Streamlining assays of glycosyltransferases activity using in vitro GT-array (i-GT-ray) platform: Application to family GT37 fucosyltransferases

Abstract

Numerous putative glycosyltransferases (GTs) have been identified using bioinformatic approaches. However, demonstrating the activity of these GTs remains a challenge. Here, we describe the development of a rapid in vitro GT-array screening platform for activity of GTs. GT-arrays are generated by cell-free in vitro protein synthesis and binding using microplates precoated with a N-terminal Halo- or a C-terminal GST-tagged GT-encoding plasmid DNA and a capture antibody. These arrays are then used for screening of transferase activities and the reactions are monitored by a luminescence GLO assay. The products formed by these reactions can be analyzed directly from the microplates by mass spectrometry. Using this platform, a total of 280 assays were performed to screen 22 putative fucosyltransferases (FUTs) from family GT37 (seven from Arabidopsis and 15 from rice) for activity toward five acceptors: non-fucosylated tamarind xyloglucan (TXyG), arabinotriose (Ara₃), non-fucosylated rhamnogalacturonan I (RG-I), and RG-II from the mur1-1 Arabidopsis mutant, and the celery RG-II monomer lacking Arap and MeFuc of chain B and l-Gal of chain A. Our screen showed that AtFUT2, AtFUT5, and AtFUT10 have activity toward RG-I, while AtFUT8 was active on RG-II. Five rice OsFUTs have XyG-FUT activity and four rice OsFUTs have activity toward Ara₃. None of the putative OsFUTs were active on the RG-I and RG-II. However, promiscuity toward acceptors was observed for several FUTs. These findings extend our knowledge of cell wall polysaccharide fucosylation in plants. We believe that in vitro GT-array platform provides a valuable tool for cell wall biochemistry and other research fields.

All cells have an extracellular matrix constructed from various types of carbohydrates and glyco-conjugates. These glycans are assembled by enzymes referred to as glycosyltransferases (GTs) (1). A large number of putative GTs have been identified using bioinformatic approaches. The carbohydrate-active enzymes database (CAZy, www.cazy.org) currently classifies GTs into 117 families (as of November 2023) and also contains many GTs assigned to no family at all. However, only a few members in each family have had their biochemical function(s) confirmed experimentally. For example, the GT2 family, which includes synthases involved in the synthesis of cellulose, chitin, mannan, mixed-linkage-glucan, and glycosylated secondary metabolites, contains 10,731 proteins from eukaryotes, of which, less than 2% have been functionally characterized. This is due in large part to several technical difficulties. Most eukaryotic GTs are integral membrane proteins and are difficult to purify (2, 3). The amounts of GT proteins and their corresponding mRNAs are low in cells. Predicting the donor/acceptor specificities of GTs from their sequences alone is not reliable. In addition, high-throughput tools for screening of GT activities are lacking.

The extracellular matrix of plants is referred to as the cell wall. This wall is considered to be a dynamic structure and have an essential role in plant growth and development, as well as in a plants ability to adapt to changing environmental conditions (4, 5). Cellulose, hemicellulose, and pectin account for the bulk of the polysaccharides present in the primary wall that surrounds growing plant cells. The main monosaccharides used to form these polysaccharides are glucose (Glc), galactose (Gal), arabinose (Ara), galacturonic acid (GalA), xylose (Xyl), and rhamnose (Rha). Lower amounts of fucose (Fuc, 6-deoxy-L-Gal), methyl fucose, apiose (a branched-chain pentose with a tertiary alcohol), methyl Xyl, aceric acid (a branched acidic hexose), glucuronic acid (GlcA), 3-keto-d-manno-octulosonic acid (Kdo), and 3-keto-d-lyxo-heptulosaric acid (Dha) are also present (6). Fucosylated cell wall polysaccharides are believed to have a role in many physiological processes. For example, virtually no Fuc is present in the walls of the Arabidopsis mur1 mutant, which has lost the ability to produce GDP-l-Fuc, the activated sugar donor used by fucosyltransferases (FUTs). These walls have reduced mechanical strength, which leads to reduced growth (7, 8, 9). Cell wall fucosylation may be associated with tolerance to freezing temperatures (10), influences control of leaf water loss, and alters stomatal development and mechanical properties (11). Fucosylation may also have a role in the formation of lignified tissue through activation of jasmonic acid signaling pathway (12). Many of these effects have been attributed to reduced fucosylation of cell wall pectins and the inability of the plants to form normal amounts of the borate crosslinked rhamnogalacturonan (RG) dimer (9).

It has been estimated that at least 67 different GTs including several FUTs are required for pectin biosynthesis (13). These FUTs add Fuc residues to RG-I and RG-II. Fuc is α(1,2)-linked to Gal in type-I and type-II arabinogalactan (AG) side chains (14). In RG-II, MeFuc is α(1,2)-linked to Gal in side chain B, whereas the Fuc in side chain A is α(1,4)-linked to Rha (13, 15). No FUTs specific for RG-I and RG-II have been identified. The CAZy database classifies FUTs into 7 GT families according to the type of linkage formed by these enzymes (16). Families GT37 and GT11 contain mostly α(1,2) FUTs. Family GT10 contains α(1,3)/(1,4) FUTs. Family GT23 contains α(1,6) FUTs. Protein O–fucosyltransferases (POFUTs) have been assigned to families GT41, GT65, and GT68. Only FUTs from families GT10, GT37, GT41, and GT65 are present in plants. The only plant cell wall polysaccharide FUTs with known functions are from family GT37. Arabidopsis AtFUT4 and AtFUT6 catalyze fucosylation of AG proteins (AGPs) (17), whereas Arabidopsis AtFUT1 (18), pea (PsFUT) (19), and rice OsMUR2 (Os02g0764200) (20) catalyze fucosylation of XyG. It has been suggested that all the putative FUTs of family GT37 catalyze the formation of a α(1,2)-linkage in plant cell wall polysaccharides. Advances in bioinformatic based on evolutionary relationships (phylogeny) (21, 22, 23, 24) have assisted functional predictions for many GTs. However, this approach failed in predicting the function or substrate specificity of the FUTs in family GT37 because they cluster in clades by species rather than by similarity of enzyme activity (20, 25, 26, 27). Thus, the function of a FUT from GT37 family must be determined by defining enzyme activity in vitro or by mutant complementation.

In this work, we developed a platform in vitro GT-array (i-GT-ray) that simplifies the screening of transferase activity in vitro. The i-GT-ray platform is based on a recently developed method called nucleic acid programmable protein array (NAPPA), which has been widely used in protein-protein interactions (PPIs) studies in animals (28, 29, 30, 31), Arabidopsis (32), and more recently, was applied to demonstrate PPIs between GTs on microplates (33). In the original NAPPA method, glass slides are coated with N-terminal Halo- or C-terminal GST-tagged GT plasmid DNA and an anti-tag capture antibody (CAb), which allows protein production using cell-free in vitro protein synthesis and protein capture on glass slide. Since GTs produced on microplates can engage in PPIs, strongly suggests that these GTs have proper folding in vitro and should maintain their transferase activity. Thus, we sought to adapt this method for GT assays on microplates and named it i-GT-ray platform. GT arrays were generated by precoating microplates with plasmid DNA and CAb (anti-tag antibody), which can be used for the production of tagged GTs (via cell-free in vitro protein synthesis) and capture on the microplate (Fig. 1). The i-GT-ray screening platform is more efficient and rapid than heterologous expression and purification of proteins for screening GT activities. The possible use of desalting paper spray-mass spectrometry (DPS-MS) detection method (in addition to GLO system) was demonstrated through detection of the products formed by transferase reactions of three GTs (AtXXT1, AtGUX1, and AtFUT6) directly from the microplates (34) (Fig. 1).

Figure 1.

Schematic presentation of the i-GT-ray screening platform coupled with DPS-MS detection method.A, capture antibody and biotinylated plasmid DNA are immobilized on microplates, which is used to couple protein synthesis (via IVTT system) and capturing on microplate to create a GT-array. B, GT-array is then used for transferase assay using NDP-sugar donors and acceptors. C, products generated by transferase reactions can be detected through DPS-MS. DPS-MS, desalting paper spray-mass spectrometry; GT, glycosyltransferase; i-GT-ray, in vitro GT-array; IVTT, in vitro transcription/translation; NDP, nucleotide diphosphate.

Using i-GT-ray platform, we performed a total of 280 assays (including the controls and duplicates) to screen 22 putative FUTs (seven from Arabidopsis and 15 from rice) from family GT37 for activity toward five acceptors: non-fucosylated tamarind XyG (TXyG), arabinotriose (Ara₃), non-fucosylated RG-I(mur1) and RG-II(mur1) prepared from the Arabidopsis mur1-1 mutant, and RG-II (-l-Gal,-MeFuc), obtained by removing Arap and MeFuc from chain B and l-Gal from chain A of the de-esterified celery RG-II monomer. Our screen showed that AtFUT2, AtFUT5, and AtFUT10 have activity toward RG-I, while AtFUT8 is active on RG-II. Five rice OsFUTs have XyG-FUT activity and four rice OsFUTs have activity toward Ara₃. None of the putative OsFUTs were active on the RG-I and RG-II acceptors used in this study. Screening of enzyme activities of some putative rice OsFUTs using GLO system is a starting point for detailed studies to improve and expand our understanding of plant cell wall fucosylation mechanisms, which should allow comparisons between rice and Arabidopsis and help elucidate how FUTs from GT37 family function.

Results

Implementing the i-GT-ray platform for high-throughput in vitro screening of nonprocessive GTs

Our goal was to develop a pipeline for high-throughput screening of GT activities by combining in vitro transcription/translation (IVTT)-based protein synthesis and enzyme assays on microplates with product detection (Fig. 1). To achieve this goal, optimization at several levels were performed. Since most eukaryotic GTs are membrane proteins and often difficult to produce in vitro as soluble, full-length active forms (2, 3), we reasoned that adding a GST or haloalkane dehalogenase (Halo) tag would improve solubility without impacting enzyme activity. This approach was validated previously with the nonprocessive Arabidopsis GTs AtXXT1 and AtGUX1 (34). Here, we used three additional nonprocessive GTs (AtFUT1, AtFUT6, and AtMUR3) to optimize our i-GT-ray platform (Table S1 lists all GTs used in the optimizations). We first compared the efficiency of two cell-free coupled IVTT expression systems (TNT Quick System and 1-Step Human Coupled IVT Kit.) in producing the tagged GTs. Western blotting analysis was performed to confirm the presence and estimate the amounts of fusion proteins produced by each system. Our results indicated that the 1-Step Human Coupled IVT system produced similar amounts of fusion proteins (∼25 ng/μl) for each GT without substantial degradation (Fig. S1). The GTs produced by the TNT Quick expression system were degraded and Western blot analysis showed high background (Fig. S1). Thus, the 1-Step Human Coupled IVT system was used in all further experiments.

The next step was to determine the optimal dilutions of the CAb (anti-GST or anti-Halo) needed in microwells to capture the fusion proteins in amounts sufficient for detection of transferase activity. Thus, we precoated microplates with three different dilutions (1/50, 1/200, and 1/400) of the CAb. The tagged GTs were then added to the microwells. AtFUT1, AtXXT1, and AtGUX1 (as N-terminal Halo- or C-terminal GST-tagged) were used, and their transferase activities analyzed according to published protocols (see Table S1 for references) using the GLO system, which is a sensitive bioluminescent assay for the detection of UDP or GDP released from nucleotide diphosphate (NDP)-sugars by a transfer reaction. The acceptors used for AtFUT1, AtXXT1, and AtGUX1 were TXyG, cellohexaose (C₆), and xylohexaose (X₆), respectively. Our findings show that the anti-GST and anti-Halo GTs gave higher activity in microwells precoated with CAb diluted to 1/50 and 1/200 (Fig. 2A). The position of the tag in fusion proteins had no discernible effect on the activity of AtXXT1 and AtGUX1 attached to microwells. By contrast, the C-terminal GST-tagged version of AtFUT1 showed no transferase activity (Fig. 2A). This result was consistent with data obtained using an assay performed with radiolabeled donor substrate (Fig. S2).

Figure 2.

Optimizations of i-GT-ray platform for screening of GT activity on microplates. A, microwells were coated with three different dilutions of capture antibody (anti-Halo or anti-GST) and the activity of captured GTs was tested using GLO system. B, correlation of enzyme activity of AtMUR3-GST with increasing amount of free plasmid DNA used in the [¹⁴C]radioactive assay in a tube. C, microwells were coated with various amounts of streptavidin to capture biotinylated plasmid DNA containing AtXXT1-GST and Halo-GUX1 constructs. Two micrograms of streptavidin were necessary to achieve immobilization of ∼500 ng plasmid DNA on the microwells. All assays were performed in duplicate + SD of the means. GTs, glycosyltransferases; i-GT-ray, in vitro GT-array.

High-throughput screening requires the use of comparable assay conditions (buffer and ions) for each GT. Thus, we next determined whether specific buffer/ion conditions can be used for most GTs. For this experiment, we tested the activity of N-terminal Halo-tagged AtFUT1, AtGUX1, and AtMUR3 in 50 mM Hepes or Tris–HCl, pH 7, containing Mg⁺⁺ and Mn⁺⁺ (1 mM each). For AtMUR3, partially degalactosylated nasturtium XyG was used as the acceptor (35). Hepes buffer is widely used in GT assays, but causes ion suppression in mass spectrometry (MS). Tris–HCl buffer is compatible with MS detection (34, 36), but not commonly used for GT assays. All three GTs were active in both buffers (Fig. S3). Although the activity levels were somewhat lower in Tris–HCl compared to Hepes, we concluded they were sufficient for high-throughput screening.

Next, we optimized the immobilization of both plasmid DNA and CAb (at 1/200) to allow simultaneous protein synthesis and capturing on the microwells (Fig. 1). DNA immobilization was carried out using a “streptavidin-biotin” system (37). We choose this system because streptavidin has four subunits into which four biotin molecules are strongly and noncovalently bound. Plasmid DNA is readily linearized and biotinylated. The C-terminus of streptavidin subunit D has a cysteine that enables covalent immobilization on a functionalized surface by a maleimide-polyethylene glycol (PEG) linker.

To determine the amounts of plasmid DNA needed to produce fusion proteins in amounts sufficient for transferase activity detection, we tested increasing amount of plasmid DNA (containing AtMUR3-GST) for protein synthesis in a test tube. We then used a [¹⁴C]radioactive assay according to (35) using partially degalactosylated nasturtium XyG. Our results indicate that 300-500 ng of plasmid DNA were sufficient to produce fusion proteins for detection of 75 to 95% of transferase activity obtained in solution (Fig. 2B). To immobilize ∼500 ng of biotinylated plasmid DNA, it was necessary to precoat the microwells with 2-3 μg of streptavidin (Fig. 2C). For the remaining experiments, GT-arrays were prepared by precoating the microwells with streptavidin (2 μg) and CAb (1/200 dilution). Somewhat unexpectedly, we found that transferase activities were higher when using microwells prepared by precoating with streptavidin and then with CAb, compared to microwells prepared by applying a mixture of streptavidin and CAb. Once all these optimizations were completed, we generated GT-arrays that were successfully used to streamline protein synthesis, immobilization, enzyme assay, and monitoring with GLO system.

i-GT-ray platform can be connected to downstream analytical methods such as MS

One of our goals is to integrate i-GT-ray platform with a downstream analysis/detection of the products formed during the transferases reactions. Typically, these products are present in very small amounts and their detection necessitates the use of radioactive NDP sugars as the sugar donors. The use of radioactive materials is not compatible with high-throughput strategy of i-GT-ray. Thus, there is a need for an alternative method that is rapid and sensitive to monitor transferase reaction products (i.e., oligosaccharides). We previously optimized a mass spectrometry method called DPS-MS that fits these requirements (34, 38, 39). Thus, we used microplates precoated with biotinylated plasmid DNA and CAb to generate GT-arrays to streamline protein synthesis, immobilization, enzyme assay, and detection of products of Halo-tagged AtXXT1, AtFUT6, and AtGUX1. Arabinotriose (Ara₃, α-l-Araf-(1,5)-α-l-Araf-(1,5)-l-Ara) from arabinan was used as a proxy for an AGP acceptor for AtFUT6, since AGPs from tobacco cells have arabinotriose side chains with a terminal α(1,5)-linked arabinofuranose (Araf) residue. We showed previously that AtFUT6 catalyzes Fuc transfer from GDP-Fuc onto Araf residues of these tobacco AGPs (17).

We successfully measured the activity of the three GTs using the GLO system (Fig. 3A) and detected the products of the reactions using DPS-MS (Fig. 3, B–D). Our results show that Ara₃ is an acceptor for AtFUT6. The enzyme catalyzed the formation of fucosylated Ara₃ (Fuc-Ara₃), which was detected by DPS-MS as an [M + Na]⁺ ion at m/z 583. The acceptor Ara₃ gives an [M + Na]⁺ ion at m/z 437 (Fig. 3D). DPS-MS analysis of the products formed by AtGUX1 reaction with X₆ (m/z 833, Fig. 3B), gave an ion at m/z 1009 corresponding to the formation of GlcA-X₆. Reacting AtXXT1 with C₆ (m/z 1013, Fig. 3C) gave an ion at m/z 1145 corresponding to the formation of Xyl-C₆. Although the intensities of the product ions were somewhat low, none of the oligosaccharides formed by the GTs were detected in the control reactions lacking donor substrates (GDP-Fuc, UDP-GlcA, or UDP-Xyl). Furthermore, DPS-MS detected these ions with high accuracy (<3 ppm, Table S2). Mass error values lower than 5 ppm is an indication of detection accuracy.

Figure 3.

Validation of i-GT-ray-DPS-MS platform using Halo-tagged AtXXT1, AtFUT6, and AtGUX1. A, transferase activities (in Tris–HCl buffer) of these GTs were measured using GLO system (expressed as luminescence unit, RLU). Assays were performed in triplicates and values are the averages ± SD. DPS-MS was used to detect the products generated by Halo-AtGUX1 (B), Halo-AtXXT1 (C), and AtFUT6 (D). All detections were carried out using 10 μl from transferase reactions performed in microwells. Zoom in spectra are shown as 5X, and 500X. The substrate acceptors used are G₆, Ara₃, and X₆, which stand for cellohexaose, arabinotriose, and xylohexaose, respectively. Asterix indicates the peaks are contaminants from the acceptor or donor substrates. DPS-MS detection was performed in duplicate, and representative spectra are shown. DPS-MS, desalting paper spray-mass spectrometry; GTs, glycosyltransferases; i-GT-ray, in vitro GT-array.

Screening of putative Arabidopsis AtFUTs members of the GT37 family for fucosyltransferase activity

We used our i-GT-ray platform to screen seven (of the ten) Arabidopsis family GT37 FUTs with no established function for their activity toward five acceptors. Non-fucosylated RG-I(mur1) and RG-II(mur1) acceptors were isolated from the cell walls of Arabidopsis mur1-1 mutant plants. This mutant is unable to produce GDP-l-Fuc as it carries a mutation in the GDP-mannose 4,6-dehydratase (GMD2) gene (9, 40, 41). The Fuc and MeFuc residues in RG-II(mur1) are partially replaced with l-Gal and l-MeGal (40, 42). Thus, we used as an alternative acceptor, a deesterified RG-II monomer from celery, that had been treated with glycosyl hydrolases that specifically remove the MeFuc and Arap from side chain B and the l-Gal from side chain A (RG-II(-l-Gal,-Fuc), Fig. 5 for structures).

Figure 5.

Activity of AtFUT8 on modified rhamnogalacturonan (RG)-II monomers from celery. The acceptors used are deesterified RG-II monomers lacking Arap (B5 in the illustration) and MeFuc (B4′) of side chain B and l-Gal of side chain A (A5 in the illustration) [RG-II(-l-Gal,-Fuc)], deesterified RG-II lacking Arap and MeFuc of side chain B but has unchanged side chain A [RG-II(+l-Gal,-Fuc)], RG-II lacking l-Gal of side chain A with unchanged side chain B [RG-II(-l-Gal,+Fuc)]. Illustrations of the RG-II structure is indicated on the right using standard drawing symbols of monosaccharides (61). The position of the Fuc (red triangle) in the acceptor are highlighted in yellow boxes. RG, rhamnogalacturonan.

Arabidopsis has ten members in the GT37 family and seven of them are currently annotated as putative FUTs with no known enzyme activity. All AtFUTs were predicted to have one transmembrane domain, except AtFUT8 and AtFUT9, which were predicted to have two transmembrane domains and AtFUT10 predicted to be secreted soluble protein. AtFUT1 and AtFUT6 fucosylate XyG and AGP, respectively, and were used as positive controls. We performed 100 assays using i-GT-ray platform with GLO assays to screen the nine AtFUTs against five non-fucosylated potential acceptors with negative controls (no acceptors added) to determine whether the AtFUTs have hydrolysis activity of GDP-Fuc. None of the putative AtFUTs hydrolyzed GDP-Fuc in the absence of acceptors. AtFUT2, AtFUT5, and AtFUT10 showed a more than three-fold increase in the activity for RG-I(mur1) acceptor compared to the control (Fig. 4A) AtFUT8 acted on celery RG-II(-l-Gal,-MeFuc), whereas RG-II(mur1), in which side chain A is largely truncated to a tetrasaccharide (42) was not an acceptor for AtFUT8. The controls AtFUT1 and AtFUT6 acted on TXyG and Ara₃, respectively as expected (Fig. 4A). Fucose is α(1,2)-linked to Gal in the type-I AG of RG-I whereas MeFuc is α(1,2)-linked to Gal in side chain B of RG-II (44, 45, 43). Since all FUTs from GT37 family thus far characterized catalyze α(1,2) linkages in various plant cell wall polysaccharides, we anticipated that these AtFUTs would also catalyze α(1,2) linkages on RG-I(mur1) and RG-II(-l-Gal,-Fuc).

Figure 4.

Fucosyltransferase activity of nine Arabidopsis AtFUTs from GT37 family on i-GTray platform. A, N-terminal Halo-tagged AtFUT1, AtFUT2, AtFUT3, AtFUT5, AtFUT6, AtFUT7, AtFUT8, AtFUT9, and AtFUT10 were synthesized and immobilized on microwells and their activity was monitored in the presence or absence of the acceptor and GDP-Fuc using GLO system. Acceptors used are non-fucosylated rhamnogalacturonan (RG) from mur1 mutant [RG-I(mur1) and RG-II(mur1)], deesterified celery RG-II monomer lacking Arap and MeFuc of side chain B and l-Gal of side chain A [RG-II(-Gal,-Fuc)], arbinotriose (Ara₃), and tamarind xyloglucan (TXyG). Halo-tagged AtFUT1 and AtFUT6 were used as positive controls. B, the layout for fucosyl transfer assays on i-GTray platform. A total of 100 FUT assays were performed in duplicates and values are the averages ± SD. FUT, fucosyltransferase; i-GT-ray, in vitro GT-array; RG, rhamnogalacturonan.

To further our understanding of RG-II fucosylation, we used the celery monomer lacking the MeFuc and Arap of side chain B or the celery monomer lacking the l-Gal of side chain A as acceptors (Fig. 5 for structure). The presence of l-Gal in side chain A strongly reduced the ability of AtFUT8 to act on the RG-II(+l-Gal,-Fuc) acceptor (Fig. 5). AtFUT8 was able to act on RG-II(-l-Gal,+MeFuc) acceptor (Fig. 5). However, the RG-II isolated from the walls of the atfut8 single mutant (SALK_010981) existed predominantly as the dimer (Fig. S4) and had a monosaccharide composition similar to RG-II from WT plants (Tables S3 and S4).

Screening of putative rice OsFUTs members of the GT37 family for fucosyltransferase activity

No fucosyltransferase activity has been demonstrated for any of the grass members of family GT37. To address this, we screened 15 rice FUTs for activity toward the same five substrates used with the Arabidopsis FUTs. With the exception of Os02g0763900, which is predicted to be soluble protein with no secretion signal peptide, all the other putative OsFUTs are predicted to be type II membrane proteins.

We used Halo-tagged versions of the rice OsFUTs, since all N-terminal Halo-tagged AtFUTs were active. AtFUT1 and AtFUT6 were included as positive controls as were controls lacking the acceptors to determine if the OsFUTs just hydrolyze GDP-Fuc. None of the putative OsFUTs hydrolyzed GDP-Fuc in the absence of acceptors. As indicated in Figure 6B, AtFUT1 and AtFUT6 had the expected XyG-FUT and Ara₃-FUT activities, respectively. OsFUTs (Os02g0763200, Os02g02763900, Os02g0764200 (OsMUR2), Os04g0449100, and Os06g0212100) were active against TXyG (Fig. 6B). OsMUR2 is annotated as a XyG-FUT (20), and our data confirm its enzyme activity in vitro. Os02g0763200 acted on TXyG in vitro, but did not restore fucosylation of XyG in the Atfut1 mutant (20). It is not known if Os02g02763900, Os04g0449100, and Os06g0212100, which also showed activity on TXyG, can complement the Atfut1 mutant. Neither Os02g0764400 nor Os06g0211600 showed activity on TXyG in vitro and they did not complement the Atfut1 mutant (20). Therefore, rice may have only one true XyG-FUT (OsMUR2). The remaining OsFUTs may fucosylate other cell wall polymers but have broad acceptor substrates in vitro and thus show activity on TXyG. It is not known if these OsFUTs use GDP-l-Gal as well as GDP-Fuc as donor substrates, as has been demonstrated for AtFUT1 (46).

Figure 6.

Fucosyltransferase activity of 15 putative rice OsFUTs members of the GT37 family on i-GT-ray platform. A, N-terminal Halo-tagged GTs were synthesized and immobilized on microwells and their activity was monitored in the presence GDP-Fuc and in the presence or absence of the acceptor using GLO system. Acceptors used are nonfucosylated rhamnogalacturonan (RG) from mur1 mutant [RG-I(mur1) and RG-II(mur1)], deesterified celery RG-II monomers lacking Arap and MeFuc of side chain B and l-Gal of side chain A [RG-II(-l-Gal,-Fuc)] (see illustration in Fig. 5), arbinotriose (Ara₃), and tamarind xyloglucan (TXyG). B, Halo-tagged AtFUT1 and AtFUT6 were used as positive controls. A total of 180 assays were performed (including duplicates) and values are the averages ± SD. “∗” indicates these FUTs were previously tested by overexpression in Arabidopsis axy2.2/fut1/mur2 mutant background (20). FUT, fucosyltransferase; GTs, glycosyltransferases; i-GT-ray, in vitro GT-array; RG, rhamnogalacturonan.

Four OsFUTs (Os02g0764200, Os06g0211600, Os06g0211700, and Os06g0212600) could act on Ara₃, with Os02g0764200 (OsMUR2), Os06g0211700, and Os06g0212100 having the highest activity (Fig. 6B). Surprisingly, OsMUR2, which is a XyG-FUT, could also act on Ara₃ demonstrating that OsMUR2 can recognize two completely different acceptors in vitro. It is not known if OsMUR2 can fucosylate AGP in vivo (for example, by introducing OsMUR2 gene in Arabidopsis atfut4/atfut6 double mutant background, which lacks Fuc in its AGPs (47). None of the rice OsFUTs showed activity toward RG-I(mur1) and RG-II(-l-Gal,-Fuc) acceptors (Fig. 6B).

Discussion

Determining the enzyme activities of new GTs will improve our understanding of the mechanisms of plant cell wall polysaccharides biosynthesis and the physiological role of a particular sugar/polymer in plant development and fitness. However, determining GTs enzyme activities in vitro is time-consuming and the least adapted to large-scale screening of GTs (48). This work describes the implementation of a powerful screening platform (i-GT-ray) based on NAPPA method. i-GT-ray simplifies the testing of enzyme activity in vitro of many GTs simultaneously (at lower costs and in less time). Our platform uses GT-arrays, generated by precoating microplates with plasmid DNA and CAb (anti-tag antibody). GT-arrays can be stored or used immediately for in vitro transferase assays. Four benchmarked nonprocessive GTs (AtFUT1, AtXXT1, AtGUX1, and AtMUR3) were used to optimize the amounts of plasmid DNA and CAb needed on the microplates and to identify a buffer suitable for the GTs. We were able to streamline the production of tagged GTs directly from plasmid DNA, their capture by the immobilized CAb, performing enzyme assays, and detection of the generated products in a single experiment and in a shorter time (48 h) compared to heterologous expression of GTs and purification. Importantly, we demonstrated that GST- or Halo-tagged GTs can be produced in a soluble and active form (in the absence of detergent) using cell-free IVTT systems. Our i-GT-ray screening platform has several advantages for GTs activity testing in vitro. Traditionally, GT must be expressed in heterologous systems and then purified before in vitro assays (may take at least 1 week). These procedures cannot be adapted to high-throughput. In i-GT-ray, cell-free production of GTs and immobilization on the surface of the microwells combines production and purification steps, which saves time (∼48 h). The transferase reactions can then be monitored directly on microplates using the GDP/UDP-GLO systems. The products of these reactions can be detected directly through DPS-MS, provided a suitable acceptor is available. All these processes can be performed in a single experiment. The i-GT-ray platform overcomes the problem of NDP-sugar donor interconversion (49), which often occurs when using Golgi-enriched membranes (or solubilized and partially purified GTs) as enzyme sources. Such interconversion typically results in the formation of multiple products, which complicates analysis of the reaction products.

For validation, i-GT-ray platform was used to screen 22 FUTs from GT37 family (seven Arabidopsis and 15 rice putative FUTs) for activity toward five non-fucosylated potential acceptors, including TXyG, Ara₃, RG-I, and RG-II. None of these FUTs have known acceptor/donor substrates, although members of family GT37 are predicted to catalyze the addition α(1,2)-linked Fuc residues onto plant cell wall glycans. Our data suggest that AtFUT2, AtFUT5, and AtFUT10 are RG-I-FUTs since they acted on the RG-I(mur1) acceptor, but not on WT RG-I. AtFUT8 may fucosylate RG-II, as it acted on celery RG-II(-l-Gal,-MeFuc), but not on RG-II(mur1) acceptor because the B side chain MeFuc residues is replaced with l-MeGal (40). Interestingly, we observed no activity of AtFUT8 toward RG-II(+l-Gal, -MeFuc) acceptor, which has the terminal l-Gal in side chain A but no MeFuc in side chain B, suggesting that the addition of terminal sugars in side chains A and B may proceed in an ordered manner (Fuc first and then l-Gal), and the order of synthesis of side chain A might be dependent on order of synthesis of side chain B. We also observed that AtFUT8 had some activity toward the RG-II(-l-Gal,+MeFuc) acceptor, which has the terminal MeFuc in side chain B but no terminal l-Gal of side chain A. Thus, AtFUT8 may transfer Fuc to the GlcA residue of the side chain A. This was surprising because there are no reports that a terminal l-Gal is replaced by l-Fuc on side chain A of naturally occurring RG-II. Furthermore, an Arabidopsis galactosyltransferase (l-GalT called Cdi, At1g64980) responsible for the addition of l-Gal to GlcA in side chain A was recently identified through genetic analysis of cdi mutant (50). RG-II from this mutant lacked the terminal l-Gal. Interestingly, while AtFUT8 showed activity on RG-II acceptor used in vitro, the monosaccharide composition and dimerization status of RG-II from atfut8 mutant plants were similar to WT plants (Fig. S4, Tables S3, and S4). One possible explanation is that AtFUT8 may have broader acceptor substrates (promiscuity) in vitro, but may have a different acceptor substrate specificity in vivo. Thus, the RG-II-FUT that adds the terminal Fuc/MeFuc of side chain B may be from a different GT family or a family not yet identified. Promiscuity toward acceptor substrates was also observed for AtFUT1, as analysis of XyG from root hair of Arabidopsis atfut1 mutant led Peña et al., (2012) (51) to conclude that AtFUT1 may be also responsible of the fucosylation of the GalA residues of root hair XyG.

Our screening of putative rice OsFUTs showed that five OsFUTs acted on TXyG. OsMUR2 (Os02g0764200) has been previously assigned XyG-FUT function through genetic complementation of axy2.2/fut1/mur2 mutant plants that lack fucosylated XyG (20). Our data provide direct evidence that OsMUR2 has XyG-FUT activity in vitro. Three other putative OsFUTs (indicated by “∗” in Fig. 6) were also tested in the same genetic complementation experiment but failed to rescue the chemotype, and one of those OsFUTs (Os02g0763200) was active on TXyG in our platform (Fig. 6), but did not complement the axy2.2/fut1/mur2 mutant, which also suggests that Os02g0763200 might be a FUT for a different cell wall polymer in vivo. OsMUR2 was found to be also active on Ara₃ (Fig. 6). In addition to OsMUR2, three additional putative OsFUTs (Os06g0764200, Os06g0211700, and Os06g0212600) could also act on Ara₃. Recently, AtFUT7 was assigned AGP-FUT activity (52), but this AtFUT did not show AGP-FUT activity in our screening platform. Taken together, these findings suggest that several FUTs from GT37 family have promiscuity toward acceptor/donor substrates in vitro. Nevertheless, if this promiscuity does exist it must be controlled in vivo. One possibility could be simply a tight regulation at the gene expression level to coincide with the presence of the appropriate acceptors. Another possibility is the presence of endogenous cofactors (in Golgi or ER) that allow FUTs to acquire substrate specificity. The structural conformation of the acceptor present during the formation of the polymer may also have a role in controlling the type of glycosidic bond that can be formed (53).

Perhaps the most intriguing finding is that none of the rice enzymes acted on non-fucosylated RG-I(mur1) and RG-II(-l-Gal,-Fuc) acceptors. The absence of FUT activity on RG-I(mur1) in rice is in agreement with previous reports indicating that RG-I in the Poaceae is not fucosylated (42, 45). However, the possibility remains that the non-fucosylated RG-I(mur1) acceptor used in this study may lack the appropriate fucosylation sites for rice RG-I-OsFUTs. On the other hand, the fact that none of putative rice OsFUTs acted on RG-II(-l-Gal,-Fuc) acceptor was surprising considering that the structure of RG-II is largely conserved in vascular plants (42, 54). Furthermore, there are still six putative rice OsFUTs (Os02g0275200, Os02g0764400, Os04g0449200, Os06g0212300, Os08g0334900, and Os09g0458100) and two putative Arabidopsis AtFUTs (AtFUT3 and AtFUT9) with unknown enzyme activity. The implication of this result is that either (i) some fucosylated polymers (to be identified) might be present in the cell wall at very low amounts, and the appropriate acceptors for these polymers are not currently available; (ii) these FUTs may require additional factors to fucosylate some of the acceptors used in this work; (iii) these FUTs may use a different donor substrate (other than GDP-Fuc); or (iv) the lack of activity of these tagged GTs is due to misfolding, which we believe unlikely.

In conclusion, we have developed a platform for high-throughput screening of GT activities. The platform allowed the generation of GT-arrays that can be used to screen substrate specificity for putative GTs whose functions are currently unknown or incompletely characterized. However, it is important to emphasize that i-GT-ray should be considered as a first step for rapid screening of GTs, as the platform alone does not provide information about the type of linkage (alpha versus beta) nor about the position of the linkage (between sugars). Therefore, the platform developed can contribute in facilitating the progress in determining the biochemical function of GTs in plants and other organisms and has the potential to be extended to enzymes other than GTs (i.e., hydrolases). The platform has some limitations such as possible misfolding of some proteins during the binding to microplate or no production of the proteins. The next challenge would be to assemble functional complexes of multiple GTs on GT-arrays that can produce full polymers in vitro. Furthermore, since we optimized the binding of DNA plasmid to microwell surface, it would be possible to adapt this platform as an additional method to investigate DNA–protein interactions associated with transcription factors for plant cell wall biosynthesis. Using i-GT-ray platform, we presented the first testing of enzyme activity of rice and Arabidopsis FUTs of the GT37 family against five non-fucosylated acceptor substrates, which expanded our view of the fucosylation landscape of plant cell wall polysaccharides in a dicot and a grass.

Experimental procedures

Chemicals and plant materials

Manganese (II) chloride, magnesium chloride, potassium chloride, sodium bicarbonate, detergents, and DOWEX 1X8-100 resin (Cl) were obtained from Sigma-Aldrich. HPLC-grade Tris, Hepes, acetic acid, potassium phosphate, formic acid, scintillation liquid, Immobilon membrane, SuperSignal West Femto Maximum Sensitivity Substrate, PureLink RNA Mini Kit, streptavidin, and biotin were from Thermo Fisher Scientific. GST was from GenScript. GDP-[³H]Fucose (299.7 GBq/mmol), GDP-[¹⁴C]Mannose (5.123 GBq/mmol), UDP-[¹⁴C]Xylose (7.141 GBq/mmol) and UDP-[¹⁴C]Glucuronic acid (9.213 GBq/mmol) were purchased from PerkinElmer Life Sciences. UDP-[¹⁴C]Galactose (1.85 GBq/mmol) was purchased from American radiolabeled chemicals Inc. UDP-Xylose was purchased from CarboSource (University of Georgia). GDP-fucose and UDP-glucuronic acid were purchased from Sigma-Aldrich. GDP or UDP-GLO glycosyltransferase assay, Halo-GST tag fusion protein, and protein expression kit (TNT Quick Coupled Transcription/Translation System) were purchased from Promega. 1-Step Human Coupled IVT Kit was purchased from Invitrogen. β-galactosidase, tamarind xyloglucan (TXyG), cellohexaose (C₆), xylohexaose (X₆), and arabinotriose (Ara₃) were purchased from Megazyme. X-ray film was purchased from Research Products International Corp. pCR8/GW/TOPO TA Cloning Kit and Gateway LR Clonase II enzyme mix kit were purchased from Invitrogen. The expression vector pJFT7_nHALO and pANT7_cGST were purchased from DNASU Plasmid Repository housed at The Biodesign Institute/Arizona State Universitye. NucleoBond Xtra Midi kit was purchased from Macherey-Nagel. A 96-well microplate (clear, flat bottom, half area, high binding, and polystyrene) was purchased from Corning. Anti-GST antibody was purchased from GE HealthCare Life Sciences, and anti-Halo antibody was purchased from Promega. Secondary mouse anti-goat immunoglobulin G (IgG) horseradish peroxidase (HRP) conjugated was purchased from Santa Cruz Biotechnology and secondary anti-rabbit IgG HRP conjugated was purchased from Promega. One-step ultra TMB-ELISA and sulfuric acid was purchased from Thermo Fisher Scientific.

Cloning of GT genes

Full-length complementary DNA (cDNA) clones of AtFUT1 (At2g03220), AtFUT6 (At1g14080), AtXXT1 (At3g62720), AtXXT2 (At4g02500), AtXXT5 (At1g74380), AtGUX1 (At3g18660), and AtMUR3 (At2g20370) were obtained from the Arabidopsis Biological Resource Center (ABRC) DNA Stock Center in the pDONR223 vector. The cloning steps are as follow: full-length protein coding sequences were amplified using gene-specific primers (Table S5) and HiFi polymerase, resulting in one PCR products with a stop codon and one without a stop codon. Both forms of PCR products were then cloned into the pCR8/GW/TOPO vector following the manufacturer’s protocol. Sanger sequencing was used to determine the correct orientation of the genes and confirm the absence of any mistakes in the sequences. The pCR8/GW/TOPO vector containing the PCR product with the stop codon was transferred to the pJFT7_nHALO expression vector using the Gateway LR Clonase II enzyme mix kit (Invitrogen) to create an N-terminal Halo-tagged fusion protein. The PCR product without a stop codon was transferred to the pANT7_cGST expression vector, generating a GST-tagged fusion protein using the Gateway LR Clonase II enzyme mix kit (Invitrogen). Gateway LR reactions were carried out in 8 μl total volume containing 150 ng of destination vector (either pANT7-cGST or pJFT7-nHalo), 150 ng of entry vector (pDONR223 or pCR8/GW/TOPO vectors containing the GT genes) and 2 μl of LR Clonase II enzyme mix. The Gateway LR reactions were performed according to the manufacturer’s instructions.

For the 15 rice genes from GT37 family, four clones were obtained from Dr Zeng’s lab (Zhejiang A&F University, Sino-Australia Plant Cell Wall Research Centre, and State Key Laboratory of Subtropical Silviculture). The remaining rice genes were cloned via RT-PCR. mRNAs were extracted from one month-old rice plants (100 mg). cDNAs were synthesized from mRNAs using Superscript III reverse transcriptase according to the manufacturer’s instruction (Life Technologies). cDNAs for FUT genes were amplified using gene-specific primers (Table S5) and HiFi polymerase. The PCR products (with and without a stop codon) were cloned into pCR8/GW/TOPO vector before transfer to pANT7-cGST or PJFT7-nHalo vectors using gateway technology as described above. Large amount of purified plasmid (mg) was prepared from 200 ml bacteria cultures (LB broth at 37 °C for overnight) using the NucleoBond Xtra Midi kit (Macherey-Nagel). All rice GT37 FUT genes were cloned except Os10g0125100, which was the only FUT annotated as a O-fucosyltransferase-like protein and Os06g0212300, which failed to amplify.

Preparation of GT-arrays: binding of plasmid DNA and CAb to microplates

Microplates with clear, flat bottom, half area, and high binding polystyrene (Corning Inc) were coated with streptavidin (2 μg in 50 μl 50 mM sodium bicarbonate, pH 9.6) for overnight at 4 °C. Unbound streptavidin was removed by washing with 10 mM phosphate buffered saline (PBS) (3 times, 10 min each). Linearized and biotinylated plasmid DNA (1 μg) in 10 mM (PBS, 50 μl) was applied to each well and the plate kept for at least 4 h at 20 °C with shaking at 250 rpm. Unbound plasmid DNA was removed by washing 3 times (10 min each) with10 mM PBS. The amounts of attached plasmid DNA was estimated using the AccuBlue NextGen dsDNA quantitation kit (Biotium). Our protocol resulted in the attachment of ∼500 ng plasmid DNA to the microwell of a microplate.

The plasmid-precoated microplates were incubated overnight at 4 °C with 50 μl of 50 mM sodium bicarbonate buffer, pH 9.6 containing anti-GST or anti-Halo antibody at various dilutions (as specified in the text). After incubation, the wells were washed with PBS buffer (3 times, 10 min each) to remove unbound CAb, and then microwells were blocked for at least 6 h at 4 °C with 5%(w/v) fat-free dry milk in PBS. The microplates can be used immediately for protein synthesis and capturing or stored at 4 °C for up to 60 days.

Linearization and biotinylation of plasmid DNA

To generate linearized and biotinylated DNA, 60 μg of plasmid DNA was linearized using PfoI and freeze-dried. The pellet was dissolved in 10 μl water and then 7.5 μl (40–50 μg) of the linearized plasmid DNA solution was mixed with ∼1.25 mg (6.52 μmol) of 1-ethyl-3-[3-dimethylaminopropyl] carbodiimide hydrochloride. Biotin hydrazide (0.25 mM in 0.1 M imidazole, 5 μl) was immediately added and the mixture vortexed, and 20 μl of 0.1 M imidazole was added before incubation overnight at 37 °C. The nonreacted 1-ethyl-3-[3-dimethylaminopropyl] carbodiimide hydrochloride and its by-products were removed using a spin desalting column (Zeba Spin Desalting Column) using 10 mM PBS containing 150 mM NaCl and 10 mM EDTA. The biotinylated plasmid DNA can be stored at −80 °C until use.

In vitro production of GT proteins in tubes and on microplates

Tagged GT proteins were produced using 25 μl TNT Quick Coupled Transcription/Translation System (Promega) or 1-Step Human Coupled IVT Kit (Thermo Fisher Scientific) either in 1.5 ml Eppendorf tubes using 1 μg GT plasmid DNA (two-step approach) or on 96-well microplates (one-step approach). The reactions were incubated for 4-6 h at 30 °C without shaking.

In the two step approach, the tagged GT proteins were synthesized in tubes using one of the IVTT systems and the synthesized fusion proteins were applied to microplates that were precoated with CAb (dilutions used are specified in text). The capturing of tagged GT proteins by CAb was carried out as follows: tagged proteins (usually in 25 μl) were mixed with 5% (w/v) fat-free milk in PBS buffer in 1:3 ratio (final volume 100 μl), and each well receives 50 μl of this mixture. After incubation overnight at 14 °C with shaking at 250 rpm, noncaptured tagged GT proteins were removed by washing with PBS buffer (3 times, 10 min each). The microplates having tagged proteins attached to the CAb can be used immediately for transferase assays or stored sealed at 4 °C for up to 60 days.

In the one-step approach, tagged GT proteins were produced directly on microplates using 50 μl of one of the IVTT systems according to the manufacturer’s recommendations and microplates incubated for 4 to 6 h at 30 °C. The synthesized fusion proteins were allowed to attach to CAb by incubation for at least an additional 4 h at 14 °C with shaking at 250 rpm. Unbound fusion proteins were removed by washing with PBS buffer (3 times, 10 min each). These microplates, which are now precoated with “fusion protein-CAb” complex (in addition to “plasmid DNA-streptavidin” complex) can be used immediately for transferase assays (see section below) or stored at 4 °C for up to a week (depending on the stability of the GT).

Immunoblotting analysis

To determine the efficiency of protein expression in vitro, 4 μl of expression reaction were analyzed by Western blotting. The produced proteins were separated on SDS-polyacrylamide gels and then transferred onto Immobilon membranes (Thermo Fisher Scientific) using the mini-protein tetra cell system (Bio-Rad). After transfer, the membranes were blocked overnight at 4 °C with 5% (w/v) fat-free dry milk in PBS. The membranes were incubated for 1 to 3 h with the primary antibody (anti-GST or anti-Halo) at a 1:10,000 dilution in PBS containing 5% (w/v) fat-free milk and 0.05% (v/v) Tween 20. After washing (three times, 10 min each) with PBS containing 0.05% (v/v) Tween 20 to remove unbound primary antibody, the membranes were kept for 1 h at room temperature with the secondary antibody (mouse anti-goat IgG or anti-rabbit IgG fused to HRP at a 1:15,000 dilution in PBS containing 5% (w/v) fat-free milk with 0.05% (v/v) Tween 20. Excess of secondary antibody was removed by several washes (5–8 times, 15 min each) with PBS containing 0.05% (v/v) Tween 20. The presence of the fusion protein was detected using SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific) and exposure to X-ray film (Research Products International Corp) according to the manufacturer’s recommendation. In some experiments, the detection of the bands was performed using ChemDoc MP (Bio-Rad).

Transferase assays

Substrates acceptors used in this work were tamarind xyloglucan (TXyG), cellohexaose (C₆), xylohexaose (X₆), arabinotriose (Ara₃), RG-I(mur1), and RG-II(mur1) both prepared from Arabidopsis mur1-1. RG-II(-l-Gal,-Fuc), RG-II(+l-Gal,-Fuc), and RG-II(-l-Gal,+Fuc) were prepared by treating de-esterified celery RG-II with chains A- and B-specific exoglycanases. Nasturtium seed XyG was degalactosylated by treatment with β-galactosidase from Aspergillus niger (E-AGLAN, Megazyme) as described in (35). Non-fucosylated AGPs were prepared from the fut4/6 Arabidopsis double mutant (47, 55) or from tobacco BY2 cell culture, which produce AGPs lacking terminal Fuc (17).

Nonradioactive assays in 50 μl reaction volume on microplates using NDP-sugars and acceptors according to published data. Transferase activity was monitored by measuring the release of UDP or GDP using the UDP-GLO or GDP-GLO glycosyltransferase assay, respectively, (Promega) according to the manufacturer’s protocol. The relative luminescence signal was measured with a Synergy/HTX multi-mode reader (BioTek). Reaction products were also analyzed by DPS-MS (34). The reactions were collected from the microwells, stopped by adding 0.3 ml water, freeze-dried, and then solubilized at the desired concentration for DPS-MS. All the assays were performed in technical duplicates.

Radioactive assays were performed as indicated for nonradioactive assays, except that the appropriate NDP-[¹⁴C] sugars were used. If the acceptors were oligosaccharides (i.e., for testing AtGUX1, AtXXT1), excess of unused NDP-[¹⁴C] sugars were removed either by chelation with DOWEX 1X8-100 resin (Cl) 1:1 (v/v) or by TLC, according to the published work (56, 57). If the acceptors were polysaccharides (i.e., for testing AtFUT1 and AtMUR3), the reactions were collected from the wells in a tube, and 1 ml of cold 70% ethanol was added and kept overnight at 4 °C. The precipitate that formed was collected by centrifugation and the pellet rinsed five times with 1 ml of cold 70% ethanol to remove excess of NDP-[¹⁴C] sugars. The dried pellets were resuspended in 0.3 ml water and mixed with 3-5 ml of liquid scintillation solution. The radioactivity was measured as cpm using a LS 6500 multi-purpose scintillation counter (Beckman Coulter).

Acid hydrolysis and monosaccharide composition analysis

Solutions of polysaccharide (∼250 μg) in 2M TFA (350 μl) were kept for 1.5 h at 120 °C. The acid was then removed under a flow of air. The released monosaccharides were converted to their alditol acetate derivatives and analyzed by GC-MS as described in (58)

Isolation of RG-I and RG-II

RG-I and RG-II was solubilized from the destarched alcohol insoluble residue of Arabidopsis leaves and flowers and from celery petioles using endopolygalacturonase (M1, Megazyme 5 units/g) as described in (58) The RG-I and RG-II was then isolated using Superdex G-75 size-exclusion chromatography. Neutral carbohydrate contaminants of RG-I and RG-II were removed using DEAE-Superose anion-exchange chromatography as described in (58)

Enzyme treatments of RG-II

A solution of celery RG-II (5 mg/ml) in 0.1 M HCl was kept for 1 h at room temperature to hydrolyze the borate diester and generate the monomer. The solution was then dialyzed (3500 molecular weight cutoff) against deionized water. An equal volume of cold 200 mM NaOH was added and the mixture kept for 16 h at 4 °C to deesterify the RG-II. The solution was adjusted to pH 5 by dropwise addition of glacial acetic acid, dialyzed (3500 molecular weight cutoff) against deionized water, and freeze dried.

Recombinant 2-O-methyl α-l-fucosidase (BT0984; MeFucase), α-l-arabinopyranosidase (BT0983; Arapase), and α-l-galactosidase (BT1010; l-Galase) were obtained from NZYtech. Solutions of the de-esterified monomer (6 mg) in 50 mM NaOAc, pH 5.2 (500 μl), were treated for 24 h at 37 °C with the l-Galase (10 μg), the Arapase (10 μg), and the MeFucase (10 μg). A separate solution of celery monomer was also treated with the Arapase and MeFucase. The reactions mixtures were chromatographed on a Superdex 75 size-exclusion chromatographic column and the monomeric RG-II collected. The solutions were repeatedly freeze dried to remove the ammonium formate. Portions of each product (∼250 μg) were then treated for 1 h with 0.1 M TFA at 80 °C to release side chains A and B or for 16 h at 40 °C with 0.1 M TFA to preferentially release side chain B. The acid was removed under a flow of warm air and the residue dissolved in water (∼1 mg/ml). MALDI-TOF-MS analysis of the products confirmed that only the expected glycoses had been removed from side chains A and B.

Desalting paper spray-mass spectrometry

A high-resolution Q Exactive Orbitrap mass spectrometer (Thermo Fisher Scientific) was used throughout this study. The commercial electrospray ionization ion source was removed to accommodate DPS. Data analysis was acquired by Thermo Xcalibur (3.0.63). The filter paper was cut into triangles (10 × 5 mm, height × base) after sequential sonication-assisted cleaning in acetone, methanol, and methanol/water (50/50 v/v, 15 min each). As reported previously in (34), for DPS-MS analysis, a 10 μl sample solution was dropped onto the paper triangle that was placed on top of a Kimwipe to facilitate the absorption by capillarity. Desalting was achieved by loading 10 μl of acetonitrile/H₂O solution (90/10 v/v) onto the paper placed on top of another Kimwipe to wick the eluent containing salts and other chemicals. The paper triangle was then held in front of the MS inlet (8 mm away) using a high-voltage cable alligator clip, and 10 μl of acetonitrile/H₂O/formic acid solution (10/90/1 v/v/v) was added directly onto the paper triangle to elute the target compounds for ionization upon application of a high voltage (3.5 kV) to the wetted paper.

Data availability

The authors confirm that the data supporting the findings of this study are available within the article [and/or] its supplementary materials. Share upon Request.

Supporting information

This article contains Supporting information (59, 60).

Conflict of interest

The authors declare that they have no conflicts of interest with the contents of this article.

Reviewed by members of the JBC Editorial Board. Edited by Chris Whitfield

Supporting information

Figure S1.

Figure S2.

Figure S3.

Figure S4.