Improvement of Dolichol-linked Oligosaccharide Biosynthesis by the Squalene Synthase Inhibitor Zaragozic Acid*

The majority of congenital disorders of glycosylation (CDG) are caused by defects of dolichol (Dol)-linked oligosaccharide assembly, which lead to under-occupancy of N-glycosylation sites. Most mutations encountered in CDG are hypomorphic, thus leaving residual activity to the affected biosynthetic enzymes. We hypothesized that increased cellular levels of Dol-linked substrates might compensate for the low biosynthetic activity and thereby improve the output of protein N-glycosylation in CDG. To this end, we investigated the potential of the squalene synthase inhibitor zaragozic acid A to redirect the flow of the polyisoprene pathway toward Dol by lowering cholesterol biosynthesis. The addition of zaragozic acid A to CDG fibroblasts with a Dol-P-Man synthase defect led to the formation of longer Dol-P species and to increased Dol-P-Man levels. This treatment was shown to decrease the pathologic accumulation of incomplete Dol pyrophosphate-GlcNAc2Man5 in Dol-P-Man synthase-deficient fibroblasts. Zaragozic acid A treatment also decreased the amount of truncated protein N-linked oligosaccharides in these CDG fibroblasts. The increased cellular levels of Dol-P-Man and possibly the decreased cholesterol levels in zaragozic acid A-treated cells also led to increased availability of the glycosylphosphatidylinositol anchor as shown by the elevated cell-surface expression of the CD59 protein. This study shows that manipulation of the cellular Dol pool, as achieved by zaragozic acid A addition, may represent a valuable approach to improve N-linked glycosylation in CDG cells.

Congenital disorders of glycosylation (CDG) 3 are a group of inherited defects of protein glycosylation (1). Mutations in genes encoding either proteins involved in biosynthesis of lipid-linked oligosaccharide (LLO) required for N-glycosylation (2) or proteins involved in glycan processing (3,4) or transport of N-glycoproteins (5) form the molecular basis of CDG. The majority of CDG encompass disorders affecting the assembly of the LLO precursor dolichol pyrophosphate (Dol-PP)-GlcNAc 2 Man 9 Glc 3 , which lead to under-occupancy of N-glycosylation sites (6). The stepwise biosynthesis of the LLO precursor begins at the cytosolic side of the endoplasmic reticulum membrane by transfer of GlcNAc-P to Dol-P and completes at the luminal side of the endoplasmic reticulum membrane. Dol-P serves not only as a carrier of maturing LLO but also as a lipid component of Dol-P-Man and Dol-P-Glc, both donor substrates for luminally acting mannosyl-and glucosyltransferases (7).
The symptoms associated with CDG are principally of neurologic nature, such as psychomotor retardation, ataxia, and hypotonia, but also include hormonal alterations and coagulopathies (8). The clinical severity of CDG depends mainly on the degree of N-glycosylation site under-occupancy (9), which itself depends on the available pool of the complete LLO Dol-PP-GlcNAc 2 Man 9 Glc 3 . To date, only two forms of CDG can be successfully treated by oral carbohydrate supplementation. The glycosylation defects resulting from deficiency of Man-P isomerase can be corrected by Man supplementation (10,11), whereas Fuc uptake has been shown to rescue the deficiency of GDP-Fuc transport (12). Considering the involvement of Dol throughout the LLO biosynthetic pathway, we hypothesized that increased cellular levels of Dol-and Dol-P-based substrates may increase the formation of Dol-PP-GlcNAc 2 Man 9 Glc 3 in CDG.
Dol biosynthesis follows the sterol pathway up to the formation of the C 15 -intermediate farnesyl-PP (Fig. 1A) (13). Instead of squalene formation by head-to-head assembly of two farnesyl-PP molecules (14), the consecutive condensation of isopentenyl-PP units leads to the diverging synthesis of polyprenyl-PP, a pre-stage of Dol-P. The enzyme squalene synthase catalyzes the first reaction, leading exclusively to the formation of sterol compounds, such as cholesterol and steroid hormones (Fig. 1A). Inhibition of squalene synthase, e.g. by zaragozic acid A (ZGA), leads to the stimulation of prior diverging pathways and hence to increased formation of Dol and Dol-P (15). ZGA, also known as squalestatin I, was discovered by screening metabolites of filamentous fungi for cholesterol-lowering activity (16). Detailed analysis disclosed that ZGA acts as a competitive inhibitor of squalene synthase by mimicking the farnesyl-PP substrate or the stable intermediate presqualene-PP with its bicyclic, highly acidic core (Fig.  1B). In contrast to statins acting on common steps of both Dol and cholesterol biosynthesis, such as the HMG-CoA re-ductase inhibitors pravastatin (17), lovastatin (18), and rosuvastatin (19), ZGA does not negatively affect the diverging pathways generating Dol, ubiquinone, or prenylated proteins (Fig. 1A). In this study, we investigated the ability of the squalene synthase inhibitor to increase Dol biosynthesis and thereby the output of N-glycosylation in CDG fibroblasts.

EXPERIMENTAL PROCEDURES
Materials-ZGA was a gift from Merck. The C 80 -polyprenyl-P standards were purchased from Larodan Fine Chemicals AB (Malmö, Sweden), and broad-range mammalian Dol-P standards were from Sigma. Acetonitrile (Scharlau, Barcelona, Spain), dichloromethane (Sigma), and water (Sigma) were HPLC grade; other chemicals were analytical grade.
Cholesterol Analysis-Human primary skin fibroblasts (4 ϫ 10 7 cells) cultured in DMEM (Sigma) with 10% fetal calf serum (Bioconcept, Allschwil, Switzerland) at 37°C were treated for 72 h with either 100 M ZGA or dimethyl sulfoxide (DMSO) alone as a negative control. Cells were harvested by trypsinization, washed in PBS, and centrifuged at 1000 ϫ g for 5 min. To the cell pellet was added 1 ml of 4% (w/v) KOH in 90% ethanol. After vortexing, the solution was incubated for 10 min in an ultrasonic bath and transferred to a glass tube. After the addition of 80 l of internal standard (0.1 mg/ml epi-coprostanol (Sigma) in pyridine), the solution was saponified for 60 min at 60°C, mixed with 1 ml of water, and extracted three times with 2 ml of heptane. The pooled heptane extracts were dried under nitrogen and derivatized with 75 l of N,O-bis(trimethylsilyl)trifluoroacetamide (Macherey-Nagel, Oensingen, Switzerland) in 75 l of pyridine at 60°C for 60 min. For GC/MS analysis, the derivative mixture was diluted 5-fold with heptane, and 1 l was injected (injector temperature of 280°C, splitless injection). A Restek Rtx-1MS column (15 m, inner diameter of 0.25 mm; BGB Analytik AG, Böckten, Switzerland) was used for chromatographic separation of the sterols. The carrier gas was helium at a constant flow of 1.5 ml/min. After a dwell time of 3 min at 90°C, the oven temperature was raised to 200°C at 20°C/min and then to 260°C at 1.5°C/min and was finally held at 260°C for 10 min. MS was performed on a Finnigan PolarisQ ion trap mass spectrometer. Mass spectra were acquired in the mass range of m/z ϭ 50 -550.
Determination of Dol-P-Man-Approximately 4 ϫ 10 7 fibroblasts were grown for 72 h in DMEM containing low Glc (5 mM; Sigma) supplemented with 2% fetal calf serum and 100 M ZGA or DMSO as a negative control. Low Glc medium was utilized to achieve improved Man incorporation (23). Cellular Dol-P-Man was metabolically labeled, extracted, and purified according to the protocol of Körner et al. (23). Briefly, the fibroblasts were labeled by incubation in DMEM containing 0.5 mM Glc and 125 Ci of [ 3 H]Man (Hartmann Analytic GmbH, Braunschweig, Germany) for 30 min. Dol-P-Man and short LLO were extracted once with chloroform/ methanol (2:1) and twice with chloroform/methanol (3:2). The combined organic phases were dried and washed. Thinlayer chromatography on Silica Gel 60 plates was performed in chloroform/methanol/water (65:25:4). The plates were analyzed by radiography after signal enhancement with an EN 3 HANCE spray (PerkinElmer Life Sciences), and the areas containing Dol-P-Man were scraped and counted in a Packard Tri-Carb 2900TR liquid scintillation analyzer.
LLO and N-Linked Oligosaccharide (NLO) Analysis-The LLO profiles of 1.5 ϫ 10 7 fibroblasts treated for 72 h with either 100 M ZGA or DMSO were analyzed as described (25). The cells were starved for 45 min in fetal calf serum-and Glcfree DMEM (Invitrogen) and metabolically labeled for 60 min by the addition of 150 Ci of [ 3 H]Man. LLO were extracted from cell pellets, and oligosaccharides were released by mild acid hydrolysis in 0.1 N HCl. Glycoproteins recovered from the LLO extraction were denatured, and NLO were released by overnight incubation with peptide:N-glycosidase F (New England Biolabs) (26). Oligosaccharides were purified by ionexchange chromatography on AG 1-X2 and AG 50W-X8 resins (Bio-Rad) and by hydrophobic chromatography on Supelclean ENVI-Carb 120/400 beads (Supelco) and Sep-Pak C 18 columns and subjected to HPLC analysis.
Statistics-Results are expressed as means Ϯ S.E. A oneway analysis of variance test with the Bonferroni multiple comparison post-test was applied to confirm differences between groups. Significance was accepted for p Ͻ 0.05.

RESULTS
The toxicity of ZGA was first determined by incubating human primary skin fibroblasts for 10 days with increasing concentrations of 10 -500 M. ZGA was tolerated by fibroblasts up to a concentration of 125 M. Above that concentration, the rate of cell proliferation slowed down, and cell morphology was altered (data not shown).The addition of 100 M ZGA to healthy control and CDG fibroblasts led to a moderate decrease in cellular cholesterol levels by 15 and 30%, respectively (Fig. 2).
The effect of ZGA on glycosylation was determined by measuring cellular levels of Dol-P, Dol-P-Man, a glycosylphosphatidylinositol (GPI)-anchored model protein, LLO, and NLO. The impact of ZGA on Dol-P levels was first addressed using healthy control fibroblasts. After labeling with the fluorochrome 9-anthryldiazomethane, Dol-P levels were quantitated after HPLC separation (21). The resulting fluorescent HPLC profiles of untreated cells (Fig. 3A) and of cells treated with 100 M ZGA for 72 h (Fig. 3B) were compared. The pattern of Dol-P species changed upon ZGA treatment, with the amount of the longer C 100 -Dol-P and C 105 -Dol-P species increasing in ZGA-treated fibroblasts (Fig. 3C).
To address the effect of ZGA on various parameters of LLO biosynthesis in CDG, we chose fibroblasts with a deficiency of DPM1 (Dol-P-Man synthase-1) (24), which present low Dol-P-Man levels and an accumulation of the LLO Dol-PP-GlcNAc 2 Man 5 and the corresponding GlcNAc 2 Man 5 glycan structure on glycoproteins. Dol-P-Man production was measured after labeling cells with [ 3 H]Man. The [ 3 H]Dol-P-Man pool determined in DPM1-deficient fibroblasts reached 65% of the levels measured in healthy control cells (Fig. 4). When DPM1-deficient fibroblasts were treated with 100 M ZGA for 72 h, the [ 3 H]Dol-P-Man levels increased by reaching 120% of normal values. Similarly, when healthy control fibroblasts were incubated with 100 M ZGA, the levels of [ 3 H]Dol-P-Man increased by 150% (Fig. 4).
DPM1 deficiency leads to reduced cell-surface expression of GPI-anchored proteins because Dol-P-Man is required for the assembly of the GPI anchor (27). Accordingly, low levels of the GPI-anchored protein CD59 have been detected on the cell surface of DPM1-deficient fibroblasts (24). The addition of ZGA led to increased expression of CD59 in healthy control (Fig. 5A) and DPM1-deficient fibroblasts (Fig. 5B) as monitored by flow cytometry. When DPM1-deficient fibroblasts were treated with 100 M ZGA for 72 h, the cell-surface levels of CD59 returned to the levels observed in untreated control fibroblasts (Fig.  5C). A time course experiment showed that the ZGA effect was maximal by 72 h of treatment, whereas longer periods did not increase CD59 expression further (Fig. 5C). The effect on CD59 expression was also maximal upon application of  FEBRUARY 25, 2011 • VOLUME 286 • NUMBER 8 ZGA at 100 M, but changes were already visible with lower ZGA concentrations (Fig. 5D).

Zaragozic Acid Improves N-Glycosylation
The biosynthesis of the LLO Dol-PP-GlcNAc 2 Man 9 Glc 3 requires Dol as a carrier of the growing oligosaccharide and as a donor substrate for Dol-P-Man and Dol-P-Glc (7). In normal cells, only the complete LLO Dol-PP-GlcNAc 2 Man 9 Glc 3 could be detected (Fig. 6A). DPM1 deficiency led to the accumulation of the intermediate LLO Dol-PP-GlcNAc 2 Man 5 (Fig. 6C) (24,28). The amount of this incomplete LLO could be reduced by treatment of DPM1-deficient fibroblasts with 100 M ZGA. The ratio of the abnormal Dol-PP-GlcNAc 2 Man 5 peak to the mature Dol-PP-GlcNAc 2 Man 9 Glc 3 peak was decreased from 54 to 17% upon ZGA supplementation (Fig. 6D). By comparison, the addition of 100 M ZGA to control fibroblasts did not influence the quality of the LLO profile (Fig. 6B).
After transfer from LLO to proteins, the NLO GlcNAc 2 Man 9 Glc 3 is trimmed by glucosidases and mannosidases in the endoplasmic reticulum (29). The analysis of NLO after [ 3 H]Man labeling of control fibroblasts showed, as expected, GlcNAc 2 Man 8 , GlcNAc 2 Man 9 , and GlcNAc 2 Man 9 Glc 1 as main oligosaccharide structures (Fig. 7A). In DPM1-deficient fibroblasts, several intermediary oligosaccharide structures ranging from GlcNAc 2 Man 4 to GlcNAc 2 Man 7 were found, whereas the normal oligosaccharides GlcNAc 2 Man 8 , GlcNAc 2 Man 9 , and GlcNAc 2 Man 9 Glc 1 were under-represented (Fig. 7C). Although ZGA treatment did not affect the NLO profile of control fibroblasts (Fig. 7B), it decreased the occurrence of abnormal N-linked glycans GlcNAc 2 Man 4 -7 and increased the amounts of the normal NLO GlcNAc 2 Man 8 , GlcNAc 2 Man 9 , and GlcNAc 2 Man 9 Glc 1 in DPM1-deficient fibroblasts (Fig. 7D). In these cells, the ratio of the pathologic GlcNAc 2 Man 5 peak to the normal GlcNAc 2 Man 9 peak was 422%, and this ratio was decreased to 143% after the addition of ZGA (Fig. 7D). This normalization of the NLO profile demonstrated the beneficial effect of ZGA treatment on N-glycosylation.
This study shows that treatment of human CDG fibroblasts with the squalene synthase inhibitor ZGA stimulated Dol-P biosynthesis and thereby improved the level of Nglycosylation under conditions of limited substrate availability as encountered in DPM1 deficiency. In addition, supplementation with ZGA resulted in normalization of the expression of the GPI-anchored protein CD59 on DPM1-deficient fibroblasts.

DISCUSSION
Successful treatment of biosynthetic CDG is restricted so far to oral administration of Man to Man-P isomerase-deficient CDG patients (10,11). The applied Man can be phosphorylated by hexokinase and allows functional bypassing of the defective isomerization of Fru-6-P to Man-6-P. In the group of Golgi-associated CDG, a deficiency of the Golgi GDP-Fuc transporter (SLC35C1-CDG or leukocyte adhesion FIGURE 3. Analysis and quantification of Dol-P levels in mock-and ZGAtreated control fibroblasts. A, control cells were grown in the presence of DMSO. Cellular Dol-P was labeled with the fluorophore 9-anthryldiazomethane and separated by HPLC. The inset in the fluorescent profile highlights the separation of the particular Dol-P species. The elution times of mammalian Dol-P standards ranging from C 90 -to C 105 -Dol-P are indicated. The internal C 80 -polyprenol-P standard utilized for quantification eluted at a retention time of 33.5 min. B, the separation of Dol-P from ZGA-treated control fibroblasts is presented in inset, and the respective Dol-P species were assigned as in A. Quantification was accomplished by comparison with the internal standard hexadecaprenyl-P, eluting at 34.2 min. C, the Dol-P levels from four independent HPLC runs were calculated and normalized to 10 7 fibroblasts. White bars, Dol-P from mock-treated cells; gray bars, Dol-P from ZGA-treated cells. Data are means Ϯ S.E. of four independent experiments. *, p Ͻ 0.05. deficiency type II) (30) can also be treated by simple supplementation of a monosaccharide. Oral Fuc administration induces the expression of fucosylated glycoproteins, and within short terms, the clinical symptoms of a SLC35C1-CDG patient can be relieved (12). Moreover, a few elementary attempts have been conducted to treat PMM2 (phosphomannomutase-2) deficiency (PMM2-CDG), which forms by far the largest CDG subtype (31). Mutations in the PMM2 gene lead to disrupted conversion of Man-6-P to Man-1-P, which acts as precursor of GDP-Man (32). However, Man-1-P is not able to diffuse through biological membranes, and a transport system does not exist. Thus, it needs chemical modification prior to direct administration to overcome this obstacle (33)(34)(35). Membrane-permeable Man-1-P analogs are successful in re-storing LLO biosynthesis in CDG patient cells, but such prodrugs are toxic and very unstable and have to be applied in high concentrations (33).
Shang and Lehrman (36) discovered a metformin-stimulated Man-specific transport activity in human fibroblasts. The resulting increased Man uptake was shown to be able to correct artificially induced defects in LLO biosynthesis and protein N-glycosylation in control and PMM2-CDG fibroblasts. However, the observation that the clinical phenotype of PMM2 deficiency cannot be treated with dietary Man, which is actually effective in cell models (31), renders a therapeutic administration of metformin questionable.  Our approach to increase Dol-P biosynthesis by inhibition of the squalene synthase offers a therapeutic perspective for inherited biosynthetic N-glycosylation defects featuring residual enzymatic activity. Conveniently, ZGA is well tolerated in animal experiments (37,38) compared with classical statins, which interfere with various isoprenoid-based biosynthesis pathways and are associated with adverse effects (39). We could show that ZGA decreased cellular cholesterol levels, although only by 15-30%, indicating that membrane properties were likely unaltered due to this treatment. However, we cannot exclude an effect of the lowered cholesterol levels on the formation of glycolipid rafts, in which GPI-anchored proteins partition. Accordingly, the increased expression of the GPI-anchored protein CD59 observed in ZGA-treated cells could be caused by increased Dol-P availability and altered raft distribution. By contrast, the positive effects of ZGA supplementation on Dol-P-Man levels and LLO patterns in DPM1-CDG fibroblasts are unlikely to be related to decreased cholesterol biosynthesis because the endoplasmic reticulum membrane is low in cholesterol levels (40). Nevertheless, the effects of ZGA on protein modifications, such as farnesylation and geranylgeranylation, have not been investigated, so we cannot dismiss an impact of such modifications on the Nglycosylation pathway at this stage.
This study was based on fibroblasts deficient in DPM1 because this form of CDG shows several measurable abnormalities along the biosynthesis of N-glycans and GPI-anchored proteins. By contrast, deficiency of PMM2, which represents the largest group of CDG cases (6), does not feature accumulating LLO intermediates. Consequently, incomplete oligosaccharides on glycoproteins are usually not encountered in PMM2 deficiency. However, the up-regulation of Dol-P-dependent substrates by ZGA may be beneficial in normalizing the glycosylation disorders caused by deficiency of glycosyltransferases catalyzing the attachment of the last four Man and three Glc residues of the LLO precursor (7).
Over the last decade, knock-out mouse models for PMM2, Man-P isomerase, and DPAGT1 deficiency have been generated (41)(42)(43). These works clearly showed that disruption of LLO biosynthesis leads to early embryonic lethality. In this context, it would be interesting to investigate whether the embryonic lethality of the homozygotes could be delayed to a later stage of development by treating pregnant heterozygous mice with ZGA according to Keller (15). In addition, the generation of viable CDG mouse models, e.g. by introduction of selected point mutations, would allow detailed analysis of organ-specific effects of ZGA in the context of inherited N-glycosylation deficiencies. This seems to be of particular interest considering that Keller reported an enormous effect of ZGA on Dol and Dol-P pools in rat livers, but not in other organs, such as brain, kidney, intestine, and testis (15), which was similarly observed in different brain cells (44). Subcutaneously administered ZGA might preferentially be taken up by mammalian livers via a specific hepatic transport mechanism, which was likewise proposed in other studies (37,38). Considering the frequent impairment of liver function in CDG patients (8), the proposed hepatic transport system for ZGA would suggest that liver N-glycosylation may benefit from treatment with the compound. Along this line, it would be of interest to address the effect of ZGA supplementation in hepatic cell lines because fibroblasts do not represent suitable cell models to study glycoprotein secretion. The generation of hepatic cell lines with inactivated glycosylation genes will represent an opportunity to study the regulation of N-glycosylation by ZGA and similar compounds. In conclusion, this study suggests new perspectives in developing treatments of glycosylation deficiency by showing that manipulation of the dolichol biosynthesis pathway, as shown here by ZGA administration, represents a valid option.