The anion transporter SLC26A9 localizes to tight junctions and is degraded by the proteasome when co-expressed with F508del–CFTR

Mutations in the gene encoding the cystic fibrosis transmembrane conductance regulator (CFTR) disrupt epithelial secretion and cause cystic fibrosis (CF). Available CFTR modulators provide only modest clinical benefits, so alternative therapeutic targets are being explored. The anion-conducting transporter solute carrier family 26 member 9 (SLC26A9) is a promising candidate, but its functional expression is drastically reduced in cells that express the most common CF-associated CFTR variant, F508del–CFTR, through mechanisms that remain incompletely understood. Here, we examined the metabolic stability and location of SLC26A9 and its relationship to CFTR. Compared with SLC26A9 levels in BHK cells expressing SLC26A9 alone or with WT–CFTR, co-expression of SLC26A9 with F508del–CFTR reduced total and plasma membrane levels of SLC26A9. Proteasome inhibitors increased SLC26A9 immunofluorescence in primary human bronchial epithelial cells (pHBEs) homozygous for F508del–CFTR but not in non-CF pHBEs, suggesting that F508del–CFTR enhances proteasomal SLC26A9 degradation. Apical SLC26A9 expression increased when F508del–CFTR trafficking was partially corrected by low temperature or with the CFTR modulator VX-809. The immature glycoforms of SLC26A9 and CFTR co-immunoprecipitated, consistent with their interaction in the endoplasmic reticulum (ER). Transfection with increasing amounts of WT–CFTR cDNA progressively increased SLC26A9 levels in F508del–CFTR-expressing cells, suggesting that WT–CFTR competes with F508del–CFTR for SLC26A9 binding. Immunofluorescence staining of endogenous SLC26A9 and transfection of a 3HA-tagged construct into well-differentiated cells revealed that SLC26A9 is mostly present at tight junctions. We conclude that SLC26A9 interacts with CFTR in both the ER and Golgi and that its interaction with F508del–CFTR increases proteasomal SLC26A9 degradation.

The airway surface liquid (ASL) 3 is a thin layer of fluid that lines the conducting airways and plays a crucial role in the mucociliary clearance of inhaled pathogens (1). The volume and composition of the ASL determine the efficiency of mucociliary clearance and are tightly regulated by epithelial ion and fluid secretion (1,2). An important channel mediating airway anion secretion is the cystic fibrosis transmembrane conductance regulator (CFTR) (3)(4)(5), a cAMP/protein kinase Aregulated channel in the ABC transporter family that has two membrane-spanning domains (MSD1 and MSD2), two nucleotide-binding domains (NBD1 and NBD2), and a regulatory (R) domain (3, 6 -8). Loss-of-function mutations in CFTR impair anion and fluid secretion and increase the susceptibility of airways to infection (9,10).
The most common CFTR mutation is a deletion of phenylalanine at position 508 (F508del-CFTR), which causes misfolding and retention of the mutant in the endoplasmic reticulum (ER) and its rapid degradation by the ubiquitin-proteasome pathway (11,12). CFTR correctors (i.e. small molecule pharmacological chaperones that partially restore the folding and trafficking of this mutant) have been described; however, they provide modest clinical benefit and only for a subset of patients (13). Thus, there is increasing interest in alternative anion efflux pathways as potential therapeutic targets, such as the Cl Ϫ conductance SLC26A9 (14 -19). SLC26A9 activity protects mice from mucus airway obstruction, and polymorphisms in the SLC26A9 gene that reduce its expression in human airways are associated with asthma (20). Genome-wide association studies have also identified SLC26A9 as a modifier of CF severity and CFTR potentiator efficacy, and several groups have reported interactions between SLC26A9 and CFTR (21)(22)(23)(24).
SLC26A9 has a transmembrane domain with putative N-glycosylation sites at asparagine 153 and asparagine 156, a STAS (sulfate transporter and anti-factor antagonist) domain, and a C-terminal PDZ domain-binding motif (25). SLC26A9 can be coimmunoprecipitated with both WT-CFTR and F508del-CFTR using lysates of human embryonic kidney 293 (HEK293) cells or the human cystic fibrosis bronchial epithelial cell line CFBE41oϪ (14,17). Interaction between the STAS domain of SLC26A9 and the R domain of CFTR has been shown biochemically using purified recombinant domains (16), and interaction between the STAS domain of a closely-related protein (SLC26A3) and the R domain of CFTR was demonstrated using NMR (26). Importantly, Bertrand et al. (14) found that SLC26A9-dependent currents can be measured when SLC26A9 is co-expressed with WT-CFTR in HEK293 cells, but not when co-expressed with F508del-CFTR. Although whole-cell SLC26A9 levels, including the mature glycoform, were similar when SLC26A9 was overexpressed with WT or mutant CFTR in HEK cells, plasma membrane expression of SLC26A9 was reduced in the presence of F508del-CFTR, and it was co-immunoprecipitated with the Golgi-localized PDZ protein CAL (CFTR-associated ligand (27)).
Recently, CAL has also been demonstrated in the ER (28); however, potential degradation of SLC26A9 by the proteasomal pathway at the ER has not been investigated.
It is important to understand the SLC26A9 trafficking abnormality induced by F508del-CFTR as it is a hurdle for the development of SLC26A9 as a therapeutic target. Approximately 90% of individuals with CF have at least one F508del-CFTR allele. Here, we confirm that SLC26A9 surface expression is diminished by F508del-CFTR and then examine the mechanism of premature degradation using inhibitors, surface biotinylation, fluorescence microscopy, and functional assays. In addition to CAL-dependent degradation at the Golgi, as described previously (27), the present results reveal a novel mechanism in which F508del-CFTR causes the retention of SLC26A9 at the ER and degradation by the proteasome. Although interaction with WT-CFTR was observed and may normally enhance the maturation and trafficking of SLC26A9 in well-differentiated primary human bronchial epithelial (pHBE) cells, the latter was localized at tight junctions and had much faster turnover at the cell surface compared with CFTR. These findings clarify the dependence of SLC26A9 on CFTR and support the development of disruptors of the SLC26A9 -F508del-CFTR interaction as a therapeutic strategy for CF.

F508del reduces SLC26A9 expression
To examine the influence of CFTR on SLC26A9 protein expression and trafficking, we transfected 3HA-SLC26A9 into parental baby hamster kidney (BHK) cells lacking CFTR (BHKparental) and also into BHK cell lines that stably express WT-CFTR (BHK-WT) or F508del-CFTR (BHK-F508del) and then immunoblotted 48 h later for SLC26A9. SLC26A9 expression was consistently much lower in BHK-F508del cells than in BHK-WT cells and was about half that in BHKparental cells devoid of CFTR (Fig. 1, A and B). These results indicate that F508del-CFTR has a negative effect on steadystate SLC26A9 expression and is more deleterious than the complete absence of CFTR, evidence that SLC26A9 may be retained intracellularly and degraded prematurely as described for F508del-CFTR. The immunoblots revealed two SLC26A9 bands that likely represent the immature, nonglycosylated form (band B) and the mature, complex-glycosylated form (band C), as shown for CFTR (29). The immature band B glycoform of SLC26A9 was diminished in F508del-CFTR-expressing cells while band C SLC26A9 was still present, suggesting that some SLC26A9 maturation still occurs despite F508del-CFTR. For comparison, we examined SLC26A9 levels in a BHK cell line that stably expresses G551D-CFTR, a mutant with defective channel gating that traffics normally to the plasma membrane. When co-expressed with G551D-CFTR, SLC26A9 levels were intermediate between those with BHK-WT and BHK-F508del and slightly higher than in parental cells that lack CFTR completely (Fig. 1B). This further suggests that SLC26A9 may be degraded prematurely along with F508del-CFTR. SLC26A9 mRNA levels were similar in all three cell lines as measured using quantitative PCR (qPCR) (Fig. 1C); therefore, the reduction in SLC26A9 expression by F508del-CFTR is not due to decreases in transfection efficiency or transcription.
To determine whether SLC26A9 trafficking to the plasma membrane in BHK cells is adversely affected by F508del-CFTR, cell-surface biotinylation assays were performed. As with whole-cell lysate protein levels, SLC26A9 expression at the plasma membrane was lower in BHK-F508del cells compared with BHK-WT cells (Fig. 1D). The function of SLC26A9 was also assessed in these cells using the FLIPR membrane potential (FMP) assay, which reports depolarization of the membrane potential due to Cl Ϫ efflux. A reduction in the SLC26A9-dependent change in membrane potential during challenge with low-Cl Ϫ solution was evident in cells co-expressing F508del-CFTR (Fig. 1E).
To examine whether F508del-CFTR also suppresses SLC26A9 in airway epithelial cells, pHBEs from non-CF and F508del homozygote donors were studied. We first confirmed the specificity of the antibody by transfecting 3HA-SLC26A9 into BHK cells and co-immunostaining with anti-SLC26A9 and anti-HA antibody (Fig. 1F). Nearly perfect co-localization was observed (Fig. 1F, Merge), and no signal was detected in neighboring cells that had not been transfected (Fig. 1F, brightfield). Specificity was further confirmed in pHBEs by transducing them with adenoviral eGFP-SLC26A9 and comparing eGFP fluorescence and immunofluorescence signals (Fig. 1G). Immunostaining was superimposable with the eGFP fluorescence and again was detected only in pHBE cells that had been successfully transduced by the eGFP-SLC26A9 adenovirus. Despite similar levels of SLC26A9 mRNA (Fig. 1H), immunofluorescence staining of SLC26A9 was clearly reduced in F508del/ F508del pHBE cells compared with non-CF cells (Fig. 1, I and J), and similar results were obtained using a different anti-SLC26A9 antibody (data not shown). These results confirm

Localization and degradation of SLC26A9
that co-expression of F508del-CFTR reduces total and plasma membrane SLC26A9 expression when SLC26A9 is transfected into BHK cells or endogenously expressed at physiological levels in pHBE cells.

SLC26A9 is degraded by both proteasomal and lysosomal pathways
Impaired folding of F508del-CFTR leads to its retention in the ER and premature degradation by the 26S proteasome (12). Because levels of band B SLC26A9 were reduced in BHK-F508del cells, we examined whether SLC26A9 interacts with F508del-CFTR by precipitating the latter using M3A7 antibody and probing the blots for SLC26A9. F508del-CFTR seemed to interact predominantly with band B SLC26A9, suggesting an association during biosynthesis ( Fig. 2A). We then explored whether SLC26A9 is degraded by the proteasome pathway when F508del-CFTR is not present. SLC26A9 was transiently expressed in BHK-parental cells, which were then treated with the proteasome inhibitor MG132 or the lysosome inhibitor chloroquine (CQ). SLC26A9 expression was increased by CQ, suggesting it is at least partially degraded by the lysosome (Fig. 2B). Although there was no increase in soluble SLC26A9 in the lysate after treatment with MG132, there was a marked increase in the insoluble fraction due to polyubiquitination and aggregation (Fig. 2B). These results suggest that SLC26A9 undergoes degradation by both lysosomal and proteasomal pathways in the absence of CFTR. This was further indicated by an increase in overall SLC26A9 immunofluorescence upon treatment with either MG132 or CQ (Fig. 2C).
Pharmacological inhibitors were used to assess the role of CFTR in SLC26A9 degradation in BHK-WT and BHK-F508del cells. SLC26A9 fluorescence intensity and band intensity were increased in both cell lines after treatment with MG132 indicating proteasomal degradation, although it was difficult to determine whether there was more MG132-sensitive degradation in F508del-BHK cells, proba- After 48 h, cells were lysed, and protein was resolved using SDS-PAGE, and expression was assessed by immunoblotting. Figure is representative of 3-18 experiments. B, SLC26A9 protein expression quantified by densitometry using ImageJ and normalized to tubulin in each sample. One-way ANOVA and Tukey's multiple comparison test were used. **, p ϭ 0.0013; ***, p ϭ 0.0002; F ϭ 10.6 mean Ϯ S.E., n ϭ 10. C, SLC26A9 gene expression in transiently-transfected BHK cells was assessed using qPCR and quantified using ⌬CT analysis, normalized to GAPDH expression, n ϭ 6. D, membrane proteins labeled using sulfo-NHS-SS-biotin (0.5 g/ml) and pulled down on streptavidin beads. Labeled proteins were resolved by SDS-PAGE, and expression was determined by immunoblotting. Blot shown is representative of n ϭ 4 experiments. E, representative membrane potential measurements using the FMP assay in transiently-transfected BHK cells exposed to low-Cl Ϫ solution (see "Experimental procedures"). F, BHK cells were transiently transfected with 3HA-SLC26A9 and stained with anti-SLC26A9 and anti-HA to verify the specificity of the antibody. G, primary HBE cells were transduced with eGFP-SLC26A9 and immunostained using anti-SLC26A9 for comparison with the eGFP signal. H, SLC26A9 gene expression in well-differentiated pHBE cells from non-CF or F508del/F508del donors was determined using qPCR. Data represent the expression of SLC26A9 normalized to GAPDH (n ϭ 4 for each). I, well-differentiated pHBE cells were isolated from healthy or F508del/F508del donors, scraped, cytospun, and fixed on coverslips. SLC26A9 was stained using rabbit anti-SLC26A9 antibody, and confocal images were taken at the apical pole. J, intensity of SLC26A9 immunofluorescence calculated from 45 to 70 cells in at least four independent experiments using ImageJ software. Student's t test; ***, p Ͻ 0.001 (p ϭ 2.7 ϫ 10 Ϫ19 ) All scale bars are 10 m.

Figure 2. SLC26A9 is degraded by both lysosomal and proteasomal pathways.
A, BHK-WT and BHK-F508del were transfected with SLC26A9, immunoprecipitated with CFTR antibody (M3A7), and immunoblotted for CFTR or SLC26A9. Figure is representative of three blots. B, BHK parental cells were transfected with 3HA-SLC26A9 cDNA and exposed 32 h later to MG132 or CQ at the concentrations indicated. After 16 h of treatment, cells were lysed, proteins were resolved by SDS-PAGE, and expression was detected by immunoblotting. Blot is representative of three experiments. C, BHK cells stably expressing SLC26A9 were treated with MG132 or CQ, fixed, and stained with mouse anti-HA antibody followed by goat anti-mouse Alexa-488. Scale bar, 10 m. Images are representative of 18 -40 total cells from three independent experiments.

Localization and degradation of SLC26A9
bly due to the overexpression of both proteins (Fig. 3, A-D). However, when similar experiments were performed using well-differentiated non-CF and CF pHBE cells that express both proteins endogenously at their normal levels, MG132 treatment caused a more dramatic increase in SLC26A9 immunofluorescence in F508del/F508del cells (Fig. 3, E and F). These data suggest that SLC26A9 degradation by the proteasome is enhanced in cells that also express F508del-CFTR, depends on the stoichiometry of the two proteins, and is most evident when SLC26A9 and F508del-CFTR are expressed at physiological levels.

Correcting F508del-CFTR increases SLC26A9 expression
To study the dependence of SLC26A9 on F508del-CFTR trafficking, we tested whether pharmacological rescue of the mutant CFTR could restore SLC26A9 expression. BHK-F508del cells were treated with the small molecule corrector VX-809, which was developed to improve the trafficking of F508del-CFTR, or incubated at 27°C. VX-809 increased expression of mature (band C) F508del-CFTR, and this was accompanied by an increase in SLC26A9 protein (Fig. 4A). Similarly, when cells were incubated at low temperature, the amount of F508del-CFTR band C increased as did total SLC26A9 (Fig. 4A). Low temperature also enhanced the expression of both proteins at the cell surface (Fig. 4B). To test whether increased expression could be due to a direct effect of VX-809 or low temperature on SLC26A9 itself rather than F508del-CFTR, cells expressing only SLC26A9 were exposed to the corrector or 27°C. VX-809 treatments had no effect on SLC26A9, and 27°C caused only a slight increase that probably reflects modest improvement in folding efficiency at low temperature (Fig. 4C).
We noticed some mature SLC26A9 in the presence of F508del-CFTR suggesting it reaches the Golgi, whereas immature SLC26A9 was greatly reduced by F508del-CFTR compared with when SLC26A9 was expressed alone or with WT-CFTR. Together, these observations imply there are two populations of SLC26A9, one that is retained with F508del-CFTR and undergoes proteasomal degradation at the ER, and another noninteracting population that traffics to the plasma membrane. To test this hypothesis, clonal BHK cell lines stably co-expressing F508del-CFTR together with varying amounts of SLC26A9 were prepared. We selected one clone with low SLC26A9 expression (clone 5) and another clone with relatively high expression (clone 6) for further study (Fig. 4D). SLC26A9 levels were elevated slightly in both clone 5 (Fig. 4D, exposure 1) and clone 6 ( Fig. 4D, exposure 2) after exposure to VX-809 or 27°C. FMP assays confirmed that constitutive Cl Ϫ conductance was higher in clone 6 than clone 5, consistent with higher SLC26A9 protein expression. However, the additional SLC26A9 conductance induced by partial rescue of F508del-CFTR using VX-809 was not significant (Fig. 4E).
We then examined the effect of VX-809 on SLC26A9 levels in well-differentiated pHBE cells for comparison with BHK clones 5 and 6. SLC26A9 immunofluorescence increased when F508del/F508del pHBEs were treated with VX-809, consistent with elevated expression at the plasma membrane (Fig. 4, F and  G). Together, these results suggest that rescuing F508del-CFTR reduces ER retention and proteasomal degradation of SLC26A9, although the impact depends on SLC26A9/CFTR stoichiometry and is less noticeable when SLC26A9 is overexpressed.

Expressing WT-CFTR rescues SLC26A9 from down-regulation by F508del-CFTR
If WT-CFTR and F508del-CFTR interact with SLC26A9 at the same site, they might bind competitively. To test this, we transiently co-transfected BHK-F508del cells with SLC26A9 cDNA ϩ increasing amounts of WT-CFTR cDNA (Fig. 5A). Immunoblots revealed that SLC26A9 expression increased progressively as the amount of CFTR cDNA (and hence CFTR protein expression) was increased, with maximal SLC26A9 levels observed using 0.2 g of WT-CFTR plasmid. Forskolinstimulated, Cl-dependent membrane potential responses were proportional to the amount of WT-CFTR cDNA used, indicating successful transfection with WT-CFTR (Fig. 5B). SLC26A9 function (measured in the absence of forskolin) was also progressively increased as measured using the FMP assay (Fig. 5C). Although a clear increase in band C CFTR was not detected, this was likely due to the low level of CFTR expression and limited sensitivity of immunoblots. These results show that WT-CFTR can relieve the inhibition of SLC26A9 expression by F508del-CFTR suggestive of competitive binding to SLC26A9, although this remains to be confirmed using biochemical methods.

Degradation of immature SLC26A9 is CFTR-dependent
We next compared the rate of SLC26A9 degradation when it is expressed alone or together with WT-or F508del-CFTR. BHK cells were transfected with SLC26A9 and treated with the protein synthesis inhibitor cycloheximide for 0 -6 h and then metabolic stability of SLC26A9 and CFTR was assessed by immunoblotting at multiple time points. CFTR band B declined in WT-and F508del-CFTR cells consistent with partial processing to the mature form and degradation by the proteasome (Fig. 6, A and B). Band C CFTR was observed in BHK-WT cells after inhibiting translation for 6 h, consistent with stable surface expression and/or efficient recycling of WT-CFTR at the plasma membrane (Fig. 6, A and C). Interestingly, the degradation of immature SLC26A9 was faster in cells expressing F508del than in cells expressing WT-CFTR or those lacking CFTR completely (compare band B degradation rates in Fig.  6B), yet the degradation of mature SLC26A9 occurred at a similar rate in all three cell lines. These results suggest that only immature SLC26A9 has reduced metabolic stability in the presence of F508del-CFTR, consistent with its premature degradation in the ER together with immature F508del-CFTR. They also raise the question as to why SLC26A9 and CFTR degradation rates are so different.

SLC26A9 is endocytosed more rapidly than CFTR
SLC26A9 was eliminated within 4 h following treatment with cycloheximide, whereas band C CFTR remained nearly constant for at least 6 h and was still significant after 24 h (data not shown). Although CFTR is efficiently recycled from endosomes to the plasma membrane, little is known regarding SLC26A9

Localization and degradation of SLC26A9
turnover at the cell surface. We found that low SLC26A9 expression in F508del-expressing cells was due, at least in part, to enhanced degradation of immature SLC26A9 by the proteasome, reminiscent of F508del-CFTR. By contrast, the degradation of mature SLC26A9 appears independent of CFTR (compare red symbols in Fig. 6, B and C). To determine whether their very different metabolic stabilities might reflect distinct rates of endocytosis, we quantified the internalization of both proteins using a modified cell-surface biotinylation assay (see under "Experimental procedures"). There was a striking difference between their rates of endocytosis (Fig. 7). About one-fifth of the cell surface SLC26A9 was internalized within 2.5 min in cells expressing WT-CFTR or F508del-CFTR (Fig. 7, C and D), while only one-tenth of the CFTR was internalized within this time period (Fig. 7E). After 10 min, 44% of the cell surface SLC26A9 was endocytosed versus only 9.5% of WT-CFTR (Fig.  7, A, C, and E), and similar SLC26A9 internalization (50.5%) was observed in BHK cells expressing F508del-CFTR (Fig. 7, B  and D). Thus, SLC26A9 endocytosis is more rapid and is independent of CFTR, and this may account for the lower metabolic stability of mature SLC26A9.

SLC26A9 is expressed in both ciliated and goblet cells and localizes to the tight junctions
To localize SLC26A9, well-differentiated pHBE cell cultures were immunostained using antibodies against SLC26A9 and markers of cilia (acetylated-␣-tubulin) (30), goblet cells (mucin 5AC; MUC5AC), and undifferentiated cells (cytokeratin 14; Fig. 8). Confocal microscopy revealed SLC26A9 at the apical membrane in ciliated and goblet cells near their points of contact with neighboring cells (Fig. 8, A-C). Further studies confirmed expression in both these cell types but not in undifferentiated cells (Fig. 8D). Endogenous SLC26A9 immunostaining was co-localized with the tight junction protein zonula occludens-1 (ZO-1) (Fig. 9, A and B). We also observed some intracellular staining of SLC26A9; however, most was localized at the apical edge of the tight junction near ZO-1. Very similar results were obtained when these cells were reverse-transfected with 3HA-SLC26A9 and immunostained with anti-HA antibody (Fig. S1). To further establish the peri-junctional localization of SLC26A9, we exposed cells to EGTA for 15 min to disrupt the tight junction complexes. This caused both ZO-1 and SLC26A9 immunofluorescence to become more diffuse (Fig. 9, C-E). ZO-1 is a classical PDZ domain protein (PDZ; PSD-95/ SAP90, Discs-large, ZO-1), and SLC26A9 has a C-terminal PDZ-binding motif; therefore, we tested whether this motif might target SLC26A9 to the tight junctions. We mutated SLC26A9 so that it lacks four amino acids at the C terminus (SLC26A9⌬PDZ). Immunofluorescence of this mutant was still near the margins of CFBE41oϪ cells but was displaced by ϳ1 m from ZO-1 (Fig. 10B) compared with full-length SLC26A9 (Fig. 10, A and E). Nevertheless, the width of SLC26A9 immunofluorescence was not altered significantly (Fig. 10F), suggesting that the PDZ motif contributes to SLC26A9 localization near the tight junction but is not essential for its narrow distribution. By contrast, deletion of the STAS domain (i.e. SLC26A9⌬STAS) did not increase the distance from ZO-1 but did cause a more diffuse spatial distribution, as indicated by the width of SLC26A9⌬STAS immunostaining. When both deletions were combined in one mutant (SLC26A9⌬STAS⌬PDZ), we observed both abnormalities, i.e. a loss in localization to the tight junction and a more diffuse distribution. Together, these data suggest that the PDZ-binding motif helps localize SLC26A9 near the tight junction complex, whereas the STAS domain influences its spatial organization there.

Discussion
The Cl Ϫ conductance SLC26A9 has been proposed as a therapeutic target for the treatment of cystic fibrosis. However, Bertrand et al. (14) showed that the most common CF mutant F508del-CFTR reduces SLC26A9 expression and function, which would limit its utility as a drug target for most patients. In this study, we confirmed the inhibitory effect of F508del-CFTR on SLC26A9 and examined the mechanism of inhibition.
SLC26A9 cell-surface expression was higher in BHK cells when it was expressed alone or with WT-CFTR compared with when it was co-expressed with F508del-CFTR, consistent with previous results using HEK293 cells (31). This negative effect of F508del-CFTR on SLC26A9 was most pronounced in well-differentiated epithelial cells that endogenously expressed both proteins at physiological levels.
In previous studies the suppression of SLC26A9 by F580del-CFTR was found to be due, at least in part, to an increased interaction of SLC26A9 with CFTR-associated ligand (CAL) through its PDZ motif. That mechanism may account for the lysosomal (i.e. chloroquine-sensitive) component of SLC26A9 degradation that we have observed (Fig. 2). However, in addition to lysosomal degradation at the Golgi, many plasma membrane proteins are degraded by the endosome-lysosome pathway (32), and we observed relatively rapid SLC26A9 endocytosis that could explain the increase in SLC26A9 expression upon treatment with chloroquine (Fig. 7). Regardless, in this study we have identified an additional component of degradation that is mediated by the proteasome. Proteasomal deg-

Figure 3. Inhibiting the proteasome increases SLC26A9 levels more in cells that express F508del-CFTR than in WT-CFTR cells. BHK WT-CFTR (A) or BHK
F508del-CFTR (B) cells were transfected with 3HA-SLC26A9 32 h before adding MG132 or CQ for 16 h. Cells were stained with mouse anti-CFTR antibody or rabbit anti-HA antibody followed by goat anti-mouse Alexa-488 or goat anti-rabbit Alexa-596 and visualized by confocal microscopy. Images are representative of n ϭ 20 -40 cells in three independent experiments. C, SLC26A9 and CFTR fluorescence intensity analyzed using ImageJ (two-way ANOVA, Bonferroni post-tests; left: ***, p Ͻ 0.001 (WT-CFTR DMSO versus MG-132, p ϭ 1.0 ϫ 10 Ϫ24 ; F508del-CFTR DMSO versus MG-132, p ϭ 7.6 ϫ 10 Ϫ20 ; F508del-CFTR DMSO versus CQ, p ϭ 0.0000184, *, p ϭ 0.02, interaction factor ϭ 7.6; row factor ϭ 129.2; column factor ϭ 4.0; and right: ***, p Ͻ 0.001 (p ϭ 1.9 ϫ 10 Ϫ10 ), interaction factor ϭ 4.5; row factor ϭ 21.3; column factor ϭ 6.8). D, BHK cells stably expressing WT-CFTR (left) or F508del-CFTR were transiently transfected with SLC26A9 and treated with MG132 and CQ for 16 h, lysed, and subjected to SDS-PAGE and immunoblotting for SLC26A9. Blot is representative of n ϭ 3 experiments. E, well-differentiated pHBE cells from non-CF or CF (F508del/F508del) donors treated with MG132 for 16 h, scraped, cytospun, and fixed on coverslips. SLC26A9 was stained using rabbit anti-SLC26A9 antibody, and images were taken at the apical membrane. F, fluorescence intensity of SLC26A9 was calculated from 39 to 57 cells in at least three independent experiments. SLC26A9 fluorescence intensity was assessed using ImageJ. Two-way ANOVA, Bonferroni post test, ***, p Ͻ 0.001 (non-CF versus F508del/F508del, p ϭ 1.6 ϫ 10 Ϫ30 ; F508del/F508del DMSO versus MG132, p ϭ 9.3 ϫ 10 Ϫ12 ), interaction factor F ϭ 16.8; row factor (Ϯ MG132) F ϭ 200.8; column factor (non-CF versus F508del/F508del) F ϭ 2.0. Scale bars, 10 m. with SLC26A9 and exposed to VX-809 (1 M) or low temperature (27°C) and then lysed, and proteins were resolved by SDS-PAGE and immunoblotted. Blot is representative of four experiments. B, BHK-F508del cells were transiently transfected with SLC26A9 and incubated at 27°C. Cell-surface proteins were labeled using sulfo-NHS-SS-biotin (0.5 g/ml) and pulled down from lysates on streptavidin beads. Blot is representative of three independent experiments. D, double stable cell lines expressing both F508del-CFTR and SLC26A9. Two clones were selected that expressed both proteins but had very different SLC26A9 levels. Both clones were treated with VX-809 or low temperature and analyzed by immunoblotting. Blot representative of n ϭ 3 independent experiments. E, representative measurements of membrane potential using the FLIPR membrane potential (FMP) assay performed using both clones treated with VX-809. F, well-differentiated pHBE cells isolated from F508del/ F508del patients were treated with VX-809 (1 M) and stained using rabbit anti-SLC26A9 antibody observed using confocal microscopy. SLC26A9 immunofluorescence was calculated in 22-48 cells from at least three independent experiments. G, summary of SLC26A9 immunofluorescence quantified using ImageJ. One-way ANOVA, Tukey's multiple comparison test, ***, p ϭ 2.3 ϫ 10 Ϫ10 , F ϭ 65.4. Scale bars, 10 m. radation became apparent during treatment with MG132, lactacystin, or bortezomid (data not shown), which caused a dramatic increase in SLC26A9 band intensity in BHK cells (Fig.  3) that was most evident in pHBE cells homozygous for F508del-CFTR. A specific retention mechanism mediated by F508del-CFTR was suggested by enhanced SLC26A9 expression after partial correction of F508del-CFTR following exposure to VX-809 or low temperature. Indeed, it was possible to titrate the inhibition by F508del-CFTR by expressing increasing amounts of WT-CFTR (Fig. 5). These data indicate that rescue of F508del-CFTR reduces SLC26A9 degradation by the proteasome.

Localization and degradation of SLC26A9
One implication of these results is that pharmacological disruption of the SLC26A9-F508del-CFTR interaction may enhance the trafficking of mature SLC26A9 to the cell surface and increase Cl Ϫ conductance in CF cells. Clinical experience suggests that a potentiator of SLC26A9 conductance may also be required to alleviate symptoms however, because patients homozygous for G551D-CFTR still have severe CF despite relatively normal SLC26A9 trafficking (Fig. 1, A and B). WT-CFTR and F508del-CFTR have distinct interactomes (33), and the mutant probably affects many proteins besides SLC26A9. For example, in the epididymis where most CFTR localizes to tight junctions, ZO-1 levels are reduced in both CFTR knockout and F508del-CFTR mice, leading to nuclear localization of ZONAB (ZO-1 nucleic acid-binding protein), a transcription factor that controls cell growth and differentiation (34). The present results suggest that reduced expression of ZO-1 in F508del/F508del cells could also cause mislocalization or reduced expression of SLC26A9. It is interesting that PDZ interactions help co-localize CFTR and SLC26A9 with ZO-1 and that pharmacological inhibition of CFTR reduces ZO-1 expression. Together, these results suggest some role of Cl Ϫ conductance in stabilizing ZO-1, and we showed previously that the barrier function of airway epithelia depends on CFTR trafficking to cell surface (35). Sodium-hydrogen exchanger regulator factor 1 (NHERF1) expression and its distribution at the plasma membrane also depend on WT-CFTR, and NHERF1 expression is reduced in CFBE41oϪ cells that express F508del-CFTR (36). Whether F508del-CFTR downregulates NHERF1 and other proteins through a mechanism like the one described here for SLC26A9 remains to be deter- Blot is representative of three experiments. B and C, representative measurements of membrane potential using the FMP assay (see "Experimental procedures"). Cells were preincubated with dye in normal Cl Ϫ buffer, and then exposed to low Cl Ϫ buffer ϩ FSK (B) or low Cl Ϫ buffer without FSK (C), and the change in fluorescence was measured.

Localization and degradation of SLC26A9
mined. Interestingly, SLC26A4 (pendrin), another transporter hypothesized to interact with CFTR, is elevated in CF primary cells homozygous for F508del-CFTR compared with non-CF cells, probably a transcriptional up-regulation induced by proinflammatory signaling (37). Such differences highlight the complexity of F508del-CFTR interactions with SLC26A proteins. Interestingly, it was shown recently that the epithelial sodium channel is also localized at tight junctions very similar to what we have observed with SLC26A9, suggesting that tight junctions may be an important site of salt transport (38).
In addition to studying the effects of F508del-CFTR on SLC26A9 biogenesis and trafficking, we examined the behavior of SLC26A9 at the plasma membrane and found that it is both endocytosed (Fig. 7) and degraded more rapidly than WT-CFTR (Fig. 6). SLC26A9 expression in cells from CF patients appeared lower than in those from non-CF donors, despite the low level of WT-CFTR normally present in this cell type. Thus, SLC26A9 expression is reduced in F508del-CFTR cells that would normally have little if any CFTR immunostaining.
In summary, these results indicate that F508del-CFTR reduces SLC26A9 expression through SLC26A9 retention in the ER and premature degradation by the proteasome. Future studies should identify the site(s) of the SLC26A9 -F508del-CFTR interaction so that a high-throughput biochemical assay for disrupters of the interaction can be developed to increase SLC26A9 surface expression and anion secretion in CF airways.

Cell culture
Parental BHK cells were cultured in Dulbecco's modified Eagle's medium/nutrient mixture F-12 supplemented with 100 units/ml penicillin, 100 g/ml streptomycin, and 10% FBS. The medium for BHK cell lines stably expressing WT-CFTR or F508del-CFTR also contained 500 M methotrexate. The medium was changed every 2 days, and cells were maintained at 37°C in humidified air containing 5% CO 2 . Primary human bronchial epithelial cells (HBEs) were obtained from the Primary Airway Cell Biobank at McGill University, which isolated them from tissue received from the Biobank of Respiratory Tissue at the Centre Hospitalier de l'Université de Montréal. Informed written consent was obtained from patients undergoing lung transplantations, and all procedures were approved by the Institutional Review Board of McGill University (catalog no. A08-M70-14B). The studies abide by the Declaration of Helsinki principles. Cells were cultured and differentiated using methods similar to those described by Fulcher et al. (41). Briefly, they were seeded onto collagen-coated polyester mem-

Localization and degradation of SLC26A9
brane inserts (Corning) and remained submerged for 4 days. The apical medium was then removed, and the cells were allowed to differentiate at the air-liquid interface (ALI) for at least 21 days before use.

SLC26A9 construct and transfection
SLC26A9 cDNA (NM_052934.3) was excised from PCMVS-port6 and subcloned into 3HA-PCMVsport6 to generate 3HA-SLC26A9 -PCMVsport6. 3HA-SLC26A9 was then subcloned into pcDNA3.1 or pNUT. To obtain the SLC26A9 mutants (SLC26A9⌬PDZ and SLC26A9⌬STAS⌬PDZ), TGA was inserted using mutagenesis at the desired site of the 3HA-SLC26A9 -pcDNA3.1 for the truncated SLC26A9 mutants. SLC26A9⌬STAS was obtained by inserting an MluI site before and after the STAS domain and using MluI to excise the STAS domain. Constructs were transiently transfected into BHK cells using GeneJuice (Millipore) according to the manufacturer's instructions. In well-differentiated CFBE cells, polyJet (frogobio) was used to transfect the cells. Briefly, the cDNA and polyJet were added together following the manufacturer's instructions and added onto the filters. CFBE cells were then washed with PBS, trypsinized, and seeded onto the filter containing the polyJet transfection solution. The following day, cells were put in ALI conditions and cultured for 1 week.
To image well-differentiated pHBE cells, they were cultured at the air-liquid interface for 1 month, washed two times with PBS, and either fixed immediately or gently scraped and centrifuged onto coverslips at 450 rpm for 5 min using a Cytospin4, then fixed with 4% PFA, permeabilized with 0.5% Triton X-100, and blocked with 2% BSA. Cells were immunostained using anti-SLC26A9 (Novus Biological, 1:50), anti-CFTR (596, provided by T. Jensen and J. Riordan, University of North Carolina, Chapel Hill, through the CFF CFTR Antibody Distribution Program), or antibodies against tubulin, ZO1, MUC5AB, or cytokeratin 14 from Santa Cruz Biotechnology (1:200).
The intensity of SLC26A9 immunofluorescence staining was analyzed using ImageJ (40). The background was corrected, and the brightness and contrast were adjusted to visualize the feature and then kept constant between different conditions. For BHK cells, fluorescence intensity was measured in cells chosen using the "freehand selection tool." In the case of primary cells where SLC26A9 was localized to tight junctions, fluorescence intensity was measured along a line drawn across the cells, and 5-10 measurements were taken per cell and averaged to estimate the fluorescence of one cell. Images were normalized to the control condition for BHK cells or to non-CF cultures in the case of primary cells.

RNA extraction and quantitative real-time PCR
Cells were seeded on 6-well plates (Corning) at a density of 10 5 cells per well and studied at 48 h post-transfection. Alternatively, RNA was prepared from well-differentiated pHBE that had been cultured at the ALI for 1 month. RNA was extracted

Localization and degradation of SLC26A9
and purified using the RNase Easy mini kit (Qiagen) following the manufacturer's instructions. For reverse transcription, 500 ng of RNA was added to 4 l of SuperScript VILO Mastermix (Thermo Fisher Scientific) in a total volume of 20 l for 1 h at 42°C, and for 5 min at 85°C. Then, 250 ng of cDNA, 10 l of TaqMan Fast Advanced Mastermix, and 1 l of TaqMan Gene Expression Assay primers in a total volume of 20 l were added to the wells of a MicroAmp EnduraPlate TM Optical 96-Well Fast Reaction Plate. qPCR was carried out using a QuantStudio TM 7 Flex Real-Time PCR system with the follow-ing protocol: 20 s at 95°C and 40 cycles at 95°C (1 s) and 60°C (20 s). This was followed by ⌬⌬Ct analysis.

Localization and degradation of SLC26A9
secondary antibody for 1 h at room temperature, and washed four times with TBST. The bands were visualized using Amersham Biosciences ECL start Western blotting detecting reagent (GE Healthcare) and ChemiDoc Imaging system (Bio-Rad).

Cell-surface biotinylation
Transfected BHK cells were cultured in 100-mm dishes for 48 h and washed two times with ice-cold PBS. Sulfo-NHS-SSbiotin was applied onto the cell surface for 10 min at 4°C, and the reaction was stopped using PBSϩϩ (PBS containing 1 mM MgCl 2 and 0.1 mM CaCl 2 , pH 8.0). After washing cells with PBSϩϩ, they were lysed in RIPA buffer, and an equal amount of total protein was incubated with streptavidin-agarose beads for 2 h at 4°C and then eluted using 2ϫ Laemmli buffer and analyzed by immunoblotting.

Co-immunoprecipitation
Transfected BHK cells were washed two times with ice-cold PBS and lysed (150 mM NaCl, 1 mM EDTA, 50 mM HEPES, 0.1% Triton X-100). The soluble lysate was precleared with Sepharose G beads for 30 min. The precleared lysates were incubated with M3A7 CFTR antibody for 30 min at 4°C. Immunocomplexes were precipitated using Sepharose G beads. The beads were washed with lysis buffer and eluted using 2ϫ Laemmli buffer, and analyzed by immunoblotting.

Endocytosis assay
Transfected BHK cells were cultured in 60-mm dishes for 48 h and washed two times with ice-cold PBS. Sulfo-NHS-SSbiotin was applied onto the cell surface as described above and then the cells were washed and returned to the 37°C incubator for the time period indicated. All remaining surface-biotinylated protein then was stripped at different time points using GSH buffer (50 mM GSH, 75 mM NaCl, 1 mM MgCl 2 , 0.1 mM CaCl 2 , 80 mM NaOH, 10% FBS, pH 8.6). Cells were subsequently lysed in RIPA buffer, and equal amounts of total protein were incubated with streptavidin-agarose beads for 2 h at 4°C, eluted using 2ϫ Laemmli buffer, and analyzed by immunoblotting (39).

FMP assay
The FMP assay (Molecular Devices) was used to monitor changes in membrane potential. BHK cells (10 6 ) were seeded in 100-mm dishes and transfected. After 24 h, they were detached using Detachin (Genlantis, San Diego) and seeded on black 96-well, half-area, flat-bottomed microplates (Corning) 24 h prior to study. 20ϫ stock FMP dye was diluted in normal Cl Ϫ buffer (in mM: 4.5 KCl, 115 NaCl, 1.2 MgCl 2 , 1.2 CaCl 2 , 10 glucose, and 10 HEPES) and exposed to cells for 30 min. Plates were then placed in the Synergy plate reader, and basal fluorescence was measured using a 530-nm laser for excitation and collecting emission at 565 nm. FMP dye (1ϫ) in low-Cl Ϫ buffer (mM: 4.5 KCl, 115 sodium gluconate, 1.2 MgCl 2 , 4 mM CaCl 2 , 10 glucose, and 10 HEPES) was injected onto the cells to generate an outward Cl Ϫ gradient, and the total fluorescence of each well was measured at 8-s intervals.

Statistics
Data are shown as means Ϯ S.E. and were analyzed using Prism 5 for Mac (GraphPad software). To determine whether differences were significant, Student's t test and one-or twoway ANOVA with Tukey's multiple comparison post-tests were carried out as indicated in the figure legends.