Serum- and Glucocorticoid-Inducible Kinase-1 (SGK-1) Plays a Role in Membrane Trafficking in Caenorhabditis elegans

The mammalian serum- and glucocorticoid-inducible kinase SGK1 regulates the endocytosis of ion channels. Here we report that in C. elegans sgk-1 null mutants, GFP-tagged MIG-14/Wntless, the sorting receptor of Wnt, failed to localize to the basolateral membrane of intestinal cells; instead, it was mis-sorted to lysosomes. This effect can be explained in part by altered sphingolipid levels, because reducing glucosylceramide biosynthesis restored the localization of MIG-14::GFP. Membrane traffic was not perturbed in general, as no obvious morphological defects were detected for early endosomes, the Golgi apparatus, and the endoplasmic reticulum (ER) in sgk-1 null animals. The recycling of MIG-14/Wntless through the Golgi might be partially responsible for the observed phenotype because the subcellular distribution of two plasma membrane cargoes that do not recycle through the trans-Golgi network (TGN) was affected to a lesser degree. Consistently, knockdown of the ArfGEF gbf-1 altered the distribution of SGK-1 at the basolateral membrane of intestinal cells. In addition, we found that sgk-1(RNAi) induced unfolded protein response in the ER, suggesting at least an indirect role of SGK-1 early in the secretory pathway. We propose that SGK-1 function is required for lipid homeostasis and that it acts at different intracellular trafficking steps.

How SGK1 regulates these ion channels and transporters is mostly unknown. Recently, it was suggested that SGK1 is involved in the endocytosis of membrane proteins [16]. Cystic fibrosis transmembrane conductance regulator (CFTR) is a chloride channel residing on the apical plasma membrane (PM) of epithelial cells [17]. Curiously, while SGK1 inhibits the endocytosis of CFTR in human airway epithelial cells, it promotes the endocytosis of the epidermal growth factor receptor, which is also an apical plasma membrane protein [16]. Thus mammalian SGK1 may be involved in differentially regulating endocytosis of plasma membrane proteins.
Endocytosis is a key process by which cells internalize molecules [18]. Through receptormediated endocytosis, the major route in most cells, plasma membrane proteins and lipids are internalized in clathrin-coated vesicles and delivered to various destinations [19]. Once endocytosed, different cargoes are sorted in the early endosomes [20]: ligands typically enter the degradative pathway while their membrane receptors are often recycled back to the plasma membrane [21]. There are three different routes in which membrane receptors are recycled back to the plasma membrane: directly from sorting endosome through the tubular membrane structures (fast recycling), from the sorting endosome to the recycling endosomes or endosomal recycling compartments (ERC) (slow recycling), or through retrograde transport to the trans-Golgi network (TGN) followed by re-export to the plasma membrane [22][23][24].
In yeast, Ypk1-a homologue of SGK-1-activates serine palmitoyl-CoA acyltransferase (SPT) and promotes the biosynthesis of ceramide and sphingolipid [25,26]. Ceramide is synthesized at ER and transported to the Golgi for conversion to sphingomyelin (SM) [27]. Ceramide is an important structural element of cell membranes and SM is one of the major lipid species in the lipid bilayer. Disruption of the biosynthesis of ceramide affects membrane trafficking [28].

C. elegans strains
Strains of C. elegans were cultured and maintained using standard protocols. The following strains or alleles were used: the wild-type N2, pwIs765 (P vha-6 MIG-14::GFP), qxIs194 (P tat-SGK-1, using a 7.3 kb genomic sequence containing the a isoform of sgk-1 (5.1 kb) and a 2.2 kb upstream regulatory sequence. The sequence at the far 5' end is: CTCCGGTAACTTACTC ATTTTCAAC and the sequence at far 3' end is: CGTCGACACCAATCGCGTTTTGGTC. BR3063 was integrated by gamma-irradiation and renamed as hqIs150 (P sgk-1 SGK-1::GFP). hqIs150 was backcrossed to N2 for 3 times and assigned the strain name MQD862.

Microscopy and imaging analysis
Differential interference contrast (DIC) and fluorescent images were captured with a Zeiss AxioImager M1 equipped with an AxioCam monochrome digital camera. A 10 × air Plan-Neofluar objective was used for detection of P hsp-4 GFP. For confocal images, a Zeiss LSM 510 Meta inverted confocal microscope with 488 nm and 543 nm lasers was used, and images were processed and viewed using LSM Image Browser software and ZEN lite 2012 (Carl Zeiss). For SGK-1::GFP upon gbf-1 RNAi, animals were mounted on 2% agarose pads in a drop of M9 containing 10 mM Levamisole, covered with a vaseline-rimmed cover slip and imaged with a spinning-disk confocal system Andor Revolution (Andor Technologies, Belfast, Northern Ireland) mounted onto an IX-81 inverted microscope (Olympus, Center Valley, PA) equipped with an iXon EM + electron-multiplying charge-coupled device camera (Andor Technologies). Specimens were imaged using a 63×/1.42 numerical aperture oil objective. Each pixel represents 0.107 μm. Excitation was achieved using solid-state 488 nm laser. Exposure time was 100 ms.
Quantitation of co-localization of mCHERRY::RAB-7 and MIG-14::GFP, the fluorescence intensity of MIG-14::GFP, hTAC::GFP, hsp-4pr::GFP and SGK-1::GFP with gbf-1 RNAi mCHERRY::RAB-7 positive ring-like vesicles containing 0, 1 or ! 2 MIG-14::GFP puncta were counted for six N2 worms and ten sgk-1(ok538) mutant worms, within a 3,036 μm 2 area in the intestine of each worm. The cytoplasmic MIG-14::GFP was quantified by determining the average pixel intensity in non-overlapping 80 μm 2 areas (8 μm × 10 μm) in the intestine (4 areas per worm, 8 worms of N2, 10 worms of sgk-1(mg455) and 13 worms of sgk-1(ok538)). The P value was calculated using Wilcoxon rank sum test. The cytoplasmic accumulation of hTAC:: GFP was quantified by determining the average pixel intensity within 32 non-overlapping 80 μm 2 areas (8 μm × 10 μm) in the intestine (4 areas per worm; 8 worms). The intensities of hTAC::GFP on the basal membrane and the lateral membrane were each measured as an average over 32 non-overlapping 35 μm 2 areas (2.86 μm × 12.27 μm) in the intestine (4 areas per worm; 8 worms). The average pixel intensity per unit area was determined using Image J 1.46r software as described before [33]. The P value was calculated using Wilcoxon rank sum test. The expression of the hsp-4 promoter driven GFP was quantified using Image J 1.46r software as follows. Each image containing 15 worms was taken using the same exposure time (1,000 ms). The background was subtracted by deleting pixels that are nearly completely black (grey values = 0, 1, or 2). Then, the average intensity of an entire image was measured. The results of 10 independent images are shown for each condition (a total of 150 worms). Z-stack images of SGK-1::GFP upon gbf-1 RNAi were compressed in ImageJ version 1.48 (Wayne Rasband, National Institutes of Health, USA) using the average projection tool, and total fluorescence intensity and area was measured on the intestine that was outlined by the polygon selection tool.

RNAi
RNAi was performed as described [34]. For RNAi feeding experiments, the full-length cDNA of sgk-1 was cloned into the empty vector L4440, and the resulting plasmid pZM79 was used to transform the E. coli strain HT115. The cgt-1, cgt-3, gbf-1 and mon-2 RNAi bacterial strains were obtained from the Ahringer RNAi library. The arl-1, cgt-1, cgt-3 and unc-11 RNAi bacterial strains were obtained from the RCE1181 RNAi library. NGM plates containing 1 mM IPTG were inoculated with RNAi bacteria and induced for 12 hours at room temperature. Eggs were cultured at 20°C until L4. Bacteria transformed with the empty RNAi vector served as control for feeding. RNAi started from 5-10 adult worms for 3-4 days and the L4 larvae of the next generation were imaged. The only exception was the gbf-1 RNAi experiment for SGK-1:: GFP, in which RNAi started from L3 larvae for 3 days and the adult worms were imaged.

Immunoprecipitation-Mass Spectrometry (IP-MS) analysis
Chromosomally integrated transgenic worms expressing SGK-1::GFP were cultured on highgrowth (HG) plates. Lysates were made from 2 ml of packed, unsynchronized worms using FastPrep-24 (MP Biomedicals) in lysis buffer (20 mM Tris-HCl pH 8.0, 150 mM NaCl, 0.1% NP40, 2 mM EDTA, 1 × cocktail of protease inhibitors from Roche (Complete, EDTA free)). Proteins bound to the GBP (GFP binding protein) beads (ChromoTek) were eluted with 60 μl of 0.1 M glycine-HCl, pH 2.6, and neutralized with 10 μl of 1 M Tris-HCl, pH 8.0. The proteins were precipitated with 4 × volumes of cold acetone and then re-dissolved in 8 M urea, 100 mM Tris-HCl, pH 8.5. After reduction with 5 mM TCEP and then alkylation with 10 mM iodoacetamide, the samples were diluted fourfold to 2 M urea, 1 mM CaCl 2 , 20 mM methylamine, 100 mM Tris-HCl, pH 8.5, and digested with 1 μg of trypsin at 37°C overnight. The MS analysis was done in duplicates (two technical repeats). For each technical repeat, one fourth of the resulting peptides were pressure-loaded onto a fused silica capillary column packed with 5 μm Luna C18 material (RP, Phenomenex, Ventura, CA), with a Kasil frit at the end. The column was washed with a buffer containing 95% water, 5% acetonitrile, and 0.1% formic acid. After desalting, a 9 cm 100 μm i.d. capillary with a 5 μm pulled tip packed with 5 μm Luna C18 material was attached to the two-phase column with a union. The peptides were separated over a 2-h reverse phase gradient generated by an Agilent 1100 quaternary HPLC (Agilent) and sprayed directly into an LTQ Orbitrap mass spectrometer (ThermoFisher Scientific) with the application of a distal 2.5 kV spray voltage. The gradient was as follows: 5 min from 100% buffer A (5% acetonitrile, 0.1% formic acid) to 5% buffer B (80% acetonitrile, 0.1% formic acid), then increasing to 30% buffer B over 75 min, further to 80% buffer B in 10 min, to 100% buffer B in 10 min, and lastly, a 20-min 100% buffer B wash. The flow rate was maintained at 0.1 ml/min and the flow was split using a microTee that was connected, at one of its openings, to an empty 50 μm i.d. capillary. The length of the empty capillary was adjusted so that a flow rate of about 200 nl/min was achieved at the tip of the column. The mass spectrometer was operated in the data-dependent mode. Survey MS scans were acquired in the orbitrap with the resolution set to a value of 60000. Each survey scan (400−2000 m/z) was followed by 8 datadependent tandem mass (MS/MS) scans in the linear ion trap at 35% normalized collision energy. AGC target values were 200,000 for the survey scan and 10,000 for the MS/MS scan. Target ions already selected for MS/MS were dynamically excluded for 30 seconds. Proteins were identified by searching the MS/MS spectra against a C. elegans protein database using Prolucid [35] and filtering the search results with DTASelect 2.0 [36] at 0.01 false discovery rate (FDR) at the spectral level, 5 ppm precursor mass accuracy, and a minimal Z score of 4.0. The overall false discovery rate for protein identification is less than 0.5%. Using the control IP-MS results of seventeen transgenic strains expressing unrelated GFP fusion proteins, identified proteins were ranked by WD scores calculated as described previously [37]. WD scores help downgrade common contaminant proteins and enrich for specific binding proteins [37]. The RAW data of SGK-1::GFP IP-MS experiments has been deposited to the ProteomeXchange Consortium [38] via the PRIDE partner repository with the dataset identifier PXD002190.

Yeast two-hybrid analysis
We used the Matchmaker system (Clontech). The cDNA of the sgk-1 gene was cloned into a prey vector modified from the pGAD GH vector (Clontech). The cDNA of the arl-1, unc-11, cogc-3 and fragments of mon-2, gbf-1 were cloned into a bait vector modified from the pGBKT7 vector (Clontech). Bait and prey plasmids were co-transformed into the AH109 strain and transformants were selected on the double dropout medium (SD/-Leu/-Trp). The activation of the HIS3 reporter gene was assessed on the triple dropout medium (SD/-Leu/-Trp/-His).

Quantitative RT-PCR
Total RNA was extracted from synchronized L4 worms using TRIZOL (Invitrogen), followed by the removal of contaminant DNA using DNase I. cDNAs were synthesized from the total RNA templates using a reverse transcription kit (Takara

Feeding of Myriocin
A 1 mg/ml myriocin (Sigma) stock solution was made in methanol. Gravid adults (five per plate) were transferred to plates containing UV-irradiated (10 min at 100 mJ/cm 2 ) OP50 mixed with an equal volume of myriocin or solvent control (methanol). Precipitation of myriocin occurred at 25.2 μM and 50.4 μM. The first batch of L4 larvae of the next generation were imaged.

Results
In sgk-1 mutants MIG-14 failed to localize to the basolateral membrane of intestinal cells To determine whether SGK-1 plays a conserved role in endocytosis, we investigated the function of SGK-1 in recycling of the plasma membrane protein MIG-14. mig-14 encodes the C. elegans homologue of Wntless, an evolutionarily conserved multi-pass transmembrane protein, which binds Wnt in the Golgi and escorts Wnt to the plasma membrane for its release. Wnt proteins are a conserved family of secreted lipid-modified signaling glycoproteins that play a critical role in embryonic development [39]. Plasma membrane-bound MIG-14 is retrieved through endocytosis and transported to the Golgi in a retromer-dependent manner. In mutant C. elegans lacking subunits of the retromer complex, MIG-14 is mis-sorted into late endosomes and lysosomes [40][41][42].
The distribution of MIG-14/Wntless clearly changed in sgk-1 mutants. In wild-type animals, MIG-14::GFP is located on the basolateral membrane ( Fig 1A) and also on dispersed punctate structures in the cytoplasm (Fig 1A'). In two putative null mutants of sgk-1, ok538 and mg455, the basolateral membrane localization of MIG-14::GFP was hardly detectable, and most of MIG-14::GFP punctate structures localized near the apical membrane and the averaged fluorescence intensity of the cytoplasmic MIG-14::GFP decreased slightly (Fig 1B-1D and S1-S5 Figs). Thus the subcellular localization of MIG-14 is dependent on SGK-1.
MIG-14::GFP was mis-sorted into early lysosomes in sgk-1 mutant worms Cargo that cannot be retrieved to the TGN remains mostly in endosomes and hence should be transported to lysosomes [42]. To test this notion, we determined the subcellular localization of MIG-14::GFP in sgk-1 mutants. In wild-type animals, late endosomes and early lysosomes are both marked by mCHERRY::RAB-7 but can be distinguished morphologically: the former are punctate structures and the latter ring-like vesicles [33,43]. We found that mCHERRY:: RAB-7 labeled ring-like vesicles contained no or only a single MIG-14::GFP punctum in the wild type (Fig 2, panels A, A', A" and C), while most of the them contained two or more MIG-14::GFP puncta in the sgk-1 mutant, indicating that MIG-14::GFP is mis-sorted to lysosomes (Fig 2, panels B, B', B" and C). A similar phenotype of MIG-14::GFP was reported in a mutant lacking rme-8 [42], which encodes a retromer-associated protein.
Golgi-independent recycling of membrane receptors was defective in sgk-1 mutants Given that we only investigated a cargo that recycles through the Golgi so far, we asked next whether recycling to the plasma membrane was generally affected in sgk-1 mutant animals. hTAC (the a-chain of the human IL-2 receptor TAC) and GLUT1 (glucose transporter 1) are plasma membrane receptors internalized via clathrin-independent endocytosis and recycled back to the plasma membrane through the endocytic recycling compartment (ERC) and RME-1-positive basolateral recycling endosomes [44,45]. hTAC::GFP mainly localized to the basolateral membrane in the wild type (Fig 3, panels A and A'). In the sgk-1(ok538) mutant, the pool of hTAC::GFP on the basolateral membrane decreased slightly. However, strikingly, large hTAC::GFP structures, possibly aggregated vesicles, accumulated in the cytoplasm (Fig 3, panels B, B' and D). In the sgk-1(mg455) mutant, similar cytoplasmic accumulation of hTAC::GFP was observed, as well as a slight reduction of hTAC::GFP on the basal membrane (Fig 3, panels C, C', D, E and F, and S1, S2 and S6-S8 Figs). To further explore whether the internal aggregations are abnormal early endosomes, we determined the distribution of hTAC::GFP and an early endosome marker, RFP::RAB-5 in wild type and sgk-1(ok538) mutant worms. In wild type worms, only a small portion of the cytoplasmic hTAC::GFP puncta co-localized with early endosomes labeled by RFP::RAB-5 (S9 Fig) In sgk-1(ok538) mutant worms, most of the hTAC:: GFP labeled large, aggregated structures that are also labeled by RFP::RAB-5 (S9 Fig). It is unclear whether these structures were abnormal early endosomes or protein aggregates that contain both hTAC::GFP and RFP::RAB-5.
In intestinal cells of L4 larvae, GLUT1::GFP is limited to the apical membrane and punctate structures near the apical membrane ( Fig 3G). Loss of sgk-1 did not affect the apical membrane localization of GLUT1::GFP, but it increased the amount of cytosolic GLUT1::GFP near and away from the apical membrane ( Fig 3H). In young wild-type adults, GLUT1::GFP is seen on both apical and basolateral membranes [43]. We found that the basolateral membrane localization of GLUT1::GFP was abolished in young sgk-1 mutant adults (S10 Fig). The reason for the change in localization during development is unclear.
The above results suggest that sgk-1 plays a role in intracellular traffic of plasma membrane receptors. The less pronounced phenotype observed for GLUT1::GFP may suggest that apical recycling is less profoundly dependent on SGK-1 than basolateral recycling.

Reduction of glucosylceramide alleviates the sgk-1(lf) induced trafficking defects
To find out whether sgk-1 regulates membrane trafficking through sphingolipids, we first tested Myriocin, an inhibitor of the first and the key step in sphingosine biosynthesis. We fed the wild type and sgk-1(ok538) worms with 4.2 μM, 25.2 μM and 50.4 μM of Myriocin, but did not detect any obvious growth or morphological effect in either strain. Also, we did not see any effect of Myriocin on the localization pattern of either hTAC::GFP or MIG-14::GFP in wild type animals (S11 and S12 Figs). It is plausible that the drug may not accumulate efficiently in cells to show an effect. Therefore, we knocked-down cgt-1 and cgt-3, which encode two of the three ceramide glucosyltransferases that are important for the synthesis of glucosylceramide and glycosphingolipids. Similar to the cgt-1; cgt-2; cgt-3 triple mutant, cgt-1; cgt-3 double mutant animals have reduced glucosylceramide levels and arrest at the L1 larval stage [46]. In about 50% of the wild-type worms, cgt-1/3 double RNAi reduced the expression of hTAC::GFP to an undetectable level. In the remaining 50% wild-type worms, cgt-1/3 RNAi significantly decreased the amount of small cytoplasmic hTAC::GFP puncta (Fig 4). Larger-sized hTAC:: GFP puncta or aggregates were seen in the cytoplasm of intestinal cells in the sgk-1 mutant, but they were reduced in number upon cgt-1/3 RNAi (Fig 4). Therefore, cgt-1/3 RNAi alleviated the phenotype of sgk-1 null with respect to hTAC::GFP. Similarly, cgt-1/3 double RNAi suppressed sgk-1 with respect to MIG-14::GFP. sgk-1 null nearly abolished MIG-14::GFP on the basolateral membrane, but this was restored by cgt-1/3 double RNAi (Fig 5). It has been reported that worms treated with cgt-1/3 double RNAi arrested at the L1 stage [46], but in our hands, only some of the cgt-1/3 double RNAi animals arrested at L1, probably due to a weaker RNAi effect. The sgk-1 mutant worms develop somewhat more slowly than the wild type. Interestingly, sgk-1 mutant worms treated with cgt-1/3 double RNAi developed very slowly (Fig 6). Altogether, our results show that sgk-1 and cgt-1/3 have complex genetic interactions. Our data indicate that SGK-1 plays a conserved role in lipid homeostasis.
sgk-1(lf) did not affect the morphology of early endosomes, the Golgi and ER So far we provided evidence that plasma membrane proteins are mis-localized in sgk-1 mutants. One explanation for the phenotype is that the proteins do not reach the plasma membrane efficiently. Therefore, we asked next, whether the morphology of the organelles along the secretory pathway were affected in sgk-1 mutant animals. First, we examined early endosomes by checking the distribution of GFP::RAB-5 in the wild type and sgk-1(ok538) backgrounds (Fig 7, panels A and B). No obvious difference was detected, suggesting that SGK-1 function is not required for the formation of early endosomes. In earlier experiments we observed large  aggregates in the sgk-1(ok538) mutant that were positive for RFP::RAB-5 and hTAC::GFP (S9 Fig), but it was unclear whether these were protein aggregates or abnormal early endosomes. Given that sgk-1(ok538) had no effect on GFP::RAB-5, they were most likely protein aggregates that contained both hTAC::GFP and RFP::RAB-5. Alternatively, loss of sgk-1 may affect early endosomes only in a sensitized background such as when the trafficking system is overloaded with cargoes. Similarly, loss of sgk-1 did not change the morphology of the Golgi compartment labeled by mCHERRY::MANS (mannosidase), which appeared as dispersed punctate structures in the intestinal cells (Fig 7, panels C and D). Next, we examined the morphology of ER using a GFP-tagged TRAM (translocating chain-associating membrane protein), an ER- specific marker, and found no obvious difference between wild type and the sgk-1 mutant (Fig 7, panels E and F). However, loss of sgk-1 induced unfolded protein response in the ER (UPR-ER) as indicated by an elevated expression of hsp-4pr::GFP (Fig 8, panels A and B). We confirmed the UPR induction by detecting increased hsp-4 mRNA levels in sgk-1(ok538) L4 mutant worms (Fig 8C). Together, these results indicate that the morphology of early endosomes, Golgi and ER is not grossly altered and that sgk-1 may be important for protein homeostasis.

SGK-1::GFP is distributed at the cortex of the plasma membrane and throughout the cytoplasm of intestinal cells
To find out where SGK-1 may physically interact with the membrane trafficking system, we examined the distribution of SGK-1::GFP driven by its own promoter. SGK-1::GFP was expressed at a high level in the intestine, as well as the head and tail neurons (S13 Fig). In the intestinal cells, SGK-1::GFP was seen at the cortex of the plasma membrane (most prominently the apical membrane) (Fig 9A' and S13 Fig) and throughout the cytoplasm in a diffuse pattern (Fig 9, panels B' and C', and S13B Fig), occasionally in punctate structures over a diffuse background (not shown). The cortical SGK-1::GFP pool did not colocalize with RFP::RAB-5 (Fig 9,  panels A, A' and A"). The diffused appearance of SGK-1::GFP in the cytoplasm precludes the conclusion of co-localization of SGK-1::GFP and mCHERRY::MANS or mCHERRY::TRAM, an ER marker (Fig 9). However, it is possible that SGK-1 may interact with membrane proteins or membrane-associated proteins and is transiently recruited to organelles.

SGK-1 interacts with proteins involved in the regulation of intracellular transport
To identify potential SGK-1 binding proteins on membranes, we performed co-immunoprecipitation experiments. Using GFP binding protein (GBP), which is a high-affinity single-chain antibody of GFP [47], we immunoprecipitated (IP) a SGK-1::GFP fusion protein expressed in C. elegans under its own promoter. Co-immunoprecipitated proteins were identified using mass spectrometry (MS). After subtracting background binding proteins and non-specific interacting proteins that co-immunoprecipitated with GFP alone or a number of unrelated GFP fusion proteins [37], we found five SGK-1 associated proteins with an established function in intracellular transport: ARL-1, COGC-3, GBF-1, MON-2 and UNC-11 (Table 1). Interestingly, homologs of all but UNC-11 have been reported to be localized to the Golgi apparatus [48][49][50][51][52]. For example, the human homologue of ARL-1 is localized to the Golgi and is a  [48,49]. By sequence similarity, GBF-1 and MON-2 are classified as ArfGEFs and both are shown to reside on the Golgi, although MON-2 lacks a domain responsible for guanine nucleotide exchange [50,52]. In addition, COGC-3 is part of a tethering complex at the Golgi [51].
MON-2 and COGC-3 also interacted with SGK-1 in yeast two-hybrid assay, validating these two hits (S14 Fig). To find out whether SGK-1 regulates membrane trafficking through interaction with our candidate proteins, we assessed the localization of MIG-14::GFP and hTAC::GFP after knock-down of the individual candidates. We obtained RNAi bacterial strains of arl-1, gbf-1, mon-2 and unc-11. We found that gbf-1(RNAi) caused an accumulation of hTAC::GFP in the cytoplasm (S15 Fig) and about 60% of the arl-1 RNAi treated worms showed a large decrease in the expression level of hTAC::GFP. The basolateral membrane localization of hTAC::GFP were not affected by any of them (S15 Fig). mon-2(RNAi) abolished  specifically the lateral membrane localization of MIG-14::GFP along the circumferential (S16 Fig; vertical lines). RNAi of the other three genes had no obvious effect on MIG-14::GFP (S16 Fig). Together, these results neither prove nor reject the possibility of sgk-1 regulating membrane trafficking through these four candidates.

GBF-1 function is required for proper SGK-1 localization
Among the five candidate proteins, the ArfGEF protein GBF-1 has been characterized in intracellular trafficking in C. elegans [50]. The largest pool of GBF-1 resides at the cis-Golgi, and GBF-1 is required for endosome-mediated membrane traffic and the normal morphology of the Golgi and ER [50]. Thus, like SGK-1, GBF-1 acts at different stages in trafficking pathways. Therefore, we checked whether GBF-1 and SGK-1 would affect each other. sgk-1(RNAi) had no effect on steady-state GBF-1 localization (data not shown). In contrast, gbf-1(RNAi) affected the localization of SGK-1. SGK-1::GFP displayed a smooth, continuous distribution along the cortex of basolateral membrane of the intestinal cells. Upon gbf-1 RNAi, SGK-1::GFP at the cortex of basolateral membrane changed to small punctate structures (Fig 10). Likewise the apical localization of SGK-1::GFP was affected by gbf-1 RNAi, since the staining was strongly reduced in the RNAi-treated animals (Fig 10, panels B and B', pointed arrows). Meanwhile, there were more bright, large-sized SGK-1::GFP aggregates on the basolateral side of the intestinal cells (Fig 10C', open arrows). Such aggregates were observed in control animals only occasionally (Fig 10C, open arrows). The overall fluorescence intensity of SGK-1::GFP appeared not affected by gbf-1 RNAi (S17 Fig), probably because the increase of SGK-1::GFP aggregates offset the decrease in the plasma membrane associated SGK-1::GFP. Considering that GBF-1 co-immunoprecipitated with SGK-1 (Table 1), this result suggests that GBF-1 may be required additional images, nine for control RNAi and nine for the sgk-1 RNAi. The average intensity is denoted with a red line. There were fifteen L4 worms in each image and ten images for each treatment. *** P value <0.001 (Student's t-test). (C) Relative mRNA levels of hsp-4 in wild type and sgk-1(ok538) mutant animals. ** P value <0.01, * P value <0.05 (Student's t-test).
doi:10.1371/journal.pone.0130778.g008 for SGK-1 function near the basolateral membrane by either being involved in the exocytosis of a SGK-1 binding protein at the plasma membrane or through a role in lipid homeostasis. Alternatively, SGK-1 may directly require GBF-1 GEF function.

Discussion
In this study, we have found that SGK-1 associates in vivo with five proteins functioning in membrane trafficking. In C. elegans, both COGC-3 and GBF-1 reside on the Golgi, and the same can be said for ARL-1 and MON-2 in other species [48,50,52,54]. The N-terminal region of Mon2p, which is conserved from yeast to human, is sufficient to bind the Golgi membrane where Mon2p acts as a scaffold protein [52]. Thus, we speculate that SGK-1 may be recruited onto the Golgi membrane to regulate membrane trafficking, although no apparent co-localization was detected between SGK-1::GFP and the Golgi marker mCHERRY::MANS. Further studies are needed to determine whether SGK-1 is recruited to the Golgi membrane under certain conditions. The ArfGEF protein GBF-1 is localized in close proximity to the Golgi and the ER-exit sites, and is required for the normal morphology of ER [50]. In this study, we found that GBF-1 coimmunoprecipitated with SGK-1 (Table 1) and the loss of sgk-1 induced unfolded protein response in ER (Fig 8). Therefore, although SGK-1::GFP is seen diffusely distributed in the cytoplasm for the most part (S13 Fig), we speculate that SGK-1 may be transiently recruited onto intracellular membranes by GBF-1 and functions with or through GBF-1 to regulate trafficking of cargo proteins in or out of ER. Supporting this idea, the smooth distribution of SGK-1::GFP along the basolateral membrane was disrupted by gbf-1 RNAi (Fig 10). It is conceivable that SGK-1 may phosphorylate GBF-1 and alter its activity at the periphery of ER. More studies are needed to characterize the interaction between SGK-1 and GBF-1 and dissect the mechanism by which these two proteins regulate protein trafficking around ER. Homologous to human ARF1, a small GTPase that localizes to the Golgi apparatus and plays a central role in intra-Golgi transport [48].
GBF-1 1 4 Guanine nucleotide exchange factor of the Arf family small GTPases (ArfGEF). GBF-1 localizes to the cis-Golgi, is part of the t-ER-Golgi elements, required for secretion and the integrity of the Golgi and ER. [50] MON-2 3 3 ArfGEF-like protein orthologous to Saccharomyces cerevisiae Mon2p and the vertebrate MON2 proteins, predicted to function in the endosome-to-Golgi retrograde transport in C. elegans [52].
Spectral count: number of spectra that identify a protein, suggestive of the abundance of this protein in the sample. The spectral counts of SGK-1 are 302 and 412 in two technical repeats (Rep 1 and Rep 2). The spectral counts of the five membrane trafficking proteins are low, indicating their low abundance in the SGK-1::GFP immunoprecipitates. However, these proteins were highly specific for SGK-1::GFP, for they were not seen in the immunoprecipitates of seventeen other unrelated GFP fusion proteins (not shown). doi:10.1371/journal.pone.0130778.t001

SGK-1 Plays a Role in Membrane Trafficking in Caenorhabditis elegans
Transcytosis allows for the exchange of cargoes between the apical and basolateral membranes and it is critical for the establishment and maintenance of the cell polarity [55]. Although there is no obvious membrane-targeting signal on SGK-1, this protein is seen near or on the plasma membrane in intestinal cells (S13 Fig). The concentration of SGK-1::GFP is higher in the cortex of apical membrane than that of basolateral membrane. Curiously, deletion of sgk-1 affected basolateral recycling more severely than apical recycling, suggesting that SGK-1 activity is more important at basolateral membrane than apical membrane where there is more SGK-1 under normal situations. Also, gbf-1 RNAi disrupted the smooth appearance of SGK-1::GFP lining the basolateral membrane, but not that lining the apical membrane (Fig 10). It would be interesting to determine whether SGK-1 and GBF-1 are both involved in transcytosis to coordinate directed movement of cargoes.
Ypk1-the yeast homologue of SGK-1-activates serine palmitoyl-CoA acyltransferase (SPT), which catalyzes the first and the key step of de novo synthesis of ceramide [25,26]. Recently, Ypk1 is also reported to activate ceramide synthase [26]. We suspect that similar to yeast Ypk1, C. elegans SGK-1 may play a positive role in ceramide biosynthesis. In C. elegans, two types of complex sphingolipids, sphingomyelins and glycosphingolipids, are derived from ceramide and both contain an unusual, branched-chain sphingonoid base known as d17iso [56]. It is suggested that a particular species of sphingolipid called d17iso-GluCer or its derivative activates the TORC1 complex to promote larval development [57]. Inactivating the enzymes along the biosynthesis pathway of d17iso-GluCer arrests C. elegans at the L1 stage [57] or causes an uncharacterized lethal phenotype [58]. Our observation that sgk-1(null) exacerbated the developmental phenotype of cgt-1/cgt-3 RNAi (Fig 6) is consistent with SGK-1 playing a positive role in ceramide synthesis, because this phenotype can be explained by a reduction of d17iso-GluCer by cgt-1/cgt-3 RNAi, which is moderate in the wild-type background but greatly worsened by a reduction of ceramide, the precursor of d17iso-GluCer, in sgk-1 null animals. Also, we speculate that cgt-1/cgt-3 RNAi alleviated the trafficking phenotype of sgk-1(null) animals (Figs 4 and 5) by diverting the limited supply of ceramide to synthesizing sphingomyelins. In sgk-1 mutants, sphingomyelins may be reduced the most among all complex sphingolipids, which leads to impaired membrane trafficking. Alternatively, all complex sphingolipids may be reduced to similar levels, but the trafficking phenotype may be more sensitive to a reduction in sphingomyelins than in glycosphingolipids. Following the hypothesis that SGK-1 regulates membrane trafficking by regulating ceramide synthesis, we expected that Myriocin, an inhibitor of SPT, would alter the membrane localization of MIG-14::GFP and hTAC::GFP, but it did not (S11 and S12 Figs). Possibly, Myriocin was not delivered efficiently enough to show an effect due to the tight barrier of the worm cuticle. It remains to be tested whether inactivating sptl-1, sptl-2 and sptl-3, the SPT genes in C. elegans, would cause the same trafficking phenotype observed in sgk-1 mutants.  :GFP expression is seen in embryos, all larval stages and throughout adulthood. SGK-1::GFP is highly expressed in the intestine and the head and tail neurons, as described previously [28]. The apical and basolateral membranes were indicated by arrow heads and arrows. Scale bars: 10 μm. (EPS) S14 Fig. Yeast two-hybrid assay of SGK-1 and candidate binding proteins. Yeast cells were co-transformed with a GAL4BD-bait plasmid and a GAL4AD-prey plasmid as indicated. Among the double transformants (grown in the absence of Leu and Trp, or-LW) only two (AD-SGK-1/BD-COGC-3 and AD-SGK-1/BD-MON-2N) displayed positive interactions as they grew in the absence of Leu, Trp and His (-LWH). The apparent positive interaction of AD-SGK-1/BD-UNC-11 was false because BD-UNC-11 showed self-activating activity independent of the prey (AD-SGK-1 or AD-Xrc4). Xrc4 is a negative control; it is a nonhomologous end joining factor from S. pombe [59] and is not expected to interact with C. elegans  Table. Summary of pinhole size, detector gain, offset, and laser intensity settings for each figure.