Hyperosmotic Stress Induces Phosphorylation of CERT and Enhances Its Tethering throughout the Endoplasmic Reticulum

The ceramide transport protein (CERT) delivers ceramide from the endoplasmic reticulum (ER) to the Golgi apparatus, where ceramide is converted to sphingomyelin (SM). The function of CERT is regulated in two distinct phosphorylation-dependent events: multiple phosphorylations in a serine-repeat motif (SRM) and phosphorylation of serine 315 residue (S315). Pharmacological inhibition of SM biosynthesis results in an increase in SRM-dephosphorylated CERT, which serves as an activated form, and an enhanced phosphorylation of S315, which augments the binding of CERT to ER-resident VAMP-associated protein (VAP), inducing the full activation of CERT to operate at the ER–Golgi membrane contact sites (MCSs). However, it remains unclear whether the two phosphorylation-dependent regulatory events always occur coordinately. Here, we describe that hyperosmotic stress induces S315 phosphorylation without affecting the SRM-phosphorylation state. Under hyperosmotic conditions, the binding of CERT with VAP-A is enhanced in an S315 phosphorylation-dependent manner, and this increased binding occurs throughout the ER rather than restrictedly at the ER–Golgi MCSs. Moreover, we found that de novo synthesis of SM with very-long acyl chains preferentially increases via a CERT-independent mechanism under hyperosmotic-stressed cells, providing an insight into a CERT-independent ceramide transport pathway for de novo synthesis of SM.


Introduction
Intracellular trafficking of lipids is an essential event for the metabolism of lipids and membrane biogenesis in cells. In eukaryotic cells, the endoplasmic reticulum (ER) is the center of de novo synthesis of various lipid types, and lipid transport mediated by lipid transfer proteins (LTPs) at organelle membrane contact sites (MCSs) is the predominant mechanism to transport lipid from the ER to other organelles [1,2]. However, how the function of LTPs and the formation of MCSs are regulated in cells remains poorly understood.
In the synthesis of sphingomyelin (SM) in the Golgi apparatus, the ceramide transport protein (CERT), a typical LTP, delivers the precursor ceramide synthesized in the ER to the Golgi apparatus, where ceramide is converted to SM [3]. CERT contains a steroidogenic acute regulatory protein-related lipid transfer (START) domain, which catalyzes the inter-membrane transfer of ceramide as its lipid-transfer domain. In addition, to execute rapid and accurate ceramide transport at the ER-Golgi contact sites, CERT possesses the pleckstrin homology (PH) domain that preferentially recognizes phosphatidylinositol 4-monophosphate [PtdIns(4)P], which is the predominant phosphoinositide in the Golgi membranes, and two phenylalanines in an acidic tract (FFAT) motif that associates with VAMP-associated protein (VAP), an ER-resident protein [4]. Of note, de novo synthesis

Phosphorylation State of S315 but Not SRM in CERT Is Affected by Hyperosmotic Stress
In this study, to employ retroviral transfection, we used the HeLa mCAT#8 cell line, which stably expresses the mouse ecotropic retroviral receptor mCAT-1. We employed a HeLa CERT knockout (KO) cell line to prevent any interference with the endogenous wild-type (WT) CERT. In addition, when various CERT variants were expressed, CERT KO/shCERT cells that stably express a short hairpin RNA matching to human CERT were used as the parent HeLa KO cells to prevent overexpression artifacts unless otherwise noted. Using the cell system, we could control the protein expression levels of CERT variants in the CERT KO background near to the endogenous CERT level in wild-type HeLa cells ( Figure 1A). To examine whether the phosphorylation states of CERT are altered in response to hyperosmotic stress, we used 10% serum-containing culture media supplemented with various concentrations of sorbitol as hyperosmotic media unless otherwise noted. Upon incubation of cells in a 600 mM sorbitol-supplemented medium for 60 min, the phosphorylation level of S315 in HA-CERT wild-type (WT) was markedly elevated ( Figure 1B). This elevation was not due to any non-specific reactivity of the antibody used because no signal was detected in cells expressing HA-CERT S315A, which is incapable of being phosphorylated at the S315 position ( Figure 1B). In addition, Western blotting for [8]. When the SRM receives multiple phosphorylations, the activities of the PH domain, FFAT motif, and START domain are simultaneously repressed, and the function of CERT is consequently inactivated [9,10]. The S132A mutant, which mimics the dephosphorylated SRM form of CERT, exhibits a constitutively active feature [7,9]. Another phosphorylation site in CERT is serine 315 (S315), which resides upstream of the FFAT motif. The phosphorylation of S315 results in an increase in the FFAT motif-dependent affinity of CERT for VAP, enhancing the ER-to-Golgi ceramide trafficking function of CERT [10], although it remains unclear whether CERT is capable of binding to VAP anywhere in the ER or at the ER-Golgi MCSs. When de novo synthesis of SM is pharmacologically blocked, the SRM of CERT is dephosphorylated while S315 is phosphorylated [10], providing evidence that the two events of the SRM dephosphorylation and S315 phosphorylation are coordinated to fully activate CERT when the cells require SM. However, it remains unknown whether these two events are always coordinated.
In the present study, we found that hyperosmotic stress induces the phosphorylation of CERT S315 residue while the phosphorylation state of SRM remains multiply phosphorylated, which indicates that the two phosphorylation events can occur independently. Under hyperosmotic conditions, the interaction of CERT with VAP-A is enhanced in a CERT S315 phosphorylation-dependent manner, and this increased binding occurs throughout the ER rather than at the ER -Golgi MCSs. In addition, we unexpectedly found that de novo synthesis of SM with very-long acyl chains increases in a CERT-independent manner in hyperosmotic stressed cells, providing a feature of a CERT-independent ceramide transport pathway for de novo synthesis of SM.

Phosphorylation State of S315 but Not SRM in CERT Is Affected by Hyperosmotic Stress
In this study, to employ retroviral transfection, we used the HeLa mCAT#8 cell line, which stably expresses the mouse ecotropic retroviral receptor mCAT-1. We employed a HeLa CERT knockout (KO) cell line to prevent any interference with the endogenous wild-type (WT) CERT. In addition, when various CERT variants were expressed, CERT KO/shCERT cells that stably express a short hairpin RNA matching to human CERT were used as the parent HeLa KO cells to prevent overexpression artifacts unless otherwise noted. Using the cell system, we could control the protein expression levels of CERT variants in the CERT KO background near to the endogenous CERT level in wild-type HeLa cells ( Figure 1A). To examine whether the phosphorylation states of CERT are altered in response to hyperosmotic stress, we used 10% serum-containing culture media supplemented with various concentrations of sorbitol as hyperosmotic media unless otherwise noted. Upon incubation of cells in a 600 mM sorbitol-supplemented medium for 60 min, the phosphorylation level of S315 in HA-CERT wild-type (WT) was markedly elevated ( Figure 1B). This elevation was not due to any non-specific reactivity of the antibody used because no signal was detected in cells expressing HA-CERT S315A, which is incapable of being phosphorylated at the S315 position ( Figure 1B). In addition, Western blotting for the HA-tag showed that hyperosmotic stress did not affect the protein expression level of HA-CERT, ruling out the possibility that the elevated signal of S315 is due to increased expression of CERT ( Figure 1B). Notably, the doublet pattern of CERT did not discernibly alter ( Figure 1B). In the doublets, the upper band represents multiply phosphorylated forms in the SRM, while the lower band represents hypo-phosphorylated and/or de-phosphorylated form(s) [9,10]. The minor possibility that the upper band of the CERT doublet in cells exposed to the hyperosmotic stress did not represent the SRM-hyperphosphorylated form was eliminated because protein phosphatase treatment of cell lysate fractions shifted the upper band to the lower band independent of whether there was or was not an osmotic stimulus ( Figure 1C). Thus, the characterization of the phosphorylation states of CERT within cells exposed to hyperosmotic stress let us notice that the events of SRM dephosphorylation and the S315 phosphorylation might not necessarily be coordinated. To detect the S315 phosphorylated form by Western blotting analysis with anti-CERT S315p polyclonal antibodies, we had to concentrate CERT from the cell lysate fraction by immunoprecipitation ( Figure 1D). HA-CERT was immunoprecipitated with an anti-HA-epitope monoclonal antibody but not with commercially available anti-CERT polyclonal antibodies. Thus, for this technical reason, we could not verify that the hyperosmotic stress affected the level of phosphorylation of S315 of the endogenous CERT.
A previous study using CERT-overexpressing HeLa cells showed that when cells were treated with ISP-1/myriocin, a potent inhibitor of serine palmitoyltransferase (the enzyme catalyzing the first step in de novo synthesis of all sphingolipid types), S315 of HA-CERT WT was shifted to be phosphorylated [10]. In agreement with this previous study, the S315 phosphorylation was discernibly enhanced upon ISP-1-treatment with HeLa cells expressing HA-CERT at a near endogenous level, but the elevated level was far less than that seen in response to the hyperosmotic treatment ( Figure 1E). In contrast with the case of the hyperosmotic treatment, the SRM dephosphorylation form (the lower band of the doublet) increased in response to ISP-1 treatment, in line with previous studies [9,10]. These results indicate that responses in the phosphorylation states of the SRM and S315 of CERT vary depending on stimulus types. the HA-tag showed that hyperosmotic stress did not affect the protein expression level of HA-CERT, ruling out the possibility that the elevated signal of S315 is due to increased expression of CERT ( Figure 1B). Notably, the doublet pattern of CERT did not discernibly alter ( Figure 1B). In the doublets, the upper band represents multiply phosphorylated forms in the SRM, while the lower band represents hypo-phosphorylated and/or de-phosphorylated form(s) [9,10]. The minor possibility that the upper band of the CERT doublet in cells exposed to the hyperosmotic stress did not represent the SRM-hyperphosphorylated form was eliminated because protein phosphatase treatment of cell lysate fractions shifted the upper band to the lower band independent of whether there was or was not an osmotic stimulus ( Figure 1C). Thus, the characterization of the phosphorylation states of CERT within cells exposed to hyperosmotic stress let us notice that the events of SRM dephosphorylation and the S315 phosphorylation might not necessarily be coordinated. To detect the S315 phosphorylated form by Western blotting analysis with anti-CERT S315p polyclonal antibodies, we had to concentrate CERT from the cell lysate fraction by immunoprecipitation ( Figure 1D). HA-CERT was immunoprecipitated with an anti-HAepitope monoclonal antibody but not with commercially available anti-CERT polyclonal antibodies. Thus, for this technical reason, we could not verify that the hyperosmotic stress affected the level of phosphorylation of S315 of the endogenous CERT.
A previous study using CERT-overexpressing HeLa cells showed that when cells were treated with ISP-1/myriocin, a potent inhibitor of serine palmitoyltransferase (the enzyme catalyzing the first step in de novo synthesis of all sphingolipid types), S315 of HA-CERT WT was shifted to be phosphorylated [10]. In agreement with this previous study, the S315 phosphorylation was discernibly enhanced upon ISP-1-treatment with HeLa cells expressing HA-CERT at a near endogenous level, but the elevated level was far less than that seen in response to the hyperosmotic treatment ( Figure 1E). In contrast with the case of the hyperosmotic treatment, the SRM dephosphorylation form (the lower band of the doublet) increased in response to ISP-1 treatment, in line with previous studies [9,10]. These results indicate that responses in the phosphorylation states of the SRM and S315 of CERT vary depending on stimulus types.

Severe Hyperosmotic Stress, but Not Mild Hyperosmotic nor Hypoosmotic Stress, Induces Reversible CERT S315 Phosphorylation
We next examined whether the elevation of CERT S315 phosphorylation depended on the strength and duration of hyperosmotic stress. When cells were treated with various concentrations (100-800 mM) of sorbitol, and for different time periods, the phosphorylation states of the SRM were not altered ( Figure 1F). In contrast, the phosphorylation states of S315 clearly increased in a sorbitol concentration and exposure period-dependent manners ( Figure 1F): At 800 mM sorbitol, the level of the S315 phosphorylation gradually increased and reached a plateau at 60 min. At 400 mM sorbitol, the level of the S315 phosphorylation peaked at 30 min-but the degree of the peak was weaker than that of 800 mM sorbitol-and then reduced to the untreated control level. Significant changes were not observed at 100 mM sorbitol. Importantly, when NaCl was used as an osmolyte in place of sorbitol, the phosphorylation of CERT S315 was also elevated, ruling out the possibility that the elevation of the S315 phosphorylation was a sorbitol-specific event ( Figure 1G).
The osmolarity of the medium supplemented with 800 mM sorbitol is about 1100 mOsm, which should burden cells with severe hyperosmotic stress. This invoked the concern that the enhanced S315 phosphorylation was simply an irreversible event occurring in dying cells. However, the S315 phosphorylation elevation was found to be reversible ( Figure 2A): When cells were cultured in isosmotic media after exposure to 800 mM sorbitol for 60 min, the S315 phosphorylation level of the cells gradually returned to the near-basal level ( Figure 2A). These results demonstrate that the elevation of the CERT S315 phosphorylation in response to the severe hyperosmotic stress is not a response occurring only in dying cells. hyperosmotic conditions (500 mOsm for 24 h). However, both the states of multiply phosphorylated SRM and the S315 phosphorylation did not change ( Figure 2B).

The FFAT Motif-Dependent Interaction of CERT with VAP-A Is Enhanced under Hyperosmotic Conditions
We next examined impacts of hyperosmotic stress on the CERT-VAP interaction by a co-precipitation assay. When cells expressing HA-CERT WT at a nearly endogenous level were used, we failed to detect any discernible co-precipitation signal. Thus, we employed new cell lines in which CERT variants with an N-terminal mVenus-tag were ectopically expressed within the parental HeLa KO cell line ( Figure 3A). In these cells, two bands with apparent molecular masses of ~100 kDa were detected using anti-CERT and anti-GFP antibodies, and these two bands were also detected with an anti-C terminus CERT antiserum ( Figure 3A), suggesting that they are the products translated from two possible initiation codons in the open-reading frame of the N-terminally tagged-mVenus. The cells were incubated in the isosmotic or hyperosmotic medium for 1 h and treated with the membrane-permeable chemical crosslinker dithiobis (succinimidyl propionate) (DSP) to stabilize the CERT-VAP complex. Then, cell lysate treated with a mild detergent was subjected to co-immunoprecipitation with anti-GFP monoclonal antibody. A larger amount of VAP-A was co-precipitated with mVenus-CERT S315E compared with the CERT WT control ( Figure 3B), in agreement with a previous study that showed that the phosphorylated S315-mimetic mutation S315E augments the FFAT motif-dependent binding of CERT to VAP [10]. When cells were exposed to hyperosmotic stress, co-precipitation of CERT WT with VAP-A was augmented coincidently with the elevation of the S315 phosphorylation ( Figure 3B). Such augmentation did not occur in the S315A mutant (Figure 3B). For the mVenus-CERT ΔFFAT mutant, while its S315 phosphorylation occurred We also tested whether the phosphorylation states of CERT were responsive to hypoosmotic stress. When cells were exposed to a 1.5-fold diluted culture medium for various time periods, the states of multiply phosphorylated SRM and the S315 phosphorylation in CERT were unchanged ( Figure 2B). A previous study showed that hyperosmotic stress (550 mOsm for 48 h) induced the upregulation of SM synthesis in MDCK cells [11]. Therefore, we challenged the cells' prolonged exposure under relatively mild hyperosmotic conditions (500 mOsm for 24 h). However, both the states of multiply phosphorylated SRM and the S315 phosphorylation did not change ( Figure 2B).

The FFAT Motif-Dependent Interaction of CERT with VAP-A Is Enhanced under Hyperosmotic Conditions
We next examined impacts of hyperosmotic stress on the CERT-VAP interaction by a co-precipitation assay. When cells expressing HA-CERT WT at a nearly endogenous level were used, we failed to detect any discernible co-precipitation signal. Thus, we employed new cell lines in which CERT variants with an N-terminal mVenus-tag were ectopically expressed within the parental HeLa KO cell line ( Figure 3A). In these cells, two bands with apparent molecular masses of~100 kDa were detected using anti-CERT and anti-GFP antibodies, and these two bands were also detected with an anti-C terminus CERT antiserum ( Figure 3A), suggesting that they are the products translated from two possible initiation codons in the open-reading frame of the N-terminally tagged-mVenus. The cells were incubated in the isosmotic or hyperosmotic medium for 1 h and treated with the membrane-permeable chemical crosslinker dithiobis (succinimidyl propionate) (DSP) to stabilize the CERT-VAP complex. Then, cell lysate treated with a mild detergent was subjected to co-immunoprecipitation with anti-GFP monoclonal antibody. A larger amount of VAP-A was co-precipitated with mVenus-CERT S315E compared with the CERT WT control ( Figure 3B), in agreement with a previous study that showed that the phosphory-lated S315-mimetic mutation S315E augments the FFAT motif-dependent binding of CERT to VAP [10]. When cells were exposed to hyperosmotic stress, co-precipitation of CERT WT with VAP-A was augmented coincidently with the elevation of the S315 phosphorylation ( Figure 3B). Such augmentation did not occur in the S315A mutant ( Figure 3B). For the mVenus-CERT ∆FFAT mutant, while its S315 phosphorylation occurred in response to hyperosmotic stress, its co-precipitation with VAP-A was not detected. These results demonstrate that the augmented binding of CERT to VAP-A in response to hyperosmotic stress depends on the S315 phosphorylation and the FFAT motif in CERT.
We also performed an in situ proximity ligation assay (PLA) to assess the interaction between CERT and VAP-A. PLA detects the physical closeness between two proteins within 40 nm of each other as spot-like signals [12,13]. Although these signals do not necessarily represent the direct interaction between two proteins, the number of spot-like PLA signals can serve as a semi-quantitative measure for assessing the complex formation frequency of the two proteins. Moreover, the subcellular distribution of these signals enables one to evaluate where the interaction between two proteins occurs in cells. In HeLa cells expressing the mVenus-CERT WT, the number of PLA signals per cells increased in response to hyperosmotic stress ( Figure 3C), while PLA signals decreased by hyperosmotic stress in the CERT S315A-expressing cells. Hyperosmotic stress did not affect the numbers of the signals in cells expressing CERT ∆FFAT and CERT S315E mutants. These PLA experiment results validated that hyperosmotic stress-induced phosphorylation of S315 enhances the CERT-VAP interaction in cells.
Notably, the PLA signals representing the CERT-VAP association were distributed throughout the cytoplasm, and the signal distribution patterns did not differ between isosmotic and hyperosmotic conditions. This result raised the possibility that the S315phosphorylated CERT binds to VAP throughout the ER and not at limited regions at ER-Golgi MCSs. However, the results might be accounted for by another possibility: The Golgi apparatus was fragmented under hyperosmotic conditions, thereby enabling ER-Golgi contact sites to occur throughout the cytoplasm and not to be restricted to the perinuclear regions to which the Golgi is typically localized under normal culture conditions. To examine this possibility, subcellular distributions of VAP-A and the cis-Golgi marker GM130, or the trans-Golgi network marker TGN46, were analyzed by coimmunostaining. We observed no marked differences in the subcellular distribution of these Golgi markers between the osmotically stressed and the isosmotic control HeLa cells ( Figure 3D-G), although hyperosmotic stress was reported to induce fragmentation of the Golgi apparatus in human corneal cells [14]. It should also be noted that hyperosmotic stress increased the perinuclear signals of VAP-A, which were well colocalized with GM130 or TGN46 ( Figure 3D,E). The perinuclear redistribution of VAP-A occurred in both the HeLa WT and CERT KO cells, indicating that the VAP-A redistribution is induced in a CERT-independent manner ( Figure 3D,E). Sec61β, another ER-resident membrane protein, did not exhibit hyperosmotic treatment-dependent changes in its subcellular distribution ( Figure 3F,G). Although it remains unknown how the perinuclear redistribution of VAP-A occurs under hyperosmotic conditions, it may imply an increased association of VAP-A with the perinuclear Golgi complex. Collectively, these results show that the S315 phosphorylated CERT binds to VAP throughout the ER and not at limited regions at the ER-Golgi MCSs.

Effects of Hyperosmotic Stress on the Synthesis of SM
We next examined effects of hyperosmotic stress on de novo synthesis of SM. Cells were labeled with [ 14 C]-serine under isosmotic or hyperosmotic conditions for 1 h, and the metabolically labeled lipids were quantified ( Figure 4A,B). Under isosmotic conditions, the patterns of SM synthesis among cell types used were as expected: Compared with the level of SM synthesis in HeLa WT cells, the level in CERT KO was far less. Ectopic expression of mVenus-CERT WT and S315A in CERT KO cells reversed the SM synthesis levels with that in HeLa WT cells, while the ectopic expression of mVenus-CERT ∆FFAT and CERT S315E constructs reversed the levels below and beyond, respectively, the WT control level ( Figure 4A). Hyperosmotic stress unexpectedly enhanced the synthesis of SM in CERT KO cells ( Figure 4A,B), demonstrating that CERT-independent synthesis of SM is induced under hyperosmotic conditions. In all other cell lines except for the CERT S315E-expressing cell line, hyperosmotic stress enhanced the level of SM synthesis, but this enhancement could be accounted for by the induction of the CERT-independent pathway ( Figure 4A). In HeLa CERT KO/mVenus-CERT S315E cells, the SM synthesis was augmented even under isosmotic conditions, and the level was not further enhanced by hyperosmotic stress ( Figure 4A). These results suggest that hyperosmotic stress does not upregulate the CERT-dependent synthesis of SM even when CERT-VAP interaction is enhanced by S315 phosphorylation and that hyperosmotic stress somehow renders de novo SM synthesis in a CERT-independent manner.

Effects of Hyperosmotic Stress on the Synthesis of SM
We next examined effects of hyperosmotic stress on de novo synthesis of SM. Cells were labeled with [ 14 C]-serine under isosmotic or hyperosmotic conditions for 1 h, and the metabolically labeled lipids were quantified ( Figure 4A,B). Under isosmotic conditions, the patterns of SM synthesis among cell types used were as expected: Compared with the level of SM synthesis in HeLa WT cells, the level in CERT KO was far less. Ectopic expression of mVenus-CERT WT and S315A in CERT KO cells reversed the SM synthesis levels with that in HeLa WT cells, while the ectopic expression of mVenus-CERT ΔFFAT and CERT S315E constructs reversed the levels below and beyond, respectively, the WT control level ( Figure 4A). Hyperosmotic stress unexpectedly enhanced the synthesis of SM in CERT KO cells ( Figure 4A,B), demonstrating that CERT-independent synthesis of SM is induced under hyperosmotic conditions. In all other cell lines except for the CERT S315Eexpressing cell line, hyperosmotic stress enhanced the level of SM synthesis, but this enhancement could be accounted for by the induction of the CERT-independent pathway ( Figure 4A). In HeLa CERT KO/mVenus-CERT S315E cells, the SM synthesis was augmented even under isosmotic conditions, and the level was not further enhanced by hyperosmotic stress ( Figure 4A). These results suggest that hyperosmotic stress does not upregulate the CERT-dependent synthesis of SM even when CERT-VAP interaction is enhanced by S315 phosphorylation and that hyperosmotic stress somehow renders de novo SM synthesis in a CERT-independent manner.

Hyperosmotic Stress Induces the Preferential Synthesis of the Very-Long-Chain SM via a CERS2-Dependent and CERT-Independent Manner
Scrutinizing the metabolic labeling of TLC images, we noticed that molecular species of de novo synthesized SM were likely altered under hyperosmotic conditions. Metabolic labeling under isosmotic conditions gave typical doublet patterns of SM ( Figure 4A

Hyperosmotic Stress Induces the Preferential Synthesis of the Very-Long-Chain SM via a CERS2-Dependent and CERT-Independent Manner
Scrutinizing the metabolic labeling of TLC images, we noticed that molecular species of de novo synthesized SM were likely altered under hyperosmotic conditions. Metabolic labeling under isosmotic conditions gave typical doublet patterns of SM ( Figure 4A), in which the upper bands represent SM with very-long N-acyl chain lengths (20 or more car-bons), and the lower bands represent SM with long N-acyl chain lengths (16 or 18 carbons). Under hyperosmotic conditions, the upper bands of SM were preferentially produced in the metabolic labeling ( Figure 4A). Hyperosmotic stress did not affect the doublet patterns of phosphatidylethanolamine ( Figure 4A), indicating that the hyperosmotic stress-induced alteration in acyl chain lengths is a sphingolipid-specific event.
Acyl chain lengths of de novo synthesized SM are determined in the step of ceramide production. There are six isozymes in ceramide synthase (CERS1-6) that exhibit different substrate specificity with distinct lengths of acyl-CoA, and CERS2 mainly catalyzes the formation of ceramide species with very-long acyl chains in HeLa cells [15]. To elucidate the contribution of CERS2 to the preferential synthesis of very-long-chain SM (VL-SM) under hyperosmotic conditions, metabolic labeling analysis was performed with CERS2 KO HeLa cells [6]. Long-chain SM (L-SM) was predominantly synthesized in hyperosmotic-stressed CERS2 KO cells as well as in the non-treated cells, and the amounts of synthesized L-SM increased compared with those of parent cells ( Figure 4B). This finding is consistent with a previous report that found that in CERS2-deficient cells, the amount of L-ceramides increased to compensate for the shortage of VL-ceramides [6]. This result indicated that the preferential VL-SM synthesis in hyperosmotic environments was dependent on CERS2 and suggested that the VL-ceramides synthesis was upregulated under these conditions. Furthermore, we performed an in vitro ceramide synthesis assay to monitor CERS2 activity. After exposure of the parental HeLa mCAT, CERS2 KO, and CERT KO cells to the isotonic control or hyperosmotic conditions, membrane fractions were prepared from the cells. The membrane preparations as the CERS enzyme source were incubated with deuteriumlabeled (d7)-sphingosine in the presence of C24:1-or C16:0-acyl CoA, and the quantities of d7-sphingosine-labeled C24:1-and C16:0-ceramides were determined by LC-MS. The membrane fractions from parental HeLa cells and CERT KO cells exhibited a hyperosmotic stimulus-dependent increase in the synthesis of C24:1-ceramide, while no synthesis of C24:1-ceramide was detected in CERS2 KO cell-derived membranes ( Figure 4C). On the other hand, none of the parental, CERS2 KO, or CERT KO cell-derived membranes exhibited significant changes in the synthesis of C16:0-ceramide in response to the hyperosmotic stimulus ( Figure 4D). These results indicate that hyperosmotic stress selectively upregulates CERS2 activity, which presumably underlies the hyperosmotic stress-induced shift from L-SM to VL-SM, although it remains unclear how CERS2-selective activation occurs under hyperosmotic conditions.

Pharmacologically Induced Merging of the Golgi Apparatus with the ER Does Not Abrogate the Hyperosmotic Stress-Induced Preferential Synthesis of VL-SM
Glucocylceramide (GlcCer) synthase, which catalyzes the conversion of ceramide to GlcCer, has been suggested to be localized to the cis/medial Golgi regions in mammalian cells [16], and transport of ceramide from the ER to the Golgi site for the synthesis of GlcCer does not require CERT. Because glycosphingolipids have relatively longer acyl chains than SM, it has been proposed that a CERT-independent transport pathway of ceramide for the synthesis of GlcCer preferentially delivers VL-ceramides over L-ceramides [6]. We hypothesized that a VL-ceramides-selective and CERT-independent pathway for delivering ceramide from the ER to the Golgi apparatus might participate in the hyperosmotic stressinduced preferential synthesis of VL-SM even in the absence of CERT. To test this hypothesis, we employed the pharmacological tools Brefeldin A (BFA) and Golgicide A (GCA) (both of which are inhibitors of GBF1: guanine nucleotide exchange factor for ARF [17][18][19]). Because treatment of cells with these drugs induces absorption of the Golgi apparatus into the ER, ceramide is efficiently converted to SM in the drug-treated cells even without any transport pathways to mediate the ER-to-Golgi transport of ceramide. Consistent with previous studies [3,6], when CERT KO cells were treated with BFA or GCA under isosmotic conditions, the cells produced high amounts of both VL-SM and L-SM ( Figure 5A). If hyperosmotic stress induces a CERT-independent pathway to deliver ceramide from the ER to the Golgi apparatus, treatment of cells with the ER-Golgi merging drugs should abrogate the preferential synthesis of VL-SM. Nevertheless, the drug treatment of CERT KO cells did not abrogate the preferential synthesis of VL-SM ( Figure 5B). These results rule out the possibility that the hyperosmotic stress-induced preferential synthesis of VL-SM is due to a possible upregulation of the putative CERT-independent ceramide transport pathway and further support the notion that the hyperosmotic stress-induced preferential synthesis of VL-SM is due to preferential synthesis of VL-ceramides.
Int. J. Mol. Sci. 2022, 23, x FOR PEER REVIEW 13 of 20 5B). These results rule out the possibility that the hyperosmotic stress-induced preferential synthesis of VL-SM is due to a possible upregulation of the putative CERT-independent ceramide transport pathway and further support the notion that the hyperosmotic stressinduced preferential synthesis of VL-SM is due to preferential synthesis of VL-ceramides.

Discussion
Both the SRM dephosphorylation and S315 phosphorylation occur to activate CERT fully [9,10]. Nevertheless, it remained unknown whether these two phosphorylation events are always induced simultaneously. In this study, we serendipitously found that hyperosmotic stress evokes CERT S315 phosphorylation without affecting the SRM-phosphorylation state. This demonstrates that the two phosphorylation events can be separated depending on the types of stimuli. Under hyperosmotic conditions, the binding of CERT with VAP-A is enhanced in an S315 phosphorylation-and FFAT motif-dependent manner, and this increased binding occurs widely on the ER ( Figure 6A), indicating that CERT has the potential to bind VAP anywhere on the ER and that it is not restricted to the ER-Golgi MCSs. When the PH domain of ER-scattering CERT meets PtdIns(4)P of trans Golgi regions, which appose in close proximity to subregions of smooth ER [20], CERT might be trapped at the ER-Golgi MCSs.

Discussion
Both the SRM dephosphorylation and S315 phosphorylation occur to activate CERT fully [9,10]. Nevertheless, it remained unknown whether these two phosphorylation events are always induced simultaneously. In this study, we serendipitously found that hyperosmotic stress evokes CERT S315 phosphorylation without affecting the SRMphosphorylation state. This demonstrates that the two phosphorylation events can be separated depending on the types of stimuli. Under hyperosmotic conditions, the binding of CERT with VAP-A is enhanced in an S315 phosphorylation-and FFAT motif-dependent manner, and this increased binding occurs widely on the ER ( Figure 6A), indicating that CERT has the potential to bind VAP anywhere on the ER and that it is not restricted to the ER-Golgi MCSs. When the PH domain of ER-scattering CERT meets PtdIns(4)P of trans Golgi regions, which appose in close proximity to subregions of smooth ER [20], CERT might be trapped at the ER-Golgi MCSs. conditions. Under isosmotic conditions, a major population of CERT, multiply phosphorylated in the SRM, localizes to the ER or the cytosol, while a minor population of CERT, which is de/hypophosphorylated in the SRM, may localize at the ER-Golgi MCSs. Under hyperosmotic conditions, a majority of CERT are phosphorylated at the S315 residue with SRM-hyperphosphorylation resulting in enhancement of tethering to the ER, rather than localization at the ER-Golgi MCSs, via the increased binding of CERT to VAP-A. (B) A hypothetical model that may account for the increased VL-SM synthesis under hyperosmotic conditions in a CERT-independent manner. Ceramide synthesized in the ER is transported to the trans-Golgi, where SM synthase 1 (SMS1) locates, via a CERT-dependent (solid line) or -independent (dotted line) pathway. In wild-type HeLa cells under isosmotic conditions, long-chain ceramides (L-ceramides), which are mainly catalyzed by CERS5, are delivered from the ER to the distal Golgi regions preferentially via the CERT-dependent pathway, whereas very-long-chain ceramides (VL-ceramides), which are mainly catalyzed by CERS2, are presumably transported to the proximal or more broad Golgi regions via a CERT-independent pathway(s), of which the entity remains elusive. Under hyperosmotic conditions, the synthesis of VL-ceramides is enhanced, and therefore, a greater amount of VL-ceramides is delivered to the Golgi via the CERT-independent pathway, thereby enhancing the synthesis of VL-SM. conditions. Under isosmotic conditions, a major population of CERT, multiply phosphorylated in the SRM, localizes to the ER or the cytosol, while a minor population of CERT, which is de/hypophosphorylated in the SRM, may localize at the ER-Golgi MCSs. Under hyperosmotic conditions, a majority of CERT are phosphorylated at the S315 residue with SRM-hyperphosphorylation resulting in enhancement of tethering to the ER, rather than localization at the ER-Golgi MCSs, via the increased binding of CERT to VAP-A. (B) A hypothetical model that may account for the increased VL-SM synthesis under hyperosmotic conditions in a CERT-independent manner. Ceramide synthesized in the ER is transported to the trans-Golgi, where SM synthase 1 (SMS1) locates, via a CERT-dependent (solid line) or -independent (dotted line) pathway. In wild-type HeLa cells under isosmotic conditions, long-chain ceramides (L-ceramides), which are mainly catalyzed by CERS5, are delivered from the ER to the distal Golgi regions preferentially via the CERT-dependent pathway, whereas very-long-chain ceramides (VL-ceramides), which are mainly catalyzed by CERS2, are presumably transported to the proximal or more broad Golgi regions via a CERT-independent pathway(s), of which the entity remains elusive. Under hyperosmotic conditions, the synthesis of VL-ceramides is enhanced, and therefore, a greater amount of VL-ceramides is delivered to the Golgi via the CERT-independent pathway, thereby enhancing the synthesis of VL-SM.
Although hyperosmotic stress invokes an increase in de novo SM synthesis in HeLa cells, this increase occurs independently of CERT ( Figure 4). Thus, the hyperosmotic stressinduced enhancement of CERT binding to the ER is unlikely to enhance CERT's ER-Golgi ceramide trafficking function. We speculate that the enhanced binding of CERT to the ER is somehow beneficial for hyperosmotic stressed cells. Previous studies demonstrated that aberrant trafficking of ceramide to mitochondria causes cell death [21,22]. Cells undergo shrinkage under hyperosmotic conditions, which might generate otherwise absent emergent physical contacts between the ER and other organelles.
Moreover, hyperosmotic stress impinges on various types of membrane vesicle trafficking, such as the ER-to-Golgi transports or endosomal pathways [23,24]. These trafficking defects may also disrupt cellular homeostasis, including flows of sphingolipids. Therefore, it would be conceivable that trapping CERT on the ER prevents CERT from mis-sorting ceramide to non-Golgi organelles when abnormal organelle MCSs are generated or lipid metabolic enzymes are not appropriately sorted.
In addition, we found that hyperosmotic stress induces preferential biosynthesis of VL-SM in a CERT-independent manner ( Figures 4A and 5A). The hyperosmotic stressinduced synthesis of VL-SM is not abrogated, even when the ER and Golgi are merged using pharmacological tools ( Figure 5B), which eliminated the possibility that the phenomenon is due to an upregulation of a CERT-independent pathway(s) for ER-to-Golgi transport of ceramide. CERS2 is the predominant ceramide synthase (also known as dihydrosphingoine-N-acyltransferase) isoform to produce VL-ceramides (more precisely, VL-dihydroceramides) in HeLa cells [6]. In CERS2 KO HeLa cells, L-SM, instead of VL-SM, is predominantly synthesized, and the level of L-SM synthesis is not affected by hyperosmotic stress ( Figure 4A). To explain these results together with those of a previous study [6], we propose the model illustrated in Figure 6B. In wild-type HeLa cells under isosmotic conditions, L-ceramides are delivered from the ER to the distal Golgi regions preferentially via the CERT-dependent pathway, whereas VL-ceramides are presumably transported to the proximal or broad Golgi regions via a CERT-independent pathway(s), the entity of which remains elusive. Under hyperosmotic conditions, the synthesis of VL-ceramides is enhanced, and therefore, a greater amount of VL-ceramides is delivered to the Golgi via the CERT-independent pathway, thereby enhancing the synthesis of VL-SM. This model would account for that hyperosmotic stress inducing the synthesis of VL-SM even in the absence of CERT and also would explain that hyperosmotic stress does not enhance L-SM synthesis in the absence of CERS2. More studies will be needed to address unresolved questions-for example, how does hyperosmotic stress enhance the CERS2-dependent synthesis of VL-ceramides, and what is the entity of the CERT-independent pathway?
The osmolarity of the medium supplemented with 800 mM sorbitol was about 1100 mOsm. The renal inner medulla is exposed to such severe hyperosmotic environments [25]. SM molecular species with different acyl-chain lengths were shown to be distributed among different subregions of the kidney [26]. It could be possible that the hyperosmotic stressinduced phosphorylation of CERT S315 partly underlies the region-dependent distribution of various SM molecular species in the kidney.
Although the biological importance of VL-SM under osmotic stress has not yet been elucidated, a previous study showed that an increase in VL-sphingolipids in yeast cells may strengthen membrane integrity and may endow some tolerance to hyperosmotic stress [27]. By analogy, the increased VL-SM synthesis in mammalian cells might be an acute response for adapting to hyperosmotic conditions. Considering that ceramides act as modulators of various proteins [28], the shift from L-ceramides to VL-ceramides might affect cellular signaling events. For example, C16-ceramide, but not other ceramide species, stabilizes the p53 tumor suppressor, a key regulator in various fundamental cellular events such as the cell cycle, apoptosis, and survival in response to diverse stimuli [29]. On the other hand, acute hyperosmotic stress induces caspase-mediated apoptosis [30]. Thus, it is conceivable that a shift from L-ceramides to VL-ceramides may destabilize p53 and consequently attenuate p53-mediated hyperosmotic stress-induced cell death. Further studies will be needed to address these possibilities.
In conclusion, hyperosmotic stress evokes CERT S315 phosphorylation without affecting the SRM-phosphorylation state, showing compelling evidence that the two phosphorylation events can be separated. Under hyperosmotic conditions, the binding of CERT with VAP-A is enhanced in an S315 phosphorylation-and FFAT motif-dependent manner, and this increased binding occurs widely on the ER. This reveals that CERT has the potential to bind VAP anywhere on the ER and that it is not restricted to the ER-Golgi MCSs.

HeLa Cell Lines and Cell Culture
The HeLa-mCAT#8 cell line that stably expresses the mouse ecotropic retroviral receptor mCAT-1 was used as the HeLa WT cell line, and its CERT1-disrupted mutant cell line was used as the HeLa CERT KO cell line [5] unless otherwise noted. To avoid unwanted overexpression of CERT cDNAs, the HeLa CERT KO/shCERT cell line was used. The HeLa CERT KO/shCERT cell line was established as described in a separate publication (Goto et al., submitted). In brief, the CERT KO cells were transfected using a retroviral vector containing a short hairpin RNA against CERT1 RNA to interfere with the expression of CERT1 mRNA. Then, a stable clone was isolated and used as the parental HeLa CERT KO/shCERT cell line. The CERS2 KO cell line (#16) was established as described previously [6]. HeLa cells were cultured in high glucose Dulbecco's modified Eagle medium (D-MEM; 044-29765, Wako, Osaka, Japan) supplemented with 10% heat-inactivated fetal calf serum.

Plasmids
cDNA fragments encoding the full-length human CERT wild-type and its mutants were digested from previously described vectors [4,9] and subcloned into pMX-IB vectors containing the human influenza HA tag or the mVenus tag.

Retroviral Transfection and Establishment of Stable Cell Lines
Retroviral transfection of HeLa cells with various cDNA constructs was performed using the pMXs-IB-based retroviral vector as described previously [5]. After selecting transfected cells with 7.5 µg/mL blasticidin (#029-18701, Wako), cell clones were purified by limited dilution. Cell cloning was not performed for blasticidin-selected cells expressing mVenus-tagged CERT; instead, a population of cells expressing the fluorescent protein in a limited range was selected by FACS. After propagation, the FACS-selected cells were used for analysis.