Glucosylated cholesterol in mammalian cells and tissues: formation and degradation by multiple cellular β-glucosidases

The membrane lipid glucosylceramide (GlcCer) is continuously formed and degraded. Cells express two GlcCer-degrading β-glucosidases, glucocerebrosidase (GBA) and GBA2, located in and outside the lysosome, respectively. Here we demonstrate that through transglucosylation both GBA and GBA2 are able to catalyze in vitro the transfer of glucosyl-moieties from GlcCer to cholesterol, and vice versa. Furthermore, the natural occurrence of 1-O-cholesteryl-β-D-glucopyranoside (GlcChol) in mouse tissues and human plasma is demonstrated using LC-MS/MS and 13C6-labeled GlcChol as internal standard. In cells, the inhibition of GBA increases GlcChol, whereas inhibition of GBA2 decreases glucosylated sterol. Similarly, in GBA2-deficient mice, GlcChol is reduced. Depletion of GlcCer by inhibition of GlcCer synthase decreases GlcChol in cells and likewise in plasma of inhibitor-treated Gaucher disease patients. In tissues of mice with Niemann-Pick type C disease, a condition characterized by intralysosomal accumulation of cholesterol, marked elevations in GlcChol occur as well. When lysosomal accumulation of cholesterol is induced in cultured cells, GlcChol is formed via lysosomal GBA. This illustrates that reversible transglucosylation reactions are highly dependent on local availability of suitable acceptors. In conclusion, mammalian tissues contain GlcChol formed by transglucosylation through β-glucosidases using GlcCer as donor. Our findings reveal a novel metabolic function for GlcCer.

endoplasmic reticulum in hepatocytes ( 22 ), at the endoplasmic reticulum and Golgi apparatus in HEK293 cells overexpressing enzyme ( 25 ), and at the endosomes in fibroblasts and COS-7 cells ( 23 ). GBA2 degrades GlcCer without need for an activator protein, and further differs from GBA in noted artifi cial substrate and inhibitor specifi city ( 21 ). Finally, some tissues express the enzyme GBA3, also referred to as broad-specifi c cytosolic ␤ -glucosidase ( 26 ). This enzyme shows relatively poor in vitro hydrolytic activity toward GlcCer and is thought to be primarily involved in detoxifi cation of glucosylated xenobiotics ( 26 ). All three human retaining ␤ -glucosidases employ the double displacement mechanism in catalysis. There are many documented examples of transglucosylation mediated by retaining glycosidases ( 27 ). Therefore, theoretically, in addition to GBA, GBA2 and GBA3 might also generate GlcChol .
Modifi cation of cholesterol by glucosylation changes the physicochemical properties of the sterol, rendering it far more water soluble. Given the potential physiological relevance, the natural occurrence of GlcChol and its metabolism in cells and tissues are of interest. We therefore studied the existence of the glucosylated sterol in mammalian tissues. For this purpose 13 C 6 -isotope-labeled GlcChol was synthesized to be used as internal standard in sensitive quantitative detection of GlcChol by LC-MS/MS. Here we demonstrate the natural occurrence of GlcChol in mammalian cells and tissues. Moreover, we document the ability of both GBA and GBA2 to degrade, as well as synthesize, GlcChol. The importance of substrate and acceptor concentrations regarding the action of GBA and GBA2 in GlcChol metabolism is experimentally demonstrated. Our investigation demonstrates the surprising versatility of ␤ -glucosidases, a fi nding discussed in relation to metabolism of sphingolipids and sterols in health and disease. Glucosylceramide (GlcCer), the intermediate in biosynthesis and degradation of more complex glycosphingolipids (GSLs), is ubiquitous in mammalian cells, particularly located in the cell membrane ( 1 ). Its presence in plants and some fungi is also documented. Glucosyldiacylglycerol has been identifi ed in various plants, but its presence in mammalian cells is comparatively poorly documented ( 2,3 ). Likewise, sterol-glucosides are known to occur in plants and fungal species ( 4 ), but their existence in mammalian cells has not been extensively studied. Indications of the existence of glucosyl-␤ -D-cholesterol or 1-O -cholesteryl-␤ -D-glucopyranoside (GlcChol) in mammalian cells were fi rst provided by Murofushi and coworkers ( 5,6 ). They described its occurrence in cultured human fi broblasts and gastric mucosa ( 5,6 ). Heat shock was found to increase biosynthesis of GlcChol and, subsequently, induce HSP70 ( 7 ). GlcCer is formed by the enzyme GlcCer synthase (GCS, EC2.4.1.80). This transferase, fi rst cloned by Hirabayashi and colleagues ( 8 ), is located at the cytosolic leafl et of the Golgi apparatus where it transfers the glucosemoiety from UDP-glucose to ceramide ( 9 ). In a recent study, Akiyama et al. ( 10 ) showed that GCS does not synthesize GlcChol. They noticed that GM-95 cells defi cient in GCS are unable to synthesize GlcChol without the addition of exogenous GlcCer. Furthermore, the same researchers demonstrated that, at least in vitro, the lysosomal enzyme glucocerebrosidase (GBA; E.C.3.2.1.45) generates, through transglucosylation, 25-[ N -[(7-nitro-2-1,3-benzoxadiazol-4-yl) methyl] amino]-27-norcholesterol (25-NBD-cholesterol)glucoside from GlcCer and artifi cial 25-NBD-cholesterol ( 11 ). Such ability of GBA to perform transglucosylation was earlier demonstrated by Glew and coworkers, showing catalyzed transfer of the glucose moiety from 4methylumbelliferyl-␤ -glucoside to retinol and other alcohols ( 12 ).

Materials
The enzyme GBA is well-studied because its defi ciency underlies Gaucher disease (GD), a relatively common lysosomal storage disease ( 13 ). Assisted by the small activator protein, saposin C, GBA degrades GlcCer to ceramide and glucose in lysosomes, the penultimate step in GSL catabolism ( 13 ). Defi cient GBA activity in GD patients consequently results in accumulation of GlcCer in lysosomes, most prominently in macrophages. These "Gaucher cells" secrete specifi c proteins, as well as glucosylsphingosine (GlcSph), the deacylated form of GlcCer (14)(15)(16). The non-neuronopathic (type 1) variant of GD is presently treated by enzyme replacement therapy (ERT), implying chronic twice-weekly intravenous infusion of macrophagetargeted recombinant enzyme ( 17 ). An alternative treatment of type 1 GD, named substrate reduction, is based on oral administration of an inhibitor of GCS (18)(19)(20).
Mammalian cells and tissues contain other ␤ -glucosidases besides GBA that degrade GlcCer. All cells express the membrane-associated nonlysosomal glucosylceramidase, named GBA2 (21)(22)(23). This enzyme is not defi cient in GD patients. In fact, a compensatory overexpression of GBA2 in materials of GD has been reported ( 24 ). GBA2 has been found to be located outside lysosomes, being noted at the room with a 12 h light/dark cycle and given access to food and water ad libitum. All animal protocols were approved by the Institutional Animal Welfare Committee of the Academic Medical Centre Amsterdam in the Netherlands (DBC101698, DBC100757-115, DBC100757-125, and DBC17AC) The generation of the GD1 mouse model has been described previously ( 32,33 ). Mice were maintained in individually ventilated cages with ad libitum food and water in the animal facility at Lund University Biomedical Center. Breeding and experimental procedures were approved by the Committee for Animal Ethics in Malmö/Lund, Sweden.
Animals were fi rst anesthetized with a dose of Hypnorm (0.315 mg/ml phenyl citrate and 10 mg/ml fl uanisone) and Dormicum (5 mg/ml midazolam) according to their weight. The given dose was 80 l/10 g bodyweight. Animals were euthanized by cervical dislocation. Organs were collected by surgery, rinsed with PBS, directly snap-frozen in liquid nitrogen, and stored at Ϫ 80°C. Later, homogenates were made from the frozen material in 25 mM potassium phosphate buffer (pH 6.5), supplemented with 0.1% (v/v) Triton X-100 and protease inhibitors (4 l of buffer per 1 mg of tissue).
according to the method we established previously ( 11 ) with slight modifications. The reaction mixture, in a total volume of 20 l, contained 100 pmol C6-NBD-GlcCer, 50 mM citratephosphate buffer (pH 6.2), and a desired amount of enzyme. After incubation at 37°C for 30 min, the reaction was terminated by adding chloroform/methanol (2:1, v/v), and the lipids were extracted and analyzed as reported before ( 11 ).

Analysis of GlcChol by LC-MS/MS
A Waters Acquity TM TQD instrument was used in all experiments. The instrument consisted of a UPLC system combined with a tandem quadrupole mass spectrometer as mass analyzer. Data were analyzed with Masslynx 4.1 software (Waters Corporation, Milford, MA). GlcChol and 13 C 6 -GlcChol (internal standard) were separated using a BEH C18 reversed-phase column (2.1 × 50 mm, particle size 1.7 m; Waters Corporation), by applying an isocratic elution of mobile phases, 2-propanol:water 90:10 (v/v) containing 10 mM ammonium formate (eluent A) and methanol containing 10 mM ammonium formate (eluent B). The UPLC program was applied during 5 min consisting of 10% A and 90% B. The divert valve of the mass spectrometer was programmed to discard the UPLC effl uent before (0-0.25 min) and after (4-5 min) the elution of the analytes to prevent system contamination. The fl ow rate was 0.250 ml/min and the retention time of both GlcChol and the internal standard was 1.43 min ( Fig. 1C ). The column temperature and the temperature of the autosampler were kept at 23°C and 10°C, respectively, during the run.
Solutions of GlcChol and 13 C 6 -GlcChol and a mixture of both compounds were prepared with concentrations of 1 pmol/ l in 5 mM ammonium formate in methanol. The compounds were introduced in the mass spectrometer by direct infusion and the optimal tuning conditions for both compounds in ES + (electrospray positive) mode were determined ( Table 1 ) Confi rmation of compounds with m/z 566.6 > 369.4 being GlcChol was performed by demonstration of complete digestion to cholesterol using excess rGBA at pH 5.2 (McIlvaine buffer). The release of glucose was confi rmed by a glucose oxidase assay.
with 5% FBS and penicillin/streptomycin under 5% CO 2 at 37°C. Cells were seeded at 75% confl uence in 6-well plates and transfected using FuGENE ® 6 transfection reagent (Promega Benelux, Leiden, The Netherlands) according to the manufacturer's instructions, at a FuGENE:DNA ratio of 3:1. After 24 h, inhibitors of GBA (CBE, 300 M) or GBA2 (AMP-DNM, 20 nM) were added and 48 h later, the medium was removed, cells were washed three times with ice-cold PBS and harvested by scraping in 25 mM potassium-phosphate buffer (pH 6.5). CHO-K1 cells (RCB0285, established by T. T. Puck) were purchased from RIKEN BioResource Center (Ibaraki, Japan) and cultured in Ham's F-12 medium (Nissui) supplemented with 10% FBS under 5% CO 2 at 37°C. cDNA transfection for CHO-K1 cells was carried out using Lipofectamine ® 2000 transfection reagent (Life Technologies) according to the manufacturer's instructions. After 24 h, medium containing transfection reagents was removed and cells were incubated with lysis buffer [50 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 1 tablet/10 ml Complete Protease Inhibitor Cocktail Tablets (Roche, Basel, Switzerland) (pH 7.4)] for 15-30 min at 4°C after washing with PBS. The cells were harvested and centrifuged at 12,000 g for 10 min at 4°C. The obtained supernatants were collected for in vitro enzyme assays.
Identifi cation of newly formed fl uorescent lipid in transglucosylation assays with 25-NBD-cholesterol as acceptor was performed following its isolation by scraping from plates by demonstration of complete digestion to NBD-cholesterol using excess rGBA at pH 5.2 (McIlvaine buffer) in the presence of 0.2% (w/v) sodium taurocholate and 0.1% (v/v) Triton X-100.
Lysates of CHO-K1 cells were used to access the transglucosylase activity and the ␤ -glucosidase activity of GBA3. The assay for transglucosylase activity was performed according to the method we established previously ( 11 ) with slight modifi cations. The reaction mixture, in a total volume of 20 l, contained 40 M 25-NBD-cholesterol, 80 M C16:0-GlcCer, 50 mM citrate-phosphate buffer (pH 6.2), 0.5% CHAPS, 2% ethanol, and the desired amount of enzyme. After incubation at 37°C for 20 h, the reaction was terminated by adding chloroform/methanol (2:1, v/v), and the lipids were extracted and analyzed as reported before ( 11 ). The assay for ␤ -glucosidase activity was performed

Analysis of GlcChol in animal tissues by LC-MS/MS
13 C 6 -labeled GlcChol (3 pmol) in methanol was added to 30 l of mouse tissue homogenate. Next, lipids were extracted according to the method of Bligh and Dyer ( 34 ) and GlcChol was analyzed by LC-MS, as described above.

Analysis of GlcChol in COS-7 cells by LC-MS/MS
COS-7 cells overexpressing GBA2 or GCS were homogenized by sonication on ice. Prior to extraction, 2 pmol of 13 C 6 -labeled GlcChol in methanol (used as an internal standard) was added to 180 l of homogenate. Next, lipids were extracted according to the method of Bligh and Dyer ( 34 ) by addition of methanol, chloroform, and water (1:1:0.9, v/v/v) and the lower phase was taken to dryness under a stream of nitrogen. Isolated lipids were purifi ed by water/butanol extraction (1:1, v/v) and GlcChol was analyzed by LC-MS, as described above.

Analysis of GlcCer and ceramide in COS-7 cells by HPLC
COS-7 cells, overexpressing GBA2 or GCS, were homogenized by sonication on ice. Prior to extraction, 1 nmol of C17-sphinganine in methanol (used as an internal standard) was added to 100 l of homogenate. Next, lipids were extracted according to the method of Bligh and Dyer ( 34 ) by addition of methanol, chloroform, and water (1:1:0.9, v/v/v) and the lower phase taken to dryness under a stream of nitrogen. Isolated lipids were deacylated in a microwave oven, derivatized, and analyzed by HPLC, as described before ( 36 ).

Analysis of GlcChol in RAW264.7 cells by LC-MS/MS
13 C 6 -labeled GlcChol (3 pmol) in methanol was added to 100 l of RAW264.7 cell lysate. Next, lipids were extracted according to the method of Bligh and Dyer ( 34 ) and GlcChol was analyzed by LC-MS, as described above.

Protein concentration
Protein concentration was determined using the Pierce BCA protein assay kit (Thermo Scientifi c) by the microplate procedure. Absorbance was measured in an EL808 Ultra microplate reader (BIO-TEK Instruments Inc.) at 550 nm .

Statistical analysis
Values in the fi gures are presented as a mean ± SD. Data were analyzed by unpaired Student's t -test or Mann-Whitney U test. P values <0.05 were considered signifi cant (* P < 0.05, ** P < 0.01, and *** P < 0.001).

Quantifi cation of GlcChol by LC-MS/MS
To establish whether glucosyl-␤ -D-cholesterol (GlcChol) physiologically occurs in mammals, we fi rst developed a LC-MS/MS procedure for its quantitative detection. For this purpose, a 13 C 6 isotope-labeled GlcChol was synthesized ( 13 C 6 -GlcChol). The use of the isotope-labeled compound as internal standard avoids the need for corrections for extraction effi ciency, chromatographic behavior, and ionization effi ciency during quantifi cation of GlcChol. To prevent undesired adduct formation, lipids were extracted in the absence of additional salts. To stimulate formation of desired ammonium adduct, we incorporated 10 mM ammonium in the eluent.

Collection of Niemann-Pick type C and GD patient plasma
EDTA plasma (15 males and 3 females) was collected prior to therapy from Dutch patients suffering from type 1 GD, known by referral to the Academic Medical Center. Diagnosis of GD in patients was confi rmed by genotyping and demonstration of deficient GBA activity in leukocytes or fi broblasts. Plasma samples were stored frozen at Ϫ 20°C until further use. EDTA plasma samples of 42 male and 47 female control subjects were collected at the Academic Medical Center. The EDTA plasma samples of 15 Niemann-Pick type C (NPC) patients and 9 NPC carriers were collected at the Unidad de Investigación Traslacional in Zaragoza, Spain.
The status of affected or carrier of NPC disease was determined after the exomic sequencing of NPC1 and NPC2 genes, according to the presence of two mutations or one mutation, respectively. Filipin stainings of fi broblasts were conducted to complete the diagnosis study. Plasma samples were stored frozen at Ϫ 20°C until further use. Approval had been obtained from the institutional ethics committee and informed consent according to the Declaration of Helsinki.

Quantifi cation of total GlcChol in human plasma
For quantitative analysis of GlcChol in samples of plasma, we developed a LC-MS/MS method using the multiple reaction monitoring (MRM) mode of the transitions mentioned above . First, GlcChol was extracted from plasma from a healthy individual according the method of Bligh and Dyer ( 34 ) with a few modifi cations. Plasma (20 l) was pipetted in an Eppendorf tube (2 ml) and 20 l of an internal standard solution, containing 0.1 pmol/ l of 13 C 6 -GlcChol in methanol, was added, followed by addition of 280 l methanol and 150 l of chloroform. After brief mixing, the sample was left at room temperature for 30 min, mixed occasionally, and centrifuged for 10 min at 15,700 g to spin down precipitated protein. The supernatant was transferred to an Eppendorf tube and 150 l chloroform and 250 l water were added to induce separation of phases. After centrifugation (5 min at 15,700 g ), the lower hydrophobic phase was transferred to a clean Eppendorf tube and the upper phase was washed by addition of 300 l of chloroform. Lower phases were pooled and taken to dryness at 35°C under a nitrogen stream. Next, the residue was dissolved in 700 l of butanol and 700 l of water, mixed well, and centrifuged for 10 min at 15,700 g . The upper phase (butanol) was transferred to a 1 ml tube with screw cap and the sample was dried under a gentle stream of nitrogen at 35°C. Subsequently, the residue was dissolved in 150 l of eluent B by mixing and sonication, and after centrifugation (5 min at 15,700 g ), an aliquot of 100 l was transferred into an UPLC vial with insert. Ten microliters of the solution were injected for analysis.
Second, for the quantifi cation of GlcChol in plasma, the sample was spiked with GlcChol (concentrations: 0-2.5-5-10-50-100-200-1,000 pmol GlcChol/ml of plasma), internal standard was added and samples were extracted. The ratio, the area from transition GlcChol over the area from the transition 13 C 6 -GlcChol, was plotted against the concentration of GlcChol spiked in the plasma samples. A linear response was obtained over the entire concentration range (y = 0.0108x + 1.9188, R 2 = 0.998). The within-run variation (164.2 ± 4.3 pmol/ml with coeffi cient of variation % 2.6) and between run variation (166.8 ± 3.6 pmol/ml with coeffi cient of variation % 2.2), was determined in plasma of a healthy volunteer by 10 repetitive measurements .
The limit of detection was 0.5 pmol/ml plasma with a signalto-noise ratio of three and the limit of quantitation was 0.9 pmol/ ml plasma with a signal-to-noise ratio of 10. Calculation of the signal-to-noise ratio was done using the peak-to-peak method. ( 22 ) and LIMP-2 KO mice with markedly reduced GBA due to impaired transport to lysosomes ( 31 ). As shown in Fig. 2B-D , the GlcChol concentration was markedly lower in all the tissues of GBA2-defi cient animals analyzed compared with WT controls, especially in thymus. In contrast, in the GBA-defi cient LIMP-2 KO mice, no reduction in GlcChol, but rather a signifi cant increase in liver and plasma levels was observed ( Fig. 2C, D ; P < 0.001 and P < 0.05, respectively). Mice with an induced GBA defi ciency in the white blood cell lineage showed, upon treatment with GCS inhibitor Genz-112638 (eliglustat tartrate, Genzyme), partial reduction in plasma GlcCer ( 37 ) and an almost statistically signifi cant reduction of elevated plasma GlcChol ( Fig. 2E , P = 0.07). An increase in GlcChol was also observed in liver ( Fig. 2F , P < 0.001), spleen, and bone marrow (see supplementary Fig. 1) of mice with induced GD. Correction of GBA defi ciency by lentiviral gene therapy under the control of different promoters led, in all cases, to a concomitant reduction of GlcChol in these tissues ( Fig. 2F ).

In vitro transglycosylation by ␤ -glucosidases
The enzymes GBA and GBA2 were both found able to hydrolyze GlcChol at conditions optimal for degradation of 4MU-Glc (supplementary Table 1).
Sensitive quantitative measurement of GlcChol proved feasible with 13 C 6 isotope-labeled GlcChol as internal standard, as shown in Fig. 1A-C . The limit of detection was 0.5 pmol/ml plasma, with a signal-to-noise ratio of three and a limit of quantitation of 0.9 pmol/ml plasma with a signal-to-noise ratio of 10. GlcChol was found to be an excellent substrate for rGBA (Cerezyme ® ), even at sub-optimal conditions (absence of Triton X-100 and sodium taurocholate) ( Fig. 1D ).

Demonstration of natural occurrence of GlcChol in mice
Next, we examined various tissues of WT mice on the presence of GlcChol. Relative high amounts of glucosylated sterol were noted for thymus, sciatic nerve, brain, and lungs (see Fig. 2A ). The identity of the quantifi ed structure ( m/z 566.6) in thymus (and other tissues) was confi rmed by its digestion by rGBA, showing that in WT animals more than 90% of the lipid measured is indeed glucosyl-␤ -D-cholesterol. However, in sciatic nerve and brain, a signifi cant proportion of m/z 566.6 was not digested by rGBA, suggesting that it may represent another glycosylated cholesterol, for example galactosyl-␤ -D-cholesterol.
The concentration of GlcChol in liver and thymus was determined for tissues of WT mice, animals lacking GBA2  ( Fig. 3A ). Transglucosylation was hardly observed for cell lysates with overexpressed GBA3 and GCS ( Fig. 3A ), recapitulating the fi ndings by Akiyama and coworkers concerning GCS ( 10 ). We repeated the experiment using natural cholesterol as acceptor and determined the levels of formed GlcChol by LC-MS ( Fig.  3B ). Again, GlcChol formation occurred in the presence of GBA (inhibitable by CBE) and in cell lysates with high GBA2 (inhibitable by AMP-DNM) ( Fig. 3B ). Lysates of cells overexpressing GBA3 and GCS showed no additional GlcChol formation (supplementary Fig. 2). The pH optimum of GBA and GBA2 to generate GlcChol was next determined (see supplementary Fig. 3). In the case of GBA, optimal activity was seen between pH 4.5 and 5.5 and for GBA2 between pH 6.0 and 7.0. We investigated, with the same assay, whether GlcSph, GalCer, and GalSph also act The fi ndings on GlcChol levels in WT, GBA-defi cient, and GBA2-defi cient mice prompted us to study the ability of the three ␤ -glucosidases, GBA, GBA2, and GBA3, to generate glucosylated cholesterol by transglucosylation. We fi rst studied this ability in vitro and reproduced the assay of Akiyama and coworkers ( 11 ) using 25-NBD-cholesterol as acceptor and detection of 25-NBD-glucosyl-cholesterol formation by TLC and fl uorescence scanning . As the source of enzyme, we used rGBA and individually overexpressed GBA2, GBA3, and GCS in COS-7 cells. Overexpression of enzymes was checked by measuring enzymatic activity with appropriate assays (not shown).
Recombinant enzyme and COS-7 cell lysates were incubated with natural GlcCer (C18:1-GlcCer) as donor and 25-NBD-cholesterol as acceptor. Following incubation at optimal conditions for each enzyme with inclusion of UDP-Glc for GCS (see Materials and Methods section), lipids were extracted and subjected to HPTLC. As shown in Fig. 3A , rGBA and cell lysates with overexpressed GBA2
We next studied whether natural GlcChol (100 M) can also act as donor in transglucosylation by incubating rGBA and lysates of cells overexpressing GBA2 or GBA3 in the presence of NBD-ceramide (40 M) as acceptor. Lysates of cells overexpressing GBA2 and rGBA showed formation of fl uorescent NBD-GlcCer ( Fig. 3C ). This was not observed for lysates with overexpressed GBA3 ( Fig. 3C ). Transglucosylation by both GBA and GBA2 occurs as an ␤ -Glucosidase activity in lysates of CHO-K1 cells overexpressing GBA3. The activity was measured in the absence or presence of the specifi c ␤ -glucosidase inhibitors: 0.5 mM CBE and/or 0.3 mM NB-DGJ. CBE-and NB-DGJ-sensitive activities were defi ned as activities derived from GBA (black box) and GBA2 (gray box), respectively. CBE-and NB-DGJ-insensitive activity was defi ned as activity derived from GBA3 (dotted box). Mock represents the cells transfected with empty vector. E: Transglucosylase activity in lysates of cells with overexpression of GBA3. Data (n = 3, mean ± SD) were analyzed using an unpaired t -test. * P < 0.05 and ** P < 0.01. Black, gray, and dotted lines show the signifi cant difference of the activity derived from GBA, GBA2, and GBA3, respectively. measured glucosylated sterol was examined by digestion with rGBA. While more than 90% of the GlcChol in the livers of WT mice was digested to cholesterol, in the case of material from NPC mice, this was only around 70% (not shown). Based on this fi nding, it seems likely that part of the elevated compound with m/z 566.6 > 369.4 in NPC liver consists of cholesterol molecules modifi ed differently with sugar, indistinguishable from glucosyl-␤ -D-cholesterol with the LC-MS method.
To experimentally substantiate the observations made for GlcChol in the livers of NPC mice, we induced impaired cholesterol export from lysosomes in murine macrophage RAW264.7 cells by exposure to U18666A ( 42 ). Following lysosomal cholesterol accumulation, cells showed elevated GlcChol. Concomitant inhibition of lysosomal GBA by CBE prevented the increase of GlcChol in U18666A-exposed cells ( Fig. 5C ). Formation of excessive GlcChol was also prevented by the presence of 1 mM ␤ -methyl-cyclodextrin, an agent known to reduce intralysosomal cholesterol in NPC cells ( 43 ). This indicates that during extreme intralysosomal accumulation of cholesterol, GBA actively generates GlcChol. In normal lysosomes GBA most likely largely degrades the glucosylated sterol.

GlcChol in NPC and GD patients
Finally, we determined GlcChol levels in the plasma of untreated symptomatic type 1 GD patients, as well as in NPC patients, carriers, and healthy controls. As shown in Fig. 6A , GlcChol tends to be increased in the plasma of symptomatic GD patients and less prominently in that of NPC patients. The abnormalities in GD patients are more pronounced when plasma GlcChol is related to cholesterol ( Fig. 6B ). Investigation of plasma specimens of type 1 GD patients treated with the GCS inhibitor, eliglustat ( Fig. 6C ), showed a prominent reduction upon inhibition of GSL synthesis. Treatment of matched patients with the weaker GCS inhibitor, miglustat (Zavesca ® , Actelion) also led to a reduction of GlcChol, albeit more slowly ( Fig. 6C ). Of note, treatment of matched patients with rGBA Cerezyme (ERT) did not reduce GlcChol to the same extent, Molecular docking of GlcChol in the GBA crystal structure was performed (see supplementary Fig. 4). The ligand GlcChol was built and regularized with ligand and superimposed on the bicyclic nojirimycin analog ligand that was crystallized in complex with GBA (protein data bank code 2XWE) using the program, Coot (see supplementary Methods ). GlcChol was found to be positioned such that the glucosidic bond is accessible to the catalytic residues Glu235 and Glu340. The cholesterol moiety concomitantly interacts with aromatic side chains.

Metabolism of GlcChol in cultured COS-7 cells
Next we studied factors infl uencing the formation of GlcChol content in cultured green monkey kidney COS-7 cells. We fi rst studied the impact of overexpressed GBA2 and GCS. Figure 4 shows the effect on cellular GlcChol and GlcCer levels. Only overexpression of GCS led to increased levels of GlcCer ( Fig. 4A ). GlcChol was not changed by overexpression of GBA2, but overexpression of GCS caused a 12-fold increase in this lipid. Importantly, inhibition of GBA2 activity with low nanomolar AMP-DNM ( 38 ) resulted in reduced cellular GlcChol. Even in cells with overexpressed GCS, the elevation in GlcChol was prevented ( Fig. 4B ). In contrast, inhibition of GBA with CBE hardly diminished the increased GlcChol level in cells with overexpressed GCS. These fi ndings suggest that GCS does not generate GlcChol itself, but is required to generate suffi cient GlcCer to be used as donor in formation of GlcChol by transglucosylation. This transglucosylation in COS-7 cells is particularly mediated by GBA2, and not GBA. The latter notion is consistent with the fi nding that GBA2 defi ciency in mice, and not that of GBA, is accompanied with reduced GlcChol levels of tissues.

GlcChol in NPC mice and U18666A-treated cells
In NPC disease, cholesterol accumulates prominently in lysosomes as the result of impaired export from the compartment due to defects in either Npc1 or Npc2 . In the livers of spontaneous Npc1 nih/nih ( 29,39,40 ) and Npc1 spm/spm mice ( 30,41 ), we observed a remarkable 25-fold increase in GlcChol content ( Fig. 5A, B ). The identity of the in vitro. Artifi cial ␤ -glucosides, like 4MU-Glc, may act as donor in this reaction in vitro, as well as natural GlcCer . GlcChol is, on the other hand, also an excellent substrate for hydrolysis by GBA. Our fi ndings suggest that GBA normally lowers GlcChol levels. GBA-defi cient LIMP2 KO mice show modestly elevated GlcChol in several tissues and the same is observed in GBA null/fl ox mice with induced type 1 GD ( 32 ). Finally, plasma GlcChol is elevated in symptomatic type 1 GD patients. The actual role played by GBA in GlcChol metabolism, degradation versus synthesis, might be highly dependent on local concentrations of donors (GlcCer and GlcChol) and acceptors (ceramide and cholesterol) in the transglucosylation equilibrium of the enzyme. The importance of this is suggested by some observations made in the course of our investigation. In the fi rst place, high intralysosomal cholesterol concentrations appear to favor formation of GlcChol by GBA. This is indicated by the 25-fold elevated GlcChol in the liver of two different models of NPC disease in mice. In accordance with this, induction of lysosomal cholesterol accumulation with U18666A in cells causes a rapid increase in GlcChol, which was noted to be abolished by inactivation of GBA. Consequently, under normal conditions, GBA seems to promote GlcChol degradation, but under excess cholesterol accumulation in lysosomal membrane, such as that found in NPC, the equilibrium of the metabolism is altered to favor GlcChol formation by GBA.
We observed that the nonlysosomal ␤ -glucosidase, GBA2, can also generate GlcChol through transglucosylation in vitro. Again, GlcCer was found to be an excellent donor for this reaction. GBA2 is equally able to degrade GlcChol. Our fi nding of reduced GlcChol levels in GBA2 KO mice suggests, but does not entirely prove, that GBA2 in vivo contributes to the presence of GlcChol. The enzyme GBA2 is located differently in cells from lysosomal GBA, with its despite impressive clinical improvements in these patients ( Fig. 6C )

DISCUSSION
Recently, Akiyama and colleagues ( 11 ) demonstrated that GBA, either pure recombinant protein or enzyme in fi broblast lysates, can generate GlcChol by transglucosylation of cholesterol when provided with GlcCer as donor . The physiological signifi cance of glucosylation of cholesterol is hypothetically great because it renders a molecule far more water soluble. To establish the natural occurrence of GlcChol in mammals, we fi rst needed to develop a quantitative and sensitive assay for quantifi cation of GlcChol in plasma, cells, and tissues. The method developed by us makes use of newly synthesized 13 C 6 -GlcChol as internal standard. Regarding extraction effi ciency, chromatographic behavior, and ionization characteristics, the natural and isotope-labeled compounds are identical, so no correction for the sample matrix (ion suppression or ion enhancement) is required.
With sensitive quantifi cation of GlcChol in place, we next observed that almost all tissues of mice show GlcChol. The relative high amount of the glucosylated sterol in the thymus, several nanomoles per gram wet weight, is striking and deserves special attention in view of noted abnormalities in NKT and B-cells in GBA-defi cient GD patients ( 44 ). It has been speculated by Mistry and colleagues that elevated GlcCer or GlcSph, via binding to CD1, may cause this phenomenon ( 44 ). It will now be of interest to study whether GlcChol interacts with CD1 because abnormalities in concentration of this lipid in GBA-defi cient GD patients are likely.
GlcChol is far more water soluble than cholesterol and, therefore, more suited for transport. The relatively low steady concentration of GlcChol does not exclude a vital role as intermediate in a transport pathway. Tentatively, water soluble GlcChol formed by transglucosylation at one cellular site would be transported and reconverted at the destination site back to more inert cholesterol. The combination of the two enzymes, GBA and GBA2, could provide such a mechanism without need for ATP. Of interest, in view of this speculation, is that LIMP2, the membrane protein interacting with GBA, has recently been shown to have a cholesterol binding site and potentially even a tunnel/channel function ( 45 ).
A fi nal consideration concerns the possible pathophysiological consequences of disturbed GlcChol metabolism. It seems likely that in individuals with abnormal GlcCer metabolism, as for example GBA-defi cient GD patients, secondary abnormalities in GlcChol occur. Future research will need to reveal whether such abnormalities in GlcChol or in other glucosylated metabolites contribute to particular symptoms associated with GD. In conclusion, GlcCer has earlier been identifi ed as an important structural membrane component and intermediate in GSL biosynthesis. In addition, it is known to act as an important sink for pro-apoptotic ceramide ( 46,47 ). Our study suggests catalytic pocket exposed to the cytoplasmic leafl et of membranes. It likely encounters different concentrations of GlcCer and cholesterol than GBA, given its location close to the cellular sites of de novo synthesis of these lipids. To maximally form GlcChol through transglucosylation, high concentrations of GlcCer as donor and high concentrations of cholesterol as acceptor are optimal. Vice versa, low concentrations of GlcCer and cholesterol will reduce net GlcChol formation. This consideration holds equally for GBA2 and GBA. Fluctuations in sterols and sphingolipids conceivably occur in cells, for example, after uptake of cholesterol-rich lipoproteins or upon release of ceramide from sphingomyelin. The ability to maintain some equilibrium between (glucosylated) sphingolipids and sterols by transglucosylating ␤ -glucosidases may have benefi cial buffering effects for cells. Of interest in this respect is our fi nding that inhibition of GCS leads to reduction of GlcChol in cultured cells, plasma of mice, and plasma of GD patients. This strongly suggests that the availability of GlcCer is an important driver of formation of GlcChol through transglucosylation. Because the ␤ -glucosidases, GBA and GBA2, also hydrolyze GlcCer, and thus tend to lower its concentration, the exquisite balance of various GlcCer-metabolizing enzymes and local cholesterol concentrations will determine GlcChol formation in subcellular compartments. Percentage reduction in plasma GlcChol of matched type 1 GD patients following eliglustat, miglustat, and ERT treatment compared with pretreatment values. Data were analyzed using the Mann-Whitney U test. ** P < 0.01, and *** P < 0.001.