The phospholipase PNPLA7 functions as a lysophosphatidylcholine hydrolase and interacts with lipid droplets through its catalytic domain

Mammalian patatin-like phospholipase domain–containing proteins (PNPLAs) are lipid-metabolizing enzymes with essential roles in energy metabolism, skin barrier development, and brain function. A detailed annotation of enzymatic activities and structure–function relationships remains an important prerequisite to understand PNPLA functions in (patho-)physiology, for example, in disorders such as neutral lipid storage disease, non-alcoholic fatty liver disease, and neurodegenerative syndromes. In this study, we characterized the structural features controlling the subcellular localization and enzymatic activity of PNPLA7, a poorly annotated phospholipase linked to insulin signaling and energy metabolism. We show that PNPLA7 is an endoplasmic reticulum (ER) transmembrane protein that specifically promotes hydrolysis of lysophosphatidylcholine in mammalian cells. We found that transmembrane and regulatory domains in the PNPLA7 N-terminal region cooperate to regulate ER targeting but are dispensable for substrate hydrolysis. Enzymatic activity is instead mediated by the C-terminal domain, which maintains full catalytic competence even in the absence of N-terminal regions. Upon elevated fatty acid flux, the catalytic domain targets cellular lipid droplets and promotes interactions of PNPLA7 with these organelles in response to increased cAMP levels. We conclude that PNPLA7 acts as an ER-anchored lysophosphatidylcholine hydrolase that is composed of specific functional domains mediating catalytic activity, subcellular positioning, and interactions with cellular organelles. Our study provides critical structural insights into an evolutionarily conserved class of phospholipid-metabolizing enzymes.

Mammalian patatin-like phospholipase domain-containing proteins (PNPLAs) 2 constitute a family of lipid-metabolizing enzymes with critical roles in energy metabolism, skin barrier development, and brain function (1)(2)(3)(4). The patatin-like phospholipase domain (Pfam01734) is named after a homologous domain in the potato protein patatin, which adopts a threelayer ␣/␤/␣ architecture and harbors an enzymatic active site composed of a Ser-Asp catalytic dyad (1,5). In accordance with a primarily enzymatic function of the patatin-like phospholipase domain, mammalian PNPLAs have been assigned hydrolase, transacylase, or acyltransferase activities with diverse lipid substrates such as phospholipids, acylglycerols, and retinoids (6 -9). These reactions have been proven essential for the turnover of cellular membranes, mobilization of storage lipids, and generation of signaling molecules (10 -12). The physiological relevance of PNPLAs is illustrated by a diverse spectrum of inherited disorders that have been associated with mutations in human PNPLA genes including neutral lipid storage disease, non-alcoholic fatty liver disease, ichthyosis, hereditary spastic paraplegia, and other neurodegenerative syndromes (2,3,13,14). Of note, defects in enzymatic function and subcellular localization have been identified as common molecular mechanisms in the onset and progression of PNPLA-related disorders (4,15,16).
PNPLA6 and PNPLA7 constitute a subgroup within the PNPLA family that has been remarkably conserved during evolution with orthologous proteins in yeast, nematodes, and flies (17). PNPLA6 acts as a (lyso)phospholipase and is involved in the degradation of membrane lipids such as phosphatidylcholine (PC) and lysophosphatidylcholine (LPC) (6,11). Human PNPLA6 is a primary substrate of organophosphates (OPs) that cause a neurotoxic syndrome termed OP-induced delayed neuropathy. Inhibition of PNPLA6 activity and the resultant disruption of neuronal phospholipid homeostasis initiate OP-induced delayed neuropathy. Accordingly, PNPLA6 has been traditionally referred to as neuropathy target esterase (18,19). Recently, mutations in the human PNPLA6 gene have been linked to a complex spectrum of neuroendocrine disorders including ataxia, spastic paraplegia, chorioretinopathy, and hypopituitarism (4,14,20,21). Similar neurodegenerative phenotypes have been observed in PNPLA6 mutant animals from diverse phyla, indicating an evolutionarily conserved role of PNPLA6-mediated phospholipid catabolism in brain function (22)(23)(24).
In contrast to PNPLA6, little is known about the molecular and physiological function(s) of the closely related PNPLA7 (also termed neuropathy target esterase-related esterase). Both proteins share a highly conserved domain architecture that is assembled of the enzymatic patatin-like phospholipase domain and extensive "non-enzymatic" segments of poorly defined function including three putative cyclic nucleotide monophosphate (cNMP)-binding sites (1,25). Initial in vitro studies identified PNPLA7 as a lysophospholipase whose transcript expression is highly responsive to feeding/fasting transitions and insulin concentrations (25,26). In mammalian cells, PNPLA7 localizes to the ER and lipid droplets (LDs), which are cellular lipid storage organelles with pivotal functions in energy metabolism and lipid trafficking (25,27). Although these observations closely link PNPLA7 to lipid and energy metabolism, it is presently unknown how fluctuations in PNPLA7 expression or subcellular distribution affect lipid homeostasis of cells or tissues (25).
In this study, we further characterized the enzymatic function of PNPLA7 in cellular lipid metabolism and established detailed structure-function relationships among domain architecture, subcellular positioning, and enzymatic activity of the protein. We confirm that PNPLA7 acts as a potent intracellular lysophospholipase and identify LPC as a major substrate of PNPLA7 in living cells. Moreover, we demonstrate that PNPLA7 is composed of specific functional parts mediating ER targeting, interactions with LDs, and substrate hydrolysis. Our study provides novel structural insights into an evolutionarily conserved class of phospholipid-metabolizing enzymes.

PNPLA7 expression affects lysophospholipid metabolism in mammalian cells
To assess possible function(s) of PNPLA7 in cellular lipid metabolism, we first created cell lines stably expressing PNPLA7-EGFP or EGFP, respectively, and subjected them to enzyme activity assays and lipid analysis. Consistent with previous studies, cell homogenates expressing PNPLA7-EGFP exhibited increased hydrolytic activity toward several lysophospholipid species as compared with EGFP-expressing control homogenates including C18:1 LPC (4.6-fold), C18:1 lysophosphatidylethanolamine (LPE; 4.0-fold), and C18:1 lysophosphatidylserine (LPS; 2.7-fold) (25). In addition, PNPLA7-EGFP-expressing homogenates showed a low but significant increase in the hydrolytic activity toward C18:1/C18:1 PC (1.5fold) and C18:1/C18:1 phosphatidylethanolamine (PE; 1.2fold), whereas the activity toward C18:1/C18:1 phosphatidylserine (PS) was not different from EGFP-expressing controls (Fig. 1A). The lipid hydrolase activity of PNPLA7-EGFP was 2.6-fold higher toward C18:1 LPC as compared with C18:0 LPC, suggesting a preference of the enzyme for unsaturated compared with saturated species (Fig. 1B). Lipid analysis revealed that stable expression of PNPLA7-EGFP significantly reduced cellular LPC levels (Ϫ34%) as compared with EGFP-expressing control cells but did not affect total cellular levels of LPE, LPS, PC, PE, or PS (Fig. 1C). The decrease in LPC levels observed in PNPLA7-EGFP-expressing cells was mainly due to a reduction in unsaturated LPC species containing C16:1 (Ϫ35%), C18:1 (Ϫ43%), and C18:2 (Ϫ52%) fatty acids (FAs) (Fig. 1D). Although a trend toward decreased cellular levels was also observed for other LPC species, the difference was not significant when compared with controls (Fig. 1D). To further investigate the function of PNPLA7 in cellular LPC metabolism, we depleted endogenous PNPLA7 in murine AML12 cells by RNAi and assessed cellular LPC hydrolase activities and LPC levels. Stable expression of two independent shRNA constructs targeting PNPLA7 (shRNA1 and shRNA2) decreased cellular Pnpla7 A and B, lipid hydrolase activities of PNPLA7-EGFP. PNSs of COS-7 cells stably expressing PNPLA7-EGFP or EGFP were incubated with glycerophospholipid substrates containing C18:1 or C18:0 FAs as indicated, and the release of FAs was determined. C and D, lipid levels of COS-7 cells stably expressing PNPLA7-EGFP or EGFP. Total lipids were extracted, and the cellular levels of the indicated lipids were analyzed by LC/MS. Data are representative of two independent experiments and are expressed as means (n ϭ 3). Error bars represent S.D. Statistical significance was determined using Student's unpaired t test. *, p Ͻ 0.05. AU, arbitrary units.

Functional domains of PNPLA7
mRNA concentrations by 59 and 63%, respectively, as compared with cells expressing scrambled shRNA ( Fig. 2A). This was accompanied by decreased PNPLA7 protein levels in cells expressing either shRNA1 or shRNA2, whereas expression of the closely related PNPLA6 was unchanged (Fig. 2B). The reduction in PNPLA7 expression was associated with a 16 -20% decrease in cellular LPC hydrolase activity (depending on the shRNA used; Fig. 2C) but did not affect total cellular LPC levels (Fig. 2D). Taken together, these data provide evidence that PNPLA7 acts as an intracellular LPC-preferring lysophospholipase that contributes to cellular LPC hydrolase activity in mammalian cells.

PNPLA7 is an integral membrane protein with a luminal N and a cytosolic C terminus
To link the lipid-metabolizing function of PNPLA7 to a specific cellular site, we next assessed its subcellular localization and the topological orientation of its domains. The arrangement of possible functional domains within the PNPLA7 polypeptide is shown in Fig. 3A. The N-terminal half of the protein includes a single predicted transmembrane (TM) domain close to the N terminus and a putative "regulatory" (R) domain, which harbors three predicted cNMP-binding domains (CBDs). The C-terminal half (C domain) includes the patatin-like phospholipase domain predicted to mediate enzymatic activity (Fig.  3A). Subcellular fractionation of Neuro-2a postnuclear supernatants (PNSs) and subsequent immunoblotting showed that endogenous PNPLA7 was enriched in the membrane fraction but undetectable in the cytosol (Fig. 3B). Likewise, a PNPLA7-EGFP fusion protein transiently expressed in COS-7 cells was exclusively recovered in the membrane fraction (Fig. 3B), indicating a similar subcellular distribution of endogenous PNPLA7 and ectopically expressed PNPLA7-EGFP. Membrane-associated PNPLA7-EGFP was partially solubilized with Triton X-100 but not with NaCl or Na 2 CO 3 , similar to the integral membrane protein Calnexin but unlike the peripheral membrane protein DDHD2 (28,29) (Fig. 3C). Next, we used proteinase K protection assays to assess the topological orientation of PNPLA7 within cellular membranes. Incubation of isolated microsomes with increasing concentrations of proteinase K provoked a dose-dependent depletion of the signal corresponding to full-length PNPLA7-EGFP in our immunoblots (Fig. 3D). This was accompanied by the generation of a ϳ25-kDa fragment likely representing EGFP, which is known to be resistant to proteinase K digestion, indicating proteolytic degradation and therefore cytosolic orientation of the C terminus of PNPLA7 (Fig. 3D). Using antibodies directed toward the C domain or the R domain of PNPLA7, we failed to detect additional protein bands eluding proteolytic digestion of PNPLA7-EGFP, suggesting that these domains are also largely exposed to the cytosol and thus accessible to proteolytic degradation (Fig.  3D). Integrity of the isolated microsomes was confirmed by the apparent protection of the 66-kDa luminal domain of Calnexin from proteolytic degradation (28) (Fig. 3D). To address whether the N terminus of PNPLA7 is oriented toward the cytosol or the lumen, we repeated the proteinase protection assay with microsomes isolated from COS-7 cells expressing PNPLA7 harboring an N-terminal HA tag (HA-PNPLA7). As shown in Fig. 3E, proteinase K treatment depleted the signal corresponding to full-length HA-PNPLA7 and led to the accumulation of an immunoreactive peptide of less than 10 kDa consistent with a luminal orientation of the N terminus. Addition of Triton X-100 to solubilize microsomal membranes rendered this peptide susceptible to proteolytic digestion, suggesting that microsomal integrity is indeed required to protect the N-terminal part of PNPLA7 from proteolysis. In an independent approach, we permeabilized COS-7 cells expressing HA-PNPLA7 either with Triton X-100 or digitonin and costained the cells with antibodies directed toward the HA tag and the C domain (Fig. 3F). Permeabilization with Triton X-100 rendered both epitopes accessible to antibody staining, resulting in a reticular staining pattern reminiscent of the ER and an apparent overlap of both signals. In contrast, digitonin treatment, which is used to selectively permeabilize the plasma membrane, rendered only the C region but not the N-terminal HA tag accessible to antibody staining (Fig. 3F). Taken together these results suggest that PNPLA7 is a bilayer-spanning TM protein that orients its N terminus toward the lumen but exposes its R and C domains largely toward the cytosol.

Functional domains of PNPLA7 The N-terminal region targets PNPLA7 to the ER
We next addressed the functional contribution of individual protein domains to the subcellular positioning of PNPLA7. To this end, we constructed EGFP-tagged PNPLA7 variants lacking individual domains (Fig. 4A) and assessed their subcellular localization in living COS-7 cells by confocal fluorescence microscopy. Consistent with the results described above, ectopically expressed full-length PNPLA7-EGFP exhibited a reticulate pattern that colocalized with an ER marker protein in COS-7 cells (Fig. 4B). A PNPLA7 mutant lacking the C domain (PNPLA7-N-EGFP) exhibited a similar subcellular distribution as full-length PNPLA7-EGFP and colocalized with the ER, indicating that the C domain is dispensable for ER association of PNPLA7. Furthermore, sole expression of the N-terminal TM domain (PNPLA7-TM-EGFP) was sufficient to target EGFP to the ER, whereas deletion of the TM domain (PNPLA7-⌬TM-EGFP) compromised ER localization of PNPLA7, suggesting an essential requirement of the TM domain for ER targeting of PNPLA7. Notably, expression of a construct harboring an internal deletion of the R but not the TM domain resulted in aggregation of the recombinant protein that failed to properly distribute to the ER network. This suggests that both the TM Figure 3. Domain architecture, subcellular localization, and membrane topology of PNPLA7. A, predicted domains of PNPLA7. The N-terminal region of PNPLA7 comprises a putative transmembrane domain (black) and three putative CBDs, which are designated as the R domain. The C domain contains a patatin domain implicated in the enzymatic function. B, distribution of endogenous PNPLA7 and ectopically expressed PNPLA7-EGFP in subcellular fractions. PNSs of Neuro-2a cells and COS-7 cells expressing PNPLA7-EGFP were fractionated into cytosol and membranes, and the distribution of proteins was assessed by immunoblotting using antibodies against the R domain of PNPLA7, GFP, IRE1␣, Calnexin, and GAPDH. C, membrane association of PNPLA7-EGFP. Membranes of COS-7 cells expressing PNPLA7-EGFP were treated with the indicated agents, and soluble (S) and pellet (P) fractions were obtained by ultracentrifugation and analyzed by immunoblotting using antibodies against GFP, Calnexin, and DDHD2. D and E, sensitivity of PNPLA7 epitopes to protease treatment. Membrane fractions of COS-7 cells expressing PNPLA7-EGFP (D) or HA-PNPLA7 (E) were incubated in the presence of 0 -10 g/ml proteinase K (D) or 10 g/ml proteinase K (PK) and 1% Triton X-100 (TX) as indicated (E) and subjected to immunoblotting using antibodies against GFP, the R domain of PNPLA7, the C domain of PNPLA7, HA, or Calnexin as indicated. F, immunostaining of PNPLA7 epitopes in COS-7 cells. COS-7 cells expressing HA-PNPLA7 were fixed, permeabilized with Triton X-100 or digitonin as indicated, and costained with antibodies directed toward the N-terminal HA tag and the C domain of PNPLA7. Scale bars, 10 m.

Functional domains of PNPLA7
domain and the R domain contribute to the positioning of fulllength PNPLA7 at the ER. Consistent with these findings, a mutant lacking both the TM and R domains (PNPLA7-C-EGFP) distributed diffusely in the cytoplasm and showed only minimal overlap with the ER marker (Fig. 4B). As a complementary approach, we used subcellular fractionation and immunoblotting to assess membrane association of PNPLA7 mutant proteins. As shown in Fig. 4C, PNPLA7-EGFP, PNPLA7-N-EGFP, PNPLA7-TM-EGFP, and PNPLA7-⌬R-EGFP were recovered exclusively in the membrane fraction, whereas PNPLA7-⌬TM-EGFP and PNPLA7-C-EGFP distributed between the soluble and particulate fractions, suggesting impaired membrane association of these mutants. Thus, our data suggest that the N-terminal TM domain mediates efficient membrane tethering of PNPLA7 and that both the TM and R domains but not the C domain control the subcellular positioning of PNPLA7 at the ER.

The C domain of PNPLA7 associates with LDs
In addition to the ER, PNPLA7 has been shown to localize to a subset of cellular LDs upon incubation of cells with FAs (25). To address the structural requirements for LD association of PNPLA7, we induced LD formation in COS-7 cells transfected with EGFP-tagged PNPLA7 or PNPLA7 truncation mutants and reassessed subcellular localization of the recombinant proteins by confocal fluorescence microscopy. PNPLA7-EGFP exhibited a reticular distribution and only marginal overlap with LDs, suggesting that PNPLA7 remains predominantly ERassociated also upon increased LD formation (Fig. 5A). Likewise, neither PNPLA7-N-EGFP nor PNPLA7-⌬TM-EGFP exhibited apparent colocalization with LDs under these conditions. In contrast, PNPLA7-⌬R-EGFP showed punctate and semi-ringlike enrichments in close proximity to LDs, indicating a possible interaction of this truncation mutant with LDs. This phenomenon was even more pronounced for PNPLA7-C-EGFP, which presented as numerous ring-shaped structures surrounding cellular LDs (Fig. 5A). Moreover, PNPLA7-C-EGFP colocalized with ectopically expressed PLIN2-mCherry, a bona fide LD protein, suggesting that it indeed associates with the LD surface (30, 31) (Fig. 5B).
To further study the interactions of PNPLA7 protein domains with LDs, we fractionated transfected COS-7 cells by ultracentrifugation and assessed the abundance of each PNPLA7 variant in the LD fraction by immunoblotting. Consistent with the microscopy data, ectopically expressed PNPLA7-C-EGFP was markedly enriched in the LD fraction, resembling the distribution of endogenous PLIN2 (Fig. 5C). Low levels of PNPLA7-EGFP, PNPLA7-⌬TM-EGFP, and PNPLA7-⌬R-EGFP were also detected in the LD fraction. However, the abundance of these proteins was much lower as compared with PNPLA7-C-EGFP.
To identify the minimal domain required for LD targeting of PNPLA7-C-EGFP, we further truncated the protein and assessed the ability of these mutants to colocalize with LDs. Schematic drawings of the truncations and their effects on LD targeting are summarized in Fig. 6A. Mutants containing truncations at the C terminus retained the ability to localize to LDs, suggesting that the patatin-like phospholipase domain is largely dispensable for LD targeting. In contrast, even minor trunca-

Functional domains of PNPLA7
tions at the N terminus of the C domain compromised LD targeting. As shown in Fig. 6, A and B, a stretch between amino acids 681 and 967 was sufficient for LD targeting, whereas further N-or C-terminal truncations abolished LD association of the C domain. Thus, a stretch of 287 amino acids is essential for the interactions between the C domain of PNPLA7 and LDs.
Because the N-terminal region contains three putative CBDs (Fig. 7A), we next asked whether the presence of cNMP analogs would affect the subcellular distribution of PNPLA7. Incubation of cells with 8-CPT-cAMP provoked a subcellular redistribution of ectopically expressed PNPLA7-EGFP, which manifested as ring-and semi-ringlike enrichments of the protein close to the LD surface on top of its reticulate ER-like distribution (Fig. 7B, inset). A similar response was elicited by 8-CPT-cGMP (data not shown). In contrast to PNPLA7-EGFP, PNPLA7-N-EGFP was not enriched at LDs under identical conditions. Furthermore, deletion of CBD3 but not CBD1 or CBD2 abolished LD interactions of PNPLA7-EGFP (Fig. 7, A  and B). In summary, these data show that the C domain localizes to cellular LDs and suggest an involvement of both the C domain and CBD3 in LD interactions of PNPLA7 in response to elevated cNMP levels. Noteworthy, the specific C18:1 LPC hydrolase activity of PNPLA7 was largely unchanged upon incubation of cells with FAs and/or cNMPs (Fig. 7C) or starvation (data not shown), suggesting that alterations in subcellular distribution do not alter PNPLA7 activity.

The C domain mediates catalytic activity of PNPLA7
We finally aimed to understand how specific protein domains regulate the enzymatic function of PNPLA7. To do so, we expressed PNPLA7-EGFP or truncated PNPLA7-EGFP mutant proteins in COS-7 cells and compared cellular lysophospholipase activities. Immunoblotting revealed similar expression levels of PNPLA7-EGFP and truncated versions of the protein (Fig. 8A). As shown in Fig. 8B, ectopic expression of PNPLA7-EGFP or PNPLA7-C-EGFP increased cellular LPC hydrolase activity to a similar extent (2.1-2.3-fold) as compared with EGFP-expressing controls, suggesting that the C domain maintains full catalytic competence also in the absence of TM and R domains. The specific activities of PNPLA7-⌬TM-EGFP and PNPLA7-⌬R-EGFP were 3.0-and 3.3-fold higher than the activity of EGFP-expressing controls, thereby moderately exceeding the activity of PNPLA7-EGFP. In contrast, expression of PNPLA7-N-EGFP failed to increase cellular lysophospholipase activity, which is consistent with a requirement of the C domain for substrate hydrolysis (Fig. 8B). To assess whether the catalytic C domain is functional also in intact cells, we generated COS-7 lines stably expressing PNPLA7-C-EGFP or EGFP, respectively, and performed lipid analyses. Expression of PNPLA7-C-EGFP reduced cellular levels of LPC by 42% com-

Functional domains of PNPLA7
pared with EGFP-expressing control cells but did not affect cellular levels of PC, PE, PS, LPE, and LPS (Fig. 8C). Thus, the C domain is both required and sufficient to mediate catalytic activity of PNPLA7 in mammalian cells.

Discussion
In this study, we assessed the role of PNPLA7 in cellular lipid metabolism and analyzed the structure-function relationships between domain architecture, subcellular distribution, and enzymatic activity of the protein. Ectopic expression of PNPLA7 markedly increased cellular hydrolase activities toward lysophospholipids including LPC, LPE, and LPS. In contrast, phospholipids with two acyl chains were poor substrates for PNPLA7, suggesting a preference of the enzyme for mono-over diacylglycerophospholipids. In line with this observation, stable expression of PNPLA7 in COS-7 cells decreased cellular levels of the monoacylglycerophospholipid LPC but not of diacylglycerophospholipids such as PC, PE, and PS. Despite increased LPE and LPS hydrolase activities, the cellular levels of these lysophospholipid classes were also not altered upon PNPLA7 expression, arguing for a preference of PNPLA7 for LPC in a cellular context. Moreover, stable reduction of PNPLA7 expression by RNAi decreased cellular LPC hydrolase activity in AML12 cells, further supporting an enzymatic function of PNPLA7 as LPC hydrolase. PNPLA7 is closely related to PNPLA6, a protein involved in axon maintenance and brain function, which has been previously shown to regulate hydro-

Functional domains of PNPLA7
lysis of cellular LPC and PC (11,14,22). Our data classify PNPLA7 also as LPC hydrolase, suggesting a function in PC/LPC turnover similar to that of PNPLA6. PC is the most abundant glycerophospholipid in mammalian membranes and serves a plethora of functions including maintenance of membrane structure, vesicle transport, and lipid signaling (32)(33)(34). Cellular PC levels are tightly regulated by synthetic and degradative processes (34). In addition, PC species composition is continuously adjusted by a remodeling process termed Lands' cycle that involves deacylation of PC to LPC and subsequent reacylation of LPC by acyltransferases (35). PNPLA6 and PNPLA7 may thus affect both bulk PC/LPC degradation and PC remodeling in the Lands' cycle by controlling the availability of LPC precursors. The striking similarities in the enzymatic activities of PNPLA6 and PNPLA7 raise the question as to what extent both enzymes fulfill distinct or redundant roles in a physiological context. Stable knockdown of PNPLA7 expression did not provoke an accumulation of cellular LPC in AML12 cells. This is likely due to the presence of additional LPC hydrolases that contribute substantial "PNPLA7-independent" LPC hydrolase activity. Notably, both PNPLA6 and PNPLA7 are expressed in AML12 cells, raising the possibility that these enzymes act in redundancy to control LPC catabolism in this particular cell line. Preliminary data from our laboratory suggest differential expression of PNPLA6 and PNPLA7 in various murine cell lines and tissues, suggesting that the contribution of each enzyme to cellular LPC hydrolysis depends on the specific cellular context. Moreover, the neurodegenerative phenotypes provoked by PNPLA6 deficiency in humans and mice that cannot be compensated by PNPLA7 argue against a redundant function of both proteins in vivo (14,22,36). Noteworthy, PNPLA7 is expressed at high levels in several non-neuronal tissues such as muscle, adipose tissue, and testis (25). Thus, PNPLA6 and PNPLA7 may perform similar biochemical functions but act in different cell types or tissues.
We initially described PNPLA7 as a lipid hydrolase with a dual subcellular localization at the ER and LDs (25). Our new data show that PNPLA7 interacts with both organelles through distinct domains. ER targeting of PNPLA7 relies on a single bilayer-spanning TM domain close to the N terminus, which is both required and sufficient for ER membrane association. In the context of the full-length protein, also the R domain contributes to the ER positioning of PNPLA7 as its deletion results in protein aggregation and perturbed distribution within the ER network. The interaction of PNPLA7 with LD does not require N-terminal ER targeting information but is instead mediated by a 287-amino acid region within its C domain. Sole expression of this region conferred LD targeting information to EGFP, whereas loss of the C domain precluded LD interactions of PNPLA7. Notably, the isolated C domain exhibited higher affinity to LDs than full-length PNPLA7. Upon increased FA supply, the C domain translocated to the LD surface, whereas full-length PNPLA7 largely retained its ER-bound configuration. Unlike the C domain, full-length PNPLA7 only partially redistributed toward LDs under specific conditions, e.g. in response to cNMP analogs. It is reasonable to assume that the bilayer-spanning topology of full-length PNPLA7 firmly anchors the protein within the ER and prevents translocation of the C domain to the LD surface. The domain architecture of full-length PNPLA7 may thus restrict LD interactions to specific stimuli (e.g. cNMP) and/or a specific subset of LDs (e.g. ER-associated or "nascent" LDs).
Additional structure-function analyses revealed that the C domain constitutes the major catalytic part of PNPLA7 and thus serves a dual role in subcellular localization and substrate hydrolysis. Expression of full-length PNPLA7 or the isolated C domain increased cellular lysophospholipase activity to a similar extent, suggesting that N-terminal domains neither restrict nor promote catalytic activity of PNPLA7. Consistent with this finding, it has been shown that the C-terminal esterase domain of the homologous human protein PNPLA6 acquires catalytic competence also in the absence of N-terminal regions (37). Sole expression of the C domain was sufficient to provoke a decrease in cellular LPC levels, suggesting enzymatic activity of this domain also in living cells. The combination of LD targeting information and enzymatic activity within the C domain indicates that PNPLA7 may directly act at the LD surface to degrade excess (lyso)phospholipids or antagonize phospholipid (re)synthesis. Although the functional relevance of this reaction is currently unknown, it is possible that (lyso)phospholipid catabolism contributes to distinct aspects of LD biology such as "budding" of LDs from the ER membrane or degradation of the LD monolayer in times of increased LD catabolism. In summary, our study annotates PNPLA7 as an intracellular LPC hydrolase with specific functional domains that coordinate substrate hydrolysis, ER targeting, and interactions with LDs.

Functional domains of PNPLA7
ing Technology (Danvers, MA). Antisera against the R domains of murine PNPLA7 and PNPLA6 were generated by repetitive immunizations of rabbits with the peptides CSVPLPSNH-GEVDELRQSQGSGSNT and PAGDPVKPTSLEAPPAPLL-SRC, respectively, which were coupled to keyhole limpet hemocyanin as carrier (PiChem, Graz, Austria). An ECL TM sheep anti-mouse IgG antibody (NA931) linked to horseradish peroxidase (HRP) was purchased from GE Healthcare. An HRPlinked goat anti-rabbit IgG antibody (PI-1000) was purchased from Vector Laboratories (Burlingame, CA). An HRP-linked goat anti-guinea pig IgG antibody (6090-04) was purchased from Southern Biotechnology (Birmingham, AL). A Rhodamine Red TM -X-linked goat anti-mouse IgG antibody was obtained from Jackson ImmunoResearch Laboratories (West Grove, PA), and DyLight 488-conjugated goat anti-rabbit IgG antibody was obtained from Thermo Fisher Scientific.

Protein sequence analysis
Protein domains were analyzed using the protein family database (Pfam), an online database containing collections of protein domains and families (38). TM domains were predicted using Hidden Markov Models (39).

Generation of lentivirus and stable infection of COS-7 and AML12 cells
Lentiviral particles harboring pLVX IRES Puro vectors for the expression of EGFP-tagged proteins or pLKO.1 vectors for the expression of shRNAs were generated in HEK293T cells (ATCC CRL-3216) according to the manufacturer's instructions (Clontech). Before transduction, COS-7 or AML12 cells were seeded into 6-well plates at a density of 300,000 cells/well. Cells were incubated for 24 h with lentivirus-containing supernatants in the presence of 8 g/ml Polybrene. To select for stable expression, cells were maintained for 7 days in medium containing 2 g/ml puromycin.

Subcellular fractionation
COS-7 cells were washed with phosphate-buffered saline (PBS), scraped from tissue culture plates, and collected by brief centrifugation. For the preparation of membrane and cytosol, cells were homogenized in 10 mM HEPES, 0.25 M sucrose, 1 mM EDTA, pH 7.4 (buffer A), by passing 30 times through a 26-gauge needle on ice. PNSs were obtained by centrifugation at 4°C and 1,000 ϫ g for 10 min. PNS was further fractionated into cytosol and membrane fractions by ultracentrifugation at 4°C and 100,000 ϫ g for 60 min. For the isolation of lipid droplets, cells were homogenized on ice with a Dounce homogenizer in 20 mM Tris/HCl, 1 mM EDTA, pH 7.4, and lipid droplets were isolated by ultracentrifugation using a discontinuous sucrose gradient according to Brasaemle and Wolins (42) with minor modifications. Protein concentrations of cell extracts were determined with the Bio-Rad Protein Assay kit according to the manufacturer's instructions (Bio-Rad) using BSA as standard. Alternatively, protein concentrations were determined with the Pierce BCA TM Protein Assay kit according to the manufacturer's protocol (Thermo Scientific).

Membrane extraction and proteinase K protection assays
Membrane fractions of COS-7 cells were resuspended in buffer A and incubated in the presence of 1 M NaCl, 0.1 M Na 2 CO 3 , or 1% Triton X-100 for 30 min on ice. Afterward, soluble and insoluble components were separated by centrifugation at 4°C and 100,000 ϫ g for 30 min and subjected to immunoblotting. For membrane protection assays, isolated membrane fractions were incubated in the presence of 0 -10 g/ml proteinase K in the absence or presence of 1% Triton X-100 for 15 min at room temperature. The reactions were stopped by the addition of PMSF at a final concentration of 2 mM, further incubated for 5 min, and subjected to immunoblotting.

Immunoblotting
Samples were mixed with Laemmli buffer, denatured for 5 min, subjected to SDS-PAGE, and electroblotted onto PVDF membranes. Membranes were incubated for 1 h with 20 mM Tris/HCl, 150 mM NaCl, 0.1% Tween 20, pH 7.5 (TBST), containing 10% milk to block unspecific binding sites. Primary and secondary antibodies were diluted in TBST containing 5% milk and incubated with the membrane for 1 h at room temperature followed by extensive washing with TBST.

RNA extraction and RT-qPCR
Cells were harvested and snap frozen in liquid nitrogen. RNA was extracted using the TRIzol reagent (Invitrogen TM , Life Technologies) according to the manufacturer's instructions. RT-qPCR was performed as described previously (43) using the StepOnePlus TM Real-Time PCR System (Applied Biosystems, Life Technologies) and the Maxima TM SYBR Green/ROX PCR reaction mixture (Thermo Scientific Fermentas, Waltham, MA). The following primers were used: Pnpla7 fw, 5Ј-CGT GTT TTC CAA CGA CCA CC-3Ј; Pnpla7 rv, 5Ј-TCT GCT AGT GCC CTG AGG AT-3Ј; 36B4 fw, 5Ј-GCT TCA TTG TGG GAG CAG ACA-3Ј; 36B4 rv, 5Ј-CAT GGT GTT CTT GCC CAT CAG-3Ј. Relative mRNA levels were quantified according to the ⌬⌬Ct method using 36B4 as reference gene.

Lipid hydrolase assays
Protein samples for lipid hydrolase assays were prepared by sonication of COS-7 or AML12 cells in 0.25 M sucrose, 1 mM EDTA, 1 mM DTT containing 20 g/ml leupeptin, 2 g/ml antipain, and 1 g/ml pepstatin (solution A) followed by centrifugation at 4°C and 1,000 ϫ g. Lipid substrates were prepared by sonication of glycerophospholipids in 200 mM Bistris propane buffer, pH 7, containing 10 mM CHAPS, 1 mM EDTA, and 600 mM NaCl. Unless otherwise indicated, substrates contained lysophospholipids at final concentrations of 3 mM or diacylglycerophospholipids at final concentrations of 1 mM. Assays were started by the addition of 50 l of substrate to 50-l samples in solution A and incubated under steady shaking at 37°C in a water bath. Substrate hydrolysis is expressed as release of

Functional domains of PNPLA7
FAs, which was determined routinely using a commercially available kit (non-esterified FA (NEFA) reagent, Wako, Neuss, Germany) according to the manufacturer's instructions. For the determination of endogenous LPC hydrolase activities in AML12 cells, 3 H-labeled LPC at a specific activity of 0.33 Ci/ mol was included in the substrate. After incubation, assays were extracted with 0.5 ml of chloroform/methanol/acetic acid (66:33:1, v/v/v), and extracts were separated by thin layer chromatography using Silica G plates and hexane/diethyl ether/acetic acid (70:30:1, v/v/v) as solvent. FAs were identified by comigrating authentic standards, and the associated radioactivity was determined by liquid scintillation counting. All assays were performed in triplicates.

Immunostaining and confocal fluorescence microscopy
COS-7 cells were seeded in chambers mounted onto coverslips (Sarstedt, Nümbrecht, Germany) and transfected as described above. Cells expressing EGFP-tagged proteins were fixed with 4% paraformaldehyde for 20 min at room temperature, and LDs were counterstained using HCS LipidTOX Deep Red. Cells expressing HA-PNPLA7 were fixed with 4% paraformaldehyde for 20 min at room temperature and permeabilized for 15 min with PBS containing either 0.1% Triton X-100 or 0.0005% digitonin. Cells were washed three times with PBS and incubated at room temperature with PBS containing 1% BSA for 1 h to block unspecific binding. Primary antibodies were diluted in blocking solution and incubated with cells overnight at 4°C. Cells were washed with PBS and incubated with secondary antibodies diluted in blocking solution at room temperature for 1 h followed by extensive washing with PBS. Cells were imaged using a Leica SP5 confocal microscope equipped with a Leica HCX 63ϫ 1.4 numerical aperture oil immersion objective. EGFP and DyLight 488-conjugated antibodies were excited at 488 nm, and emission was detected between 500 and 530 nm. DsRed2, mCherry, and Rhodamine Red were excited at 561 nm, and emission was detected between 580 and 610 nm. HCS LipidTOX Deep Red was excited at 633 nm, and emission was detected between 650 and 700 nm.

Statistical analysis
All measurements were performed in triplicates. Data are presented as means Ϯ S.D. Statistical significance was determined by the Student's unpaired t test. Group differences were considered statistically significant for p Ͻ 0.05 (*).