Preparative Scale Purification of Natural Glycans by Closed-loop Recycle HPLC

Abstract

While glycoscience has become well recognized as an indispensable area in biomedical research, studies on the function of individual glycans remains a great challenge due to the lack of tools and methods. One of the greatest impediments to progress in this area is the lack of biomedically relevant complex glycans in sufficient quantity and purity for structural and functional analysis. Despite recent advances in chemoenzymatic synthesis of complex glycans, generating significant amounts of pure glycans is limited to laboratories with specialized expertise. We have previously reported the oxidative release of natural glycans (ORNG) using household bleach, which provides large quantities of biologically relevant glycans that can be a source of glycans in quantities (>mg scale) suitable for functional studies. However, the preparative scale separation of complicated glycan mixtures has not been studied due largely to the fact that gram quantities of starting glycans have not been available until now. Here we report the adoption of closed-loop, recycle HPLC to resolve closely related glycan structures, including complex glycan isomers at preparative scale (10-100 mg).

Graphical Abstract

In recent years, glycoscience, the study of chemistry and biochemistry of glycans or carbohydrates, has attracted great attention owing to more and more biological pathways being discovered to involve glycans or glycoconjugates.[1-3] Glycans play important roles in normal physiology and disease, including protein folding, cell adhesion, and host-pathogen interaction. Despite its recognized importance in biomedical study, glycoscience lags far behind the study of other major biomolecules such as nucleic acids and proteins/peptides, largely due to the structural complexity of glycans.[4] While the biosynthesis of nucleic acids and proteins are template-driven based on genetic codes, the biosynthesis of glycans can be co-translational or post-translational and regulated by multiple glycosyltransferases with distinct substrate and linkage specificity.[5] As a result, structures of glycans are defined by many features, including monosaccharide composition and sequence, linkage position and anomeric stereochemistry, unlike linear sequences of nucleic acids and proteins which define their structures. Thus, the structural complexity of glycans brings unique challenges to glycoscience. While high throughput sequencing of nucleic acids and proteins/peptides through next generation sequencing platforms and modern mass spectrometry (MS) based proteomics has become routine practices readily available through commercial vendors and widely available core facilities, sequencing of glycans is still low throughput and involves special expertise limited to analytical glycoscientists.[6] Similarly, while automatic synthesis of nucleic acid and peptides and recombinant expression of proteins are all widely available commercially, the synthesis of complex glycans is still a laborious and costly process limited to a small number of laboratories.[7] The synthetic challenges greatly limit the generation of a large number of biomedically relevant glycans, which are essential for functional studies and also serve as structural standards for the development of advanced sequencing techniques.

Despite the limited access to complex glycans, the functional study of glycans has advanced significantly in the last decades.[8] Since functions of many glycans are realized through their interactions with glycan binding proteins (GBPs), glycan microarray, in which multiple glycan structures are immobilized onto a microscope slide for interrogation with fluorescently labeled GBPs, has become the most successful platform for screening protein-glycan interactions and generating hypotheses regarding glycan functions.[9, 10] The effectiveness of a glycan microarray, however, is dependent on the size, diversity, and biological relevance of the corresponding glycan library. As the synthetic approach is yet to meet the need for more complex glycans, nature is an alternative source of biomedically relevant complex glycans often occurring as heterogeneous mixture attached to other biomolecules. These glycans can be released from glycoproteins and/or glycolipids, tagged with chromophore, and separated by chromatography. Due to their low natural abundance and absence of methods from amplification, it is necessary to work with large amounts of starting material to obtain significant quantities of glycans, which is not always practical using traditional enzymatic release or harsh chemical conditions. We have developed oxidative release of natural glycans (ORNG), which enabled us to process kilogram scale animal/plant tissues directly to obtain gram quantities of natural glycans.[11-13] To resolve these glycans into individual glycans with significant purity, they need to be separated by multi-dimensional high performance liquid chromatography (HPLC).[4, 14] While many separation techniques have been developed for glycomics study using analytical HPLC systems, the preparative purification of natural glycans remains a less explored area since gram scale complex glycans are only now available. Due to the high heterogeneity of natural glycans, especially widely existing isomers, it is very challenging to separate pure individual components. This problem is seriously aggravated when the preparative scale is applied. To address this problem, special HPLC techniques are needed. In this manuscript, we report the separation and purification of multi-milligram amounts of natural glycans generated by our oxidative release method and derivatized with bifunctional tags [11] using closed-loop recycle HPLC on several commonly used stationary phases.

RESULTS

Alternative and Closed-loop Recycle HPLC.

To increase the theoretical plate number of a column and then to achieve separation of closely related compounds, many parameters can be optimized, including stationary phase particle size and column length. While increasing column length directly increases column plate number, it is associated with higher pressure and higher manufacturing cost and becomes impractical at a certain length. To circumvent this limitation, recycle HPLC systems have been designed so that partially separated compounds can be re-injected back into the column for further separation. Repeated recycling essentially extends the column length many times without increasing system pressure and resolve hard-to-separate compounds. Two common recycle HPLC systems, alternative and closed-loop, have been designed involving different mobile phase flow (Fig. 1).[15, 16] Alternative recycle HPLC uses a multiport two position valve to control solvent flow through two identical columns alternatively.[15, 17] Closed-loop recycle HPLC simply uses a 3-way valve to re-direct mobile phase back into pump.[18] While alternative recycle HPLC offers higher resolution and lower diffusion, it generates higher system pressure and consumes a significant amount of solvents. In closed-loop recycle HPLC, the mobile phase is redirected back into the pump, which introduces more dead volume into the system. To reduce this negative effect, the pump head and flow path need to be specially designed to remove as much dead volume as possible. Nevertheless, multiple cycles of recycling can still be used to improve the separation. It is important to note that the closed-loop recycle system conserves solvents and extends column life, which are essential for preparative scale HPLC separation.

Figure 1.

Two mechanical designs of recycle HPLC. a) alternative recycle HPLC. a 10-port/2 positive valve is installed between two columns. Upon switching valve positive, eluent from one column can be reinjected to the other column and this process can be repeated until desired resolution is reached; b) closed-loop recycle HPLC. Eluent from column is simply redirected back to pump and reinjected to the column.

In our preliminary study, we tested alternative recycle HPLC on the purification of several natural glycans previously isolated by two-dimensional HPLC. While alternative recycle HPLC clearly resolve hard-to-separate isomers, we adopted closed-loop HPLC system for preparative scale separation. A recycle HPLC kit from Shimadzu was used to easily modify an existing semipreparative pump to build a closed-loop recycle HPLC system, which reduces the void volume introduced to each cycle as the mobile phase flows through the pump. Due to the nature of recycle HPLC, only isocratic elution is possible. The normal mode and recycle mode can be easily controlled by switching the 3-way valve installed between the detector and pump inlet (Fig. 1).

During recycle mode, solvents flow through the column in the closed-loop continuously, resolving closely eluted compounds as monitored by the detector. Once a satisfactory resolution is reached, the 3-way valve is switched back to normal mode for fraction or waste collection. One concern is that the yiled of an analyte might diminish after multiple cycles in the column. To address this question, we examined the recovery yield of lactose-AA conjugate after 10 cycles using recycle RP-HPLC (Fig. S1). The recovery yield of 92.8% suggests that the loss during recycle HPLC run is minimum. Using this system, we explored preparative scale recycle HPLC purification of natural glycans using several combinations of stationary/mobile phases.

Closed-loop Reverse Phase Recycle HPLC on C18 Column.

As a proof-of-principle test of the preparative scale closed-loop recycle HPLC system in the separation of glycans, we examined the separation of mixtures of monosaccharide-anthranalic acid (AA) conjugates. When a mixture of Galactose (Gal)-AA/mannose (Man)-AA (1/1) were injected into the system (Fig. 2a), no apparent separation was observed without recycling. However, two distinct peaks were apparent after the 2nd cycle. These peaks are gradually resolved when more cycles are carried out. After 10 cycles, Gal-AA and Man-AA were well-separated from each other and baseline separation were achieved after several more cycles. To further explore the separation by recycle HPLC, we injected a mixture of three components: Gal-AA/Man-AA/Glucose (Glc)-AA. Interestingly, Glc-AA is well separated from Gal-AA/Man-AA even without recycling. When more cycles of separation are carried out, better resolution among the three components was achieved. Importantly, starting from the 6th cycle, Glc-AA conjugate peak (at 171 min) is merging with Gal-AA conjugate peak (at 174 min) eluted in 7th cycle. In 8th recycling circle, peaks from Glc-AA and Gal-AA overlapped with each other completely. While this seems to be a problem, in practice, the well-separated Glc-AA can be released from the system during the run by simply switching the 3-way valve (Fig. 1b) from recycle mode to normal mode during 85-95min, collecting the puridied Glc-AA, while Gal-AA/Man-AA stay in the column for further separation. This “shaving” process can be easily calculated based on the retention time differences between each cycle for individual components, which are generally very consistent.

Figure 2.

Closed-loop recycle HPLC separation of monosaccharide-AA conjugate mixtures: a) Gal-AA(1)/Man-AA(2); b) Gal-AA(1)/Man-AA(2)/Glc-AA(3). Column: Luna C18(2) 21.2 x 250mm; Mobile phase: 8% acetonitrile/0.1% TFA; Flow rate: 12 mL/min. Injection amount: 1 mg each in 1mL.

We then applied the closed-loop recycle HPLC to the separation of complex natural glycans. During the development of shotgun glycomics,[19] we have used multi-dimensional HPLC for the separation of natural glycans released from cells/tissues/organs. While most glycans can be fractionated to sufficient purity (>90%) for microarray printing, poorly resolved isomeric mixtures are not uncommon. At analytical scale (10-100 μg), long separation time and repeated separation can be used to eventually resolve these glycans to >95% purify with significant loss. However, this becomes impractical at preparative scale (10-100 mg). With the development of oxidative release of natural glycans (ORNG)[11] where gram quantities of glycans are in the process stream, the need to develop preparative scale purification is necessary. We found multi-dimensional HPLC remains an effective way to resolve and purify glycans from gram scale of natural glycan mixture, using appropriate tagging strategy and various separation mechanisms.[20] Nevertheless, even after multidimensional HPLC, some fractions contain mixtures of isomers that are very difficult to separate. We envisioned that the closed-loop recycle HPLC could be an ideal solution for this problem. Using similar conditions, we applied this approach to the separation of a Man8GlcNAc2-AA (Man8-AA) fraction obtained from egg yolk using ORNG followed by 2D HPLC (Fig. 3). The starting material of Man8-AA was a single peak by analysis on C18 analytical HPLC (Fig. 3a, insert), although the peak shape and width indicated that it could be a mixture of structures [12, 20-22]. Upon preparative closed-loop recycle HPLC separation, it was detected as a single peak on the first cycle. However, the single peak started to split into two peaks after only 3 cycles (Fig. 3a, retention time 60~70 min). After 6 recycling steps, another two minor peaks became visible (Fig. 3a, labeled with black arrows), which were discarded during the 8th and 9th cycles respectively to avoid later contamination with the major peaks. Because of the inherent diffusion in the HPLC system, there was an expansion in the peak width while the recycling continues. At the 13th cycle, the two major components are sufficiently separated for collection as pure fractions without losing significant amounts of material. While the MALDI-TOF-MS spectra of the starting material and the two separated components (Fig. 3c) all matched the Man8-AA molecular weight, the MS/MS analysis revealed distinct pattern of fragmentation of the isomers. While Man8-AA-1 showed fragmentation peaks at m/z 1036.9 and 1216.6, Man8-AA-2 showed a different set of fragmentation peaks at m/z 1054.6 and 1198.4. Their structures were therefore assigned as shown in Fig. 3b. Interestingly, the fragmentation pattern of the starting Man8-AA mixture is also shown as a mixture of the fragmentation patterns of the two separated isomers.

Figure 3.

Closed-loop recycle HPLC separation of two Man8-AA isomers on C18 column. a) The recycle HPLC profile; Insert: the analytical HPLC profile of starting mixture of Man8-AA isomers; b) Deduced structures of two Man8-AA isomers; c) MALDI-TOF-MS of starting mixture (top) and the two separated isomers (middle and bottom); d) MALDI-TOF-MS/MS profiles of the starting mixture (top) and the two separated isomers (middle and bottom). Column: Luna C18(2) 21.2 x 250mm; Mobile phase: 9% acetonitrile/0.1% TFA; Flow rate: 10 mL/min. Injection: 12 mg in 1mL water. Yield: Man8-AA-1: 6.5 mg (3.53 μmol) and Man8-AA-2: 1.8 mg (0.98 μmol) (>70%).

As a result, this closed-loop HPLC recycling procedure enabled us to separate and identify the accurate structures of Man8-AA isomers and purify 6.5 mg (3.53 μmol) and 1.8 mg (0.98 μmol) of 1 and 2 from 12 mg (6.51 μmol) starting mixture, respectively (>70%yield).

Preparative Scale Closed-loop Recycle Hydrophilic Interaction Liquid Chromatography (HILIC) HPLC on Amino Column.

HILIC mode HPLC is used to provide efficient separation of polar compounds on a polar stationary phase.[23, 24] HILIC-HPLC on amine columns has been widely applied to the separation of glycans, mostly based on their size and charge.[25, 26] To demonstrate the compatibility of HILIC-HPLC with closed-loop recycling HPLC, we explored the separation of a glycan fraction from egg yolk glycan-AA conjugates partially purified by 2D-HPLC (Fig. 4). This fraction contains three glycans with compositions H8N1, H9N1 and H10N1 (Fig. 4b, top) and shows a major peak with two shoulder peaks on analytical HILIC-HPLC (Fig. 4a, insert). When this mixture was applied to a closed-loop recycle HPLC on amine column, three peaks were well-separated after 6 cycles (Fig. 4a). The three peaks were collected and reanalyzed on MALDI-TOF, which showed three distinct compositions at high purity (Fig. 4b). The proposed structures of the three glycans were proposed to be Man8-AA, Man9-AA and Man9Glc-AA in Fig. 4c. While Man9-AA and Man9Glc-AA are assumed to be single components, Man8-AA may contain isomers that require reverse phase recycle HPLC for further purification.

Figure 4.

Closed-loop HPLC separation of Man8-AA, Man9-AA and Man9Glc-AA. a) Closed-loop recycle HPLC profile of the mixture. Insert: Analytical HILIC-HPLC profiles of the starting mixture; Column: Luna NH2 10μm, 21.2 x 250mm; Mobile phase: 48% acetonitrile/52% water with 60 mM ammonium acetate/0.6% acetic acid; Flow rate: 12 mL/min. Injection amount: 50 mg in 1mL water. b) MALDI-MS profiles of the starting mixture and the three fractions collected from recycle HILIC-HPLC; c) The assigned structure of the three purified glycans.

Preparative Scale Closed-loop Recycle Size Exclusive Chromatography (SEC)-HPLC.

We then further explored the closed-loop recycle HPLC on the separation of glycans of different sizes using Size exclusion chromatography (SEC). SEC has been a traditional way to fractionate glycans, such as human milk (HM) glycans. However, complete separation of glycans based on size difference is fairly challenging, even with very long columns. Based on the success with RP and HILIC modes, we reasoned that the application of recycle HPLC on SEC may greatly improve its resolution. Using a previously isolated human milk glycan fraction containing mostly neutral tetra-to hexasaccharides[27, 28], we demonstrated the improvement of resolution in SEC with a preparative PL aqua gel-OH 20 (100Da to 20kDa range) column (Fig. 5). As shown in Fig. 5a, resolution of three major peaks was continuously improved with more cycles through the column, eventually, nearly baseline separation of the three major peaks were achieved and the three fractions were collected and analyzed by MALDI-MS, each of which showed a distinct m/z without nearly no crossover (Fig. 5b). The monosaccharide compositions of the three major fractions are assigned to be H3N1F2, H3N1F, and H3N1. While each fraction surely contains multiple isomers that need to be separated by other methods, the complete separation of glycans with a single fucose difference using a simple recycle HPLC system is remarkable.

Figure 5.

Closed-loop recycle HPLC separation of a human milk glycan-AEAB conjugates neutral fraction. a) Closed-loop recycle HPLC profile; Flow rate: 4mL/min; Column: PL aquagel-OH 20, 25 x 300mm; mobile phase: 0.2M ammonium bicarbonate. UV 280nm is used to avoid detector signal saturation. b) MALDI-MS profiles of the three separated fractions. H: Hexose; N: HexNAc, F: Fucose. Injection: 100 mg in 2mL water.

DISCUSSION

Here we report the exploration of a simple and affordable closed-loop recycle HPLC system for the preparative scale separation/purification of complex glycans. While recycle HPLC has been known for many years, the application of this technique to the separation of glycans has been scarce. Recently alternative recycle-HPLC has been applied to the separation of glycans on an analytical scale for the analysis of glycan purity[15, 17]. Since our oxidative release method can generate process streams containing grams of released glycans, we are developing preparative methods for their purification that can be applied at a commercial scale. In our attempt to use alternative recycle HPLC for this purpose, we encountered high system pressures and the need for large volumes of expensive solvents. We, therefore explored the closed-loop recycle HPLC, which we found to be a system more compatible with preparative scale separation of natural glycans at the multimilligram scale. For closed-loop recycle HPLC, the eluate from column is reintroduced into the pump and back through the column for a continuous flow resulting in higher resolution due to the elimination of the dead volume and subsequent sample diffusion during separation.

We have reported using preparative scale multi-dimensional HPLC method to collect N-glycans from soy proteins and egg yolks[12, 29]. Many of the fractions we generated from preparative scale multi-dimensional HPLC are a mixture of closely related structures showing no or insufficient separation (Fig. 3a, 4a, inserts) even on analytical columns. Although recycle HPLC is not suitable to separate a complex mixture due to potential re-mixing of fractions during recycle, it can be used to further polish fractions isolated from an initial separation by preparative gradient HPLC. When these fractions are subjected to recycle HPLC, they are well separated so that the collection of high purity glycans is possible (Fig. 3a, 4a). Recycle HPLC can be applied to all the separation mechanisms examined here, including reverse phase on C18 column, HILIC on amino column, and SEC. Importantly, the separation of isomers on the C18 column represents an orthogonal method to the separation of glycans based on size and charge on amino column.

The combination of the two separation mechanisms meets nearly all of our need to purify large complex natural glycans. Furthermore, the recycle HPLC on SEC column showed a great separation of human milk glycans that differ by a single fucose residue in size. This will provide another orthogonal separation mechanism to resolve a complex mixture of glycans before RP and/or HILIC separation. It is also worth note that although we only discussed glycan-AA and AEAB conjugates, the recycle HPLC mechanism can be applied to other glycoconjugates with different tags [21].

One inherent drawback of recycle HPLC is the time needed for multiple cycles, which are necessary to achieve satisfactory separation. Despite this, we believe closed-loop recycle HPLC provides a major advancement in the preparative separation of glycans and will enable the purification of significant amounts (10-100 mg) of natural glycans for the study of their detailed biological functions.

EXPERIMENTAL

All chemicals and HPLC solvents were purchased from Sigma-Aldrich, Acros, Oakwood Chemicals, and Fisher Scientific. Milli-Q water was used to prepare all aqueous solutions. C18 Sep-pak was from Waters, Inc. Sodium hypochlorite solutions are from Pure Bright (6% NaOCl), and prepared freshly by addition of water.

Mass Spectrometry (MS).

A Bruker Daltonics Ultraflex-II MALDI-TOF/TOF system and an anchorchip target plate were used for MS analysis. Reflective positive mode was used for glycans. 2,5-dihydroxybenzoic acid (DHB) (10 mg/mL in 50% acetonitrile with 0.1% trifluoroacetic acid) was used as matrix.

High-performance Liquid Chromatography Analyses.

A Shimadzu HPLC CBM-20A system with LC-20AT pump, UV detector SPD-20A and fluorescence detector RF-10Axl was used for analytical HPLC profiles. A Shimadzu HPLC CBM-20A system with LC-20AR semi-preparative pump and UV detector SPD-20A was converted to recycle HPLC system with Shimadzu’s recycle HPLC kit (including pump heads, switching valve and other connection components). UV absorption at 330 nm or fluorescence at 330 nm excitation and 420 nm emission was used for detection of anthranilic acid (AA) tag. Phenomenex amino columns were used for normal phase HPLC. Phenomenex C18(2) columns were used for reverse-phase HPLC separation. PL aquagel-OH 20 column (25 x 300mm) was used for SEC-HPLC. HPLC conditions are described in figure legends.

Recycling HPLC Operation.

Recycle HPLC is installed with a manual injector (2mL sample loop). Before injection of sample, the recycle valve is open so that the elution goes to waste. After sample injection, the recycle valve is kept open. When the peak of interest emerges, the recycle valve is closed. When the peak of interest is eluted, the recycle valve is open again to release other impurities. The recycle valve is closed again when the peak of interest emerges at the 2nd cycle and kept closed for several cycles until satisfactory resolution is achieved. Then the recycle valve is open and peaks are collected manually. Flow rate from 10-15 mL/min for column size 21.2 x 250 mm was used. While higher flow rate generates higher pressure in the HPLC system, lower flow rate increases run time per cycle.

• Closed-loop recycle HPLC is adopted for glycan separation for the first time.

• Closely related glycan structures including isomers can be well resolved.

• Preparative scale purification of glycans (10-100mg) can be achieved.

• Multiple separation mechanisms, including reverse phase (RP), Hydrophilic interaction chromatography (HILIC) and size exclusive chromatography (SEC) are compatible.