Isolation and Characterization of a Novel Rebaudioside M Isomer from a Bioconversion Reaction of Rebaudioside A and NMR Comparison Studies of Rebaudioside M Isolated from Stevia rebaudiana Bertoni and Stevia rebaudiana Morita

A minor product, rebaudioside M2 (2), from the bioconversion reaction of rebaudioside A (4) to rebaudioside D (3), was isolated and the complete structure of the novel steviol glycoside was determined. Rebaudioside M2 (2) is considered an isomer of rebaudioside M (1) and contains a relatively rare 1→6 sugar linkage. It was isolated and characterized with NMR (1H, 13C, COSY, HSQC-DEPT, HMBC, 1D-TOCSY, and NOESY) and mass spectral data. Additionally, we emphasize the importance of 1D and 2D NMR techniques when identifying complex steviol glycosides. Numerous NMR spectroscopy studies of rebaudioside M (1), rebaudioside D (3), and mixture of 1 and 3 led to the discovery that SG17 which was previously reported in literature, is a mixture of rebaudioside D (3), rebaudioside M (1), and possibly other related steviol glycosides.


Introduction
Sweetness is universally regarded as pleasant and preferred taste for beverages, food, pharmaceuticals, and oral hygiene/cosmetic products.To provide sweet taste to consumer products the most commonly used natural caloric sugars are sucrose, fructose, and glucose.Since these natural sugars provide calories, alternative sources must be utilized when the consumer desires a sweet taste with low to no calories.Artificial and natural sweeteners have been developed to fulfill both criteria [1,2].It is not a simple task to create a non-caloric or low calorie sweetener because they exhibit a temporal profile, maximal response, flavor profile, mouth feel, and/or adaptation behavior that differ from sugar [3,4].Steviol glycosides isolated from Stevia rebaudiana Bertoni have been explored to produce an ideal sweetener (sweet, low to no calorie, and natural) [5][6][7][8][9][10][11][12].

Rebaudioside M2 (2)
To our knowledge this is the first report of isolation and complete characterization using NMR ( 1 H, 13 C, COSY, HSQC-DEPT, HMBC, 1D-TOCSY and NOESY) (Supplementary, Figure S1) and high resolution mass spectral data of rebaudioside M2 (2).Compound 2 was isolated as a white powder and accurate mass measurement using High Resolution Mass Spectrometry (HRMS) provided the exact mass m/z of 1289.5299,[M-H] í , in it's negative ESI-TOF mass spectrum corresponding to a molecular formula of C 56 H 90 O 33 .
Correlations observed in the NOESY spectrum were used to assign the relative stereochemistry of the central diterpene core.In the NOESY spectrum, NOE correlations were observed between H-14 and H-20 indicating that H-14 and H-20 are on the same face.Additionally, NOE correlations were observed between H-5 and H-9 but NOE correlations were not clearly observed between H-9 and H-14 indicating that H-5 and H-9 were on the opposite face of the rings compared to H-20 and H-14 as presented in Figure 3. Due to data overlap NOE correlations between H-5 and H-18 could not be confirmed, however the carbon chemical shifts support the relative stereochemistry as presented in Figure 3 [9,10,21].These data thus indicated that the relative stereochemistry was retained during the glycosylation step., confirmed the presence of six sugar units in the structure.The complete assembly of the glycoside structure was done on the basis of correlations observed in the 2D and 1D-TOCSY NMR data.Thus, long range 1 H- 13 C correlations observed in the HMBC experiment from the anomeric proton at į H 5.65 to a carbonyl carbon at į C 181.5 (C-19) allowed its assignment as the anomeric proton (H-1') of sugar I. Similarly, HMBC correlation from the anomeric proton observed at į H 4.85 to a quaternary carbon at į C 90.9 (C-13) allowed it to be assigned as the anomeric proton (H-2'') of sugar II (Figure 3).
Further analysis of the 1D and 2D NMR data allowed the assignment of the remaining four sugars in 2. The relatively downfield chemical shift of C-2' (į C 80.5) and C-6' (į C 70.9) in sugar I suggested a 2,6-branched-D-glucotriosyl substituent at C-19.Long range 1 H- 13 C correlations observed in the HMBC experiment from the anomeric proton observed at į H 4.83/4.84(H-1''''') to the carbon at į C 80.5 (C-2') and from H-2' at į H 3.96 to an anomeric carbon at į C 105.3/105.0(C-1''''') confirmed the substitution at C-2' in sugar I. Additionally, HMBC correlations observed from the anomeric proton at į H 4.50 (H-1'''''') to the carbon at į C 70.9 (C-6') and from the methylene protons of sugar I at į H 4.00 and 4.15 to the anomeric carbon (į C 105.7) of C-1'''''' confirmed the presence of a 1ĺ6 sugar linkage between sugar VI and sugar I.
The remaining two glucose moieties were assigned in a similar manner.The relatively downfield chemical shift of C-2'' (į C 81.7) and C-3'' (į C 88.0) in sugar II suggested a 2,3-branched-D-glucotriosyl substituent at C-13.Long range 1 H- 13 C correlations observed in the HMBC experiment from the anomeric proton observed at į H 4.92 (H-1''') to the carbon at į C 81.7 (C-2'') and from H-2'' at į H 3.75 to an anomeric carbon at į C 104.9 (C-1''') confirmed the sugar substitution at C-2' in sugar II.Similarly, the sugar substituent at C-3'' in sugar II was also corroborated by HMBC correlations observed from the anomeric proton at į H 4.84/4.83(H-1'''') to the carbon at į C 88.0 (C-3'') and from H-3'' at į H 3.98 to the anomeric carbon (į C 105.3/105.0) of sugar IV confirmed the presence of a 1ĺ3 sugar linkage between sugar IV and sugar II.
The 1 H and 13 C chemical shifts for the glycoside at C-13 and C-19 are found in Table 1 and a summary of the key HMBC, COSY, and 1D-TOCSY correlations used to assign the glycoside are provided in Figure 3.
Thus the structure of rebaudioside M2 (2), containing a relatively rare 1ĺ6 sugar linkage, was established as ( 13

NMR Study of Rebaudioside M, Rebaudioside D and SG17
During isolation (HPLC-MS) of compounds 1 and 2 it was revealed that more than one peak provided the molecular weight of 1290, and thus NMR studies were critical for complete structure determination.To determine the complete structure of rebaudioside M2 (2) we compared the NMR spectral data of rebaudioside M (1) (reported in Prakash et al. 2013) [21] and the NMR spectral data for SG17 (reported in Ohta et al. 2010) [24].Unfortunately rebaudioside M2 (2) was not soluble in any of the solvents screened except for D 2 O and thus its data could not be directly compared to previously reported data which were acquired in different solvent systems.This led to further examination of the two reported NMR spectral data for rebaudioside M, Ohta et al. [21] and Prakash et al. [24], and noticeable differences in spectral data, acquired in the same solvent system (Pyridine-d 5 + TMS) were observed.
Thus, a series of NMR experiments including 1 H, 13 C, 1 H-1 H COSY, 1 H-13 C HSQC-DEPT, and 1 H- 13 C HMBC were performed in pyridine-d 5 + TMS (Supplementary, Figure S2) to allow assignment of >95% rebaudioside M (1) in this solvent system.The 1 H-and 13 C-NMR data of >95% rebaudioside M (1) in pyridine-d 5 + TMS, were compared to the data of SG17 (isolated from S. rebaudiana Morita) in pyridine-d 5 + TMS, which was previously reported by Ohta et al. [24].As presented in Tables 2 and 3, more than half of the reported 1 H and 13 C chemical shifts of SG17 are not consistent with the 1 H and 13 C chemical shifts of compound 1 indicating that the data presented for SG17 by Ohta et al. [24] are most likely for a mixture of steviol glycosides.Therefore, NMR analysis of related samples such as >95% rebaudioside D (3) (Exp.2), 80% rebaudioside M (1) (Exp.5), and mixture of >95% rebaudioside M and >95% rebaudioside D (Exp. 3 and Exp.4) were carried out (Supplementary, Figures S3-S6) to compare their data to the data of SG17.The 13 C-and 1 H-NMR assignments for compounds 1 and 3 in experiments 3 and 4 (Tables 2 and 3) were confirmed on the basis of HSQC-DEPT and HMBC data.All compounds used in this NMR study were isolated from S. rebaudiana Bertoni.
The data presented by an asterisk for SG17 in Table 2 (with the exception of assignments for C9, C11 and C12) differ by about 1 ppm from the data of rebaudioside M (1) clearly indicate that these data do not belong to rebaudioside M. Instead, the NMR data of SG17 presented by an asterisk are consistent with the data of rebaudioside D (3).Some of the data of SG17, however, match with the data of compound 1 but some do not match with either the data of rebaudioside M (1) or rebaudioside D (3).Similarly, in the 1 H-NMR spectrum of SG17 (Table 3), the data of sugar III H-1' and sugar IV H-1' are most likely swapped, otherwise are consistent with the data of rebaudioside D (3).
Furthermore, as reported in Ohta et al. [24], the structure characterization of SG17 was performed in part by partial hydrolysis of the glycoside rather than using the modern 1D and 2D NMR techniques which are very commonly and widely used methods for structure elucidation of unknown compounds such as rebaudioside M. As reported in Morita et al. [25], rebaudioside M and rebaudioside D were separated by HPLC as a combined peak, and the rebaudioside M structure was deduced from mass spectrometry data, rather than using purified rebaudioside M for characterization by the modern 1D and 2D NMR techniques.Further, complete 1 H-and 13 C-NMR spectral assignments of the glycoside SG17 is not provided in the paper or patent application [24,25].For a compound as complex as rebaudioside M (1), 2D NMR data is needed to establish the specific sugar linkages and complete assignment of the structure, however in the publications of Ohta et al. [24], and Morita et al. [25] there was no indication of any 2D NMR or advance 1D NMR experiments utilized.In the absence of such NMR data, complex structures such as rebaudioside M can be misidentified.
In summary, based on the numerous NMR studies it was inferred that SG17 could be a mixture of rebaudioside D (3), rebaudioside M (1), and possibly related steviol glycosides, with rebaudioside D (3) as the major compound.This work not only uncovers a novel steviol glycoside (2), containing a relatively rare 1ĺ6 sugar linkage, it also illustrates how critical 1D and 2D NMR techniques are when identifying complex steviol glycosides.

Mass Spectrometry
The ESI-TOF mass spectra and MS/MS data were generated by a Waters QTof Premier mass spectrometer (Waters Corp., Manchester, UK) equipped with an electrospray ionization source.Samples were analyzed by negative ESI.Samples were diluted with H 2 O:MeCN (1:1) by 50-fold and introduced via infusion using the onboard syringe pump.

Nuclear Magnetic Resonance
The sample of Rebaudioside M2 (2) (~1.0 mg in 150 μL of D 2 O) was prepared and NMR data were acquired on Bruker Avance 500 MHz instrument (Bruker BioSpin Corp., Billerica, MA, USA) with a 2.5 mm inverse detection probe and 5 mm broad band probe.The 1 H-NMR and 13 C-NMR spectra were referenced to the residual solvent signal HDO (į H 4.79 ppm) and TSP (į C 0.00 ppm), respectively.

Bioconversion Reaction
Rebaudioside M2 (2) was isolated from bioconversion reaction of rebaudioside A (4) to rebaudioside D (3) by a proprietary glucosyltransferase from PureCircle Ltd.In vivo production of glycosylation enzymes were expressed in yeast.
Rebaudioside A to rebaudioside D conversion with glucosyltransferase UGTSL2 experiment condition are as follows: 430 ȝL of a reaction mixture containing 0.5 mM rebaudioside A, 3 mM MgCl 2 , 50 mM sodium phosphate buffer at pH 7.2 and 2.5 mM of UDP-Glucose was added to a 1.5 mL sterile microtube.52 ȝL of the enzyme expressed medium was added and the resulting mixture was allowed to react at 30 °C for 24 h.125 ȝL samples were taken after 2 h, 16 h, and 24 h and added to a 115 ȝL of 60% methanol and 10 ȝL of 2 N sulfuric acid.The quenched sample was centrifuged at 18,000× g for 2 min at room temperature.200 ȝL was transferred to a HPLC vial and analyzed by the method described below.

General Experimental Procedures for Rebaudioside D (3)
Rebaudioside D (3), bioconversion desired product, was isolated and characterized by NMR and MS which allowed a full assignment confirmation.

Mass Spectrometry
The ESI-TOF mass spectra and MS/MS data were generated with a Waters Q-Tof Premier mass spectrometer equipped with an electrospray ionization source.Samples were analyzed by negative ESI.Samples were diluted with H 2 O:MeCN (1:1) by 50-fold and introduced via infusion using the onboard syringe pump.

Nuclear Magnetic Resonance
The sample of rebaudioside D (3) was prepared by dissolving 10.7 mg in 200 μL of pyridine-d 5 + TMS and NMR data were acquired on Bruker Avance 500 MHz instruments with a 2.5 mm inverse detection and 5 mm broad band probes.The 1 H-NMR and 13 C-NMR spectra were referenced to TMS resonance (į H 0.00 ppm and į C 0.00 ppm).

Conclusions
To the best of our knowledge this is the first report of full isolation and spectral characterization of possesses a 1ĺ6 sugar linkage between sugar VI and sugar I, making its structural properties unique.Continued discovery in the area of novel steviol glycosides provides great opportunity to find novel sweeteners or sweetener enhancers that can improve sweet taste.In addition to the discovery of rebaudioside M2 (2), we have reiterated the importance of multiple 1D and 2D NMR techniques when identifying complex steviol glycosides.Thus extensive NMR studies of rebaudioside M (1), rebaudioside D (3), and mixtures of 1 and 3 led to the discovery that SG17 is not a single compound but is a mixture of rebaudioside D (3), rebaudioside M (1), and possibly other related steviol glycosides.
a Assignments were made on the basis of COSY, HSQC-DEPT, HMBC, and 1D-TOCSY correlations; chemical shift (į) values are in ppm; and coupling constants are in Hz; * 1 H and13C values can be exchangeable.
NR: Not reported; NA: Not applicable for Reb D; * = Similar spectral data.