GOPC facilitates the sorting of syndecan-1 in polarized epithelial cells

Abstract

The trans-Golgi network must coordinate sorting and secretion of proteins and lipids to intracellular organelles and the plasma membrane. During polarization of epithelial cells, changes in the lipidome and the expression and distribution of proteins contribute to the formation of apical and basolateral plasma membrane domains. Previous studies using HeLa cells show that the syndecan-1 transmembrane domain confers sorting within sphingomyelin-rich vesicles in a sphingomyelin secretion pathway. In polarized Madin–Darby canine kidney cells, we reveal differences in the sorting of syndecan-1, whereupon the correct trafficking of the protein is not dependent on its transmembrane domain and changes in sphingomyelin content of cells during polarization. Instead, we reveal that correct basolateral targeting of syndecan-1 requires a full-length PDZ motif in syndecan-1 and the PDZ domain golgin protein GOPC. Moreover, we reveal changes in Golgi morphology elicited by GOPC overexpression. These results suggest that the role of GOPC in sorting syndecan-1 is indirect and likely due to GOPC effects on Golgi organization.

INTRODUCTION

The ability to correctly process, sort, and traffic proteins and lipids is a fundamental determinant of cell function. The trans-Golgi network (TGN) is an integral protein and lipid sorting center and mediates the recognition, packaging, and targeting of cargoes for delivery to the plasma membrane or intracellular organelles. The heterogeneity of proteins and destinations inevitably requires multiple cargo recognition modules, signaling pathways, and the formation of coated and uncoated transport carriers at the TGN (Ford et al., 2021; Ramazanov et al., 2021).

In polarized cells, such as those that make up epithelia, integral membrane proteins are differentially targeted to apical or basolateral plasma membrane domains (Weisz and Fölsch, 2020). In these cells, several signals have been identified that confer sorting. For apically targeted proteins, these can include a GPI-anchor and/or glycan modifications to proteins, and it seems that there may be a more generalized requirement for clustering to facilitate apical sorting (Paladino et al., 2004, 2008, 2014; Delacour et al., 2007; Mishra et al., 2010; Lebreton et al., 2021). Amino acid sequence–based signals for basolateral sorting of proteins are slightly better defined: tyrosine- or dileucine-based sorting motifs present in the cytoplasmic tails of cargo proteins (YXXΦ or [DE]XXXL[LI] [where X is any amino acid and Φ is a bulky hydrophobic amino acid]) (Tan and Gleeson, 2019; Ford et al., 2021). However, the full cohort of molecular determinants for selection of protein and lipids to be incorporated into TGN-derived secretory carriers and how these are coupled to coordinate cargo sorting with vesicle budding and fission is unknown.

Recently, we identified a class of sphingomyelin (SM)-rich, Golgi-derived transport carriers in HeLa cells that mediate sorting of a specific subset of protein cargoes, including syndecan-1 (Sdc1), upon exit from the TGN, which we termed the SM secretion (SMS) pathway (Deng et al., 2018). In HeLa cells, it was revealed that the sorting and secretion of Sdc1 within the SM-rich vesicles of the SMS pathway is dependent on the transmembrane domain of Sdc1 (Sundberg et al., 2019). In Madin–Darby canine kidney (MDCK) epithelial cells, Sdc1 is targeted to the basolateral plasma membrane (Miettinen et al., 1994; Maday et al., 2008). However, the structural features of Sdc1 that confer basolateral targeting are unknown. Several Sdc1-interacting partners contain PSD-95/discs large/ZO-1 (PDZ) domains that typically bind PDZ binding motifs present in the 4–10 carboxy-terminal residues of client proteins (Lee and Zheng, 2010). All four syndecan family members (Sdc1–4) share the same carboxy-terminal PDZ motif sequence, and several PDZ domain–containing proteins are thought to bind to this region, including CASK, synectin, synbindin, and syntenin (Grootjans et al., 1997; Hsueh et al., 1998; Ethell et al., 2000; Gao et al., 2000), although specificity for specific syndecans and/or posttranslational modifications of the PDZ-binding motif are not clear (Cheng et al., 2016). Nevertheless, mutant Sdc1 (Sdc1∆YA), lacking the last two amino acids in the four-residue PDZ binding motif, was incorrectly sorted to both plasma membrane domains (Miettinen et al., 1994; Maday et al., 2008). This suggests that PDZ domain proteins may be involved in post–Golgi sorting of Sdc1, particularly as the protein is delivered directly to the basolateral plasma membrane and does not utilize a transcytotic delivery route (Maday et al., 2008). Nevertheless, known PDZ motif–binding partners of Sdc1 do not appear to be involved in basolateral targeting of the protein (Maday et al., 2008). Though PDZ domain–containing proteins have been shown to be required for the polarized distribution of membrane proteins through retention at the plasma membrane and/or via endocytic recycling (Perego et al., 1999; Shenolikar et al., 2002; Swiatecka-Urban et al., 2002; Milewski et al., 2005), a link between PDZ domain–containing proteins and sorting in the Golgi apparatus is less clear. GOPC (also known as PIST, CAL, or FIG) is a trans-Golgi localized PDZ domain–containing protein that is reported to bind via its PDZ domain numerous disparate transmembrane plasma membrane receptors, presumably either en route during biosynthesis or following internalization from the plasma membrane (Yao et al., 2001; Cheng et al., 2002; Swiatecka-Urban et al., 2002; He et al., 2004; Wente et al., 2005; Li et al., 2006 ). Some reports suggest that GOPC interaction is associated with protection from and conversely promotion of lysosomal degradation, while others link GOPC with the autophagic machinery within cells (Cheng et al., 2004, 2010b; Joubert et al., 2009; Meiffren et al., 2010; Luo et al., 2019; Wilhelmi et al., 2021). Multiple reports link GOPC overexpression with retention of PDZ motif–containing proteins intracellularly (Cheng et al., 2002; He et al., 2004; Wente et al., 2005; Nie et al., 2016), which has been suggested to result from an imbalance between competitive PDZ-motif interactors that ultimately control protein localization (Cheng et al., 2002). However, most of these phenomena are based on ectopic expression of putative GOPC secreted clients in cultured cell lines of unclear physiological relevance. In addition to the PDZ domain–mediated interactions, the coiled-coil domains of GOPC are implicated in interactions with syntaxin-6, golgin-160, and the Rho GTPase Tc10 (Charest et al., 2001; Neudauer et al., 2001; Hicks and Machamer, 2005).

We sought to clarify how Sdc1 sorting in polarized cells relates to the SMS pathway defined in HeLa cells. This is particularly relevant given the physiological role of Sdc1 at the basolateral plasma membrane, where it has significant roles in cell signaling and cell–matrix and cell–cell interactions (Teng et al., 2012; Palaiologou et al., 2014). Moreover, we wanted to examine how the SMS pathway might function within polarized epithelial cells, which by necessity incorporate higher sorting complexity. This analysis revealed differences in Sdc1 sorting in HeLa versus polarized epithelial cells, where the trafficking of Sdc1 to the basolateral plasma membrane is not dependent on the sequence of its transmembrane domain, as it is in HeLa cells. Instead, basolateral trafficking of Sdc1 is dependent on the golgin, GOPC. The results reveal the importance of the Golgi “matrix” in the sorting of integral membrane secretory protein cargo.

RESULTS AND DISCUSSION

Sorting of EQ-SM within mesenchymal and polarized MDCK cells

To visualize SM trafficking in cells, we previously engineered equinatoxin II, a SM-binding protein produced by the marine organism Actinia equina, into a nontoxic SM reporter protein, termed “EQ-SM” and a non–SM binding control protein, “EQ-Sol” (Deng et al., 2016). Within the secretory pathway of HeLa cells, EQ-SM is exported from the TGN in vesicular carriers that constitutively fuse with the plasma membrane releasing the protein. Once released at the plasma membrane, EQ-SM remains bound to SM in the outer leaflet (Deng et al., 2016). In HeLa cells, EQ-SM and Sdc1 are delivered to the plasma membrane in the same vesicles (Sundberg et al., 2019). However, in polarized epithelial cells, Sdc1 localizes to the basolateral domain of the plasma membrane (Miettinen et al., 1994; Maday et al., 2008). Therefore, it was our interest to determine whether EQ-SM and Sdc1 are transported in SM-rich vesicles in polarized epithelial cells, as they are in nonpolarized HeLa cells. To this end, we used MDCK (clone II) cells for these experiments because Sdc1 trafficking has been previously examined in this cell type, though little is known about how features of Sdc1 or Sdc1-containing vesicles confer the basolateral localization of the protein.

To track EQ-SM and EQ-Sol as they are trafficked through the secretory pathway, we utilized the “retention using selective hooks” (RUSH) system (Boncompain et al., 2012) to synchronously release EQ-SM and EQ-Sol from the endoplasmic reticulum (ER). In this study, all exogenous proteins were expressed in MDCK cells by lentiviral transduction to yield stably expressing mesenchymal (i.e., nonpolarized) and epithelial cell lines (Figure 1). Owing to decreases in expression of transgenes with advancing passage number of transduced cells, heterogeneous (i.e., not clonal) populations of cells were used in all experiments. A RUSH-release time course was carried out before fixation and analysis by confocal microscopy. In these experiments, both EQ-SM and EQ-Sol were observed to be retained in the ER and upon biotin addition to the cell culture medium were transported to and through the Golgi, into post-Golgi compartments and vesicles, and finally to the plasma membrane. In agreement with our previous findings in HeLa cells (Deng et al., 2016), EQ-SM remained bound to the plasma membrane of mesenchymal MDCK cells, whereas EQ-Sol, which does not bind SM, was released into the cell culture supernatant (Figure 1A).

FIGURE 1:

RUSH EQ-SM and EQ-Sol trafficking in MDCK cells. (A) Example micrographs of mesenchymal MDCK cells expressing RUSH-EGFP-EQ-SM or RUSH-EGFP-EQ-Sol at 0, 10, 30, 60, and 90 min after the addition of biotin, imaged by deconvolution microscopy. Scale bars represent 20 μm. (B) Example micrographs of polarized MDCK cells expressing RUSH-EGFP-EQ-SM or RUSH-EGFP-EQ-Sol at 0, 30, 60, and 90 min after the addition of biotin. Both single XY slices and XZ stacks are shown. Cells were fixed before staining with anti-NaK ATPase and anti-GP135 primary antibodies, followed by DyLight 550 (red)- and AlexaFluor 633 (blue)-conjugated secondary antibodies, respectively. Scale bars represent 20 μm. (C) Representative immunoblots of secreted RUSH-EGFP-EQ-SM or RUSH-EGFP-EQ-Sol collected from cell culture medium. MDCK cells expressing RUSH-EGFP-EQ-SM or RUSH-EGFP-EQ-Sol were polarized, and RUSH proteins were released from the ER by the addition of biotin. Apical and basolateral media as well as respective whole cell lysates were collected at 0 and 90 min postrelease. Media and lysates were processed for immunoblotting with an anti-GFP antibody. (D) Graph showing quantification of band intensity of three representative immunoblots of RUSH-EGFP-EQ-SM or RUSH-EGFP-EQ-Sol secretion into apical and basolateral medium as described in C. Blots were analyzed using ImageJ.

In epithelial MDCK cells, trafficking of EQ-SM and EQ-Sol were similar; both proteins were observed to traffic from the ER (0 min) to the Golgi (<30 min) and subapical post-Golgi vesicles (<60 min) before delivery to the plasma membrane (<90 min; Figure 1B). In contrast to mesenchymal MDCK cells, however, EQ-SM failed to remain bound to the cell surface after exocytosis, similar to EQ-Sol. Consistent with this, EQ-SM and EQ-Sol were recovered from the apical and basolateral cell culture media (Figure 1C). The proportions of EQ-SM and EQ-Sol recovered from the apical or basolateral media were 73 ± 7% and 80 ± 5% (mean ± SD), respectively (Figure 1D). In fact, the ratio of apical to basolateral secretion of these exogenous proteins closely resembled the distribution of an exogenous soluble secreted protein (Cp) in MDCK cells, which occurs via a bulk flow route (Thor et al., 2009).

During polarization, the sphingolipid profile of MDCK cells changes, shifting from majority SM and ganglioside GM3, to SM, galactosylceramide-sulfate (sulfatide), and the apically targeted GalNAcα1-3GalNAcβ1-3Galα1-4Galβ1-4Glcβ1-Cer (a.k.a., Forssman antigen) (Hansson et al., 1986; van Genderen et al., 1991; Sampaio et al., 2011). Given the differences in EQ-SM binding in mesenchymal versus epithelial MDCK cells, it is likely that the decrease in abundance of SM from ∼6 mol% to ∼5 mol% that occurs during polarization (Sampaio et al., 2011) no longer supports probe binding to cellular membranes. As a result, EQ-SM cannot be utilized as a tool to study SM trafficking in epithelial MDCK cells and instead, like EQ-Sol, appears to follow a “bulk-flow” secretion route. Nevertheless, it is intriguing that an apparently minor decrease in SM abundance should prevent EQ-SM binding, which has implications for sorting of proteins that are transported within SM-rich vesicles in nonpolarized cells.

Basolateral sorting of Sdc1 is independent of its transmembrane domain

Our previous studies identified Sdc1 as a cargo of post-Golgi, SM-rich secretory vesicles (Deng et al., 2018; Sundberg et al., 2019). Sdc1 is a single-pass transmembrane heparin sulfate proteoglycan expressed chiefly in epithelial cells. It acts as a coreceptor for numerous ligands of physiological significance including growth factors, cytokines, and chemokines through which it has diverse roles in cell signaling and cell–matrix and cell–cell interactions (Teng et al., 2012; Palaiologou et al., 2014). Moreover, increased expression of Sdc1 in pancreatic ductal adenocarcinoma cells has been proposed to support cell growth by up-regulating nutrient uptake by macropinocytosis (Yao et al., 2019). In our prior studies, we reported that the physical properties (i.e., not its sequence per se) of the transmembrane domain of Sdc1 confer its sorting into SM-rich vesicles of HeLa cells (Sundberg et al., 2019). As the amount of SM in the plasma membrane of MDCK epithelial cells is too small to support binding of EQ-SM, we hypothesized that Sdc1 could serve as a proxy to explore factors important for sorting into the SMS pathway in epithelial cells. The physical properties of the transmembrane domain, important for the SMS pathway, might also constitute elements required for basolateral sorting of the protein. We engineered MDCK cells expressing RUSH constructs pHluorin-Sdc1 WT (pHl-Sdc1) and pHluorin-Sdc1-All(Leu) (pHl-Sdc1-All(Leu), in which the transmembrane domain of the protein has been modified to encode an equivalent number of leucine residues (Lorent et al., 2017) that does not support sorting at the Golgi into the SMS pathway of HeLa cells (Sundberg et al., 2019). The RUSH system was used to retain these proteins in the ER before release into the secretory pathway and examination of protein localization. Initially, we sought to confirm previous reports that in epithelial cells Sdc1 localizes exclusively to the basolateral plasma membrane, which it traffics to directly from the secretory pathway, without a transcytotic route observed for some basolateral proteins (Miettinen et al., 1994; Maday et al., 2008). Indeed, by confocal microscopy we observed pHl-Sdc1 to localize to the basolateral membrane (Figure 2A), as indicated by the colocalization between Sdc1 and the sodium potassium ATPase (NaK ATPase), a classical basolateral marker protein (Caplan et al., 1986), but not the apical marker GP135 (Ojakian and Schwimmer, 1988). At no time point was Sdc1 observed on the apical membrane; instead it transverses the Golgi and subapical post-Golgi compartments (30–60 min) before delivery to the basolateral membrane (90 min). We also observed that mutant pHl-Sdc1-All(Leu) was delivered to the basolateral membrane, appearing to take a route identicalto that of the wild-type (WT) protein, indicating that the sequence of the transmembrane domain is not responsible for the polarized delivery of Sdc1 (Figure 2B). We confirmed the presence of pHl-Sdc1 and pHl-Sdc1-All(Leu) on the basolateral membrane by cell surface biotinylation of polarized cells, confirming that Sdc1 was chiefly accessible only for biotinylation on the basolateral membrane (Figure 2D; pHl-Sdc1 92 ± 2% and pHl-Sdc1-All(Leu) 100 ± 5% [mean ± SD from two representative experiments]). Moreover, analysis of Sdc1 by immunoblotting revealed an obvious electrophoretic mobility shift associated with the release of the protein from the ER. Before release of Sdc1 from the ER, the protein appears as two distinct bands. At 90 min postrelease, when Sdc1 is present at the plasma membrane, it appears as a “smear” (Figure 2D), likely due to the addition of glycosaminoglycan chains during transport through the early secretory pathway that are essential for Sdc1 function (Eriksson and Spillmann, 2012).

FIGURE 2:

Physical properties of Sdc1 transmembrane domain not required for polarized delivery of Sdc1. Example micrographs of polarized MDCK cells expressing RUSH-pHluorin-Sdc1 (A), RUSH-pHluorin-Sdc1-AllL (B), or RUSH-pHluorin-Sdc1∆YA (C) at 0, 30, 60, and 90 min after the addition of biotin. Both single XY slices and XZ stacks are shown. Cells were fixed before staining with anti-NaK ATPase and anti-GP135 primary antibodies, followed by DyLight 550 (red)– and AlexaFluor 633 (blue)-conjugated secondary antibodies, respectively. Scale bars represent 20 μm. (D) MDCK cells expressing RUSH-pHluorin-Sdc1 or RUSH-pHluorin-Sdc1-AllL were polarized, and RUSH proteins were released from the ER by the addition of biotin. Biotin was added to the apical domain (AP) or the basolateral domain (BL) at 0 or 90 min postrelease. Biotinylated domain-specific surface protein was collected and processed for immunoblotting with an anti-GFP antibody. (E) Cultured MDCK cells expressing RUSH-pHluorin-Sdc1∆YA were collected at 0 or 90 min after release by the addition of biotin. Samples were collected and processed for immunoblotting with an anti-GFP antibody.

Given that determinants required for Sdc1 sorting into the SMS pathway of HeLa cells are dispensable for its polarized distribution in epithelial MDCK cells, we sought to identify factors that are required for trafficking of Sdc1 to the basolateral plasma membrane domain. The C-terminal cytoplasmic segment of Sdc1 contains information for basolateral targeting of the protein (Miettinen et al., 1994; Maday et al., 2008), including a conserved sequence at its C-terminus, Glu-Phe-Tyr-Ala (EFYA), that is present in all syndecan proteins (Sdc1–4) and conforms to a type II PDZ domain–binding motif (Bass and Humphries, 2002; Maday et al., 2008; Cheng et al., 2016). Previously, it was shown that deletion of the two carboxy-terminal residues of Sdc1 (Tyr-Ala) resulted in missorting of Sdc1 to the apical and basolateral membrane of MDCK cells (Miettinen et al., 1994; Maday et al., 2008), implicating PDZ motif recognition in basolateral targeting of Sdc1. We confirmed these results, transducing MDCK cells with a mutant pHluorin-Sdc1 (Sdc1∆YA) under the control of the RUSH system and following the progress of Sdc1∆YA through the secretory pathway via a RUSH-release time course. For unknown reasons, Sdc1∆YA was released from the ER slowly compared with the full-length Sdc1, evidenced by a clear ER signal that could still be seen at 60 and 90 min postrelease and a lack of clear Golgi signal as the RUSH-release time course progresses (Figure 2C). Nonetheless, as described in previous reports, we observed by confocal microscopy that Sdc1∆YA accumulated on the apical plasma membrane without evidently visiting the basolateral plasma membrane en route (Figure 2C) and Sdc1∆YA in mesenchymal cells undergoes an obvious electrophoretic mobility shift associated with the release of the protein from the ER (Figure 2E) akin to Sdc1 (Figure 2D).

Sdc1 sorting in polarized epithelial cells is dependent on GOPC

What is responsible for this polarized delivery of Sdc1? The C-terminal residues of Sdc1, Sdc2, Sdc3, and Sdc4 are identical and constitute PDZ motifs that are reported to be recognized in vitro by four different proteins containing a PDZ domain: CASK, syntenin, synectin, and synbindin. Depletions of these proteins by a targeted short hairpin RNA (shRNA) in MDCK cells, either alone or in pairs, were unable to recapitulate missorting of Sdc1∆YA to the apical plasma membrane (Maday et al., 2008). This raises the possibility that an unidentified protein(s) containing a PDZ domain may mediate basolateral sorting of Sdc1 at the Golgi. Notably, one Golgi-localized PDZ domain protein that is implicated in the trafficking of numerous disparate ligands through PDZ domain–mediated recognition is GOPC (also known as FIG, CAL, and PIST) (Charest et al., 2001; Neudauer et al., 2001; Yao et al., 2001; Cheng et al., 2002). GOPC is a golgin that has been shown to interact with other golgins, as well as a variety of secreted cargo proteins, so it is a candidate for mediating Sdc1 trafficking.

To determine whether GOPC has a role in Sdc1 trafficking, we utilized a previously validated shRNA sequence (Lu et al., 2015) to deplete MDCK cells of GOPC by lentiviral transduction and selected these cells with puromycin before transducing cells a second time with a pHluorin-Sdc1 RUSH construct, such that MDCK cells were doubly transduced with GOPC shRNA and pHluorin-Sdc1 transgenes. GOPC levels in GOPC shRNA–expressing cells were confirmed by Western blotting (Figure 3A). RUSH-based trafficking experiments were then carried out to observe the trafficking of Sdc1 in these GOPC shRNA–treated cells by cell surface biotinylation and confocal fluorescence microscopy (Figure 3, B and C). Interestingly, at 90 min post–release from the ER we observed Sdc1 to reside in both the , apical and basolateral plasma membrane domains in cells expressing GOPC-shRNA. Plasma membrane localization of Sdc1 was confirmed by cell surface biotinylation of apical and basolateral plasma membrane domains separately. Whereas Sdc1 was not biotinylated while Sdc1 was retained in the ER, biotinylated Sdc1 was detected on both the apical and basolateral membranes at 90 min postrelease in GOPC knockdown (KD) cells (Figure 3B; 39 ± 15% and 60 ± 15% [mean ± SD], respectively, from two representative experiments). In these cells, pHl-Sdc1 could be observed in the apical and basolateral plasma membrane of GOPC KD cells and to colocalize with GP135 and NaK ATPase, apical and basolateral cell surface markers, respectively (Figure 3C).

FIGURE 3:

GOPC is required for basolateral delivery of Sdc1. (A) Immunoblot showing WT MDCK cells or MDCK cells stably expressing GOPC shRNA that were collected and processed for immunoblotting with an anti-GOPC antibody, revealing KD of GOPC. Two nonspecific bands are indicated (ns). (B) Polarized MDCK cells expressing GOPC shRNA and RUSH-pHluorin-Sdc1 were subjected to cell surface biotinylation following release of RUSH-pHluorin-Sdc1 from the ER. Biotin was added to the apical domain (AP) or the basolateral domain (BL) at 0 or 90 min postrelease. Biotinylated domain-specific surface protein was collected and processed for immunoblotting with an anti-GFP antibody. (C) Example micrographs of polarized MDCK cells expressing GOPC shRNA and RUSH-pHluorin-Sdc1 at 0, 45, and 90 min post–release from the ER. Both single XY slices and XZ stacks are shown. Cells were fixed before staining with anti-NaK ATPase and anti-GP135 primary antibodies, followed by DyLight 550 (red)- and AlexaFluor 633 (blue)-conjugated secondary antibodies, respectively. Scale bars represent 20 μm. (D) Micrographs showing the presence of Sdc1 on the apical membrane of MDCK cells following GOPC KD, as visualized by differential secondary antibody addition to the apical or basolateral membrane of nonpermeabilized cells. WT and GOPC shRNA cells expressing RUSH-pHluorin-Sdc1 were polarized before being fixed at 0 and 90 min post–release of pHlourin-Sdc1. After fixation, an anti-GFP antibody that recognizes pHluorin was added to both the apical and basolateral Transwell chambers. Cells were then washed, a DyLight 550 (red) secondary antibody was added to the apical chamber, and an AlexaFluor 633 (blue) secondary antibody was added to the basolateral chamber. Apical, basolateral and total pHlourin-Sdc1 signal can thus be distinguished by red, blue, and green fluorescence, respectively.

We further confirmed these results by immunodetection of the pHluorin moiety of pHl-Sdc1 in the absence of cell permeabilization, taking advantage of the filter-grown MDCK cell system that separates the apical and basolateral chambers. WT and GOPC KD MDCK cells expressing RUSH-pHl-Sdc1 were grown for 5 d before being fixed at 0 and 90 min post–release from the ER. After fixation, an anti-GFP antibody that recognizes pHluorin was added to both the apical and basolateral Transwell chambers simultaneously. Cells were then washed and incubated with two distinctly labeled secondary antibodies such that the pHl-Sdc1 in the apical plasma membrane versus that on the basolateral surface were detected as distinct fluorescence channels. In both WT and GOPC KD cells at 0 min postrelease, pHl-Sdc1 was intracellular (Figure 3D). In WT cells at 90 min postrelease, pHl-Sdc1 was present on the basolateral membrane, indicated by the presence of signal from both AlexaFluor 633, which was added to the basolateral membrane, and pHl-Sdc1 (Figure 3D). No DyLight 550 signal was present in the apical membrane, indicating that no pHl-Sdc1 was present. In the GOPC-depleted cells at the same time point, pHl-Sdc1 was observed in both the apical and basolateral membranes of cells, DyLight 550 signal was visible apically, and AlexaFluor 633 labeling was observed in the basolateral membrane (Figure 3D).

These data revealed that the correct basolateral targeting of Sdc1 is dependent on GOPC. Like other PDZ-domain proteins, GOPC has a proposed scaffolding role in cells, potentially bridging together different effectors through its multiple protein-binding domains. The coiled-coil region of GOPC is reported to confer Golgi localization and to bind syntaxin-6, golgin-160, TC10, and Stargazin (Charest et al., 2001; Neudauer et al., 2001; Cuadra et al., 2004; Hicks and Machamer, 2005; Huttlin et al., 2017, 2021), whereas the PDZ domain has been implicated in the trafficking of numerous integral membrane proteins, ion channels, pumps, and adhesion molecules (Yao et al., 2001; Cheng et al., 2002; Gentzsch et al., 2003; Xu et al., 2010; Nie et al., 2016). We sought to determine which domains of GOPC are required for the basolateral sorting of Sdc1 in polarized MDCK cells. To do this we constructed shRNA-resistant forms of GOPC that lack its putative coiled-coil motif (termed “∆CC-GOPC-EGFP”) or two missense mutations of conserved amino acid residues in the PDZ motif binding pocket that in other PDZ domains are essential for recognition of C-terminal PDZ motifs (termed “PDZ mut-GOPC-EGFP”) (Songyang et al., 1997; Li et al., 2006). A shRNA-resistant form of WT GOPC-EGFP (termed “WT GOPC-EGFP”) served as a control. GOPC mutants lacking the coiled-coil domains were no longer Golgi localized, as has been observed previously, whereas PDZ mutant GOPC was still present at the Golgi but no longer had the same continuous Golgi ribbon localization, instead appearing more particulate (Figure 4, A and B).

FIGURE 4:

The PDZ domain of GOPC is required for Sdc1 sorting. (A) MDCK cells expressing GOPC shRNA and RUSH-Ruby-Sdc1 were complemented with either GOPC-WT-EGFP or GOPC-PDZm-EGFP and polarized. Cells were fixed at 0 and 90 min post–release of Ruby-Sdc1 from the ER. Cells were stained with anti-NaK ATPase and AlexaFluor 633 (blue)-conjugated secondary antibody. Images show representative fields of view as XY slices and XZ stacks. (B) MDCK cells expressing GOPC-∆CC-EGFP. Cells were fixed before being stained with anti-NaK ATPase and anti-GP135 primary antibodies, followed by DyLight 550 (red)- and AlexaFluor 633 (blue)-conjugated secondary antibodies, respectively. Images show a max projection of the mid three z slices of polarized MDCK cells to indicate lack of GOPC-∆CC-EGFP signal within the Golgi region. Scale bars represent 20 μm.

We then examined trafficking of Ruby-Sdc1 in complemented MDCK epithelial cells using RUSH trafficking assays. In cells that had been complemented with WT GOPC-EGFP, Ruby-Sdc1 signal was restricted to the basolateral plasma membrane following Ruby-Sdc1 release, confirming that WT GOPC-EGFP is functional when expressed in this manner. In contrast, in cells complemented by PDZ mut-GOPC-EGFP (Figure 4A), Ruby-Sdc1 localized to the basolateral and apical plasma membrane domains 90 min postrelease. This was similar to its localization in GOPC-depleted cells, although pHl-Sdc1 fluorescence is easier to visualize. These results establish that the GOPC coiled-coil motif and a functional PDZ domain are required for GOPC function in basolateral targeting of fluorescent Sdc1 fusion proteins.

Golgi localization of GOPC is sensitive to brefeldin A

One Golgi-localized protein previously shown to bind to GOPC is golgin-160 (Hicks and Machamer, 2005; Huttlin et al., 2017, 2021), a protein reported to associate with the motor protein dynein and through this interaction to control Golgi positioning in cells (Yadav et al., 2012). The coiled-coil domain of GOPC has been reported to mediate the interaction with golgin-160 (Hicks and Machamer, 2005). Along with a direct interaction between the proteins, both GOPC and golgin-160 have been implicated in the cell surface delivery of the β1-adrenergic receptor (Hicks et al., 2006) and are implicated in acrosome biogenesis during spermatogenesis (Yao et al., 2002; Bentson et al., 2013). We therefore considered whether the role of GOPC in Sdc1 sorting could be attributed to an interaction with golgin-160. Initially, we sought to determine whether golgin-160 was required for GOPC targeting to the Golgi, as this dependence has not been shown previously, although golgin-160 is reported to bind the GOPC coiled-coil domain (essential for Golgi targeting of the GOPC) (Figure 4B). For this purpose, we used transiently transfected HeLa cells due to the availability of reagents to detect human golgin-160.

In golgin-160 KD cells, there is an obvious lack of golgin-160 signal in the perinuclear region in comparison to control cells, and anti–golgin-160 immunoblotting of lysates from small interfering (siRNA)-treated cells indicated a near complete loss of detectable golgin-160 protein (Supplemental Figure 1, A and B). In these siRNA-treated cells, GOPC is clearly observed at the Golgi, suggesting that golgin-160 is not required for GOPC Golgi localization. Nevertheless, we could not rule out the possibility that residual golgin-160 is sufficient for Golgi localization of GOPC and so we utilized golgin-160 knockout (KO) HeLa cells to eliminate this possibility (Figure 5B). Indeed, in the golgin-160 KO cells, GOPC was still observed to be present at the Golgi, indicating that golgin-160 is not responsible for GOPC localization at the Golgi (Figure 5A). Therefore, it is not clear how GOPC is targeted to the Golgi, as both golgin-160 (this study) and syntaxin-6 KD had no influence on its localization (Cheng et al., 2010a).

FIGURE 5:

GOPC localization is influenced by cell cycle and ARF GTPase signaling but not golgin-160. (A) Representative images showing endogenous golgin-160 and GOPC in control and golgin-160 CRISPR KO HeLa cells. Cells were cultured for 24 h before fixation, permeabilization, and staining with an anti–golgin-160 or anti-GOPC antibody (as indicated) and a DyLight 550–conjugated secondary antibody and anti-GM130 and AlexaFluor 488–conjugated secondary antibody. Nuclei were identified by staining with 4’-6-diamidino-2-phenylindole (DAPI). Imaged by deconvolution microscopy. (B) Immunoblot showing WT or golgin-160 CRISPR KO HeLa cells that were collected and processed for immunoblotting with an anti–golgin-160 antibody. (C) Representative images showing endogenous golgin-160 and GOPC in cells undergoing mitosis. Cultured HeLa cells were fixed and permeabilized before being stained with an anti–golgin-160 or anti-GOPC antibody (as indicated) and a DyLight 550–conjugated secondary antibody and anti-GM130 and an AlexaFluor 488–conjugated secondary antibody. DAPI was used to stain the nucleus. Cells with condensed nuclei, indicated by DAPI staining, are outlined with white dashes. During mitosis, close association between GM130 and golgin-160 or GOPC is lost, and localization becomes distinct. Imaged by deconvolution microscopy. (D) Representative images showing endogenous golgin-160 and GOPC in control or brefeldin A (BFA)-treated cells. Cultured HEK cells were treated with either mock vehicular control or 1 µg/ml BFA for 5 min before fixation. Fixed cells were permeabilized before being stained with an anti–golgin-160 or anti-GOPC antibody (as indicated) and a DyLight 550–conjugated secondary antibody and anti-GM130 and an AlexaFluor 488–conjugated secondary antibody. Nuclei were stained with DAPI. Short BFA incubation times do not grossly disrupt Golgi morphology but reveal proteins that are dependent on Arf for localization. Imaged by deconvolution microscopy. Scale bars represent 20 μm.

In the course of these experiments, we noted that in cells with condensed nuclei indicative of cells in mitosis, both GOPC and golgin-160 are mainly cytosolic and remain distinct from punctate Golgi fragments, identified by GM130 staining (Figure 5C). Previously it was shown that localization of golgin-160 is disrupted by brefeldin A (BFA), an inhibitor of guanine nucleotide exchange on some Arf GTPases, suggesting that targeting of golgin-160 to the Golgi apparatus requires GTP-bound Arf GTPases (Yadav et al., 2012). Similarly, incubation of cells with BFA resulted in the dispersion of GOPC throughout the cytosol, indicating that targeting of GOPC to the Golgi requires Arf^GTP (Figure 5D [Yao et al., 2001]).

The results of these experiments and what is known regarding golgin-160 and GOPC suggested that the proteins may function together or in adjacent pathways to mediate their cellular roles. We next tested whether golgin-160 directly influenced Sdc1 trafficking. To do this, shRNA sequences were designed against the Canis lupus familiaris gene encoding golgin-160 (NCBI Taxonomy ID 9615), inserted into the pLKO vector, and ultimately used to transduce MDCK cells expressing RUSH-pHl-Sdc1. Golgin-160 KD was confirmed by immunoblotting (Supplemental Figure 2A). In these cells, RUSH-pHl-Sdc1 was observed to be trafficked to the basolateral plasma membrane, identically to WT MDCK cells (Supplemental Figure 2B). Therefore, the roles of these proteins in the context of Sdc1 sorting appear to be unrelated; we cannot rule out the possibility that residual golgin-160 is sufficient for Sdc1 sorting in MDCK cells. Consequently, mechanistic insights into how KD of Golgi-localized GOPC results in Sdc1 missorting to the basolateral plasma membrane remain outstanding.

GOPC overexpression causes Golgi compaction

Through the course of our investigations, we revealed unexpected changes in the appearance of the Golgi associated with GOPC-EGFP overexpression that are uniquely visualized within the physical constraints of the Golgi in polarized cells. In polarized epithelial cells expressing GOPC-EGFP, the Golgi, visualized with the cis Golgi marker GM130, appears to “bloat” and condense, becoming more spherical compared with the convoluted ribbon morphology observed in WT cells (Figure 6A). These morphological effects on global Golgi architecture were quantified, using GM130 fluorescence signals to create three-dimensional (3D) surface renderings of the Golgi in these cells, which allowed measurement of surface area and volume of this 3D structure (Figure 6B). The surface areas of the Golgi in WT and GOPC-EGFP–expressing cells were essentially identical (Figure 6C), while a small increase in Golgi volume was noted in the GOPC-EGFP– and GOPC-PDZm-EGFP–expressing cells (Figure 6D). However, GOPC-EGFP–expressing cells had an almost 50% decrease in surface area:volume ratio compared with the WT cells (Figure 6E). These results suggest that elevated expression of GOPC-GFP results in swelling of Golgi cisternae. These effects were only partially reliant on the PDZ domain, as GOPC-PDZm-EGFP–expressing cells caused a modest decrease in surface area:volume ratio of the Golgi in comparison to WT GOPC (Figure 6E). Interestingly, observing the localization of the WT and PDZm GOPC in comparison to the WT, the PDZm shows less continuous Golgi localization, appearing fragmented or in small domains along the ribbon (Figures 4A and 6, A and B). Along with the fact that deletions in the coiled-coil of the protein no longer support GOPC Golgi localization, this suggests that these effects on Golgi morphology likely require the full-length WT protein.

FIGURE 6:

GOPC overexpression induces changes in Golgi morphology. (A) Representative micrographs from WT MDCK cells or MDCK cells expressing GOPC-EGFP or GOPC-PDZm-EGFP showing Golgi morphology. Cells were polarized and fixed before being stained with an anti-GM130 primary antibody and an AlexaFluor 546–conjugated secondary antibody. Scale bar represents 50 μm. (B) 3D surfaces of GM130 staining generated from WT or GOPC-EGFP– or GOPC-PDZm-EGFP–expressing MDCK cells using IMARIS software. Scale bar represents 20 μm. Analysis of Golgi surface area (C), volume (D), and the surface area:volume ratio (E) of cells expressing GOPC-EGFP or GOPC-PDZm-EGFP. Cells expressing GOPC-EGFP or GOPC-PDZm-EGFP were polarized and fixed before being stained as in A. GM130 signal was then analyzed using IMARIS, rendering GM130 as a 3D surface for measurement. Quantification shows the average ± SD and individual values obtained for 50 Golgi per condition. ** p ≤ 0.01, **** p ≤ 0.0001.

The effects of GOPC overexpression on Golgi morphology are not readily explained solely by the retention of integral membrane protein secretory cargo in Golgi cisternae, in part because overexpression of GOPC does not result in retention of Sdc1 in the Golgi. Morphological changes in the Golgi have been reported with the use of various pharmacological agents. For example, ultrastructural analysis of Golgi morphology in cycloheximide-treated cells (clearing the Golgi of secretory cargoes) shows Golgi stacks that are more tightly packed and circular (Taylor et al., 1997). Moreover, pharmacological agents that disrupt the cytoskeleton such as toxins that depolymerize the actin cytoskeleton result in the formation of compact Golgi (Lázaro-Diéguez et al., 2006), whereas depolymerization of microtubules leads to Golgi membrane fragmentation and dispersal (Cole et al., 1996). It may be that an imbalance between GOPC effectors induced by overexpression of GOPC results in changes to cytoskeleton-Golgi dynamics. Furthermore, both the actin- and microtubule-based cytoskeletons are implicated in apical and/or basolateral targeting of certain cargoes, although mechanistic insights into these differences have not been elucidated (Cao et al., 2012).

In this study, we show that basolateral targeting of Sdc1 in epithelial MDCK cells is independent of the sequence of its single transmembrane domain. In HeLa cells, previous studies have shown that the Sdc1 transmembrane domain confers sorting within SM-rich vesicles in a SMS pathway. Instead, we show that GOPC, a golgin family protein that is associated with trans and TGN Golgi compartments, plays a key role in sorting of newly synthesized Sdc1 from the Golgi apparatus to the basolateral plasma membrane of MDCK cells. Syndecans are enriched in discrete features of the plasma membrane, such as intercellular junctions, and this has been attributed to retention by PDZ domain–containing scaffolding proteins, which localize to these sites (Cheng et al., 2016; Afratis et al., 2017). Although GOPC is reported to retain numerous integral membrane secretory cargoes in the Golgi via recognition of their PDZ motifs (Yao et al., 2001; Cheng et al., 2002; Gentzsch et al., 2003; Xu et al., 2010; Nie et al., 2016), several observations suggest that it is unlikely that GOPC exerts its effects on Sdc1 basolateral sorting via recognition of the C-terminal PDZ motif of Sdc1 in the Golgi apparatus. In RUSH format Sdc1 trafficking experiments, we did not observe an accumulation of Sdc1 in Golgi, which has been reported for some other PDZ motif–containing integral membrane proteins when GOPC is overexpressed (Cheng et al., 2002; He et al., 2004; Wente et al., 2005; Nie et al., 2016). Second, unbiased selections of sequenced randomized C-terminal PDZ motifs by the GOPC PDZ domain in vitro identified the sequence Thr-Ser-Ile-Ile as optimal (i.e., binds with the highest affinity of all tested sequences) (Vouilleme et al., 2010). This sequence conforms to a type I PDZ motif, whereas the Sdc1 PDZ motif (EFYA) conforms to a type II motif. The atomic structure of a Sdc1 PDZ motif peptide bound to the PDZ domain of TIAM1 (Liu et al., 2013), a PDZ domain–containing component of intercellular tight junctions, identified amino acid residues in the PDZ motif–binding pocket that confer specificity for the Sdc1 PDZ motif, and these residues are not present in the GOPC PDZ motif–binding pocket. We conclude that the requirement for the GOPC PDZ domain in Sdc1 sorting is not mediated by recognition of Sdc1 by GOPC. Nevertheless, our data and those of Maday and colleagues (Maday et al., 2008) unequivocally reveal the importance of the PDZ-binding motif for correct basolateral targeting of Sdc1. While we and others have not identified the PDZ domain protein(s) responsible, there are likely additional Sdc1-PDZ-motif–binding-partner(s) that remain to be identified that likely participate in sorting and retention of Sdc1 at the basolateral plasma membrane. By virtue of the GOPC-PDZm remaining Golgi localized and also the fact that neither syntaxin-6 (Cheng et al., 2010a) nor golgin-160 (this study) is responsible for GOPC Golgi localization, it remains to be determined how GOPC maintains its Golgi localization. We consider it likely that GOPC domains recognize an as-yet-unidentified component(s) of the Golgi matrix that influence polarized sorting. The roles of GOPC in this context are independent of golgin-160, although Arf activity is required for GOPC localization. Loss of this interaction(s) likely causes pleiotropic effects on Golgi function, resulting in missorting of Sdc1. Consistent with this, depletion of GOPC from HeLa cells results in a decrease in cellular levels of sphingolipids, to a level similar to that of the sphingolipid transfer protein FAPP2 (Pothukuchi et al., 2021). Moreover, depletion of GRASP55, another golgin containing a PDZ domain, results in increased sphingolipid levels, suggesting global golgin-mediated influences on sphingolipid metabolism within the Golgi. The effects of GOPC depletion on sphingolipid homeostasis and protein sorting could be indirect due to disruption of the overall integrity of numerous Golgi functions, resulting in impaired trafficking of lipid-modifying enzymes and other cargoes. Identifying the ligand(s) of the GOPC PDZ domain should be key for elucidating the role of GOPC in the golgin network.

MATERIALS AND METHODS

Antibodies

The rabbit anti–sodium potassium ATPase (clone EP1845Y; ab76020) monoclonal antibody was purchased from Abcam. The mouse anti-podocalyxin (GP135) (clone 3F2:D8; MABS1327) monoclonal antibody was purchased from EMD Millipore. The rabbit anti-GFP (PABG1) polyclonal antibody was obtained from Chromotek. The rabbit anti-GOPC (A13436) polyclonal antibody was purchased from ABclonal. The rabbit anti–golgin-160 antibody (PA5-59041) polyclonal antibody was purchased from Invitrogen. Alexa- and DyLight-labeled secondary antibodies were purchased from Invitrogen. Horseradish peroxidase–conjugated secondary antibodies were purchased from Cell Signaling.

Cell lines

MDCK II, HEK293T, and HeLa cells were cultured in DMEM (high glucose) supplemented with 10% fetal bovine serum (FBS) (Life Technologies, Invitrogen). All cell lines were maintained at 37°C in a 5% CO₂ incubator and routinely checked for mycoplasma contamination using the Universal Mycoplasma Detection Kit (American Type Culture Collection).

For RUSH assays, MDCK cells expressing RUSH vectors were plated onto 12-mm Transwell filter supports (0.4-μm pore size; Corning Life Sciences) at a density of 1.75 × 10⁵ cells per well and cultured for 5 d in the presence of avidin (1 mg/ml). On day five, RUSH assays were performed. Cells were washed 2× with phosphate-buffered saline (PBS) before incubation with DMEM 10% FBS containing 400 μM d-biotin (Sigma) for the indicated time points.

Constructs for lentiviral transduction

To generate the RUSH-pHluorin-Sdc1, RUSH-pHluorin-Sdc1-AllL, and RUSH-Ruby-Sdc1 pLenti plasmids for lentiviral production, previously described constructs (Str-KDEL-stp-ss-pHluorin-Sdc1, Str-KDEL-stp-ss-pHluorin-Sdc1-AllL, and Str-KDEL-stp-ss-Ruby-Sdc1) (Sundberg et al., 2019) were used as a templates for PCR, and constructs were cloned using the Gateway (Thermo Fisher) system according to the manufacturer instructions: first into the pDONR221 donor vector and then a pLenti CMV destination vector. To generate the RUSH-pHluorin-Sdc1ΔYA truncation mutant, a stop codon was placed after the C-terminal phenylalanine at amino acid position 309.

GOPC constructs were generated by PCR using human GOPC cDNA as a template (GenBank accession no. NM_001017408). To generate the GOPC-∆CC mutant, amino acids (86–211) corresponding to the coiled-coil regions of the protein were removed by overlap extension PCR. To generate the GOPC-PDZm containing three amino acid residues in the carboxylate-binding region were introduced: codon Gly290 “GCC” was mutated to Ala “GCA,” codon Leu291 “CTT” was mutated to Ala “GCA,” and codon LY292 “GGC” was mutated to Ala “GCA.”

pLKO GOPC (shRNA: CTGGAGAAGGAGTTCGACAAA) was obtained from the Broad Institute, Harvard Medical School. KD was confirmed and quantified by immunoblotting using anti-GOPC antibody. GOLGA3 (golgin-160) shRNA (pLV shRNA-Puro-U6-Golga3 GCAGAAGAGCAGTACCAAAGA) was designed against the C. lupus familiaris genome and assembled by and purchased from VectorBuilder.

Lentiviral production and transduction

HEK293T cells were cotransfected with the pLenti/pLKO vector, along with a vector expressing the VSV-G envelope protein and the viral packaging vector pCMV ∆8.91 and incubated at 37°C in a 5% CO₂ incubator. Twenty-four hours posttransfection the cell medium was replaced. Medium containing recombinant lentivirus was collected 48 and 72 h posttransfection, pooled, and passed through a 0.45-μm filter unit. MDCK cells were infected with recombinant lentivirus–containing medium, supplemented with 6 µg/ml polybrene (Sigma). Twenty-four hours postinfection the medium was changed, and 48 h postinfection stable cell lines were selected with the appropriate antibiotics.

Cell surface biotinylation

Cells were placed on ice and washed 3× with ice-cold PBS⁺⁺ (containing Mg²⁺ and Ca²⁺). Either the apical or basolateral domain was then biotinylated using a solution of 250 ng/ml EZ-Link-Sulfo-NHS-LC-LC-Biotin (Thermo Fisher) diluted in PBS⁺⁺. The biotinylation reaction was allowed to proceed for 20 min in the dark on ice with rocking. Cells were then washed 2× with PBS⁺⁺ and incubated in PBS⁺⁺ containing 50 mM NH₄Cl for 20 min in the dark on ice with rocking. Filters were washed 3× with PBS⁺⁺, and then cells were harvested in 1 ml lysis buffer (1% Triton X-100, 0.1% SDS, 20 mM glycine, 1× complete protease inhibitor cocktail tablet with EDTA [Roche Diagnostics] and PBS). Cells were scraped into lysis buffer and incubated for 30 min on ice with occasional vortexing. The samples were then spun 18,000 × g for 25 min at 4°C. After centrifugation, 900 µl of the supernatant was transferred to a 35 μl bed volume of NeutrAvidin beads (Thermo Fisher). The beads were rotated end over end overnight at 4°C. Samples were then washed 1× in lysis buffer and 2× in PBS. Samples were then prepared for Western blotting and detection with anti-GFP antibody.

Immunoblotting

Samples were incubated in SDS–PAGE and incubated at 96°C for 10 min before being loaded onto an SDS–PAGE gel. In some cases, proteins were first precipitated using 10% trichloroacetic acid. Proteins were transferred to nitrocellulose membrane, which was then blocked in PBS-T containing 5% milk. After immunoblotting, proteins were detected by chemiluminescence. Images of immunoblots were acquired by the ChemiDoc Imaging System (Bio-Rad). For the cell surface biotinylation experiments, 3% input, 3% output, and 50% IP are loaded in the gel.

Fluorescence microscopy

MDCK cells were plated onto 12-mm Transwell filter supports (0.4-μm pore size; Corning Life Sciences) at a density of 1.75 × 10⁵ cells per well and cultured for 5 d. On the fifth day, cells were analyzed and/or RUSH assays were carried out. HeLa and HEK cells were plated in Mattek 35 mm glass-bottom (no. 1.5) microwells at a density of 5 × 10⁵ cells per dish. Cells were fixed in 4% paraformaldehyde for 20 min at room temperature before neutralization with 50 mM NH₄Cl in PBS for 20 min. Following fixation, cells were permeabilized in 0.05% Triton X-100 for 10 min and washed 2× in PBS. Cells were then incubated with primary antibodies diluted in PBS/5% (wt/vol) bovine serum albumin (BSA) for 2 h. Cells were then washed 2× in PBS and incubated with secondary antibodies (AlexaFluor 488 and 633 and DyLight550) diluted in PBS/5% (wt/vol) BSA for 1 h. If staining with DAPI, cells were washed again and incubated with DAPI diluted in PBS/5% (wt/vol) BSA for 20 min. Cells were then mounted with Mowiol (Sigma) and observed using a confocal microscope (TCS Sp5 AOBS; Leica) or deconvolution microscope (DeltaVision Elite; Applied Precision). Unless otherwise specified, all microscopy images were confocal. For deconvolution microscopy, image stacks were collected at 0.33 mm z increments using a 60×, 1.4 NA oil immersion lens. Images were captured with a sCMOS camera (CoolSnap HQ; Photometrics) and deconvolved with softWoRx (v.6.0) software using the iterative-constrained algorithm and the measured point spread function. For confocal microscopy, image stacks were collected with 0.33 µm z increments with an oil-immersion objective (63×, 1.4 NA; Leica). Laser excitation wavelengths of 488, 568, and/or 633 nm were used for sequential mode two- or three-channel imaging. Single z slices in polarized cells were chosen to highlight the organelle(s) of interest; otherwise approximate midsection z slices were chosen unless otherwise specified. Images were processed using Adobe Photoshop software (Adobe Systems), IMARIS (Oxford Instruments), or FIJI/ImageJ (Schindelin et al., 2012).

Abbreviations used:

AP

apical plasma membrane domain

BL

basolateral plasma membrane domain

DAPI

4’-6-diamidino-2-phenylindole

EGFP

enhanced green fluorescent protein

ER

endoplasmic reticulum

FBS

fetal bovine serum

GAPDH

glyceralde-hyde-3-phosphate dehydrogenase

KD

knockdown

KO

knockout

MDCK

Madin-Darby canine kidney

NaK ATPase

Na ²⁺/K ⁺-ATPase

PBS

phosphate buffered saline

PDZ

structural domain shared by the postsynaptic density protein (PSD95), Drosophila disc large tumor suppressor (DlgA), and zonula occludens-1 protein (ZO-1) proteins

pHl

phlourin green fluorescent protein

RUSH

retention using selective hooks

shRNA

short hairpin RNA

SM

sphingomyelin

TGN

trans-Golgi network.