Human Molecular Genetics, 2003, Vol. 12, No. 22 2981–2991 DOI: 10.1093/hmg/ddg321 NSDHL, an involved in biosynthesis, traffics through the Golgi and accumulates on ER membranes and on the surface of lipid droplets

Hugo Caldas1 and Gail E. Herman1,2,* Downloaded from https://academic.oup.com/hmg/article/12/22/2981/606598 by guest on 27 September 2021 1Center for Molecular and Human Genetics, Columbus Children’s Research Institute, Columbus, OH 43205, USA and 2Department of Pediatrics, The Ohio State University, Columbus, OH 43205, USA

Received July 15, 2003; Revised August 29, 2003; Accepted September 11, 2003

NSDHL, for NAD(P)H steroid dehydrogenase-like, encodes a sterol dehydrogenase or decarboxylase involved in the sequential removal of two C-4 methyl groups in post-squalene cholesterol biosynthesis. Mutations in this are associated with human CHILD syndrome (congenital hemidysplasia with ichthyosiform nevus and limb defects), an X-linked, male lethal disorder, as well as the mouse mutations bare patches and striated. In the present study, we have investigated the subcellular localization of tagged encoded by wild-type and selected mutant murine Nsdhl alleles using confocal microscopy. In addition to an ER localization commonly found for of post-squalene cholesterol biosynthesis, we have identified a novel association of NSDHL with lipid droplets, which are endoplasmic reticulum (ER)-derived cytoplasmic structures that contain a neutral lipid core. We further demonstrate that trafficking through the Golgi is necessary for ER membrane localization of the and propose a model for the association of NSDHL with lipid droplets. The dual localization of NSDHL within ER membranes and on the surface of lipid droplets may provide another mechanism for regulation of the levels and sites of accumulation of intracellular cholesterol.

INTRODUCTION The 27-carbon cholesterol molecule is synthesized from acetate in a series of approximately 30 enzymatic reactions. Cholesterol is an essential component of the membranes of The first sterol intermediate in the pathway, lanosterol, is animal cells and a major determinant of membrane fluidity. In formed by the condensation of squalene, a 30-carbon isopre- addition, cholesterol is the substrate for synthesis of steroid noid. This and subsequent reactions in ‘post-squalene’ hormones and bile acids. Isoprenoid intermediates in early cholesterol biosynthesis are known, or believed, to occur in steps of the cholesterol biosynthetic pathway are precursors for the membranes of the endoplasmic reticulum (ER), while the synthesis of isopentyl tRNAs for protein synthesis; dolichol earlier steps in the biosynthetic pathway occur in the cytosol, for N-linked glycosylation of proteins; ubiquinone and heme A peroxisome and/or ER membrane (4–6). The compartmentali- that are cofactors for respiratory chain enzymes in the zation of enzymes involved in the early steps of the cholesterol mitochondria; and farnesyl and geranylgeranyl ligands that biosynthetic pathway may provide a mechanism to help result in the prenylation of key cellular proteins and anchor regulate its synthesis. them to cell membranes. Cholesterol and selected sterol inter- Nsdhl [NAD(P)H steroid dehydrogenase-like] encodes a mediates can be converted to oxysterols that act as regulatory 3b-hydroxysterol dehydrogenase that was identified as part of a signaling molecules and can bind to orphan nuclear receptors positional cloning effort; mutations in the murine gene are including LXRa (reviewed in 1,2). Finally, hedgehog signaling responsible for the X-linked dominant, male lethal mouse proteins that are involved in many important developmental mutations bare patches (Bpa) and striated (Str) (7). Bpa females pathways are modified by the addition of cholesterol in auto- have a skeletal dysplasia and are dwarfed compared with processing reactions that are required for their proper function normal littermates. They develop a hyperkeratotic skin eruption in vivo (reviewed in 3). on postnatal day 5–7 that resolves producing a striping of the

*To whom correspondence should be addressed at: Columbus Children’s Research Institute, 700 Children’s Drive, Room W403, Columbus, OH 43205, USA. Tel: þ1 6147222848/9; Fax: þ1 6147222817; Email: [email protected]

Human Molecular Genetics, Vol. 12, No. 22 # Oxford University Press 2003; all rights reserved 2982 Human Molecular Genetics, 2003, Vol. 12, No. 22 adult coat consistent with random X-inactivation (8, reviewed To examine the subcellular localization of the protein, GFP- in 9). Milder Bpa alleles, originally thought to be a distinct tagged NSDHL was co-expressed transiently in COS7 cells locus called striated (Str) (10), are indistinguishable from with an ER marker, DsRed2-ER (Fig. 1). NSDHL was detected normal littermates until day 12–14 when the striped coat is first in the ER as evidenced by the complete overlap of the green apparent. Mutations in Nsdhl have been identified in seven Bpa (NSDHL) and red (DsRed2-ER) signals (Fig. 1F), while the and Str alleles, including a nonsense mutation (K103X) in the GFP vector alone produced diffuse nuclear and cytoplasmic original Bpa1H mutant mouse that produces a truncated protein staining (Fig. 1A). Similar results were obtained using that lacks the putative catalytic site and is predicted to be a null N-terminal FLAG tag and C-terminal myc tag NSDHL fusion allele (7,11). Heterozygous mutations in the human NSDHL proteins, demonstrating that the N-terminal GFP tag did not gene were subsequently detected in several females with perturb the three-dimensional structure or the localization of CHILD syndrome (congenital hemidysplasia with ichthyosi- the NSDHL protein. Further, the presence of the C-terminal

form nevus and limb defects, MIM 308050) (12–14), a rare myc tag did not prevent proper ER membrane targeting (data Downloaded from https://academic.oup.com/hmg/article/12/22/2981/606598 by guest on 27 September 2021 X-linked, male-lethal malformation syndrome characterized by not shown). The localization was distinct from that of the unilateral ichthyosiform skin lesions and limb reduction defects peroxisomal integral membrane protein PMP70 (20) (data not (reviewed in 9). shown). Most of the experiments described below were The function of the enzyme as a sterol dehydrogenase involved performed using the GFP-Nsdhl construct, which produced in the removal of C-4 methyl groups from the cholesterol the best signals for immunofluorescence. precursor lanosterol was initially suggested by the accumulation In most cells, NSDHL also accumulated on the surface of of 4-methyl and 4,4-dimethyl sterol intermediates in tissue spherical cytoplasmic structures that were often clustered samples and cultured skin fibroblasts from Bpa/Str mutant together. To determine whether the NSDHL-positive structures females and affected male embryos (7) (G. Herman and were LDs, cells were stained with Oil Red O, which stains their R. Kelley, unpublished data). Further support for this role of the neutral lipid core. As shown in Figure 2, the green fluorescence enzyme is our recent demonstration that the murine NSDHL from GFP-NSDHL coated the red core of the LDs stained with protein can rescue the lethality of erg26 deficient S. cerevisiae Oil Red O, demonstrating the presence of the protein on their that lack the yeast ortholog (11). Finally, human X-linked surface. In most successfully transfected cells, all Oil Red dominant chondrodysplasia punctata (CDPX2; MIM 302960) O-positive LDs were also positive for NSDHL. and the X-linked, male-lethal tattered (Td) mouse, whose phenotypes are very similar to that of Bpa mice, result from 8 7 Subcellular localization of selected mutant mutations in the human and murine D –D sterol isomerase NSDHL proteins , respectively, encoding the protein for the next step in the cholesterol biosynthetic pathway (15,16). Recently, we have described two ENU-induced alleles, Bpa7H Based on homology with yeast (17,18), it is predicted that and Bpa8H, that result from missense mutations in the NSDHL would be part of a multisubunit ER membrane conserved amino acids V53D and A94T, respectively, yet complex that also contains proteins for a sterol C-4 methyl demonstrate the more severe ‘bare patches’ phenotype in oxidase, a 3-keto sterol reductase, and possibly one or more surviving females (11). Two previous missense mutations, regulatory scaffold proteins. We describe here the subcellu- P98L and V109M, as well as a 3 bp in-frame insertion of a lar localization of the NSDHL protein, confirming its ER tyrosine residue within the predicted membrane spanning localization. Surprisingly, the protein also appears to accu- domain, had all been associated with the milder ‘striated’ mulate on the surface of lipid droplets (LDs) that function as phenotype (7). Thus, we were interested to examine the in vivo intracellular storage compartments for neutral lipids and choles- effects of these new mutations, including possible perturbations terol esters. We further demonstrate that association of NSDHL in proper protein folding. We compared the subcellular with ER membranes and LDs requires prior trafficking through localization of mutant proteins for the Bpa7H and Bpa8H alleles the Golgi, suggesting an additional possible mechanism for with those of wild-type NSDHL and the Bpa1H nonsense allele. the regulation of cellular cholesterol content. Each mutation was generated from the wild type GFP-Nsdhl expression construct by site-directed mutagenesis. The protein representing the Bpa1H allele was found throughout the cyto- plasm and to some extent in the nucleus (Fig. 1G–I), similar to RESULTS the GFP vector itself (Fig. 1A–C). It did not co-localize with the ER, as evidenced by a lack of substantial overlap with the Subcellular localization of the NSDHL protein DsRed2-ER marker. There was also some accumulation in unidentified cytoplasmic inclusions that did not stain with Oil The predicted amino acid sequence of murine NSDHL (7) Red O (not shown) and, hence, were not LDs. The Bpa7H includes 362 amino acids and contains an N-terminal NADH (V53D) protein demonstrated a diffuse cytoplasmic and nuclear 1H cofactor binding site and a central Tyr–X3–Lys motif that is subcellular localization very similar to that of the Bpa protein present at the active site of all known 3b-hydroxysteroid (not shown). The protein produced from the Bpa8H allele dehydrogenases (3b-HSDs) (19). The SOSUI system, hydro- (A94T) correctly localized to the ER (Fig. 1J–L) and also phobic cluster analysis (HCA), and DNASTAR Protean 4.05 accumulated on the surface of LDs (Fig. 2G–I), similar to the each predict that the enzyme contains a single 22-amino acid wild type NSDHL protein. hydrophobic transmembrane domain between residues 285 and To assess whether the mutant Bpa1H and Bpa7H proteins were 307 organized in an a-helical conformation. degraded and the resulting immunofluorescence was derived Human Molecular Genetics, 2003, Vol. 12, No. 22 2983 Downloaded from https://academic.oup.com/hmg/article/12/22/2981/606598 by guest on 27 September 2021

Figure 1. Subcellular localization of murine wild type and mutant GFP-NSDHL fusion proteins. GFP–Nsdhl fusion constructs were transiently transfected into COS7 cells together with an ER-localization marker, DsRed2-ER. Cells were visualized by direct fluorescence using confocal microscopy. GFP fluorescence results in green pixels (A, D, G and J), DsRed2 produces red pixels (B, E, H and K) while co-localization of GFP and DsRed2 results in yellow pixels (C, F, I and L). (A, B, C) Empty GFP vector; (D, E, F) wild-type GFP-Nsdhl; (G, H, I) fusion construct containing the Bpa1H nonsense mutation; (J, K, L) fusion construct contain- ing the Bpa8H missense mutation. Fusion proteins containing the wild type and Bpa8H mutation correctly localize to the ER as evidenced by the overlap of the two fluorescent channels. Protein containing the Bpa1H mutation presents a pattern of predominantly nuclear staining, similar to that observed with GFP vector alone.

solely from GFP and whether any of the mutant proteins were membrane fraction and were detected by both anti-GFP membrane-bound, we performed western blotting on mem- antibodies and polyclonal antibodies directed against a brane and soluble cytosolic fractions isolated from cells C-terminal epitope of the murine NSDHL protein (Fig. 3 top transiently transfected with wild-type and mutant GFP-Nsdhl and middle). The truncated GFP-NSDHL protein correspond- constructs (Fig. 3). As expected, both the wild-type and Bpa8H ing to the Bpa1H allele, with a predicted size of 39 kDa, GFP-NSDHL fusion proteins were found exclusively in the produced an appropriately sized protein that was found 2984 Human Molecular Genetics, 2003, Vol. 12, No. 22 Downloaded from https://academic.oup.com/hmg/article/12/22/2981/606598 by guest on 27 September 2021

Figure 2. Localization of GFP–NSDHL in LDs and requirement of C-terminal RKDK sequence for ER and LD targeting. The transfected constructs were the GFP vector (A–C), wild-type GFP-Nsdhl (D–F, J), GFP–Nsdhl containing the Bpa8H mutation (G–I), or GFP–Nsdhl DRKDK containing a stop codon at amino acid 359 (K, L). NSDHL protein localized to the ER is not well seen in (A–I) since images were optimized for visualization of LD targeted protein. Cells in (J–L) were counterstained with DAPI to visualize the cell nuclei.

primarily in the cytosolic fraction, consistent with the lack of a The C-terminal amino acids RKDK transmembrane domain (Fig. 3 bottom). This fusion was act as an ‘ER retrieval’ signal detected solely by the anti-GFP antibody since the anti-NSDHL C-terminal epitope is not present in the Bpa1H mutant protein. The C-terminal amino acid sequence ‘KKXX’ has been The GFP-NSDHL protein encoded by the Bpa7H allele was demonstrated to act as a consensus signal sequence for not detected with either antibody, suggesting that this retrograde transport of ER membrane proteins from the Golgi fusion protein is not stable, at least under the conditions by the COPI complex (21–23). Murine NSDHL contains a studied here. modified version of this sequence, RKDK. To test the Human Molecular Genetics, 2003, Vol. 12, No. 22 2985

Figure 4. Subcellular fractionation and western blotting of NSDHL protein Figure 3. Subcellular fractionation and western blotting of COS7 cells trans- lacking the putative C-terminal ER targeting signal. COS7 cells were trans- Downloaded from https://academic.oup.com/hmg/article/12/22/2981/606598 by guest on 27 September 2021 fected with GFP–Nsdhl fusion protein constructs. Cells were harvested for sub- fected with plasmids encoding either wild-type GFP–NSDHL or GFP– cellular fractionation 48 h post-transfection. Equal amounts (50 mg) of NSDHL–DRKDK. After 48 h, protein extracts were subjected to subcellular membrane and cytosolic protein fractions were subjected to 12% SDS– fractionation, and 25 mg of membrane or cytosolic proteins analyzed as in PAGE, electroblotted and immunoprobed with polyclonal anti-NSDHL anti- Figure 3 with an anti-GFP antibody (top) or anti-NSDHL antibody (bottom). body (top), or polyclonal anti-GFP antibody (middle and bottom). Molecular NSDHL–DRKDK is not detected by the anti-NSDHL antibody due to loss of size markers (Kaleidoscope pre-stained markers, BioRad) were used to estimate part of the C-terminal epitope. molecular weights. The GFP protein is predicted to have a molecular weight of 28 kDa, while the N-terminal 102 amino acids of NSDHL have a predicted molecular weight of 11 kDa. The full length, unmodified NSDHL protein has a predicted molecular weight of 41 kDa. M, membrane fraction, C, cytosolic fraction. Adequate separation of soluble and membrane fractions as well as Surface probability plots predict the N-terminal major portion equal protein loading was assessed by stripping and reprobing membranes with of NSDHL flanking the hydrophobic transmembrane domain to an anti-calnexin antibody (data not shown), which detects an abundant 90 kDa be localized on the cytoplasmic side of the ER membrane, while ER-resident transmembrane protein (25). predictions for the smaller C-terminal region were uncertain (data not shown). To determine whether NSDHL may be recruited to LDs as they bud from the ER, we determined hypothesis that the C-terminal RKDK motif is responsible for whether the C-terminus of the protein is exposed to the the ER targeting of the wild-type NSDHL protein, we cytoplasm or ER lumen by exploiting the differential sensitivity performed site-directed mutagenesis and converted the first of the plasma and ER membranes to permeabilization with non- amino acid in this motif to a stop codon (creating the mutant ionic detergents. As shown in Figure 6, when HeLa cells are protein designated GFP–NSDHL–DRKDK). The GFP– permeabilized with saponin, the ER membrane protein calnexin NSDHL–DRKDK protein remained membrane-bound (Fig. 4); with a C-terminal epitope exposed to the cytoplasm is readily however it was localized exclusively in cytoplasmic perinuclear detected (25), while a C-terminal epitope of protein disulfide structures consistent with the Golgi apparatus (Fig. 2K and L). isomerase (PDI), which resides in the ER lumen, is only exposed These results are consistent with the function of the C-terminal upon harsher treatment with Triton X-100. The carboxy- RKDK as a COPI-complex mediated ER retrieval signal. terminus of the NSDHL protein detected with an antibody to There was also no accumulation of the GFP–NSDHL– either a C-terminal myc tag or the C-terminal amino acids of the DRKDK protein on the surface of cytoplasmic LDs. LDs are murine protein is readily accessible in saponin-treated cells, known to derive from the ER, where triacylglycerides and strongly favoring a topology of the protein in the ER membrane cholesteryl esters are synthesized, and recent data are consistent as shown in Figure 5 in which both the N- and C-termini are with the hypothesis that the lipid droplet surface is derived exposed to the cytoplasm. from the cytoplasmic leaflet of the ER membrane (24). As discussed below, some proteins found on LDs are believed to bud from the ER membrane with the droplet as it forms. Thus, we believe that the lack of accumulation of GFP–NSDHL– DISCUSSION DRKDK on LDs strongly suggests that a prior ER membrane The conversion of lanosterol to cholesterol involves the localization is required for this association. However, we removal of three methyl groups, one at the C-14 position and cannot completely exclude the possibility that NSDHL two at the C-4 position; the reduction of double bonds at the accumulates on LDs without prior retrieval to the ER, that C-24 and C-14 positions; and the ‘migration’ of the C-8(9) LD localization precedes incorporation into the ER membrane, double bond to the C-5 position (reviewed in 5) (Fig. 7). Nine and/or that the C-terminal RKDK sequence is specifically enzymes complete this series of reactions, including three required for LD association of NSDHL. enzymes involved in the sequential removal of the two C-4 Some of the known proteins found on the surface of methyl groups—a sterol-C4-methyl-oxidase (encoded by the LDs demonstrate an ER membrane topology in which both SC4MOL gene in human) (26), the NSDHL 3b-hydroxysterol the N- and C-termini are exposed to the cytoplasmic side of the dehydrogenase and a 3b-ketosteroid reductase (encoded by the membrane. Movement of such proteins onto the surface of LDs HSD17B7 gene) (27). It has been suggested (4,5) that all of occurs as the droplet buds from the ER membrane (see Fig. 5). these enzymatic reactions occur on the ER membrane, due in In this model, ER lumenal proteins and those spanning the ER part to the poor solubility of their sterol substrates. An ER membrane would be excluded from entry into the LD, although membrane localization has been suggested based on partial exceptions do occur (22). enzyme purification for some of the enzymes (summarized in 5); 2986 Human Molecular Genetics, 2003, Vol. 12, No. 22

Figure 5. Model of trafficking of selected ER membrane proteins to the surface of LDs (modified from 32). Neutral lipids accumulate in the interior of the ER Downloaded from https://academic.oup.com/hmg/article/12/22/2981/606598 by guest on 27 September 2021 bilayer, making a bulge that eventually buds into the cytoplasm to form a droplet surrounded by an ER-derived phospholipid monolayer. This process would drive opposite ER membrane monolayers apart, thickening the hydrophobic portion of the membrane. ER membrane proteins, such as NSDHL (blue), with N- and C-terminal surfaces exposed to the cytoplasm can easily diffuse laterally between the ER membrane proper and the monolayer surrounding the nascent droplets. Other proteins on the surface of the LD are shown in green and orange.

however, direct demonstration of a subcellular localization associate with LDs, including S3-12 (36) and TIP47 (34), using tagged constructs from cloned genes has been reported initially identified as a protein involved in trafficking of the for only four of the mammalian enzymes: 3b-hydroxysterol- mannose-6-phosphate receptor. In addition, caveolins that are D14-reductase activity has recently been demonstrated in the overexpressed or modified by truncations or the addition of ER Lamin B receptor (28,29), which is localized on the inner retrieval sequences have been found on the surface of LDs, nuclear membrane (29). GFP-tagged 3b-ketosterol reductase although the physiologic significance of these findings is not protein has been localized to the ER membrane (27). The known (32,37). Caveolins are found primarily in invaginations D8–D7–sterol isomerase has been localized to ER membranes of the plasma membrane called caveolae that are enriched in as well as the nuclear membrane (30). Finally, human 3b- cholesterol and glycosphingolipids and represent a subset of hydroxysteroid-D7-reductase (DHCR7) has been localized to lipid rafts (38). They are involved in a variety of cellular the ER membrane by immunofluorescence of tagged protein in processes including endocytosis, cell signaling and intracellular COS7 cells (29). Subcellular localizations for the remainder of trafficking, suggesting that their presence on LDs could have the mammalian proteins have not been directly determined, physiologic importance. LDs in some plant seeds contain a although analysis of predicted protein sequences suggests an continuous coating of proteins called oleosins (39). integral membrane location for all of them. As shown in Figure 5, LDs are believed to bud from the ER We have demonstrated here that exogenously expressed membrane and contain a single phospholipid bilayer derived murine NSDHL is localized on ER membranes and also from the outer (cytoplasmic) ER membrane surrounding a accumulates on the surface of LDs. This is the first known central lipid core. Proteins are excluded from the core but can association of a cholesterol biosynthetic enzyme with this coat the surface of the droplet. The mechanism of association ubiquitous cytoplasmic organelle. LDs are involved in the of the known lipid droplet proteins with the organelle is not storage of neutral lipids, specifically, triacylglycerols and completely understood. Perilipins and ADRP appear to be cholesterol esters. They are found in selected prokaryotes and translated on free ribosomes and associate with preformed in a wide variety of cell types of virtually all eukaryotes, LDs (35). Oleosins and caveolins have N- and C-terminal including fungi and plants (reviewed in 31–33). However, we hydrophilic domains that face the cytoplasm flanking a central are only beginning to understand their biogenesis, as well as hydrophobic one and could bud from the ER with the newly their possible functions in intracellular cholesterol transport formed lipid droplet (32,35,40). and signaling, in addition to lipid storage. We have demonstrated that ER retrieval and trafficking is likely In animal cells, LDs are most abundant in adipocytes, where necessary for the accumulation of NSDHL protein on LDs. they range from 2 to 100 mm in diameter, and in steroidogenic Protein lacking the ER retrieval signal that is sequestered in the tissues such as the adrenal cortex and liver. The major proteins Golgi complex does not coat LDs, despite the presence of the coating the surface of LDs in animal cells demonstrate regions transmembrane domain and all but the last four amino acids of of homology in their amino acid sequences and have been the protein (Fig. 2J–L). Further, like oleosins and caveolins, called PAT proteins [for perilipins, adipose-differentiation- NSDHL demonstrates a topology in which both the N- and related-protein (ADRP), and TIP47 (34)]. The proteins are C-terminal domains extrude from the ER membrane into the tissue-specific with ADRP found in LDs in all tissues as well as cytoplasm and flank a single transmembrane region. We hypo- in pre-adipocytes, while perilipins are found only in mature thesize that NSDHL in the ER membrane can diffuse laterally to adipocytes and in steroidogenic cells (35). In addition to their nascent LDs. Despite having no transmembrane domain, it is structural role, altered gene expression during differentiation or known that the yeast ortholog of NSDHL (ERG26) is intimately after hormone-stimulated lipolysis suggests that the association associated with ER membranes (17,41,42), suggesting that other of these proteins with LDs may have regulatory roles in cellular regions of the protein may facilitate lipid binding. lipid metabolism. Recently, additional proteins that demon- We have also examined the subcellular localization of strate homology to perilipins and ADRP have been noted to selected mutant NSDHL proteins. NSDHL protein from the Human Molecular Genetics, 2003, Vol. 12, No. 22 2987 Downloaded from https://academic.oup.com/hmg/article/12/22/2981/606598 by guest on 27 September 2021

Figure 6. Analysis of the topology of NSDHL in the ER membrane by differential permeabilization of transiently transfected HeLa cells. Transfected constructs are shown to the left and the non-ionic detergent used for permeabilization is shown at the top. NSDHL protein was detected with antibody to the C-terminal myc tag (myc) or antibody to a C-terminal NSDHL epitope (NSDHL). The blue signal represents DAPI staining and nuclear fluorescence. PDI ¼ protein disulfide isomerase, a protein found in the ER lumen.

Bpa1H allele that is missing the C-terminal 260 amino acids, for LD association. Protein from the Bpa7H allele, which contains including the 22 amino acid hydrophobic domain, is not the missense mutation V53D, retains both the central hydro- membrane associated and does not associate with LDs phobic domain and the ER retrieval signal, yet had a diffuse (Figs 1G–I and 3). These data suggest that this domain is nuclear and cytoplasmic localization similar to Bpa1H and necessary for ER membrane association and, hence, indirectly the empty GFP vector (not shown). The immunofluorescence 2988 Human Molecular Genetics, 2003, Vol. 12, No. 22 Downloaded from https://academic.oup.com/hmg/article/12/22/2981/606598 by guest on 27 September 2021

Figure 7. Schematic representation of ‘post-squalene’ cholesterol biosynthesis. Individual enzymes, denoted by numbers in parentheses, are: (1) 3b-hydroxysteroid- D24-reductase, (2) lanosterol 14a-demethylase, (3) 3b-hydroxysteroid-D14-reductase, (4) 3b-hydroxysteroid C-4 methyl oxidase, (5) NSDHL, 3b-hydroxysteroid dehydrogenase, (6) 3b-keto-steroid reductase, (7) 3b-hydroxysteroid–D8–D7–sterol isomerase, (8) 3b-hydroxysteroid-D5-desaturase (lathosterol dehydrogenase), and (9) 3b-hydroxysteroid-D7-reductase (DHCR7). Reduction of the C-24 double bond can occur at several points along the pathway, but is shown only as the first step for simplification. likely derives from GFP plus variable portions of an unstable the mutation results in the substitution of a charged aspartate mutant protein since we could not detect it on western blots residue within a small cluster of hydrophobic residues, with antibodies to either GFP or NSDHL (Fig. 3). This breaking the predicted secondary conformation of a b-strand. instability could be related to improper protein folding since In addition, the preceding amino acid, T52, is predicted to be Human Molecular Genetics, 2003, Vol. 12, No. 22 2989 phosphorylated; after introduction of the mutation, this MATERIALS AND METHODS phosphorylation site is lost. In contrast, protein for Bpa8H allele demonstrates proper targeting to the ER as well as to LDs Materials and plasmid constructs (Figs 1J–L and 2G–I). The mechanism by which the A94T mutation produces a non-functional protein remains unknown The pDsRed2-ER subcellular localization vector was obtained and may require determination of the three-dimensional from Clontech. Oil Red O and saponin were purchased from structure of the protein. Sigma Chemical Co. Custom synthetic oligonucleotides were Finally, we have shown that the RKDK sequence found at the purchased from Sigma-Genosys. GFP-tagged NSDHL was C-terminus of the NSDHL protein acts as an ER retrieval signal as generated by subcloning the murine Nsdhl cDNA (7) into the evidenced by the retention in the Golgi complex of protein lacking BamHI site of pEGFP-C1 (Clontech) yielding an in-frame these four amino acids (Fig. 3J–L). The traditional consensus fusion 30 to GFP. Mutations to generate the Bpa1H, Bpa7H and 8H signal sequence for retrograde transport of ER membrane Bpa alleles were introduced into the wild-type fusion Downloaded from https://academic.oup.com/hmg/article/12/22/2981/606598 by guest on 27 September 2021 1 proteins from the Golgi by the COPI complex is the C-terminal construct using the Quikchange XL site-directed mutagenesis amino acids KKXX (21–23). Our data suggest that, despite kit (Stratagene) and specific complementary oligonucleotides divergence in sequence, the variant RKDK motif is still sufficient designed to contain the desired mutations. NSDHL–DRKDK, to promote retrieval of NSDHL to the ER via the retrograde lacking the C-terminal residues 359–RKDK–362, was similarly pathway. generated by introducing a stop codon at the position of R359. The functional significance of the association of NSDHL All mutations were confirmed by sequencing. with LDs is not known. Unesterified cholesterol has been reported on the surface of LDs, but not within its core (43,44), Antibodies raising the possibility that one or more steps of cholesterol biosynthesis could occur on the LD surface. In S. cerevisiae, Polyclonal rabbit antisera were raised against a C-terminal squalene epoxidase encoded by the erg1 gene, lanosterol peptide VERTVQSFHHLRKDK corresponding to amino acids synthase or oxidosqualene cyclase encoded by the erg7, and 348–362 of the mouse NSDHL protein (Cocalico Biologicals) erg6 encoded sterol-D24-methyltransferase are located primarily and affinity purified using a Sulfolink kit (Pierce). Rabbit in LDs (33,45,46) suggesting a possible evolutionarily anti-GFP(FL) (Santa Cruz Biotechnology Inc.), rabbit anti- conserved role for the LD localization of enzymes involved calnexin (Stressgen), mouse anti-myc (Invitrogen), mouse anti- in sterol biosynthesis. The yeast NSDHL ortholog (ERG26 PDI (Affinity Bioreagents), goat anti-mouse FITC and goat protein) has not been found on lipid droplets; however, the anti-rabbit FITC (Sigma), goat anti-rabbit Texas red (Jackson erg27 encoded 3-ketosterol reductase has recently been Laboratories), and goat anti-rabbit IgG horseradish peroxidase demonstrated to interact with and facilitate association of the conjugate (Cell Signalling) were used at the dilutions described yeast oxidosqualene cyclase with LDs and is necessary for below. The antibody to the peroxisomal membrane protein enzymatic activity of the cyclase in this organelle (46). PMP70 was the generous gift of D. Valle and G. Jimenez- Recent demonstrations of distinct subcellular locations for Sanchez, The Johns Hopkins University. Vectashield and several of the mammalian enzymes of post-squalene choles- Vectashield-DAPI mounting media were obtained from terol biosynthesis—the ER, the nuclear membrane and LDs— Vector Labs. could provide another level of complexity in regulating the amount of cholesterol synthesized in the cell. Alternatively, Cell culture and immunofluorescence since the pattern of staining of NSDHL on LDs suggests that the protein coats most, if not all, of the surface of droplets, COS7 and HeLa cells (American Type Culture Collection) similarly to PAT proteins, it could have a function unrelated to were maintained in Dulbecco’s modified Eagle’s media its role as a cholesterol biosynthetic enzyme. To help differ- (Mediatech) supplemented with 10% BCS (bovine calf serum, entiate between these two possibilities, it will be important to defined/supplemented; HyClone) containing 2 mML-glutamine determine whether the other proteins of the C-4 demethylase (Invitrogen) at 37 C and 5% CO2. Transient transfections with complex co-localize with NSDHL on LDs. Effectene transfection reagent were performed as recom- Finally, just as we completed these studies, Ohashi et al. (47) mended by the supplier (Qiagen), and cells were studied 48 h reported a similar localization of endogenous and C-terminal after transfection. VSV-G-tagged human NSDHL protein on the surface of LDs Transiently transfected COS7 cells grown on 13 mm glass of CHO cells. The protein demonstrated a co-localization coverslips were fixed in a solution of 2% sucrose and 1.8% with ADRP and TIP47. They found significant localization in formaldehyde in PBS for 10 min at room temperature and the ER only in cells expressing high levels of the protein. mounted with Vectashield. For staining of LDs, fixed cells were Treatment with sterol-depleted medium reduced the develop- briefly rinsed in 60% propan-2-ol followed by incubation ment of LDs in cells, resulting in re-distribution of NSDHL to with 0.6% Oil Red O in 60% propan-2-ol for 2 min at the ER. Their studies complement those reported here that room temperature, rinsed in 60% propan-2-ol, PBS and water, demonstrate the likely mechanism of LD targeting of NSDHL. and mounted as above. For selective permeabilization of Together, these results stress the importance of and need the plasma membrane, HeLa cells grown on 13 mm glass for further studies to understand the biological processes coverslips were fixed in 2% paraformaldehyde for 30 min, in which LDs participate and to examine whether other permeabilized with either 20 mg/ml saponin (selectively cholesterol biosynthetic or regulatory proteins are found on permeabilized cells) or 0.2% Triton X-100 (permeabilized these organelles. cells). The cells were washed with PBS, blocked with PBS/5% 2990 Human Molecular Genetics, 2003, Vol. 12, No. 22

FBS and incubated with primary antibodies (anti-myc, 1 : 500; time (1 s to 1 min 30 s). Prior to reprobing the membranes, they anti-NSDHL, 1 : 200; anti-PDI, 1 : 100; anti-calnexin, 1:200). were stripped by treatment with 100 mM 2-mercaptoethanol, 2% The cells were washed and incubated with a secondary SDS, 62.5 mM Tris pH 6.7 for 30 min at 55C with agitation. antibody (anti-mouse FITC, 1 : 32; anti-rabbit Texas Red, The membranes were washed with large volumes of TBST 1 : 200; anti-rabbit FITC, 1 : 80), washed and mounted as 2 10 min. Blocking and immunodetection of stripped mem- above. For selectively permeabilized cells, all incubations and branes was performed as above. washes were performed in the presence of 20 mg/ml saponin. Confocal microscopy was performed on a Zeiss LSM510 Laser Scanning Microscope. Image analysis was performed using the Protein motif and structure prediction standard system operating software provided with the micro- DNASTAR family of programs, Hydrophobic Cluster Analysis scope. GFP was detectable in the FITC channel (green), DsRed2 (HCA) (49), and SOSUI system (50) were used for secondary and Oil Red O in the Texas Red channel (red), and DAPI structure prediction, hydrophobicity and surface probability Downloaded from https://academic.oup.com/hmg/article/12/22/2981/606598 by guest on 27 September 2021 in the near infrared channel (blue). Co-localization of green plots. NetPhos 2.0 was used for the prediction of phosphoryla- (GFP/FITC) and red (DsRed2/Texas Red) signals in a single tion sites (51). pixel produced yellow, while separated signals remained green or red. ACKNOWLEDGEMENTS Membrane and cytosol separation The authors thank R. Burry, The Ohio State University for helpful discussions concerning cell permeabilization. This Membrane and cytosolic protein separation was performed work was supported by NIH R01 HD38572 to G.E.H. DNA with modifications of a procedure described by Martinez-Botas sequences were determined with the help of the Sequencing et al. (48). Briefly, transiently transfected 100 mm dishes of Core Facility at Columbus Children’s Research Institute and cells from different experimental conditions were rinsed with confocal microscopy was performed in a Core Facility of the cold PBS and collected in sonication buffer [2 mM EDTA, Comprehensive Cancer Center of The Ohio State University. 2mM PMSF, and complete protease inhibitor cocktail (Roche) in PBS]. Cell suspensions were subjected to three freeze– thaw cycles, followed by a 5 min sonication in an ice-bath (2 s REFERENCES pulse, 50% amplitude) using a 100 W high intensity ultrasonic processor (Sonics and Materials Inc.). Cell membranes were 1. Fitzgerald, M.L., Moore, K.J. and Freeman, M.W. (2002) Nuclear hormone receptors and cholesterol trafficking: the orphans find a new home. J. Mol. pelleted by centrifugation at 50 000g for 30 min at 4 Cina Med., 80, 271–281. Beckman Optima TLX ultracentrifuge using a TLA100.2 rotor. 2. Edwards, P.A., Kennedy, M.A. and Mak, P.A. (2002) LXRs; The supernatant containing the cytosol fraction was removed oxysterol-activated nuclear receptors that regulate genes controlling and the membrane pellet dissolved in resuspension buffer lipid homeostasis. Vascul. Pharmac., 38, 249–256. (100 mM Tris, pH 7.5, 300 mM NaCl, 1% Triton X-100, 0.1% 3. Jeong, J. and McMahon, A.P. (2002) Cholesterol modification of Hedgehog family proteins. J. Clin. Invest., 110, 591–596. SDS, 2 mM EDTA, 2 mM PMSF, complete protease inhibitor 4. Goldstein, J.L. and Brown, M.S. (1990) Regulation of the mevalonate cocktail). The amount of protein in the membrane and cytosol pathway. Nature, 343, 425–430. extracts was quantified using the BCA method (Pierce). Both 5. Gaylor, J.L. (2002) Membrane-bound enzymes of cholesterol synthesis fractions were boiled in Laemmli sample buffer [100 mM Tris– from lanosterol. Biochem. Biophys. Res. Commun., 292, 1139–1146. 6. Kovacs, W.J., Olivier, L.M. and Krisans, S.K. (2002) Central role of M HCl pH 6.8, 4% (w/v) SDS, 200 m b-mercaptoethanol, peroxisomes in isoprenoid biosynthesis. Prog. Lipid. Res., 41, 369–391. 20% (v/v) glycerol, 0.2% (w/v) bromophenol blue] and stored 7. Liu, X.Y., Dangel, A.W., Kelley, R.I., Zhao, W., Denny, P., Botcherby, M., at 20C. Cattanach, B., Peters, J., Hunsicker, P.R., Mallon, A.M. et al. (1999) The gene mutated in bare patches and striated mice encodes a novel 3beta- hydroxysteroid dehydrogenase. Nat. Genet., 22, 182–187. 8. Happle, R., Phillips, R.J., Roessner, A. and Junemann, G. (1983) Immunoblot analysis Homologous genes for X-linked chondrodysplasia punctata in man and mouse. Hum. Genet., 63, 24–27. Equal amounts of protein were electrophoretically resolved on 9. Herman, G.E. (2000) X-Linked dominant disorders of cholesterol 12% SDS-polyacrylamide gels at 150 V for 1 h 30 min using biosynthesis in man and mouse. Biochim. Biophys. Acta, 1529, 357–373. 10. Phillips, R.J.S. (1963) Striated, a new sex-linked gene in the house mouse. vertical slab units, and transferred onto Hybond-ECL mem- Genet. Res., 4, 151–153. branes (Amersham) for 1 h at 180 mA. Non-specific sites on 11. Lucas, M.L., Ma, Q., Cunningham, D., Peters, J., Cattanach, B., Bard, M., the membrane were blocked by incubation for 1 h at room Elmore, B.K. and Herman, G.E. (2003) Identification of two novel temperature in 10% non-fat dry milk powder (Carnation) in TBS mutations in the murine Nsdhl sterol dehydrogenase gene and development (20 mM Tris–HCl, 137 mM NaCl, pH 7.6) containing 0.1% of a functional complementation assay in yeast. Mol. Genet. Metab. (in press). Tween-20 (TBST). Membranes were probed with NSDHL 12. Konig, A., Happle, R., Bornholdt, D., Engel, H. and Grzeschik, K.H. antibody (1 : 500), GFP antibody (1 : 200) or calnexin antibody (2000) Mutations in the NSDHL gene, encoding a 3beta-hydroxysteroid (1 : 5000) in TBST containing 5% non-fat dry milk. Detection dehydrogenase, cause CHILD syndrome. Am. J. Med. Genet., of the antigen–antibody complexes was performed with horse- 90, 339–346. 13. Konig, A., Happle, R., Fink-Puches, R., Soyer, H.P., Bornholdt, D., radish peroxidase conjugated secondary antibody (1 : 1000) and Engel, H. and Grzeschik, K.H. (2002) A novel missense mutation of the enhanced chemiluminescence kit (Amersham). Membranes NSDHL in an unusual case of CHILD syndrome showing bilateral, almost were exposed to X-Omat film S (Kodak) for variable lengths of symmetric involvement. J. Am. Acad. Dermatol., 46, 594–596. Human Molecular Genetics, 2003, Vol. 12, No. 22 2991

14. Hummel, M., Cunningham, D., Mullett, C.J., Kelley, R.I. and Herman, G.E. 34. Miura, S., Gan, J.W., Brzostowski, J., Parisi, M.J., Schultz, C.J., Londos, C., (2003) Left-sided CHILD syndrome caused by a nonsense mutation in the Oliver, B. and Kimmel, A.R. (2002) Functional conservation for lipid NSDHL gene. Am. J. Med. Genet., 122A, 246–251. storage droplet association among Perilipin, ADRP, and TIP47 15. Braverman, N., Lin, P., Moebius, F.F., Obie, C., Moser, A., Glossmann, H., (PAT)-related proteins in mammals, Drosophila, and Dictyostelium. Wilcox, W.R., Rimoin, D.L., Smith, M., Kratz, L. et al. (1999) Mutations J. Biol. Chem., 277, 32253–32257. in the gene encoding 3 beta-hydroxysteroid-delta 8, delta 7-isomerase 35. Londos, C., Brasaemle, D.L., Schultz, C.J., Segrest, J.P. and Kimmel, A.R. cause X-linked dominant Conradi–Hunermann syndrome. Nat. Genet., (1999) Perilipins, ADRP, and other proteins that associate with intracellular 22, 291–294. neutral lipid droplets in animal cells. Semin. Cell. Dev. Biol., 10, 51–58. 16. Derry, J.M., Gormally, E., Means, G.D., Zhao, W., Meindl, A., Kelley, R.I., 36. Wolins, N.E., Skinner, J.R., Schoenfish, M.J., Tzekov, A., Bensch, K.G. and Boyd, Y. and Herman, G.E. (1999) Mutations in a delta 8-delta 7 sterol Bickel, P.E. (2003) Adipocyte protein S3-12 coats nascent lipid droplets. isomerase in the tattered mouse and X-linked dominant chondrodysplasia J. Biol. Chem., 278, 37713–37721. punctata. Nat. Genet., 22, 286–290. 37. van Meer, G. (2001) Caveolin, cholesterol, and lipid droplets? J. Cell Biol., 17. Mo, C., Valachovic, M., Randall, S.K., Nickels, J.T. and Bard, M. 152, F29–34. (2002) Protein–protein interactions among C-4 demethylation enzymes 38. Zajchowski, L.D. and Robbins, S.M. (2002) Lipid rafts and little caves. involved in yeast sterol biosynthesis. Proc. Natl Acad. Sci. USA, 99, Compartmentalized signalling in membrane microdomains. Eur. J. Downloaded from https://academic.oup.com/hmg/article/12/22/2981/606598 by guest on 27 September 2021 9739–9744. Biochem., 269, 737–752. 18. Baudry, K., Swain, E., Rahier, A., Germann, M., Batta, A., Rondet, S., 39. Abell, B.M., Holbrook, L.A., Abenes, M., Murphy, D.J., Hills, M.J. and Mandala, S., Henry, K., Tint, G.S., Edlind, T. et al. (2001) The effect of the Moloney, M.M. (1997) Role of the proline knot motif in oleosin erg26-1 mutation on the regulation of lipid metabolism in Saccharomyces endoplasmic reticulum topology and oil body targeting. Plant Cell, 9, cerevisiae. J. Biol. Chem., 276, 12702–12711. 1481–1493. 19. Penning, T.M. (1997) Molecular endocrinology of hydroxysteroid 40. Ostermeyer, A.G., Paci, J.M., Zeng, Y., Lublin, D.M., Munro, S. and dehydrogenases. Endocr. Rev., 18, 281–305. Brown, D.A. (2001) Accumulation of caveolin in the endoplasmic 20. Kamijo, K., Taketani, S., Yokota, S., Osumi, T. and Hashimoto, T. (1990) reticulum redirects the protein to lipid storage droplets. J. Cell Biol., The 70-kDa peroxisomal membrane protein is a member of the Mdr 152, 1071–1078. (P-glycoprotein)-related ATP-binding protein superfamily. J. Biol. Chem., 41. Gachotte, D., Barbuch, R., Gaylor, J., Nickel, E. and Bard, M. (1998) 265, 4534–4540. Characterization of the Saccharomyces cerevisiae ERG26 gene encoding 21. Teasdale, R.D. and Jackson, M.R. (1996) Signal-mediated sorting of the C-3 sterol dehydrogenase (C-4 decarboxylase) involved in sterol membrane proteins between the endoplasmic reticulum and the golgi biosynthesis. Proc. Natl Acad. Sci. USA, 95, 13794–13799. apparatus. Annu. Rev. Cell. Dev. Biol., 12, 27–54. 42. Gachotte, D., Eckstein, J., Barbuch, R., Hughes, T., Roberts, C. and 22. Andersson, H., Kappeler, F. and Hauri, H.P. (1999) Protein targeting to Bard, M. (2001) A novel gene conserved from yeast to humans is endoplasmic reticulum by dilysine signals involves direct retention in involved in sterol biosynthesis. J. Lipid Res., 42, 150–154. addition to retrieval. J. Biol. Chem., 274, 15080–15084. 43. Kraemer, F.B., Laane, C., Park, B. and Sztalryd, C. (1994) Low-density 23. Letourneur, F., Gaynor, E.C., Hennecke, S., Demolliere, C., Duden, R., lipoprotein receptors in rat adipocytes: regulation with fasting. Am. J. Emr, S.D., Riezman, H. and Cosson, P. (1994) Coatomer is essential for Physiol., 266, E26–32. retrieval of dilysine-tagged proteins to the endoplasmic reticulum. Cell, 79, 44. Prattes, S., Horl, G., Hammer, A., Blaschitz, A., Graier, W., Sattler, W., 1199–1207. Zechner, R. and Steyrer, E. (2000) Intracellular distribution and 24. Tauchi-Sato, K., Ozeki, S., Houjou, T., Taguchi, R. and Fujimoto, T. (2002) mobilization of unesterified cholesterol in adipocytes: triglyceride The surface of lipid droplets is a phospholipid monolayer with a unique droplets are surrounded by cholesterol-rich ER-like surface layer fatty acid composition. J. Biol. Chem., 277, 44507–44512. structures. J. Cell Sci., 113, 2977–2989. 25. Elagoz, A., Callejo, M., Armstrong, J. and Rokeach, L. (1999) Although 45. Milla, P., Athenstaedt, K., Viola, F., Oliaro-Bosso, S., Kohlwein, S.D., calnexin is essential in S. pombe, its highly conserved central domain is Daum, G. and Balliano, G. (2002) Yeast oxidosqualene cyclase (Erg7p) is a dispensable for viability. J. Cell Sci., 112, 4449–4460. major component of lipid particles. J. Biol. Chem., 277, 2406–2412. 26. Li, L. and Kaplan, J. (1996) Characterization of yeast methyl sterol oxidase 46. Mo, C., Milla, P., Athenstaedt, K., Ott, R., Balliano, G., Daum, G. and (ERG25) and identification of a human homologue. J. Biol. Chem., 271, Bard, M. (2003) In yeast sterol biosynthesis the 3-keto reductase protein 16927–16933. (Erg27p) is required for oxidosqualene cyclase (Erg7p) activity. Biochim. 27. Marijanovic, Z., Laubner, D., Moller, G., Gege, C., Husen, B., Adamski, J. Biophys. Acta, 1633, 68–74. and Breitling, R. (2003) Closing the gap: Identification of human 47. Ohashi, M., Mizushima, N., Kabeya, Y. and Yoshimori, T. (2003) 3-ketosteroid reductase, the last unknown enzyme of mammalian Localization of mammalian NAD(P)H steroid dehydrogenase-like protein cholesterol biosynthesis. Mol. Endocrinol., 17, 1715–1725. on lipid droplets. J. Biol. Chem., 278, 36819–36829. 28. Silve, S., Dupuy, P.H., Ferrara, P. and Loison, G. (1998) Human lamin B 48. Martinez-Botas, J., Ferruelo, A.J., Suarez, Y., Fernandez, C., Gomez- receptor exhibits sterol C14-reductase activity in Saccharomyces cerevisiae. Coronado, D. and Lasuncion, M.A. (2001) Dose-dependent effects of Biochim. Biophys. Acta, 1392, 233–244. lovastatin on cell cycle progression. distinct requirement of cholesterol and 29. Holmer, L., Pezhman, A. and Worman, H.J. (1998) The human lamin B non-sterol mevalonate derivatives. Biochim. Biophys. Acta, 1532, 185–194. receptor/sterol reductase multigene family. Genomics, 54, 469–476. 49. Callebaut, I., Labesse, G., Durand, P., Poupon, A., Canard, L., Chomilier, J., 30. Dussossoy, D., Carayon, P., Belugou, S., Feraut, D., Bord, A., Goubet, C., Henrissat, B. and Mornon, J.P. (1997) Deciphering protein sequence Roque, C., Vidal, H., Combes, T., Loison, G. et al. (1999) Colocalization of information through hydrophobic cluster analysis (HCA): current status sterol isomerase and sigma(1) receptor at endoplasmic reticulum and and perspectives. Cell. Mol. Life Sci., 53, 621–645. nuclear envelope level. Eur. J. Biochem., 263, 377–386. 50. Mitaku, S., Hirokawa, T. and Tsuji, T. (2002) Amphiphilicity index of 31. Murphy, D.J. and Vance, J. (1999) Mechanisms of lipid-body formation. polar amino acids as an aid in the characterization of amino acid Trends Biochem. Sci., 24, 109–115. preference at membrane–water interfaces. Bioinformatics, 32. Brown, D.A. (2001) Lipid droplets: proteins floating on a pool of fat. 18, 608–616. Curr. Biol., 11, R446–449. 51. Blom, N., Gammeltoft, S. and Brunak, S. (1999) Sequence and 33. Zweytick, D., Athenstaedt, K. and Daum, G. (2000) Intracellular lipid structure-based prediction of eukaryotic protein phosphorylation sites. particles of eukaryotic cells. Biochim. Biophys. Acta, 1469, 101–120. J. Mol. Biol., 294, 1351–1362.