Male-specific Modification of Human CD52*

CD52 is an unusually short, bipolar glycopeptide bearing a highly charged N-linked carbohydrate moiety and a glycosylphosphatidylinositol membrane anchor. It is exclusively expressed on lymphocytes and in the male genital tract where it is shed into the seminal plasma and inserts into the sperm membrane. The sperm surface molecule has potential significance as a target for antibodies that inhibit sperm function and gamete interaction. Western blot analyses suggested cell type-specific modifications of the antigen. It was purified from seminal plasma and a detailed structural analysis performed. The majority of anchor structures in male genital tract CD52 showed 2-inositol palmitoylation, rendering molecules insensitive toward phospholipase C, and asn-1-alkyl-2-lyso-glycerol structure in place of the diacylated anchor described by Treumann et al.(Treumann, A., Lifely, M. R., Schneider, P., and Ferguson, M. A. (1995) J. Biol. Chem. 270, 6088–6099).N-Glycans of the male genital tract product were based on bi-, tri-, and tetraantennary structures of highly charged (up to -7), terminally sialylated complex-type sugars. A substantial proportion carried varying numbers of lactosamine repeats of which nearly 30% were branched. Different from lymphocytes, 10–15% of allN-glycans of the male genital tract antigen also contained peripheral fucose. These data confirm that male genital tract CD52 is distinct from the lymphocyte form by both N-linked glycans and COOH-terminal attached lipid anchor.

. CD52 monoclonal antibodies are remarkably effective for complement-mediated lysis and have proved to be very potent for lymphocyte depletion both in vitro and in vivo (2,5). Differential screening of a human epididymal cDNA library yielded a highly abundant cDNA, named HE5, which turned out to be colinear to the CD52 cDNA (6). Epididymal and lymphocyte cDNAs predict the same precursor peptide, including NH 2 -terminal and COOH-terminal signals typical of GPIanchored proteins (6). Both mRNAs are derived from the same single-copy gene (6,7) which has been localized on chromosome 1.
CD52 transcripts are found in the epithelial cells of the distal epididymis and deferent duct but not in either spermatogenetic cells or on spermatozoa, nor in the more proximal parts of the epididymis (6,8). Anti-CD52 antibodies react strongly with parts of the male genital tract, particularly with the cauda epididymal epithelium and cauda fluid, and with epididymal but not testicular spermatozoa. CD52 reactivity persists in the ejaculate and on ejaculated sperm, implying an exogenous, post-testicular origin of the sperm membrane antigen (7,9,10). This finding of sperm membrane "painting" by CD52 in vivo was unexpected considering the well established biosynthetic pathway of GPI-anchored membrane proteins, and that a transfer in vivo of the lymphocyte antigen has not been observed. Meanwhile, evidence has been provided that a massive cell-to-cell transfer of GPI-anchored HE5/CD52 occurs in vivo within the male genital tract (10,11). Shedding from the epididymal epithelium of vesicles or of micellar intermediates may be involved interacting with the spermatozoa (12).
Compared with the large body of information available on the structure and function of the zona pellucida (for review, see Ref. 13) very little is known about the glycocalyx of spermatozoa, although a vital component during the fertilization process as well. Surface carbohydrate labeling experiments on intact rat spermatozoa from the cauda epididymidis (14 -16) show a very high selectivity: the only accessible molecule is the rat counterpart of CD52 (7,16,17). Since it appears on the sperm surface coinciding with the acquisition of fertilizing capacity, it was also called the "major maturation-associated antigen" (18,19). Thus, acquisition of the CD52 antigen seems to be one of the major events forming the glycocalyx of mammalian spermatozoa (7,10,17).
Sperm surface glycoproteins are targets for immune recognition, and the possibility has been discussed of using them as targets for contraception (for review, see Ref. 20). CAMPATH-1 antibodies can agglutinate and completely immobilize sperm in the presence of complement (9). However, the finding that the CD52 antigen is produced by the male genital tract epithelium on the one hand and by lymphocytes on the other, raises some issues related to its use in interventional therapy, unless cell type-specific modifications could be found which would enable a clear differentiation between male genital tract and lymphocyte antigen. To explore the extent of specific modifications occurring in the male genital tract, we performed a structural analysis of seminal plasma CD52, i.e. of its N-linked glycans and the GPI-anchor, as purified from human ejaculates.
Folch Extraction and Gel Filtration Chromatography-CD52 was extracted from human semen samples and from peripheral blood lymphocytes according to a modified Folch extraction as described previously (4,9). For large scale preparation of seminal plasma antigen, pooled ejaculates were centrifuged to remove spermatozoa, lymphocytes, and cell debris. Cell-free supernatants were stirred with methanol and chloroform (4:11:5.4) and the extraction continued as described (4). Freeze-dried extracts were resolved in water, incubated for 5 min at 40°C, and loaded onto a 90 ϫ 2.5-cm Sephadex G-50 column equilibrated with 120 mM (NH 4 ) 2 CO 3 , pH 8.0. The CD52-containing fractions forming micellular structures of high molecular weight were recovered from the flow-through.
CAMPATH-1G Affinity Chromatography-Purified CAMPATH-1G mAb was coupled to epoxy-activated Sepharose 6B according to the suggestions of the supplier and affinity chromatography performed as described (1), however, using SB10 as a detergent instead of deoxycholate. The antigen was eluted after extensive washing with 0.05 M diethylamine, pH 11.5, neutralized with 3 M NaOAc, pH 4.0, and dialyzed against deionized (Milli-Q) water.
HPLC Reversed Phase Chromatography-Reversed phase chromatography was performed using a HPLC system (Beckmann, Mü nchen, Germany) equipped with a C 8 -column (20 ϫ 1 cm, Macherey Nagel, Dü ren, Germany). The column was eluted at 2 ml/min with a mixture of 100 mM (NH 4 ) 2 CO 3 , pH 8.0, including 5% methanol (solution A) and 60% methanol including 3% chloroform (solution B). Starting conditions were 2 min 100% A, followed by a linear gradient rising to 100% B within 40 min. CD52-containing fractions were pooled and purity of the final preparation confirmed by NH 2 -terminal amino acid sequence analysis using a Procise TM instrument (Perkin-Elmer, Foster City, CA).
Enzymatic Release of N-Linked Glycans by N-Glycosidase F-On an analytical scale, Folch extracts were digested for 2 h at 37°C with 5 milliunits/l of PNGase F (4). On a preparative scale, 50 nmol of purified CD52 glycopeptide were incubated with 6 units of PNGase F (22).
Recovery of Oligosaccharides and Deglycosylated GPI Peptide-Released oligosaccharides and deglycosylated GPI peptide were recovered by HPLC on a C 8 -reversed phase column (12.5 ϫ 0.4 cm, Macherey-Nagel). The reaction mixtures were adjusted to 5% 1-propanol and 100 mM ammonium acetate, pH 5.5, and incubated for 5 min at 37°C before loading. The column was eluted at 0.5 ml/min with 100 mM ammonium acetate, pH 5.5, containing 5% 1-propanol (solution A) and 80% 1-propanol (solution B). Starting conditions were 5 min 100% A, followed by a linear gradient to 100% B in 40 min; elution was monitored at 220 nm. N-Glycans were recovered from the flow-through. CD52 GPI peptide containing fractions were separated into two pools (pool 1 and pool 2) and purity proven by NH 2 -terminal sequence analysis. Oligosaccharides were desalted by fast protein liquid chromatography on a Sephadex G-25 Superfine Fast-Desalting HR 10/10 column (Amersham Pharmacia Biotech).
Generation of CD52 PI Fractions-Deglycosylated CD52 GPI peptide was extracted 3 times with equal volumes of water-saturated 1-butanol to remove residual contaminations. Approximately 0.5 nmol of CD52 pool 1 and 10 nmol of CD52 pool 2 were deaminated as described (1). Released phosphatidylinositol residues were extracted with water-saturated 1-butanol and dried under nitrogen.
Electrospray Ionization Tandem Mass Spectrometry (ESI-MS/ MS) 1 -Deglycosylated CD52 was dissolved in methanol containing 10 mM ammonia to a concentration of approximately 10 pmol/l, and 3 l filled into gold-coated nanospray glass capillaries. The tip of the capillary was placed orthogonally in front of the entrance hole of a Quadrupole time-of-flight (QTOF) mass spectrometer (Micromass, Manchester, United Kingdom) equipped with a nanospray ion source, and a voltage of Ϫ1200 V was applied. For collision-induced dissociation experiments, parent ions were selectively transmitted from the quadrupole mass analyzer into the collision cell. Argon was used as the collision gas and the kinetic energy was set at around ϩ40 eV. The resulting daughter ions then were separated by an orthogonal time-of-flight mass analyzer. The PI moiety of CD52 was analyzed similarly on a TSQ 700 triple quad mass spectrometer (Finnigan MAT, Bremen, Germany).
Preparative MonoQ Anion-exchange Chromatography of N-Glycans-Desalted oligosaccharides were dried and redissolved in 0.5 ml of MilliQ water. N-Glycans were applied to a MonoQ HR 5/5 column (Amersham Pharmacia Biotech) and separated into charge groups according to Nimtz et al. (21). Oligosaccharides were detected by their ultraviolet absorption at 206 nm. Peaks were pooled and desalted as described.
High pH Anion-exchange Chromatography (HPAEC-PAD) of Oligosaccharides-Analysis was performed according to Grabenhorst et al. (23) using a Dionex Bio LC system (Dionex, Sunnyvale, CA) equipped with a CarboPac PA1 column (0.4 ϫ 25 cm) in combination with pulsed amperometric detection (PAD). Oligosaccharides were chemically desialylated prior to HPAEC-PAD analysis by incubation in 100 l of 0.2% trifluoroacetic acid for 1 h at 85°C and the reaction mixture neutralized with 12.5% NH 4 OH. Elution was performed using a slightly modified gradient: a 5-min isocratic run with 100% solvent A followed by a linear gradient from 0 to 20% solvent B over a period of 30 min, a linear gradient to 30% solvent B within 10 min, and a linear gradient to 100% B within 2 min. Solvent A ϭ 0.2 M sodium hydroxide, solvent B ϭ 0.6 M sodium acetate in solvent A, flow rate ϭ 1 ml/min.

Matrix-assisted Laser Desorption/Ionization Time of Flight Mass Spectrometry (MALDI/TOF-MS)-N-Glycans
were reduced, permethylated, and extracted as described (24). Determination of the molecular masses of permethylated oligosaccharides was carried out by positiveion MALDI/TOF-MS using a Bruker REFLEX time of flight instrument according to Grabenhorst et al. (23).
Methylation Analysis-N-Glycans were reduced, permethylated, extracted, hydrolyzed, reduced again, and peracetylated as described (24). Separation and identification of partially methylated alditol acetates were performed using a gas chromatograph equipped with a 30-m DB5 capillary column connected to a GCQ ion-trap mass spectrometer (Finnigan MAT) running in the electron-impact mode.

RESULTS
Western Blot Analysis of CD52 Antigens-Taking advantage of the glycolipid-like properties of CD52, the antigen was en-riched from whole ejaculates, washed sperm, and cell-free seminal plasma by modified Folch extraction (9). Western blot analysis employing the CAMPATH-1G mAb revealed a complex pattern of at least five immunostained bands of molecular mass Ϸ15-20 kDa, the pattern of single donor samples being as heterogeneous as that of pooled sperm samples (Fig. 1a). Sialidase treatment did not change this pattern (data not shown), suggesting that the negative charge of substituting sialic acids did not affect electrophoretic mobility. Microheterogeneity patterns of sperm and seminal plasma CD52 were similar (Fig.  1a), however, the staining intensity of comparable amounts of sperm extract was always much lower (approximately 10%) than that of cell-free seminal plasma, corroborating previous enzyme-linked immunosorbent assay data that there is substantially more CD52 antigen in seminal plasma than on sperm (9,12). PNGase F-digestion converted the heterogeneous Ϸ15-20 kDa immunoreactivity into a single band of Ϸ6 kDa (Fig. 1b), suggesting that a PNGase F-sensitive site was present in male genital tract CD52 as shown earlier for the lymphocyte antigen (1,3,4). Cell-free preparations of seminal plasma were used during the entire following study, also to get rid of the majority of contaminations originating from cellular elements other than spermatozoa, including leukocytes which are present in normal human ejaculates in varying numbers (25). Seminal plasma and lymphocyte extracts were compared by Western blot analysis (Fig. 1, b and c). Before deglycosylation, antigens from both sources showed considerable microheterogeneity. Electrophoretic mobility and microheterogeneity, however, were higher in seminal plasma (5 bands) than in the lymphocyte extracts (3 bands), suggesting cell type-specific differences of glycans (Fig. 1b). After PNGase F digestion the electrophoretic mobilities were still different suggesting cell type-specific modifications also of the lipid anchor ( Fig. 1, b and c). This was emphasized by an analysis employing three types of mAbs directed against different epitopes of the CD52 molecule (Fig. 1c), (i) the 097 mAb (26), recognizing a peptide epitope (27); (ii) the CF1D12 mAb (5), binding to an epitope predominantly formed by the N-glycan; and (iii) the CAM-PATH-1G mAb, binding to an epitope formed by three amino acids of the carboxyl terminus and the lipid anchor (27).
Purification of CD52-The cell-free fraction of human ejac-ulates was a convenient source of large amounts of male genital tract CD52. Approximately 700 g, corresponding to 100 nmol of CD52 antigen, were recovered from approximately 300 ml of pooled seminal plasma (see "Experimental Procedures"). Purity of the preparation was confirmed by sequential Edman degradation. No amino acid residues other than that of the known mature CD52 peptide backbone (3,4) were detected. A blank at Edman cycle 3 (data not shown, compare Fig. 2c for sequence) was most probably due to a completely occupied N-glycosylation site at Asn-3 on the basis that glycoamino acids are not recovered to a significant degree. The two serines near the carboxyl terminus, by comparison, were clearly detected and thus must have been essentially unoccupied, suggesting that there was no O-glycosylation present. After preparative PN-Gase F digestion, the released N-glycans were separated from the GPI peptide by C 8 -reversed phase chromatography and recovered from the flow-through for subsequent analysis. Edman degradation after cleavage of N-glycans again showed no foreign amino acid residues, except that the former asparagine had been quantitatively transformed into aspartic acid. Positive mode MALDI/TOF-MS measurements of the native glycopeptide preparation (not shown) yielded a very broad asymmetric molecular ion signal ranging from approximately m/z ϭ 7000 to 8300 with an apex around m/z ϭ 7600. This is considerably less than the size apparent during SDS-polyacrylamide gel electrophoresis (Fig. 1). The most likely cause for the occurrence of this very broad signal is the extensive microheterogeneity of N-linked oligosaccharides. The calculated mass of the GQDNTSQTSSPS dodecapeptide backbone is 1174 Da, that of the expected GPI peptide ranged from 2642 to 3031 Da (compare Fig. 2c: fragmentation scheme) suggesting that nearly two-thirds of the molecular mass of the glycosylated molecule are accounted for by the N-linked carbohydrate moiety.
Analysis of the Deglycosylated GPI Peptide of CD52-During reversed phase HPLC, the CD52 antigen eluted in two partly overlapping pools of different hydrophobicity. Dot blot analyses employing CAMPATH-1G as well as Edman sequencing confirmed that both pools contained deglycosylated CD52. Approximately 40 times more material was contained in the later eluting and hence more hydrophobic pool 2. Considering the very small amounts of material in pool 1 and the considerable overlap of both pools, a separate analysis of pool 1 was not attempted. Compositional analysis by GC/MS of the deglycosylated GPI-anchor peptide in pool 2 after methanolysis and pertrimethylsilylation (not shown) revealed mannose, GlcNH 2 , and inositol as monosaccharide components. In addition, two peaks of approximately equal intensity could be assigned to the trimethylsilylated derivatives of C16:0 and C18:0 alkylated glycerol by their characteristic fragmentation pattern, indicating the presence of an alkylated PI anchor moiety instead of the acylated structures isolated from lymphocyte CD52 (1).
phodiester bridge. The third and fourth molecular ion species differed from the first and second species by mass increments of 123.0 Da, corresponding to the mass of one additional phosphoethanolamine moiety. The last four molecular ion species, being nearly five times as intense, could be explained by the occurrence of an additional C16:0 fatty acid. No indication for the presence of substantial amounts of structures bearing a further fatty acid residue could be detected.
These assignments were confirmed by tandem mass spectrometric analysis of the triple charged dominant species detected at m/z ϭ 968.7. From the resulting daughter ion spectrum (Fig.  2b), the sequence of the various components of this complex molecule could be unequivocally deduced. From the fragment ion pair at m/z ϭ 423.3 and 405.3 (minus H 2 O) including the alkyl (C18:0)-substituted glycerol residue, it is obvious that the fatty acid moiety was not bound to this part of the molecule. Rather, it was linked to the inositol giving rise to a very intense primary fragment ion. The existence of an acylated (palmic acid) derivative of the molecule is also evident from the intense signal at m/z ϭ 823.6, accompanied by a relatively weak, but clearly visible secondary fragment ion at m/z ϭ 585.4, due to elimination of the fatty acid. The next building block of the molecule is a GlcNH 2 residue, as indicated by the intense fragment ion at m/z ϭ 984.7, followed by a phosphoethanolamine bearing mannose residue giving rise to the relatively weak double charged fragment at m/z ϭ 634.4 (Fig. 2b). Two further unsubstituted mannose residues were implicated by the occurrence of fragment ions at m/z ϭ 1431.9 and 715.4 (doubly charged), and 1593.9. The mannose residues were followed by another phosphate group as indicated by the fragment ion at m/z ϭ 1673. 9 showing the characteristic relative mass shift of 80 milliunits. A series of intense doubly charged ions at m/z ϭ 1295.1, 1286.1, and 1264.1, respectively, could be assigned to fragments generated by the elimination of the three amino-terminal amino acids (GQD) followed by the loss of water and CO 2 . This was confirmed by the detection of corresponding fragments of the amino-terminal tri-and dipeptides at m/z ϭ 316.1 (GQD) and 201.1 (GQ), respectively. The above assignment of the fragment ions was corroborated by the daughter ion spectrum (not shown) of the analogous triple charged species at m/z ϭ 959.3, equipped with a C16:0 alcohol linked to the glycerol residue. All fragment ions containing this alkylated glycerol moiety showed the expected mass shift. Since all other components yielded molecular ions of much lower intensity, no relevant MS/MS data could be obtained. Therefore the linkage position of the second phosphoethanolamine moiety in the higher molecular weight components could not be determined.
These results were confirmed by ESI-MS/MS analysis of the PI anchor part of the GPI peptide (not shown) which was obtained by a deamination of the intact GPI anchor (28). Major deprotonated molecular ions occurred at m/z ϭ 823 and 795. With relative intensities of approximately 20%, signals at m/z ϭ 585 and 557 were detected by negative ion mode ESI-MS. These results again suggested anchor structures with (approximately 80%) and without (approximately 20%) one additional fatty acid linked to a mono-alklylated (C16 or C18) GPI anchor. The daughter ion spectra of the signals at m/z ϭ 823 and 795 (with the expected mass shifts, not shown) containing fragment ions of m/z ϭ 423 and 405 (m/z ϭ 395 and 377, respectively) and especially 479 confirmed linkage of the fatty acid (predominantely C16, but small amounts of C14 and C18 were also detected) to the inositol moiety. The smaller molecular ions at m/z ϭ 585 and 557 were shown by ESI-MS/MS analysis to have an analogous structure, however, lacking the fatty acid component, and giving rise to an intense signal at m/z ϭ 241, due to the existence of a free hydroxyl group at position 2 of the inositol (29). In accordance, no molecular ions corresponding to a diacetylated PI anchor species with an additional fatty acid bound to the glycerol moiety could be detected.
HPAEC-PAD Mapping of CD52 N-Glycans-Preliminary lectin binding studies on intact CD52 employing biotinylated lectins from Sambucus nigra and Maackia amurensis had suggested the presence of high amounts of ␣2,3-linked as well as ␣2,6-linked sialic acid in seminal plasma CD52 (data not shown). Correspondingly, HPAEC-PAD analysis of the native CD52 N-glycans yielded a complex pattern of peaks, the majority of oligosaccharides eluting in a region that would predict highly sialylated glycans (data not shown). To improve the chromatographic resolution, an aliquot of oligosaccharides was enzymatically desialylated before HPAEC-PAD analysis. The elution profile again revealed a highly complex mixture (Fig. 3,  lowest panel), and, when compared with a standard oligosaccharide preparation containing proximally fucosylated bi-, tri-, and tetraantennary structures (Fig. 3, top panel), only about 10% of the CD52 glycans could be tentatively assigned to standard peaks by their retention times. For a more detailed HPAEC-PAD mapping, oligosaccharides were quantitatively released from approximately 95 nmol of CD52 and the native (sialo)-oligosaccharides subfractionated according to charge. As shown in Fig. 4, nine carbohydrate containing pools were collected comprising differently charged groups of glycans, their charge ranging from 0 or Ϫ1 in fraction M1 up to Ϫ6 in fractions M5/6 and M6. Fraction M1 seemed to contain only very small amounts of heterogeneous carbohydrate material and was not considered further. Prior to a detailed mass spec-trometric analysis (see below) an aliquot of each glycan-containing pool, i.e. M2, M3, M4, M4*, M5, M5*, M5/6, and M6, was desialylated and subjected to HPAEC-PAD analysis individually (Fig. 4), at the same time estimating the relative proportion of these fractions (compare Table I). The elution patterns obtained were compared with the retention times of known oligosaccharide standards (Fig. 3, top panel). The chromatograms indicated a total of at least 50 chromatographically different glycans associated with the single N-glycosylation site at Asn-3 (see Fig. 3, panels M2-M6).
Mass Spectrometric Analysis of CD52 N-Glycans-The native oligosaccharide mixtures of the individual MonoQ pools M2-M6 were subjected to MALDI/TOF-MS after reduction and permethylation, and following hydrolysis, reduction, and acetylation, also to methylation analysis for the determination of monosaccharide linkages. The results, as summarized in Table  I (MALDI analysis) and Table II Fig. 3). Since M5* to M6 contained only 10% of the total N-glycans, the amount subjected to HPAEC mapping was 10% compared with 2% for M2-M5.
oligosaccharides. The series of N-glycosidic structures bearing lactosamines and/or fucose residues were typically identified by their characteristic masses. Their approximate ratios were determined by assessment of the relative intensities of the respective molecular ion signals (sodium adducts, compare Table  I). MonoQ pool M2 contained ϳ15% of total N-glycans. MALDI/ TOF-MS of the derivatized M2 carbohydrates (Fig. 5a) revealed a dominant molecular ion at m/z ϭ 2984 (NeuAc 2 Hex 5 HexNAc 3 dHexHexNAc-ol ϩ Na ϩ ) indicating the presence of a disialylated biantennary structure bearing a proximal fucose. A series of additional molecular ions with decreasing intensities showing mass increments of 449 at m/z ϭ 3433, 3882, and 4332 can be interpreted as modifications of the basic biantennary structure with additional N-acetyllactosamine repeats. This view was supported by the methylation analysis data where almost exclusively the 2-monosubstituted outer mannose derivative was detected excluding the presence of significant amounts of isomeric structures with higher antennarity (see Table II). These interpretations were also in agreement with the elution position of peak 2a (compare Fig. 3, panel M2) in HPAEC-PAD mapping corresponding to that of the biantennary standard oligosaccharide (see legend Fig. 3). The elution positions of 2b and 2c were compatible with the presence of 1 and 2 N-acetyllactosamine repeats in the basic structure 2a. Digestion of the desialylated MonoQ pool M2 with endo-␤galactosidase from B. fragilis and subsequent reduction and permethylation (data not shown) yielded a major molecular ion at m/z ϭ 2262 (Hex 5 HexNAc 3 dHexHexNAc-ol ϩ Na ϩ ) as could be expected for a proximally fucosylated asialo biantennary N-acetyllactosamine-type oligosaccharide. The detection of two ion signals at m/z ϭ 2058 and 1854 indicated the removal of 1 or 2 N-acetyllactosamine repeats from the aforementioned glycan (data not shown). Based on the ratio of the 2,3,4-Gal and 2,4,6-Gal derivatives obtained in methylation analysis (Table  II) for the native oligosaccharide pool M2 we calculate that 30% of the NeuAc is ␣2,6-linked to galactose. These data were also corroborated by digestion experiments using ␣2,3-specific Newcastle disease virus neuraminidase (data not shown). An additional rather weak signal at m/z ϭ 2810 is compatible with the presence of small amounts of a disialylated biantennary structure lacking proximal fucose (Fig. 5a). A further series of weak molecular ions with mass increments of 174 milliunits compared with the major structures containing at least one Nacetyllactosamine repeat described above suggests the presence of approximately ϳ10% of structures bearing an additional peripheral fucose residue and thus constituting sialyl Lewis X-(sLeX-) or VIM2-type structural motifs. This was confirmed by the detection of adequate amounts of 3,4-disubstituted GlcNAc by methylation analysis.
The trisialylated MonoQ pool M3 was the largest comprising approximately 32% of the total CD52 oligosaccharide material. The structures consisted predominantly of the 2,4-branched triantennary isomeric structure as revealed by methylation data by the detection of 3,4,6-Man, 3,6-Man, and 3,4-Man derivatives at a ratio of 9:7:2 (see Table II). The series of molecular ions detected in MALDI/TOF-MS spectra (Fig. 6a) of the derivatized glycans (m/z ϭ 3796, 4245, 4695, 5144, 5594, and 6044) indicated a modification of the basic triantennary oligosaccharide with 0 -5 N-acetyllactosamine repeats. Peaks containing at least one repeat were accompanied by a second molecular ion peak increased by 174 milliunits. This indicates the occurrence of substantial amounts of peripheral fucose (ϳ10%) again suggesting the sLeX/VIM2-type motif. No 3-substituted GlcNAc derivative characteristic for the presence of type I N-acetyllactosamine antennae was detected by methylation analysis which would have lead to the Lewis A (LeA) motif after fucosylation at O-4 of GlcNAc residues. The HPAEC-PAD pattern after desialylation is in agreement with this interpretation (Fig. 3, panel M3 and Fig. 6c). Peak 3a elutes at a position identical with standard peak b (Fig. 3, top panel, 2,4-branched triantennary glycan); peaks 3a-3f are separated by roughly 2 min from each other which is expected for structures containing no, 1, 2, 3, 4, or 5 N-acetyllactosamine repeats (compare also pattern of biantennary oligosaccharides, panel M2 in Fig. 3). Peaks 3aЈ-dЈ would represent the corresponding 2,6-branched isomeric triantennary oligosaccharides, the presence of which was confirmed by the methylation data (Table II). Digestion of the desialylated oligosaccharide pool M3 with endo-␤-galactosidase and subsequent reduction and permethylation yielded the MALDI/TOF signals m/z ϭ 3795 (triantennary: NeuAc 3 Hex 6 HexNAc 4 dHexHexNAc-ol ϩ Na ϩ ), 3230, 2664, and 2099 representing 75% of M3 oligosaccharides (Fig. 6b). The latter three signals could be explained by the cleavage of 1, 2, or 3 NeuAc-Gal-GlcNac-Gal units from 1, 2, and 3 outer antennae by the endoglycosidase. The HPAEC-PAD pattern of MonoQ pool M3 dramatically changes after endo-␤-galactosidase digestion (Fig. 6d). The majority of peaks elutes much earlier than before digestion (compare Fig. 6c) due to cleaved oligosaccharide chains, only peak 3a and 3aЈ not containing lactosamine remain unchanged. However, minor peaks are revealed co-eluting with lactosamine-containing isomers. Methylation analysis of pool M3 also revealed substantial amounts of the 3,6-disubstituted Gal (Ͼ10% of total Gal derivatives). Therefore we assume that a significant portion of the structures also contain branched antennae which are resistant to endo-␤-galactosidase (30). This is in agreement with the MALDI/TOF spectra obtained after the enzyme treatment (Fig. 6b). The appearance of two additional series of molecular ions: first series with m/z ϭ 4127 (compatible with biantennary plus 3R, tri-plus 2R or tetra-plus 1R, uncleaved), 3562 (4127 minus Hex, NeuAc), 2998 (4127 minus 2 Hex, 2 NeuAc) and the FIG. 4. Anion-exchange chromatography on MonoQ of the total N-glycans from seminal plasma CD52. Oligosaccharides were released from purified CD52 by PNGase F digestion and separated from peptide by reverse phase HPLC. The pooled carbohydrate fractions obtained by separation on MonoQ were designated M1-M6. The neutral and monosialylated fraction M1 contained only very small amounts of carbohydrate material and was not further studied in detail. MALDI/ TOF data of M2-M6 indicated preponderantly hybrid-type structures with proximal and some peripheral fucose to be present (not shown). The following fractions were combined for further analysis: pool M2 (disialoglycans) contained 15%, pool M3 (trisialo glycans) contained 32%, pool M4 and 4* (tetrasialo glycans) contained 19 and 7%, respectively; pools M5 and 5* (pentasialo glycans) contained 10 and 4% of the oligosaccharide material applied. Pool M5/6 (a mixture of penta-and hexasialylated glycans) and pool M6 (hexa-and heptasialylated glycans) contained 3% each. second series with m/z ϭ 4244 (biantennary plus 2R, tri-plus R or tetra-, uncleaved), 3680 (4244 minus Hex, NeuAc), and 3114 (4244 minus 2 Hex, 2 NeuAc), partially consisting of uncleaved structures with one or more lactosamine repeats, can only be explained by these branched structures. The identification of terminal Gal in methylation analysis is consistent with the branched repeats in the trisialylated and basically triantennary oligosaccharides mentioned above.
The tetrasialylated pool M4 contained approximately 20% of total CD52 oligosaccharides. The MALDI/TOF spectrum revealed a series of molecular ions which could be explained by the presence of tetraantennary oligosaccharides with 2 to 5 N-acetyllactosamine repeats (Hex 9 -12 HexNAc 7-10 dHexHex-NAc-ol ϩ Na ϩ ). The HPAEC-PAD profile is in agreement with this interpretation (Fig. 3, panel M4, coelution of 4a and 4b with the corresponding standards f and g). Endo-␤-galactosidase digestion of the desialylated oligosaccharide mixture gave products which upon reduction and permethylation yielded 2 weak molecular ions at m/z ϭ 3162 (Hex 7 HexNAc 5 dHexHexNAc-ol ϩ Na ϩ ) and 2958 (3162 minus Hex). Intense signals were detected at m/z ϭ 2753 (3162 minus 2 Hex), 2549 (3162 minus 3 Hex), and 2345 (3162 minus 4 Hex) which we interpret to result from cleavage of no or 1 to 5 N-acetyllactosamine repeats. This is in agreement with the HPAEC-PAD mapping data where all peaks were shifted toward earlier elution times after endo-␤-galactosidase digestion. The methylation data corroborate the tetraantennary nature of the majority of oligosaccharides in the tetrasialylated pool M4. Furthermore, as can be derived from the value observed for the disubstituted galactose derivative, approximately 20% of oligosaccharides should contain a branched outer antenna.
The second tetrasialylated fraction M4* yielded three major molecular ion signals after derivatization: m/z ϭ 4604 (NeuAc 4 Hex 7 HexNAc 5 dHexHexNAc-ol ϩ Na ϩ ), 5054 (4604 plus 1R), and 5503 (4604 plus 2R). 10% of each of these ions were accompanied by molecular ions plus 174 milliunits, indicating the presence of an additional peripheral fucose. The lower proportion of N-acetyllactosamine repeats and branched antennae when compared with pool M4 is reflected also by their methylation data (see Table II).
Oligosaccharide pool M5 contained pentasialylated glycans of m/z ϭ 6313 (NeuAc 5 Hex 10 HexNAc 8 dHexHexNAc-ol ϩ Na ϩ ), 6762 (6313 plus 1R), and 7211 (6313 plus 2R). These molecular ions are consistent with the presence of tetraantennary structures bearing 3, 4, or 5 additional N-acetyllactosamine repeats. The high proportion of 3-substituted galactose derivatives corroborates the occurrence of N-acetyllactosamine repeats. The detection of approximately one 3,6-disubstituted galactose residue per 3,6-disubstituted mannose suggests the presence of at least one single branched repeat per N-glycan (pentaantennary structure). Furthermore, in the MALDI/TOF spectrum of pool M5, exclusively molecular ions containing 5 NeuAc were present. Since only small amounts of nonsubstituted terminal galactose were found during methylation analysis, we conclude that indeed most of these structures are pentaantennary Nglycans with 3 up to 5 repeats and complete terminal sialylation. When compared with the HPAEC-PAD profile of pool M4 oligosaccharides (see Fig. 3, panel M4) the elution times of peak 4b and 5a and also 4c and 5b are identical. Therefore, we conclude that m/z ϭ 6313 and 6762 represent the fully sialylated forms of structures m/z ϭ 5952 and 6401 detected in pool M4 glycans. The corresponding peripherally fucosylated structures (m/z ϭ plus 174 milliunits, 10, 11, and 3% of signal intensity, respectively) of the pentasialylated oligosaccharides were also detected. Only small amounts of heterogeneous oligosaccharide populations (4% of total carbohydrate content) were identified in MonoQ pool M5* which possessed the same characteristics as pool M5 except that the carbohydrates contained only 2 and 3 N-acetyllactosamine repeats.
Reduction and permethylation of the remaining two MonoQ pools M5/6 and M6 (approximately 6% of total oligosaccharide material) suggested the existence of hexasialylated oligosaccharides. Only M6 is discussed, since HPAEC-PAD patterns were very similar. The signal at m/z ϭ 6225 detected in the MALDI/TOF spectrum of M6 (Fig. 5b) could be explained by the hexasialotetraantennary structure containing two N-acetyllactosamine repeats (NeuAc 6 Hex 9 HexNAc 7 dHexHexNAc-ol ϩ Na ϩ ); 6674 (6225 plus 1R), 7123 (6225 plus 2R), 7572 (6225 plus 3R). A series of additional relatively weak peaks at m/z ϭ 7035, 7484, and 7933 can be explained by the addition of a further NeuAc residue to each of the three largest former structures.
The HPAEC-PAD profile (Fig. 3, panel M6) gave a pattern of repeating peaks separated by roughly 1 min suggesting structures with high amounts of N-acetyllactosamine repeats to be present, since the difference in elution time between structures decreases with growing number of repeats (compare also pattern M2 and M3 for the bi-and triantennary oligosaccharides). Endo-␤-galactosidase digestion of the desialylated fraction followed by derivatization yielded three series of molecular ions which can be explained by the tetraantennary structure with 1 or 2 repeats, each minus 0 -4 Gal.

DISCUSSION
One of the most distinctive features of the CD52 antigen is its tissue distribution: besides lymphocytes and monocytes it is only found in parts of the male genital tract where it is produced by the epithelial cells lining the distal epididymal and deferent duct (6,7,9). Its most remarkable structural property is a bipolar, glycolipid-like composition with a large hydrophilic N-linked carbohydrate moiety at the one pole and a hydrophobic GPI anchor on the other (1, 3). Our results show that the structures of both the carbohydrate and the GPI anchor moieties are distinct in seminal plasma preparations as compared with the lymphocyte CD52 antigen (summarized in Fig. 8). The  ϩ ϭ2984, 3433, 3882, 4332, and 4780). The molecular ion signals corresponding to structures with 1 or 2 additional repeats are each accompanied by a minor peak (ϳ10%) corresponding to peripherally fucosylated structures (ϩ174 milliunits). The signal at m/z ϭ 2810 is due to a disialylated biantennary structure without a fucose residue. The signal marked by an asterisk could not be assigned to an oligosaccharide. Panel b, hexa-and heptasialylated N-glycans from MonoQ fraction M6. The first peak (M ϩ Na ϩ ϭ 6225) represents a hexasialylated tetraantennary N-glycan with 2 lactosamine repeats accompanied by signals due to the addition of 1-3 further lactosamine repeats (m/z ϭ 6674, 7123 and 7572). Molecular ions at m/z ϭ 7035, 7484, and 7933 are due to heptasialylated tetraantennary structures with 3-5 lactosamine repeats. See Table I for MALDI data of all MonoQ fractions. occurrence of male-specific modifications were first suggested by a significantly reduced sensitivity of sperm and seminal plasma CD52 toward PI-specific phospholipase C 2 and by different electrophoretic mobility patterns during Western blot analysis of native and deglycosylated sperm and seminal plasma CD52 as compared with lymphocyte extracts (see Fig.  1). This was confirmed by a detailed structural analysis.
GPI Anchor Structure of Male Genital Tract CD52-During reversed phase chromatography of seminal plasma CD52 the vast majority (Ͼ95%) was contained in only one fraction which, however, during ESI-MS analyses revealed structural heterogeneity (Fig. 2a). The PI anchor of approximately 80% of molecules was characterized by an inositol acylation, mainly palmitoylation, while the remaining 20% did not contain this acylation. Obviously, absence or presence of an additional acyl group did not result in a separation into two distinct fractions as has been described earlier for spleenic antigen preparations (1,3). The acyl chain appeared to be located at the 2-position of the inositol ring (1,29;Fig. 2b), and appeared to be predominantly palmitate with trace amounts of stearate and myristate. Since the antigen was prepared from pooled human ejaculates, we cannot exclude that the structural subclasses observed represent a mixture of several distinct CD52 structures present in the male population, each individual sperm donor showing only one of them. The occurrence of a 2-inositol acylation in the majority of molecules would explain the greater resistance of seminal plasma CD52 to PI-PLC action 2 which requires a free 2-hydroxyl group on the inositol ring (31). Interestingly, high PI-PLC sensitivity has been observed for the homologous counterpart of CD52 on rat spermatozoa (16). 3 This difference between the two species counterparts might originate from differences in the GPI anchor biosynthesis of the male genital tract epithelium. Intracellular GPI donors contain predominantly acylated inositol, and inositol acylation of mature cell surface proteins is regulated via post-transfer deacylation which in general seems to be cell type-specific (32).
Spleenic CD52 antigen preparations have been described to consist of two subclasses of comparable abundance, named CD52-I and CD52-II (1,3). The only detectable structural difference between these two subclasses was in their GPI anchor structure, CD52-I containing exclusively distearoyl-PI, while the PI-PLC-resistant CD52-II contains predominantly an inositol-palmitoylated stearoyl-arachidonyl-PI (1). Unlike this, the glycerolipid component of seminal plasma CD52 was made of sn-1-alkyl-2-lyso-glycerol only (Fig. 2). This unusual monoalkyl glycerol structure has not been previously described in mammals. Lysoalkyl-or -acyl-PI forms, however, are abundant in the glycosylated phosphatidylinositols of certain parasitic protozoa (Refs. 29, 33, and 34; for review, see Ref. 35). The significance of these differences is unknown. Male genital tract CD52 is transferred from the epithelial cells to the sperm membrane during epididymal passage with its anchor intact (10,11). However, a similar transfer of lymphocyte CD52 has not been described. While Illangumaran et al. (36) suggest that a cellto-cell transfer between blood cells of GPI-anchored proteins may be inhibited by serum proteins, different anchor structures may also play a role. DAF (CD55) molecules with in vitro-modified "single-footed" lyso-glycerol lipid anchors have been reported to display differences in their ability to incorporate into the membranes of living cells as well as functional differences after their incorporation (37,38). In principle, such single-footed glycerolipid moieties could be generated from a more typical mammalian sn-1-alkyl-2-acylglycerol anchor through the action of a phospholipase A 2 which removes sn-2acyl constituents from glycerol. High levels of secretory phospholipase A 2 have indeed been observed in seminal plasma samples of healthy donors as a product of the prostate (39). On the other hand, seminal plasma also contains potent phospholipase A 2 protein inhibitors (40,41), and it is not known whether the enzyme is active in ejaculates.
N-Glycans of Male Genital Tract CD52-The CD52 peptide is among the smallest peptide backbones described for cell membrane proteins. Moreover, it is so diverse in different mammalian species (for review, see Ref. 10) that it may be just a scaffold for the presentation and orientation of the large Nlinked carbohydrate moiety. This carbohydrate, possibly masking the small GPI-anchored peptide, may well be the most important feature of CD52 with respect to possible interactions with other molecules and/or cell surfaces. The various chro-    Table I and Figs. 5 and 6) performed in this study revealed that male genital tract CD52 consists of an extremely complex and heterogeneous mixture of more than 50 different glycoforms (summarized in Fig. 7). The number of different glycoforms described by Treumann et al. (1) for spleenic CD52 preparations is much smaller. While for lymphocyte CD52 only tetraantennary glycan structures were identified (1), triantennary oligosaccharides represented the predominant class of male genital tract CD52. Examples of structures constituting the most complex N-glycans of male genital tract CD52 are given in Figs. 7 and 8. The larger oligosaccharides contained additionally branched outer antenna which were almost completely sialylated, leading to high molecular weight penta-and hexasialylated structures. In addition to their core fucosylation, roughly 10 -15% of carbohydrates were substituted with a peripheral fucose ␣1,3-linked to GlcNAc. The peripherally fucosylated structures detected contained at least one N-acetyllactosamine repeat and the proportion of peripherally fucosylated structures increased with increasing numbers of N-acetyllactosamine repeats. This would implicate the VIM2 motif. Moreover, no NeuAc-Gal-(Fuc)-Glc-NAc-Gal units could be detected during MALDI/TOF-MS of endo-␤-galactosidase digested N-glycans from MonoQ pool M2 and M3 (data not shown) which would have resulted in case of sLeX being present.
Epididymal secretions and seminal plasma are a rich source of carbohydrate-modifying enzymes (for review, see Ref. 42) which have been suggested to be able to extracellularly modify luminal glycoproteins. However, our structural data revealed intact polylactosamine repeats with Ͼ90% terminal sialic acids and would not suggest extensive CD52 carbohydrate truncation or degradation. Numerous extracellular glycosyltransferase activities have also been described in epididymal fluid as well as on the sperm surface (for review, see Ref. 42). These enzymes as well could, in principle, modify the sperm glycocalyx provided that the appropriate sugar donors and acceptors were available. The major maturation-associated antigen of rat spermatozoa has been suggested to represent an acceptor molecule for glycosyltransferases (for review, see Ref. 43). However, from the structural data presented here it seems unlikely that the extensive glycan heterogeneity observed in human seminal plasma CD52 resulted from extracellular enzyme action.
Surface charge differences, most probably attributable to the negative charge of terminal sialic acid residues, and changes in lectin-binding patterns are among the first epididymal modifications of spermatozoa that have been described (Refs. 44 and  (c and d). The oligosaccharides were subjected to MALDI/TOF-MS analysis before (a) and after (b) digestion with endo-␤-galactosidase and, after desialylation, to HPAEC-PAD mapping (c and d). Panel a, MALDI/TOF spectrum of derivatized N-glycans from MonoQ pool M3. First signal: trisialylated triantennary N-glycan (m/z ϭ 3796) followed by molecular ions due to the addition of 1-5 lactosamine repeats (m/z ϭ 4245, 4695, 5144, 5594, and 6044). The lactosamine repeat containing oligosaccharides are accompanied by structures (approximately 10%) with an additional peripheral fucose residue (ϩ174 milliunits). Panel b, M3 after digestion with endo-␤galactosidase. The triantennary structure without lactosamine repeats has remained unchanged (m/z ϭ 3795), whereas lactosamine containing structures have lost 1-3 NeuAc-Gal-GlcNAc-Gal units (m/z ϭ 3230, 2664, and 2099). The appearance of structures with undigested lactosamine repeats at m/z ϭ 4244 (biantennary with two repeats or triantennary with 1 repeat or tetraantennary) accompanied by an analogous series of structures having lost 1-2 repeats at m/z ϭ 3680 and 3114 indicates the presence of oligosaccharides with branched galactose residues, which are resistant to digestion by endo-␤-galactosidase. Panels c and d show the corresponding asialo-Dionex profiles of M3 before (c) and after (d) endo-␤-galactosidase digestion. Arrows indicate the elution positions of the proximally fucosylated triantennary isomers (2,4-or 2,6-branched) confirming the MALDI results indicating the presence of triantennary structures before and after digestion. Additional peaks corresponding to the 2,4-and 2,6-branched triantennary N-glycans with 1-3 lactosamine repeats (3b-3f) disappear to a large degree after treatment with endo-␤galactosidase and give rise to early eluting structures, which have lost 1-3 repeats. Again the presence of some material resistant to digestion can be clearly detected. 45; for review, see Ref. 46). Acquisition by spermatozoa during epididymal transit of the highly sialylated CD52 antigen would explain this. Although we cannot rule out that the population of CD52 glycans on the sperm surface might differ from the seminal plasma fraction analyzed in this study due to possible extracellular modifications, our present knowledge on intracellular glycosyltransferase action as well as lectin and antibody binding studies comparing sperm and seminal plasma CD52 (see Fig. 1) would not support this.
Possible Function(s) of CD52 Glycans-The highly sialylated polylactosamine-containing carbohydrate chains of CD52 are likely to form a negatively charged, dense glycocalyx adjacent to the sperm plasma membrane and may help to prevent sperm self-agglutination and/or unspecific binding to the genital tract epithelium during transport and storage. Other roles ascribed to sperm surface carbohydrates are a stimulation of the production of naturally occurring anti-sperm antibodies on the one hand and the "masking" of intrinsic protein antigens of the sperm membrane on the other. Sperm surface glycans may serve as "traitors" in generating an immune response and an infertile state, but may also serve in a protective or stabilizing role to prevent premature loss of the acrosome content (for review, see Ref. 47). Presently, it seems unlikely that they directly participate in gamete binding, although it cannot be excluded (compare Ref. 48).
The clinically most important role ascribed to sperm surface carbohydrates is that of antigens stimulating the production of naturally occurring anti-sperm antibodies. Only a limited number of glycans, mainly of glycolipid-like substances, seem to be immunodominant (49,50). Several monoclonal antibodies to the CD52 type of glycans have been found to bind to spermatozoa and inhibit fertilization, including H6 -3C4 which was derived from an infertile woman with sperm immobilizing antibodies (49,51,52). Although only little information is available concerning the carbohydrate epitopes involved on the sperm surface, terminally sialylated neolactosamines, either unbranched (i-antigen) or branched (I-antigen) are reactive  Table I). Suggested structures represent one example of each of the five charge groups obtained (M2-M6). Oligosaccharides carry varying numbers of N-acetyllactosamine repeats either unbranched or branched (numbers of repeats ranging from n ϭ 0 to n ϭ 4 in group M2, from n ϭ 0 to n ϭ 5 in groups M3 and M4/M4*, from n ϭ 0 to n ϭ 4 in M5/M5*, and from n ϭ 0 to n ϭ 3 in M6. More than 80% of NeuAc are ␣2Ϫ3-linked, ␣2Ϫ6-linked NeuAc present in a minority of structures are not indicated to enhance clarity. Linkage and branching positions of N-acetyllactosamine repeats were not determined. Polylactosamine stretches are represented as white beams. Lymphocyte CD52 contains only tetraantennary oligosaccharides, approximately 35% of which may be branched (1). The core structure of all N-glycans (bold black lines) is fucosylated (core fucosylation not displayed); additional peripheral fucose residues present in 10 -15% of male genital tract CD52 are drawn as triangles. Less than 8% of free terminal galactose in male genital tract CD52 suggested near complete sialylation, the majority of sialic acid residues (open circles) being ␣2,3-linked. While for lymphocyte CD52 the total amount of terminal sialic acid was not given, most of the N-glycans seemed to contain ␣2Ϫ6-linked sialic acid (1). In the lipid anchors, the ethanolamine phosphate substitution (PO 4 -EtNH 2 ) of the mannose core is similar in both modifications, in lymphocyte CD52 an additional mannose (Man) residue may be present. The fatty acid residues, however, are completely divergent between male genital tract and lymphocyte CD52 (see text). with this type of antibody (52).
To identify antigens for immunocontraceptive development, a panel of monoclonal antibodies was generated against the human sperm surface by various research groups. S19, a mAb from this panel was found to strongly agglutinate human spermatozoa, impede sperm penetration of cervical mucus, inhibit gamete interaction (53), and also inhibit sperm binding activity of the H6 -3C4 mAb (49). Periodate treatment abolished S19 immunoreactivity, suggesting that the S19 mAb reacted with a carbohydrate epitope. The cognate antigen, named sperm agglutination antigen-1 has similarities to the CD52 antigen: it is described as a highly acidic, hydrophobic, and polymorphic sperm glycoprotein with an apparent molecular mass of Ϸ15-25 kDa and a pI of 2.5 to 3.0; its solubility and phase partitioning characteristics suggested that it is GPI-anchored (54). Our present study demonstrates that an exhaustive structural study of the epitope(s) involved would be advisable before embarking on a contraceptive immunization trial with sperm agglutination antigen-1 or any of the other antibodies directed against CD52 on the sperm surface.