Introduction

Glycoscience, the study of glycans (complex carbohydrates) and glycoconjugates along with their biological roles, is a rapidly evolving and impactful field within biomedical research. Glycans, found abundantly on cell surfaces and within the extracellular matrix, play essential roles in cellular communication, immune responses, and disease processes1,2,3,4,5,6,7,8. They contribute to protein folding, stability, and trafficking and are central to key biological interactions, including cell-cell recognition and pathogen-host interactions, often through specific protein-glycan interactions. Abnormal glycosylation patterns are frequently associated with conditions such as cancer, inflammation, and infectious diseases. As such, elucidating glycan structures and functions holds great promise for developing diagnostics and targeted therapies.

Despite its importance, glycoscience remains less understood than genomics or proteomics, largely due to the structural complexity of glycans and the technical challenges involved in their study. Recent advances in high-performance liquid chromatography (HPLC)9,10,11, mass spectrometry (MS)12,13,14,15,16,17,18,19,20, and glycan array technologies21,22,23,24,25,26 are driving progress; however, a key limitation persists: the restricted access to biologically relevant glycans with sufficient structural diversity and in quantities necessary for functional analyses. Traditional chemoenzymatic approaches have enabled glycan production, but even with recent improvements27,28,29,30,31,32, the synthesis of complex glycans remains technically demanding and time-consuming.

While isolation of natural glycans is possible, it is typically done on a small scale for analytical purposes, yielding insufficient quantities for extensive functional studies. To address this, we developed a method called oxidative release of natural glycans (ORNG) using household bleach, allowing the extraction of glycans from large quantities of natural material33,34,35. However, efficiently separating heterogeneous glycan mixtures into pure individual glycans, especially at gram scale, remains challenging due to the limited availability of complex glycans at this scale in prior studies.

For both analytical and preparative purposes, natural glycans are often released from glycoconjugates as free reducing glycans, labeled with chemical tags for sensitive detection, and analyzed by HPLC or MS. Tags based on 2- or 4-aminobenzoic acid are commonly used to improve glycan separation and detection36,37,38,39,40,41,42,43,44. These tags, installed via reductive amination, provide a chromogenic or fluorescent group in the UV-Vis range, often enhancing MS sensitivity. However, this reaction also opens the reducing end ring structure, which can impact glycan function and is typically irreversible. Free reducing glycans are important intermediates for further functionalization or for constructing more complex glycoconjugates. While traditionally considered irreversible, we recently discovered that N-bromosuccinimide (NBS)45 and Oxone46 can regenerate free reducing glycans from 2-aminobenzoic acid-tagged glycans, though significant side products are formed in the process.

Our prior research focused on the development of the oxidative release of natural glycans (ORNG) method using household bleach33,34,35, which significantly advanced the gram-scale isolation of natural N-glycans from various biomaterials. To address the challenge of separating complex glycan mixtures, we introduced innovative tagging techniques using 2-aminobenzoic acid derivatives for enhanced detection and purification. Our work demonstrated the potential of our methods in overcoming the limitations posed by the low natural abundance and structural heterogeneity of glycans, setting a foundation for the current study.

In this manuscript, we present an approach using 4-aminobenzoic acid-based chemical tags—such as para-aminobenzoic acid (PABA), ethyl 4-aminobenzoate (EPAB, also known as benzocaine), and 2-(diethylamino)ethyl 4-aminobenzoate (DEAB or procaine)—as cleavable tags for the quantitative regeneration of free reducing glycans. Notably, the EPAB tag greatly facilitates glycan purification, enabling gram-scale production of tagged glycans and allowing for separation through various chromatography techniques. Using this chemistry, we demonstrate a streamlined, gram-scale preparation of complex N-glycans from abundant natural materials.

Results

Oxone-mediated regeneration of free reducing N-glycans from glycan conjugates tagged with 4-Aminobenzoic acid derivatives

Our previous study demonstrated that 2-aminobenzoic acid derivative tags, installed at the reducing end of glycans via reductive amination, can be removed through oxidation-facilitated hydrolysis using either N-bromosuccinimide (NBS) or Oxone (Scheme 1). However, each reagent produces distinct side products: NBS yields a major byproduct with a reducing-end arabinose, resulting from the loss of the anomeric carbon, while Oxone produces a modified tag side product. This difference is likely due to bromine’s strength as a leaving group, which promotes C1-C2 carbon-carbon bond cleavage through a pericyclic rearrangement—a process less prevalent in Oxone oxidation. Additionally, oxidation of the aromatic NH2 at the 2-position of benzoic acid can lead to the formation of a five-membered heterocycle. We hypothesized that substituting the 2-amino group with a 4-amino group would eliminate both side products under Oxone treatment. Notably, 4-aminobenzoic acid derivatives, such as benzocaine and procaine, have been used previously for analytical purposes, prompting us to evaluate these compounds as cleavable tags for glycan conjugation.

Scheme 1
scheme 1

Regeneration of free reducing glycans from 2-aminobenzoic acid derivative-tagged glycans using NBS/Oxone. The reaction of 2-aminobenzoic acid derivative-tagged glycans with NBS/Oxone results in the release of free reducing glycans, accompanied by the formation of side products.

Full size image

In Fig. 1a, we show that high-mannose glycan Man9 (Man9GlcNAc2-OH) can be efficiently tagged with PABA, EPAB, and DEAB under standard reductive amination conditions. Upon mixing equimolar amounts of Man9-PABA, Man9-EPAB, and Man9-DEAB, significant differences in MS sensitivity were observed: procaine was ~10 times more sensitive than benzocaine, which was about twice as sensitive as PABA. This finding is consistent with recent reports showing high sensitivity of procaine, procainamide, and related tags in LC-MS analysis of glycans43,44 (Fig. 1b). Additionally, C18-HPLC with UV detection revealed distinct retention behaviors for each tag (Fig. 1c): the excess benzocaine tag had a significantly longer retention time than Man9-EPAB; the excess PABA tag exhibited a shorter retention time than Man9-PABA; and the excess procaine tag showed a slightly longer retention time than Man9-DEAB. Overall, the tags’ retention on C18 media followed the order benzocaine > procaine > PABA, and the tagged glycans exhibited similar UV absorbance at 280 nm.

Fig. 1: Man9GlcNAc2-OH can be efficiently tagged with 4-aminobenzoic acid derivatives.
figure 1

a Reaction scheme, b MS results, and (c) HPLC results of Man9GlcNAc2-OH tagged with para-aminobenzoic acid (PABA), ethyl 4-aminobenzoate (EPAB), and 2-(diethylamino)ethyl 4-aminobenzoate (DEAB) under standard reductive amination conditions. Green circles, Man; blue squares, GlcNAc.

Full size image

Efficient tagging using all three tags was also demonstrated with the disialylated N-glycan (S2G2-M, composition Hex5HexNAc3Neu5Ac2), as expected (Fig. 2a). Subsequent treatment of S2G2-M-PABA, EPAB, and DEAB conjugates with Oxone proceeded smoothly, yielding corresponding free reducing glycans as the sole product (Fig. 2b). No significant desialylation or over-oxidation side products were observed. While all three tested tags—and potentially other 4-aminobenzoic acid derivatives—serve as effective reversible tags for free reducing N-glycans, we selected benzocaine as the optimal tag due to its stronger hydrophobicity, which enhances affinity purification of tagged glycans from untagged glycans, salts, excess tags, and other impurities. Furthermore, benzocaine is commercially available on a kilogram scale, making it feasible for large-scale applications. We anticipated that these properties would facilitate gram-scale preparation of tagged glycans, allowing efficient separation through multi-dimensional chromatography. Specifically, released natural N-glycans are not retained on C18 reverse-phase media; however, after EPAB conjugation, tagged glycans are well-retained on C18 resin and can be eluted with 15% acetonitrile. This straightforward C18 solid phase extraction (SPE) procedure yields EPAB-tagged glycans at multi-gram scales from kilogram quantities of natural material.

Fig. 2: Regeneration of free reducing glycans from 4-aminobenzoic acid derivative-tagged glycans using Oxone.
figure 2

Reaction scheme (a) and MS results (b) of free S2G2-M-OH regeneration from S2G2-M-PABA/EPAB/DEAB by Oxone. Green circles, Man; blue squares, GlcNAc; yellow circles, Gal; purple diamonds, Neu5Ac.

Full size image

For natural glycan isolation to become a viable general approach for glycan preparation, several key conditions must be met. First, the separation process must be scalable, allowing for the isolation of substantial amounts of glycans, even those with lower natural abundance. Second, the isolated glycans must achieve high purity. Third, the structures of the purified glycans need to be thoroughly characterized. Finally, the isolated glycans should be suitable for further functionalization or use in glycoconjugate synthesis. In the following section, we demonstrate that EPAB-tagged glycans, isolated on a gram scale from dry large lima beans, egg white powder, and porcine plasma powder, can be separated into highly pure, structurally defined glycans and subsequently regenerated as free reducing glycans in a streamlined process.

Dry-Load Hydrophilic Interaction Liquid Chromatography (flash-HILIC) for large-scale separation of high-Mannose Glycans from Lima Beans

While tagged glycan separation has been widely studied, most studies focus on high-performance liquid chromatography (HPLC) for analytical purposes; separation of natural glycans on a gram scale is unprecedented. HILIC separation of glycans on amino, amide, and diol HPLC columns, based on hydrophilicity, glycan size, and charge density, generally provides the best resolution. For further refinement, C18-HPLC and porous graphitized carbon (PGC)-HPLC can separate glycans with identical or similar compositions, such as isomers. While preparative NH2 and C18-HPLC columns are widely available and affordable, preparative PGC-HPLC columns are less accessible and prohibitively expensive. Additionally, complex glycan mixtures often require two-dimensional HILIC-C18 HPLC separation for adequate resolution, with HILIC as the preferred initial separation due to its higher resolution and distinct separation mechanism, simplifying downstream C18-HPLC.

However, this strategy has its limits at multi-milligram scales, as common analytical columns (4.6 × 250 mm) scale up to 50 × 250 mm preparative columns at best, representing an approximate 120-fold scale increase. The separation capacity of preparative columns is also limited by particle size; while smaller particle sizes improve resolution, they risk peak-shape distortion due to overloading. Another factor often overlooked is glycan solubility and its impact on HILIC separation. Glycans are generally soluble only in water, while HILIC-HPLC often requires high-organic starting conditions. In analytical separations, aqueous injections below a few microliters disperse quickly in the mobile phase, but at larger scales, injections of several milliliters of aqueous glycan solution can cause partial glycan breakthrough without retention and poor peak shape.

In flash chromatography, this issue is resolved by dry-loading, where samples are dissolved in a strong solvent, mixed with silica or Celite, and dried to remove the solvent. The dry sample is then packed into a dry-load cartridge for separation, minimizing the solubility issue. While straightforward in flash chromatography, this approach is technically challenging in HPLC due to the high-pressure system, which can result in leaks. Given the limitations of gram-scale glycan purification on HILIC-HPLC, we adapted dry-loading on a flash NH2 column for initial fractionation of gram-scale EPAB-tagged glycans, a method we term flash-HILIC.

Lima beans (Phaseolus lunatus, also known as butter beans), which are rich in high-mannose N-glycans47, were selected as a source for high-mannose glycan production (Fig. 3a). Following ORNG treatment, desalting, C18 flow-through, EPAB tagging, and C18-SPE purification, approximately 0.6 g of glycan-EPAB conjugates were obtained per pound of dry large lima beans. In Fig. 3b, two grams of EPAB-tagged lima bean N-glycans were dried with Celite, dry-loaded onto an 80 g NH2 column, and separated on a water/acetonitrile gradient. Unexpectedly, high-mannose N-glycan-EPABs achieved near-baseline separation based on size, from Man5 to Man9 as shown by MS (Fig. 3c). The dry-load flash-HILIC separation of EPAB-tagged glycans from egg white (Figure. S5-S8) further demonstrated the effectiveness and robustness of this separation method. While the resolution of dry-load flash-HILIC is lower than HILIC-HPLC, it is an essential and practical first-dimensional separation for gram-scale EPAB-tagged glycan mixtures.

Fig. 3: Large-scale separation of tagged high-mannose glycans from lima beans.
figure 3

a Overview of the procedure for large-scale separation of tagged high-mannose glycans from lima beans by ORNG method; b Lima beans N-glycan EPAB-conjugates flash chromatography profile on NH2 column; c MS of M5-M9 fractions from flash-HILIC purification; d Preparative HPLC separation profiles of M5-M9 fractions from flash-HILIC purification. 2 to 4 glycan structures were collected from each fractions.

Full size image

Reverse phase C18-HPLC separation of structural isomers of high Mannose N-glycans

Since HILIC separation is based on hydrophilicity, determined by the size and charge density of glycans, it generally does not effectively separate glycans with isomeric structures. High-mannose N-glycans, in particular, often occur as mixtures of isomers. In contrast, reverse-phase HPLC, such as C18 or porous graphitized carbon, is known to separate isomeric glycan structures due to subtle differences in their shapes. Therefore, we applied a second-dimensional preparative C18-HPLC to further separate high-mannose N-glycan isomers. The five distinct flash-HILIC fractions (Fig. 3b) of lima bean N-glycan EPAB conjugates were further purified using preparative C18-HPLC (Fig. 3d), resulting in 16 two-dimensional (2D) fractions, each containing a single component as confirmed by LC-MS (Fig. S1-S4) and NMR analysis (Fig. 4).

Fig. 4: NMR spectrum of 16 purified high-mannose glycan-EPAB conjugates from lima beans.
figure 4

We characterized high-mannose glycan-EPAB conjugates from lima beans using routine NMR, including the full assignment of anomeric protons. Green circles, Man; blue squares, GlcNAc; red triangles, Fuc; orange star, Xyl.

Full size image

Notably, C18-HPLC provided an orthogonal separation mechanism, with individual HILIC fractions further resolved into multiple peaks. For example, the paucimannose glycan Man3GlcNAc2XylFuc-EPAB was well-separated from Man5-EPAB despite their similar size and co-elution on flash-HILIC. The Man5 and Man6 isomers were also successfully resolved. Since a standard preparative C18-HPLC run could not fully separate the Man7 and Man8 isomers, we applied recycled C18-HPLC to achieve resolution of these isomers. A single Man9 isomer was isolated as expected, along with Man9GlcNAc1-EPAB, in which the reducing-end GlcNAc was lost during the ORNG process, as previously reported34.

This streamlined process from a single natural source yielded 16 distinct glycans, with yields up to several hundred milligrams for Man9, and can be easily repeated. These high-mannose glycan-EPAB conjugates were converted to free reducing glycans through simple Oxone treatment followed by desalting using a Sephadex G-25 column. The regenerated free reducing high-mannose N-glycans were weighed, with yields listed in Table 1.

Table 1 Isolated yields of 16 distinct high-mannose N-glycans from lima beans
Full size table

Structural Elucidation of Lima bean N-glycans using NMR spectroscopy

The substantial amounts of material obtained from the separation process allow for straightforward structural characterization of these natural glycans using routine NMR spectroscopy, without requiring specialized techniques (Fig. 4). Glycan structures can be analyzed as either glycan-EPAB conjugates or as free reducing glycans regenerated from the EPAB conjugates. Except for the innermost GlcNAc anomeric position—where distinct α- and β-signals are observed due to chemical shift differences, which are eliminated after EPAB conjugation—all other anomeric positions, which are key for glycan structure identification, are comparable between glycan-EPAB conjugates and reducing glycans.

The detailed anomeric signals and structures of these glycans, including the full assignment of anomeric protons, are presented in Fig. 4 and Table S1. The spectra of all reported structures align well with previously published literature48,49,50. Notably, several high mannose structures that are rarely reported and isolated, such as M5-F2, M5-F4, M6-F2, and M7-F4, were identified through comparison of their 1H NMR spectra with those documented in the literature. Our assumptions regarding anomeric proton shifts are grounded in well-established spectroscopic principles: typically, the anomeric proton experiences minimal chemical shift unless influenced by the addition of another saccharide. The proximity of this additional saccharide to the anomeric proton is critical; the closer the saccharide, the more pronounced the shift observed in the 1H NMR spectrum. For instance, an α2-linked saccharide typically induces a downfield shift of approximately 0.25 ppm in the 1H NMR chemical shift. This assumption is based on empirical observations and supported by published NMR data, providing a reliable framework for predicting the effects of glycosidic linkages on spectral properties. These principles guide our interpretations, helping to ensure accuracy in the assignment of complex glycan structures, although we acknowledge that deviations can occur based on the unique context of each glycan’s molecular environment.

In the M5-F2 structure, for instance, the anomeric proton assigned to Man B, observed in M6-F1, is absent, confirming the structural assignment of M5-F2 as shown in Fig. 4. Similarly, for M5-F4, the anomeric proton is also missing, and the proton at δ 5.330 is attributed to Man A α2-linked with Man D2, while the proton at δ 5.015 is assigned to Man 4, supporting the proposed structure in Fig. 4. The M6-F2 spectrum reveals that both Man A and 4 are attached with α2 mannose, indicating that M6-F2 shares the backbone structure of M5-F4 with an additional Man C. In comparing M7-F4 with M6-F3, we noted an additional α2 mannose unit and a 0.24 ppm downfield shift in the chemical shift of Man B, with minimal impact on the chemical shifts of other anomeric protons. Based on these observations, we propose the structure of M7-F4 as depicted in Fig. 4. The M7-F4 together with three Man8 structures imply the existence of 1,2-α-D-mannosidase in cis-Golgi of during N-glycan biosynthesis.

Preparation of N-Glycans from egg white powder

We extended the streamlined procedure to egg white powder, which has a considerably more complex N-glycome than lima beans. Ovalbumin, the dominant protein in egg white, is known to contain various complex N-glycan structures, including high-mannose and hybrid types, and serves as a rich source of glycans with bisecting GlcNAc. The same streamlined procedure was applied to dried egg white powder at a kilogram scale. After ORNG treatment with calcium hypochlorite, desalting by diafiltration, and C18 column purification, approximately 15–20 g of crude glycan mixture was obtained per kilogram of egg white powder.

This crude mixture was then conjugated with EPAB, and the glycan-EPAB conjugates were separated from excess tag and unconjugated glycans via C18 purification. The glycan-EPAB conjugate mixture underwent two rounds of dry-loading flash-HILIC separation using NH2-flash column, yielding 16 fractions (Figure. S5–S8). All 16 flash-HILIC fractions were subsequently purified by preparative C18-HPLC (Fig. S9–S11), resulting in 73 two-dimensional (2D) fractions, most of which contained a single component (>90% purity), as characterized by LC-MS (Table. S2).

These fractions were converted to free reducing glycans by simple Oxone treatment followed by desalting using a Sephadex G-25 column. As shown in Table 2, multiple glycans were prepared at scales ranging from low to multi-milligrams. Using this approach, we essentially achieved a parallel “synthesis” of a complex glycan library, which would be extremely challenging using any chemoenzymatic system.

Table 2 Isolated yields of N-glycans from egg white powder
Full size table

Structural Characterization of Egg White N-Glycans

A common challenge in working with natural glycans is the structural characterization of isolated glycans. In chemoenzymatic synthesis, most structural information is defined by the synthetic process, and NMR and MS characterization primarily confirm the synthesized structure51,52,53,54. However, for natural glycans, structures can be partially inferred based on known biosynthetic pathways, but many details need to be elucidated by NMR, sequential MS fragmentation, or exoglycosidase digestion. This is particularly challenging for structural isomers that differ only in the positions of one or two monosaccharides. While the structures of many glycans prepared from egg white can be inferred based on previous reports51,53,54,55,56, we have identified several complex glycan isomers not previously reported.

Fractions F29 (Hex3HexNAc5, H3N5), F33 (H3N6), and F43 (H4N6) display very similar NMR spectra, suggesting they share the same H3N5 backbone. According to a previous report51, the structure of F29, shown in Fig. 5, is one of the most abundant ovalbumin glycans. F33 contains an additional GlcNAc residue compared to F29, which is most likely placed in a β(1-4)-linkage to Man-4. This is evident from the downfield shift in the Man-4 H2 proton signal, from δ 4.152 to δ 4.196. The structure of F33 is given in Fig. 5. In F43, the additional anomeric proton at δ 4.383 (J1,2 = 7.6 Hz) indicates the presence of a β(1-4)-linked galactose residue attached to one of the terminal GlcNAc residues. A comparison of the spectra of F33 and F43 shows a downfield shift of one of the GlcNAc-7 H-1 doublets from δ 4.427 to δ 4.445 in F43, indicating that a galactose residue is linked to GlcNAc-753. Additionally, in the N-acetyl region, the signal assigned to GlcNAc-7 exhibits a small but significant upfield shift from δ 1.993 to δ 1.991, further confirming the attachment of a galactose residue to GlcNAc-753. Thus, the structure of F43 is shown in Fig. 5.

Fig. 5: NMR spectrum of selected distinct N-glycans separated from egg white powder.
figure 5

7 selected N-glycans structures and their assignment of anomeric protons, H2 protons, and NAc protons. Green circles, Man; blue squares, GlcNAc; yellow circles, Gal.

Full size image

Additionally, we analyzed the structures of fractions F71, F72, and F73, which contain minor components that do not interfere with the structural elucidation of H5N8 isomers. These fractions were digested with β-Galactosidase, yielding H3N8 glycans (Fig. S12). The resulting glycans, along with F60 (H3N8), were permethylated and analyzed by MS/MS (Fig. S13). MS/MS analysis revealed that β-Galactosidase-treated F71, F72, and F73 shared the same structure as F60. The eight GlcNAc anomeric signals were assigned based on comparison with structural reporter group signals identified for a pentaantennary oligosaccharide isolated from chicken glycoproteins. Therefore, F60 is proposed to be a pentaantennary oligosaccharide containing a bisecting GlcNAc residue, as shown in Fig. 5, which is consistent with the previous report51. The NMR spectra of Oxone-treated F71, F72, and F73 differ from that of F60, primarily due to the presence of two additional anomeric doublets between δ 4.350 to δ 4.413 (J1,2 = 7.6 Hz), indicating the presence of two β(1-4)-linked galactose residues attached to two terminal GlcNAc residues. The precise locations of these galactose residues were deduced by comparing chemical shift effects with those of fraction 60. One of the GlcNAc H-1 doublets at δ 4.424 in F60 (either GlcNAc-7 or -7’) shifted downfield to δ 4.444 in all three H5N8 fractions, indicating one galactose residue is linked to GlcNAc-7 or -7’. In the N-acetyl region, the signal assigned to GlcNAc-7 exhibited a small but significant upfield shift from δ 1.991 to δ 1.987 across all three fractions, confirming the attachment of a galactose residue to GlcNAc-7. In the spectrum of F71, the GlcNAc-7’ H-1 also shifted downfield from δ 4.424 to δ 4.444. However, no changes were observed at this position in F72 and F73, indicating that a second galactose is attached to GlcNAc-7’. For F72, the downfield shift of GlcNAc-10’ from δ 4.499 to δ 4.525 suggests the attachment of another galactose to GlcNAc-10’. Similarly, the downfield shift of GlcNAc-5 H-1 from δ 4.446 to δ 4.470 in F73 confirms the attachment of a second galactose to GlcNAc-5. The proposed structures of F71, F72, and F73 are illustrated in Fig. 5.

We also elucidated the glycan structures for all the other fractions containing one major component (Fig. S14). Determining these complex N-glycan structures presents significant challenges using only MS fragmentation. Even after permethylation, the fragmentation pattern of these glycans is overwhelmed by the simple loss of terminal GlcNAc, generating little information towards structural elucidation. However, as demonstrated above, once adequate quantities, sufficient purity, and enough variety of materials are isolated, detailed glycan structures can be readily identified using NMR, which is the gold standard for structural characterization.

Preparation of complex-type N-glycans from spray-dried porcine plasma powder

Spray-dried animal plasma has been widely used as a protein supplement in the veterinary industry and are available in multi-kilogram packages at low price. Plasma is considered as a rich source for various N-glycans, especially complex-type sialylated N-glycans. We applied the streamlined procedure to spray-dried porcine plasma (SDPP) at a multi-kilogram scale (Fig. 6). After ORNG, diafiltration-desalting, and C18 purification, crude glycans were conjugated with EPAB. C18 SPE after conjugation yielded ~1.6 g glycan-EPAB conjugates per kilogram SDPP starting material. The EPAB conjugated SDPP N-glycans were well separated into neutral and multiple charged fractions using DEAE A-25 column (Fig. 6a), as demonstrated by MS analysis (Fig. S15) and reprofiling on C18-HPLC (Fig. 6b). We targeted fractions E2 and E3 for the preparation of di- and tri-sialylated complex-type N-glycans using preparative C18-HPLC (Fig. S16). We achieved gram-scale production of the disialylated, core 6-fuctosylated biantennary glycan S2G2F. The purification is straightforward and yields ~0.2 g S2G2F/kilogram SDPP. Furthermore, we also prepared multi-milligrams of trisialylated, core-fucosylated N-glycans (yield: ~10 mg/kilogram SDPP). The conjugates are converted to free reducing glycan, and the structures are fully characterized by MS (Fig. 6c) and NMR (Table S4). To our knowledge, the “synthesis” of these complex N-glycans at this scale is unprecedented.

Fig. 6: Preparation of complex-type N-glycans from spray-dried porcine plasma powder.
figure 6

(a) Flash-DEAE separation profile of EPAB conjugated SDPP N-glycans; (b) C18-HPLC profile of E1-E5 from DEAE separation; (c) MS of purified free reducing glycan of S2G2F and S3G3F. Green circles, Man; blue squares, GlcNAc; yellow circles, Gal; red triangles, Fuc; pink diamonds, Neu5Ac.

Full size image

Discussion

In this manuscript, we aim to develop a general and streamlined method for the gram scale preparation of natural N-glycans. The first requisite is an efficient and cleavable tag for released glycans, essentially acting as a protecting group during the “reverse synthesis”. The tag should also facilitate the chromatography separation of a complex glycan mixture. Based on our previous studies, tags installed at the reducing end of glycans via reductive amination can be removed through oxidative cleavage. When NBS is used as the oxidant, N-bromination serves as a good leaving group, favoring beta-elimination and C1-C2 bond cleavage. Conversely, when Oxone is used as the oxidant, the beta-elimination process is much slower than the desired Schiff’s base hydrolysis. However, the N-hydroxyl group can act as an intramolecular nucleophile to the ortho-carboxyl group, generating a different undesired byproduct. Our rational analysis of these results suggested that 4-aminobenzoic acid and related derivatives could serve as ideal reversible tags when Oxone is used for tag removal. This was confirmed using PABA, EPAB, and procaine-tagged glycans, all of which generated intact free reducing glycans nearly quantitatively.

The selection of the EPAB tag, due to its optimal hydrophobicity for C18 retention and affordability, facilitated the acquisition of EPAB-tagged glycans at a multi-gram scale with relative ease. This larger glycan yield enabled extensive exploration of gram-scale chromatographic separation for natural glycans. Many standard analytical techniques become impractical at this scale and require adaptation. One key advancement was the use of dry-load flash HILIC on an NH2 flash column, addressing the intrinsic but often-overlooked issue of glycan solubility with HILIC loading conditions at preparative scale. Flash-HILIC fractions were further resolved using preparative C18-HPLC. Due to the orthogonal separation mechanism of the C18 stationary phase compared to flash-HILIC on NH2 stationary phase, most major fractions are obtained with sufficient purity for most applications. When needed, recycled C18-HPLC provided further resolution of closely eluting structural isomers.

Another challenge in natural glycan production is the structural characterization of purified glycans. While modern glycomics generates extensive structural data through mass spectrometry, many glycan structures remain incomplete and lack detailed resolution. For purified natural glycans to be meaningful for functional analysis, their structures must be eventually fully defined. In this study, the availability of large quantities of glycans enabled us to perform NMR analysis to unambiguously determine their complete structures, including complex tridecasaccharide isomers (Hex5HexNAc8). We anticipate that as more glycans are collected and NMR data expands, structural analysis of glycans by NMR will become increasingly straightforward and potentially automated with artificial intelligence.

By using reversible tags as reducing-end protecting groups, we developed a general and streamlined procedure to prepare libraries of free reducing natural N-glycans. Applying this method to various abundant natural sources, we obtained a comprehensive library of complex N-glycans at multi-milligram to gram scales—yields unattainable through current synthetic approaches. This study effectively addresses common challenges in natural glycan isolation, including the low abundance of minor components, heterogeneity, purity limitations, and complex structural characterization. Most of the glycans obtained in this study have not been synthesized at a comparable scale.

The approach described above represents a “reverse” synthesis of complex natural glycans, utilizing glycoprotein mixtures as starting materials and 4-aminobenzoic acid derivatives as a single protecting group. Table 3 compares three strategies for complex glycan production: chemical synthesis, enzymatic synthesis, and “reverse” synthesis. Each method has unique advantages and challenges, making them complementary to one another. With the recent advancements in ORNG, the development of removable tags, and large-scale chromatography for separation and structural characterization, we believe the “reverse” approach will substantially expand the chemical space of biomedically relevant glycans and glycoconjugates. It is certainly conceivable to use enzymatic modification to further expand the repertoire of complex glycans using the abundant natural glycans as substrates.

Table 3 Comparison of three strategies for complex glycan production
Full size table

Conclusion

In this report, we present a streamlined procedure utilizing commonly available 4-aminobenzoic acid derivatives as quantitatively reversible tags for the large-scale preparation of natural glycans. We demonstrate that a library of complex natural glycans, from milligram to gram scale, can be efficiently prepared through a streamlined procedure of ORNG, tagging, flash-HILIC, preparative C18-HPLC, and de-tagging. These purified glycans are amenable to full structural characterization. This scalable, efficient approach to generating a natural glycan library opens the door to the broader use of natural glycans in functional studies within glycoscience.

Methods

General labeling procedure with PABA, EPAB, or Procaine for small-scale glycan analysis

Glycans (0.1–1 mg) were lyophilized in a 1.5 mL Eppendorf tube. PABA, EPAB, or procaine (0.35 M) was prepared by dissolving the labeling reagent in a DMSO/acetic acid (7:3, v/v) mixture (Solution A). Separately, NaCNBH3 (1 M) was dissolved in the same DMSO/acetic acid mixture (Solution B). To the lyophilized glycans, 50 µL of Solution A and 50 µL of Solution B were added, followed by incubation at 65 °C for 2 hours. After the reaction, 1 mL of acetonitrile was added to the mixture, which was briefly shaken and then centrifuged at 15,000 g for 2 minutes. The supernatant was discarded, and the pellet was dissolved in 100 µL of water. The labeled glycans were analyzed using HPLC and LC-MS.

Labeling procedure with EPAB in large scale

Glycans were dried in a round-bottom flask and dissolved in a DMSO/acetic acid (7:3, v/v) mixture to a concentration of approximately 50 mg/mL. EPAB and NaCNBH3 were added to reach final concentrations of 50 mg/mL each. The solution was stirred in a 70 °C water bath for 3 hours. After the reaction, the mixture was poured into 10 volumes of acetonitrile, then centrifuged at 3000 g for 10 minutes. The supernatant was discarded. The pellet was resuspended in 3 volumes of acetonitrile, followed by centrifugation at 3000 g for another 10 minutes. The supernatant was discarded again, and the pellet was dissolved in water, dried, and reconstituted in water for subsequent glycan separation.

Tag removal procedure

PABA, EPAB, or procaine-tagged glycans were dissolved in 5% trifluoroacetic acid (TFA) in water at a concentration of 1–20 mg/mL. Oxone was then added to achieve a concentration of 10–50 mg/mL, and the mixture was incubated at room temperature for 1 hour. After incubation, the solution was centrifuged to remove precipitates. The resulting supernatant was loaded onto a Sephadex G-25 column and eluted with water. Fractions were collected, stained with orcinol, and/or analyzed by mass spectrometry (MS). Fractions containing glycans were lyophilized for further analysis.

Release of N-glycans from lima beans

3.63 kg of dry large lima beans were homogenized with 16 L water using a commercial Waring blender in multiple runs. The homogenate was mixed with 30 L of water with an overhead stirrer, followed by the addition of 10 L of ice. Next, 0.91 kg of calcium hypochlorite powder was added, and the mixture was stirred for 20 minutes. Afterward, 100 mL of 1-octanol and 300 g of sodium metabisulfite were added, with stirring continued for 10 minutes. Then pH was adjusted to approximately 10 using 2 M sodium hydroxide solution, and the mixture was stirred for an additional 10 minutes. The suspension was centrifuged in batches at 3000 g for 15 minutes, and the supernatants were collected and pooled, while the pellets were discarded. The pooled supernatant was then reduced in volume and desalted via tangential flow filtration using a 600–800 Da membrane. The filtrate was passed through a 3 L Sephadex A-25 DEAE column, washed with 5 L of water, and the flow-through and wash fractions were collected. These were subsequently passed through a 1500 g C18 flash column, washed with 10 L of water, and the resulting material was collected and dried. Yield: 1.6 g/ pound dry lima bean. This material is ready to be conjugated with EPAB following the described procedure.

Separation of lima beans N-glycan EPAB Conjugates

EPAB-tagged high mannose glycans from lima beans were dissolved in water and passed through a C18 flash column. The glycans were eluted using 25% acetonitrile, and the eluate was dried. The dried mixture was ground into a fine powder and mixed with four times the amount of celite powder. The resulting mixture was loaded into a Combiflash dry load cartridge and separated using a Claricep NH2 flash column (20–35 µm) with a 15–25% acetonitrile gradient. Fractions were collected and, if necessary, re-separated. The fractions were then dried and further purified by preparative C18-HPLC (Sunfire C18, 30 × 150 mm, 5 µm) using a mobile phase of 12% acetonitrile with 0.1% trifluoroacetic acid (TFA). The collected fractions were lyophilized for further analysis.

Release of N-glycans from egg white powder

2 kg of egg white protein powder were suspended in 40 L of water using an overhead stirrer, followed by the addition of 15 L of ice. 1.4 kg of bleach powder were added, and the mixture was stirred for 20 minutes. Subsequently, 150 mL of 1-octanol and 0.45 kg of sodium metabisulfite were added, and the solution was stirred for an additional 10 minutes. The pH was adjusted to approximately 10 using 2 M sodium hydroxide, and the mixture was stirred for another 10 minutes. The suspension was centrifuged in batches at 3000 g for 15 minutes, and the supernatants were collected and pooled, while the pellets were discarded. The volume of the pooled supernatant was reduced, and the solution was desalted using tangential flow filtration with a 600–800 Da membrane. The filtrate was passed through a 3 L Sephadex A-25 DEAE column, washed with 5 L of water, and the flow-through and wash fractions were collected. These fractions were then passed through a 1500 g C18 flash column, washed with 10 L of water, and the resulting material was collected, dried, and conjugated with EPAB as described.

Separation of egg white N-glycan EPAB conjugates

EPAB-tagged egg white glycans were dissolved in water and passed through a C18 flash column. The glycans were eluted with 25% acetonitrile, and the eluate was dried. The dried material was ground into a fine powder and mixed with four times the amount of celite powder. This mixture was loaded into a Combiflash dry load cartridge and separated using a Claricep NH2 flash column (20–35 µm) with a 15–25% acetonitrile gradient. Fractions were collected and, if necessary, re-separated. The fractions were then dried and further purified by preparative C18-HPLC (Sunfire C18, 30 × 150 mm, 5 µm) using a mobile phase of 12% acetonitrile with 0.1% trifluoroacetic acid (TFA). The collected fractions were lyophilized for further analysis.

Preparation of N-glycans from spray-dried porcine plasma (SDPP)

The ORNG treatment, glycan enrichment and EPAB-tagging of SDPP N-glycans are essentially the same as described for egg white glycans. After C18 purification, 8.6 g of EPAB-tagged glycans were obtained from 6 kg of SDPP. This material was dissolved in water and loaded onto a DEAE A-25 column (2.5 x 20 cm), washed with water, and then eluted with a step-gradient of ammonium acetate buffer (25 mM, 50 mM, 100 mM, 200 mM, and 500 mM). Five fractions were collected and dried. Fractions E2 and E3 were further separated on preparative C18-HPLC, and the major peaks were collected and characterized by MS. The purified EPAB conjugates are de-tagged as described above. From 6 kg of SDPP, 857 mg biantennary and 60 mg triantennary sialylated and fucosylated N-glycans were yielded.

Galactosidase treatment

To a 10 µL solution of glycan-EPAB (1 mg/mL in water), 1 µL of sodium acetate buffer (1 M, pH 5) and 0.5 µL of galactosidase (10 mg/mL in water) were added. The reaction mixture was incubated at room temperature, with progress monitored by mass spectrometry (MS) analysis.

Permethylation of Glycans for MS analysis

Glycans (10–20 µg) were lyophilized. Sodium hydroxide beads were ground in anhydrous DMSO (~100 mg/mL) to create a slurry. To the lyophilized glycans, 200 µL of the sodium hydroxide slurry and 50 µL of methyl iodide were added, and the mixture was shaken for 30 minutes. Afterward, the mixture was centrifuged at 5000 g for 1 minute, and the supernatant was transferred to a new tube. 500 µL of water and 700 µL of chloroform were added to the supernatant, followed by mixing and centrifugation at 5000 g for 1 minute. The top layer was discarded. This washing step with 500 µL of water was repeated three times, discarding the top layer each time. The bottom layer was air-dried and then redissolved in 20–50 µL of 50% acetonitrile. The resulting solution was injected for MS and MS/MS analysis.