Molecular basis for branched steviol glucoside biosynthesis

Significance

The naturally occurring noncaloric sweetener stevia is a plant natural product consisting of a core terpene structure decorated with a specific pattern of glucose molecules, including a branched three-sugar unit. Stevia and other related molecules are being explored as noncaloric dietary sweeteners because they can help maintain the health of diabetic, phenylketonuric, and obese patients. Here, we describe the three-dimensional structure of the plant enzyme (UGT76G1) that forms the branched group of sugars that defines the stevia molecule and is critical for its high-intensity sweetness. Understanding how this enzyme forms this chemical group provides insight on how the stevia plant makes this sweetener and suggests how to alter the protein to generate new versions of the noncaloric sweetener.

Abstract

Steviol glucosides, such as stevioside and rebaudioside A, are natural products roughly 200-fold sweeter than sugar and are used as natural, noncaloric sweeteners. Biosynthesis of rebaudioside A, and other related stevia glucosides, involves formation of the steviol diterpenoid followed by a series of glycosylations catalyzed by uridine diphosphate (UDP)-dependent glucosyltransferases. UGT76G1 from Stevia rebaudiana catalyzes the formation of the branched-chain glucoside that defines the stevia molecule and is critical for its high-intensity sweetness. Here, we report the 3D structure of the UDP-glucosyltransferase UGT76G1, including a complex of the protein with UDP and rebaudioside A bound in the active site. The X-ray crystal structure and biochemical analysis of site-directed mutants identifies a catalytic histidine and how the acceptor site of UGT76G1 achieves regioselectivity for branched-glucoside synthesis. The active site accommodates a two-glucosyl side chain and provides a site for addition of a third sugar molecule to the C3′ position of the first C13 sugar group of stevioside. This structure provides insight on the glycosylation of other naturally occurring sweeteners, such as the mogrosides from monk fruit, and a possible template for engineering of steviol biosynthesis.

Sweeteners derived from plant natural products have significant potential as dietary supplements because they are stable and noncaloric; can maintain good dental health by reducing sugar intake; and have possible uses by diabetic, phenylketonuric, and obese patients (1, 2). For example, the sweetener stevia is isolated from the leaves of Stevia rebaudiana (sweetleaf), a perennial herb native to Paraguay and Brazil (2, 3). The leaves of this plant contain a variety of ent-kaurene diterpenoid glycosides composed of a steviol aglycone decorated with different numbers and types of sugars attached to the C13 and C19 positions (2, 3) (Fig. 1A). The predominant steviol glucosides are stevioside (5–10% of leaf dry weight) and rebaudioside A (2–4% of leaf dry weight), which taste up to 300-fold sweeter than sucrose. Use of these compounds as naturally sourced noncaloric sweeteners has expanded globally over the last decade (1, 4). For example, rebiana, a commercially available sweetener, mainly contains rebaudioside A, which reduces bitterness and aftertaste associated with other compounds isolated from the plant (4). Moreover, recent efforts to engineer steviol glucoside production in yeast aim to provide specific types of molecules as a way to avoid variation that results from the use of different S. rebaudiana cultivars and growing conditions (5).

Fig. 1.

Role of UGT in steviol glucoside biosynthesis and the overall structure of UGT76G1. (A) Role of UGT in the steviol biosynthesis pathway. The C13 and C19 positions of the steviol aglycone are indicated. (B) Three-dimensional structure of UGT76G1. The ribbon diagram for the UGT76G1•UDP complex shows the secondary structure with α-helices (blue) and β-strands (gold). UDP is shown as a space-filling model.

The biochemical pathway for steviol glucoside biosynthesis involves formation of the core steviol diterpenoid followed by a series of glucosylations catalyzed by a set of uridine diphosphate (UDP)-dependent glucosyltransferases (UGT) (2, 6–12) (Fig. 1A). Three UDP-glucosyltransferases in the rebaudioside A biosynthesis pathway have been identified and shown to localize to the cytosol (7, 8). Glycosylation of steviol by UGT85C2 begins at the C13 hydroxyl group to yield steviolmonoside (7). Transcript levels of UGT85C2 correlate with total accumulation of steviol glucosides in plant tissue, which suggests this step is the limiting reaction in the pathway (9). The UGT that modifies the C2′ of the C13 glucose to form steviolbioside remains to be identified (6, 7). Next, UGT74G1-catalyzed glucosylation of the carboxylate completes formation of stevioside (7). Further modification by UGT76G1 at the C3′ of the C13 sugar leads to synthesis of rebaudioside A, which contains the branched glucoside (7). Variants of UGT76G1 alter levels of rebaudioside A in the plant (10, 11). The broad acceptor molecule activity of UGT76G1 has been explored for biocatalyst uses in industrial synthesis of a range of glucosylated natural products (12). Although the general sequence of modifications in the steviol glucoside pathway has been determined, the structural basis for selectivity of the UGT enzymes, in particular how UGT76G1 catalyzes branched glucoside formation, remains unclear.

Results and Discussion

Overall Structure of UGT76G1.

To understand the structural basis of branched steviol glucoside synthesis, the X-ray crystal structure of S. rebaudiana UGT76G1 was determined by single-wavelength anomalous dispersion phasing using selenomethionine (SeMet)-substituted protein (Fig. 1B and Table 1). The SeMet-substituted model was then used to solve 3D structures of the UGT76G1•UDP (1.75 Å resolution) and UGT76G1•UDP•rebaudioside A (1.99 Å resolution) complexes by molecular replacement. The overall structure of UGT76G1 is monomeric and adopts the GT-B fold, which consists of an N- and C-terminal Rossmann-fold domains (13). Each structural domain of UGT76G1 contains a central β-sheet flanked by multiple α-helices (N-terminal: β1a–β1g and α1–α9; C-terminal: β2a–β2f and α10–α17) (Fig. 1B and SI Appendix, Fig. S1). The C-terminal domain contains the UDP-binding site with a cleft between the two domains providing the sugar acceptor-binding site (13). The amino acid sequence and 3D structure of UGT76G1 indicate that this protein is a member of the carbohydrate-active enzymes (CAZy) glycosyl transferase family 1 (14), which consists of related enzymes that glycosylate a variety of plant natural products. The 3D structure of UGT76G1 shares 2.2–2.9 Ų root mean square deviations (rmsds) for 420–440 C_α atoms with the other structurally characterized plant UGT. These include enzymes that modify terpenoids (Medicago truncatula/barrel clover UGT71G1 and Orzya sativa/rice Os79), flavonoids and isoflavonoids (Vitis vinifera/grape GT1, M. truncatula UGT85H2 and UGT78G1), chlorinated phenols (Arabidopsis thaliana/thale cress UGT72B1), anthocyanin floral pigments (Clitoria ternatea/bluebell vine UGT78K6), salicylic acid (A. thaliana UGT74F2), and indoxyl dyes (Polygonum tinctorium/Japanese indigo GT-B) (15–23); however, all of these UGT transfer sugars directly to the aglycone and do not form branched natural product glucosides.

Table 1.

Summary of crystallographic data collection and refinement statistics

Data collection	UGT76G1(SeMet) •UDP	UGT76G1•UDP	UGT76G1•UDP •rebaudioside A
Space group	P3121	P3121	P3121
Cell dimensions	a = b = 97.98 Å, c = 90.62 Å	a = b = 98.45 Å, c = 90.67 Å	a = b = 98.12 Å, c = 91.52 Å
Wavelength, Å	0.979	0.979	0.979
Resolution, Å (highest shell)	38.5–1.80 (1.83–1.80)	40.0–1.75 (1.78–1.75)	42.5–1.99 (2.02–1.99)
Reflections (total/unique)	320,766/47,033	304,795/48,196	317,717/35,139
Completeness (highest shell), %	99.6 (99.1)	93.5 (96.7)	99.5 (94.1)
<I/σ> (highest shell)	13.0 (2.3)	31.8 (2.6)	25.9 (1.8)
Rsym (highest shell), %	17.4 (74.7)	4.2 (34.8)	9.32 (61.9)
Figure of Merit	0.564	—	—
Refinement
Rcryst/Rfree, %	17.1/20.0	17.2/19.8	16.4/20.0
No. of protein atoms	3,609	3,567	3,559
No. of waters	369	267	223
No. of ligand atoms	42	42	140
rmsd, bond lengths, Å	0.007	0.007	0.007
rmsd, bond angles, °	0.912	0.922	1.03
Avg. B-factor (Å2): protein, water, ligand	28.0, 36.6, 20.2	47.2, 46.1, 31.3	40.6, 67.3 43.5
Ramachandran plot: favored, allowed, disallowed, %	97.6, 2.4, 0.0	96.0, 4.0, 0.0	97.8, 2.2, 0.0

Structure of the UDP Sugar Donor-Binding Site.

Clear electron density for UDP in the C-terminal domain of UGT76G1 defines the location of the sugar donor-binding site (Figs. 1B and 2A). The uridine ring of UDP is sandwiched between Trp338 and Gln341 with hydrogen bond interactions contributed by Val339, Pro340, and a water-mediated interaction with His260 (Fig. 2B). Glu364 forms a bidentate interaction with the hydroxyl groups of the ribose. Direct interactions with Ser285, His356, Asn360, and Ser361, along with water-mediated contacts to Ser283 and Thr284, position the diphosphate group toward the cleft between the N- and C-terminal domains.

Fig. 2.

UDP binding and sugar donor-binding site of UGT76G1. (A) Electron density for UDP shown as a 2F_o − F_c omit map (1.5σ). (B) Stereoview of the UDP-binding site in the UGT76G1•UDP complex. Ligand-binding interactions are shown as dotted lines. (C) Glycerol binding in the sugar donor-binding site of the UGT76G1•UDP complex. Ligand-binding interactions are shown as dotted lines. The view is shown for comparison with D. (D) Sugar donor-binding site of rice UGT Os79 (16). The orientation is the same as the glycerol site from UGT76G1. The X-ray structure of the nonreactive UDP-glucose analog U2F is shown for comparison with C.

A structure of UGT76G1 with UDP-glucose was not obtained; however, Thr146, Trp359, and Asp380 bind a glycerol molecule in proximity to the diphosphate group of UDP (Fig. 2C). The position of this ligand mimics how the glucose moiety of the sugar donor would interact with UGT76G1. Comparison of the site where glycerol binds in UGT76G1 to the structure of the Os79 UGT from rice in complex with a nonreactive UDP-glucose analog, uridine-5′-diphosphate-2-deoxy-2-fluoro-α-d-glucose (U2F) (16), highlights the structural and sequence conservation of the sugar donor-binding site and the position of the catalytic histidine in each active site (Fig. 2 C and D). The 3D position of glycerol in the UGT76G1 sugar donor-binding site fills the same space as the C3–C5 portion of the U2F glucose group from the Os79 structure. The residues in Os79 (i.e., Ser142, Gln143, Trp369, Asp385, and Gln386) that interact with the sugar donor are either invariant or highly conserved in UGT76G1 (Thr146, Ser157, Trp359, Asp380, and Gln381). Likewise, the histidine (His25 in UGT76G1; His27 in Os79) that serves as a general base to abstract a proton from the acceptor molecule in the SN₂ mechanism and the aspartate (Asp124 in UGT76G1; Asp120 in Os79) that stabilizes the histidine in the transfer reaction are invariant (24).

Although UGT76G1 shares ∼25% sequence identity with UGT85C2 and UGT74G1 from the steviol glucoside biosynthesis pathway of S. rebaudiana, key active site residues are retained across these three enzymes (SI Appendix, Fig. S1). These features include the histidine and asparate of the catalytic dyad and residues (i.e., His260, Ser283, Thr284, Ser285, Trp338, Qln341, Asn360, and Glu364 of UGT76G1) that interact with the UDP portion of the shared sugar donor substrate (SI Appendix, Fig. S1, red and blue, respectively). Similarly, residues critical for positioning the donor glucosyl group in UGT76G1 (Thr146, Ser157, Trp359, Asp380, and Gln381) are maintained in the other UGT of this biosynthetic pathway (SI Appendix, Fig. S1, purple). A large portion of the UDP-binding and sugar donor-binding sites are part of the canonical PSPG (putative secondary plant glycosyltransferase) sequence motif found in the plant UGT (25) (SI Appendix, Fig. S1, black box), which is consistent with conserved sequence and structure to bind the UDP molecule. This analysis indicates that the UGTs in the steviol pathway retain highly conserved catalytic dyads and UDP-glucose sugar donor sites, yet each enzyme displays differences in acceptor regiospecificity.

Rebaudioside A Binding and Generation of the Branched Glucoside.

As a branched sugar side chain-forming enzyme, UGT76G1 differs from the other two UGTs in the steviol pathway, which directly glucosylate the steviol aglycone via an oxygen at either the C13 (UGT85C2) or C19 (UGT74G1) positions (Fig. 3A). For synthesis of the branched glucoside rebaudioside A, the acceptor site of UGT76G1 needs to accommodate a two-glucosyl side chain to allow for the addition of a third sugar molecule to the C3′ position of the first C13 sugar group of stevioside. To achieve this regioselectivity, UGT76G1 requires a binding site near the UDP-glucose donor site that also orients the second sugar away from the catalytic histidine.

Fig. 3.

Structural basis for branched steviol glucoside synthesis by UGT76G1. (A) Schematic of sugar donor and acceptor in steviol UGT. The UDP-glucose donor (UDP-Glc; red) and acceptor molecules (blue) are shown. Note that the oxygen in the acceptor of UGT85C2 and UGT74G1 differ. The position of the catalytic histidine is indicated by the triangle. (B) Electron density for rebaudioside A shown as a 2F_o − F_c omit map (1.5σ). (C) Stereoview of the molecular surface of the UGT76G1 acceptor site. Crystallographically determined positions of UDP and rebaudioside A are shown. The surface corresponding to His25 is colored blue. The three glycosyl units (glc1, glc2, glc3) attached at the C19 position of the steviol aglycone are labeled. The fourth sugar attached at the C13 carboxylate was not modeled because of disorder. (D) View of the rebaudioside A binding site of UGT76G1. As in C, the glc1, glc2, glc3, and steviol portions of the ligand are labeled. Hydrogen-bonding interactions are shown as dotted lines. The two α-helices (α3 and α8) that define the steviol aglycone binding site are also labeled.

The structure of UGT76G1 in complex with UDP and rebaudioside A reveals the basis for branched steviol glucoside synthesis. Electron density for both ligands was visible in the structure with the omit map for rebaudioside A showing clear density for the three glucosyl groups extending from the steviol C13, weaker density for the steviol core, and no density for the C19 glucose (Fig. 3B). Because the solvent-exposed portion of rebaudioside A was disordered, a model of the ligand lacking the C19 sugar was used for refinement. The nucleotide is bound as observed in the UDP complex and rebaudioside A binds in a pocket of the N-terminal domain (Fig. 3C).

The branched sugar portion of rebaudioside A is oriented into the interior and toward the UDP pyrophosphate. From the bottom of the binding site, the terminal glucose (glc3) fills the sugar donor portion of the pocket; the branching glucose (glc2) is positioned into a pocket away from the catalytic histidine; and the C13-linked glucose (glc1) and the steviol portion of the molecule extend out to the solvent exposed opening of the site (Fig. 3 C and D). In the acceptor site, Thr146, Trp359, Asp380, and Gln381 interact with glc3 and position the α1,3-linkage between this glucosyl unit and a glc1 in proximity to His25 (Fig. 3D). Interactions with Met145, Ser147, Asn151, His155, and Leu379 orient the α1,2-linked glc2 into a pocket away from the catalytic His25 and Asp124. The C13-linked glucose group (glc1) stacks with Phe22 and Ile90 and forms a water-mediated interaction with Gly24. The steviol group is sandwiched between Leu126 and residues from α3 (Met88 and Ile90) and α8 (Leu200, Ile203, Leu204, and Met207). The C19 carbonyl hydrogen bonds with the Pro84, which positions the crystallographically disordered sugar at this position toward solvent. The UGT76G1 structure reveals an active site that is generally divided between hydrophilic branched glucoside-binding residues and largely apolar steviol interaction residues; however, sequence comparisons also suggest that the residues forming the glc2 pocket in UGT76G1 are varied in UGT85C2 and UGT74G1 (SI Appendix, Fig. S1).

As shown schematically in Fig. 3A, the position of the diterpene core of steviol in the active site needs to vary between UGT76G1 and the other two enzymes in the stevia biosynthesis pathway. UGT85C2 and UGT74G1 directly modify the terpene at the C13 and C19 positions, respectively. In contrast, UGT76G1 needs to bind stevioside with the terpene moiety further away from the catalytic site. Sequence comparison indicates that residues in the glc2- and stevia-binding regions of UGT76G1, along with the lengths of the α3 and α8 helices defining the acceptor site, vary between the three UGT of the stevia pathway and that these extensive changes likely contribute to different substrate preferences (SI Appendix, Fig. S1, yellow). In comparison with the residues of the glc2 pocket in UGT76G1, sequence comparison suggests that larger side chains are found in UGT85C2 and UGT74G1. For example, Met145, Ser147, and Leu379 are replaced with tryptophan, isoleucine, and tryptophan in UGT85C2 or phenylalanine, glutamine, and serine in UGT74G1. Each retains a histidine at position 155 and has smaller side chains in place of Asn151 (a glycine in UGT85C2 and a valine in UGT74G1). Homology modeling of the other two UGT in the steviol biosynthesis pathway suggests that the various amino acid substitutions narrow the glc2 pocket, which likely occludes binding of longer side-chain steviol glucosides but allows for steviol and steviolbioside glucosylation in UGT85C2 and UGT74G1, respectively (SI Appendix, Fig. S2 A–C).

Structural and sequence comparison of UGT76G1, a branch-forming glycosyltransferase, to the plant UGTs that directly glycosylate a given substrate (i.e., terpenoids, flavonoids, phenols, anthocyanins, and indoxyls; refs. 15–23) highlights key differences in residues forming the glc2 site (SI Appendix, Fig. S2 D–G). For each of these enzymes, residues defining the glc3 site, in which the glucosyl group of a UDP-sugar donor binds, are highly conserved, as expected for UGTs (SI Appendix, Fig. S2G). In contrast, each of the other plant UGT examined have multiple bulky side-chain substitutions in residues corresponding to the glc2 site of UGT76G1. In particular, Leu126, Met145, and His155 of UGT76G1 are typically replaced by phenylalanine, phenylalanine/tryptophan, and phenylalanine/tyrosine, respectively (SI Appendix, Fig. S2G). Examination of the anthocyanidin glucosyltransferase UGT78K6, flavonoid glucosyltransferase GT1, and indoxyl glucosyltransferase GT-B X-ray crystal structures (SI Appendix, Fig. S2 D–F) highlight how various changes reduce the available space in the region corresponding to the glc2 site of UGT76G1. The introduction of key changes in the glc2 site of UGT76G1 are critical for the evolution of the branch chain-forming glucosyltransferase activity of this enzyme.

Biochemical Analysis of Site-Directed Mutants.

To probe the contribution of active site residues to UGT76G1 function, 21 site-directed mutants were generated and examined for biochemical activity (Fig. 4 and Table 2). Substitution of the catalytic histidine with an alanine (H25A) eliminated enzymatic activity. The analogous histidine residue in other UDP-glucosyltransferases facilitates a direct displacement SN₂-like mechanism. This residue acts as a general base by abstracting a proton from the acceptor substrate to yield an oxyanion nucleophile that reacts with the UDP-sugar acceptor (13).

Fig. 4.

Summary of wild-type and mutant UGT76G1 enzyme activities. A comparison of wild-type and mutant UGT76G1 catalytic efficiencies (k_cat/K_m) with stevioside (black) and UDP-glucose (white) is shown. Steady-state kinetic parameters are summarized in Table 2.

Table 2.

Summary of wild-type and mutant UGT76G1 steady-state kinetic parameters

Protein	Varied substrate	kcat, min−1	Km, μM	kcat/Km, M−1·s−1
WT	Stevioside	33.8 ± 0.7	360 ± 23	1,560
	UDP-Glucose	32.5 ± 1.5	943 ± 96	574
L126I	Stevioside	0.09 ± 0.01	676 ± 320	2
	UDP-Glucose	0.06 ± 0.04	12,500 ± 10,700	0.1
M145F	Stevioside	14.9 ± 1.8	3,030 ± 580	82
	UDP-Glucose	13.5 ± 1.2	5,940 ± 710	38
M145W	Stevioside	1.0 ± 0.2	7,690 ± 2,780	2
	UDP-Glucose	0.8 ± 0.1	17,900 ± 4,800	1
T146A	Stevioside	8.9 ± 0.7	7,340 ± 1,420	20
	UDP-Glucose	5.3 ± 0.2	8,770 ± 674	10
S147A	Stevioside	1.4 ± 0.1	2,570 ± 558	9
	UDP-Glucose	0.8 ± 0.1	3,750 ± 355	4
S147T	Stevioside	0.6 ± 0.1	927 ± 163	11
	UDP-Glucose	0.7 ± 0.1	477 ± 49	24
S147N	Stevioside	0.6 ± 0.1	880 ± 162	11
	UDP-Glucose	0.7 ± 0.1	2,160 ± 320	5
N151A	Stevioside	9.2 ± 0.4	1,600 ± 125	96
	UDP-Glucose	8.6 ± 0.5	1,870 ± 190	76
N151Q	Stevioside	27.8 ± 0.4	1,210 ± 34	390
	UDP-Glucose	105 ± 18	21,200 ± 4,090	83
H155A	Stevioside	11.4 ± 0.8	424 ± 69	448
	UDP-Glucose	12.8 ± 0.1	577 ± 16	370
H155R	Stevioside	3.4 ± 0.1	1,040 ± 72	54
	UDP-Glucose	4.3 ± 0.3	3,290 ± 290	22
H155W	Stevioside	7.8 ± 0.4	2,090 ± 180	62
	UDP-Glucose	6.1 ± 0.1	3,220 ± 96	32
L200I	Stevioside	43.0 ± 19	1,750 ± 1,120	410
	UDP-Glucose	30.7 ± 2.0	2,270 ± 240	225
L204I	Stevioside	27.1 ± 2.7	762 ± 150	593
	UDP-Glucose	24.7 ± 1.2	1,090 ± 113	378
M207F	Stevioside	22.9 ± 2.4	885 ± 172	431
	UDP-Glucose	19.2 ± 0.5	1,050 ± 56	305
M207W	Stevioside	34.7 ± 18.9	4,710 ± 3090	123
	UDP-Glucose	68.6 ± 22.4	15,600 ± 5,700	73
L379I	Stevioside	42.6 ± 3.5	1,120 ± 160	634
	UDP-Glucose	33.2 ± 0.7	1,420 ± 54	390

Assays were performed as described in Methods. Average values ± SD (n = 3) are shown.

Mutagenesis of residues in the rebaudioside A-binding site and biochemical analysis of the mutants reveals the importance of key residues for enzymatic activity. Removal of the side chains of Asp380 (D380A) and Gln381 (Q381A) in the glc3/sugar donor-binding site resulted in inactive enzyme. Based on their position in the conserved PSPG sequence motif (SI Appendix, Fig. S1), loss of either side chain removes key interactions with the sugar being transferred from donor to acceptor and likely results in compromised substrate binding, which prevents efficient catalysis. Likewise, alteration of Thr146, which is situated between the glc3/sugar donor-binding site and the glc2 site, either lost activity (T146N) or decreased catalytic efficiency by 80-fold (T146A). Mutations of residues in the glc2 site resulted in a range of effects. Substitutions of Ser147 yielded mutants with ∼170-fold reductions in catalytic efficiency (S147A, S147T, and S147N). Modest 3- to 15-fold changes in k_cat/K_m were observed with the N151A, N151Q, H155A, and L379I mutants. Mutations that introduce larger side chains in to the glc2 site at His155 (H155R and H155W) led to ∼25-fold less-efficient variants. Change of Met145 to either a phenylalanine (M145F) or tryptophan (M145W) resulted in 20- and 750-fold reductions in k_cat/K_m. In the apolar steviol-binding cleft, a subtle change of Leu126 to isoleucine (L126I) leads to a 750-fold decrease in catalytic efficiency. Other point mutants, such as L200I, L204I, and M207F, displayed modest threefold changes with the M207W mutant having a 13-fold effect. Overall, biochemical analysis of the UGT76G1 mutants confirms the conserved function of the catalytic histidine and identifies critical residues in the glc2- and steviol-binding sites.

Conclusion.

This first structure of a branched steviol glucoside-producing enzyme (i.e., UGT76G1) provides insight on how the enzyme accommodates large sugar side chains and differs from the other glycosyltransferases in the pathway. Importantly, the branched-chain glycosylation pattern built from the C19 position of rebaudioside A leads to rebaudiosides D and M, which are more commercially interesting steviosides because they deliver high-intensity sweetness with less off-taste than rebaudioside A (26). Similar glycosylation patterns are found in the biosynthesis of other molecules, such as the mogrosides from monk fruit (Siraitia grosvenorii), which are being explored as stevia alternates (27). Knowledge of the active site architecture also provides a template for enzyme engineering that may lead to the development of variants with altered regiospecificity and/or substrate glycosylation patterns that can be combined with varied donor substrates either in vitro or in vivo (28–31). Moreover, the amino sequence of the glc2 site of UGT76G1, which allows for formation of a branched-sugar modified product (i.e., rebaudioside A), may provide a useful “signature motif” to bioinformatically distinguish branched chain-forming UGT from those that directly glucosylate various substrates when assessing potential metabolic function. Such structure-guided efforts offer the potential for altered pathways of steviol production to generate tailored variants of this noncaloric sweetener.

Methods

Chemicals, Codon Optimized Gene Synthesis, and Site-Directed Mutagenesis.

All reagents used were purchased from Sigma-Aldrich unless noted otherwise. An Escherichia coli codon-optimized version of the gene encoding UGT76G1 was generated for protein expression. The original sequence (SwissProt Q6VAB4; ref. 7) was optimized and synthesized by GenScript. The resulting gene was inserted into pET-28a to yield the pET-28a-UGT76G1 for expression of an N-terminally His₆-tagged fusion protein. Site-directed mutants of UGT76G1 (H25A, L126I, M145F, M145W, T146A, T146N, S147A, S147T, S147N, N151A, N151Q, H155A, H155R, H155W, L200I, L204I, M207F, M207W, L379I, D380A, and Q381A) were generated using QuikChange PCR mutagenesis with the pET-28a-UGT76G1 vector as template and appropriate oligonucleotides.

Protein Expression and Purification.

The general protein expression and purification protocol for UGT76G1 uses a combination of affinity and size-exclusion chromatographies based on a previously published protocol (32). For production of SeMet-substituted protein, the pET-28a-UGT76G1 construct was transformed into E. coli BL21 (DE3) cells, which were grown to A₆₀₀ ∼ 0.8 at 37 °C in M9 minimal media supplemented with SeMet containing 50 μg·mL⁻¹ kanamycin (33). Protein expression was induced by addition of isopropyl 1-thio-β-d-galactopyranoside (IPTG; 1 mM final) with cells growth continued overnight (16 °C). Cells were pelleted by centrifugation (10,000 × g) and resuspended in lysis buffer [50 mM Tris, pH 8.0, 500 mM NaCl, 20 mM imidazole, 1 mM β-mercaptoethanol, 10% (vol/vol) glycerol, and 1% (vol/vol) Tween-20]. After sonication of the resuspended cells, debris was removed by centrifugation (30,000 × g). The lysate was passed over a Ni²⁺-nitriloacetic acid (Qiagen) column equilibrated with wash buffer (lysis buffer minus Tween-20). Bound his-tagged protein was eluted with elution buffer (wash buffer with 250 mM imidazole). The eluant was further purified by size-exclusion chromatography using a Superdex-200 26/60 HiLoad FPLC column equilibrated with 50 mM Tris, pH 8.0, 25 mM NaCl, 1 mM Tris(2-carboxyethyl)phosphine. Peak fractions were collected and concentrated to ∼10 mg·mL⁻¹ using centrifugal concentrators (Amicon). Bradford assay with BSA as the standard was used to determine protein concentration. For storage at −80 °C, purified protein was flash-frozen in liquid nitrogen. Expression of wild-type and mutant UGT76G1 proteins used similar protocols, except that Terrific broth replaced the minimal media.

Protein Crystallization and Structure Determination.

Purified UGT76G1 was concentrated to 10 mg·mL⁻¹ and crystallized using the hanging-drop vapor-diffusion method with a 2-µL drop (1:1 concentrated protein and crystallization solution). Diffraction quality crystals of both SeMet-substituted and native protein were obtained at 4 °C with 15% (wt/vol) PEG 4000, 20% 2-propanol (vol/vol), 100 mM sodium citrate tribasic dihydrate buffer (pH 5.6), and either 5 mM UDP or 5 mM UDP and 5 mM rebaudioside A. To generate the UGT76G1•UDP•rebaudioside A complex, 5 mM UDP and 5 mM rebaudioside A [in 10% (vol/vol) DMSO] were added during protein concentration. Individual crystals were flash-frozen in liquid nitrogen with the mother liquor containing 25% glycerol as a cryoprotectant. Diffraction data (100 K) was collected at the Argonne National Laboratory Advanced Photon Source 19-ID beamline. HKL3000 (34) was used to index, integrate, and scale diffraction data. The structure of SeMet-substituted UGT76G1 was determined by single-wavelength anomalous diffraction (SAD) phasing. SHELX (35) was used to determine SeMet positions and to estimate initial phases from the peak wavelength dataset. Refinement of SeMet positions and parameters was performed with MLPHARE (36). Solvent flattening using density modification implemented with ARP/wARP (37) was employed to build an initial model. Subsequent iterative rounds of manual model building and refinement, which included translation-libration-screen parameter refinement, used COOT (38) and PHENIX (39), respectively. The structures of UGT76G1 in complex with either UDP or UDP and rebaudioside A were solved by molecular replacement in PHASER (40) using the SeMet structure as a search model with refinement and building performed as above. The final model of the SeMet-substituted UGT76G1 includes residues Arg12-Pro169 and Arg174-Leu458, one UDP molecule, one glycerol molecule, and 369 waters. The UGT76G1•UDP complex includes the same residues and ligands, but with 267 waters. The UGT76G1•UDP•rebaudioside A complex includes the same residues, one UDP molecule, one rebaudioside A molecule (modeled without the C19 sugar), and 223 waters. Data collection and refinement statistics are summarized in Table 1. Coordinates and structure factors for the UGT76G1(SeMet)•UDP complex (PDB ID code: 6O86), the UGT76G1•UDP complex (PDB ID code: 6O87), and the UGT76G1•UDP•rebaudioside A complex (PDB ID code: 6O88) were deposited in the Protein Data Bank.

Enzyme Assays.

UDP-glycosyltransferase activity was monitored spectrophotometrically (A₃₄₀) using a coupled assay system (41). Standard reaction conditions were 50 mM Hepes, pH 7.5, 200 μM NADH, 500 μM phosphoenol pyruvate, 10 mM MgCl₂, two units of pyruvate kinase, and six units of lactate dehydrogenase in a 0.1-mL reaction at 25 °C. Reactions were initiated by addition of protein with changes in absorbance measured on a Tecan 96-well plate reader. Steady-state kinetic parameters were determined by initial velocity experiments with either varied stevioside concentrations and 5 mM UDP-glucose or varied UDP-glucose and 2 mM stevioside. Data were fit to the Michaelis–Menton equation, v = k_cat[S]/(K_m + [S]), using Kaleidagraph.