Convergent Emergence of Glucomannan β-Galactosyltransferase Activity in Asterids and Rosids Authors: Konan Ishida, Matthew Penner, Kenji Fukushima, Yoshihisa Yoshimi, Louis F L Wilson, Alberto Echevarría-Poza, Li Yu, Paul Dupree Published: 2024 ABSTRACT Abstractβ-Galactoglucomannan (β-GGM) is a primary cell wall polysaccharide in rosids and asterids. The β-GGM polymer has a backbone of repeating β-(1,4)-glucosyl and mannosyl residues, usually with mono-α-(1,6)-galactosyl substitution or β-(1,2)-galactosyl α-galactosyl disaccharide side chains on the mannosyl residues. Mannan β-galactosyltransferases (MBGTs) are therefore required for β-GGM synthesis. The single MBGT identified so far, AtMBGT1, lies in glycosyltransferase family 47A subclade VII and was identified in Arabidopsis. However, despite the presence of β-GGM, an orthologous gene is absent in tomato (Solanum lycopersicum), a model asterid. In this study, we screened candidate MBGT genes from the tomato genome, functionally tested the activities of encoded proteins and identified the tomato MBGT (SlMBGT1) in GT47A-III. Interestingly therefore, AtMBGT1 and SlMBGT1 are located in different GT47A subclades. Furthermore, phylogenetic and glucomannan structural analysis from different species raised the possibility that various asterids possess conserved MBGTs in an asterid-specific subclade of GT47A-III, indicating that MBGT activity has been acquired convergently among asterids and rosids. The present study highlights the promiscuous emergence of donor and acceptor preference in GT47A enzymes. The independent acquisition of the activity also suggests an adaptive advantage for eudicots to acquire β-GGM β-galactosylation and hence also suggests that the disaccharide side chains are important for β-GGM function. Introduction Mannans are a class of polysaccharides conserved from algae to eudicots but with large structural variations. The simplest structure, homomannan, consisting of β-(1,4)-linked mannosyl residues (Man), is found as a storage polysaccharide in plants such as Aloe vera and ivory nuts and as a cell wall polysaccharide in some green algae and red algae (Aspinall 1959, Preston 1979, Painter 1983, de O. Petkowicz et al. 2001, Moreira and Filho 2008). Glucomannan, whose backbone is interspersed with β-(1,4)-linked glucose (Glc) and whose mannosyl residues are often acetylated, is also seen as a storage polysaccharide (e.g. konjac and orchid) (Gille and Pauly 2012, He et al. 2017, Voiniciuc 2022). α-Galactosyl (Gal) substitution on C6 of Man is seen in some storage mannans such as in legume seeds (Meier and Reid 1982, McCleary et al. 1985). Acetylated galactoglucomannan (AcGGM) is widespread in the cell walls of woody tissues of vascular plants and is particularly abundant in conifers (Capek et al. 2002, Handford et al. 2003, Goubet et al. 2009, Terrett et al. 2019, Chernova et al. 2020). Recently, a novel type of galactoglucomannan with a repeating Man–Glc disaccharide unit in the backbone was discovered (Yu et al. 2018). This patterned galactoglucomannan can be not only monogalactosylated as found in AcGGM but also di-galactosylated on the Man residues through a β-(1,2)-linkage on α-(1,6)-Gal. This glucomannan is called β-galactoglucomannan (β-GGM) (Yu et al. 2022). β-GGM is widespread in eudicots, being found in the asterid and rosid clades, including Arabidopsis, apples, kiwifruit and tomatoes (Solanum lycopersicum) (Schröder et al. 2001, Yu et al. 2022), but has not been found in monocots (Ishida et al. 2023). Mannan structural variation is likely to be related to physiological function in ways that are only partly understood. In algae, homomannan can serve as a major fibrillar component of the cell wall similar to cellulose in land plants (Preston 1979, Painter 1983, Fernández et al. 2012). The replacement of Man backbone residues with Glc and the addition of acetyl groups on the backbone, or branching by α-Gal, are thought to affect the polysaccharide solubility (Scheller and Ulvskov 2010) and interaction with other cell wall components (Yu et al. 2018). It had been thought until recently that the importance of mannans may have decreased in land plant evolution as xyloglucan became more abundant in cell walls (Scheller and Ulvskov 2010). However, mannans are likely to be essential for some aspects of plant physiology since the putative glucomannan synthase mutant csla7 shows embryonic lethality and defective pollen tube growth (Goubet et al. 2003, 2009). Indeed, challenging the idea of a waning importance of glucomannan through vascular plant evolution is the fact that eudicots acquired the more complex β-GGM during angiosperm evolution (Yu et al. 2022, Grieß-Osowski and Voiniciuc 2023, Ishida et al. 2023). The significance of β-GGM seems to be obscured in many tissues by the presence of xyloglucan (Yu et al. 2022). As there is a substantial gap in our understanding of the relationship between mannan structures and physiological functions, it is essential to expand our knowledge about the biosynthesis of β-GGM. Three types of transferase enzymes are involved in mannan biosynthesis: CELLULOSE SYNTHASE–LIKE A (CSLA, GT2), which synthesizes the backbone; MANNAN α-GALACTOSYLTRANSFERASE (MAGT, GT34), which adds the first galactosyl side chain residue; and MANNAN O-ACETYLTRANSFERASE (MOAT), which adds acetylation (Voiniciuc 2022). For β-GGM synthesis, an additional activity MANNAN β-GALACTOSYLTRANFERASE (MBGT, GT47) is required to complete the β-(1,2)-galactosyl α-galactosyl disaccharide side chains (Yu et al. 2022). Furthermore, the participation of GT106 family putative glycosyltransferase enzymes in mannan production has been identified (Wang et al. 2013, Voiniciuc et al. 2019), but the activity and function remain unclear. The two major types of mannan are synthesized by different combinations of enzymes in Arabidopsis (Goubet et al. 2009, Voiniciuc et al. 2015, Yu et al. 2018, 2022, Zhong et al. 2019). AcGGM of primary and secondary cell walls is synthesized by AtCSLA9, AtMAGT1 and AtMOAT1–4. In contrast, β-GGM is synthesized by AtCSLA2, AtMAGT1 and, uniquely for β-GGM, the β-galactosyl transferase AtMBGT1. Among the four mannan synthesis enzymes, CSLA and MOAT are highly conserved in land plants (Liepman et al. 2007, Zhong et al. 2019). However, close orthologs of AtMAGT1 and AtMBGT1 are not found in bryophytes and vascular seedless plants (Ishida et al. 2023). Consistent with the genomic/phylogenetic information, mannan α-Gal side chains have not been found in Physcomitrium patens (Ye et al. 2022). β-GGM biosynthesis shares the same GT families as of xyloglucan (Yu et al. 2022, Grieß-Osowski and Voiniciuc 2023). The β-1,4-glucan backbone is synthesized by CELLULOSE SYNTHASE–LIKE C in the GT2 family, and xylosyl side chains are added by xyloglucan xylosyltransferase in the GT34 family (Julian and Zabotina 2022). The addition of the second side chains (equivalent to β-1,2-Gal of β-GGM) is catalyzed by enzymes in GT47A, which is divided into seven subclades (I–VII) (Yu et al. 2022). Interestingly, GT47A has many xyloglucan-modifying enzymes with varying donor preferences, but there is no clear correlation between the activity and their position in GT47A. For example, AtMUR3 and AtXLT2 xyloglucan galactosyltransferases locate to GT47A-VI and GT47A-III, respectively, albeit both with galactosyltransferase activities. SlXST xyloglucan arabinofuranosyltransferases and CcXBT and VcXBT xyloglucan β-xylosyltransferases are classified in GT47A-III along with AtXLT2 (Schultink et al. 2013, Immelmann et al. 2023, Wilson et al. 2023). Two highly similar GT47A enzymes from P. patens in GT47A-I showed distinctive activities: XLT2 activity and xyloglucan arabinopyranosyltransferase activities (Zhu et al. 2018). The range of activities suggests that there is relatively frequent evolutionary emergence of new donor preferences in GT47A enzymes that synthesize xyloglucan side chains. AtMBGT1 belongs to GT47A-VII and is the only identified MBGT to date (Yu et al. 2022). To investigate conservation of MBGT activity in GT47A-VII in monocots and Amborellales, Nymphaeales and Austrobaileyales (ANA)-grade plants, we previously studied rice (Oryza sativa) and Amborella (Amborella trichopoda) enzymes (Ishida et al. 2023). Neither candidate exhibits MBGT activity, and the rice GT47A-VII enzyme was shown to be a xyloglucan galactosyltransferase. Moreover, mannans containing β-Gal side chains were not found in more than 20 species of monocot and ANA-grade tissues studied (Ishida et al. 2023) (Fig. 1). These findings support the notion that the acquisition of MBGT activity and therefore β-GGM-containing β-Gal side chains is likely to have occurred only in eudicots. However, within eudicots, the evolutionary acquisition of MBGT activity and hence β-GGM is not yet well studied. β-GGM and MBGT distribution in the different evolutional stages. Here, we searched for an MBGT enzyme from the asterids, since many asterids synthesize β-GGM. Surprisingly, asterids were found to lack a direct ortholog of AtMBGT1, and the closest homolog in tomato failed to demonstrate MBGT activity. Instead, based on tomato genomic and transcriptomic data, one candidate was found in GT47A-III, a distant and distinct subclade from that of AtMBGT1. We showed that the candidate has MBGT activity in vitro and in vivo. Further data support the hypothesis that this MBGT GT47A-III subclade is likely to be specific and conserved throughout asterids. These findings indicate the separate yet convergent acquisition of MBGT activity in GT47 in asterids and rosids, emphasizing the importance of the MBGT-dependent digalactosyl disaccharide side chains for β-GGM functions. Results Identification of Solyc02g092840 as a candidate asterid MBGT The MBGT in Arabidopsis is in the GT47A-VII clade (Yu et al. 2022). A single GT47A-VII (SlGT11, Solyc03g115750) is found encoded in the tomato genome (Fig. 2A). However, its ortholog in Arabidopsis, AtGT11, is thought to be a pollen tube–specific xyloglucan galactosyltransferase (Wei et al. 2021). Therefore, we re-examined the previous phylogeny (Yu et al. 2022) and conducted gene expression analyses to identify any additional MBGT candidates in tomato. Tomato has nine enzymes in the GT47A clade (Fig. 2A). Three of them (SlMUR3 and SlXST1,2) have had their activities confirmed by in vivo complementation analysis of the Arabidopsis xyloglucan galactose–lacking mutant mur3-1 xlt2 (Schultink et al. 2013). The other six are SlGT11, two AtXUT1 orthologs in GT47A-II (Solyc08g080930 and Solyc12g056260), one AtXLT2 ortholog in GT47A-III (Solyc02g092840), one GT19 ortholog (GT47A-IV; Solyc08g079040) and one GT17 ortholog (GT47A-V; Solyc08g080930). Interestingly, a previous study reported that Solyc02g092840 did not complement the mur3-1 xlt2 phenotype (Schultink et al. 2013), raising the possibility that Solyc02g092840 works on a different substrate. In line with this, gene expression data from fruits showed that Solyc02g092840 was the most highly expressed GT47A gene through fruit development [that β-GGM is abundant in tomato fruits has been shown in our previous paper (Yu et al. 2022)] (Fig. 2B). In contrast, other genes encoding potential MBGT candidates, such as SlGT11, AtXUT1 orthologs and SlGT17, had lower expression levels. Hence, we selected Solyc02g092840 as an MBGT candidate. Screening for candidate tomato MBGT. Biochemical activity of Solyc02g092840. Solyc02g092840 exhibits MBGT activity To investigate β-galactosyltransferase activity of Solyc02g092840 in vitro on β-GGM, we transiently expressed the gene Solyc02g092840 in tobacco (Nicotiana benthamiana) leaves. In addition to the candidate, AtMBGT1 (a positive control) and SlGT11 (the candidate and closest orthologous to AtMBGT1 in tomato) were also expressed. Microsomal fractions prepared from the transgenic material were subjected to an immunoblot, whereby expression of all three proteins was confirmed (Supplementary Data S1, Fig. S1). The microsomal fraction was used as an enzyme source and incubated with UDP-Gal and Arabidopsis seed mucilage containing β-GGM as the substrates (mucilage β-GGM lacks β-Gal because of the trimming by β-galactosidases [Voiniciuc et al. 2015, 2016, Yu et al. 2018, 2022)]. The reaction mixtures were treated with AnGH5 mannanase for structural characterization of the products. Digestion of the initial substrate yields oligos GA and GM [Fig. 3A, where A is Gal-α-(1,6)-Man, G is Glc and M is Man] (Yu et al. 2022). Digestion of AtMBGT1-treated substrate produced the longer oligos including GBGA and GBGM [where B is Gal-β-(1,2)-Gal-α-(1,6)-Man], because the mannanase cleavage is inhibited at β-galactosylated sites. These oligos were sensitive to β-galactosidase and the main products migrated to GAGA and GAGM (Fig. 3A). Similar to AtMBGT1, digestion of Solyc02g092840-treated material also released GBGA and GBGM. In contrast, SlGT11-treated material did not produce β-galactosylated oligos. These results indicate that Solyc02g092840, but not SlGT11, possesses MBGT activity in vitro. To confirm MBGT activity in vivo, we next performed complementation analysis. Solyc02g092840 was expressed in the Arabidopsis mbgt1 mutant under a strong constitutive promoter. The mannan structure in cell walls of young stem from wild-type (WT) (positive control), mbgt1 (negative control) and complementation lines (pro35S:Solyc02g092840/mbgt1) was analyzed using CjMan26A digestion. This mannanase released larger diagnostic β-GGM-derived oligos distinct from those from digestion of the AcGGM that is also present in the stem cell walls [e.g. MM and GMM, Yu et al. (2022))]. The β-Gal-containing oligosaccharides GBGAGM and GBGM were visible in WT and the complementation line, but these β-galactosidase-sensitive structures were absent in mbgt1, which produced only the α-galactosylated oligosaccharides GAGAGM and GAGM (Fig. 3B). Together with the in vitro activity evidence, we concluded that Solyc02g092840 encodes a MBGT that can contribute to the biosynthesis of β-GGM. Thus, we name it SlMBGT1. Genes from asterids form a putative MBGT subclade in GT47A-III The discovery that SlMBGT1 in tomato (an asterid) is in the GT47A-III subclade implies that this enzyme emerged separately from AtMBGT1 in Arabidopsis (a rosid), which resides in the GT47A-VII subclade. We considered whether the GT47A-III group of putative MBGT enzymes might be present more widely in plant lineages beyond tomato. Our previous paper reported asterid enzymes in three subclades within the GT47A-III defined as GT47A-IIIa, GT47A-IIIb and GT47A-IIIc (Wilson et al. 2023). SlMBGT1 is found in the previously uncharacterized subclade GT47A-IIIb. The xyloglucan galactosyltransferases such as AtXLT2 are associated with GT47A-IIIa. Xyloglucan arabinofuranosyltransferases from tomato (SlXST1,2) and cranberry (VmXST1) as well as xyloglucan β-xylosyltransferases from Robusta coffee (CcXBT1) and blueberry (VcXBT) are found in the c subclade (Immelmann et al. 2023, Wilson et al. 2023). Based on GT47 protein structural modeling and different activities found in these subclades, the putative donor specificity of GT47A-III subclades was speculated to vary with changes in five regions of amino acid sequence that form the donor sugar binding site (Wilson et al. 2023). Importantly, and in contrast to the XST and XBT sequences, the SlMBGT-containing GT47A-IIIb subclade has greater similarity in these donor site residues to those of the characterized galactosyltransferases, consistent with the notion that the GT47A-IIIb clade consists of enzymes sharing the UDP-Gal donor specificity (Wilson et al. 2023). Genes from multiple orders of the asterid group (Ericales, Cornales, Lamiales, Solanales, Apiales and Asterales) are found in this putative MBGT subclade, suggesting that other asterids may also have MBGT enzymes and have the capacity to synthesize β-GGM. Non-Solanales asterids exhibit β-GGM As the phylogenetic analysis suggested the presence of MBGT enzymes in different orders of asterids, these plants should be capable of synthesizing β-GGM. To test this, we collected tissues from blueberry (Vaccinium corymbosum, Ericales), carrot (Daucus carota, Apiales), olive (Olea europaea, Lamiales) and wild sunflower (Helianthus tuberosus, Asterales) and analyzed their mannan structures (Fig. 4). Coinciding with our prediction, β-galactosylated β-GGM was found in blueberry and olive fruits. Since blueberry is within Ericales, an early branching order of asterids, this suggests that the MBGT activity was already present early in asterid evolution. Interestingly, the subterranean parts from carrot and H. tuberosus had unsubstituted β-GGM, which was digested by the mannanase to GM and GMGM. Perhaps the β-GGM in these tissues is trimmed by galactosidases, as seen in Arabidopsis seed mucilage β-GGM (Yu et al. 2022). Consistent with the previous identification of β-GGM in kiwifruit (Schröder et al. 2001, Yu et al. 2022), a plant in the asterid Ericales group, this mannan structural analysis suggests that asterid orders possess β-GGM and so have MBGTs, and therefore the other asterid enzymes in the GT47A-IIIb subclade may also have MBGT activity. Mannan structure analysis of four species of asterids. Functional convergence may have occurred with different amino acid substitutions Given that MBGTs originated separately in asterids and rosids, we investigated whether the two distinct GT47 MBGT subclades (GT47A-VII in rosids and GT47A-IIIb in asterids) share the same evolutionary pathway to acquire MBGT activity. If they evolved through a common pathway, convergent amino acid substitutions should be observed at the same protein sites between the two MBGT subclades to allow β-GGM acceptor specificity to replace the xyloglucan acceptor specificity. To test this hypothesis, we performed an amino acid convergence analysis (Fukushima and Pollock 2023). First, based on a phylogenetic tree containing multiple asterid and rosid species, possible asterid and rosid orthogroups with MBGT activity were identified (Supplementary Data S1 Fig. S2). In asterids, the orthogroup containing SlMBGT corresponds to the GT47A-IIIb subclade from branchpoint 1605, and we therefore used this as the asterid comparator for the convergence analysis. The rosid orthogroup cannot yet be so precisely defined. We do not know if there is MBGT activity among the Arabidopsis AtMBGT1 paralogs (AT2G32750, AT2G32740 and AT2G31990), and no other rosid MBGTs have been identified, so it is not yet clear where MBGT activity diverged from the xyloglucan galactosyltransferase activity presumably possessed by the common ancestor with GT47A-VII (Fig. 2) (Yu et al. 2022). We therefore examined each of the stem branches between divergence from OsGT47A-VII and more recent subclades to investigate any amino acid convergence with the asterid enzymes (Supplementary Data S1 Fig. S2). The results did not identify any clear convergence of amino acid sequences that could explain acquisition of the β-GGM acceptor specificity. Thus, distinct amino acid substitutions, or changes that did not occur concurrently with the origins of the subclades, have likely contributed to the convergence of GT47A-III and GT47A-VII activities to MBGTs. Discussion Recent acquisition of β-GGM In this study, we identified a tomato MBGT located in GT47A-III, which is distinct from the subclade GT47A-VII to which the rosid AtMBGT1 belongs. This suggests the functional convergence of GT47A enzymes to acquire MBGT activity and hence also convergence of the β-GGM structure, during eudicot diversification. Furthermore, we found that asterid enzymes form a putative MBGT subclade (GT47A-IIIb) near the XST/XBT subclade (GT47A-IIIc). This subclade is asterid-specific (Yu et al. 2022, Wilson et al. 2023). Therefore, asterids diversified their GT47A-III enzymes after species divergence from other eudicot families. Interestingly, there are similarities in the evolutionary processes that led to MBGT activity in rosids and asterids. In asterids, GT47A-III is likely to have undergone three necessary gene duplication events. At the first divergence in a presumed clade of XLT2 xyloglucan galactosyltransferases, ancestors of XST/XBT/MBGT emerged. Then, the XST/XBT/MBGT ancestor split to create GT47A-IIIb (SlMBGT1 subclade, β-GGM acceptor) and GT47A-IIIc (XST/XBT, xyloglucan acceptor). Interestingly, the rosid MBGT in GT47A-VII may have also evolved from a xyloglucan galactosyltransferase. AtMBGT1 diverged from the common ancestor with AtGT11 (GT47A-VII, suggested to be xyloglucan galactosyltransferase) (Wei et al. 2021) and ultimately from a common ancestor with AtMUR3 (GT47A-VI) xyloglucan galactosyltransferase (Fig. 2A). It therefore appears that the acquisition of MBGT activity each time involved a duplication event of genes encoding a xyloglucan galactosyltransferase, and subsequent accumulations of amino acid substitutions allowed the resulting β-galactosyltransferases to accept patterned galactoglucomannan in place of xyloglucan. Despite these similarities, we were unable to identify any convergent changes in amino acid sequence that might suggest critical residues for this change in acceptor binding. Interestingly, XyG fucosylation also appears to have arisen in different GT37 subclades in the course of evolution (Mikkelsen et al. 2021), suggesting that multiple occurrences of selection of GT activity to produce particular plant cell wall polysaccharide structures an important and more general phenomenon. Do MBGTs from distinct subclades have different activity? AtXLT2 and AtMUR3 are responsible for galactosylation of different positions on xyloglucan, where AtXLT2 and AtMUR3 add galactosyl residues on the second and third positions of a XXXG unit, respectively (Madson et al. 2003, Jensen et al. 2012). Given the two-fold screw conformation of the xyloglucan backbone (Levy et al. 1997), the xylosyl residues of the XXXG unit position alternately on one side to the other of the backbone. In other words, two xylosyl residues of XXXG are on the opposite face to the one in the middle (XXXG), making the backbone appear xylosylated every other glucosyl residue on one side of the polymer. In contrast, the opposite side of the backbone appears xylosylated every four residues. It is therefore likely that AtXLT2 and AtMUR3 recognize the frequency of xylosyl residues on XyG and thus bind the xyloglucan substrate very differently to position the correct xylosyl residue for galactosylation. This notion led us to speculate that since the two MBGTs from rosids and asterids inherited their substrate recognition from the different ancestral xyloglucan galactosyltransferases, they might also bind their β-GGM substrates differently. In the case of β-GGM, the α-galactosyl substitutions of mannosyl residues always face just one side of the backbone due to the Glc-Man repeating unit of the backbone. When the mannosyl residues are frequently galactosylated (e.g. GAGAGA), every other mannosyl residue of the backbone would be decorated, which structurally mirrors the side XyG backbone that AtMUR3 recognizes. Therefore, the MBGT enzymes that emerged in rosids (e.g. AtMBGT1) could better recognize the frequently galactosylated β-GGM. On the other hand, SlMBGT1, a descendant of XLT2 type galactosyltransferases, may favor the substrate decorated with less α-Gal on the β-GGM backbone such as GMGAGM patterns. Indeed, there seemed to be a difference in substitution patterns of β-GGM between Arabidopsis and tomatoes, where Arabidopsis showed a similar proportion of GBGA and GBGM, while tomatoes primarily had GBGM with little GBGA (Yu et al. 2022). It would be interesting to investigate further whether the β-GGMs in rosids and asterids have distinct substitution patterns and whether such structural differences change the functionality of β-GGM. Open question: what functions do glucomannans with different structures have? We previously noted that both β-GGM and xyloglucan have structural and functional seminaries (Yu et al. 2022). Here, we studied particularly the acquisition of the β-galactosylated disaccharide side chains in β-GGM, the synthesis of which now appears to have evolved at least twice in eudicots. Some structural variations of β-GGM have now been found, from the relatively highly β-galactosylated form in Arabidopsis primary cell wall (Yu et al. 2022) and blueberry fruit to the β-GGM from Arabidopsis mucilage that lacks almost all β-galactosylation (Yu et al. 2018) and the unsubstituted form in carrot root and wild sunflower tuber (Fig. 4). These structural differences appear to be the result of biosynthetic gene expression variation, different specificity of MAGTs and MBGTs and the production of hydrolases that trim the side chain (e.g. MUM2 β-galactosidase in mucilage biosynthesis), rather than the absence of decoration enzymes in the genomes of these different eudicots. Although these structural variations may provide physiological adaptations, their specific role in growth and development remains to be elucidated. Despite the absence of some aspects of the structure in some organs, the convergent evolution of β-GGM biosynthesis to possess a similar structure in both rosids and asterids strongly suggests that there has been an adaptive advantage to possession of the fully galactosylated β-GGM structure in eudicot cell walls. Why both xyloglucan and β-GGM have evolved this structure is an important outstanding question. Materials and Methods Plant materials Arabidopsis and tobacco (N. benthamiana) were grown at 21°C under 16-h day and 8-h night conditions (Ishida et al. 2023). The Arabidopsis mbgt1-1 mutant (SALK_065561) was as previously studied (Yu et al. 2022). Blueberry fruits and carrot roots were purchased at a local supermarket. Brined green olives were obtained from a supermarket and used after washing them with extensively with water. Helianthus tuberosus was grown in a home garden in Cambridge (UK), and subterranean tubers were harvested in January 2023. Mannan structural analysis Mannan structural analysis was conducted using Polysaccharide Analysis by Carbohydrate gel Electrophoresis (PACE) as described previously (Ishida et al. 2023). Briefly, alcohol-insoluble residues and a 4 M KOH alkali-soluble hemicellulose fraction were prepared. Then, mannan was digested sequentially by AnGH5 or CjMan26A mannanase and Aspergillus niger GH35 β-galactosidase (E-BGLAN, Megazyme, Bray, Ireland). The digestion products were labeled by reductive amination with 8-aminonaphthalene-1,3,6-trisulfonic acid and separated by carbohydrate gel electrophoresis. Molecular cloning Tomato gene expression data are from Tomato Genome Consortium (2012) accessed through the BAR expression browser https://bar.utoronto.ca/eplant_tomato/. For tobacco transient protein expression, the SlMBGT1 gene was cloned into the pHREAC vector as previously described (structure: CaMV 35S promoter + CDS + TEV + eGFP + 10× His-tag + Nos terminator) (Peyret et al. 2019). SlGT11 (structure: CaMV 35S promoter + CDS + TEV + eGFP + 6× His-tag + Nos terminator) was cloned into a vector containing a P19 suppressor of gene silencing from tomato bushy stunt virus flanked by the CaMV 35S promoter and Nos terminator, which was constructed using a Golden Gate–based system (Engler et al. 2009). The AtMBGT1-Myc expression construct was prepared previously (Yu et al. 2022). The SlMBGT1 expression vector for in vivo complementation was constructed with a Golden Gate–based system (Engler et al. 2009), in which the strategy and other cloning parts were the same as described previously (structure: CaMV 35S promoter +CDS + Myc-tag + Nos terminator) (Ishida et al. 2023). The coding sequences of SlGT11 and SlMBGT1 were prepared by a commercial gene synthesis service (Integrated DNA Technologies, Coralville, IA, USA). Nucleotide sequences of primers are described in Supplementary Data S1, Table S1. Heterologous expression and in vitro assay of MBGT Seven-week-old tobacco (N. benthamiana) leaves were infiltrated with the Agrobacterium AGL1 strain harboring the constructs described earlier to express the GTs of interest and were harvested 4 d after infiltration. Microsome preparation for enzyme assays and immunoblotting to confirm expression were performed as described previously (Ishida et al. 2023). For Myc-tagged AtMBGT1, Ab9106, a rabbit anti-Myc antibody (Abcam, Cambridge, UK), was used. For SlMBGT1 and SlGT11, Ab290 rabbit anti-GFP antibody (Abcam) was used. Goat anti-rabbit IgG-HRP conjugates (Bio-Rad, Hercules, CA, USA) were used to detect primary antibodies. The in vitro assay for MBGT activity in the microsomal membrane fractions using acceptor Arabidopsis mucilage β-GGM and donor UDP-Gal was carried out as described except that the reaction time was changed from 5 h to overnight (Yu et al. 2022). In vivo complementation Arabidopsis mbgt1 (Yu et al. 2022) was transformed by Agrobacterium-mediated floral dip using constructs to test MBGT activity as described previously (Ishida et al. 2023). Transformants were chosen by their seed eGFP signal and kanamycin resistance. Then, genotyping was carried out using primers in Supplementary Data S1, Table S1. Phylogenetic analysis The species phylogeny for 24 angiosperm species [ANA-grade: A. trichopoda, Nymphaea colorata, Monocots: O. sativa, Eudicots Rosids: Arabidopsis thaliana (Brassicales), Brassica rapa (Brassicales), Capsella rubella (Brassicales), Citrus clementina (Sapindales), Eucalyptus grandis (Myrtales), Fragaria vesca (Rosales), Gossypium hirsutum (Malvales), Populus trichocarpa (Malpighiales), Theobroma cacao (Malvales), Vigna unguiculata (Fabales), Eudicots Asterids: D. carota (Apiales), Helianthus annuus (Asterales), Hydrangea quercifolia (Rosales), Lactuca sativa (Asterales), Lindenbergia philippensis (Lamiales), O. europaea (Lamiales), Solanum lycopersicum (Solanales), Solanum tuberosum (Solanales), Vaccinium darrowii (Ericales), Vitis vinifera (Vitales) and other Eudicots: Aquilegia coerulea (Ranunculales)] with sequenced genomes was obtained from timetree.org. The phylogenetic analysis of GT47A-III was performed according to previous work (Fukushima and Pollock 2023). In brief, genes homologous to the Arabidopsis GT47A-III were retrieved from the publicly available protein-coding gene sets of the 24 species using TBLASTX v2.9.0 with an E-value cutoff of 0.01 and >50% query coverage. In-frame codon sequence alignment was performed using MAFFT v7.455 (https://mafft.cbrc.jp/alignment/software/), ClipKIT v0.1.2 (https://github.com/JLSteenwyk/ClipKIT) and CDSKIT v0.9.1 (https://github.com/kfuku52/cdskit). The gene phylogeny was reconstructed using IQ-TREE v2.0.3 (https://github.com/iqtree/iqtree2) with the general time-reversible nucleotide substitution model and four gamma categories of among-site rate variation, and it was then reconciled with the species tree using GeneRax v1.2.2 (https://github.com/BenoitMorel/GeneRax). Amino acid convergence analysis The reconciled gene tree and the codon alignment were used as inputs for CSUBST v1.1.0 (https://github.com/kfuku52/csubst) to infer the codon substitution history, which was then translated using the standard genetic code to obtain the probabilities of amino acid substitutions, as well as their convergence and divergence between focal branches. The inferred substitutions were mapped to the protein structure of AtMBGT1 predicted by AlphaFold2 (Q84R16) using the ‘site’ function of CSUBST. Posterior probabilities of convergent or divergent substitutions greater than 0.5 were reported. Supplementary Material