Accumulation of Glucosylceramide in Murine Testis, Caused by Inhibition of β-Glucosidase 2

One of the hallmarks of male germ cell development is the formation of a specialized secretory organelle, the acrosome. This process can be pharmacologically disturbed in C57BL/6 mice, and thus infertility can be induced, by small molecular sugar-like compounds (alkylated imino sugars). Here the biochemical basis of this effect has been investigated. Our findings suggest that in vivo alkylated imino sugars primarily interact with the non-lysosomal glucosylceramidase. This enzyme cleaves glucosylceramide into glucose and ceramide, is sensitive to imino sugars in vitro, and has been characterized as β-glucosidase 2 (GBA2). Imino sugars raised the level of glucosylceramide in brain, spleen, and testis, in a dose-dependent fashion. In testis, multiple species of glucosylceramide were similarly elevated, those having long acyl chains (C16–24), as well as those with very long polyunsaturated acyl chains (C28–30:5). Both of these GlcCer species were also increased in the testes from GBA2-deficient mice. When considering that the very long polyunsaturated sphingolipids are restricted to germ cells, these results indicate that in the testis GBA2 is present in both somatic and germ cells. Furthermore, in all mouse strains tested imino sugar treatment caused a rise in testicular glucosylceramide, even in a number of strains, of which the males remain fertile after drug administration. Therefore, it appears that acrosome formation can be derailed by accumulation of glucosylceramide in an extralysosomal localization, and that the sensitivity of male germ cells to glucosylceramide is genetically determined.

position in the metabolic pathways of sphingolipids, connecting "simple" sphingolipids (including ceramide, sphingosine, and sphingomyelin (1)) with complex glycosylated sphingolipids (e.g. gangliosides GM3 and GM1), as illustrated in Fig. 1. Also, in contrast to complex GSLs, which have a luminal/extracellular orientation in cellular membranes, GlcCer is in part uniquely oriented in the cell, being located in the cytosolic leaflet of cellular membranes with its carbohydrate moiety extending into the cytosolic compartment (2)(3)(4). This is a consequence of the cellular topology of GSL biosynthesis. The synthesis of GlcCer, catalyzed by the ceramide-glucosyltransferase (GlcCer synthase, Fig. 1 and Table 1), takes place in the cytoplasmic leaflet of the cis-Golgi cisternae (5)(6)(7), or potentially in (a subcompartment of) the ER (8,9). From here GlcCer is translocated to the luminal side of the Golgi, where it is converted into more complex GSLs, starting with the addition of a galactose by lactosylceramide synthase (10) (Fig. 1). Alternatively, after synthesis GlcCer can be transported by nonvesicular means to the cytoplasmic leaflet of the plasma membrane (11).
It has long been known that GlcCer is degraded in lysosomes by glucocerebrosidase/GBA1 ( Fig. 1 and Table 1), but it is increasingly evident that GlcCer can also be hydrolyzed in an extralysosomal location, through the glucosylceramidase activity of ␤-glucosidase 2 (GBA2, Fig. 1 and Table 1). This enzyme was originally identified and characterized as the microsomal bile acid ␤-glucosidase, or ␤-glucosidase 2 (12)(13)(14). It was recently found that GBA2 also has glucosylceramidase activity, it cleaves the sphingolipid glucosylceramide into glucose and ceramide (15), as illustrated in Fig. 1. In addition, the enzymatic activity of GBA2 has been identified as the non-lysosomal glucosylceramidase (16), which was previously only operationally defined (17,18).
An interesting role has been proposed for the substrate of GBA2, GlcCer. This has emerged from the analysis of male (129S6/SvEv x C57BL/6) mice that are deficient for GBA2. The GBA2-deficient male mice have increased levels of glucosylceramide in their testes, liver, and brain (15). Remarkably, the male GBA2 Ϫ/Ϫ mice produce round-headed spermatozoa that have aberrant acrosomes. The fertility of the male GBA2-deficient mice is reduced, as assessed in natural mating tests (15). Also, in wild-type testis GBA2 was detected by immunostaining in Sertoli cells (15). The data on the subcellular localization of GBA2 suggest that it is present in the ER (15) and/or in the plasma membrane (16,17). Taken together, these studies suggest that the lack of non-lysosomal glucosylceramidase activity leads to an accumulation of GlcCer in Sertoli cells, which, in turn, has a deleterious impact on spermatid development, resulting in reduced fertility. It has thus been proposed that an elevation of glucosylceramide in Sertoli cells has a disrupting effect on spermatogenesis (15).
The bile acid ␤-glucosidase activity of GBA2 can be inhibited by N-alkylated imino sugars such as N-carboxynonyldeoxynojirimycin and N-dodecyldeoxynojirimycin, having IC 50 values of 20 and 2 nM, respectively (13). Also, the nonlysosomal glucosylceramidase activity is sensitive to N-alkylated imino sugars, including N-butyldeoxynojirimycin (NB-DNJ) and N-butyldeoxygalactonojirimycin (NB-DGJ) (16,17)(see also Fig. 1). In addition, the latter two imino sugars induce infertility in C57BL/6 mice, by perturbing spermiogenesis, leading to the production of abnormal spermatozoa that lack acrosomes (19). This prompted speculation that the impairment of sperm development caused by NB-DNJ and NB-DGJ may be caused by inhibition of the non-lysosomal glucosylceramidase/GBA2 (15, 16, 19 -22). This implies that, similar to the GBA2-deficient mice, imino sugar administration to wild-type mice should result in elevation of GlcCer levels in the testes and other tissues. In turn, this would further support the notion that non-lysosomal GlcCer accumulation in the testis leads to impairment of spermatogenesis.
However, we have recently reported that administration of NB-DNJ to male mice does not necessarily result in infertility. The reproductive effect of NB-DNJ in male inbred mice is straindependent (23). We have shown that the reproductive conse-quences of NB-DNJ are much milder in male mice of various inbred strains from the Swiss/Castle lineages, judging by the acrosomal and nuclear morphology of their spermatozoa. Also, after NB-DNJ administration, FVB/N, DBA/2, and 129S1/ SvImJ males were equally fertile as controls in natural mating tests (23). Therefore, we have here investigated the biochemical consequences of administration of alkylated imino sugars to male mice, in particular the effects on GlcCer and on enzymes involved in GlcCer metabolism. The data presented here are consistent with the notion that GBA2 is the primary target of alkylated imino sugars that induce infertility in male C57BL/6 mice. We have extended previous studies with detailed analyses of the lipid structures of the GlcCer species in imino sugartreated and in GBA2-deficient mice. Also, examination of GlcCer levels in the testes of NB-DNJ-treated males, from strains that remain fertile after drug administration, revealed that GlcCer accumulation in the testis does not necessarily lead to impairment of spermatogenesis.

EXPERIMENTAL PROCEDURES
Drug Administration-Male mice from various strains (BALB/c, C3H, C57BL/6, C57BL/10, DBA2, FVB/N, and NZW) were obtained from Harlan UK (Bicester, UK). Male mice from the 129S1/SvImJ, AKR/J, C57BR/J, and MA/MyJ strains were purchased from The Jackson Laboratory (Bar Harbor, ME). Animals were housed and sacrificed in accordance with the UK Home Office Animals (Scientific Procedures) Act 1986. NB-DNJ was a gift from Oxford GlycoSciences, Abingdon, Oxfordshire, UK/CellTech UK, Slough, Berkshire, UK. NB-DGJ was purchased from Toronto Research Chemicals (Downsview, ON, Canada). To administer imino sugars, 6-or 10-week-old mice were fed a diet of powdered mouse chow (expanded Rat and Mouse Chow 1, ground, Special Diet Services, Witham, Essex, UK) containing either NB-DNJ or NB-DGJ. The diet and compounds (both as dry solids) were mixed thoroughly, and stored at room temperature. Mice were administered an imino sugar for 5 weeks, unless stated otherwise.
Imino Sugar Determination-The serum concentration of NB-DNJ was measured by high-performance cation-exchange chromatography on a BioLC system (Dionex, Camberley, Surrey, UK) with pulsed amperometric detection according to Ref. 24.
Glycolipid Extraction and Purification-For isolation of GlcCer, tissues were homogenized in deionized water (Յ250 mg wet weight/ml) with a glass/Teflon homogenizer, lyophilized, and extracted four times with chloroform/methanol (2:1, v/v) over 24 h. Combined extracts were dried under a steam of nitrogen. Aliquots of extracted lipids normalized to dry weight were dissolved in 1 ml of chloroform and loaded on silicic acid columns (prepared with 3 ml of 10% (w/v) silicic acid (mesh 200 -400; Sigma) slurry in chloroform). Columns were washed with 2 ml of chloroform, and eluted with 6 ml of acetone. Eluates were dried under nitrogen. For isolation of neutral and anionic glycolipids other than GlcCer, tissues were homogenized as above and extracted in chloroform/methanol/water (4:8:3, v/v/v). Glycolipids were processed and purified by C18chromatography as described (25). . This reaction is catalyzed by GlcCer synthase (also referred to as the ceramide-specific glucosyltransferase and UDP-glucose:N-acylsphingosine D-glucosyltransferase). At extralysosomal locations GlcCer can be cleaved by ␤-glucosidase 2 (also referred to as bile acid ␤-glucosidase and the non-lysosomal glucosylceramidase), and in lysosomes by GBA1 (glucocerebrosidase). Both GlcCer synthase and GBA2 can be inhibited by N-alkylated imino sugars such as NB-DNJ and NB-DGJ. GlcCer can be converted to lactosylceramide (LacCer) and more complex glycosphingolipids, including gangliosides. Alternatively, via deglucosylation (in lysosomes and in extralysosomal locations), GlcCer can be metabolized into simple sphingolipids, like sphingomyelin, sphingosine, and further derivatives.
Thin Layer Chromatography-Aliquots of silicic acid-and C18-purified glycolipids corresponding to 3 mg dry weight (unless stated otherwise) were spotted on HPTLC plates (VWR, Lutterworth Leicestershire, UK), which were resolved as specified in the text. Neutral sphingolipids isolated from brain samples were loaded on borate-impregnated HPTLC plates that were developed in chloroform, methanol, 25% NH 4 OH, water (130:35:4:4, v/v/v/v). Glycolipids were detected with orcinol. Immediately after staining images were recorded with an Epson Perfection 4870 scanner. Glycolipid levels were quantitated with Image J 1.36b software.
For preparative thin layer chromatography aliquots of silicic acid-purified glycolipids corresponding to 6 mg dry weight were spotted on HPTLC plates, which were developed in chloroform/methanol/water (65:30:4, v/v/v). Areas of the plate corresponding to bands of interest were scraped. The silica was extracted twice with chloroform/methanol (2:1, v/v) and once with chloroform/methanol (1:1, v/v).
Glucocerebrosidase Digestion and Glucose Detection-Glycolipids purified by preparative TLC were dried under nitrogen, and incubated with 5 l of 36 units/ml glucocerebrosidase (Ceredase, Genzyme, West Malling, Kent, UK) in 50 l of 0.1 M phosphate/citrate buffer, pH 5.0, 0.4% Triton X-100, and 0.8% sodium taurodeoxycholate for 2 h at 37°C. Reaction mixtures were diluted with H 2 O to 200 l and loaded on 10-mg Oasis HLB extraction cartridges (Waters, Elstree, Hertfordshire, UK) equilibrated with methanol and water). Columns were washed with 1 ml of 5% methanol in H 2 O, and eluted with 1 ml of methanol and 1 ml of acetone. Eluates were dried under nitrogen gas, and loaded on a HPTLC plate.
To detect enzyme-released glucose, the material in the flowthrough and wash fractions of the HLB columns was concentrated by lyophilization. Dried samples were redissolved in 50 l of deionized water and duplicate aliquots (10 l) analyzed by a BioLC system (Dionex), equipped with a PA-10 column (4 ϫ 250 mm) and an AminoTrap guard column. Following isocratic elution with 18 mM NaOH at 1 ml/min at 30°C, glucose was detected electrochemically using an ED 50 detector using a triple potential discontinued waveform 1. The column was regenerated with 200 mM NaOH following each sample analysis. The peak area corresponding to glucose was analyzed with on-line Chromeleon chromatographic software (Dionex).
Non-lysosomal Glucosylceramidase-The inhibitory effects of DNJ derivatives on the activity of the non-lysosomal glucosylceramidase were determined in membrane preparations from various tissues using a method adapted from Refs. 17 and 18. Pooled mouse tissues were homogenized in water (1:3, w/v) using a Polytron PT1000 homogenizer (VWR) for 1 min. Homogenates were centrifuged for 20 min at 15,000 ϫ g (4°C). The supernatant was removed and the pellet was resuspended in ice-cold 50 mM potassium phosphate buffer, pH 5.8. Membranes were washed three times in the phosphate buffer, resuspended in this buffer, and stored at Ϫ80°C.
For the enzyme assay aliquots of the membrane suspension were preincubated with conduritol ␤-epoxide (CBE) (Toronto Research Chemicals, Downsview, ON, Canada) at a final concentration of 2.5 mM for 15 min, supplemented with an imino sugar to the desired concentration, incubated for 15 min, then diluted 3-fold in 4.5 mM 4-methylumbelliferyl-␤-D-glucoside (Sigma) in 0.1 M citric acid, 0.2 M disodium hydrogen phosphate, pH 5.8, in a final volume of 30 l, and incubated at 37°C for 1 h. The reaction was stopped by adding 200 l of 500 mM sodium carbonate, pH 10.7. Released 4-methylumbelliferone was detected using a Fluoroskan Ascent fluorometer (Thermo Electron Corp., Basingstoke, Hampshire, UK; excitation 355 nm, emission 460 nm).
Statistics-Quantitative data were statistically evaluated with Student's t test (comparison between two groups) or one-way analysis of variance with Tukey post hoc test (more than two groups), and phenotypic parameters were tested for correlation with Pearson test, using GRAPHPAD INSTAT version 3.0b for Macintosh (GraphPad Software). Values were considered statistically significantly different when p Ͻ 0.05.

Effects of High-and Low-dose NB-DNJ on Complex Glucosphingolipids in Testis-High-dose administration of NB-DNJ
(1200 -2400 mg/kg/day) lead to a 22-36% reduction in levels of complex GSLs in the testes of C57BL/6 mice (Fig. 2, A and B). In contrast, when treated with low doses of NB-DNJ, complex GSL levels were unchanged (15-150 mg/kg/day) or 15% increased (5 mg/kg/day) (Fig. 2, A and B). Six and nine months of drug administration at 15 mg/kg/day resulted in a similar increase in testicular complex GSLs (Fig. 2C).
Effects of High-and Low-dose Imino Sugars on GlcCer in Various Tissues-In the testes from NB-DNJ-treated C57BL/6 mice a number of neutral glycolipids were increased, present in three closely migrating bands (one major and two minor) on thin layer chromatograms developed with a solvent mixture optimized for small neutral GSLs (Fig. 3A). The testicular glycolipids comigrated with GlcCer standards (Fig. 3A). In addition, the compounds isolated by preparative TLC were fully digested with glucocerebrosidase, yielding free glucose (Fig.  3B). These data identify the glycolipid species that is affected by imino sugar treatment as glucosylceramide.
The glucosylceramide level in the testes from C57BL/6 mice treated with NB-DNJ was dose-dependent (Fig. 3A). From 1 to 15 mg/kg/day GlcCer levels were positively correlated with drug dose, reaching a maximum of 6 times control value at 15 mg/kg/day (Fig. 3C). Conversely, from 15 to 2400 mg/kg/day GlcCer was negatively correlated with drug dose, and, at the highest drug dose, GlcCer was at a similar level as in untreated mice (Fig. 3, A and C). NB-DGJ administration had a comparable effect on testicular GlcCer. The level of this glycolipid was also at first positively correlated with dose, reached a maximum, and was then negatively correlated with dose (Fig. 3D). However, after NB-DGJ treatment GlcCer was maximally elevated at a higher dose, 300 mg/kg/day, compared with NB-DNJ. In addition, at 150 -600 mg/kg/day of NB-DGJ, GlcCer levels were higher than in mice treated with 5-150 mg/kg/day NB-DNJ (Fig. 3D). At 15 and 150 mg/kg/day the GlcCer levels differed significantly between the two imino sugars.
Low-dose NB-DNJ also increased GlcCer significantly in spleen, but not in liver (Fig. 3D). In addition, GlcCer could be detected in the brains of mice treated with 5 mg/kg/day NB-DNJ (Fig. 3E). Brain GlcCer increased with dose to reach a maximum at 1200 mg/kg/day (Fig. 3, D and E). However, the increase in brain GlcCer relative to control samples could not be calculated, because the GlcCer levels in brains of untreated mice were below the level of detection, even after loading 6 times more material from control brains (Fig. 3E).
NB-DNJ and Testicular GlcCer: Kinetics and Reversibility-To investigate the kinetics of the imino sugar-induced rise in GlcCer in testis, and whether this effect was reversible, C57BL/6 mice were treated with NB-DNJ for increasing lengths of time, to a maximum of 35 days, and then allowed to recover from the drug for up to 25 days. The increase in GlcCer was most rapid over the first 5 days of NB-DNJ administration, and modest during the following 6 days (Fig. 4, A and B). From day 11 to 35 of drug treatment GlcCer did not change significantly. GlcCer was also monitored following withdrawal of NB-DNJ from mice that had been treated for 35 days. GlcCer levels rapidly decreased during the first 5 days off drug, and were similar to control values 10 days after drug withdrawal (Fig. 4, A and B).
GlcCer was also measured in the testes of mice that had been treated with 15 mg/kg/day NB-DNJ for longer periods of time, up to 12 months. From 3 to 12 months of drug administration GlcCer was 5-7-fold increased, comparable with the rise in GlcCer after short-term treatment (Fig. 4C). Testicular GlcCer did not change significantly from 3 to 12 months of NB-DNJ administration. When the drug was withdrawn from long termtreated animals for 9 weeks, their testicular GlcCer levels were similar as in age-matched control animals. Taken together, NB-DNJ administration resulted in a higher steady-state level of testicular GlcCer, which was reached within 11 days, and was maintained at the same level for the duration of drug treatment. The increase in GlcCer was reversible both after short and longterm NB-DNJ regimens.
Further Analysis of Testicular GlcCer-The fact that testicular GlcCer resolved in three closely migrating bands on TLC (Figs. 3A and 4A) suggested that multiple species of GlcCer are present in this tissue, differing in their ceramide moieties. Two major types of N-acyl chains have been identified in testicular sphingolipids, differing in length and level of saturation. The acyl chains are either long (C16 -24) and saturated (or monounsaturated), or very long (C26 -32) and polyunsaturated (4 -6 double bonds) (30,31). Interestingly, we have recently found that the very long-chain polyunsaturated (VLC-PUFA) glycosphingolipids are restricted to germ cells. 4 Therefore, detailed analysis of the various testicular GlcCer species can indicate which of the testicular cell types build up GlcCer after NB-DNJ treatment. The same approach can be followed to find out which testicular cell types accumulate GlcCer in the testes from GBA2-deficient mice.
Using mass spectrometry, we found that, after imino sugar administration to C57BL/6 mice, the two classes of testicular GlcCer (having long and saturated or very long and polyunsaturated acyl chains) were equally increased ( Fig. 5 and supplementary Fig. S1). In testes that are devoid of GBA2, long-chain GlcCers were also increased to a similar extent as the very longchain GlcCers (Fig. 5 and supplementary Fig. S1). Thus, within each of these experimental models, the two types of GlcCer were comparably elevated. When contrasting drug-treated mice with GBA2 knock-out mice, GlcCers were elevated to a significantly higher degree in the knock-out mice (Fig. 5). These data suggest that, after NB-DNJ administration and in GBA2-deficient mice, GlcCer was elevated in somatic as well as germ cells, and that in each of these cases GlcCer increased to a similar extent in somatic and germ cells. Mass spectrometric analysis also showed that NB-DNJ did not affect testicular ceramide or sphingomyelin levels in C57BL/6 mice (supplementary Fig. S2).
Imino Sugar Inhibition of GlcCer Metabolizing Enzymes: Kinetics-We determined the sensitivities of two testicular enzymes to inhibition by NB-DNJ and NB-DGJ, the GlcCer synthase and the non-lysosomal glucosylceramidase. The testicular glucosylceramidase was significantly more responsive to NB-DNJ than GlcCer synthase, having IC 50 values of 0.14 and 22.9 M, respectively (Fig. 6A). Furthermore, compared with NB-DNJ, NB-DGJ was a less effective inhibitor of the testicular glucosylceramidase (IC 50 5.3 M, Fig. 6B). The glucosylceramidase activities in liver and brain were equally sensitive to inhibition by NB-DNJ as the testicular glucosylceramidase activity (Fig. 6C).
NB-DNJ Concentration in Serum-To correlate these in vitro data on drug sensitivities of GlcCer metabolizing enzymes with the levels of GlcCer found in testes from NB-DNJ-treated mice, we determined the serum levels of the drug in mice treated with 150 -2400 mg/kg/day. At 15 mg/kg/day the serum level was 0.49 M (23), higher than the IC 50 of NB-DNJ for the glucosylceramidase, but much lower than its IC 50 for the GlcCer synthase (Fig. 3D). At higher drug doses the serum level of NB-DNJ increasingly approached the IC 50 of the GlcCer synthase (Fig. 3D).
Effect of NB-DNJ on GlcCer in Testes from Mice of Various Inbred Strains-Having seen that NB-DNJ does not have a severe impact on spermatogenesis in various mouse strains, we reasoned that this might be due to the drug failing to elevate testicular GlcCer in these mice. Alternatively, the possibility could be considered that in mouse strains, which do not become infertile from NB-DNJ, the drug causes GlcCer to increase, but without this being of consequence for spermatogenesis. Therefore, we administered NB-DNJ to males of 11 strains of mice that can be divided into three groups on the basis of the impact of NB-DNJ on the morphology of their spermatozoa, having a high, medium, or low percentage of grossly abnormal acrosome-less spermatozoa (23). In strains of mice 4 R. Sandhoff and A. C. van der Spoel, submitted for publication. that have an intermediate or low sensitivity to the reproductive effects of NB-DNJ, the levels of GlcCer (measured after drug treatment) were comparable with those detected in highly sensitive strains (which include C57BL/6) (Fig. 7, A and B). Also, similar to C57BL/6 mice, the two distinct types of testicular GlcCer were equally increased in DBA2 mice, a drug-insensitive strain (Fig. 5C). The overall increase in long-chain GlcCer was significantly higher in the latter strain than in C57BL/6, when measured by mass spectroscopy (Fig. 5C).
We have also recently reported that the consequences of NB-DNJ administration are highly variable in C57BL/6 x FVB/N interstrain hybrid mice, judging by the percentage of grossly abnormal acrosome-less spermatozoa (23). We measured testicular GlcCer in eight NB-DNJ-treated interstrain hybrid mice that together span the range of responses in these mice (0 -100% aberrant spermatozoa (23)). Among these hybrid mice GlcCer was similarly increased (Fig. 7B), and the level of testicular GlcCer was not significantly correlated with the proportion of misshapen spermatozoa.

DISCUSSION
Imino Sugar Dose and Testicular GlcCer-We have found a paradoxical relationship between oral dose of NB-DNJ and NB-DGJ and the testicular level of GlcCer. We treated mice with various doses of these drugs, covering wide dose ranges. Two types of correlation between oral dose and GlcCer were seen. At first GlcCer was positively correlated with dose, whereas at higher doses this correlation was negative. Thus, for each of the two drugs, GlcCer reached a maximum level at a dose that was considerably lower than the highest dose used. To understand the imino sugar-GlcCer relationship we have examined some of the enzymes that are involved in the metabolism of GlcCer (illustrated in Fig. 1, see also Table 1).
We have shown that the non-lysosomal glucosylceramidase activity in mouse testis and brain can be inhibited by NB-DNJ and NB-DGJ. The sensitivity of this enzyme to inhibition by these imino sugars is higher than that of the GlcCer synthase. The difference in imino sugar sensitivity between these enzymes, determined in in vitro assays, most likely underlies the biphasic relationship observed in vivo between oral drug  3). C, relative testicular GlcCer levels after 3, 6, 9, and 12 months of NB-DNJ administration, and after 9 weeks of recovery (R) from long-term drug treatment (n ϭ 4). Asterisk, significantly different from control value (p Ͻ 0.05). Data are indicated as mean Ϯ S.D.
dose and abundance of GlcCer in the testis. Our value of the IC 50 of miglustat toward the testicular glucosylceramidase (0.14 M) was slightly lower than that found previously in human spleen (0.31 M (17)). We can assume that low NB-DNJ doses (up to 15 mg/kg/day, serum level Ϯ 0.5 M (23)) primarily affect the testicular glucosylceramidase, increasing the pool of its substrate. At higher drug doses (serum levels 1.7-21.5 M) the activities of both the glucosylceramidase and the GlcCer synthase (IC 50-NB-DNJ ϭ 22.9 M) will be reduced, precluding a (substantial) elevation of whole testis GlcCer levels. The latter IC 50 value was close to that reported earlier for a human myeloid cell line (29). Similarly, after administration of NB-DGJ, the maximal level of GlcCer was reached at 300 mg/kg/day. This is in agreement with the IC 50 values of NB-DGJ for the glucosylceramidase and the GlcCer synthase being 5.3 (our data) and 41.4 M (29), respectively. These values are higher than the corresponding values of NB-DNJ. Furthermore, as the low-dose imino sugar-treated mice have a net increase in testicular GlcCer, they resemble genetically modified mice that lack all non-lysosomal glucosylceramidase (GBA2) activity. The GBA2-deficient mice accumulate GlcCer in various tissues, including testis, liver, and brain (15).
Testicular GlcCer and Spermatogenesis-The increase in testicular GlcCer, caused by low-dose NB-DNJ or NB-DGJ administration, is most likely the first step in the cascade, through which these compounds cause C57BL/6 mice to produce aberrant spermatozoa without acrosomes. This notion is supported by the study of Yildiz et al. (15) who report that the GBA2deficient mice, which are on a 129S6/SvEv x C57BL/6 background, produce abnormal spermatozoa and are subfertile. The male GBA2 knock-out mouse can thus also in this respect be regarded as a phenocopy of the mice treated with low-dose alkylated imino sugars, as has previously been speculated (20).
In the imino sugar-treated mice the relationship between whole testis GlcCer and the quality of spermatogenesis is confounded by the dual impact of these drugs, targeting both the biosynthesis and extralysosomal degradation of GlcCer. Similar levels of GlcCer were achieved both with low drug doses (Ͻ15 mg/kg/day for NB-DNJ and Ͻ300 mg/kg/day for NB-DGJ), and with higher doses, whereas C57BL/6 mice only become infertile at doses of at least 15 mg/kg/day miglustat (19) or 150 mg/kg/ day NB-DGJ (21). Nevertheless, for both drugs the minimal dose required to induce infertility coincided with the dose that maximally elevated GlcCer. We can speculate that the higher amounts of GlcCer that hamper spermatogenesis are limited to a particular (sub)cellular site. It is not unlikely that at high imino sugar doses the localized elevation of GlcCer is masked by lower overall levels of GlcCer (due to reduced activity of GlcCer synthase). In this model, the localized increase of GlcCer can be thought to be sufficient to impair spermatogenesis in C57BL/6 mice, irrespective of whole testis GlcCer. Alternatively, the reduction seen in complex testicular GSLs at 1200 -2400 mg/kg/day NB-DNJ may have a negative impact on FIGURE 6. Inhibition of non-lysosomal (CBE-insensitive) glucosylceramidase activity by imino sugars. A, titration of CBE-insensitive glucosylceramidase (OE) and GlcCer synthase (f) activities with NB-DNJ. The CBE-insensitive glucosylceramidase was more sensitive to NB-DNJ than GlcCer synthase. B, titration of testicular glucosylceramidase activity with NB-DNJ (OE) and NB-DGJ (E). NB-DNJ was a more potent inhibitor of NLG activity than NB-DGJ. C, comparison of inhibition of glucosylceramidase activity by NB-DNJ in membranes prepared from brain (ࡗ), testis (ϫ), and liver (E). The CBE-insensitive glucosylceramidase activities in testis, brain, and liver membranes were 0.80, 4.82, and 0.23 pmol/mg/min, which corresponded to 6.4, 24.1, and 1.7% of total ␤-glucosidase activities, respectively. spermatogenesis, as a complete lack of complex GSLs also results in male sterility (32). Clearly the reproductive phenotype caused by reduction of complex GSLs is less severe than in the complex GSLs-negative mice, in which post-meiotic germ cell development is severely limited (32).
Cellular Localization of GBA2 and GlcCer-We have found that multiple structurally distinct forms of GlcCer accumulate in mouse testes after NB-DNJ administration, long-chain GlcCer (primarily C16:0) and very long-chain polyenoic GlcCer (primarily C30:5). Also, in C57BL/6 and DBA2 mice, these two forms of GlcCer were increased to a similar extent upon drug treatment. Considering that very long-chain polyenoic sphingolipids are associated with germ cells, 4 this indicates that NB-DNJ treatment causes an increase in GlcCer in both somatic and germ cells. In turn, taking into account that the primary target of NB-DNJ is GBA2, our mass spectrometric data most likely reflect that GBA2 also is present in germ cells. This is in agreement with expression studies, where Gba2 mRNA has been detected in spermatogonia, spermatocytes, and spermatids at similar levels as in Sertoli cells (33) (www.germonline. org). Therefore, our results extend the range of testicular cell types where Gba2 is expressed, as Yildiz et al. (15) found by immunohistochemistry that the enzyme is present in Sertoli cells.
Subcellular Localization of GBA2 and GlcCer-Clearly knowledge about the subcellular location of GBA2 will greatly add to the understanding of the process by which inhibition/ deficiency of this enzyme impairs spermiogenesis. However, current data on this issue are contradictory. After subcellular fractionation, the glucosylceramidase activity has been detected by Overkleeft et al. (17) in fractions containing cell surface markers. Similarly, Boot et al. (16) found GBA2-GFP and GFP-GBA2 fusion proteins to be present at the cells surface in transfected COS cells. However, according to Yildiz et al. (15) an alternative localization of GBA2 is the ER. Using immunostaining the enzyme has a distribution typical for the ER in COS cells that were transfected with the GBA2 cDNA (15).
Genetic Background-The genetic background of the GBA2deficient mice is in part C57BL/6 (129S6/SvEv x C57BL/6). These knock-out mice thus share many elements of their genome with C57BL/6 mice. Therefore impairment of spermatogenesis has been observed in mice that have, on the one hand, at least in part a C57BL/6 background, and, on the other hand, diminished GBA2 activity (through disruption of the GBA2 gene or via administration of imino sugars). We have found that NB-DNJ-treated inbred mice with genetic backgrounds unrelated to C57BL/6 have levels of testicular GlcCer as high as in C57BL/6 mice, but without a reproductive phenotype (23). We have also seen that NB-DNJ elevates GlcCer in FVB/N x C57BL/6 hybrid mice, whereas only in a minority of these hybrids is spermatogenesis affected (23). Thus the consequence of GBA2 deficiency and GBA2 inhibition through lowdose imino sugar treatment seems universal (higher level GlcCer), but the impact of this alteration on male germ cell development is dependent on genetic background. Examined more closely, the phenotype of the GBA2-deficient 129S6/ SvEv x C57BL/6 mice seems most