Glycosylation at an evolutionary nexus: the brittle star Ophiactis savignyi expresses both vertebrate and invertebrate N-glycomic features

Echinoderms are among the most primitive deuterostomes and have been used as model organisms to understand chordate biology because of their close evolutionary relationship to this phylogenetic group. However, there are almost no data available regarding the N-glycomic capacity of echinoderms, which are otherwise known to produce a diverse set of species-specific glycoconjugates, including ones heavily modified by fucose, sulfate, and sialic acid residues. To increase the knowledge of diversity of carbohydrate structures within this phylum, here we conducted an in-depth analysis of N-glycans from a brittle star (Ophiactis savignyi) as an example member of the class Ophiuroidea. To this end, we performed a multi-step N-glycan analysis by HPLC and various exoglyosidase and chemical treatments in combination with MALDI-TOF MS and MS/MS. Using this approach, we found a wealth of hybrid and complex oligosaccharide structures reminiscent of those in higher vertebrates as well as some classical invertebrate glycan structures. 70% of these N-glycans were anionic, carrying either sialic acid, sulfate, or phosphate residues. In terms of glycophylogeny, our data position the brittle star between invertebrates and vertebrates and confirm the high diversity of N-glycosylation in lower organisms.

The Echinodermata represent the second largest grouping within the deuterostomes, after the Chordata, which also include vertebrates. This lineage has separated from the protostomes around 500 -600 million years ago and so is at the nexus between invertebrates and vertebrates (1). Living representatives of the echinoderms include classes of Crinoidea (feather stars), Echinoidea (sea urchins), Holothuroidea (sea cucumbers), Asteroidea (starfishes), and Ophiuroidea (brittle stars). Because of their close relationship to vertebrates, members of this phylum have been intensively studied in terms of the conserved basics of development compared with vertebrates as well as species-specific fertilization and regeneration (2, 3).
The evolutionary position of Ophiuroidea, with some 2000 living species (4), makes this class of organisms an interesting target to study the phylogeny of protein-linked glycans of not only echinoderms but also of vertebrates. To date, a major focus of glycostructural studies of echinoderms has been their O-glycans and glycolipids as well as some proteoglycan-like polymers involved in species-specific induction of the acrosome reaction during fertilization (5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15). From these studies, it appears that, in contrast to protostomes, sialic acids are probably a frequent component of echinoderm glycoconjugates and, compared with "higher" vertebrates, occur also in sulfated and methylated forms (16). Furthermore, core ␣1,3/6-difucosylation is a known recurring feature of protostome N-glycomes (17) but is absent from vertebrates and has not been identified in an echinoderm species before. Although there are a number of studies regarding the importance of N-glycan biosynthesis for sea urchin development (18,19), there is little exact information regarding the N-glycosylation capacity of echinoderms. Indeed, in terms of actual structures, there is a single report regarding mass spectrometry analysis of oligomannosidic N-glycans (20).
As part of our ongoing efforts to establish a glycophylogeny of animal species, we analyzed the N-glycans of Ophiactis savignyi, an omnivorous small tropical marine organism once claimed to be the most common brittle star on the planet, potentially invasive in some areas and capable of both asexual and sexual reproduction (21). We hypothesized that echinoderm species would display a mixture of invertebrate and vertebrate structural features; indeed, we found many N-glycans with antennal sialic acid or sulfate but also others displaying difucosylation of the core. Furthermore, compared with our other studies of echinoderms (Ref. 22 and unpublished data), there are also unique motifs that define the individual N-glycome of this species.

Overall glycomic approach
Our initial hypothesis regarding echinoderm N-glycosylation was that there would be a substantial degree of modification of oligosaccharides with anionic moieties as in vertebrates but that there may be vestiges of invertebrate-type N-glycomic features, such as difucosylation of the asparagine-bound chitobiosyl core. Therefore, the glycomic workflow for O. savignyi was aimed not only at separating neutral from anionic N-gly-cans but also at releasing structures with core ␣1,3-fucose. Serial use of PNGase F and A followed by solid-phase extraction on nonporous graphitized carbon allowed isolation of three pools of glycans (Fig. S1, enriched in neutral, anionic, and core ␣1,3-fucosylated structures), all of which were fluorescently labeled and analyzed by off-line HPLC MALDI-TOF MS. The overall predicted N-glycan compositions, based on MS and MS/MS before and after chemical or enzymatic treatments, are summarized in Table S1. Furthermore, O-glycan analysis of the residual glycopeptide fraction was performed by LC-ESI 3 -MS following reductive ␤-elimination.

Neutral N-glycans
The neutral PNGase F-released pool contains primarily oligomannosidic and hybrid structures (see Fig. 1 for a normal phase profile), and their isomeric status could be evaluated on the basis of RP HPLC elution times (Fig. S2) and MS/MS fragmentation compared with previous studies. Our initial interest was in determining the nature of the potential galactosylation of an example hybrid structure (m/z 1678 as [MϩH] ϩ ) as well as of a low-abundance biantennary glycan (m/z 1865). Based on use of linkage-specific galactosidases (Fig. 2), ␤1,3-galactosyla-tion (type 1 chain, Gal␤1-3GlcNAc␤1-R), as found in a number of invertebrates, rather than the ␤1,4-form (type 2 chain, Gal␤1-4GlcNAc␤1-R) dominant in mammals, is concluded to be a feature of this echinoderm. Antennal ␤1,3-galactosylation was also a basic feature of the anionic glycans.
In later neutral normal-phase fractions, glucosylated oligomannosidic structures were also found, as were glycans with unexpected Hex 13-16 HexNAc 2 compositions. These were partially susceptible to digestion with jack bean ␣-mannosidase but also sensitive to a bacterial endo-␣-mannosidase, as judged by shifts in the MS and MS/MS spectra ( Fig.  3 and Fig. S3). As Hex 13 HexNAc 2 (m/z 2610) lost five hexoses upon either endo-mannosidase or exo-mannosidase digestion, and no glucosidase or galactosidase tested removed the extra hexose residues, the Hex 13-16 HexNAc 2 glycans are predicted to be Glc 3 Man 9 GlcNAc 2 structures carrying unknown hexose residues on the glucosylated arm.
Upon PNGase A treatment of glycopeptides remaining after PNGase F release, a mixture of neutral structures was obtained. Only fucosylated paucimannosidic glycans as well as "residual" oligomannosidic ones were present in this pool. Of the identified structures, two had elution times (8 g.u.), compositions and MS/MS patterns (i.e. a Y ion at m/z 592) indicative of core ␣1,3/␣1,6-difucosylation (Fig. S4). Because of its early elution, the monofucosylated structure at 5 g.u. (m/z 1135,

Brittle star N-glycome
Hex 3 HexNAc 2 Fuc 1 ) is also concluded to be core ␣1,3-fucosylated, which is a feature known from plants and invertebrates but not from vertebrates.

Separation of anionic N-glycans
Based on fluorescence intensity of the various N-glycan pools, it was estimated that some 70% of the structures are anionic, and initial mass spectrometric data suggested that both sulfated and sialylated glycans were present (Fig. S1). "Sialylation-friendly" RP HPLC conditions (pH 6) in a first dimension yielded 11 pools that were then subject to charge/size-based separation using a HIAX column; almost 120 2D-HPLCseparated glycan-containing fractions were isolated ( Fig. 4 and Fig. S5).

Sulfated and phosphorylated N-glycans
Although sulfation and phosphorylation both result in an 80-Da modification compared with "parental" neutral N-glycans, their properties differ (23); for instance, "in-source" loss in positive mode is a hallmark of sulfate, whereas sensitivity to hydrofluoric acid is a characteristic of phosphate. Therefore, we were able to conclude that both sulfated and phosphorylated N-glycans are synthesized by O. savignyi.
Based on negative-ion-mode MS/MS and the results of exoglycosidase treatments, three different types of sulfated monosaccharides were observed: core ␣1,6-fucose, antennal galactose, or antennal GlcNAc. Sulfation of core ␣1,6-fucose has been found previously in insects, and hallmarks are the m/z 225 sulfated fucose fragment and resistance to bovine fucosidase treatment (Fig. 5, A-E, K, and M). In comparison, an isomeric glycan with sulfation of an antennal GlcNAc was distinguishable on the basis of the m/z 282 B fragment and loss of only one HexNAc upon serial galactosidase/hexosaminidase digestion (  was sensitive to a ␤1,3-specific galactosidase (B) but resistant to a ␤1,4-specific galactosidase (C). F-H, treatment of the hybrid glycan (F) m/z 1678 (Hex 6 HexNAc 3 ) with ␤1,3-galactosidase resulted in loss of one galactose (G), whereas jack bean ␣-mannosidase slowly removed up to three mannose residues (H); the coeluting Hex 7 HexNAc 2 structure was resistant to galactosidase but sensitive to mannosidase. In the MS/MS spectra of untreated (D and I) and galactosidase-treated glycans (E and J), loss of galactose (⌬162) is indicated by red lines, e.g. shifts in B ions from m/z 366 (Hex 1 HexNAc 1 ) to m/z 204 (HexNAc 1 ) as well as in Y ions from m/z 1338 (Hex 3 HexNAc 3 Fuc 1 ) or m/z 1192 (Hex 3 HexNAc 3 ) to ones at m/z 1176 (Hex 2 HexNAc 3 Fuc 1 ) or m/z 1030 (Hex 2 HexNAc 3 ).
Other glycans with an 80-Da modification were observed in both positive and negative modes and so were concluded to be phosphorylated on HexNAc residues, as judged by the m/z 284 [MϩH] ϩ fragment ions (Fig. 6, A-E). In the case of two hybrid structures, this was verified by sensitivity to jack bean ␤-hexosaminidase only after hydrofluoric acid treatment; subsequently one or two hexose residues could be removed upon jack bean ␣-mannosidase treatment (Fig. 6, F-I).

Sialylated N-glycans
A number of glycans contained modifications of either 307 or 321 Da, as defined by results of positive-and negative-mode MS/MS. Relevant B fragments in either positive or negative ion mode indicated the presence of N-glycolylneuraminic acid or its methylated form on terminal HexHexNAc motifs (Fig. 7, A-G, and Fig. S6). As there were multiple isomers for some masses among the 2D-HPLC fractions (e.g. two, three, or even four each of mono-and biantennary structures with m/z 1807, 1821, 2172, or 2333; Fig. S5), a major question was whether there were multiple positions for sialylation. Therefore, selected glycans were incubated with sialidases and/or mild acid.
As shown for the four isomeric monosialylated biantennary Hex 5 HexNAc 4 Fuc 1 NeuGc 1 glycans (Fig. 8, 2171 Da), two structures only lost the NeuGc residue upon acid treatment, whereas two others were sensitive to ␣2,3-specific sialidase S. On the other hand, the sialidase-resistant glycans lost two hexose residues upon ␤1,3-galactosidase treatment, whereas only one hexose was removed from those sensitive to sialidase S ( Fig. 8, B, G, R, and W). Thus, as sialic acids are known to modify either terminal galactose or antennal GlcNAc residues in, e.g., mammalian fetuin, we concluded that this also occurs in the brittle star (see below); indeed some structures with NeuGc on both Gal and GlcNAc of the same antenna were also found ( Fig.  7, E-G).
To determine which antennae of the 2171-Da isomers were sialylated, serial chemical or enzymatic treatments were then performed. By comparing the RP HPLC retention time of digestion products with standards, the two isomers with sialylated GlcNAc could be distinguished regarding the NeuGc substitution on either the ␣1,6 or ␣1,3 antenna (Fig. 8, A-P). On the other hand, despite similar fragmentation patterns and only slightly different 2D-HPLC elution properties, the two isomers with sialylated galactose were shown to differ in terms of which antenna carried the sialic acid residue by performing ␤1,3-specific galactosidase treatment followed by incubation with the arm-specific insect FDL hexosaminidase (25) (Fig. 8

, Q-Z).
A similar approach was employed on two glycans with putative NeuGcMe modifications. In the case of a terminal NeuGcMe residue, sialidase S treatment resulted in loss of 321 Da, but the glycan was resistant to ␤1,3-galactosidase unless desialylated (Fig. S7, A-C and E); in contrast, the isomeric glycan with later NP-HPLC retention was resistant to sialidase S but sensitive to ␤1,3-galactosidase and acid treatment ( Some glycans were predicted to contain both sulfate and sialic acid; in the case of two isomeric structures (Hex 5 HexNAc 4 Fuc 1 NeuGc 1 S 1 ) with different 2D-HPLC elution properties, fragments at either m/z 282 or 387 were observed and correlated with loss of either 307 or 387 Da upon sialidase S treatment. This suggested that the sulfate residues were either on an antennal GlcNAc or covalently bound to the terminal sialic acid residue (Fig. 7, H-M).
As mentioned above, the resistance to hydrofluoric acid of the 80-Da moiety would indicate the occurrence of sulfate, and when treating the complete anionic pool with this reagent, major peaks at, e.g., m/z 1578 and 1943 were still dominant in the overall spectrum. However, the masses of the sialylated glycans shifted by 18 Da (Fig. S8). This change is akin to that upon lactonization and suggested that hydrofluoric acid can also stabilize specific sialic acid linkages, similar to combined 1-ethyl-3-(3-dimethylaminopropyl)-carbodiimide/1-hydroxybenzotriazole treatment; the shift in the NeuGc 1 Hex 1 HexNAc 1 positive-mode fragment from m/z 673 to 655 is also the same (Fig. S8, C-H). Therefore, there were two sets of proof that NeuGc or NeuGcMe on terminal galactose is ␣2,3-linked. On

Brittle star N-glycome
the other hand, sialic acid linked to GlcNAc is concluded to be ␣2,6-linked because of the ability to remove it with a nonspecific sialidase, 3-sialylation of GlcNAc being ruled out by the presence of the 3-linked galactose, and LC-ESI-MS 3 evidence (Fig. 7E).

Maximum number of N-glycan antennae
Not only did the results of hydrofluoric acid treatment of the entire anionic pool indicate lactonization, but we could also detect the presence of trisialylated triantennary structures (m/z 3113 in negative mode, Fig. 9). Upon desialylation of the anionic pool using 2 M acetic acid, we could also observe larger glycans (Hex 6 -7 HexNAc 5-6 Fuc 1 ) whose monosulfated forms were detected in negative mode ( Fig. 9 and Fig. S9, A-D). As MS/MS analysis indicated that these correspond to tri-and tetra-antennary structures, selected RP HPLC fractions containing mono-, bi-, and triantennary glycans were also subjected to mild acid hydrolysis prior to degalactosylation (Fig. S9, E-J) and RP HPLC, as this method is known to distinguish different triantennary isomers (24,25). The underlying structures were compared with asialoagalactoglycans derived from fetuin (Fig. S9, K-M), for which the ␤1,2-GlcNAc on the ␣1,6-mannose and the ␤1,2/␤1,4-disubstitution of the ␣1,3-mannose have been defined previously. The resulting retention time comparisons (Fig. S9N) indicated that the mono-antennary glycans were ␤1,2-substituted on the ␣1,3-mannose and the biantennary ␤1,2-substituted on both ␣-mannose residues, whereas the triantennary glycans from O. savignyi have the same configuration as those in bovine fetuin.

Discussion
Because of their close relationship to vertebrates at the nexus of protostomes and deuterostomes, echinoderms are historically useful model organisms for studies of development, fertilization, and embryogenesis. As little is known about their N-glycosylation capacity, we conducted an in-depth analysis of N-glycans from O. savignyi. Our initial hypothesis was that the N-glycome of an echinoderm would reflect the evolutionary position of this phylum and would therefore have aspects of both invertebrate and vertebrate glycosylation patterns. Indeed, this presumption, based also on older work, especially on O-glycans and glycolipids, was more than confirmed in the case of the brittle star O. savignyi and a sea cucumber, Holothuria atra (as shown in our companion report). Previously, the occurrence of calcium-binding sulfated N-glycans potentially involved in sea urchin embryonic skeleton formation was postulated (29,30), but their exact structures were never defined, whereas the occurrence of classical oligomannosidic structures in echinoderms was confirmed in a mass spectrometry study on egg jelly glycans from the sea urchin Paracentrotus lividus (20).
Therefore, based on our reading of the literature, our two studies are indeed the first in-depth N-glycomic analyses of any echinoderm species. As summarized in Fig. 10A, in the brittle star, we found complex and hybrid N-glycans featuring the type 1 antennal motif, a minor degree of the ␣1,3/␣1,6-difucosylation core modification, glucosylated oligomannosidic structures with an unknown hexose cap, and various anionic modifications, specifically sulfate, phosphate, and sialic acid.
Invertebrate characteristics of both the brittle star and sea cucumber glycomes are reflected in the occurrence of some core difucosylated N-glycan structures of the type known from a wide range of insect, mollusk, trematode, and nematode species (31). Modification of antennal GlcNAc residues with ␤1,3linked galactose is also shared with oysters and insects (17,(32)(33)(34), but LacdiNAc (GalNAc␤1,4GlcNAc, found in many invertebrates) was lacking (Fig. 10B). Although ␤1,3-galactose is also present in mammals, the typical vertebrate modification with ␤1,4-galactose is absent in the brittle star, as are digalactose motifs, whether these be Gal␣1,3/4Gal or Gal␤1,4Gal, as found in many mammals, birds, and fish (35)(36)(37), and histo- In case of isomers with internal NeuGc, ␤1,3-galactosidase was able to cleave two residues (B and G) compared with one galactose for those with one terminal sialic acid residue (R and W); only after ␤1,3-galactosidase digest could the key fragment for the NeuGc-HexNAc modification, m/z 511 (insets in B, C, G, and H) be detected. The four isomers also differed in their susceptibility to ␣2,3-sialidase S, which only cleaves off terminal NeuGc (S and Y), whereas internal NeuGc was removed with acid hydrolysis (D and I). To determine which arms were sialylated, two of the m/z 1338 digest products were reinjected on HPLC (K and L) and coeluted with two standards, A and B, which also showed different key fragment ions (M-P) at m/z 1135 as in standard A and digest product H, 33 min, and m/z 1176 as in standard B and digest product I, 35 min; the other two m/z 2172 structures were subject to lower arm-specific FDL hexosaminidase (T and X), where the respective resistance or sensitivity proves substitution of the lower or upper arm.

Brittle star N-glycome
blood group motifs. In contrast, the sea cucumber H. atra has sulfo-and sialyl-Lewis A motifs (22). On the other hand, there are certainly triantennary N-glycans and a trace of tetra-antennary glycans indicative of the presence of a number of branching N-acetylglucosaminyltransferases, including GlcNAc-TIV (38), an enzyme also predicted to occur in cnidarians and some nematodes as well as insects and vertebrates (Fig. S11). Interestingly, when there is only one Gal␤1,3GlcNAc antenna, this is always on the ␣1,3-mannose, which is an indication that there is no dominant FDLlike hexosaminidase (25) removing the ␤1,2-linked GlcNAc from the lower arm, as known from many invertebrates.
The high degree of "charged" N-glycans in O. savignyi (estimated as 70%, as summarized in Fig. 10A) results in a more "vertebrate-like" characteristic, although we previously found an approximately 50% degree of N-glycan sulfation in oyster plasma (33). Indeed, the predominance of sialylation differentiates it from typical invertebrates, but the sialic acid modifications (methylation and sulfation) are different as compared with vertebrates. N-glycolylneuraminic acid is the most abundant sialic acid on N-glycans of the brittle star, with either ␣2,3-sialylated ␤1,3-linked galactose residues or ␣2,6sialylation of antennal GlcNAc. The latter modification is known from bovine fetuin (39), and corresponding mammalian GlcNAc-6-sialyltransferase activity has been reported (40). Recently this "internal" sialylated motif has also been found in the mouse brain and named "sialyl-Lewis C" (41). This epitope, which interacts with Siglec-H, has been found A, based on fluorescence intensities, 30% of N-glycans are neutral, and 70% carry an anionic moiety. Within these classes, 92% of the neutral N-glycans are classical oligomannosidic Man 5-9 GlcNAc 2 structures, but neutral glycans with extra glucose and unknown hexose residues are also present, in addition to neutral core ␣1,3-fucosylated, hybrid, and biantennary glycans; antennal galactosylation was solely detected in a ␤1,3 linkage. Within the acidic N-glycan pool, sialylation (solely N-glycolylneuraminic acid) is the major anionic modification, with up to three sialylated antennae being detected; 20% of the sialylated structures are disialylated (one on Gal, one on GlcNAc) to result in sialyl-Lewis C epitopes (gray box). 5% of sialylated structures are also sulfated, and 20% of NeuGc residues are methylated. Glycans with just sulfation are less abundant, and those displaying phosphorylation account for just 1% of the acidic pool. B, simplified glyco-evolutionary scheme showing the occurrence of selected antennal elements in protostomes and deuterostomes. The phylogeny is adapted from Vaughn et al. (69). Depicted are Lewis X; sulfated blood group A; glucuronylated LacdiNAc (also modified by phosphorylcholine); glucuronylated/sulfated type I; sulfated Lewis A; NeuGcMe-modified type I and NeuAc-modified type II antennae from Schistosoma mansoni, Crassostrea virginica, Dirofilaria immitis, Anopheles gambiae, H. atra, O. savignyi, and Homo sapiens (17,22). Note that anionic N-glycans are unknown in S. mansoni; D. immitis is the only nematode for which glucuronylated N-glycans have been proven.

Brittle star N-glycome
in developing mouse brains and may mediate interactions between dendritic spines and microglia (42). Although glycoproteins and their binding partners still need to be identified in echinoderms, it is known that their neuronal systems display similarities to those in chordates (43).
We also examined O-glycans and found a number of simple structures in the brittle star, some of which are sialylated. The previous literature on echinoderm O-glycans is a bit richer than for N-glycans, and di-, oligo-, or polysialylated structures have been found in sea urchins. Specific examples include mucintype O-glycans from sperm flagella containing terminal 8-Osulfated ␣2,9-linked polyNeu5Ac (12,13) or other sperm and egg O-linked glycoproteins with ␣2,8-sialic acid and ␣2,5-sialic acid (14,15,44). Also, sialylated glycolipids from various echinoderms have been described (14,(45)(46)(47). Therefore, a range of sialyltransferases should be present in echinoderms, and indeed a number of ␣2,3-, ␣2,6-, and ␣2,8-sialyltransferase homologs are encoded by the genome of the sea urchin Strongylocentrotus purpuratus (48 -50). However, considering the lack of enzymatic characterization of the various homologs in any echinoderm, it is difficult to predict which sialic acid linkages are theoretically possible in each species. Thus, we do not know whether our detection of only monosialylated motifs in the brittle star is either due to lack of ␣2,5/8/9-sialyltransferases in the brittle star or due to preparative/analytical factors.
In terms of types of sialic acid, it is remarkable that the NeuGc and its methylated form are the only sialic acid residues on the N-glycans of O. savignyi, whereas O-glycans carry both NeuAc and NeuGc. As CMP-NeuGc biosynthesis requires prior formation of CMP-NeuAc, absence of NeuAc from N-glycans seems curious. However, methylation and sulfation of NeuGc match reports of starfish and sea cucumber O-and lipid-linked glycans (8,12,(51)(52)(53), and two starfish enzymes required for generation of methylated NeuGc (CMP-NeuAc hydroxylase and 8-O-methyltransferase) have already been identified (54,55). O-acetylation of sialic acid is known from other echinoderms but is notoriously difficult to detect because of instability, and we did not define any acetylated structures, whereas sulfated sialic acid was found only on a small subset of brittle star N-glycans. As NeuGc and acetylated NeuAc have been reported to be less easily cleaved by bacterial and viral sialidases than NeuAc (56 -58), it can be speculated that the abundance of NeuGc or its methylated or sulfated forms on the N-glycans of O. savignyi may provide a first line of defense against pathogens.
In terms of anionic modifications other than sialylation, sulfation and phosphorylation were detected; as mentioned above, there was some sulfation of sialic acid residues, but sulfated galactose, GlcNAc, and fucose were also observed. Here there are parallels to the sulfation of galactose in oysters and of core fucose in insects (32,33). LC-MS suggested that the galactose residues are 4-O-sulfated, unlike the 3-O-sulfation more common in mammals; nevertheless, 4-O-sulfo-modifications of fucose and GalNAc are known as modifications of sea cucumber chondroitin chains (59), and 4-sulfated GalNAc is a component of mammalian chondroitin and dermatan sulfates (60). On the other hand, phosphorylation of GlcNAc was found, rather than of mannose, as known from N-glycans of mamma-lian lysosomal enzymes or slime mold glycoproteins (61,62), but no zwitterionic modifications were detected (23).
Glycosylation is a species-specific process, which is especially reflected in the distinct set of glycans involved in the fertilization process (63). These differences were also reflected in our studies, as only low amounts of Man 1-3 GlcNAc 2 -based paucimannosidic glycans and almost no pseudohybrid glycans (i.e. with an antenna only on ␣1,6-mannose) were found in O. savignyi. It is noteworthy that the anionic N-glycome of the sea cucumber H. atra is rather biased toward hybrid structures based on Man 5 GlcNAc 2 , including sulfated/fucosylated, but more rarely sialylated, HexHexNAc motifs (22). Additionally, our preliminary data on two sea urchin and two starfish species revealed a distinct set of N-glycans. This highlights that the individual echinoderm species analyzed have a quite different overall N-glycome, even though some motifs overlap. Thus, we conclude that the N-glycome of O. savignyi is unique compared with other species, even compared with that of H. atra. Indicative that there is a push to "final products" of certain biosynthetic pathways is that the hybrid and complex glycans of this species are almost all sulfated and/or sialylated (accounting for 70% of the N-glycome), leaving a neutral subset dominated by oligomannosidic structures, including some with unusual hexose modifications on the "lower arm" triglucose motif. In general, except for the presence of some core difucosylated structures, the O. savignyi N-glycome is probably nearer to those of vertebrates; further comparative studies will certainly show variations and similarities in the glycomotifs associated with emergence of the deuterostome lineage before the Cambrian era.

Biological material and glycan preparation
The brittle stars O. savignyi (Müller and Troschel, 1842) were sourced from Haus des Meeres -Aqua Terra Zoo, a public aquarium in Vienna, Austria. There, upon first detection in 2013, they were already living in a total of four coral reef exhibition tanks pertaining to two independent seawater systems. Therefore, only gross reconstruction of origin and date of introduction appears possible; during the last decade, all imports of coral reef materials came from the central Indo-Pacific Ecoregion (Philippines, Indonesia, and northeast Australia) via a Dutch wholesaler (De Jong Marinelife Inc.). The imports of live corals usually comprised small portions of foundation rocks. This suggests that the O. savignyi founder population had hitchhiked in shelters found in crevices of such rocks and that they then populated the open-celled polyurethane foam, which served as the filter of the aquarium system. A regimen of 10 h of dim light/14 h of darkness prevailed in the service room. The inorganic composition of the seawater was routinely analyzed employing inductively coupled plasma -optical emission spectrometry (ICP-OES) and ion chromatography-variable wavelength detection/electrochemical detection (IC-VWD/ECD) methodology (Oceamo Inc., Vienna, Austria; Table S3). Freshly collected animals were put in small Petri dishes for microphotography ( Fig. 1 and Fig. S12).

Brittle star N-glycome
10 g of starved brittle stars were handpicked from aquarium substrates, washed once with deionized water, and stored at Ϫ80°C. After thawing, the material was heat-inactivated for 10 min in boiling water and lyophilized prior to grinding in liquid nitrogen. The powder was suspended in deionized water, and the pH was adjusted with 100 mM ammonium carbonate buffer to 8.2; CaCl 2 was added to a final concentration of 0.5 mM prior to addition of 10 mg of thermolysin (Promega). Proteolysis was allowed to proceed for 2 h at 70°C and 37°C overnight for completion. The glycopeptides were purified using standard laboratory protocols (64) prior to glycan release with PNGase F (Roche) overnight at 37°C. After cation exchange chromatography on Dowex, the unbound material was subjected to solidphase extraction on nonporous graphitized carbon (ENVIcarb, Supelco) and eluted with 40% acetonitrile or 40% acetonitrile with 0.1% TFA. These fractions were subjected to a further solid-phase extraction step on C18 reverse-phase resin (LiChroprep, Merck), and the glycans were eluted with water and stepwise increases in methanol (15%, 40%, and 100% (v/v)) concentration. Glycan-containing fractions, as judged by MALDI-TOF MS, were fluorescently labeled with 2-aminopyridine. The remaining glycopeptides bound to the Dowex resin were gel-filtrated (Sephadex G25) prior to incubation with PNGase Ar (recombinant, New England Biolabs) overnight at 37°C to facilitate release of core ␣1,3-fucosylated glycans. The glycans released with this enzyme were then no longer bound by Dowex and were subject to the same purification and fluorescent-labeling steps as for the PNGase F released ones. The residual glycopeptides remaining after PNGase Ar digestion were subject to reductive ␤-elimination. As the animals were starved and no traces of algal glycans (which would be expected to contain pentose residues) were detected, no food contamination of the glycomes occurred.

Glycan MS
Monoisotopic MALDI-TOF MS was performed using an Autoflex Speed (Bruker Daltonics) instrument in either positive or negative reflection mode with 6-aza-2-thiothymine or 2,5dihydroxybenzoic acid as matrix. In general, MS/MS was performed by laser-induced dissociation of [MϩH] ϩ or [M-H] Ϫ ions (except for derivatized sialylated glycan masses detected as [MϩNa] ϩ with 2,5-dihydroxybenzoic acid as matrix); typically 2000 shots were summed for MS (reflector voltage, lens voltage, and gain of 27 kV, 9 kV, and 2217 V, respectively) and 4000 for MS/MS (reflector voltage, lift voltage, and gain of 27 kV, 19 kV, and 2174 V, respectively). 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 (smoothed four times). Glycan spectra were manually interpreted on the basis of the masses of the predicted component monosaccharides; differences of mass in glycan series; comparison with coeluting structures from other insects, nematodes, and other marine organisms; and fragmentation patterns before and after chemical treatment or exoglycosidase digestion. A list of theoretical m/z values for each glycan composition is presented in Table S1.
Selected RP HPLC fractions of pyridylaminated N-glycans as well as O-glycans prepared by ␤-elimination were also analyzed by online LC-MS/MS using a 5-m porous graphitized carbon column (10 cm ϫ 150 m, prepared in-house) coupled to an LTQ ion trap mass spectrometer (Thermo Scientific, Waltham, MA). Glycans were eluted using a linear gradient from 0 -40% acetonitrile in 10 mM ammonium bicarbonate over 40 min at a flow rate of 10 l/min. The eluted N-glycans were detected in negative ion mode with an electrospray voltage of 3.5 kV, capillary voltage of Ϫ33.0 V, and capillary temperature of 300°C. Specified ions were isolated for MS n fragmentation by collision-induced dissociation with the collision energy set to 30%. Air was used as a sheath gas, and mass ranges were defined depending on the specific structure to be analyzed. The data were processed using Xcalibur software (version 2.0.7, Thermo Scientific). Glycans were identified from their MS/MS spectra by manual annotation using the fragment nomenclature of Domon and Costello (65).
For removal of phosphate groups, selected fractions were dried and incubated for 48 h at 0°C with 3 l of 48% (v/v) hydrofluoric acid prior to evaporation in a centrifugal concentrator. The samples were diluted in water and re-evaporated before dissolving again. Under these conditions, N-glycolylneuraminic acids were also stabilized by loss of a water molecule. Otherwise, shrimp alkaline phosphatase (Fermentas) was employed to remove phosphate residues. As appropriate, treated glycans were rechromatographed by RP HPLC (pH 6) to ascertain retention time shifts prior to MALDI-TOF MS; otherwise, an aliquot (generally one-fifth) of any digest was analyzed by MALDI-TOF MS without further purification.

Linkage-specific sialic acid derivatization and mild acid hydrolysis
Sialic acid-specific derivatization was performed as described previously (67). Around one-fifth of a sample was mixed with the carboxylic acid activator 1-ethyl-3-(3-dimethylamino)-propyl)carbodiimide together with the catalyst 1-hydroxybenzo-triazole as follows: a 0.5 M solution of 1-ethyl-3-(3-dimethylamino)-propyl)carbodiimide and 1-hydroxybenzo-triazole in 100% (v/v) ethanol was added to the sample at a 1:1 ratio, and incubation was performed for 1 h at 37°C. After adding 100% (v/v) MeCN to a final concentration of around 85% (v/v), the reaction mixture was purified using a Hydrophilic Interaction Liquid Chromatography-Solid Phase Extraction (HILIC-SPE) followed by MALDI-TOF analysis using 2,5-dihydroxybenzoic acid as matrix. Because of resistance of some sialic acid linkages toward sialidases, chemical hydrolysis was employed. Around one-fifth of a selected HPLC peak was mixed with 2 M acetic acid and incubated for 2 h at 80°C (68). To remove the acetic acid, the samples were dried using a SpeedVac and washed twice by adding water and re-evaporation. The desialylated samples were directly analyzed by MALDI-TOF MS.