Novel class of glycosphingolipids involved in male fertility.

Mice require testicular glycosphingolipids (GSLs) for proper spermatogenesis. Mutant mice strains deficient in specific genes encoding biosynthetic enzymes of the GSL pathway including Galgt1 (encoding GM2 synthase) and Siat9 (encoding GM3 synthase) have been established lacking various overlapping subsets of GSLs. Although male Galgt1-/- mice are infertile, male Siat9-/- mice are fertile. Interestingly, GSLs thought to be essential for male spermatogenesis are not synthesized in either of these mice strains. Hence, these GSLs cannot account for the different phenotypes. A novel class of GSLs was observed composed of eight fucosylated molecules present in fertile but not in infertile mutant mice. These GSLs contain polyunsaturated very long chain fatty acid residues in their ceramide moieties. GSLs of this class are expressed differentially in testicular germ cells. More importantly, the neutral subset of this new GSL class strictly correlates with male fertility. These data implicate polyunsaturated, fucosylated GSLs as essential for spermatogenesis and male mouse fertility.

Mice require testicular glycosphingolipids (GSLs) for proper spermatogenesis. Mutant mice strains deficient in specific genes encoding biosynthetic enzymes of the GSL pathway including Galgt1 (encoding GM2 synthase) and Siat9 (encoding GM3 synthase) have been established lacking various overlapping subsets of GSLs. Although male Galgt1؊/؊ mice are infertile, male Siat9؊/؊ mice are fertile. Interestingly, GSLs thought to be essential for male spermatogenesis are not synthesized in either of these mice strains. Hence, these GSLs cannot account for the different phenotypes. A novel class of GSLs was observed composed of eight fucosylated molecules present in fertile but not in infertile mutant mice. These GSLs contain polyunsaturated very long chain fatty acid residues in their ceramide moieties. GSLs of this class are expressed differentially in testicular germ cells. More importantly, the neutral subset of this new GSL class strictly correlates with male fertility. These data implicate polyunsaturated, fucosylated GSLs as essential for spermatogenesis and male mouse fertility.
Mammalian spermatogenesis, the transformation of diploid spermatogonial stem cells into haploid spermatozoa, is a complex process starting in the basal compartment of the seminiferous tubules where spermatogonial stem cells undergo asymmetric division, proliferation, and differentiation into preleptotene and leptotene spermatocytes. Next, the spermatocytes must traverse the blood-testis barrier, built up by junctions between Sertoli cells, and migrate into the adluminal compartment to complete meiosis and to develop into haploid round spermatids.
All except the earliest spermatogonial division result in a growing "string of beads" composed of interconnected germ cells with a continuous cytosol and plasma membrane caused by incomplete cytokinesis. During the final step of spermato-genesis leading to the release of spermatozoa (spermiation), spermatids undergo cytological transformations including shape change, nuclear condensation, and the development of specialized structures, such as an acrosome, and a flagellum with a mitochondrial sheath. The continuous spermiation is maintained by junctions between Sertoli cells and germ cells.
Glycosphingolipids (GSLs) 1 are amphipathic molecules present on cell membranes of the endocytotic and exocytotic pathway in mammalian cells where they influence membrane function. They also appear to play a role in cytological germ cell transformations because male mice lacking complex gangliosides, i.e. sialic acid-containing GSLs, resulting from Galgt1 deficiency (also known as Galgt1 encoding Gg3Cer/GM2/GD2synthase (EC 2.4.1.92) are infertile ( Fig. 1) (1,2). In these mice, spermatogenesis proceeds only until the stage of round spermatids, and then the string of beads composed of spermatids is thought to lose Sertoli cell contact and round up to multinuclear giant cells (1). The significant drop of serum testosterone observed in these mice was explained by a lack of complex gangliosides such as GT1b, which bind to and thereby possibly transport testosterone from Leydig cells to the seminiferous tubules (1). In a later study, infertility of male mice was induced by treatment with an inhibitor of GSL synthesis. The treated mice showed normal serum testosterone but abnormal spermatozoa in their testes (3). Recently a Siat9Ϫ/Ϫ mouse (encoding GM3-synthase; EC 2.4.99.-) was generated, which was fertile (4), although it lacks a-and b-series complex gangliosides (Fig. 1), a deficiency thought to cause infertility in male Galgt1Ϫ/Ϫ mice.
We searched for gangliosides and other GSLs present in fertile Siat9Ϫ/Ϫ and wild type testes but missing in infertile Galgt1-deficient mice. In wild type testes we uncovered a far more complex structural GSL pattern than described thus far (1,(5)(6)(7). A set of novel neutral fucosylated GSLs (FGSLs) was discovered whose presence strictly correlated with fertility. Gangliosides, structurally related to the neutral FGSLs by the addition of one sialic acid, were also found. The FGSLs are the major glycolipids of wild type mouse testes. They belong to the ganglio series; the four neutral FGSLs were determined to be IV 2 -␣-Fuc-Gg 4 Cer; IV 3 -␣-Gal,IV 2 -␣-Fuc-Gg 4 Cer; IV 3 -␣-Gal-NAc,IV 2 -␣-Fuc-Gg 4 Cer; and IV 3 -␣-GalNAc␤3Gal,IV 2 -␣-Fuc-Gg 4 Cer. Interestingly the ceramide moieties of the novel FGSLs are highly unsaturated and contain polyenoic (four to six double bonds) very long chain fatty acid residues with at least 26 carbon atoms and mostly 2-hydroxylation.
Purification of Compounds A and D from Testes Tissue-Neutral and acidic GSLs from 26-g testes, wet weight, from wild type mice were extracted according to Sandhoff et al. (9). From the flow-through of the DEAE A-25 column, the neutral GSLs A and B were purified further by repeated silica gel flash column chromatography with the appropriate mixtures of n-hexane/isopropyl alcohol/water or CHCl 3 /CH 3 OH/H 2 O (v/v) as running solvents. The enriched compounds A and D were peracetylated, then separated on quantitative high performance TLC plates with chloroform/acetone (1/1, v/v), scraped off the plates, eluted with methanol, deacetylated with methanol and 25% aqueous NH 4 OH (1/1, v/v) for 3 h at 50°C, and dried with a N 2 stream.
GSL Extraction for TLC and ESI-MS/MS Analysis-GSLs were extracted from testes according to Sandhoff et al. (9). Detection of neutral GSLs by ESI-MS/MS was greatly improved by its peracetylation, separation on Florisil column, and subsequent deacetylation. Therefore the method of Saito and Hakomori (10) was modified as described previously (11).
Detection of acidic GSLs was greatly improved by extracting the acidic GSL fraction further by the method of Folch et al. (12). No changes resulting from Folch extraction were detected in the fatty acid patterns of gangliosides.
NanoESI-MS/MS-Analysis was performed with a triple quadrupole instrument (VG micromass (Cheshire, UK) model Quattro II) equipped with a nanoelectrospray source and gold-sputtered capillaries as described previously (9). Parameters for cone voltage and collision energy of the different scan modes are listed in Table I.
Carbohydrate Constituent Analysis-Carbohydrate constituents were released by acid hydrolysis, converted into their corresponding alditol acetates, and analyzed by capillary GC/MS as detailed elsewhere (15).
Carbohydrate Permethylation Analysis-For determination of linkage positions of monosaccharide constituents, glycolipids were permethylated and hydrolyzed (16). Partially methylated alditol acetates obtained after sodium borohydride reduction and peracetylation were analyzed by GC/MS using the instrumentation and microtechniques described previously (17).
Immuno-overlay Technique-Detection of GSLs with anti-blood group A and B antibodies (Seraclone anti-A clone A003 and anti-B clone B005, Biotest AG Dreieich, Germany) was performed using the immuno-overlay technique (18).
Sphingolipid Ceramide N-Deacylase (SCDase) Treatment-Lyso compounds A and C were obtained by treatment of the purified mixture of compounds A and C with SCDase (Takara Shuso, Otsu, Shiga, Japan) according to Kurita et al. (19).
Perhydrogenation of GSLs-Dry neutral GSLs corresponding to 100 mg of wild type testes, wet weight, were dissolved under N 2 in 500 l of methanol and 250 l of tetrahydrofuran. Pd/C was added, and N 2 was exchanged by H 2 . After 24 h of stirring at room temperature the mixture was filtered and dried.
Isolation of Detergent-insoluble Complexes-"Rafts," i.e. detergentinsoluble complexes, were isolated as described by Ottico et al. (20) with minor adjustments. Testes from two mice (390 mg, wet weight) were combined, Dounce homogenized in lysis buffer containing 1% Triton X-100 (10 strokes, tight), and incubated at 4°C for 1 h. The lysate was centrifuged at 1,300 ϫ g for 5 min, and the supernatant was mixed 1:1 with 80% sucrose in TNE buffer (20 mM Tris-HCl, 150 mM NaCl, 2 mM EDTA, pH 7.4). 30% and 5% sucrose solutions were loaded on top. After ultracentrifugation (100,000 ϫ g, 20 h, 4°C) 12 fractions of 1 ml were collected from the top, fraction 3 marking the border between 5 and 30% sucrose, which was identified as the raft-containing fraction. For lipid determination the fractions were dialyzed and lipids extracted as described, proteins were isolated using trichloroacetic acid precipitation, separated by SDS-PAGE, and blotted on polyvinylidene difluoride membranes (Millipore). Caveolin was detected with a purified polyclonal rabbit anti-mouse caveolin antibody (BD Biosciences) followed by a horseradish peroxidase-labeled swine anti-rabbit immunoglobulin antibody (DakoCytomation, Glostrup, Denmark) and ECL detection (Amersham Biosciences).
Immunohistochemistry-Immunostaining of 5-m cryosections was carried out with the alkaline phosphatase anti-alkaline phosphatase method (22) and with the immunofluorescence technique (23). A confocal laser microscope (Leica TCS SL) was used for documentation of the immunofluorescence results. Immunostaining of 0.6-m ultracryo semithin sections (Leica Ultracut UCT equipped with Leica EMFCS) was performed using a fluorescent microscope (Zeiss Axiovert 35). Testes were fixed in 2% paraformaldehyde by cardiac perfusion, cryoprotected with 2.3 M sucrose, and frozen in liquid N 2 .
Wild type mouse testes contained complex neutral GSLs and complex gangliosides. The quantitatively major neutral GSL components migrated on TLC as two double bands with a slower migration velocity than Gg 4 Cer. The major gangliosides had TLC migration rates between GM1a and GD1a. Faint ganglioside bands migrated similarly to GM3, GM2, GM1a, and GT1b on the plate, and one band moved even slower than GT1b.
The ganglioside profiles of the mutant testes differed from those of wild type mice. Minor changes were observed in Siat8aϪ/Ϫ mice (lacking GD1b, GT1b, and GQ1b), whereas the ganglioside profiles of infertile Galgt1Ϫ/Ϫ and fertile Siat9Ϫ/Ϫ mice testes differed drastically from those of wild type mice. They lacked all complex wild type gangliosides as well as GM3 and GM2 in Siat9Ϫ/Ϫ mice. Both GM3 and GD3 accumulated in testes of Galgt1Ϫ/Ϫ mice, whereas Siat9Ϫ/Ϫ mice produced mainly the 0-series gangliosides GM1b and GD1a (Fig. 3B). The presence of GD1a was also confirmed by specific fragment m/z 475 as described by Yamashita et al. (4). In contrast to the multiple changes of the ganglioside patterns observed in fertile and infertile mutant mice, only the neutral, testicular GSL pattern of infertile Galgt1Ϫ/Ϫ mice differed from that of wild type testes: two wild type double bands migrating in TLC below Gg 4 Cer were missing (Fig. 3A). They represent the major complex GSLs (later described as FGSLs) of wild type testes and are present in the fertile mutant mice.
Structural Analysis of Oligosaccharide Moieties of Testicular Complex Neutral GSLs-Because the presence of the complex neutral GSL components (FGSLs) of the mouse testes correlates with fertility, their chemical structures were established. Tandem mass spectrometry revealed the presence of four different complex neutral GSL compounds, each represented mainly by two peaks with a difference of m/z 28, corresponding to a C 2 H 4 unit (Fig. 4A 1 ). The smallest compound (compound A) had protonated molecular masses of 1,548 and 1,576. Compound B followed in a distance of m/z 162 corresponding to an additional Hex residue. Compound C was 203 atomic mass units bigger than compound A corresponding to one HexNAc   (Fig. 5A, lane 2). Treatment of this product with ␤-galactosidase yielded a GSL migrating as Gg 3 Cer (Fig. 5A, lane 5). The sequential loss of one Fuc and one Gal residue was confirmed by mass spectrometry (data not shown). The combination of carbohydrate constituent and glycosidic linkage analysis revealed compound A to contain one 4-substituted Glc, one 4-substituted Gal, and one 2-substituted Gal, a 3-substituted GalNAc, and one terminal Fuc residue (data not shown). Combining these results with its ganglio series origin, compound A was characterized as Fuc␣2Gal␤3-GalNAc␤4Gal␤4GlcCer (IV 2 -␣-Fuc-Gg4Cer or Fuc-GA1) containing surprisingly large ceramide moieties with protonated molecular masses of 712 and 740 atomic mass units. This was supported by data from partially hydrolyzed compound A (data not shown).
The combination of carbohydrate constituent and glycosidic linkage analysis revealed compound D to contain one 4-substituted Glc, one 4-substituted Gal, one 3-substituted Gal, and one 2,3-disubstituted Gal, a 3-substituted GalNAc, a terminal GalNAc, and one terminal Fuc residue (data not shown). Compound D could be degraded to a GSL with migration identical to compound A by treating it first with Hex B and subsequently with ␣-galactosidase (Fig. 5B, lanes 1 and 3). The corresponding sequential loss of one GalNAc and one Gal residue was confirmed by mass spectrometry (data not shown). With a ganglio series backbone, these data fit the following structure for compound D: GalNAc␤3Gal␣3(Fuc␣2)Gal␤3GalNAc␤4Gal␤-4GlcCer (IV 3 -␣-GalNAc␤3Gal-compound A). Compound D contained ceramide moieties identical to that of compound A.
Compound B was solely susceptible to ␣-galactosidase, converting it to a GSL migrating like compound A (data not shown). Because compound A is the biosynthetic precursor of compound B, compound B should end with the terminal Gal␣3(Fuc␣2)Gal structure (also known as blood group B determinant). However, compound C resisted the combined treatment of ␣-, ␤-galactosidase, Hex B, and ␣-fucosidase (data not shown), suggesting a terminal ␣-glycosidically linked HexNAc residue as is the case for the blood group A epitope, GalNAc␣3(Fuc␣2)Gal. Indeed compounds B and C could be stained with blood group B-and blood group A-specific antibodies, respectively (Fig. 5C). These blood group antibodies could not detect compounds B and C antigens in ganglio series-deficient Galgt1Ϫ/Ϫ testes, thus proving their ganglio series backbone (Fig. 5D). These results suggest that compound B with the structure IV 3 -␣-Gal,IV 2 -␣-Fuc-Gg 4 Cer is the precursor of compound D, and compound C is IV 3 -␣-GalNAc,IV 2 -␣-Fuc-Gg 4 Cer. The large ceramide moieties of compounds B and C are identical to those of compounds A and D as discussed below.
Each blood group antibody stained a different, weaker TLC band from Galgt1Ϫ/Ϫ testes migrating slightly faster than the corresponding wild type lipids compounds B and C (Fig. 5D). These TLC bands may correspond to two of the three TLC bands still visible in Galgt1Ϫ/Ϫ mice (Figs. 3 and 5D).
The Ceramide Residues of Compounds A to D Contain Polyenoic, Very Long Chain Fatty Acids (VLCFAs)-Incubation of compounds A and C with SCDase resulted in the appearance of a single lyso compound for each of these glycolipids, i.e. lyso compound A, 1,136 atomic mass units and lyso compound C, 1339 atomic mass units. According to the molecular mass the two signals corresponded to d18:1 sphingosine-containing lyso compound A and lyso compound C, respectively (Fig. 4B). The SCDase-induced loss of either 412 or 440 atomic mass units appeared to correspond to a release of fatty acids of 430 or 458 atomic mass units (412 or 440 atomic mass units ϩ 18 atomic mass units from the addition of H 2 O). Because calculations suggested the presence of polyenoic fatty acids, a perhydrogenation experiment was performed. It resulted in a mass shift of 12 atomic mass units for the main two peaks of all four compounds A, B, C, and D, corresponding to six carbon to carbon double bonds (Fig. 4A 2 ). Compound A, with a ceramide-containing hydroxylated palmitic acid, showed a mass increase by perhydrogenation of 2 atomic mass units (m/z 1,389 to 1,391) because of hydrogenation of the sphingosine double bond (Fig.  4, A 2 and C 2 ). Taking this into account, the main fatty acids of compounds A, B, C, and D with 430 and 458 atomic mass units each contained five double bonds. In addition, their molecular mass suggested the presence of one hydroxyl group. The presence of hydroxylated polyenoic VLCFAs with compositions of h28:5 and h30:5 as major constituents was confirmed by GC/MS analysis, as well as the presence of a hydroxyl group in the 2-position by identification of the specific fragments [M-59] ϩ (loss of the carbomethoxy radical CH 3 OCϭO ⅐ ) (27) and at m/z 90, resulting from the McLafferty rearrangement and 2,3-bond cleavage using the electron impact ionization mode (data not shown) (28).
GM3 and GD3 of Galgt1-deficient Mouse Testes Do Not Contain Significant Amounts of Polyenoic VLCFAs-As revealed by MS/MS, GD3 showed a fatty acid composition analogous to the wild type GD1, i.e. without polyenoic VLCFAs. The fatty acid pattern of GM3 was similar to that of GD3. However, this ganglioside also contained very small amounts of h28:5 and h30:5 fatty acid residues (data not shown).

Siat9Ϫ/Ϫ Mice Testes Lack the Monosialocompounds A-D-
The ceramide compositions of GM1b, GD1a, and GT1 from testes of Siat9Ϫ/Ϫ mice corresponded to that of wild type GM1a, GD1a, or GT1b, respectively. The sialo derivatives of compounds A-D could not be detected. This was confirmed with immuno-overlays: Fuc-GM1 (monosialocompound A) detected with monoclonal antibody F12 in wild type disappeared in Siat9Ϫ/Ϫ testes. The blood group A (sialocompound C) and B (sialocompound B) positive compounds of the wild type testes were absent in Siat9Ϫ/Ϫ as well as in Galgt1Ϫ/Ϫ mice. No additional blood group A-or B-reactive bands were found in either of these mutant mice (data not shown).
Testes Polyenoic FGSLs Are Detergent-soluble at 4°C-GSLs are believed to be members of detergent-insoluble signaling . C, a sample containing compounds A-D was separated as one broad band on a TLC plate and cut into three parts. The left was stained with orcinol/sulfuric acid, the middle developed with an anti-blood group A, and the right part with anti-blood group B antibody. D, comparison of wild type and Galgt1Ϫ/Ϫ testes. Neutral GSLs from wild type and Galgt1Ϫ/Ϫ testes corresponding to 10 mg of tissue, wet weight, were separated on TLC. They were stained with either orcinol/sulfuric acid or the anti-blood group A or B antibodies; neutral GSL standard is as in B.

FIG. 5. Characterization of compounds A and D by exoglycosidase-digest (A and B) and compounds B and C (Cp.B and Cp.C) by immuno-overlay (C and D). A, TLC GLS patterns of compound
platforms of the plasma membrane. Hydrogen bonding between their oligosaccharide head groups as well as tight packing together with cholesterol caused by the incorporation of saturated long chain fatty acids into their lipophilic anchor are believed to be parameters resulting in detergent insolubility of GSLs at 4°C (29). We hypothesized that the polyenoic VLCFAs would give neutral and acidic FGSLs of testes physicochemical properties different from those normally ascribed to GSLs. Indeed, in a sucrose gradient, neither the neutral (Fig. 6) nor the acidic (Table II) polyenoic FGSLs of mouse testes floated with the detergent-insoluble fraction. In the insoluble fraction, caveolin and GSLs with saturated fatty acid moieties such as GlcCer (Fig.  6) and the "brain type" gangliosides GM1a, GD1a, and GT1b (Table II) were found. Interestingly, detectable amounts of sialocompound C containing the saturated fatty acid h16:0 did float into the detergent-insoluble fraction 3, which was in contrast to the polyenoic species of sialocompound C (Table II).
Immunohistological Localization of the FGSLs-Compounds A and D were detected by antibody staining in maturing spermatozoa, round and elongated spermatids, and pachytene sper-matocytes, but not in spermatogonia, Sertoli cells, or interstitial tissue (Fig. 7, A and B). In maturing spermatozoa compound A was localized to the tail (Fig. 7E).
A monoclonal anti-sialocompound A/Fuc-GM1 antibody stained a specific cell population residing within the basal compartment adjacent to the basal lamina, where Sertoli cells as well as spermatogonia are situated. In addition, particularly elongated spermatids and residual bodies in the adluminal compartment were sialocompound A-positive, whereas interstitial cells and adluminal spermatocytes appeared to be negative (Fig. 7C). By confocal microscopy no colocalization of sialocompound A with the Sertoli cell marker vimentin was found. Neighboring spermatogonia intensely expressed sialocompound A (Fig. 7, I-L).
These results were confirmed using semithin cryosections (0.6 m thick) stained for nuclei and for sialocompound A. Exclusive cytoplasmic organellar/vesicular localization of sialocompound A in spermatogonia was seen, whereas the Sertoli cells (vimentin-positive and with characteristic nuclei) lacked positive cytoplasmic structures (Fig. 7, M-P). As expected, tes-FIG. 6. Raft gradient of homogenized tissue from wild type mice testes. Rafts, i.e. detergent-insoluble complexes, were separated at 4°C with 1% Triton X-100 from detergent-soluble material by floating up in a sucrose gradient as described. Numbers 1-12 at the bottom of A and B indicate the corresponding gradient fractions. A, testicular neutral GSLs from these fractions were separated on TLC and stained with orcinol/sulfuric acid. GlcCer shifted to the raft fraction 3 as indicated by the red box, whereas the majority of the polyenoic FGSLs stayed in the detergent-soluble fractions 7-12, as indicated by the blue box. Cp.A, compound A. B, caveolin as a marker for detergent-insoluble complexes shifted to fractions 3-7. C, analysis of the ceramide moieties from GlcCer (C 1 , precursor ion scan m/z 264) of raft fraction (fraction 3) and from compound A of non-raft fraction (fraction 10) (C 2 , precursor ion scan m/z 512) by nanoESI-MS/MS. The ceramide moieties of GlcCer from the raft fraction contain predominantly saturated fatty acid residues, whereas those of compound A from the non-raft fraction incorporate solely polyenoic VLCFAs.

TABLE II
Distribution of testicular brain type versus fucosyl gangliosides in a raft gradient Values shown are the percent intensity of the base peak in the spectrum (GD1-16:0). The relative intensity of mass spectrometric signals derived from gangliosides with different oligosaccharide head groups does not necessarily correspond to their relative concentration.  tes of Galgt1Ϫ/Ϫ mice did not stain for sialocompound A (Fig. 7D).

DISCUSSION
The deficiency of brain type gangliosides in Galgt1Ϫ/Ϫ mice have been linked to male infertility. Round spermatids did not elongate, and this defect was accompanied by reduced serum testosterone (1). Despite this striking observation, data on the molecular complexity of GSLs in mouse testes are still sparse. Nakamura and colleagues demonstrated the occurrence of more than 15 different gangliosides by TLC. Structurally identified GSLs of mice testes are Forssman lipid, GM3, and the brain type gangliosides GM1a, GD1a, GD1b, and GT1b, i.e. members of the a-and b-series (see Fig. 1) (1,5,6). Synthesis of the brain type gangliosides depends on the expression of Galgt1 (Fig. 1). It was, therefore, assumed that the brain type gangliosides would be essential for spermatogenesis and testosterone transport (1). However, Siat8aϪ/Ϫ mice, lacking b-series and Siat9Ϫ/Ϫ mice lacking both, a-and b-series gangliosides (Fig.  1), remained fertile (4,8,30).
Based on the previously established pathways of GSL biosynthesis, compound B as well as C may be derived directly from compound A, and compound D from compound B. In contrast, their monosialo derivatives cannot be directly biosynthesized from either of these neutral GSLs ( Fig. 1) (33).
The FGSLs from mouse testes, as shown here, are distinguished from all other testes GSLs, e.g. Forssman lipid, GM3, GM2, GM1a, GD1a, GD1b, GT1b, and GQ1, by their high content of polyenoic VLCFAs, characterizing them as a unique GSL class, possibly synthesized at a different cellular site.
Based on immunohistochemistry Fuc-GM1/sialocompound A could be detected in spermatogonia. In addition, it occurred in elongating spermatids and maturing spermatozoa of wild type mouse testes. Pachytene spermatocytes appeared to lack this ganglioside. The neutral FGSL compounds A and D also localized to spermatids and maturing spermatozoa, but not to spermatogonia by immunohistochemistry. In maturing spermatozoa compound A appeared on the tail, implicating its function in spermatid elongation or motility. Because in wild type testes the occurrence of compounds A, D, and sialocompound A could be immunohistologically linked to testicular germ cells, it is assumed that the other five FGSLs are also expressed by these cells. Supporting this view, cultured Leydig cells as well as Sertoli cells lack FGSLs, as well as GSLs with polyenoic VL-CFAs (data not shown).
The observation that testes of Galgt1Ϫ/Ϫ mice still contain spermatogonia, spermatocytes, and Sertoli cells, but only subtle amounts of GSLs such as GM3 with polyenoic VLCFAs suggests that maturing spermatids are the main source of polyenoic VLCFA-containing sphingolipids.
In the infertile Galgt1Ϫ/Ϫ mice maturing spermatids ap-pear to lose their contact with Sertoli cells, pinching off as multinucleated giant cells and suggesting an arrest in spermatogenesis. It may be surmised that the FGSLs assure attachment of germ cells to Sertoli cells. It is not clear whether this is because of a direct cell-cell interaction with a corresponding ligand of Sertoli cells or via indirect mechanisms. Because spermatocytes and round spermatids still appear in infertile Galgt1Ϫ/Ϫ mice, primary stages of germ cell differentiation do not depend on FGSLs (Fig. 7H). Nevertheless, some apoptotic germ cells of early stages were visible in Galgt1Ϫ/Ϫ testes (data not shown).
In fertile Siat9Ϫ/Ϫ testes fucosylated a-series gangliosides are not replaced by fucosylated gangliosides of other series. Instead, corresponding neutral GSLs accumulate. Hence, sialic acid residues attached to fucosylated, polyunsaturated GSLs do not appear to be crucial for overall spermatogenesis.
In infertile Galgt1Ϫ/Ϫ mice all polyunsaturated FGSLs disappear, and only a minor amount of corresponding LacCer accumulates. Although in these mice testicular polysialo gangliosides may be functionally replaced by GD3, similar to that shown in brain (8), there is no comparable replacement of polyunsaturated FGSLs. Therefore, we conclude that ongoing spermatogenesis depends on the presence of neutral FGSLs that are fucosylated and bear a polyenoic VLCFA but not on the total level of lipid-bound sialic acid in germ cells.
Mouse fucosyltransferases (MFUT-I and -II) fucosylate GA1 or GM1 (34), leading to compound A and sialocompound A, respectively. MFUT-I and -II are expressed in a tissue-specific manner. Deficiency of FUT-II results in lack of Fuc-GA1 and Fuc-GM1 in stomach and colon, whereas deficiency of FUT-I causes loss of FGA1 in pancreas. Surprisingly, in both FUT-Iand FUT-II-deficient mice the testicular expression of Fuc-GA1/compound A and Fuc-GM1/sialocompound A remained normal (7). Redundancy of fucosylation in testes caused by the expression of both FUT-I and -II may guarantee robustness of male fertility. These findings point to the relevance of testicular GSL fucosylation for fertility. Fucosylation of GSLs may also be important in other mammals because some related compounds have been described already in the testes of boar (Fuc-GM1/sialocompound A) and rat (compound D) (32,35).
In addition to the previously published data on compound D in rat testes, we have shown that polyenoic VLCFAs (h28:4, h30:5, and h30:4) represent a major ceramide constituent of neutral and monosialocompounds B, D, and hexosyl D of rat testes (data not shown). These data indicate a species-independent significance of these compounds for spermatogenesis with respect to their exclusive ceramide composition. Therefore, loss of similar GSLs may be important in disorders of human infertility.
In contrast to sphingomyelin, the mass spectra revealed polyenoic VLCFA residues to represent at least 75-90% of the total fatty acids of testicular compound A (Fig. 4, A and C). Because of the lack of internal standards it was not possible to quantify the amount more precisely. GSLs containing polyenoic VLCFAs, as detected in the testes, have not been reported to date. Most of the GSLs reported previously contain more than 90% saturated fatty acid residues. Therefore, they are believed to participate in the formation of lipid-ordered domains in plasma membranes, so-called rafts; these are thought to act as signaling platforms on the cell surface (29). With such a high content of polyenoic fatty acid residues it is likely that the polyenoic FGSLs of testes have different physicochemical membrane properties. Raft molecules and "classical" GSLs are thought to be enriched within the detergent-insoluble fraction at 4°C. Under these conditions, however, none of the FGSLs containing the highly unsaturated fatty acids did float with the raft fraction; in contrast the small amount of FGSLs with saturated fatty acids (sialocompound C (h16:0)) as well as the other GSLs (GlcCer, GM1a, GD1a, and GT1b) containing saturated fatty acids migrated with the detergent-insoluble fraction. Hence, detergent insolubility at 4°C is not a general feature of GSLs.
Compared with classical GSLs, the polyenoic FGSLs found in testes of mice and rat have different physicochemical properties; therefore it is expected that they serve different cellular functions that remain to be uncovered. These functions obviously are particularly important in spermatogenesis. Thus, inhibition of biosynthetic pathways leading to GSL fucosylation and GSL polyunsaturation are potential targets to inhibit spermatogenesis selectively and achieve male contraception.