Unfortunately, misannotations of GM35E as GalE or the related GDP-4-keto-6-deoxy-D-mannose-3,5-epimerase-4-reductase (GMER, also known as GDP-l-fucose synthase, GFS) (or vice versa) still occur for many automatic annotations of enzymes in databases. Our recent review and in silico analysis of NDP-sugar active SDR (NS-SDR) enzymes (Da Costa et al., 2021), also highlighted additional motifs and residues that can be applied to correctly identify and annotate novel GM35E. Inclusion of this heptagonal box model in the algorithms could improve the automated annotations. Indeed, GM35E and GMER clearly differ from GalE by their presence of an additional catalytic acid and base, which are needed for the double epimerization. The difference between GM35E and GMER is that they employ slightly different residues; the catalytic acid is in both cases a Cys (yellow wall), but the catalytic base is a Lys for the epimerase but a His for the epimerase-reductase (purple wall). In addition, also the residue in charge of acidifying the catalytic Cys is different, namely an Arg for GM35E and a Lys for GMER (green wall). They also show different Glycine motifs and different residues in between the catalytic Tyr and Lys (cyan wall). In addition, also the other three walls (red, purple and grey) of the heptagonal box model show to be slightly different for GM35E and GMER (Figure 3).

In terms of rare sugar production, epimerases play a key role as they can bridge between abundant d-sugars and their rare l-counterparts (Beerens et al., 2017) and this also holds true for the GM35E enzyme. Indeed, GM35Es (no matter their origin, microbial or plant) can theoretically be applied in a biocatalytic pathway to produce l-Gal, l-Gul and/or d-Alt. However, it has until now only been shown in practice for l-Gal starting from GDP-Man (Figure 6, in black) (Gevaert, 2020). After initial epimerization of GDP-Man to GDP-l-Gal (and side products GDP-l-Gul and GDP-d-Alt), the free l-Gal monosaccharide can be obtained by the addition of two other enzymes from the Smirnoff-Wheeler pathway. Starting from GDP-Man, Gevaert and co-workers achieved l-Gal with good yield (75%) and high purity (∼94%), due to the high selectivity of the additional enzyme(s) (Gevaert, 2020). Unfortunately, the GDP-Man used as substrate here remains a scarce and expensive substrate to produce l-Gal. Nonetheless, the Nidetzky group published an interesting pathway to produce GDP-Man from cheap sucrose and mannose in a kinase-independent one-pot multi-enzyme cascade (Pfeiffer et al., 2016), which could allow cheaper synthesis of GDP-Man (Figure 6, in blue). This setup for GDP-Man synthesis only requires one expensive NTP molecule (1 GTP), in comparison to the natural pathway from Glc or Man that requires 2 NTP molecules (1 ATP and 1 GTP) (Figure 1). Furthermore, another promising biocatalyst that could further improve GDP-Man (and thus l-Gal) synthesis is the recently discovered promiscuous CDP-paratose 2-epimerase (CPa2E), which is able to epimerise GDP-Glc to GDP-Man (Rapp et al., 2021). Hence, in combination with the sucrose synthase (SuSy) enzyme, this 2-epimerase could produce GDP-Man starting from sucrose and GDP (over GDP-Glc as pathway intermediate) (Figure 6, in red). Unfortunately, the CPa2E’s low activity currently remains a bottleneck. On the other hand, an additional benefit from this multi-enzyme pathway using SuSy and CPa2E is that the GDP released by the GDP-l-galactose phosphorylase can be recycled by the SuSy enzyme and thus requires only catalytic GDP quantities. In a similar manner the phosphate needed for the l-galactose 1-phopshate phosphatase would also be recycled (Figure 6, dashed arrows). However, implementation of the steps to produce l-Gal directly from sucrose remains to be investigated.

Another potential exploitation area of GM35E is situated in the technology called “glycorandomization” (sometimes also called “glycodiversification”, Figure 7). Since the majority of epimerases (such as GM35E) are active on NDP-sugars, their products can subsequently be valorized via for example glycosylation reactions. Glycosylation is the attachment of a carbohydrate (glycon) to another molecule (aglycon), yielding so-called glycosides. The aglycones can be very diverse and range from small molecules (e.g., vitamins, antibiotics, flavors and fragrances, etc.) to macromolecules (e.g., proteins, lipids and cell wall glycans) (De Bruyn et al., 2015; Desmet et al., 2012). Glycosides are widespread in nature and display a variety of interesting applications (Keegstra and Raikhel, 2001; Brandle and Telmer, 2007) and can show altered features compared than the free aglycon (e.g., increased water solubility or reduced chemical reactivity) but also based on the glycol attached (e.g., improved chemical stability or altered biological activity) (Jones and Vogt, 2001). Hence, glycosides can be applied in many industrial disciplines as for instance nutraceuticals, therapeutics or food additives (De Bruyn et al., 2015; Desmet et al., 2012). As pharmaceuticals, glycosides can both be used as an innovative approach for targeted drug delivery (Lepenies et al., 2010; Chen and Huang, 2019) or as an active moiety for therapeutic activity (Palcic, 2011). Furthermore, numerous antibiotics are glycosides and in many cases, the sugar entity is crucial to their activity and selectivity (Thorson et al., 2001; Kren and Rezanka, 2008; Song et al., 2013). For example, the antitumor antibiotic bleomycin contains a disaccharide moiety composed of l-Gul and 3-O-carbamoyl-d-Man, of which the l-Gul sugar unit possessed tumor cell targeting properties as it enables the uptake of the drug (Schroeder et al., 2014). Another example are macrolide antibiotics, such as erythromycin, which operate by binding bacterial ribosomes, thereby inhibiting protein synthesis. This binding is realized by their sugar moieties (Schroeder et al., 2014). Furthermore, also the l-Gal subunit of the nucleoside antibiotic A201A was found to be essential for the drug’s bioactivity (Zhu et al., 2017) and l-sugar containing analogs of the phytosteroid digitoxin showed an improved anti-cytomegalovirus activity (Cai et al., 2014). This effect of the carbohydrate moiety on therapeutics led to an increasing interest in unusual glycosides. On that account, glycorandomization has become a valuable tool for the expansion and optimization of various glycoside antibiotics. In this technology, the carbohydrate moiety of antibiotics is altered in order to tailor their pharmacological properties and/or biological activity (De Bruyn et al., 2015; Thibodeaux et al., 2007; Desmet et al., 2012). This approach facilitates drug screening and discovery and might be a powerful strategy in the battle against antibiotic resistance (Ventola, 2015; Mohs and Greig, 2017). Exotic NDP-sugar libraries can be created by a collection of epimerase specificities. These can be combined with aglycon libraries, consisting of both natural products and synthetic compounds, to generate novel glycorandomized libraries (Figure 7). The practice of the glycorandomization technology resulted in unnatural sugar analogues of vancomycin and the macrolide antibiotic YC-17 with improved antibacterial activity as compared to the natural product (Fu et al., 2005; Shinde et al., 2013). Moreover, erythromycin variants that regained activity against an erythromycin-resistant strain have been obtained (Zhang G. et al., 2015).

GM35E shows excellent capacity to make such NDP-sugar libraries as it would make a mixture of GDP-Man, GDP-l-Gal, GDP-l-Gul and GDP-d-Alt. Coupling to glycosyltransferases (GT, EC 2.4), which perform the majority of glycosylation reactions in nature (Desmet et al., 2012), is of course necessary since GTs catalyze the sugar transfer from a sugar donor to an acceptor molecule to generate glycosides with high efficiency and enantioselectivity (Lim, 2005). Within this large enzyme family, the Leloir glycosyltransferases use NDP-sugars as sugar donors (Lairson et al., 2008; Mestrom et al., 2019). GTs have been applied for the synthesis of natural glycosides and artificial derivatives (Luzhetskyy and Bechthold, 2008; Palcic, 2011), and have contributed to successful in vivo and in vitro glycodiversification projects (Blanchard and Thorson, 2006; Thibodeaux et al., 2008). Despite most GT specificities discovered to date are highly specific towards both the donor and acceptor (Henrissat and Davies, 2000; Bowles et al., 2005), several promiscuous GTs have also been reported and could thus be used for glycodiversification. For example, SorF and OleD display flexible donor and acceptor specificities, respectively (Bolam et al., 2007; Kopp et al., 2007). In the case of OleD, its promiscuity is presumed to derive from its natural function as a defense mechanism against macrolide antibiotics. More interestingly, YjiC is a promiscuous GT with regard to both donor and acceptor molecules, spanning various chemical classes (Pandey et al., 2014). In addition, enzyme engineering already extensively contributed to the expansion of GT promiscuity (Williams et al., 2007; Jakeman, 2008; Williams et al., 2008; Chen, 2011; Gantt et al., 2011).

Similar as for several epimerases, the large-scale application of GTs is restricted by their need for NDP-sugars as substrates, which are expensive and hard to acquire in large quantities (Masada et al., 2007). Furthermore, it was suggested that NDP released during the transfer reaction can act as an inhibitor of GT activity, resulting in lower yields (Masada et al., 2007; Terasaka et al., 2012; Huang et al., 2016). Systems biocatalysis, the concept of creating cell-free “artificial metabolisms” for preparative multi-enzymatic synthesis (Fessner, 2015; Tessaro et al., 2015), might be applied to tackle these issues. Indeed, the use of regeneration systems allows the in situ regeneration of NDP-sugars, resulting in reduced costs and increased productivity (Ichikawa et al., 1994; Elleuche, 2015). Several of these systems including sucrose synthase (SuSy) have been developed to date, yielding for example glucosides of curcumin, resveratrol and quercetin (Masada et al., 2007; Terasaka et al., 2012; Lepak et al., 2015; Schmölzer et al., 2016). The NDP is recycled and the NDP-sugar regenerated by SuSy, yielding an overall reaction that would only require sucrose and catalytic amounts of NDP to proceed (Lim, 2005; Masada et al., 2007). Such cascade reaction could be expanded with epimerases (like GM35, but also CPa2E and/or GalE) to obtain various rare NDP-sugars as starting point for glycosylation, similar to the pathway shown in Figure 6, however, with promiscuous GTs at the end to achieve synthesize of a library of glycosides instead of releasing l-Gal. In a pathway, the equilibrium of the epimerization reactions is not an issue since the irreversible GT reaction pulls the overall cascade towards product formation.

Since l-ascorbic acid (also known as vitamin C) is an important plant antioxidant with important metabolic functions in both plants and animals (obtained via a plant-based diet), its plant biosynthesis has been studied intensively (Smirnoff and Wheeler, 2000) and plant engineering had already early on been suggested to achieve increased l-ascorbic acid production to increase both plant stress tolerance and nutritional value of humans and animals (Wheeler et al., 1998). Since both l-galactose (Wheeler et al., 1998) and l-gulose (Valpuesta and Botella, 2004) are important intermediates for l-ascorbic acid biosynthesis and both are derived from GDP-activated d-mannose, their synthesis is strongly dependent on the GDP-mannose-3,5-epimerase (and downstream enzymes). Hence, it is not surprising that GM35E and its pathway-connected enzymes have been studied in different plant species, including rice (Watanabe et al., 2006; Zhang G.-Y. et al., 2015), peach (Imai et al., 2009), cabbage (Ren et al., 2013), tomato (Mounet-Gilbert et al., 2016; Li et al., 2019), amongst others. Indeed, transgenic tomato plants over-expressing GM35E exhibited significantly increased l-ascorbic acid levels in leaves and red fruits compared with wild-type plants, which led to enhanced tolerance of different types of abiotic plant stress, incl. herbicide, oxidative stress, cold, and salt stress (Zhang et al., 2011). Similarly, transgenic GM35E overexpression in Arabidopsis also enhanced acid, drought and salt tolerance by increased l-ascorbate accumulation (Ma et al., 2014) and manipulation of the rice l-galactose pathway (incl. GM35E overexpression) also revealed enhanced salt stress tolerance (Zhang G.-Y. et al., 2015). However, this study by Zhang et al. (2015) suggest that another enzyme involved in the pathway, namely the GDP-l-galactose phosphorylase (GGP), may be a key rate-limiting step l-ascorbic acid biosynthesis, at least in rice. This, in fact, highlights that the entire l-ascorbic acid biosynthesis pathway is important and not only the GM35E. Indeed, studies with transgenic tobacco revealed that the equilibrium of the GM35E reaction is unfavorable to forward l-ascorbic acid biosynthesis, thus indicating a complex modulation (Imai et al., 2012). On the other hand, it has also been found that GM35E plays a key role beyond l-ascorbic acid, namely in non-cellulosic cell-wall biosynthesis. Indeed, GM35E silencing in tomato affected rhamnogalacturonan II (RG-II) structure (approx. 60% decrease in terminal l-Gal content in RG-II’s side chain A), hereby leading to a lower cross-linking capacity of the pectic polysaccharide RG-II and hindered normal plant growth and development (Voxeur et al., 2011). Similarly, earlier studies with transgenic tomato lines also showed growth defects affecting both cell division and cell expansion, as well as altered cell-wall monosaccharide content, especially mannose and l-galactose that is directly linked to GM35E activity, leading to plant fragility and loss of fruit firmness (Gilbert et al., 2009).

Considered all together, these findings highlight that GM35E activity is very important in plants, not only for l-ascorbic acid biosynthesis and the linked induced stress tolerance but also for its involvement in RG-II biosynthesis (Gilbert et al., 2009; Voxeur et al., 2011), needed for strong healthy plants and firm fruit. Hence, future applications of plant-derived GM35Es will most likely occur via plant engineering studies (transgenic overexpression) to enhance stress tolerance, obtain stronger plants and/or healthier and firmer fruit. Here, we recommend considering all enzymes involved in the l-ascorbic acid pathway instead of only the GM35E. Similarly, future studies focused on targeting RG-II biosynthesis should also focus beyond only GM35E and target the entire pathway.

In a similar fashion as described for the l-galactose production (as a monosaccharide) and glycans biosynthesis in vivo in plants, GM35E has also been applied for the synthesis of l-galactose-containing N-glycans in vitro (Ohashi et al., 2017). In this case, Ohashi et al. carried out a preparative scale GDP-l-Gal synthesis starting from GDP-Man by using recombinant A. thaliana GM35E followed by desalting, recycling high performance liquid chromatography and lyophilization steps. The produced and purified GDP-l-Gal was consecutively used as substrate for l-galactosylation, which was achieved using mouse α1,6-fucosyltransferase (MmFUT8) or A. thaliana α1,3-fucosyltransferase (AtFucTA) (Ohashi et al., 2017). Indeed, it had already been shown that multiple α-l-fucosyltransferases (FucT) were promiscuous towards GDP-l-Fuc analogs like GDP-l-Gal, leading to the formation of l-galactosylated Lewis structures (Stangier et al., 1997; Düffels et al., 2000). In these earlier cases, the GDP-l-Gal was synthesized either chemically or enzymatically using l-fucokinase/GDP-l-fucose pyrophosphorylase (FKP) rather than utilizing GM35E and starting from GDP-Man. The promiscuous nature of FucT has also been observed in vivo in the l-fucose deficient mur1 mutant of A. thaliana (Zablackis et al., 1996), in which l-galactose replaces the absent l-fucose as a required component of the biologically active cell wall xyloglucan-derived oligosaccharides. l-Galactosylation of these xyloglucan oligosaccharides by AtFucT has recently also been demonstrated in vitro (Ohashi et al., 2019). In vivo, the GDP-l-Gal is most likely obtained via de novo synthesis over GDP-Man and thus utilizing the GM35E. Likewise, future production of other l-galactosides and l-galactose-containing glycans can utilize a combination of both GM35E and promiscuous l-fucosyltransferases. This setup is both achievable for biotransformations (in vivo) and biocatalytic production (in vitro). However, especially for the in vitro setup, it will be important to keep in mind the most efficient production route for GDP-l-Gal, either over GDP-Man using GM35E or methods starting from l-Gal directly.

Instead of using the GM35E’s major product (GDP-l-Gal), one could also aim at utilizing its minor side products (e.g., GDP-l-Gul). By using GTs that would (specifically) transfer l-gulose to acceptors, the same setup of GM35E coupled to GT could be used to produce l-gulosides. An example of a natural l-guloside is the main polar lipid of Thermoplasma acidophilum. Also here, the biosynthesis route for the activated l-gulose is expected to pass over an GM35E or homologous enzyme (Yamauchi and Nakayama, 2013). These natural sources of l-Gal and l-Gul could also be interesting targets to search for novel GM35E’s and/or GT enzymes that would be able to utilize GM35E’s products as substrate for glycoside production.