Mechanistic Insights into Validoxylamine A 7'-Phosphate Synthesis by VldE Using the Structure of the Entire Product Complex

The pseudo-glycosyltransferase VldE catalyzes non-glycosidic C-N coupling between an unsaturated cyclitol and a saturated aminocyclitol with the conservation of the stereochemical configuration of the substrates to form validoxylamine A 7′-phosphate, the biosynthetic precursor of the antibiotic validamycin A. To study the molecular basis of its mechanism, the three-dimensional structures of VldE from Streptomyces hygroscopicus subsp. limoneus was determined in apo form, in complex with GDP, in complex with GDP and validoxylamine A 7′-phosphate, and in complex with GDP and trehalose. The structure of VldE with the catalytic site in both an “open” and “closed” conformation is also described. With these structures, the preferred binding of the guanine moiety by VldE, rather than the uracil moiety as seen in OtsA could be explained. The elucidation of the VldE structure in complex with the entirety of its products provides insight into the internal return mechanism by which catalysis occurs with a net retention of the stereochemical configuration of the donated cyclitol.


Introduction
Glycosyltransferases comprise one of the most numerous and diverse groups of enzymes in nature. They are responsible for the formation of oligo/polysaccharides, glycoproteins, glycolipids, and many other glycosylated natural products by transferring a sugar moiety from an activated donor sugar to a sugar (or non-sugar) acceptor. This abundant group of proteins consists of 92 families encoded by more than 83,400 genes [1]. However, only a fraction of those genes has actually been functionally characterized. Our comparative bioinformatics studies suggest that among those reported as glycosyltransferases are also pseudo-glycosyltransferases (such as VldE, EC 2.x.x.x), which do not recognize sugars as substrates but rather catalyze the formation of non-glycosidic C-N bonds in the biosynthesis of C 7 N-aminocyclitol-containing natural products such as acarbose and validamycin A ( Figure 1) [2][3][4]. Acarbose, an a-glucosidase inhibitor, has been proven useful in the treatment of type II insulin-independent diabetes, whereas validamycin A, a natural trehalase inhibitor, is an antifungal antibiotic that has long been used to protect crops from soil borne diseases such as rice sheath blight and the dumping-off of cucumber seedlings [5][6][7][8].
The mechanism by which non-glycosidic C-N bond is formed by a pseudo-glycosyltransferase is not entirely understood. However, because of the structural similarity of validoxylamine A 79-phosphate to trehalose 6-phosphate (Figure 1), it has been speculated that the mechanism of the pseudo-glycosyltransferase VldE is similar to that of the glycosyltransferase, trehalose 6phosphate synthase (OtsA, EC 2.4.1.15) [4]. VldE and Escherichia coli OtsA only share a modest 19% sequence identity (29% homology) (Figure 2), but they are both catalogued as members of the GT20 glycosyltransferase family by the CAZy database (www. cazy.org) [15]. OtsA mediates the transfer of glucose moiety from UDP-glucose to glucose 6-phosphate to form trehalose 6phosphate ( Figure 1). Similar to VldE, the product of OtsA conserves the anomeric configuration of the donor moiety. Glycosyltransferases have been shown to both retain and invert the anomeric state of the carbon C-1 of the donor moiety. The inversion of the anomeric center by glycosyltransferases has been well explored and is known to be carried out by a simple nucleophilic substitution. However, the underlying catalytic mechanism of glycosyltransferases that retain the anomeric configuration of the donated moiety within the product is not as well understood. Catalysis by a retaining glycosyltransferases is thought to occur through either a double displacement (S N 2 X2) or internal return mechanism (S N i) [16]. For a double displacement nucleophilic substitution reaction to occur, a nucleophilic catalytic base must be available in close vicinity to the sugar anomeric carbon to form a covalent intermediate. OtsA apparently lacks such a residue in the catalytic site and analyses of theoretical energy profiles along with recent studies of kinetic isotope effects have substantiated this unusual enzymatic reaction mechanism within OtsA [17][18][19]. Lastly, structural studies of OtsA using a bi-substrate inhibitor as a product mimic place the atoms involved in catalysis in an orientation favorable for an internal return mechanism and show that the leaving phosphate acts as a general base to deprotonate the incoming nucleophile of the acceptor group [20]. Within the S N i mechanism of OtsA, a carbocation is developed upon the detachment of the nucleotide phosphate which is stabilized by resonating into an oxonium ion-like transition state [18,20]. However, no oxonium ion formation is possible in the VldE-catalyzed non-glycosidic C-N coupling reaction. Alternatively, it is proposed that the olefinic moiety of GDP-valienol may play a critical role in facilitating the coupling reaction [4]. Coupling reactions involving an allylic moiety have been demonstrated in other biosynthetic enzymes, e.g., farnesyl diphosphate (FPP) synthases [21,22]. However, in FPP synthases, a nucleophilic substitution reaction takes place at a carbon center with a diphosphate acting as a leaving group instead of nucleotidyl diphosphate. In addition, mechanistically FPP synthases adopt a stereospecific S N 1 reaction with an inversion of the configuration, leaving the actual mechanism behind the unique catalytic function of the retaining VldE enzyme unclear. Hopefully, structural studies would clarify whether or not a S N i mechanism is conserved within the retaining pseudo-glycosyltransferase, VldE.
Here, we report the three-dimensional structures of VldE in various liganded states using X-ray crystallographic techniques. The structure of VldE was solved by molecular replacement using the structure of OtsA as a search model. We have elucidated the structures of the unliganded VldE, in complex with guanosine 59- The product and substrates of the VldE and OtsA catalyzed reactions are shown. Note the considerable similarity between the ligands of VldE and OtsA, and the conservation of the anomeric centers. The remainders of both biosynthetic pathways are then drawn to completion. VldE catalyzes the formation of validoxylamine A 79-phosphate via a non-glycosidic C-N bond between GDP-valienol and validamine 7-phosphate. After the validoxylamine A 79-phosphate has been produced; VldH and VldK complete the catalytic synthesis of Validamycin A. OtsA catalyzes the formation of trehalose 6-phosphate via a glycosidic bond between UDP-glucose and glucose 6-phosphate. OtsB dephosphorylates trehalose 6-phosphate to produce trehalose. doi:10.1371/journal.pone.0044934.g001 diphosphate (GDP), in complex with GDP and Trehalose (TRE), and in complex with GDP and VDO. Similar to OtsA, VldE is comprised of two Rossman b/a/b domains which are oriented in a GT-B configuration ( Figure 3). The active site, which is located at the interface of the Rossman domains, and the selective interactions allowing for the binding of GDP but not uridine 59diphosphate (UDP) are described in our study ( Figure 4A-B). The crystallographic investigation of VldE while binding the complete product complex, VDO and GDP, supports the proposed conservation of a S N i catalytic mechanism analogous to the mechanism of OtsA. Both the ''open'' and ''closed'' conformations of the catalytic site are also described within this study.

Purification and Crystallization
VldE from S. hygroscopicus subsp. limoneus was expressed and purified with methods similar to those previously described [4]. (His)6-tagged VldE was expressed in E. coli BL21 (DE3) pLysS and was purified using Ni-NTA affinity chromatography. Affinity Shown is the protein sequence alignment of VldE and OtsA as generated by ClustalW2 [54]. Residues whose side-chains are involved in ligand binding are indicated with shaded boxes. Conserved residues are indicated with unfilled boxes. The two-dimensional secondary structure of VldE and OtsA are also illustrated next to the corresponding sequence with a-helixes represented by spirals and b-strands represented by arrows. Numerical values for helices and strands are assigned. doi:10.1371/journal.pone.0044934.g002 chromatography was performed using 40 mM Hepes buffer, pH 7.5 with 300 mM NaCl, 20 mM imidazole, 5 mM bmercaptoethanol, and 10% glycerol. Elution from the Ni-NTA resin was achieved using the same buffer but with 175 mM imidazole. Using dialysis, the affinity chromatography buffer was exchanged with VldE storage buffer consisting of 10 mM Tris-HCl, pH 7.5 with 5% glycerol, 0.1 mM dithiothreitol, and 1 mM MgCl 2 . VldE was concentrated to 10 mg/mL, and crystals were grown using sitting drop vapor diffusion.
In the case of VldE and VldENGDP crystals, the mother liquor consisted of 100 mM Tris-HCl, pH 8.0 with 30-35% polyethylene glycol 3,350. GDP, at a concentration of 3 mM, was added to protein stocks which were intended to grow VldENGDP crystals by co-crystallization. Crystals with a size of 0.160.160.5 mm were grown over a period of four to twelve weeks. Crystallization took place at a temperature of 293 K. Cryoprotectant solutions, consisting of artificial mother liquor enriched with 20% ethylene glycol, were added to crystal drops at a 1:1(v/v) ratio. In the case of VldENGDP crystals, 5 mM GDP was added to the cryoprotectant solution. Crystals were flash-cooled in liquid nitrogen after soaking in the cryoprotectant solution for 20-40 minutes.
In the case of VldENGDPNVDO crystals, the mother liquor consisted of 100 mM Tris-HCl, pH 8.0 with 25-30% polyethylene glycol 3,350. The drop was formed through a mixture of mother liquor, protein, and seed stock in a ratio of 1.6:2.0:0.4 (v/ v). Seed stocks were prepared from crystal grown in similar conditions using Seed Beads (Hampton Research) in a stabilization buffer of 100 mM Tris-HCl, pH 8.0 with 31% polyethylene glycol 3,350. The Crystals with a size of 0.160.160.5 mm were grown over a period of four to twelve weeks at a temperature of 293 K. Cryoprotectant solutions, consisting of artificial mother liquor enriched with 20% ethylene glycol, were added to crystal drops at a 1:1(v/v) ratio. 5 mM GDP and 10 mM VDO were added to the cryoprotectant solutions. Crystals were flash-cooled in liquid nitrogen after soaking in the cryoprotectant solution for 20-40 minutes.
In the case of VldENGDPNTRE crystals, the mother liquor consisted of 50 mM Tris-HCl, pH 8.0, 18-28% polyethylene glycol 3,350, 200-500 mM NaCl, and 20 mM MgCl 2 . VldE was kept at a concentration of 10 mg/mL, and crystals were grown using sitting drop vapor diffusion with a 1:1(v/v) mixture of protein and mother liquor. Before the drops were set, 3% trehalose and 1 mM GDP were added to the protein stocks. The Crystals with a size of 0.160.160.1 mm were grown over a period of four to eight weeks at a temperature of 293 K. Cryoprotectant solutions, consisting of reservoir solution enriched with 20% ethylene glycol, 3% trehalose, and 1 mM GDP were added to crystal drops at a 1:1(v/v) ratio. Crystals were flash-cooled in liquid nitrogen after soaking in the cryoprotectant solution for 20-40 minutes.

Data Collection and Processing
All diffraction data except for the VldENGDP dataset was collected at the Northeastern Collaborative Access Team (NE-CAT) beamline at the Advanced Photon Source (Argonne National Laboratory). Data was recorded at 100 K on an ADSC Q315 (315 mm6315 mm) detector and processed by NE-CAT's RAPD automated processing (https://rapd.nec.aps.anl.gov/rapd), which uses XDS [23] for integration and scaling. A 2.0 Å VldE dataset was collected at a wavelength of 0.9479 Å while oscillating the crystal 0.5 degrees for each frame. A 2.11 Å VldENGDPNVDO dataset was collected at a wavelength of 0.9792 Å while oscillating the crystal 1.0 degrees for each frame. A 2.81 Å VldENGDPNTRE dataset was collected at a wavelength 0.9792 Å while oscillating the crystal 1.0 degrees for each frame. The 2.15 Å VldENGDP dataset was collected at NSLS Beamline X6A at a wavelength 1.0 Å while oscillating the crystal 1.0 degrees for each frame. Xray data was collected on a ADSC Quanta CCD detector at 100 K and was integrated, merged, and scaled using HKL2000 [24]. Although the cell dimensions were similar, the unliganded VldE crystal belonged to the P2 space group, while the VldENGDP and VldENGDPNVDO crystals belonged to the P2 1 space group.  OtsA and UDP within the OtsA active site (gray). The protein interactions with ribose and phosphate are conserved between VldE and OtsA. However, there are differing interactions with the nucleotide base groups. The large purine makes interactions with the residues Arg321, Asn323, Asn361, and Thr366. The ribose and phosphate moieties interact with Arg290, Lys295, Leu387, Ser388, and Glu391. Within OtsA, Leu344, Ile295, Pro297, and His338 only allow for the binding of the smaller pyrimidine. The ribose and phosphate moieties make interactions with residues Arg262, Lys267, Leu365, and Glu369. The mesh represents the |F o | 2 |F c | electron density omit map of the GDP binding site. The map is contoured at 3.0s levels. (C) Shown is a superimposition of the VldE cyclitol binding sites in the presence (yellow) and absence (green) of validoxylamine A 79-phosphate. The mesh represents the |F o | 2 |F c | electron density omit map of the VDO binding site. The map is contoured at 3.0s levels. VDO makes interactions with residues the side-chains of residues Asp158, His182, Arg12, Asn325, Arg326, and Asp383. VDO also makes interactions with the backbones of residues 384-387. Binding of the acceptor cyclitol is recognized by conformation changes by the side-chains of residues Asp158, Tyr159, and Arg326. (D) Shown is a comparison by superimposition of VDO binding within the catalytic sites of VldE (yellow) and OtsA (gray). Note the strong conservation of residue and ligand positions. doi:10.1371/journal.pone.0044934.g004 The unit cell dimensions of the VldENGDPNTRE dataset were dissimilar and belonged to the C2 space group. Diffraction data statistics are summarized in Table 1.

Structure Determination and Refinement
To determine the structure of VldE, molecular replacement (MR) was performed. Generation of a search model from the OtsA structure (PDB accession code 1UQU [25]) and molecular replacement was carried out using MrBUMP [26] to orchestrate molecular replacement via the MOLREP [27] module in the CCP4 [28] program suite. The resulting MR model was used to build a partially sequenced model (,400 residues) using the AUTOBUILD [29][30][31][32] function of the PHENIX [33] software suite. Further automated model building was carried out using the ARP/wARP [34] module within the CCP4 [28] software suite. The model was finished by manual building within COOT [35]. The model was refined using restrained refinement by REFMAC5 [36] with TLS parameters defined by the TLSMD server [37,38]. Two monomers were modeled within the asymmetric unit, and non-crystallographic symmetry (NCS) restraints were used during refinement. Waters were incorporated into the models by referring to the |Fo| 2 |Fc| omit maps. Out of the possible 994 residues, 957 residues were modeled with 97.8% within the most favored regions of the Ramachandran plot. Although 0.5% are reported as outliers, they are all within well-defined electron densities. The final model, which lacked residues 1, 2, and 482-497, has a R work of 0.190 and an R free of 0.222. The structure refinement statistics are summarized in Table 1.
The VldENGDP, VldENGDPNVDO, and VldENGDPNTRE structures were solved using the unliganded VldE structure as a search model within the AUTOMR [39] function of the PHENIX [33] software suite. The models were finished by manual building within COOT [35]. The models was refined the PHENIX.RE-FINE [29] function of the PHENIX [33] software suite. Ligands and waters were incorporated into the models by referring to the |Fo| 2 |Fc| omit maps. Two monomers were modeled within the VldENGDP the asymmetric unit, six monomers were modeled within the VldENGDPNTRE the asymmetric unit, and two monomers were modeled within the VldENGDPNVDO the asymmetric unit. NCS restraints were used during all refinements. In these structures, 0.4% of residues are reported as Ramachandran outliers; however, they are all within well-defined electron densities, and outliers at residues Asp158 and Asn386 were common within all the monomers of all the models of VldE. Diffraction and structure refinement statistics are summarized in Table 1

Overall Structure
As was described in a recent crystallographic study of ValL [40], a VldE homolog from S. hygroscopicus subsp. jinganggensis, VldE is a homodimer. Each monomer consists of twin Rossman b/a/b domains in the GT-B configuration [41] (Figure 3A) with the terminal helix (a16) extending back across the C-terminal Rossman-like domain to reach the N-terminal Rossman-like domain with a turn at Trp458 as it crosses the interface of the two domains. Despite a modest 19% sequence identity (29% homology) (Figure 2), the overall fold of VldE is remarkably similar to OtsA ( Figure 3B), a fellow member of the GT20 family. Unlike the previous study [40] however, the core of the N-terminal domain of our apo enzyme is found with the catalytic site in an ''open'' conformation instead of ''closed''. In this ''open'' conformation, the core b-sheet of the N-terminal Rossman-like domain consists of only ten strands instead of twelve. Two strands described as part of the core sheet in the previous study now are observed in a unique b-hairpin motif (b2 and b3) extending away from the catalytic site. Simultaneously, there is a dramatic shift of residues 33-47 away from the catalytic center. Unfortunately, coordinates from the structural study of ValL were not publicly available at the time of this study, so a complete comparison could not be performed. However, the closed conformation of VldE will be described in later sections of this article using our own experimental data. The GDP bound structure of VldE (VldENGDP) described here is also in an ''open'' conformation and the binding of GDP did not have an effect on the global conformation of the apo enzyme (RMSD C a = 0.350 Å ). Surprisingly, the structure of VldE in complex with the entirety of its products (VldENGDPNVDO), GDP and VDO was also in an open conformation and a comparison to the VldENGDP model showed that the binding of VDO also did not have an effect of the global conformation of VldE (RMSD C a = 0.262 Å ).
Lastly, we were also elucidated the structure of VldE in complex with both GDP and TRE (VldENGDPNTRE). Within the VldENGDPNTRE structure, six monomers were modeled within the asymmetric unit (ASU). The six monomers of the ASU assumed three alternative conformers. The monomers of the first conformation contained both GDP and trehalose, and this first conformer was found to also have an ''open'' catalytic site with similar global conformation to the VldENGDPNVDO complex (RMSD C a = 0.705 Å ). The last two conformers within the VldENGDPNTRE structure had GDP but not trehalose bound, and these two conformers contained a ''closed'' catalytic site as described in the previous studies [40]. These alternative conformations are significantly different when compared the ''open'' VldENGDP structure (RMSD C a = 2.304 Å , RMSD C a = 2.395 Å ) and are further discussed in later sections.

The Nucleotide Binding Site
Due to the conservation of key residues within their binding pockets, the binding of ligands within VldE are thought to be analogous to OtsA [19]. Here we provided experimental support for the analogous binding of GDP within VldE and UDP within OtsA. (Figure 4A) Within VldE, the ribose and phosphate moieties of GDP make interactions with the side-chains of the residues Arg290, Lys 295 and Glu391; as well as with the backbone atoms of residues Leu387 and Ser388 ( Figure 5A). These interactions are conserved within OtsA ( Figure 4B). The a-phosphate of GDP does make a single unique interaction with Ser388 within VldE. This interaction is not possible within OtsA as there is a valine at the same position. Nonetheless, the ribose and phosphate moieties of GDP within the VldE catalytic pocket are nearly superimposable (RMSD = 0.868) onto the ribose and phosphate moieties of UDP within the OtsA catalytic pocket. The conserved position of the bphosphate of GDP is critically important because the bond between this phosphate and the donated cyclitol is broken during the transfer to the acceptor. As will be described later in this study, the b-phosphate of GDP maintains an interaction with the formerly bound carbon of validoxylamine A 79-phosphate in the complete product complex (VldENGDPNVDO). This interaction is similarly found within OtsA and similarly provides strong evidence in support of an S N i mechanism [18].
The guanine moiety within the VldENGDP complex makes interactions with the hydrophilic residues Arg321, Asn323, and Thr366 ( Figure 5A). The preference of a GDP activated donor sugar by VldE (instead of UDP) is immediately apparent when comparing the catalytic site to that of OtsA. The preference of guanine by VldE is mediated by these three hydrophilic residues, while the preference for a uracil moiety is mediated by three hydrophobic residues at the same positions. Within OtsA, the uracil is surrounded by the hydrophobic residues Ile295, Pro297, and Leu344. This small hydrophobic binding pocket facilitates the  Note that this region of residues is capable of 10.9 Å shift towards the catalytic center and that strands of the b-hairpin motif (b2 and b3) move simultaneously with strands b6 and b5 to extend the core b-sheet from ten to twelve strands. (C) A view of helix a3 for each conformer is shown. Note that residues at the kink of the helix are capable of reorganizing. doi:10.1371/journal.pone.0044934.g007 binding of a uracil moiety. Interactions between VldE and GDP are summarized in Table 2.

Cyclitol Binding Sites
The first crystallographic structure of VldE in complex with the entirety of its products, VDO and GDP, was elucidated ( Figure 4C). The interactions between GDP and VldE are described within the previous section, but here the interactions between VDO and VldE are shown to be analogous to the interactions previously described by the OtsANVDO complex [20]. The hydroxyls of the donated cyclitol make numerous interactions with the backbone amides of VldE at residues Gly384, Gln385, Asn386, and Leu387. The donated cyclitol also makes interactions with the side-chains of residues His182 and Asp383 ( Figure 5B). The interactions by donated cyclitol are conserved within the catalytic site of OtsA ( Figure 4D). Also conserved is the interaction of the b-phosphate with the formerly bound carbon of the donated cyclitol. As in OtsA, the phosphate of the leaving nucleotide is involved in the stabilization of the transition state during the concerted, front-faced nucleophilic attack of a S N i reaction, and at a distance of 2.6 Å from the attacking nucleophile, the phosphate can act as a general base to deprotonate the amine [20].
The binding of the acceptor cyclitol within VldE is recognized by a shift in side-chain conformations within the catalytic site. Asp158 swings 62-83u towards the catalytic center in recognition of the hydroxyls of the acceptor cyclitol. Upon the binding of the phosphorylated acceptor group, the guanidine moiety of Arg326 enters into an interaction with the phosphate oxygens. This interaction allows Tyr159 to swing 110-112u to make an interaction with Arg326. The acceptor cyclitol also interacts with the side-chain of Asn325. These interactions are conserved within OtsA except at Asn325. At the same position within OtsA there is Ser299 which does not make an interaction with VDO ( Figure 4D) [4].
Despite the conservation of the interactions between the catalytic site and VDO, there is no observed conformational closing of the catalytic site upon the binding of the acceptor cyclitol as was witnessed in structural studies of OtsA [19,20,25]. However, a closed conformation was described in the recently published structural studies of ValL [40]. In later sections, such a conformational shift will be described within our structural studies of VldE, but this shift was not triggered by the binding of the acceptor cyclitol.  Table 3. DALI Server Results. We were also able to elicit the structure of VldE in complex with both GDP and trehalose (VldENGDPNTRE). ( Figure 6A) Within this model, GDP maintains the interactions previously described elsewhere within this study. The binding pose of trehalose is similar to what was described by the recent structural study of VldE [40]. The sugar moiety of trehalose within donor cyclitol binding site makes interactions with the backbone amides of residues Gly384, Gln385, Asn386, and Leu387 as well with the side-chains of Asp383 and His182. Unsurprisingly, this preserves the interactions between the enzyme and the structurally similar donor cyclitol as was described in the VldENGDPNVDO model. However, the sugar moiety of trehalose occupying the acceptor cyclitol binding site makes only a single hydrogen bond with Arg290 in our study. A superimposition of trehalose and VDO within the VldE catalytic site shows that trehalose does not assume a binding pose comparable to VDO ( Figure 6B). This is most likely due to the absence of a phosphorylated ligand. Within the VldENGDPNVDO model, Arg326 and Asn325 interact with the phosphorylated cyclitol and most likely help to ensure the proper position of validamine 7-phosphate for catalysis. Indicatively, Arg326 and Asn325 swing out of the catalytic site within our trehalose bound structure. Interactions between VldE and VDO and TRE are summarized in Table 2.

Observed Changes in Conformation
While solving the VldENGDPNTRE structure, six monomers of VldE were placed within the asymmetric unit. Represented within the six monomers were three differing conformations of the Nterminal Rossman-like domain (two monomers per conformation) ( Figure 7A). However, no interesting conclusions regarding possible allosteric interactions in this enzyme could be derived from these observations alone.
The first conformer had GDP and trehalose bound and is described in the previous section. This conformer is quite similar to the other models with ''open'' catalytic sites expect for residues 325-328 which swing out of the catalytic site upon the binding of trehalose. The second and third conformers have only GDP within the catalytic site and both of these conformers share a dramatic shift in residues 11-50 which result in the ''closing'' of the catalytic site ( Figure 7B). Resultantly, Gly33, Gly34, and Thr35 move 10.9 Å towards the catalytic center, and within 4 Å of the ligand atoms. Additionally, helix a1, which originally consisted of residues 36-41, coils further to include residues 42-47. This shift occurs concurrently with the rotation of the b-hairpin motif (b2 and b3) unique to VldE. Additionally, the last two of ten strands (b6 and b5) comprising the core b-sheet of the N-terminal Rossman-like b/a/b domain move towards the compacting catalytic site. The movements of the unique b-hairpin motif and the two strands of the core b-sheet narrow the distance between them and allow the necessary interactions to unite them into a single twelve-stranded core b-sheet. The ''closed'' conformation described here is similar to what was described in the structural study of ValL. [40].
A conformational change which resulted in the closing of the catalytic site was described within the structural studies of OtsA. Residues were observed to make a .10 Å movement towards the catalytic center upon binding of the accepting sugar in OtsA [19,25]. The conformational shift was thought to be due to the recognition of the phosphate moiety of the accepting glucose 6phosphate by Arg9. However, with the lack of an acceptor group within the catalytic site, the conserved arginine at position 12 of VldE cannot be responsible for the closing of the catalytic site. Therefore, the closing of the VldE catalytic site is more likely to work at least partially through a more dynamic, equilibrium based mechanism.
An additional conformation change was also observed within the third conformer found within the VldENGDPNTRE crystal ( Figure 7C). Within the a3, there is a kink at residues 96-102. Within the third conformer, the kinked residues shift to form a more organized secondary structure. Again, the observed conformational change is not concerted with the binding of ligand and further supports the notion that VldE moves through equilibrium based conformations. Conformational change at this position was observed in the previous structural studies, and similarly the change seems to be focused around Gln96 [40].

Mechanistic Considerations
The net retention of the substrate stereochemistry is thought to occur through either a double displacement (S N 2 X2) or internal return mechanism (S N i) [16]. Recent evidence substantiates claims that the conservation of the anomeric center configuration within OtsA is due to an internal return mechanism [18,20]. If the reaction of VldE were to alternatively go through a double displacement nucleophilic substitution mechanism, a residue must be available to make a covalent intermediate. Comparatively, the hydrolysis of glycosides is known to occur through a nucleophilic substitution involving a covalent intermediate [42,43]. The most likely candidate residue within VldE would be a nucleophilic His182. As can be seen in Figure 4D, this residue is also conserved within the catalytic site of OtsA (His154). At a 5.1 Å from the stereochemical center of the substrate, His182 is not observed to be immediately available for nucleophilic substitution in our structural studies, and as of yet there has been no significant, direct evidence to support a mechanism by which the non-glycosidic coupling reaction of VldE occurs through a covalent intermediate.
Our structural evidence demonstrates a strict conservation of nucleotide and (pseudo)sugar orientation within VldE and OtsA and suggests that the glycosyltransferase reaction of OtsA and the pseudo-glycosyltransferase reaction of VldE do occur through similar, concerted but asynchronous internal return mechanisms. Therefore, the olefinic moiety of GDP-valienol must play a critical role in facilitating the coupling reaction through a mechanism similar to the formation of an oxonium ion-like transition state upon the detachment of the nucleotide phosphate within OtsA as was previously suggested [4]. Through such a mechanism, the transition state is formed by the simultaneous binding of the allylic carbon on donated cyclitol, the bridging amine of the acceptor cyclitol-phosphate and the leaving oxygen of the donor diphosphate-nucleotide. As in OtsA [20], the leaving phosphate is in a position to deprotonate the incoming nucleophile of the acceptor cyclitol ( Figure 8A-B).

Significance
The biosynthetic pathways of natural products exhibiting antibiotic-like properties have recently become of great interest. As seen in the discovery of the antibacterial erythromycin, the antihelmintic avermectins, and the antitumor indolocarbozes, the genetic manipulation of biosynthetic gene clusters along with the use of alternative biosynthetic precursors, and recombinant proteins for chemoenzymatic synthesis can result in the generation of an entire array of biologically active compounds [50][51][52][53]. The gene cluster responsible for the biosynthesis of validamycin A has only been recently discovered, this study and others are underway to elucidate the individual processes within this biosynthetic pathway of this C 7 N-aminocyclitol [2]. The C 7 N-aminocyclitols (e.g., acarbose, validamycin, cetoniacytone, and salbostatin) belong to a class of natural products which include the aminoglycosides (e.g., streptomycin, hygromycin, butirosin, and neomycin), and the five membered aminocyclitols or cyclopentitols (e.g., pactamycin, trehazolin, and allosamidin). The aminoglycosides are a diverse group of natural products which were amongst the first clinical antibiotics. The C 7 Naminocyclitols are an even more diverse group of natural products with an entire range of biological activities [10]. A better understanding of the biosynthesis of C 7 N-aminocyclitols at the molecular level could allow for the generation of novel, natural product analogs which could include an array of structurally altered antibiotics. The line of inquiry provided within this article hopes to add to that understanding.