Defective lipid remodeling of GPI anchors in peroxisomal disorders, Zellweger syndrome, and rhizomelic chondrodysplasia punctata.

Many cell surface proteins in mammalian cells are anchored to the plasma membrane via glycosylphosphatidylinositol (GPI). The predominant form of mammalian GPI contains 1-alkyl-2-acyl phosphatidylinositol (PI), which is generated by lipid remodeling from diacyl PI. The conversion of diacyl PI to 1-alkyl-2-acyl PI occurs in the ER at the third intermediate in the GPI biosynthetic pathway. This lipid remodeling requires the alkyl-phospholipid biosynthetic pathway in peroxisome. Indeed, cells defective in dihydroxyacetone phosphate acyltransferase (DHAP-AT) or alkyl-DHAP synthase express only the diacyl form of GPI-anchored proteins. A defect in the alkyl-phospholipid biosynthetic pathway causes a peroxisomal disorder, rhizomelic chondrodysplasia punctata (RCDP), and defective biogenesis of peroxisomes causes Zellweger syndrome, both of which are lethal genetic diseases with multiple clinical phenotypes such as psychomotor defects, mental retardation, and skeletal abnormalities. Here, we report that GPI lipid remodeling is defective in cells from patients with Zellweger syndrome having mutations in the peroxisomal biogenesis factors PEX5, PEX16, and PEX19 and in cells from patients with RCDP types 1, 2, and 3 caused by mutations in PEX7, DHAP-AT, and alkyl-DHAP synthase, respectively. Absence of the 1-alkyl-2-acyl form of GPI-anchored proteins might account for some of the complex phenotypes of these two major peroxisomal disorders.

enzymes (such as alkaline phosphatase and 5 ′ -nucleotidase), and complement regulatory proteins (CD55 and CD59), playing roles in cell-cell and cell-environment interactions (13)(14)(15). GPI, consisting of phosphatidylinositol (PI), glucosamine (GlcN), three mannoses (Man), and three ethanolamine phosphates, is synthesized from PI in the ER, and preassembled GPI is transferred to the protein's carboxyl terminus (14)(15)(16). In mammalian cells, the majority of protein-linked GPI contains 1-alkyl-2-acyl PI, whereas most free PIs from which GPIs are generated are in the diacyl form. The fi rst and second intermediates of GPI precursors, N -acetyl glucosaminyl (GlcNAc)-PI and GlcN-PI, are mostly diacyl forms, whereas about 60% of the third intermediate, GlcN-(inositol-acylated)-PI (GlcN-acyl-PI), contain 1-alkyl-2-acyl PI, indicating that the 1-alkyl chain is generated after the third biosynthetic step in the ER ( 17 ). We reported that generation of the 1-alkyl-2-acyl form of GPI is dependent upon the peroxisomal alkyl-phospholipid biosynthetic pathway ( 18 ): Two different Chinese hamster ovary (CHO) mutant cell lines, NRel-4 and NZel-1 cells, which were defective in DHAP-AT or alkyl-DHAP synthase, respectively ( 19,20 ), did not produce a detectable amount of the 1-alkyl-2-acyl form of GPI ( 18 ). It is conceivable that a putative alkyl-containing lipid donor is generated in the ER from alkyl-DHAP and is used in the remodeling reaction that converts the diacyl form GlcN-acyl-PI to the 1-alkyl-2-acyl form ( Fig. 1 ).
These results suggested that remodeling of the diacyl form of GPI to the 1-alkyl-2-acyl form would not occur in cells from patients with RCDP types 2 and 3. Cells from In rhizomelic chondrodysplasia punctata (RCDP) type 1, which is characterized by lethality between 2-3 years of age, a typical facial appearance, cataracts, skeletal dysplasia, microcephaly, and severe psychomotor defects, biosynthesis of alkyl-phospholipids and plasmalogens, and degradation of phytanic acid derived from phytol in chlorophyl are defective, whereas peroxisome membrane biogenesis per se and incorporation of PTS1-bearing proteins are not severely affected. RCDP type 1 is caused by mutation in PEX7 that is essential for transferring the minor group of enzymes bearing a PTS2, such as 3-ketoacyl-CoA thiolase of fatty-acid ␤ -oxidation pathway, alkyl-dihydroxyacetone phosphate synthase (alkyl-DHAP synthase), which is required for synthesis of alkyl-phospholipids, and phytanoyl-CoA 2-hydroxylase, which is required for degradation of phytanic acid ( 3 ). Specifi c defects in DHAPacyltransferase (DHAP-AT) and alkyl-DHAP synthase (also called alkylglycerone phosphate synthase), the fi rst two enzymes in the alkyl-phospholipid biosynthetic pathway, cause disorders similar to RCDP type 1, termed RCDP type 2 and type 3, respectively (10)(11)(12), indicating that the major symptoms of RCDP are due to defective biosynthesis of plasmalogens and/or other alkyl-phospholipids, such as platelet activating factor.

Mass spectrometric analysis of PI from GPI-AP and HFGF-CD59
We used approximately 1 nmol (50 g) of HFGF-CD59 for mass analysis of the PI moiety. The PI portion was isolated from affi nitypurifi ed HFGF-CD59 by deamination with sodium nitrite as described previously ( 18 ). PI derived from HFGF-CD59 was stored at -80°C until MS analysis. In CHO cells, data are cited from reference 18, and the method was described in same reference. For analysis of PI in ZPG207 cells, Nano ESI-MS/MS analysis was performed in negative ion mode using a 4000Q TRAP (AB Sciex, Foster City, CA) with a chip-based ionization source, TriVersa NanoMate (Advion BioSystems, Ithaca, NY). The ion spray voltage was set at -1.2 kV, the gas pressure at 0.3 psi, and the fl ow rate at 200 nl/min. The parameter settings were m/z 400-1,200 for scan range, -100 V for declustering potential, -50 to -60 volts for collision energy, and Q1/Q3 "unit" resolution. Samples were dissolved in C/M (1:2) containing 5 mM ammonium formate for injection into the mass spectrometer. The molecular species of PI that were liberated from HFGF-CD59 were directly subjected to fl ow injection and selectively analyzed by precursor ion scanning of the phosphoryl inositol part ( 59 ). The structure of each PI molecular species was confi rmed by MS/MS analysis of the precursor ion. 3

H-mannose and a test for the alkali resistance of GPI
Before labeling, cells (3 × 10 6 in a 60-mm dish) were cultured in medium containing 10 M BE49385A/YW3548, a PIG-N patients with RCDP type 1 might also be defective in this remodeling. It is also likely that cells from patients with ZS are defective in generating 1-alkyl-2-acyl GPI. In this study, we investigated these hypotheses using CHO cells defective in PEX7, PEX5, or PEX19 and fi broblasts from patients with RCDP types 1, 2, or 3, or ZS (21)(22)(23). Our results indicate that these cells generate only the diacyl form of GPI.

Cells and materials
Fibroblast cell lines defective in PEX7, termed Gifu R01 and R03, were derived from two patients with RCDP type 1; those defective in PEX5, termed 2-07, and those defective in PEX16, termed D-01, were from patients with ZS; those defective in DHAP-AT, termed NL-#2 and #3, and those defective in alkyl-DHAP synthase, termed NL-#5 and #6, were from patients with RCDP type 2 and patients with RCDP type 3, respectively ( 24,54 ). This study was approved by internal review committees in the Research Institute for Microbial Diseases of Osaka University, Gifu University School of Medicine, and University of Amsterdam. Written informed consent for specifi c research use was not obtained because all patients died within a few years after birth. The research use of their fi broblast cells that were taken for diagnostic purposes and stored frozen was approved by the internal review committees because these are rare lethal diseases for which investigation of the disease mechanism is important for the wellness of patients with ZS and RCDP and their families. To generate immortalized cell lines, we introduced cDNAs of human telomerase reverse transcriptase and SV40-T into fi broblast cell lines using a retrovirus vector ( 55-58 ). The immortalized cells were cultured in MEM medium with 10% FBS supplemented with 100 g/ml of G418 and 0.5 g/ml of puromycin to maintain plasmids. To complement RCDP fi broblasts, human PEX7 and DHAP-AT cDNAs were transduced using a retrovirus vector with a blasticidin resistance gene, and the cells were maintained in 10 g/ml of blasticidin (InvivoGen, San Diego, CA). The PEX mutant and complemented CHO cells have been reported previously ( 21,22,34 ). ZPG207, ZP139, and ZP119 are defective in PEX7, PEX5, and PEX19, respectively. We used ZPG207 complemented with human PEX7 cDNA, termed ZPG207P7 ( 22 ), and ZP119 complemented with human PEX19 , termed 119/P19 ( 34 ). ZPG207 complemented with FLAG-tagged hamster Pex7 , termed

Fig. 2.
Transfection of human PEX7 restored the localization of PTS2-bearing protein at the peroxisome in PEX7-defective CHO cells. PTS2-GFP is diffuse in PEX7-defective ZPG207 cells (top panels), whereas a punctate distribution of PTS2-GFP was restored by transfection with human PEX7 (bottom panels). Cells were observed for GFP (left-hand panels) and stained with anti-catalase antibody (middle panels). Merged profi les are shown in the right-hand panels. Scale bars, 20 m. Fig. 3. GPI lipid remodeling to generate 1-alkyl-2-acyl GPI is defective in RCDP type1 cells. A: Schematic showing the alkali-sensitive and -resistant characteristics of diacyl and 1-alkyl-2-acyl GPI intermediates, respectively. After KOH treatment, 1-alkyl-2-acyl GPI is converted to 1-alkyl-2-lyso GPI, which is retained in the lipid fraction. B: Alkali-sensitivity of GPI mannolipids derived from PEX7-defective ZPG207 cells. by human PEX7 cDNA was confi rmed by the punctate pattern of PTS2-GFP staining that coincided with the staining pattern of catalase, an endogenous peroxisome enzyme ( Fig. 2 ). We then analyzed production of the 1-alkyl-2-acyl form of GPI by testing alkali sensitivity ( Fig. 3A ). Due to the sensitivity of ester linkage to KOH treatment, diacyl form GPI is lost from the organic phase, whereas 1-alkyl-2lyso GPI generated from 1-alkyl-2-acyl from GPI remained in the organic phase and can be detected by TLC analysis. In ZPG207 cells, all spots of GPI biosynthetic intermediates labeled with mannose disappeared after KOH treatment ( Fig. 3B , lane 4), and expression of PEX7 restored the alkali-resistant 1-alkyl-2-lyso forms of GPI intermediates (Fig. 3B, lanes 7 and 8). To determine the structure of the PI moiety of protein-linked GPI, we generated ZPG207 cells stably expressing CD59, a human GPI-anchored protein, bearing tandem His-, FLAG-, glutathione S -transferase (GST)-, and FLAG-tags at the N-terminus (HFGF-CD59). We affi nity-purifi ed HFGF-CD59 with glutathione-Sepharose and released the PI moiety from approximately 1 nmol of protein by treatment with sodium nitrite ( 28,29 ) and subjected it to nano ESI-MS analysis ( Fig. 3C ). All PIs in HF-GF-CD59 from ZPG207 cells were the diacyl form (  ( 18 ). Therefore, PEX7 is required for the production of the 1-alkyl-2-acyl form of GPI anchors.
We next analyzed fi broblast cell lines previously established from patients with RCDP type 1. As in CHO cells, signifi cant fractions of GPI intermediates in wild-type fibroblasts were resistant to alkali, indicating generation of 1-alkyl-2-acyl GPI ( Fig. 3D , lanes 1 and 4). Alkali-resistant GPI was greatly decreased in patients' cells ( Fig. 3D, lanes 5 and 6). After complementation with human PEX7 cDNA, the alkali-resistant GPIs were generated at levels comparable to those in wild-type cells (Fig. 3D, lanes 9 and 10), establishing that generation of 1-alkyl-2-acyl GPI anchors is defective in RCDP type 1.
inhibitor (a gift from Banyu Pharmaceutical) for 12-16 h ( 60, 61 ). When PIG-N, which transfers ethanolamine-phosphate sidebranch to the fi rst Man, is inhibited, Man-containing GPI intermediates lacking the side-branch are accumulated, and effi cient radio-labeling is achieved ( 60 ). Cells were then incubated in 2.5 ml of glucose-free RPMI-1640 medium (Gibco/Invitrogen) containing 10 g/ml tunicamycin (Wako Pure Chemical Industries, Ltd.), 10% dialyzed FBS, 20 mM HEPES, and 100 g/ml D-glucose for 1 h. After tunicamycin treatment, 40 Ci/ml for CHO cells or 10 Ci/ml for fi broblasts of D-[2-3 H(N)]mannose (American Radiolabeled Chemicals) was added, and incubation was continued for 1 h. Lipids were extracted from the cell pellet using two 300-l aliquots of water-saturated butanol (BuOH). The extracted lipids were treated with 500 l of 0.1 N KOH in methanol (MeOH) or MeOH only for 1 h at 37°C. To neutralize the solutions, 50 l of 1 M acetic acid or water was added, and the samples were evaporated to dryness using a speed-vac. The lipids were extracted using two 300-l aliquots of water-saturated BuOH and back-washed with 200 l of BuOH-saturated water. After evaporation, the extracted lipids were used as samples for TLC on activated high-performance TLC silica gel 60 plates (Merck) with a developing solvent system of chloroform/MeOH/H 2 O (10:10:3).

GPI lipid remodeling to generate 1-alkyl-2-acyl GPI is defective in PEX7 mutant CHO cells and fi broblasts from patients with RCDP type 1
PEX7 is required for transporting proteins bearing a PTS2, such as alkyl-DHAP synthase, into the peroxisome and is a causal gene for RCDP type 1 (24)(25)(26)(27). We tested whether a defect in PEX7 affects the lipid remodeling of GPI. We fi rst tested the PEX7 -mutant CHO cell line ZPG207 and ZPG207 transfected with human PEX7 cDNA ( 22 ). Normalization of the PTS2-bearing protein distribution

Fibroblasts from patients with RCDP types 2 and 3 are defective in 1-alkyl-2-acyl GPIs
We next assessed generation of alkali-resistant GPI in fibroblast cell lines from patients with RCDP types 2 and 3. As expected from our previous results with mutant CHO cells defective in DHAP-AT and alkyl-DHAP synthase, all spots of GPI-mannolipids generated in these cells from RCDP type 2 and type 3 patients disappeared after alkali treatment ( Fig. 4 , lanes 7-10). When the responsible DHAP-AT cDNA was transfected into RCDP type 2 fi broblasts, spots of the alkali-resistant lyso-forms appeared (Fig. 4, lanes 13 and  14). It is, therefore, evident that generation of 1-alkyl-2-acyl GPI-APs is defective in RCDP types 2 and 3.

Surface expression of urokinase-type plasminogen activator receptor, a GPI-AP, signifi cantly increase when alkylphospholipid biosynthesis is defective
We next asked if a defect in peroxisome-dependent alkylphospholipid biosynthesis affects any properties of GPI-APs other than lipid remodeling to 1-alkyl-2-acyl form. We focused on urokinase-type plasminogen activator receptor (uPAR), an endogenous GPI-AP of CHO cells. We used NRel-4 and NRel-1 cells defective in DHAP-AT and alkyl-DHAP synthase, respectively ( 19,20 ). The surface expression of uPAR on NRel-4 cells was 4-times of that on wild-type CHO-K1 cells and returned to 1.5-times of normal level by transfection of DHAP-AT cDNA ( Fig. 6A ). Similarly, uPAR on NZel-1 cells was 2-times of that on CHO-K1 cells and returned to 1.4-times of normal level by transfection of alkyl-DHAP synthase cDNA ( Fig. 6A ). To see if the increased surface expression of uPAR was due to an increase in total cellular uPAR level or to a skew in a ratio of surface to intracellular uPAR fractions, we used confocal fl uorescence microscopy. The total cellular level of uPAR was higher in NRel-4 and NZel-1 cells than in CHO-K1 cells and was substantially decreased after transfection of DHAP-AT and alkyl-DHAP synthase cDNAs, respectively (Fig. 6B). We assessed the amounts of uPAR and some nonGPI-APs, such as calnexin, ␣ -tubulin, glyceraldehydephosphate dehydrogenase, and caveolin-1, with Western blotting before and after PI-PLC treatment ( Fig. 6C ). Most of the uPAR was lost by PI-PLC treatment, indicating that most cellular uPAR was on the cell surface (Fig. 6C, top panel, lanes 1-5  vs. lanes 6-10). Levels of uPAR were clearly higher in removal of the inositol-linked acyl chain from nascent GPI-APs does not occur due to a defect in the inositol-deacylase PGAP1, transport of GPI-APs from the ER to the Golgi apparatus is delayed, and sensitivity to PI-specifi c phospholipase C is lost ( 37 ). Pgap1 -knockout mice have high rates of perinatal death due to craniofacial deformity (in particular otocephaly), growth retardation, and male infertility ( 38 ). The mutant mouse strain oto characterized by otocephaly is due to mutation in Pgap1 ( 39 ). Another example is GPI fatty acid remodeling occurring in the Golgi, in which an unsaturated fatty acid at the sn2 position is exchanged with saturated, stearic acid. When the fatty acid remodeling does not occur due to a defect in PGAP3, which is involved in removal of the sn2-linked unsaturated fatty acid, GPI-APs are present in the detergent-sensitive fraction rather than in the detergent-resistant membrane fraction after extraction with a cold nonionic detergent such as Triton X-100 ( 40 ). Pgap3 -knockout mice have a number of abnormal phenotypes, including growth retardation and immune dysfunction ( 41 ). Overall, it is possible that a lack of 1-alkyl-2-acyl GPI-APs accounts for some of the clinical symptoms of ZS and RCDP.
Mammalian cells elaborate 1-alkyl-2-acyl GPI-APs from diacyl PI via at least two events. The fi rst event, which occurs in the third biosynthetic intermediate GlcN-acyl-PI, is replacement of a diacyl glycerol moiety or a diacyl glycerol phosphate moiety ( 17,18 ). In CHO cells and mouse Tlymphoma cell lines, approximately 60% of GPIs are converted to the 1-alkyl-2-acyl form at this stage ( 17 ). Because NRel-4 and NZel-1 cells than in parental CHO cells (Fig.  6C, lanes 2 and 4 vs. lane 1) and were substantially decreased after transfection of cDNAs responsible for the mutants (Fig. 6C, lanes 3 and 5). In contrast, expression levels of calnexin, ␣ -tubulin, and glyceraldehydephosphate dehydrogenase were not changed with or without alkyl phospholipids synthesis (Fig. 6C, three middle panels). Consistent with a previous report that levels of GPI-APs and caveolin-1 are inversely correlated in CHO cells ( 35 ), we found that levels of caveolin-1 were decreased in NRel-4 and NZel-1 cells and were increased after transfection of responsible cDNAs (Fig. 6C, bottom panel). These results suggest that the defective alkylphospholipid synthesis causes increased surface expression of uPAR probably due to slowed turnover. Whether this remarkable phenotype is due to the diacyl form GPI anchor itself, to a lack of alkyl/alkenyl phospholipids such as plasmalogens, or both is yet to be determined.

DISCUSSION
The primary fi nding of this study is that cells from patients with RCDP and ZS are defective in biosynthesis of 1-alkyl-2-acyl GPIs and express only diacyl GPI-APs. Many proteins are anchored by GPI to the cell surface, and the majority of proteins in mammalian cells contain 1-alkyl-2acyl GPI ( 36 ). It is known that changes in the fi ne structure of the lipid portion of GPI-APs lead to altered properties of the GPI-APs and various abnormalities. When phosphatase, a GPI-AP that is expressed on the membrane of osteoblasts and that is secreted to the blood during bone formation ( 14,(46)(47)(48)(49)(50), might be responsible for the abnormal bone formation or chondrodysplasia. Delayed cerebellar development is a hallmark in RCDP and ZS patients, and these patients suffer from severe neonatal neurodegenerative disorders. The severity is related to the importance of peroxisomes in the maturation of the central nervous system. Peroxisomes are abundant at the termini of developing neurons and have been implicated in the early determination of neural polarity ( 51,52 ). Although the mechanisms of neuropathology are unknown, ZS leads to loss of essential peroxisomal function and causes problems in cerebellum formation. Neural cell adhesion molecule, a GPI-AP, is expressed on the surface of neural cells and contributes to cell polarization ( 53 ). The abnormal diacyl form GPI might be causally related to the neurodegenerative phenotypes. Although the functional importance of the alkyl-acyl form of GPI-APs is yet to be determined, our results suggest the possibility that the absence of alkyl-acyl GPI contributes to some of the symptoms of RCDP and ZS. fatty acid composition of the diacyl form of GlcN-acyl-PI was different from those of earlier GPI intermediates and PI, even the diacyl form must be subjected to replacement reaction ( 18 ). Therefore, a putative donor lipid for the remodeling presumably comprises 60% of the 1-alkyl-2acyl form and 40% of the diacyl form ( 18 ). The exact reaction mechanism of this GPI lipid remodeling is yet to be elucidated. The second event occurs during the following biosynthetic step(s), in which enrichment of the 1-alkyl-2acyl form proceeds further, leading to occupation of 80-90% of cell surface GPI-APs in CHO cells ( 18 ). A possible mechanism might be selection by biosynthetic enzymes, selective compartmentalization of the 1-alkyl-2-acyl form, or, although less likely, degradation of diacyl GPI intermediates.
To understand the functional signifi cance of 1-alkyl-2acyl GPI-APs, the enzyme required for the GPI lipid remodeling needs to be identifi ed and disrupted. If the gene for the putative enzyme that catalyzes diacyl to 1-alkyl-2acyl exchange in GlcN-acyl-PI is disrupted in mice, such mice would not express 1-alkyl-2-acyl GPI-APs but would have normal plasmalogens; this should allow us to determine the phenotypes that result from defective generation of 1-alkyl-2-acyl GPI-APs. Until this is achieved, information derived from the cells of patients with RCDP types 2 and 3 and CHO mutant cell lines defective in DHAP-AT and alkyl-DHAP synthase is useful. Cells from patients with RCDP types 2 and 3 are defective only in alkyl-phospholipid biosynthesis, resulting in defects in several end-products, namely, plasmalogens, platelet activating factor, and 1-alkyl-2-acyl GPI-APs, whereas cells from patients with RCDP type 1 and ZS are also defective in other peroxisomal pathways. We found that the surface expression of uPAR is 2-to 4-times elevated on mutant cells defective in DHAP-AT or alkyl-DHAP synthase ( Fig. 6 ). It was reported that these mutant CHO cells are ineffi cient in the transport of cholesterol for the cell surface or endocytic compartment to the ER ( 42 ). Fibroblasts from patients with RCDP types 2 and 3 display a number of morphological and cell biological abnormalities, such as the common presence of enlarged and multipolar cells, dilated ER and Golgi cisternae, a reduced number of caveolae associated with a 50-60% reduction in caveolin 1, accumulation of cholesterol in perinuclear structures, an increased level of clathrin, and ineffi cient endocytosis of transferrin ( 43 ). Whether the elevated uPAR expression is related to these cellular abnormalities, including affected cholesterol traffi cking, is unclear.
Dhap-at -knockout mice display a variety of phenotypes, including male infertility, defects in eye development, cataract, optic nerve hypoplasia, and cerebellum anomalies including impaired Purkinje cell innervation, defective myelination, and disturbances in paranode organization ( 44,45 ). Patients with RCDP types 2 and 3 have craniofacial dysmorphism, cataract, severe psychomotor retardation, as well as rhizomelia and chondrodysplasia punctata ( 3,(10)(11)(12). It is tempting to speculate that the lack of a particular 1-alkyl-2-acyl GPI-AP is responsible for some of these abnormalities. For example, abnormal alkaline