A characterization of recombinant Arabidopsis FRIABLE1 (FRB1) reveals robust rhamnogalacturonan-I rhamnosyltransferase activity and critical catalytic residues

Abstract

Plant cell walls are glycan-rich extracellular matrices that fundamentally impact essential cellular processes, such as growth, adhesion, and cell shape acquisition. Understanding plant cell wall glycans requires the identification and characterization of the biosynthetic enzymes that produce these polymers. Most successful in vitro protein expression studies of plant cell wall glycosyltransferases have relied on insect, fungal/yeast, or human cell expression systems, whereas prokaryotic expression systems have been generally unsuccessful. Here, we show that Arabidopsis FRIABLE1 (FRB1)/rhamnogalacturonan-I rhamnosyltransferase 8 (RRT8) can be produced in Escherichia coli RosettaGami2 cells as N-terminal maltose-binding protein fusion proteins containing C-terminal 6X-His-tags. We also report the catalytic constants of FRB1/RRT8 with apparent K_M and K_cat values of 226 μM and 33 min^-1 for UDP-Rhamnose and 117 μM and 28.7 min^-1 for rhamnogalacturonan-I (RG-I), respectively. We examine the catalytic activities of mutated FRB1/RRT8 proteins based on an AlphaFold 3-generated FRB1/RRT8 protein structural model with a virtually docked UDP-Rha donor. Enzymatic characterization of the mutated and wildtype FRB1/RRT8 protein confirmed that mutation of predicted catalytic site amino acid residues resulted in a 20-fold reduction in RRT activity. FRB1 also robustly polymerizes RG-I in combination with RG-I galacturonosyltransferase 1. These results show how a robust E. coli expression system combined with artificial intelligence tools can be used to increase understanding of plant cell wall glycosyltransferase structure and function.

Plant cell walls are complex polysaccharide-rich extracellular matrices that control critical aspects of plant growth, such as cell division, cell expansion, and the acquisition of cell shape (1, 2, 3). Although cell wall structure varies across cell types and organisms, it is generally accepted that primary cell walls are composed of four major glycan networks: cellulose, hemicelluloses, pectins, and glycoproteins. Pectins account for approximately one-third of primary cell wall polymers in dicots and nongrass monocots (4) and are comprised of discrete domains, which include homogalacturonan (HG), rhamnogalacturonan-I (RG-I), and rhamnogalacturonan-II (RG-II). HG consists of repeating α-(1→4)-linked-d-galacturonic acid (GalA) residues, which can be acetylated at the O2 and O3 hydroxyl moieties as well as methyl esterified at the carboxyl group. RG-I is defined by a repeating backbone structure comprised of α-(1→4)-d-GalA-α-(1→2)-l-rhamnose disaccharide units, which can be elaborated with structurally complex galactan, arabinan, or arabinogalactan side chains (5, 6). RG-II is a structurally complex polysaccharide domain of pectin that contains 12 different monosaccharides organized in greater than 20 different linkages (7). RG-II monomers undergo boron-mediated crosslinking in cell walls to form dimerized RG-II, and it is postulated that this modification participates in increasing cell wall rigidity (8). Although the structure of various cell wall pectic domains has been intensively investigated, many of the glycosyltransferases (GTs) that synthesize these polysaccharides remain unidentified or incompletely characterized.

Among the first pectin biosynthetic enzymes to be biochemically evaluated are the homogalacturonan-synthesizing galacturonosytransferase (GAUT) gene family (9, 10, 11, 12). The Arabidopsis genome contains 15 GAUT genes (13). It was initially shown that GAUT1 and GAUT7 form a heteromeric complex that elongates α-(1→4)-linked GalA oligosaccharides to form high molecular weight homogalacturonan (11). Subsequent studies demonstrated that additional GAUT family members also have HG galacturonosyltransferase activity, but that catalytic rates and substrate requirements vary substantially among GAUT family members (9, 10). For example, the GAUT1:7 complex and GAUTs 4, 10, 11, 13, and 14 have HG acceptor–dependent activity, but GAUT13, GAUT14, and the GAUT1:7 complex can also synthesize homogalacturonan from only UDP-GalA, demonstrating an additional de novo synthesis activity (9). These observations suggest considerable functional diversity of GTs in pectin synthesis gene families.

More recently, two enzyme classes have been implicated in RG-I backbone biosynthesis. The RG-I rhamnosyltransferase (RRT) GT family catalyzes the transfer of rhamnose from UDP-β-l-rhamnose onto the nonreducing terminus of RG-I oligosaccharides of degrees of polymerization (DP) of seven or greater to extend this oligosaccharide by one α-(1–4)-linked l-rhamnose residue, indicating that RRTs use an inverting GT mechanism (14). A second GT class called the RG-I galacturonosyltransferase (RGGAT) family uses UDP-GalA to transfer GalA onto the nonreducing end of RG-I oligosaccharides terminating in rhamnose (15). These enzymes serve as founding members of the Carbohydrate Active Enzyme Database (CAZY) GT106 and GT116 families, respectively. Together, RRT and RGGAT recombinant enzymes in the presence of UDP-Rha and UDP-GalA were sufficient to elongate RG-I oligosaccharides to a DP of >60, suggesting that these enzyme classes are responsible for RG-I backbone synthesis in vivo.

Both RRT and RGGAT enzymes are deeply conserved in the green plant lineage (16), and in the model plant Arabidopsis thaliana, these enzymes are encoded by 10- and 8-member gene families, respectively. Genetic disruption studies have implicated both RRTs and RGGAT enzymes in RG-I biosynthesis during Arabidopsis seed coat mucilage deposition (14, 17). In addition, Arabidopsis FRIABLE (FRB1)/RRT8 was initially described as a Golgi-localized GT that causes severe cell adhesion defects in hypocotyls and cotyledons of Arabidopsis seedlings when mutated (18). These developmental phenotypes suggest that FRB1 may participate in the biosynthesis of cell wall polysaccharides that mediate cell adhesion in Arabidopsis seedlings. However, the enzymatic function of this GT has been somewhat controversial. Various cell wall compositional analyses of frb1 mutants initially suggested structural alterations in xyloglucan, RG-I, and RG-I side chains, but the biochemical activity of this enzyme could not be assessed (18). Other studies have assigned potential regulatory functions of FRB1 based on its annotated function as a putative protein O-fucosyltransferase (POFUT) that is distantly related to metazoan enzymes with this activity (19). Recently, another group provided preliminary evidence that Arabidopsis FRB1/RRT8 exhibits RRT activity when immunoprecipitated from transiently infiltrated tobacco plants, but it could not be ruled out that other enzymes were coprecipitated in these experiments (16). Thus, the function of FRB1 and its mechanistic contribution to plant cell adhesion remain unclear.

In this work, we perform a detailed enzymatic characterization of Arabidopsis FRB1/RRT8. For brevity, we will refer to this enzyme as FRB1 throughout the remainder of this article. We demonstrate that FRB1, similarly to other RRTs, specifically rhamnosylates RG-I oligosaccharides terminating in GalA at the nonreducing end. Through molecular modeling and mutagenesis studies, we also identify important amino acid residues required for FRB1 catalysis. Finally, we demonstrate that FRB1 can polymerize RG-I oligosaccharides in conjunction with Arabidopsis RGGAT1. Overall, these results suggest that biochemical characterization of pectin biosynthetic gene families is necessary to understand the functional nuances of individual gene family members and to understand how the RRT subfamily members may be subfunctionalized for different RG-I biosynthetic processes.

Results

Recombinant expression and activity characterization of AtFRB1

Prior work has indicated that FRB1 displayed RRT activity in single-point enzymatic activity assays (16). However, this study relied upon heterologous expression of FRB1 or other RRTs in Nicotiana benthamiana and subsequent immunoprecipitation of the enzymes for enzymatic characterization. While this approach yielded important insights, it is challenging to compare kinetic parameters or confidently assign enzymatic activities in the absence of purified proteins. A recent report demonstrated that Arabidopsis RRT4 could be produced in human embryonic kidney 293F (HEK293F) cells in an active state (15), but the authors of this study were unable to completely purify RRT4 because of significant protein aggregation during expression and purification. Thus, we sought to develop an expression system that would facilitate recombinant RRT production for in vitro characterization of purified proteins to define their catalytic properties.

Because of the significant cell adhesion defects associated with Arabidopsis frb1 mutants, we chose Arabidopsis FRB1 as an initial expression target. FRB1 (At5g01100; AtRRT8) is a 631 amino acid protein that contains a 135-residue amino terminus, a single predicted transmembrane domain, an extended stem domain, and a GT106 domain, which is responsible for GT activity (Fig. 1A). Therefore, we attempted to express a recombinant protein consisting of amino acids 145 to 631 of AtFRB1/RRT8 fused to maltose-binding protein (MBP) on the N terminus for increased protein solubility and a C-terminal 6X-His tag. This recombinant protein is referred to as MBP-FRB1ΔN144-His throughout the remainder of the text. Our initial attempts to produce this fusion protein in Escherichia coli BL21 DE3 were unsuccessful, but we reasoned that expression trials in bacterial strains, which facilitate disulfide bond formation and culture conditions that slow expression, may be more successful. After surveying a variety of expression conditions and host strains, we found that MBP-FRB1ΔN144-His was robustly expressed in E. coli RosettaGami2 cells under conditions outlined in the Experimental procedures section and that the resulting protein could be purified to near homogeneity via sequential nickel–nitrilotriacetic acid (Ni–NTA) and amylose affinity chromatography (Fig. 1B). To confirm the identity of the purified recombinant protein, MBP-FRB1ΔN144-His was subjected to SDS-PAGE. The band corresponding to the predicted protein molecular weight of this protein was dissected from the gel and subjected to tandem mass spectrometry as described in the Experimental procedures section (Fig. S1). This experiment confirmed that the major species was MBP-FRB1ΔN144-His. A lower molecular weight minor contaminant was observed occasionally in preparations of MBP-FRB1ΔN144-His (Fig. S1). This protein was also subjected to tandem mass spectrometry, which revealed that it was a degradation product lacking the C-terminal half of the FRB1 catalytic domain.

Figure 1.

Purification and activity of recombinant FRB1. A, the domain architecture of FRB1, relevant deletions, and recombinant proteins used in this study are indicated. Residue indices are shown below each protein sequence. The predicted GT106 domain (blue), TM domain (black), MBP fusion (red), and 6x-His tag (green) are also shown. B, a typical MBP-FRB1ΔN144-His purification is shown. Proteins were separated by SDS-PAGE on a 4% to 20% gradient acrylamide gel. Lanes are 1: total extract, 2: Ni–NTA flowthrough, 3: Ni–NTA eluate, 4: amylose resin flowthrough, 5: amylose resin eluate, and 6: concentrated amylose resin eluate. The red carat indicates the position of purified MBP-FRB1ΔN144-His. C, 50 ng of recombinant MBP-FRB1ΔN144-His was incubated with 20 μM RG-I oligosaccharides (oligos) and 100 μM UDP-Rha for the indicated time. UDP production was measured by the UDP-Glo assay as described in the Experimental procedures section. Error bars represent SEM (n = 4). D, the structure of RG-I (top) and OGA (bottom) acceptor substrates used in E is shown. Sugars are drawn according to Symbol Nomenclature for Glycans (SNFG) guidelines (GalA = yellow/white diamond; Rha = green triangle; and DAB end label = black circle). Linkages between monosaccharides are indicated. E, 50 ng of MBP-FRB1ΔN144-His was incubated with substrates at the indicated concentrations for 1 h at 25 °C. The resulting UDP concentrations were measured by UDP-Glo assay as described in the Experimental procedures section. Error bars represent SEM (n = 9; all points shown). Statistical significance was analyzed by one-way ANOVA and Tukey’s HSD post hoc analysis. Letters above the samples indicate experimental groups that are statistically different at a p value threshold of p < 0.0001. Groups that are labeled with the same letter are not significantly different from one another. DAB, 3,3′-diaminobenzidine; FRB1, FRIABLE1; HSD, honest significant difference; MBP, maltose-binding protein; Ni–NTA, nickel–nitrilotriacetic acid; OGA, O-GlcNAcase; RG-I, rhamnogalacturonan-I; TM, transmembrane.

To assess the solution-state molecular weight of MBP-FRB1ΔN144-His, we subjected this protein to mass photometry (Fig. S2). Recombinant MBP-FRB1ΔN144-His at a concentration of 20 nM produced a single peak with an average mass of 96 ± 12 kDa, consistent with the predicted monomeric molecular weight of this protein (97.3 kDa). As a control, recombinant MBP-His was also assayed by mass photometry, and this protein yielded an average molecular weight of 58 ± 14 kDa, consistent with its monomeric molecular weight.

Next, we asked whether the MBP-FRB1ΔN144-His recombinant protein exhibited RRT activity. RG-I was isolated from Arabidopsis Col-0 seed coat mucilage as previously described (20). The resulting polysaccharides were subjected to a brief weak acid hydrolysis to produce RG-I oligosaccharide fragments that terminated in GalA residues at the nonreducing end, and the oligosaccharides were labeled with 2-aminobenzamide at their reducing ends by reductive amination as previously described (15). These oligosaccharide fragments exhibited a DP range of 10 to 18 residues when analyzed by MALDI-TOF mass spectrometry (Fig. S3). This oligosaccharide mixture was used as an acceptor substrate for subsequent experiments. Fifty nanograms of recombinant MBP-FRB1ΔN144-His was incubated in the presence of 20 μM RG-I oligosaccharide mixture and 100 μM UDP-Rhamnose (UDP-Rha) to assess RRT activity, and UDP production was quantified using the UDP-Glo GT assay as described in the Experimental procedures section. The RRT activity of MBP-FRB1ΔN144-His was robust and linear with respect to time for up to an hour in the presence of 20 μM RG-I oligosaccharides and 100 μM UDP-Rha (Fig. 1C), indicating that the purified MBP-FRB1ΔN144-His was enzymatically active.

We also explored the donor and acceptor substrate selectivity of this enzyme. MBP-FRB1ΔN144-His was robustly active in the presence of 100 μM UDP-Rha as a sugar-nucleotide donor and 20 μM RG-I oligosaccharides as acceptor substrates (Fig. 1D). However, MBP-FRB1ΔN144-His exhibited no activity over background controls when 100 μM UDP-Glucose was used as a donor substrate in the presence of 20 μM RG-I oligosaccharides or when this enzyme was incubated in the presence of 100 μM UDP-Rha and 20 μM oligogalacturonide fragments of a similar DP (Fig. 1D). These results suggest that FRB1 selectively catalyzes rhamnosyl transfer onto RG-I oligosaccharides terminating in GalA on the nonreducing end and that this enzyme does not exhibit activity against the alternative donor and acceptor substrates that were tested.

We next evaluated FRB1 divalent cation dependence. Divalent cations, such as MgCl₂ and MnCl₂, are typically required for GTs of the GT-A family, whereas GT-B family members are often divalent cation independent (21). MBP-FRB1ΔN144-His RRT activity was not altered by the addition of 1 mM MgCl₂ or MnCl₂ (Fig. 2A), suggesting that FRB1 RRT activity is divalent-cation independent. The stoichiometry of rhamnosyl transfer was also investigated. Recombinant MBP-FRB1ΔN144-His was incubated with UDP-Rha and a purified DP12 RG-I oligosaccharide acceptor terminating in GalA. The resulting products were analyzed by MALDI-TOF mass spectrometry (Fig. 2B). Within 20 min, the initial acceptor was converted to a new product that exhibited a mass shift of 146 Da from the initial acceptor substrate, consistent with the addition of one deoxyhexose residue.

Figure 2.

Characterization of FRB1 RRT activity. Recombinant MBP-FRB1ΔN144-His was produced in Escherichia coli as described in the Experimental procedures section, and the purified enzyme was used to assess various kinetic properties of FRB1. A, the divalent cation dependence of FRB1 was investigated in the presence of 1 mM MgCl₂ or MnCl₂. Error bars represent SEM (n = 9; all data points are shown). Statistical significance was analyzed by one-way ANOVA and Tukey’s HSD post hoc analysis. Letters above samples indicate significantly different groups. B, recombinant MBP-FRB1ΔN144-His (0.5 μM) was incubated with a purified 100 μM DP12 RG-I oligosaccharide acceptor in the presence of 1 mM UDP-Rha, and the reaction products were analyzed by MALDI-TOF-MS. The addition of one rhamnose moiety is indicated by the mass increase of 146 Da. The kinetic parameters of FRB1 were evaluated by assaying 5 ng of recombinant MBP-FRB1ΔN144-His with varying (C) RG-I oligosaccharide or (D) UDP-Rha concentrations, whereas the other substrate was held constant. Error bars represent SEM (n = 3). Lineweaver–Burke plots of each Michaelis–Menten dataset are shown as insets. FRB1, FRIABLE1; HSD, honest significant difference; MBP, maltose-binding protein; RG-I, rhamnogalacturonan-I; RRT, RG-I rhamnosyltransferase.

We then measured the kinetic parameters of MBP-FRB1ΔN144-His. Fifty nanograms of purified recombinant enzyme were incubated with increasing concentrations of either UDP-Rha or RG-I oligosaccharides, whereas the other substrate was held at saturating concentrations (Fig. 2, C and D). MBP-FRB1ΔN144-His exhibited typical Michaelis–Menten saturating kinetics with respect to both substrates. The K_m(apparent) values for UDP-Rha and RG-I are 226 μM and 117 μM, respectively (Table I). In addition, the apparent K_cat values for UDP-Rha and RG-I were approximately 33 and 28.7 min^-1, respectively. Overall, these results suggest that FRB1 can function as a monomeric, metal-independent RRT that selectively transfers a single rhamnose residue to the nonreducing terminus of RG-I oligosaccharides that terminate in GalA.

Table 1.

Kinetic parameters of recombinant MBP-FRB1ΔN144-His

Substrate	Kmapp (μM)a	Vmaxapp (pmol UDP/min/ng)a	Kcatapp (min-1)a
UDP-Rha	226.2 ± 46.4	0.332 ± 0.05	33.0 ± 4.75
RG-oligo	117.3 ± 51.8	0.285 ± 0.12	28.7 ± 11.8

^aKinetic parameter values are reported as average ± standard deviation. Data are representative of three independent experiments.

Identification of amino acids underlying FRB1 catalysis

There are currently no experimentally determined protein structures of GT106 family members. However, these enzymes bear weak amino acid sequence homology to metazoan POFUT enzymes (22). Structure–function studies of metazoan POFUTs from Caenorhabditis elegans (CePOFUT1) (23) and Mus musculus (24) have defined a network of amino acid residues that are critical for the binding and transfer of GDP-β-l-fucose to form an ⍺-linkage of Fuc to epidermal growth factor or thrombospondin repeat backbones. In the CePOFUT1 catalytic pocket, these residues include R240, W245, and N43. R240 hydrogen bonds with the anomeric oxygen and the beta phosphate of GDP-Fuc to facilitate glycosidic bond cleavage, leading to the formation of an oxocarbenium ion, followed by proton transfer from the incoming acceptor molecule to the leaving phosphate group. This proposed inverting S_NI-like mechanism is critical to CePOFUT1 activity. N43 hydrogen bonds to the incoming acceptor substrate as well as an ordered water molecule in the active site that may serve as a catalytic base during fucose transfer. It has also been proposed that N43 serves as a catalytic base that accepts a proton from the acceptor by assuming its tautomeric form (25). W245 forms contacts with the fucose moiety of GDP-Fuc and may play a structural role in the active site by hydrophobically stacking against R240. Finally, CePOFUT1 D244 forms stabilizing contacts with R240 upon sugar nucleotide binding. Based on mutagenesis studies, these residues were shown to be largely essential for CePOFUT1 catalysis, with the exception of D244 mutations, which caused a modest 40% reduction in activity (23).

We initially aligned CePOFUT1 with the GT106 domains of Arabidopsis RRT1–10 using the ClustalX identity matrix (Fig. 3A). This analysis revealed clear alignment of CePOFUT1 R240 and FRB1 R386, suggesting that FRB1 R386 is the critical sugar nucleotide coordinating residue. Due to the very low sequence identity between CePOFUT1 and Arabidopsis RRTs, gaps were observed in other critical regions of the alignment, complicating the assignment of analogous residues for D244, W245, and N43. To circumvent these issues and to gain further structural resolution of the FRB1 putative active site, AlphaFold 3 (26) was used to generate a predicted structure of the FRB1 GT106 domain (residues 187–631) with a docked UDP-Rha molecule. The lowest energy model resulting from this prediction is shown in Figure S4A colored by predicted local distance difference test values. This analysis indicated high-confidence structural prediction throughout the FRB1 GT-B fold as well as within the docked UDP-Rha molecule (Fig. S4B). A positional alignment error (PAE) 2D map similarly predicted low PAE scores between the majority of atoms in the predicted structure (Fig. S4C). A superposition of the FRB1 GT106 domain and the experimentally determined X-ray structure of CePOFUT1 bound to GDP-Fuc (Protein Data Bank [PDB] ID: 3ZY5) is shown in Figure 3B. The FRB1 GT106 structure showed a clear GT-B fold with an RMSD of 2.9 Å across only 140 aligning Cα atoms. Further analysis indicated that the core Rossmann fold surrounding the sugar nucleotide-binding site was structurally well conserved between the POFUT1 experimental structure and the FRB1 model (Fig. S5). Peripheral regions of the FRB1 model and the putative acceptor binding site showed poor structural alignment statistics compared with the POFUT1 experimental structure, which is plausibly expected because these enzymes bind very different acceptor substrates.

Figure 3.

Sequence and structural analysis of the FRB1 putative active site. A, the GT106 domain protein sequences of the Arabidopsis RRT subfamily were aligned with the Caenorhabditis elegans POFUT1 fucosyltransferase domain. Black boxes surrounding residue columns indicate regions that align with critical CePOFUT1 catalytic residues. The corresponding residue numbers in Arabidopsis FRB1 are indicated above each alignment position. B, AlphaFold 3 was used to create a model of the FRB1 GT106 domain (orange) with a docked UDP-Rha molecule. This structure was superimposed on the GDP-Fuc bound structure of CePOFUT (green; Protein Data Bank ID: 3ZY5). The structural poses of GDP-Fuc (green) and UDP-Rha (orange) are shown to the right. Residues implicated from the sequence alignment in A are shown in the FRB1 AlphaFold model (C) in comparison to the residues contacting GDP-Fuc in the CePOFUT1 experimental structure (D). Residue indices are shown in red. CePOFUT1, POFUT1 from Caenorhabditis elegans; FRB1, FRIABLE1; POFUT1, protein O-fucosyltransferase 1; RRT, RG-I rhamnosyltransferase.

The AlphaFold 3 predicted structure placed UDP-Rha in a nearly identical location of the FRB1 GT106 domain model compared with the location to GDP-Fuc in the experimentally determined CePOFUT1–GDP-Fuc complex with high confidence based on the predicted local distance difference test scores (Fig. S4B). We surveyed the FRB1 amino acid residues that surrounded UDP-Rha using LigPlot+ (27) (Figs. 3C and 4B) and found that FRB1 R386 formed close hydrogen bonding contacts with the beta phosphate of UDP-Rha, as described for CePOFUT1 R240 (Figs. 3D and 4A). FRB1 M391 projects its R group under the rhamnose ring. FRB1 D390 forms stabilizing contacts with R386, similar to the R240–D244 ion pair that forms upon GDP-Fuc binding in CePOFUT1 (Fig. 3, C and D). Finally, FRB1 Q230 appears to play an analogous role to N43 in CePOFUT1 by simultaneously hydrogen bonding to the sugar and the alpha phosphate to potentially coordinate sugar transfer (Figs. 4C and 5B). LigPlot analysis also revealed backbone amide hydrogen bonding interactions contributed by FRB1 G277 and N523 to the alpha and beta phosphate (Fig. 4B), respectively, similar to the G42 and T356 interactions observed in the POFUT1–GDP-Fuc structure (Fig. 4A). Overall, these structural predictions and corresponding analysis suggest that FRB1 may engage UDP-Rha in a manner that is similar to POFUT1–GDP-Fuc and further suggest that these enzymes may use analogous transferase mechanisms.

Figure 4.

Sugar nucleotide ligand interactions of CePOFUT1 and FRB1. Two-dimensional ligand interaction plots are shown for the CePOFUT1–GDP-Fuc complex (Protein Data Bank ID: 3ZY5) (A) and the AlphaFold 3 model of the FRB1–UDP-Rha complex (B). In both plots, amino acid three-letter codes and residue indices are indicated. Residues contributing polar interactions are shown as ball-and-stick representations with each atom labeled as described in the study by Wallace et al., 1995. Distances between interacting atoms are indicated by dashed lines. Nonpolar interactions are indicated by eyelash symbols. The components of the sugar nucleotide ligands are labeled as follows: Fucose (Fuc), Rhamnose (Rha), Ribose (Rib), Guanosine (Gua), Uridine (Ura), α-phosphate (α), and β-phosphate (β). CePOFUT1, protein O-fucosyltransferase 1 from Caenorhabditis elegans; FRB1, FRIABLE1.

Figure 5.

Enzymatic activity analysis of putative FRB1 active site mutations. A, 150 ng of each FRB1 mutant was separated on an SDS-PAGE (4–20% gradient acrylamide) gel. B, activity of FRB1 site-directed mutants was evaluated by incubating 100 ng of each enzyme with 20 μM RG-I oligos and 100 μM UDP-Rha for 30 min. UDP production was measured by the UDP-Glo assay. Error bars represent standard deviation (n = 12 for UDP-Rha and FRB1(WT); n = 9 for site-directed mutants). Statistical significance was analyzed by one-way ANOVA and Tukey’s HSD post hoc analysis. Letters above samples indicate significantly different groups at a p value threshold of p < 0.0001. Groups sharing the same letter label are not significantly different from one another. FRB1, FRIABLE1; HSD, honest significant difference; RG-I, rhamnogalacturonan-I.

Based on these structural predictions, we generated site-directed mutants of the parent MBP-FRB1ΔN144-His expression construct described above. All mutants were produced, and recombinant proteins were purified via sequential affinity chromatography (Fig. 5A). These mutant proteins were assayed for RRT activity in the presence of 20 μM RG-I oligosaccharides and 100 μM UDP-Rha, and UDP production was quantified using the UDP-Glo assay. Compared with MBP-FRB1ΔN144-His, RRT activity in the Q230A and R386A mutants was completely abolished (Fig. 5B), supporting their essential proposed roles in sugar transfer and acceptor substrate coordination. Interestingly, a Q230N mutant of FRB1 was also inactive, suggesting an absolute requirement for glutamine at this position. These residues are conserved in all Arabidopsis RRT subfamily members (Fig. 3A). Surprisingly, a conservative substitution of D390E also abolished FRB1 RRT activity, suggesting that conformational stabilization of R386 may be essential for catalysis in contrast to the relatively weak impact that the analogous D244 residue in CePOFUT1 has on catalysis. Finally, the M391A mutant lost 80% of the MBP-FRB1ΔN144-His RRT activity, suggesting that this residue may be crucial for sugar nucleotide binding or positioning during catalysis. We considered the possibility that the loss of activity in these mutants could simply be the consequence of protein misfolding. Unfortunately, we were not able to produce sufficient recombinant protein to directly measure mutant stability via differential scanning fluorimetry or other experimental approaches that would enable stability measurements. Thus, we used predictions from DynaMut2 (28), CupSAT (29), and mCSM (30) to predict the folding stability changes of each mutant (Fig. S6). This analysis revealed that alanine substitutions at Q230 and R386 as well as the asparagine substitution at Q230 were predicted to change the ΔG of protein folding by less than 1 kcal/mol. Similar results were predicted for the D390E mutant. Substitution of M391 to alanine was consistently predicted to lower the stability of FRB1 by 3 to 5 kcal/mol, but this was the only mutant that retained partial activity in the assay. Overall, these data suggest that this constellation of residues is important for FRB1 catalysis. However, we acknowledge that without structural or biochemical studies to determine how these mutations impact FRB1 folding stability, it is possible that these activity changes could be due to defects in FRB1 folding.

Coordination of activity between FRB1 and RGGAT1

RGGAT1 transfers GalA from UDP-GalA to an RG-I oligosaccharide terminating in l-rhamnose (15). Prior work demonstrated that RG-I backbone synthesis could be reconstituted in vitro by incubating RGGAT1 and partially purified RRT4 in the presence of UDP-GalA, UDP-Rha, and an RG-I oligosaccharide acceptor. Under these conditions, the initial DP12 acceptor was elongated to a DP of approximately 60 over the course of 12 h. While this important result indicates that RRT4 and RGGAT1 were sufficient to generate RG-I polymers in vitro, the resulting products were far smaller than RG-I extracted from Arabidopsis seed coat mucilage, which is practically devoid of branched side chains (20). Thus, we sought to understand whether FRB1 could accomplish a similar activity and whether conditions exist to increase the size distribution of the resulting RG-I product (Fig. 6A).

Figure 6.

In vitro synthesis of RG-I by FRB1 and RGGAT1. A, a cartoon to illustrate the proposed mechanism of RG-I backbone biosynthesis is shown. RGGAT1 transfers GalA to the nonreducing end of the growing RG-I polymer, and this product serves as a substrate for FRB1 rhamnosylation. Sugar symbols conform to Symbol Nomenclature for Glycans (SNFG). B, FRB1 and RGGAT1 were recombinantly expressed and purified from Escherichia coli RosettaGami2 cells and HEK293 cells, respectively, as described in the Experimental procedures section. 0.5 μM MBP-FRB1ΔN144-His and 2 μM RGGAT1ΔN77 were incubated with 5 mM sugar nucleotide and 50 μM (100:1 ratio) or 2.5 μM (2000:1 ratio) DP12 RG-I oligosaccharide. Time points of these reactions were taken and separated on a polyacrylamide carbohydrate electrophoresis (PACE) gel that was stained with alcian blue. Products from each reaction are indicated by red and blue carats. White and black carats indicate reference of DP12 at 400 ng (lane 3) and 40 ng (lane 5), respectively. Arabidopsis seed mucilage RG-I is shown in the first lane as a size reference. FRB1, FRIABLE1; HEK293, human embryonic kidney 293 cell line; HSD, honest significant difference; MBP, maltose-binding protein; RGGAT1, RG-I galacturonosyltransferase 1; RG-I, rhamnogalacturonan-I.

To address this question, samples containing 0.5 μM MBP-FRBΔN144-His and 2 μM RGGAT1ΔN77 were incubated with 100:1 and 2000:1 donor:acceptor concentrations and incubated for 12 h (Fig. 6). The resulting products from each reaction were analyzed by polyacrylamide carbohydrate gel electrophoresis. The mucilage control represents a polysaccharide containing ∼2000 GalA:Rha disaccharides. The reaction containing a 2000:1 donor:acceptor ratio synthesized an RG-I polysaccharide longer than all reactions tested with RRT4 from HEK293F cells (15). In reactions performed with RRT4, increasing the donor:acceptor ratio did not result in longer polysaccharides, presumably because of the limited RRT activity. Increasing the donor:acceptor ratio to 2000:1 resulted in increased RG-I product size relative to the 100:1 ratio used in most previous assays (15), suggesting that substrate availability may be a critical factor controlling the ultimate length of the polymerized product. Overall, these results suggest that FRB1 synthesizes RG-I polymers with the assistance of RGGAT1 to a length that is more similar to native seed mucilage RG-I.

Discussion

In this study, we developed a bacterial expression system to produce FRB1/RRT8 and generate pure recombinant proteins for biochemical characterization. Our work demonstrates that FRB1 is a robust RRT that adds a single rhamnose residue to RG-I oligosaccharides at the nonreducing end in a metal-independent manner (Figs. 1 and 2). We further used AlphaFold 3 modeling to identify critically important amino acid residues that participate in RRT catalysis (Figure 3, Figure 4, Figure 5) and show that FRB1 can elongate RG-I polymers in collaboration with Arabidopsis RGGAT1 (Fig. 6).

Eukaryotic GTs are notoriously difficult to produce recombinantly, and this challenge has hampered the field of plant GT biochemistry. Various mammalian and insect cell expression platforms have been used to circumvent this issue (31), but these approaches are costly and require specialized facilities. Here, we demonstrate that an optimized expression platform consisting of an E. coli expression strain that accommodates rare eukaryotic codons and enables the formation of disulfide bonds in the cytoplasm improved GT expression of FRB1, which is predicted to form three disulfide bonds. The attachment of MBP and low-temperature induction were also factors that may improve RRT expression. This system is part of a growing repertoire of approaches that enable prokaryotic recombinant expression of eukaryotic GTs (32) and thus enables high-throughput approaches, such as directed evolution and high-throughput expression of protein variants. It will be important to determine the limits and benefits of this expression system in future work.

The RRT/GT106 family is part of a large group of plant GTs that were initially annotated as protein O-fucosyltransferases because of their very low sequence similarity to metazoan GTs (22). This larger group also includes ESMERELDA1 (19), MANNAN SYNTHESIS RELATED1 (33, 34), and O-FUCOSYLTRANSFERASE1 (22), which have markedly different proposed functions based on Arabidopsis mutant genetic analyses. Still, multiple studies have demonstrated that some mechanistic aspects of these proteins can be elucidated through careful structural comparisons with metazoan POFUTs (22, 34). Here, AlphaFold 3 modeling with simultaneous UDP-sugar docking allowed us to predict multiple catalytically important residues that putatively engage UDP-Rha in similar ways to CePOFUT1, suggesting that aspects of the sugar transfer mechanism are conserved between these two protein classes. It is important to note that conservation of other residues in this region may predict sugar nucleotide specificity of other members of this family.

Many polymeric glycans, including the RG-I backbone, contain repeating disaccharide motifs, which increase the complexity of their biosynthesis. Mammalian glycosaminoglycans (hyaluronan, heparan sulfate, and dermatan sulfate) and bacterial peptidoglycan represent other well-studied examples in which repeating disaccharides are synthesized. In hyaluronan synthesis, a unique hyaluronan synthase transmembrane GT enzyme sequentially catalyzes the formation of GlcNAc and GlcA linkages to form the repeating [GlcNAc-β-(1→3)-GlcA-β-(1→4)]_n motif by binding sugar nucleotides in alternating and mutually exclusive catalytic cycles (35). Heparan ([-GlcA-β-(1→4)-GlcNAc-α-(1→4)-]_n) is synthesized by the exostosin (EXT) 1/2 complex. This complex consists of an EXT 1/2 heterodimer in which each EXT subunit contains both a GT47 domain postulated to catalyze β-(1→4)-GlcA transfer and a GT64 domain postulated to catalyze α-(1→4)-GlcNAc transfer. However, structural analyses of the complex showed that only two of the proposed four GT domains actually contributed to the bulk of copolymer heparan synthesis, with the EXT1 GT-B fold domain contributing β-(1→4)-GlcA transferase activity and the EXT2 GT-A fold domain contributing α-(1→4)-GlcNAc transferase activity (36). In peptidoglycan synthesis, N-acetylmuramic acid–N-acetylglucosamine disaccharide units are synthesized as lipid-linked precursors, translocated across the membrane, and then polymerized by transglycosylase enzymes (37). Thus, glycan heteropolymers can be synthesized by multiple mechanisms, including one enzyme catalyzing the formation of multiple linkages, polyfunctional proteins with multiple GT domains, or the polymerization of disaccharide units by transglycosylases. The coordination of RRTs and RGGATs in RG-I backbone biosynthesis provides yet another solution to this problem in eukaryotic organisms by catalyzing independent glycosylation reactions using two different polypeptides. It will be interesting to investigate whether these proteins form multimeric complexes either directly or through their interactions with accessory proteins in the future.

Experimental procedures

Cloning methods

The coding sequence for FRB1/RRT8 (At5g01100) was amplified from complementary DNA clones obtained from the Arabidopsis Biological Resource Center by PCR using Phusion DNA polymerase and gene-specific primers (Table S1). A reverse primer encoding a C-terminal 6x-His tag was used to install this sequence. Forward primers were designed to remove the predicted transmembrane domain of FRB1 based on sequence analysis using transmembrane prediction using hidden Markov models (https://services.healthtech.dtu.dk/services/TMHMM-2.0/) (38). PCRs were cycled using the following conditions: 95 °C initial denaturation for 5 min, followed by 35 cycles of 95 °C (30 s), 52 °C (30 s), and 72 °C (2 min 30 s), and a final elongation for 7 min at 72 °C. The resulting DNA fragment was resolved on a 1% (w/v) low-melting-point agarose gel, excised, and gel purified with the QiaQuick gel extraction kit (Qiagen). Extracted DNA fragments were cloned into the pENTR-D-TOPO vector according to the manufacturer’s instructions (ThermoFisher Scientific). Subsequent entry clones were transferred into the bacterial expression vector pKM596 (39) using LR Clonase II according to the manufacturer’s instructions.

Site-directed mutants were generated using the FRB1ΔN144-His/pENTR-D-TOPO construct described above as a template and mutation-specific primers (Table S1). PCRs were cycled under the following conditions: 95 °C initial denaturation for 5 min, followed by 35 cycles of 95 °C (30 s), 60 °C (30 s), and 72 °C (5 min), and a final elongation for 10 min at 72 °C. PCR products were digested with DpnI for 3 h at 37 °C, and the resulting plasmids were confirmed by Sanger sequencing at the Nevada Genomics Center.

Protein purification methods

Expression plasmids encoding MBP-FRB1ΔN144-His and site-directed mutants of this parent construct were chemically transformed into E. coli RosettaGami2. One-liter expression cultures were grown to an absorbance of 0.6 to 0.8 at 600 nm and induced with 250 μM IPTG. Induced cultures were incubated for 16 h at 18 °C with shaking and pelleted by centrifugation at 3000g for 15 min at 4 °C. The resulting pellets were stored at −80 °C.

Cell pellets were resuspended in 1x Wash Buffer (25 mM Tris–HCl [pH 7.5], 150 mM NaCl, 10 mM MgCl₂, and 20 mM imidazole) containing a Pierce EDTA-free protease inhibitor tablet (Thermo). Pellet suspensions were incubated on ice for 20 min with 1 mg/ml lysozyme, followed by six cycles of sonication for 20 s (one pulse per second, 4 V) and 20 s on ice. Cell lysates were subjected to ultracentrifugation at 100,000g for 1 h at 4 °C. Clarified supernatants were applied to a 1 ml HisPur Ni–NTA superflow agarose (ThermoFisher) equilibrated in 1x Wash Buffer and washed with 350 ml 1x Wash Buffer. Proteins were eluted from the Ni–NTA resin in 1x Wash Buffer supplemented with 300 mM imidazole. Ni–NTA eluates were applied to 1 ml amylose resin columns (NEB) equilibrated with 1x Wash Buffer and washed with 350 ml 1x Wash Buffer. Proteins were buffer exchanged on a column in 20 column volumes of 2x Hepes buffer (50 mM Hepes–NaOH [pH 7.0], 300 mM NaCl, 6 mM KCl, 10 mM MgCl2, and 0.1% v/v Tween-20). Proteins were eluted in 2x Hepes buffer supplemented with 10 mM maltose into 8 × 1 ml fractions. Protein quality and quantity of elution fractions were evaluated by 4% to 20% SDS-PAGE gels (Bio-Rad) and Pierce BCA Protein Assay Kit (Thermo), respectively. Assays conducted to assess divalent cation dependence used proteins purified with 1x Wash Buffer and buffer exchanged in 2x Hepes and eluted in 2x Hepes buffer lacking MgCl₂. Typical recombinant protein yields were between 0.5 and 1 mg of protein per liter of culture.

Sample preparation for mass spectrometry

MBP-FRB1ΔN144-His samples were separated in triplicate on a 4% to 20% gradient acrylamide SDS-PAGE gel, stained with GelCode blue, and destained in water. Proteins of interest were excised from the gel, chopped into fine pieces with a razor blade, and transferred to Eppendorf tubes. The resulting gel pieces were washed for 20 min in 500 μl of 100 mM ammonium bicarbonate, and the wash solution was discarded. The gel pieces were then incubated in 150 μl of 100 mM ammonium bicarbonate containing 3 mM dithiothreitol for 15 min at 50 °C. The supernatant was discarded, and the gel pieces were washed with 500 μl of 50:50 acetonitrile (ACN):100 mM ammonium bicarbonate for 20 min at 25 °C. Fifty microliters of ACN were added to shrink the gel pieces. After 15 min, the solvent was removed, and the samples were dried in a speed vac. Ten micrograms of mass spectrometry–grade trypsin (Promega) in 25 mM ammonium bicarbonate were added to the gel slices, and after 15 min, an additional 20 μl of 25 mM ammonium bicarbonate was added to the samples to cover the gel pieces. Samples were incubated for 16 h at 37 °C. The resulting supernatant was transferred to a new Eppendorf tube. The remaining peptides were extracted using 50 μl of 70% ACN:0.1% trifluoroacetic acid for 20 min at 25 °C. The supernatant was pooled with the respective sample, and this process was repeated using 25 μl of 100% ACN. The combined supernatants were dried in a speed vac and resuspended in 20 μl of 0.1% formic acid. The resulting sample was subjected to tandem mass spectrometry on a Thermo Orbitrap Eclipse Tribrid mass spectrometer at the Nevada Proteomics Center. Mass spectrometry data were searched against the Arabidopsis and E. coli proteomes with Proteome Discoverer utilizing SEQUEST.

Rhamnosyltransferase assays

UDP-β-l-rhamnose was obtained from Peptide Institute, Inc. Other sugar nucleotides were obtained from Sigma–Aldrich. Oligogalacturonides were obtained from Carbosynth. Samples to be assessed by the UDP-Glo GT Assay (Promega) were prepared according to respective concentrations of sugar nucleotide substrates (100 μM), acceptor molecules (20 μM), and enzyme concentrations and brought to a final volume of 25 μl in 1x Hepes buffer. For assays to measure kinetic constants, RG-I oligos or UDP-Rha were varied between 0 and 300 μM, whereas the other substrate was held at a constant 300 μM concentration. Samples were incubated for 30 min at 25 °C, heat-inactivated at 70 °C for 5 min, and cooled on ice for 5 min prior to the addition of 25 μl of Nucleotide Detection Reagent (Promega). Relative luminescence units were measured after 1 h by a Spectramax M5 Microplate Reader. With respect to divalent cation dependency, samples were prepped in the same manner with the inclusion of either 1 mM MgCl₂ or 1 mM MnCl₂ and incubated for 45 min at 25 °C, heat-inactivated at 70 °C, and cooled on ice for 5 min prior to evaluation by the UDP-Glo GT Assay (Promega). UDP-Glo product formation was calculated using standard curves of relative luminescence units produced by known concentrations of UDP.

RG-I polymerization assays

RGGAT1 was produced in HEK293F cells and purified as described previously (15). UDP-β-l-Rha and UDP-α-d-GalA were obtained from CarboSource Services. Samples containing 0.5 μM MBP-FRB1ΔN144-His and 2 μM RGGAT1ΔN77 were incubated with 100:1 donor:acceptor concentrations (5 mM UDP-GalA, 5 mM UDP-Rha, and 50 μM DP12 RG-I oligosaccharide acceptor) or 2000:1 donor:acceptor concentrations (5 mM UDP-GalA, 5 mM UDP-Rha, and 2.5 μM acceptor) and incubated for 12 h at 30 °C. The resulting reaction products were separated on 20% (v/v) polyacrylamide gels and stained with alcian blue/silver as previously described (15). Bovine serum albumin (BSA) and apyrase were included at 0.05% (w/v) and 1 unit, respectively, for their positive effects on enzyme activity in similar reactions (15).

RG-I oligosaccharide purification

RG-I oligosaccharides were prepared from RG-I extracted from Arabidopsis seed coat mucilage as previously described (15, 20). Briefly, 2 g of Arabidopsis Col-0 seeds were placed in a 50 ml conical tube and mixed with 40 ml of deionized water. Nonadherent mucilage was extracted for 3 h with gentle mixing, and seeds were removed by centrifugation at 2500g for 5 min. The supernatant was transferred to a new tube, filtered through a polycarbonate 3 μM filter (Osmonics), and the filtrate was lyophilized. The resulting mucilage polysaccharides were resuspended at a concentration of 10 mg/ml in water and digested by acid hydrolysis in 0.1 M hydrochloric acid at 80 °C for 48 h. RG-I oligosaccharides from this mixture were visualized on 20% PAGE gels as described above. Digested oligomers were neutralized by the addition of 0.1 M sodium hydroxide, dialyzed against water using a 3500 Da cutoff membrane (SpectraPor), and lyophilized. The lyophilized product was labeled on the reducing end by resuspending 10 mg/ml in 10% acetic acid with 0.2 M 2-aminobenzamide and 1 M sodium cyanoborohydride to facilitate quantification of the oligosaccharides. The mixture was incubated at 45 °C for 16 h, dialyzed as described above, and lyophilized.

Sequence alignment and structural analysis

RRT family protein sequences were obtained from The Arabidopsis Information Resource Center, truncated in silico to remove sequences outside the GT domain, and aligned in reference to the CePOFUT1 protein sequence using ClustalX and SnapGene (Dotmatics) software (40). Sequence alignments using the ClustalX identity matrix algorithm identified conserved amino acid residues among these proteins. A model of the FRB1/RRT8 GT106 domain with associated bound UDP-Rha was created using the FRB1/RRT8 GT106 domain primary sequence (amino acids 187–631) and a CCD file of UDP-Rha harvested from the structure of Arabidopsis UGT89C1 (PDB ID: 6IJA) (41). Models were built using a local copy of AlphaFold 3 parameterization file (26) on the UGA Sapelo 2 cluster. The resulting models were compared with GDP-Fuc bound CePOFUT1 (PDB ID: 3ZY5) (23) using PyMOL (Schrodinger, Inc). The PAE plot of the AlphaFold 3-generated model presented in Figure S4C was rendered using the PAE viewer (42). Ligand interaction plots were generated using LigPlus, version 2.3 (27).

Mass photometry

All mass photometry experiments were performed with a Refeyn TwoMP instrument. BSA (Thermo Scientific) and thyroglobulin (Sigma–Aldrich) standards were prepared at 100 nM in sterile diH₂O. Recombinant proteins were diluted in 1X Tris-buffered saline (TBS) to a final concentration of 100 nM for mass photometry. These stocks were centrifuged for 5 min at 14,000g to remove any insoluble material. 1X TBS buffer was filter-sterilized using 0.2 μm Choice PES Syringe Filters (Thermo Scientific) to serve as the buffer blank and droplet dilution for the mass photometer. Sample slides were prepared by applying a 20 μl 6-well sample cassette to a sample carrier slide and placing it on the stage with a drop of immersion oil (Refeyn). The full sample volume (20 μl) of filtered 1X TBS was applied to the sample slide, and a droplet dilution was calibrated before recording the buffer background. BSA and thyroglobulin standards were diluted in 1X TBS droplets on the sample slide to a final concentration of 20 nM, and their molecular masses were measured. Samples were then prepared in triplicate by diluting in 1X TBS droplets to a final concentration of 20 nM. The resulting mass photometry data were analyzed using Refeyn Discover software.

Data availability

All data supporting this article are contained within the article and the corresponding supporting information. Raw data supporting these findings are available from the authors upon reasonable request.

Supporting information

This article contains supporting information.

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 Joseph Jez

Supporting information

Supplementary Material

1