Isomeric Separation and Recognition of Anionic and Zwitterionic N-glycans from Royal Jelly Glycoproteins*

Royal jelly has special properties as the sole food for queen honeybees, but is also considered to be bioactive in other animals. Using an off-line LC-MALDI-TOF MS approach, we reveal a highly complex N-glycome with novel anionic and zwitterionic N-glycans being present on specific royal jelly glycoproteins. These glycans are shown in microarray format to bind protein ligands, revealing that there exists potential for interactions with the human innate immune system. Graphical Abstract Highlights The N-glycome of honeybee royal jelly is more diverse than previously thought Antennal phosphoethanolamine and glucuronic acid are among the special features Anionic/zwitterionic glycans are recognized by a pentraxin and anti-HNK-1 antibody Royal jelly has received attention because of its necessity for the development of queen honeybees as well as claims of benefits on human health; this product of the hypopharyngeal glands of worker bees contains a large number of proteins, some of which have been claimed to have various biological effects only in their glycosylated state. However, although there have been glycomic and glycoproteomic analyses in the past, none of the glycan structures previously defined would appear to have potential to trigger specific biological functions. In the current study, whole royal jelly as well as single protein bands were subject to off-line LC-MALDI-TOF MS glycomic analyses, complemented by permethylation, Western blotting and arraying data. Similarly to recent in-depth studies on other insect species, previously overlooked glucuronic acid termini, sulfation of mannose residues and core β-mannosylation of the N-glycans were found; additionally, a relatively rare zwitterionic modification with phosphoethanolamine is present, in contrast to the phosphorylcholine occurring in lepidopteran species. Indicative of tissue-specific remodelling of glycans in the Golgi apparatus of hypopharyngeal gland cells, only a low amount of fucosylated or paucimannosidic glycans were detected as compared with other insect samples or even bee venom. The unusual modifications of hybrid and multiantennary structures defined here may not only have a physiological role in honeybee development, but represent epitopes recognized by pentraxins with roles in animal innate immunity.


Isomeric Separation and Recognition of Anionic and Zwitterionic N-glycans from Royal
Jelly Glycoproteins* □ S Royal jelly is a product of the hypopharyngeal and mandibular glands of worker honeybees (1); its natural role is as the food source for worker and drone larvae for the first 3 days but induces queen development when fed beyond that time and remains, thereafter, the food for a queen's entire life. The observed fertility and longer lifespan of queens, as compared with worker bees, has been correlated with this special diet. However, royal jelly has attracted especial attention because of its putative effects on human health. Indeed, it has been used since ancient times as a food supplement or cosmetic and is considered to have a range of antibiotic, anti-inflammatory and other health-relevant effects (2), although these are controversial and allergic reactions are also known (3). Royal jelly contains a wide range of compounds including carbohydrates, lipids, vitamins, and proteins, many of which are post-translationally modified (4,5). Because of their multiple and variable monosaccharide units and their size (typically 2000 Da), N-glycans represent the most diverse and complex set of protein modifications known (6) and modulate a wide range of biological processes.
Based on proteomic and glycoproteomic studies of royal jelly, up to 100 different proteins have been identified including the "major royal jelly proteins" (MRJPs) 1 . Of these, MRJP1 (also known as apisin or royalactin) has a controversial role in queen development (7,8), but is reported to extend life-span in other invertebrates (9). Furthermore, MRJP1 and MRJP2 (apalbumin2) have been shown to have antihypertensive and antibiotic activity in their glycosylated state (10). Although the sites of glycosylation on MRJPs have been analyzed, there is less information regarding the structures of the glycans on the specific glycoproteins. Although mass spectrometric analyses (11) indicated potentially complex N-glycosylation of MRJP2, HPLC and NMR data showed the presence of oligomannosidic and hybrid glycans (12,13) on monomeric and oligomeric forms of MRJP1. There have also been a series of glycan analyses on royal jelly, culminating in a report that Gal␤1,3GalNAc␤1,4GlcNAc units are detectable on N-glycans and that there is a relevant galactosyltransferase activity (14 -17). However, none of the hybrid or biantennary glycans described were especially unusual and so it was concluded HPLC Fractionation-Complete pyridylaminated N-glycomes (10% of the neutral and 90% of the anionic pools) were fractionated by reversed-phase HPLC (Ascentis Express RP-amide; 150 ϫ 4.6 mm, 2.7 m; Sigma-Aldrich) and a gradient of 30% (v/v) methanol (buffer B) in 100 mM ammonium acetate, pH 4 (buffer A) was applied at a flow rate of 0.8 ml/min as follows: 0 -4 min, 0% B; 4 -14 min, 0 -5% B; 14 -24 min, 5-15% B; 24 -34 min, 15-35% B; 34 -35 min, return to starting conditions (20). The RP-HPLC column was calibrated daily in terms of glucose units using a pyridylaminated dextran hydrolysate and the degree of polymerization of single standards was verified by MALDI-TOF MS. Selected RP-HPLC fractions were subject to HIAX-HPLC as a second dimension as described (20) with an IonPac AS11 column (Dionex, Sunnyvale; 4 ϫ 250 mm, combined with a 4 ϫ 50 mm guard column). A two solvent gradient with buffer A (0.8 M ammonium acetate, pH 3) and buffer B (80% acetonitrile; LC-MS grade) was applied at a flow rate of 1 ml/min: 0 -5 min, 99% B; 5-50 min, 90% B; 50 -65 min, 80% B; 65-85 min, 75% B. Detection for both columns was by fluorescence (Shimadzu RF-20A XS detector; excitation/emission at 320/400 nm). All manually collected HPLC glycan fractions were lyophilized, redissolved in water and analyzed by MALDI-TOF MS and MS/MS.
Glycan Mass Spectrometry-Monoisotopic MALDI-TOF MS was performed using an Autoflex Speed (Bruker Daltonics, Bremen, Germany) instrument in either positive or negative reflectron modes with 6-aza-2-thiothymine (ATT; Alfa-Aesar, Thermo Scientific) as matrix. MS/MS was in general performed by laser-induced dissociation of the [MϩH] ϩ or [M-H] Ϫ pseudomolecular ions (except for permethylated structures analyzed as [MϩNa] ϩ with 2,5-dihydroxybenzoic acid as matrix); typically 2000 shots were summed for MS (reflector voltage, lens voltage and gain respectively 27 kV, 9 kV and 2217 V) and 4000 for MS/MS (reflector voltage, lift voltage and gain respectively 27 kV, 19 kV and 2174 V). Spectra were processed with the manufacturer's software (Bruker Flexanalysis 3.3.80) using the SNAP algorithm with a signal/noise threshold of 6 for MS (unsmoothed) and 3 for MS/MS (four-times smoothed). Glycan spectra were manually interpreted based on the masses of the predicted component monosaccharides, differences of mass in glycan series, fragmentation patterns, comparison with coeluting structures from other insects and nematodes and chemical treatments or exoglycosidase digestions (see also pages S2-S4 of the Supplement). Negative-ion mode LC-MS n of a 2D-HPLC-enriched glycan was performed as previously described, using a 5 m porous graphitized carbon column (10 cm ϫ 150 m) coupled to a Thermo Scientific LTQ ion trap mass spectrometer (22); refer also to page S5 of the Supplement for further details. A list of theoretical m/z values for each glycan composition is presented in the supplemental Table S1; mzXML files of raw MS/MS data are available as supplementary information.
Enzymatic and Chemical Treatments-Glycans were treated, before re-analysis by MALDI-TOF MS, with ␣-fucosidase (bovine kidney from Sigma-Aldrich), ␣-mannosidases (jack bean from Sigma-Aldrich, Aspergillus ␣1,2-specific from Prozyme, Hayward, CA, and Xanthomonas ␣1,2/3-specific from New England Biolabs), ␤-galactosidase (␤1,3-specific from New England Biolabs), ␤-glucuronidases (E. coli from Megazyme, Bray, Ireland, and Helix pomatia from Sigma-Aldrich; desalted and concentrated before use), ␤-N-acetylhexosaminidases (jack bean from Sigma-Aldrich, Xanthomonas ␤1,2specific N-acetylglucosaminidase from New England Biolabs, Streptomyces ␤1,3/4-specific N-acetylhexosaminidase (chitinase) from New England Biolabs or in-house produced recombinant forms of Caenorhabditis elegans HEX-4 specific for ␤1,4-linked GalNAc residues or Apis mellifera FDL specific for the product of GlcNActransferase I (23)) in 50 mM ammonium acetate, pH 5, at 37°C overnight (except for pH 6.5 in the case of HEX-4, pH 7 in the case of E. coli ␤-glucuronidase or an incubation time of only 3 h in the case of 1 The abbreviations used are: MRJP, major royal jelly proteins. FDL); these incubations were performed in PCR tubes with a final volume of 3 l (for further details about conditions and specificities, refer to the supplement). Hydrofluoric acid was used for removal of phosphoethanolamine or ␣1,3-linked fucose (20). As appropriate, treated glycans were re-chromatographed by RP-HPLC to ascertain retention time shifts before MALDI-TOF-MS; otherwise, an aliquot (generally one-fifth) of any digest was analyzed by MALDI-TOF-MS without further purification.
Protein-specific Glycan Analysis-After performing SDS-PAGE (10 g/lane) and staining with Coomassie Blue, protein bands were excised and washed/destained (twice with 50% acetonitrile in water and successively once with 1:1 0.1 M ammonium bicarbonate/acetonitrile and 100% acetonitrile only) before drying (26). The proteins in the gel pieces were reduced for one hour at 56°C with 10 mM dithiothreitol and alkylated for 45 min with 55 mM iodoacetamide in the dark. After a second round of washing (as above), the gel pieces were dried, covered with a 1:2 mixture of ammonium bicarbonate/trypsin (100 ng/l) and incubated at 37°C overnight. The (glyco)peptides were extracted with a mixture of acetonitrile/water/trifluoroacetic acid (660/ 330/1 (v/v/v)), dried and dissolved in water before MALDI-TOF MS using either ␣-cyanocinnamic acid (ACH) or 6-aza-thiothymine (ATT) as matrices, using similar mass spectrometer settings as above. Tryptic peptide fingerprint data were analyzed using the Mascot webserver (version 2.6.0) and the Swissprot database (release 2017_06) as described on page S5 of the Supplement.
For protein-specific glycan analysis, the (glyco)peptides were first heat-treated to inactivate trypsin and 90% of the samples were subject to PNGase Ar treatment (2.5 U; New England Biolabs) in 20 mM ammonium acetate, pH 5 overnight at 37°C. Whereas the deglycosylated peptides were analyzed by MALDI-TOF MS, the released N-glycans were purified using two different columns packed respectively with Lichroprep C18/Dowex AG50 and nonporous graphitized carbon/Lichroprep C18. The C18/AG50 column was washed with 2% acetic acid and 60% isopropanol; after equilibration of the column with 2% acetic acid, the N-glycans were acidified with 10% acetic acid before application. The N-glycans were eluted (three column volumes of 2% acetic acid) before use of the carbon/C18 column, which was pre-washed with 100% acetonitrile and equilibrated with water. The N-glycans (combined neutral and anionic pools) were then eluted with 40% acetonitrile containing 0.1% trifluoro-acetic acid. After drying, the glycans were fluorescently labeled by reductive amination using 2-aminopyridine and then analyzed with MALDI-TOF MS and HPLC as above.

RESULTS
Strategy for the N-glycan Analysis-Analysis of glycanbased post-translational modifications of proteins remains a challenge and experience has shown that a thorough analysis of unknown invertebrate glycomes requires multiple fractionation steps in order to prevent suppression effects and to enable separation of isomeric/isobaric structures of the same or similar mass (29). Thus, we first separated the free Nglycans of honeybee royal jelly into "neutral" and "anionic" pools, before fluorescent labeling, fractionation on an RPamide HPLC column (calibrated in terms of glucose units) and MALDI-TOF-MS/MS in combination with chemical and enzymatic treatments (20). Based on the fluorescence intensities of the two glycan pools, it is estimated that at least 3% of the N-glycans are "anionically" modified.
The initial screen, before HPLC, of the pyridylaminated pools by MALDI-TOF MS indicated that the neutral pool contained, in accordance with previous studies (14), a range of oligomannosidic and potentially hybrid or complex structures as judged by mass (supplemental Fig. S1). On the other hand, MALDI-TOF MS of the anionic pool suggested the presence of glycans potentially modified with sulfate or glucuronic acid, neither of which have been reported for royal jelly glycans, but of a type familiar from our recent data on dipteran and lepidopteran N-glycomes (18,19). However, unusually for Nglycans from an insect source, fucosylation was not obvious from the complete spectra despite using both PNGase F and PNGase A in series to release the glycans. Relatively low amounts of core mono-and difucosylation were found only upon HPLC analysis of the glycans; also, paucimannosidic glycans (Man 2-4 GlcNAc 2 ), generally dominant in insect glycomes (18,30), were present at only a low level.
The subsequent RP-amide fractionation was necessary to resolve the fine detail of the N-glycome of royal jelly; nearly 70 structures were detected in the neutral pools (see Fig. 1) with a partial overlap with the over 80 structures in the various anionic-enriched pools; in total, over 100 different compositions could be verified (see the supplemental Table S1 for the theoretical m/z values). Considering a widely held belief that MALDI-TOF-MS of nonpermethylated glycans leads to "in source" artifacts, the major neutral RP-HPLC peaks were subject to both NP-HPLC (i.e. HIAX; Fig. 2) and permethylation; also, aliquots of the whole N-glycome were permethylated (supplemental Figs. S2 and S3). The results of these additional procedures confirmed the validity of our general approach. Furthermore, the analysis was extended to the individual glycoproteins by using Western blotting and tryptic peptide mapping as well as release of the N-glycans from the single protein bands followed by fluorescent labeling and HPLC fractionation.
Oligomannosidic N-glycans-In two independent preparations, the most abundant N-glycans in royal jelly were found to be, according to mass and retention time data ( Fig. 1 and supplemental Figs. S1 and S2), either the typical "Golgi-processed" isomer of Man 5 GlcNAc 2 or Man 9 GlcNAc 2 (respec-FIG. 1. RP-HPLC fractionation of the neutral N-glycans from royal jelly. Glycans were released by PNGase F and A before solid phase extraction (to separate the neutral and anionic glycans), pyridylamination and finally fractionation on an RP-amide column (refer to supplemental Fig. S1 for overall MALDI-TOF MS profiles). The chromatogram is annotated with glycan structures (each with the positive mode m/z value for the protonated ion in order of abundance for each fraction) according to the Symbolic Nomenclature for Glycans (circles, galactose or mannose; squares, GalNAc or GlcNAc; triangles, fucose; PE, phosphoethanolamine; see also inset for symbols and linkage information). The HPLC column was calibrated in terms of glucose units (g.u.). Oligomannosidic glycans are annotated by comparison to previous studies on insect glycomes using the same column (unusual isomers were also subject to specific ␣-mannosidase digestion), whereas other glycans are defined based on MS/MS and digest data. The PNGase F alone and PNGase Ar (after F) digests of the second preparation yielded qualitatively similar chromatograms, but Man 9 GlcNAc 2 was more dominant than Man 5 GlcNAc 2 (see also supplemental Fig. S1A and S1B). The hymenoptera-specific zwitterionic, ␤-mannosylated and triantennary glycans are highlighted in red, green or blue boxes; the presence of phosphoethanolamine (rather than phosphorylcholine) and of ␤-mannose, but the relative lack of fucose in this specific glycome, contrasts to the larval N-glycomes of lepidopteran species (19).

FIG. 2. Isomeric separation of hybrid N-glycans by two-dimensional HPLC.
The major RP-HPLC fraction (7.2 g.u.) was analyzed by MALDI-TOF MS (A) and subfractionated by HIAX-HPLC (B); thereby, the HIAX column resolves glycans not just based on mass, but even with a different isomeric composition of mannose and galactose residues. This fractionation results in peaks containing single glycans with no evidence of in source MS fragmentation with even the fluorescent intensities correlating to those for the MALDI-TOF MS data. The pairs of isomeric glycans (Hex 4 -5 HexNAc 4 with m/z 1557 and 1719; C/F and I/L) display contrasting susceptibilities to ␤1,3-specific galactosidase tively eluting at 7.2 or 5.5 g.u. with m/z 1313 or 1961). The MALDI-TOF MS analysis of individual HPLC fractions revealed the significant presence of several other oligomannosidic structures (Man 6 -8 GlcNAc 2 ) as well as one other isomer of Man 5 GlcNAc 2 (6.4 g.u.) and low amounts of paucimannosidic glycans (Man 2-4 GlcNAc 2 ). In total, around 20 distinct pauci-and oligomannosidic isomers can be proposed based on MS/MS data, comparisons to retention times of such glycans from other species (19,31) and selected ␣-mannosidase digests (see, e.g. supplemental Fig. S4). Comparison of the MALDI-TOF MS data on the major RP-amide HPLC fraction with HIAX size fractionation (Fig. 2) and permethylation (data not shown) prove that no "in source" fragmentation of oligomannosidic N-glycans occurs in our MALDI-TOF MS analyses.
Neutral Hybrid and Complex N-glycans-About 50 of the proposed royal jelly neutral glycans were predicted on the basis of mass to contain three or more N-acetylhexosamine residues; the initial mass spectrometric screen of the neutral pool revealed a rich set of hybrid and complex structures ranging from Hex 3 HexNAc 3 (m/z 1192) through to To resolve the individual neutral glycans in terms of the modifications and isomeric status, selected HPLC fractions were treated in a targeted manner with a number of exoglycosidases. For glycans containing four or five hexose residues, a major question was as to whether these hybrid structures were terminally galactosylated as previously suggested (14). However, isomers of the same observed masses (e.g. two isomers each of Hex 4 -5 HexNAc 4 with m/z 1557 and 1719) were identified by their different 2D-HPLC elution and MS/MS properties (Fig. 2). Man 4 -5 -based hybrid glycans lost maximally three residues with jack bean ␣-mannosidase (dependent on the efficiency of removal of the "core" ␣1,6-mannose), whereas the pattern of ␤1,3-specific galactosidase and HEX-4 ␤1,4-specific N-acetylgalactosaminidase (23) sensitivity depended on the isomer ( Fig. 2 and supplemental Fig. S4).
Successful galactosidase digestion correlated with the distinct presence of an MS/MS fragment at m/z 569, whereas HEX-4 sensitivity was observed for glycans with fragments at m/z 407 (Fig. 2C-2R). The presence of either a fourth/fifth mannose or a galactose substitution of antennal GalNAc was corroborated by MS/MS of permethylated glycans showing Hex 4 -5 HexNAc 2 Y-fragments at m/z 1361 and 1565 or Hex 0 -1 HexNAc 1-2 B-fragments at m/z 486, 527 and 731 (supplemental Fig. S3A and S3B). For those hybrid glycans containing a sixth mannose (one isomer each of m/z 1881 and 2043; 6.4 g.u.), four mannose residues were lost upon jack bean mannosidase digestion, whereas a single one was removed by an ␣1,2-specific mannosidase (data not shown).
In the case of biantennary glycans, the definition of which antenna was elongated was based on the retention time (isomers with longer lower arms, i.e. on the ␣1,3-mannose, elute earlier (32)), fragmentation pattern and enzyme sensitivity (supplemental Fig. S5), whereby the FDL N-acetylglucosaminidase is specific for removal of unsubstituted ␤1,2linked GlcNAc from the lower arm (23). As for the hybrid glycans, the longer antennae (with m/z 569 B-fragments for, e.g. isomers of m/z 1760) were digested by a combination of ␤1,3-galactosidase and HEX-4 as for the hybrid forms, whereas glycans with a terminal GalNAc were sensitive to HEX-4 alone (with m/z 407 B-fragments as for isomers of m/z 1598). The cumulative data show that LacdiNAc (GalNAc␤1, 4GlcNAc), with or without a ␤1,3-linked galactose cap, is a frequent motif of royal jelly N-glycans and that galactose was present on HexdiNAc units.
In the case of triantennary glycans eluting at 11 g.u. (see Fig. 1), a combination of galactosidase and specific hexosaminidase digests together with reinjection onto the HPLC column revealed that there could be a "free" GlcNAc on the upper ␣1,6-mannose and two branches on the lower ␣1,3mannose (supplemental Fig. S6A-S6E). On treatment with Xanthomonas ␤1,2-specific N-acetylglucosaminidase, removal of the GlcNAc on the ␣1,6-arm resulted in conversion of the m/z 1030 Y-fragment to one at m/z 827; the third antenna is based on a ␤1,4-linked GlcNAc as shown by co-elution (9.5 g.u.) of the m/z 1598 galactosidase/HEX-4 digestion product of the 11 g.u. fraction with a degalactosylated bovine fetuin N-glycan. Later eluting glycans (12 g.u.) having two or three fully galactosylated antennae displayed intense MS/MS fragment ions at m/z 1233 or 1395 resulting from loss of the "heavy" ␣1,3-arm as indicated (supplemental Fig. S6F and S6G). All defined hybrid, bi-and triantennary structures were the basis for various anionic/zwitterionicmodified glycans subsequently found in the anionic pool.
Core Mannosylated N-glycans-A hallmark of many invertebrate N-glycans, when pyridylaminated, is the presence of an m/z 446 (Fuc 1 GlcNAc 1 -PA) fragment indicative of core fucosylation (see below); recently, in an oyster and a marine snail (22,33) we have found glycans with an m/z 462 fragment (Hex 1 HexNAc 1 -PA). In the snail this fragment was abolished on ␤-mannosidase digestion of relevant glycans; this rules out that this is a Y-fragment of an endoglycosidase digestion product, which would elute before 5 g.u. rather than at ϳ8 g.u.
In the present study, the m/z 462 fragment was also found for the first time in an insect species and could also be verified by the MS/MS pattern of 2D-HPLC-purified glycans followed by glycosidase digestion of selected structures including one carrying antennal glucuronic acid ( Fig. 3 and supplemental  Fig. S7). Initially, we presumed that the ␤-mannose in royal jelly would also be 1,3-linked to the proximal GlcNAc as in molluscs (22), but parallel re-injection experiments with two RP-HPLC columns (supplemental Fig. S7A and S7B) as well as the LC-MS n fragmentation pattern ( Fig. 3C and 3D) indicated that the royal jelly m/z 1557 mannosylated structure was not identical to that from the snail; thus, we conclude that the ␤-mannose in royal jelly glycans is 1,6-linked.
Zwitterionic N-glycans-On analysis of the individual "neutral" fractions, it was observed that some glycans could be detected in both positive and negative ion modes and had compositions suggestive of a modification with a moiety of 123 Da, which is reminiscent of the presence of phosphoethanolamine on N-glycans from Trichomonas vaginalis or Penicillium species (34,35). The most distinct fragment ions ( Fig. 4 and supplemental Fig. S8 and S9) were previously unknown ones at m/z 1153, 1315, 1477, 1518 and 1639 in positive mode (Hex 2-5 HexNAc 3-4 PE 1 -PA Y-fragments) or at m/z 528/530 and 690/692 in negative/positive modes (Hex 0 -1 HexNAc 2 PE 1 B-fragments, which reflect the antennal capping by LacdiNAc with or without galactose). To resolve isomers, HIAX separation of the RP-HPLC 6.6 -6.8 g.u. was performed (supplemental Fig. S8). Thereafter, it was found that the Hex 4 -5 HexNAc 4 PE 1 isomers (m/z 1678/1680 and 1840/1842) display different MS/MS spectra because of variations in the positions of galactose and mannose residues, similarly to the "parental" later-eluting nonzwitterionic glycans (compare with Fig. 2). The position of the phosphoethanolamine on "internal" GlcNAc, rather than on GalNAc, became obvious on removal of antennal galactose and GalNAc residues as judged by the alterations in the B-fragments, i.e. conversion to m/z 325/327 or 528/530 (Fig. 4A-4F and supplemental Fig. S9).
Further proof for the presence of the phosphodiester was provided by sensitivity toward hydrofluoric acid (Fig. 4H-4K) and a subsequent shift to higher RP-HPLC retention time correlating with the loss of the hydrophilic moiety (Fig. 4G). Thereby, rather than the phosphorylcholine substitution found on N-glycans from two lepidopteran species (19), we conclude that several hybrid and complex glycans in royal jelly are modified with phosphoethanolamine, a moiety which also decorates in 6-linkage internal GlcNAc residues of dipteran glycolipids (36 -38). Despite an in-depth examination and the ability to find phosphoethanolamine-modified glycans when permethylating Penicillium glycomes (unpublished data) which display a 30% degree of such structures, no zwitterionic glycans were found on permethylation of the royal jelly glycome, which may be because of losses of these lowabundance structures (Ͻ0.5%) during the chemical or sepa-  procedures; thus, the off-line LC-MALDI approach was necessary to detect this class of glycans, which were not detected in any spectrum of an entire glycome (supplemental Figs. S1 and S2).
Glucuronylated N-glycans-One of the key steps in our analyses to reveal the fine details of royal jelly glycans was the separation on graphitized carbon of the neutral from the anionic species. Thus, after pyridylamination, the anionic pools (ϳ3% of the total) obtained from different preparations were separately applied to the RP-amide HPLC column (Fig. 5); obvious was the presence of glycans modified with glucuronic acid, which (as in our previous studies of insect N-glycans (18,19)) could be detected in both positive and negative ion modes, but fragment best in the positive mode to yield MS/MS fragments at, e.g. m/z 542 and 745 (HexA 1 Hex 1 HexNAc 1-2 ; see Figs. 6-8 as well as supplemental Figs. S6H-S6K and S10 -S12). On this basis, mono-, diand triantennary glycans with up to three glucuronic acid residues were identified in the PNGase F-released anionic pool (Hex 3-6 HexNAc 4 -8 Fuc 0 -1 GlcA 1-3 ; Fig. 5A). Glucuronylated glycans were also found on permethylation of the anionic pool and displayed relevant HexA 1 Hex 1 HexNAc 0 -2 fragments at m/z 476, 704 and 949 indicative of linear antennae (supplemental Figs. S2C and S3F).
Verification that the 176 Da modification is glucuronic acid rather than methylhexose was also offered by sensitivity to ␤-glucuronidases of three HPLC fractions containing monoglucuronylated hybrid and biantennary glycans; this treatment resulted in a shift to later RP-amide retention times (typically an increase of about 1 g.u.; see chromatograms in  Fig. S10A-S10L). As deglucuronylation resulted in retention times and MS/MS spectra akin to those for glycans found in the neutral pools, it is assumed that the structures with antennal Gal␤1, 3GalNAc␤1,4GlcNAc underlie the glucuronylated forms.
Glucuronylated N-glycans Modified With Phosphoethanolamine-A further set of modifications is represented by glucuronylated structures also carrying phosphoethanolamine present in the anionic pool. Such glycans elute one HPLC fraction earlier than the "parent" structure lacking the zwitterion; hydrofluoric acid treatment to release the phosphoethanolamine results in a shift backwards in terms of retention time (Fig. 6A). This treatment also results in altera-tions in the MS/MS spectra (i.e. losses of negative or positive mode HexA 1 Hex 1 HexNAc 2 PE 1 B-fragments at m/z 866/868 and appearance of positive ones at m/z 745; Fig. 6B-6G and supplemental Fig. S11A-S11D) indicative of removal of phosphoethanolamine (123 Da) from glucuronylated antennae. As judged by re-application onto HPLC, phosphoethanolaminecontaining structures account for some 5% of the anionic pool.
Sulfated N-glycans-In previous studies on dipteran and lepidopteran glycomes, the sulfated N-glycome within the anionic pool was dominated by fucosylated structures such as Hex 3 HexNAc 2-3 Fuc 1 S 1 (m/z 1213 and 1416); these were not present in royal jelly. Rather nonfucosylated Hex 3 HexNAc 3-4 S 1 (m/z 1270 and 1473; see Fig. 5) were the major sulfated glycans detected, reflecting the general lack of fucose in royal jelly. Serial digestions of larger structures (Hex 3-4 HexNAc 4 -5 S 1 ) with ␤1,3-galactosidase and jack bean ␤-hexosaminidase removed the antennae; as only one underlying mannose was sensitive to both the unspecific jack bean ␣-mannosidase and ␣1,3-specific mannosidase, it was possible to define the ␣1,6-mannose as being sulfated (Fig.  7A-7F). Confirming our assignments were the typical negative mode fragments at m/z 606, 768 and 971 (Hex 2-3 HexNAc 1-2 S 1 ), as seen for other cases of sulfation of mannose residues, including glycans of the same elution time previously identified in lepidopteran species (18,19). Finally, on using solid phase extraction following permethylation of the anionic pool, it was also possible to detect certain sulfated glycans; MS/MS thereof supported the conclusion that the ␣1,6-linked mannose is sulfated (supplemental Figs. S2B and S3C). Otherwise, sulfated forms of the fucosylated and glucuronylated glycans were only observed in the off-line LC-MS analysis (see below).
Core Mono-and Di-fucosylated Anionic N-glycans-In the combined PNGase A and F-released neutral pools or when PNGase F alone was employed, only traces of a variety of singly fucosylated glycans were detected (Fig. 1); however, most fucosylated glycans were glucuronylated and therefore present in the anionic pool (Fig. 5), including the largest structures observed in this study, i.e. triantennary monofucosylated forms without and with sulfate (m/z 3366 and 3446 as [M-H] Ϫ ). The presence of a core ␣1,6-fucose on these can be inferred from their release by PNGase F and late elution time as well as by the m/z 446 Y-fragment absent after bovine ␣-fucosidase treatment (supplemental Fig. S12F-S12J and S12K-S12N). However, when the glycans were released from (circles, galactose or mannose; diamonds, glucuronic acid; squares, GalNAc or GlcNAc; triangles, fucose; PE, phosphoethanolamine; S, sulfate; see inset for symbol and linkage information). The HPLC column was calibrated in terms of glucose units (g.u.). The chromatogram A (of the PNGase F/A-released anionic glycans) is annotated with all structures determined in that pool, whereas chromatograms B and C (i.e. the PNGase Ar and PNGase F released anionic glycans from a second preparation) highlight the fucosylated glycans (the elution positions of other detected glycans being indicated by the m/z values alone); in the case of PNGase Ar release, the low abundance fucosylated structures can be core difucosylated and even carry one Lewis-like antennal fucose, whereas PNGase F can only release glycans with core ␣1,6and up to two antennal (␣1,3-) fucose residues. the second sample of royal jelly using PNGase Ar, we noted traces of glycans with two or three fucose residues in the anionic pool (Fig. 5B). A key Y-fragment of m/z 592 (Fuc 2 GlcNAc 1 -PA; Fig. 8 and supplemental Figs. S11E and S11H and S12A-S12E), as well as a degree of anti-horseradish peroxidase staining (supplemental Fig. S13A), indicates corresponding to the HF products of the 6.0 g.u. fraction, is also indicated; see supplemental Figs. 10 and 11 for further MS and HPLC data on glucuronidase and HF digests. B-G, The MALDI-TOF MS/MS of untreated glycans in negative (B and E) and positive modes (C and F) and of the HF-treated products (positive mode only; D and G) highlight changes in the fragmentation patterns because of removal of phosphoethanolamine residues (loss of m/z 868 fragments correlating to the later retention times; indicated with red arrows). (H-K) Positive mode MALDI-TOF MS/MS of the 6.4 and 7.2 g.u. glucuronylated glycans before and after ␤-glucuronidase digestion indicating loss of the m/z 542 and 745 fragments and appearance of ones at m/z 366 and 569. For space reasons, the phosphoethanolamine group is depicted underneath the antenna, although a 6-linkage is assumed based on analogy to insect glycolipids (36 -38); the greyscale structures indicate the original elution times, those in color the products'. Based on these analyses, glycans carrying both phosphoethanolamine and glucuronic acid are estimated to represent 30% of these particular anionic fractions. It is to be noted that sulfated glycans (e.g. m/z 1473) are resistant to hydrofluoric acid treatment, unlike those modified with phosphomono-or diesters.

FIG. 8. HPLC and MS/MS analysis of N-glycans modified with both glucuronic acid and fucose residues.
A, RP-HPLC of three hydrofluoric acid treated anionic fractions (originally eluting at 4.6, 5.8 and 7.5 g.u.) highlighting the shifts to higher retention times because of the loss of core and antennal ␣1,3-fucose and of phosphoethanolamine (either -F, -2F or -PE; respective shifts of 2, 1 and 0.5 g.u. for the removal of core ␣1,3-fucose, antennal ␣1,3-fucose and phosphoethanolamine); the loss of antennal ␣1,3-fucose or phosphoethanolamine is more efficient, whereas the loss of the core ␣1,3-fucose is partial for the employed hydrolysis time. The greyscale structures indicate the original elution times, those in color the products'; note that sulfated difucosylated glycans merely lose a fucose, but not the sulfate, as previously seen for the same structures isolated from High Five cells (19). that there are core ␣1,3/␣1,6-difucosylated N-glycans; these glycans elute earlier than those carrying a single core ␣1,6fucose. The third fucose residue is present on the antennae as seen, e.g. on bee venom glycoproteins (39), and is defined by Y-fragments such as m/z 1630, 1792 and 1954 (Hex 3-5 HexNAc 3 Fuc-PA) or losses of 891 mass units from the parent; hydrofluoric acid treatments and relevant shifts to later elution time or alterations in MS/MS spectra confirmed the presence of core and/or antennal ␣1,3-fucose ( Fig. 8 and Supplemental Fig. S11E-S11J). Another isomeric trifucosylated variation, because of a single fucose on the core and two antennal fucoses, was found in the PNGase F-released pool (compare the two glycans of m/z 2916; supplemental Fig.  S12E and S12I). Some of the fucosylated/glucuronylated glycans even contained phosphoethanolamine (Fig. 8E and supplemental Fig. S12B and S12G).
Rare, but detectable, were fucosylated glycans containing both glucuronic acid and sulfate observed as [M-H] Ϫ ions in negative mode and [M-SO 3 ϩH] ϩ in positive mode (Fig. 7G-7J); here also the fragmentation patterns indicate the sulfation of an ␣-mannose, as serial losses of an glucuronic acid antenna and then of sulfate are observed in negative mode MS/MS spectra, whereas the positive mode MS/MS show the nature of the "backbone" core fucosylated structure (Fig.  7K-7N). Smaller sulfated difucosylated glycans akin to those in High Five cells (19) were also detected (m/z 1359 and 1562; Fig. 8A) and one fucose could be removed from these with hydrofluoric acid; as this treatment is also known to remove phosphoesters (29) but not sulfate (40), the resistance of the 80 Da moiety is another confirmation that these glycans are sulfated.
Analysis of Protein-specific Glycosylation and Binding to Glycan Arrays-To test whether there are variations in the glycosylation of individual glycoproteins in royal jelly, SDS-PAGE was performed before either Western blotting or release of glycans from trypsin-treated gel bands. In terms of blotting, we used reagents capable of binding core ␣1,3fucose, zwitterionic or antennal modifications. Thereby, the presence of core ␣1,3-fucose and of phosphoethanolamine as judged by the glycomic data correlates with the staining with anti-horseradish peroxidase and serum amyloid P; correspondingly the binding to C-reactive protein, which recognizes preferentially phosphorylcholine over phosphoethanolamine (25), is low (supplemental Fig. S13A). Also, reactivity toward fucose-specific Aleuria aurantia lectin, HexNAc-specific wheat germ agglutinin and Gal␤1,3GalNAc-specific peanut agglutinin (41) was detected. The AEAB-labeled neutral and anionic pools of royal jelly N-glycans (Fig. 9A) were immobilized and specifically recognized in microarray format by peanut agglutinin (42), serum amyloid P (25) and an anti-L2/ HNK-1 antibody (43,44), whereby the reactivity to peanut agglutinin and anti-L2/HNK-1 was reduced on respective ␤-galactosidase or ␤-glucuronidase treatment ( Fig. 9B and supplemental Fig. S14) and is compatible to the MS-based glycomic analyses.
For analysis of protein-specific glycosylation, the five most dominant Coomassie Blue-stained protein bands were excised as indicated and subject to trypsin peptide mass fingerprinting to verify their identity (coverage of 29 -46%; supplemental Fig. S13B-S13D and Supplemental Table 2); MRJP1 is the major band, MRJP2 the second most dominant component and MRJP3 present in three protein bands, consistent with its heterogeneity as reported in the literature (45). We could detect forms of the arginine-containing peptide 136 -146 from MRJP2 modified by complex N-glycans (Fig.  9C). The tryptic peptides were treated with PNGase Ar to release the N-glycans and these protein-specific glycomes were separately pyridylaminated, HPLC fractionated and analyzed by MALDI-TOF MS and MS/MS (supplemental Fig.  S15). Considering the difference in the amount of each protein in the sample, it is not possible to conclude that there is much variation in the types of glycans on the individual glycoproteins. Each protein (MRJP1, MRJP 2 and MRJP 3) is modified with a range of glycans despite having each a limited number (n ϭ 1-3) of glycosylation sites with the most dominant glycan being Man 9 GlcNAc 2 (in terms of MALDI-TOF MS as well as fluorescence intensity at 5.7 g.u.); on the other hand, the heterogeneous sets of hybrid and complex glycans probably account for 50% of the structures as is the case in the overall royal jelly glycome. The ability to detect phosphoethanolamine (correlating with the binding to serum amyloid P), glucuronic acid and core ␤-mannose in the MRJP1 glycome (supplemental Fig. S15D-S15K) is possibly a result of this protein being the most abundant.

N-glycan Analysis of Insects-The
Insecta is one of the most diverse classes of invertebrates with about thirty different orders, four of which contain the vast majority of insect species: Coleoptera (beetles), Diptera (true flies), Hymenoptera (wasps, bees and ants) and Lepidoptera (moths and butterflies). Among these, scientifically, economically and medically significant species include mosquitoes (e.g. Anopheles and Aedes spp.) which transmit pathogens, the model fruit fly Drosophila melanogaster and the honeybee Apis mellifera as well as Trichoplusia ni and Spodoptera frugiperda whose cell lines are used for baculovirus-based production of recombinant proteins. N-glycan structures from all these species have been described, but until recently only "simple" paucimannosidic and oligomannosidic glycans were reported (30,46,47) except for the more complex fucosylated antennae of bee venom glycoproteins (39). Of other insects, there are singular studies on glycomes or glycoproteins isolated from the silkworm Bombyx mori (48,49), the flour beetle Tribolium castaneum (50) and the locust Locusta migratoria (51), of which the latter was most noteworthy as aminoethylphosphonate was detected as an antennal modification.
The most recent data indicate that the N-glycomic potential of insects is indeed underestimated. For many years, a major question (now resolved (52)) was whether insects sialylate their glycoconjugates, but perhaps the search for sialylated structures distracted attention from considering other anionic modifications. Our workflow based on separating the neutral and anionic fractions by solid phase extraction on graphitized carbon followed by off-line MALDI-TOF MS (in both modes) is relatively unbiased, circumvents suppression-dependent underestimation of anionic modifications and allows for further FIG. 9. A, MALDI-TOF MS of AEAB-labeled royal jelly neutral and anionic N-glycans; refer to supplemental Fig. S14 for further selected MS and MS/MS data. B, The binding of peanut aggulutin (PNA) to neutral and anionic N-glycans and of the human pentraxin serum amyloid P (SAP) and the anti-L2/HNK-1 antibody clone 412 (HNK-1) to anionic N-glycans is proven by probing the AEAB-labeled N-glycan pools on NHS-modified glass slides (see corresponding MALDI-TOF MS); the chart indicates the uncorrected fluorescence values with the standard deviations (green or red channels; mean of 10 spots) as compared with concanavalin A (as a nonspecific lectin) and to negative controls (spotting buffer or no primary lectin, antibody or pentraxin). The recognition by PNA, SAP and anti-L2/HNK-1 correlates, respectively, with the presence of "pseudo" T antigen epitopes in both pools and of phosphoethanolamine and terminal glucuronic acid in the anionic pool as proven by the analyses of the PA-labeled glycans, but the binding of the glycans to wheat germ agglutinin (WGA) is minimal. As shown in supplemental Fig. S14, binding to PNA decreases and to WGA increases after ␤-galactosidase treatment, whereas binding to anti-L2/HNK-1 is ␤-glucuronidase sensitive. Refer to supplemental Fig. S13A for Western blotting data with PNA, WGA and SAP. C, A zoomed-in region of the MRJP2 spectrum (see supplemental Fig. S13D) containing ions corresponding to a peptide modified with different complex N-glycans, including one with a T antigen epitope of the type recognized by PNA; the insets show respectively a region of the spectrum after PNGase Ar-treatment and one example MS/MS showing the 2,4 A cross-ring GlcNAc-containing fragment for the m/z 2935 glycopeptide. Raw mzXML files corresponding to the full MRJP2 peptide spectra before and after PNGase Ar treatment as well as two MS/MS glycopeptide spectra are included in the supplementary zip archive. chemical or enzymatic treatments in order to confirm fine details of glycan structures (20,29). Not only does our approach result in detection of typically one hundred different glycans but has vastly increased the range of structures now known in different species including molluscs, nematodes, slime molds and insects (18,19,22,33,53,54). We have also examined binding of three lectins (Con A, PNA and WGA), one antibody (anti-L2/HNK-1) and one pentraxin (serum amyloid P) to the neutral and anionic N-glycomes of royal jelly in a microarray format.
The Variety of N-glycans in Royal Jelly-Some of the basic backbones of the neutral glycans were shown to be biantennary, others hybrid: isomeric variation could be detected because of the 1D-or 2D-HPLC separation approach. In terms of the neutral N-glycome, our data refines previous interpretations of those structures beyond the oligomannosidic, specifically showing that terminal galactosylation in royal jelly is in the context of LacdiNAc units, resulting in the previouslyreported "pseudo" T-antigen Gal␤1,3GalNAc␤1,4GlcNAc (16), but not as Gal␤1,3GlcNAc as proposed in two other studies (15,55); our data is also compatible with the presence of terminal LacdiNAc as well as of ␤1,4-linked branching GlcNAc (15,56). The galactosylation of LacdiNAc, rather than also of GlcNAc, is a hallmark of royal jelly as opposed to honeybee larvae (unpublished data), whereas the venom is well known for both core and antennal ␣1,3-fucose (39). Interestingly, we detect core ␤-mannosylation in an insect for the first time, whereas fucosylation is at a low level in royal jelly as compared with other insect glycomes.
In the present study on royal jelly, the variety of N-glycan modifications is again larger than previously appreciated and this "expansion" is primarily because of the detection of some new and unusual anionic and zwitterionic structures (summarized in Fig. 10). We have detected anionic hybrid and bi/triantennary glycans with sulfate, phosphoethanolamine and glucuronic acid singly or in combination, some of which also have multiple fucose residues; indeed, unlike previously-defined di-and trifucosylated glycans from insects (19,39), all such structures in the royal jelly glycome were glucuronylated and/or sulfated. Although phosphoethanolamine has been found on wasp and mosquito O-glycans (18,57) and insect glycolipids (38), this is the first demonstration of this modification on any N-glycan from an animal species. Otherwise, phosphoethanolamine is a component of trichomonad, fungal and bacterial N-glycans (34,35,58), protist and bacterial lipid-linked oligosaccharides (59,60) and glycosylphosphatidylinositol membrane anchors from various species (61); sulfation of mannose is known in, e.g. slime molds and insects (18,54,62), but sulfates are linked to other monosaccharides in mammalian glycoconjugates (e.g. Gal on thyroglobulin (63), GalNAc on pituitary glycoproteins (64), GlcNAc in the context of sulfo-Lewis X (65) and GlcA of the HNK-1 epitope (66)). In all animals, glucuronic acid is a component of chondroitin and heparan sulfates (67), whereas it is also present in mammals within the extended matriglycan chains on dystroglycan (68) as well as on various neural N-glycans (66); the binding of some anti-L2/HNK-1 monoclonals to blowfly glycoproteins and glycolipids, presumed to be because of nonsulfated GlcA (69), would correlate with our structural and array data for royal jelly N-glycans.
Previously, a minor MRJP2 homologue was concluded to have Hex 9 HexNAc 2 on one glycosylation site and Hex 4 -5 HexNAc 3-4 on the other (11) and here we obtained a similar result for the second site, whereas the N-terminal glycosylation site of MRJP1 is apparently modified by Hex 3-5 HexNAc 4 -6 and the other two sites by oligomannosidic structures (16), which may be because of differential exposure to Golgi processing enzymes. Because of the off-line LC-MS approach, we have revealed a heterogeneity in the proteinspecific glycosylation pattern of MRJP1 and MRJP2 previously not observed by MS or other analyses. The proteinspecific glycans identified reflect the higher amount of Man 9 GlcNAc 2 in the second preparation as compared with the first (compare supplemental Fig. S1A and S1B); this could be because of variations in hypopharyngeal gland glycan processing in individual bees or the presence of ␣-mannosidase, a proteomically determined royal jelly component (5). Furthermore, the 3D-structures of the major glycoproteins present in royal jelly may play a role in determining the accessibility of glycans to Golgi enzymes and so effect the overall status of the secreted glycome. It is also of interest that some of these royal jelly proteins are also present in honeybee venom and, in recombinant form, are recognized by the IgE of allergic patients (70), whereas MJRP1 in honey possesses glycan-dependent IgE epitopes (71). On the other hand, in contrast to our data showing a low degree of core ␣1,3fucosylation, Western blotting with an anti-plant-glycoprotein antiserum was considered negative in a previous study (14), whereas wheat germ and peanut agglutinin binding (respectively indicative for terminal HexNAc and Gal␤1,3GalNAc modifications) to royal jelly proteins was positive (the latter correlating with the glycan array data). Certainly, modifications with fucose, glucuronic acid, core mannose or phosphoethanolamine were previously not defined on single royal jelly glycoproteins.
In contrast to mosquitoes and lepidopterans (18,19), sulfation and glucuronylation were almost mutually exclusive in royal jelly, with the latter again being the dominant "charged" modification; parallel (unpublished) analyses of honeybee venom and larvae indicate that these glycomes are different in terms of the degree and type of antennal modifications, which suggests that royal jelly glycans are the product of a tissuespecific glycosylation pattern. Thereby, whereas fucosyltransferases (FucTA, FucTC and FucT6) are predicted to be expressed at low levels in hypopharyngeal glands, other gland-specific transcription events may be responsible for the increase in anionic/zwitterionic glycans. Indeed, FucTA core ␣1,3-fucosyltransferase mRNA levels are seemingly far lower in hypopharyngeal as compared with venom glands (72).
Potential Biological Relevance of Royal Jelly N-glycans-As anionic and zwitterionic moieties have different biophysical properties and so be differently recognized as compared with neutral saccharides, it is to be expected that they have dis-tinct bioactivities. For instance, sulfation of mannose of the N-glycans of a lepidopteran protein has been claimed to have a pro-inflammatory role in leukocyte activation in dermatitis caused by setae of a tropical moth (62), whereas in various species, sulfation of glycosaminoglycans, Lewis X or terminal GalNAc is associated with processes such as blood clotting, FIG. 10. Summary of the N-glycome of honeybee royal jelly. The N-glycans are categorized based on structural type with color code indicative of either oligomannosidic, paucimannosidic, hybrid, biantennary and triantennary structures; a further highlighting is used for the ␤-mannose, sulfate, glucuronic acid or phosphoethanolamine motifs. Approximate abundances were calculated based on HPLC and MS intensities. The order of abundance of the glycans decreases from the center (with less intense background colors for the least abundant), whereby the oligomannosidic Man 5-9 GlcNAc 2 are estimated as accounting for 37% of the total N-glycome. For simplicity not all glycans of a certain type are shown, rather a representative major example, whereby, e.g. the biantennary class (27%) includes structures with Gal and GalNAc residues or the hybrid class (17%) pertains also to Man 3-5 -based isomers without Gal/GalNAc modification. morphogenesis, cell adhesion or hormone clearance (73)(74)(75)(76). Furthermore, the glucuronic acid-containing L2/HNK-1-type epitope (present in nonsulfated form on royal jelly glycans as verified by the array data in Fig. 9B) may be recognized by the human cytokine IL-6, so having relevance to interleukin signaling (77), as well as by HMGB1 or amphoterin (78) which is also known to interact with Toll-like receptors (79). Phosphoethanolamine is one of the ligands for serum amyloid P protein, a pentraxin known to bind, e.g. bacteria, but also to activate complement (80) and here we show by blotting and array experiments that glycoproteins and the anionic N-glycans thereof in particular in royal jelly are indeed recognized by serum amyloid P (Fig. 9B and supplemental Fig. S13A); it would certainly be of interest to know which ligands are recognized by insect members of the pentraxin family (81), including a homologue in the honeybee (gene LOC102656294), as such proteins from another arthropod (the horseshoe crab) do bind zwitterions (82). Phosphoethanolamine is the nonmethylated form of phosphorylcholine, which is present on N-and O-glycans of insect cell lines used for recombinant protein production (19,83), but which is also known to possess immunomodulatory function as a modification of nematode glycoproteins (84). Thus, although there are no systematic studies as to the protein ligands or receptors for such moieties, for each "charged" modification of royal jelly N-glycans there is potential for bioactivity, immunoreactivity or immunoactivity.