2 by clamping the intracellular pH between 5 and 7.8 in K ϩ -rich medium (

Flow cytometry
To measure transferrin uptake by flow cytometry, HeLa cells were transfected with GFP alone or the NHE6 GFP constructs using FuGENE6 (Promega). Twenty four hours after transfection, the cells were serum-depleted for 2 h, then incubated with Alexa Fluor 633-conjugated transferrin (Tf-AF 633 , 10 g/ml) for 5 min at 37°C, and followed by washes to remove unbound transferrin. Cells were detached from the plates by trypsinization, and 5 l of the cell viability dye 7-amino-actinomycin D (7-AAD, eBioscience) was added to each cell suspension. Cells were analyzed by flow cytometry using a FACS Aria Sorter (BD Biosciences). A gate was set around the GFP-positive cells, and the amount of Tf-AF 633 taken up by 10 4 GFP-expressing live cells (i.e. 7-AAD negative) was measured using the FACS Diva software (BD Biosciences).

Apoptosis assays
Apoptosis in AP-1 cells was measured using a flow cytometry assay that measures changes in plasma membrane asymmetry binding of annexin V-allophycocyanin conjugate (annexin V-APC) to phosphatidylserine and permeability to propidium iodide (PI) as described previously (53). Briefly, AP-1 cells were grown in 6-cm dishes and transfected with 1 g of GFP or 4 g of NHE6v1 GFP WT or mutant constructs using Lipofec-tamine2000 (Invitrogen). Forty eight hours post-transfection, the cells were washed twice with warm (37°C) PBS, detached using Cell Dissociation Buffer (Gibco), and then collected by centrifugation. The cell pellets were resuspended in PBS and labeled using the annexin V-APC apoptosis detection kit (eBioscience) according to the manufacturer's instructions. After labeling, cells were placed on ice, and 1 ϫ 10 4 GFP-positive cells were examined on a BD TM LSR II flow cytometer, and the percentage of annexin V-positive cells (i.e. early apoptotic), PI-positive (i.e. necrotic), and annexin V/PI-double positive cells (i.e. late apoptotic) was determined.

Characterization of Christianson syndrome SLC9A6 variants MD simulation
The construction of a homology structure model of monomeric human NHE6 (transcript splice-variant 1) was described in our previous publication (57). The objective of the simulation was to characterize the potential effects of the L188P mutation located in membrane-spanning helix M4 and the G383D mutation in the re-entrant loop between helices M8 and M9 of NHE6. The mutations were introduced into the homology model by using the program Coot (133). Energy minimization and MD simulation were carried out with the GROMACS package (134). First, the topology for the NHE6 model was generated with the GROMACS pdb2gmx module and the OPLS all-atom force field parameters (135). The protein was then placed in a periodic cubic simulation box, with the minimum distance of 1.0 nm to the box edge. For system solvation, the SPC/E model was used (136). After charge neutralization with counterions, the system was energy-minimized to the target F max of no greater than 1000 kJ mol Ϫ1 nm Ϫ1 by using the steepest descent algorithm implemented in GROMACS. To constrain covalent bond lengths, the LINear Constraint Solver (LINCS) algorithm was applied. To model long-range electrostatic interactions, the Particle Mesh Ewald method was used. Short-range van der Waals and Coulomb cutoff distances were both 1.0 nm. For the production run, the V-rescale thermostat and the Parrinello-Rahman barostat were used to couple the system to a constant temperature of 300 K and a pressure of 1 bar. The total simulation period was 10 ns, and the equations of motion were integrated with the leap-frog algorithm in time steps of 2 fs.

Statistical analyses
Statistical analyses were performed using OriginPro software. The data are presented as the mean Ϯ S.D. unless stated otherwise. Means of multiple samples were compared using either a one-way, one-way repeated measures, or two-way ANOVA followed by post hoc Tukey or Dunnett's test, as indicated. Data (where n Ն 6) were analyzed for statistical outliers using a Grubbs test. A few outliers were detected for some experiments that did not significantly affect the mean, so all the data were retained in the final calculations. A minimum p value of Ͻ0.05 was considered significant.