Sphingomyelin synthase-related protein generates diacylglycerol via the hydrolysis of glycerophospholipids in the absence of ceramide

Diacylglycerol (DG) is a well-established lipid second messenger. Sphingomyelin synthase (SMS)-related protein (SMSr) produces DG and ceramide phosphoethanolamine (CPE) by the transfer of phosphoethanolamine from phosphatidylethanolamine (PE) to ceramide. We previously reported that human SMSr overexpressed in COS-7 cells significantly increased DG levels, particularly saturated and/or monounsaturated fatty acid–containing DG molecular species, and provided DG to DG kinase (DGK) δ, which regulates various pathophysiological events, including epidermal growth factor–dependent cell proliferation, type 2 diabetes, and obsessive–compulsive disorder. However, mammalian SMSr puzzlingly produces only trace amounts of CPE/DG. To clarify this discrepancy, we highly purified SMSr and examined its activities other than CPE synthase. Intriguingly, purified SMSr showed a DG-generating activity via hydrolysis of PE, phosphatidic acid (PA), phosphatidylinositol (PI), and phosphatidylcholine (PC) in the absence of ceramide. DG generation through the PA phosphatase (PAP) activity of SMSr was approximately 300-fold higher than that with PE and ceramide. SMSr hydrolyzed PI ten times stronger than PI(4,5)bisphosphate (PI(4,5)P2). The PAP and PC-phospholipase C (PLC) activities of SMSr were inhibited by propranolol, a PAP inhibitor, and by D609, an SMS/PC-PLC inhibitor. Moreover, SMSr showed substrate selectivity for saturated and/or monounsaturated fatty acid–containing PA molecular species, but not arachidonic-acid-containing PA, which is exclusively generated in the PI(4,5)P2 cycle. We confirmed that SMSr expressed in COS-7 cells showed PAP and PI-PLC activities. Taken together, our study indicated that SMSr possesses previously unrecognized enzyme activities, PAP and PI/PE/PC-PLC, and constitutes a novel DG/PA signaling pathway together with DGKδ, which is independent of the PI(4,5)P2 cycle.

Sphingomyelin synthase (SMS) is a DG-generating enzyme (17,18). There are three isoforms, SMS1, SMS2, and SMSrelated protein (SMSr) ( Table 1). Huitema et al. (19) identified SMS genes via a homology search for the sequences encoding the membrane proteins containing active site motifs common to PAP2/LPP. The SMS1 and SMS2 proteins generate DG and SM via the transfer of phosphocholine from PC to ceramide (Table 1). SMSr is an endoplasmic reticulum (ER)-resident, six-transmembrane protein and has no SMS activity but displays ceramide phosphoethanolamine (CPE) synthase (CPES) activity via the transfer of phosphoethanolamine from PE to ceramide (20) (Table 1). Puzzlingly, mammalian SMSr produces only trace amounts of CPE (approximately 300-fold lower than SM levels produced by SMS1) (20) (Table 1). Moreover, CPE levels in mammalian cells are extremely low (0.002%-0.005 mol% of total phospholipids) (21). Hence, it is supposed that SMSr has little or no ceramide-dependent DG-generating activity.
Taken together, SMSr can supply DG to DGKδ (31), whereas SMSr has little or no ceramide-dependent DGgenerating activity (CPES activity) (20,21). These contradictory results prompted us to postulate that SMSr possesses other DG-generating activity in addition to CPES activity. Here, we report that human SMSr, which was expressed in Sf9 cells and is highly purified, displayed DG-generating activity via hydrolysis of PA, PI, PE, and PC in the absence of ceramide. The PAP, PI-phospholipase C (PLC), PE-PLC, and PC-PLC activities were much stronger than the CPES activity (300-fold, 10-fold, 4-fold, and 3-fold, respectively). Intriguingly, SMSr exhibited a substrate specificity for SFAand/or MUFA-containing PA and PC molecular species, but not polyunsaturated fatty acids (PUFA)-containing PA and PC. Taken together, we revealed the previously unrecognized enzyme activities of mammalian SMSr, novel PAP, PI-PLC, PE-PLC, and PC-PLC activities, to produce SFA and/or MUFA-containing DG molecular species independent of ceramide in the ER membrane.

Purification of human SMSr
To determine whether SMSr possesses other DG-generating activity in addition to CPES activity, we first expressed Nterminal hexahistidine (6×His)-tagged human full-length (FL) SMSr and SMSr-ΔSAMD, which lacks an oligomer formation domain, SAMD (32,33), in Sf9 cells and purified them from Sf9 cell membranes using Ni-affinity chromatography (Fig. 1A). SMSr-ΔSAMD showed no significant difference in DG production levels compared with FL-SMSr (Fig. 1B). However, the SMSr-ΔSAMD yield was approximately six-fold higher than that of FL-SMSr (data not shown). Therefore, we employed SMSr-ΔSAMD for further investigation.
Next, to further purify 6× His-SMSr-ΔSAMD, sizeexclusion chromatography was conducted after Ni-affinity chromatography. SMSr-ΔSAMD was eluted as a single peak in a volume of 13.9 ml (56.1 kDa) ( Fig. 2A). Thus, the purified 6× His-SMSr-ΔSAMD is likely to be a monomer because its calculated molecular mass is 40.8 kDa (SMSr-ΔSAMD: 39.2 kDa +6×His: 1.6 kDa). Immunoblotting and silver staining showed that a single protein band at approximately 40 kDa was detected (Fig. 2B), indicating that 6× His-SMSr-ΔSAMD was obtained with a high purity. The yield was 6 μg per 1 L of an Sf9 cell culture. We confirmed that the purified SMSr-ΔSAMD slightly produced DG in the presence of PE and ceramide micelles (Fig. 2C). However, the DG-generating activity was exceedingly low (3 pmol/mg/min). Moreover, only trace amounts of CPE were generated by the purified SMSr-ΔSAMD (Fig. 2D).
Because DG production activity by SMSr-ΔSAMD in the presence of d18:1/18:0-ceramide and PE was lower than that in presence of PE alone (Fig. 3A), we assumed that ceramide affects DG-generating activity of SMSr-ΔSAMD. Thus, the effects of d18:1/18:0-ceramide on hydrolysis activities of SMSr-ΔSAMD for other glycerophospholipids (PA, PI, and PC) were determined (Fig. 4). We confirmed that PE-PLC activity of SMSr-ΔSAMD was significantly decreased in presence of d18:1/18:0-ceramide (Fig. 4A). However, ceramide had no detectable effects on hydrolysis of PA, PI, and PC (Fig. 4, B-D), suggesting that the inhibitory effect of ceramide is PE selective.
To test whether human SMSr expressed in mammalian cells, not insect cells, also displayed PAP and PI-PLC activities, C-terminal V5-tagged SMSr was expressed in COS-7 cells. The V5-tagged SMSr was immunoprecipitated using an anti-V5 tag antibody (Fig. 6A), and PAP, PI-PLC, and CPES activities in the immunoprecipitates were measured (Fig. 6, B-D). The SMSr expressed in the COS-7 cells exhibited PAP and PI-PLC activities (Fig. 6, B and C). However, a CPE synthase-inactive mutant of SMSr (D348E) (20) failed to show PAP and PI-PLC activities, indicating that Asp348 is also important for PAP and PI-PLC activities. On the other hand, the CPES activity was too low to detect in the precipitates (Fig. 6D), suggesting that human SMSr lacks the ability to synthesize considerable amounts of CPE, as Vacaru et al. reported (20).
We previously reported that SMSr expressed in COS-7 cells generated DG (30). To determine whether SMSr indeed utilizes PA to produce DG in mammalian cells, we analyzed the changes in the amounts of PA in COS-7 cells overexpressing SMSr-V5 (Fig. 7A). The total PA levels in COS-7 cells overexpressing SMSr-V5 were significantly decreased (Fig. 7B). This result suggests that SMSr produced DG, at least in part, by hydrolyzing PA in mammalian cells.

Discussion
In the present study, we answered a major question: why does SMSr only exhibit negligible CPES activity? We found

SMSr acts as PAP and PLC
that SMSr is a new type of enzyme that is able to hydrolyze multiple glycerophospholipids, such as PA, PI, PE, and PC, to produce DG (Fig. 8). SMSr is a novel PAP that primarily acts in the ER lumen unlike the previously reported PAP/LPP (42,43). Intriguingly, SMSr is true mammalian PI-PLC, while the previously found so-called PI-PLC is PIP 2 -PLC (10,15,35). Moreover, our data indicate that SMSr is a strong candidate for the elusive mammalian PE-PLC and PC-PLC activities, which were first described about 40 years ago (44)(45)(46). Furthermore, we answered another question: why does DGKδ, which does not show selectivity to DG species in vitro (31), apparently utilize limited DG species, 16:0 and/or 16:1containing DG. The intrinsic selectivity of SMSr, which functions upstream of DGKδ, for 16:0 and/or 16:1-containing substrates would determine the limitation of DG species utilized by DGKδ.
The purification of six-transmembrane proteins, SMS and SMSr, is generally difficult (18). However, in order to analyze enzymatic activities of SMSr beyond its previously reported CPES activity, purification was first needed. Thus, we expressed human SMSr using the baculovirus-insect cell system and, through trial and error, highly purified it via affinity and size-exclusion chromatography (Fig. 2, A and B). The purified enzyme gave almost a single band with a molecular mass of 43,000 via SDS gel electrophoresis. It is likely that sizeexclusion chromatography with a buffer containing a nonionic, nondenaturing detergent, Nonidet P-40 (NP-40), at critical micelle concentration (0.018%) was critical for the successful purification.
Using the highly purified enzyme, we demonstrated for the first time that SMSr displayed not only CPES activity but also PAP, PE-PLC, PI-PLC, PC-PLC, and PG-PLC activities in vitro (Fig. 3, A and B). Interestingly, compared with the DGgenerating activity in the presence of PE and ceramide (substrates of CPE synthase), purified human SMSr showed a 340-fold stronger PAP activity, even in the absence of ceramide. Considering the report that SMSr metabolized approximately 300-fold less ceramide than SMS1 (20) (Table 1), it is likely that the PAP activity of human SMSr is comparable with the SM synthase activity of SMS1. Moreover, we determined the K m values for 16:0/18:1-PA and 16:0/18:1-PI of human SMSr (3.92 and 6.83 mol%) (Fig. 3, C and D), which are almost the same as those of other PAPs, LPP1 (PAP2a) (36), and lipin-1 (37) ( Table 2). We demonstrated that Asp348 of SMSr is part of the enzyme's active site, which faces the ER lumen (19,20,47), was important for not only for CPES, but also for PAP and PI-PLC activities (Fig. 6, B and C). Therefore, SMSr probably shows CPES, PAP, PE-PLC, PI-PLC, PC-PLC, and PG-PLC activities through the same catalytic site.
In this study, we found that ceramide selectively affected PE-PLC activity of SMSr-ΔSAMD (Fig. 4A), but not its PAP, PI-PLC, or PC-PLC activity (Fig. 4, B-D). It is possible that PE-PLC activity of SMSr is competitive with CPE synthase activity of SMSr. SMSr has CPE synthase activity and, consequently, ceramide can occupy the substrate (ceramide/PE) binding site of SMSr (20). Therefore, it is likely that d18:1/18:0-ceramide interferes with PE-PLC activity by blocking access of PE to the active site. It was reported that mammalian SMSr acts as not only CPE synthase, but also ceramide sensor in the ER to

SMSr acts as PAP and PLC
protect cells against ceramide-induced mitochondrial apoptosis (20,48). Taken together, SMSr modulates ceramide levels in ER via its SAMD, and vice versa, ceramide may selectively modulate the function (PE-PLC activity) of SMSr at least in part. The PI-PLC, PE-PLC, and PC-PLC activities of purified SMSr were 28100-fold weaker than the PAP activity (Fig. 3A). However, two lines of evidence support that the PI-PLC activity of SMSr is comparable with its PAP activity in cells. First, the amount of PI, PC, and PE in the ER membrane was more than 20-fold, 100-fold, and 40-fold higher than that of PA, respectively (11%, 54%, and 20% versus less than 0.5% of total lipids in the ER membrane, respectively) ( Fig. 8) (49,50). As shown in Figure 3, C and D, SMSr showed a PAP activity of 60 pmol/ mg/min at 0.5 mol% PA and 18 pmol/mg/min at 8 mol% PI. Second, the SMSr expressed in mammalian cells showed a PI-PLC activity as strong as (only 3.5-fold lower) the PAP activity (Fig. 5, B and C). These results suggest that the PI-PLC activity of SMSr expressed in mammalian cells is comparable with its PAP activity. Therefore, PAP and PI-PLC (and PE-PLC and PC-PLC) are likely to represent the substantial activity of SMSr.
What are the differences between SMSr as PAP and the already-identified PAPs (PAP2/LPP and lipin-1) (see Table 2)? LPPs have broad substrate specificity (see Table 2). However, they do not hydrolyze PC, PE, PI, or phosphoinositide (PI(4,5) P 2 ) (11,14,51). Interestingly, SMSr hydrolyzed PA, PI, PI(4,5) P 2 , PE, PG, and PC (Fig. 3, A and B). LPPs contain six transmembrane domains and are primarily localized to the plasma membrane (14,42,51). It was reported that LPP2 and LPP3 were only partly localized to the ER in addition to the plasma membrane (52,53). SMSr is an exclusive ER membrane protein (20,33). Asp348 of SMSr is one of the catalytic triad on the ER luminal side (Fig. 8). It is possible that SMSr as PAP hydrolyzes PA and PI/PC/PE in the ER lumen, in contrast to other PAPs (LPPs and lipins, which are cytosolic proteins).
It has been suggested that PC-specific phospholipase C (PC-PLC) plays pivotal roles in signal transduction pathways that are involved in atherogenesis, inflammation, and carcinogenesis (12,40,59). However, the molecular entity of mammalian PC-PLC has not yet been identified since the activity was reported 40 years ago (44). Moreover, we reported that coimmunoprecipitates using an anti-DGKδ antibody from C2C12 myoblast cells contained PC-PLC activity and that DGKδ interacted with SMSr (30,31). Furthermore, D609, a PC-PLC inhibitor (40) ( Table 2), attenuated the PC-PLC activity of SMSr (Fig. 5E). Therefore, it is likely that SMSr is PC-PLC, which has been awaited for molecular entity identification for quite some time (40 years).
PE-PLC activity in mammalian cells was discovered approximately 30 years ago (16,45,46,60) and is inhibited by D609 (13). Although bacterial and plant PE-PLC has been cloned, mammalian PE-PLC molecules have not yet been A B  In the present study, we demonstrated that SMSr displayed PAP, PI-PLC, PE-PLC, PC-PLC, and PG-PLC activities that were stronger than the CPE synthase activity (20). Moreover, the SMSr exhibited a substrate specificity for SFAand/or MUFA-containing PA and PC molecular species, but not PUFA-containing PA and PC. It is possible that SMSr preferably hydrolyzes SFA-and/or MUFA-containing glycerophospholipids independent of the ceramide. Previously, we reported that SMSr and DGKδ functionally interact via their SAMDs (30). Taken together, it is possible that there is a new mammalian DG-signaling pathway consisting of SMSr and DGKδ independent of the PI(4,5)P 2 cycle.

SMSr acts as PAP and PLC
identified. In the present study, we, for the first time, found the mammalian protein that displays PE-PLC activity. Therefore, it is possible that SMSr is the previously reported PE-PLC (46). We previously reported that in C2C12 myoblasts, DGKδ apparently prefers to phosphorylate palmitic acid (16:0) and/or palmitoleic acid (16:1)-containing DG molecular species, but not 20:4-containing DG delivered from the PI(4,5)P 2 cycle under high-glucose stimulation (31). However, DGKδ enigmatically fails to possess such DG species selectivity in vitro. Intriguingly, we recently found that SMSr interacted with DGKδ via their SAMDs and provided 16:0 and/or 16:1containing DG species to the enzyme (30). However, it remains unclear whether SMSr itself determines DG species that are supplied to DGKδ. In the present study, SMSr preferentially produced 16:0 and/or 16:1-containing DG species in vitro. Therefore, SMSr likely limits DG species utilized by DGKδ. Thus, the molecular machinery of the novel DG supply pathway independent of the PI(4,5)P 2 cycle becomes clearer.
In summary, the current study, for the first time, showed that human SMSr displayed PAP, PI-PLC, PE-PLC, PG-PLC, and PC-PLC activities in vitro, in addition to CPES activity. In particular, PAP (300-fold), PI-PLC (10-fold), PC-PLC (3fold), and PE-PLC activity (4-fold) were stronger than the CPES activity. Thus, beyond the enzyme's previously established CPES activity, SMSr can be categorized as a new type of PLC with a broad head group specificity, i.e., a multiglycerophospholipid PLC hydrolase (MG-PLC). The current results allow us to propose the new DG/PA metabolic pathway "Luminal PA, PI, PC and PE → SMSr (MG-PLC) → DG → flipflop → DGKδ → cytoplasmic PA" in the ER membrane, which is independent of the PI(4,5)P 2 cycle (Fig. 8). Future studies exploring the enzymatic property of SMSr may contribute to understanding of the novel DG/PA metabolic pathway in the ER membrane and the pathogenetic mechanisms of various diseases related to DGKδ, type 2 diabetes (25, 26) and obsessivecompulsive disorder (27,28) and, probably, to PC-PLC, atherogenesis, inflammation, and carcinogenesis (12,40,59). Detergents: n-dodecyl-β-D-maltoside (DDM) was obtained from Cayman Chemical Company. Cholesteryl hemisuccinate (CHS) was purchased from Sigma-Aldrich.
Cell culture COS-7 cells were maintained on 150-mm dishes (Thermo Fisher Scientific) in Dulbecco's Modified Eagle's Medium (DMEM) (Wako Pure Chemicals) containing 10% fetal bovine serum (Thermo Fisher Scientific), 100 units/ml penicillin G (Wako Pure Chemicals), and 100 μg/ml streptomycin (Wako Pure Chemicals) at 37 C in an atmosphere containing 5% CO 2 as described (30). For quantitation of PA and DG levels, 1 × 10 6 cells were plated on 100-mm dishes. After 24 h, plasmid cDNAs were transfected using PolyFect (Qiagen) according to the manufacturer's instruction manual. After transfection, the cells were cultured for an additional 24 h and were used for the experiments.
Sf9 cells were maintained in Sf-900 II serum-free medium (Invitrogen) in sterile Erlenmeyer flask at 120 rpm and 28 C without CO 2 in the dark. Volume of the medium was kept at 20 to 30% of flask volume.
Baculoviral stocks were amplified for seven rounds of amplification at 2 × 10 6 cells/ml, multiplicity of infection (MOI) of 0.1. Viral titer (plaque forming unit (pfu)/ml) was determined via plaque assay as described in flashBAC System Transfection Protocol. Viral stocks were stored at −80 C until use.
To express recombinant proteins, Sf9 cells were seeded at 3 × 10 6 cells/ml and infected cells with recombinant baculovirus at MOI of 2. Infected cells were harvested after 24 h and the pellets were resuspended in 40% (v/v) glycerol diluted in phosphate-buffered saline. The cell samples were flash-frozen in liquid nitrogen and stored at −80 C until use.

Immunoblot analysis
The purified proteins were mixed with a 5×SDS sample buffer and incubated at 37 C for 2 h instead of boiling. The samples were analyzed by SDS-PAGE and an immunoblot analysis, as described (30).

Preparation of DDM/CHS stock solutions
To prepare detergents stock solution (10% (w/v) n-dodecylβ-D-maltoside (DDM) and 2% (w/v) cholesteryl hemisuccinate (CHS)), 5 g of DDM was added into 40 ml of 200 mM Tris-HCl (pH 8.0), followed by gently rotation until DDM goes into solution. The detergent (1 g of CHS) was added to DDM solution and sonicated until solution becomes translucent. After sonication, 10 ml of 200 mM Tris-HCl (pH 8.0) was added and incubated at 25 C with gently rotation until the solution becomes transparent. The detergents solution was stored at −25 C until use.

Preparation of micelles
The reaction solution for PLC and CPES activity assay was prepared as we reported (34). Phospholipids dissolved in chloroform/methanol (2:1 (v/v)) and d18:1/18:0-ceramide in chloroform were dried under N 2 gas to yield a lipid film on the wall of the vial. The lipid film was resuspended in the 2×reaction buffer (40 mM Tris-HCl (pH 7.4), 30 mM KCl, 200 mM NaCl, 600 mM sucrose, 0.2% (w/v) DDM, 0.04% (w/v) CHS) to a final total lipid concentration of 0.1 mM each. The 2×reaction solutions containing lipids were vortexed for 3 min and sonicated in a bath sonicator (Sonifilter model 450) four times for 2 min at 65 C.

Mixed micellar assay
The DG-generating activities (PLC, PAP, or CPES) of purified proteins were evaluated via a DDM mixed micellar phospholipase C assay previously published (66, 67) with a modification. Because DDM/CHS (5:1) micelles provide a more membranelike environment, we adopted DDM/CHS (5:1) for a mixed micellar DG-generating enzyme activity assay.

Statistical analysis
The data are represented as the means ± SD and were analyzed using Student's t-test or one-way ANOVA using GraphPad Prism 8 (GraphPad).

Data availability
All data are contained within the article.  Conflict of interest-The authors declare no conflicts of interest associated with the contents of this article.