********

FIELD OF THE INVENTION

The present invention relates to the field of analytical determinations of substances in complex mixtures; in particular, it refers to a method that uses ¹ H NMR CSSF- TOCSY spectroscopy to identify and quantify a predetermined series of substances, particularly carbohydrates.

STATE OF ART

Simple carbohydrates are among the most important components of food products and ingredients. Particular mono-, di-, and oligosaccharides may be naturally present or added to the final product for technological, nutritional or hedonistic purposes. They are a major source of calories and potentially the cause of some diseases like obesity and diabetes. To guarantee some important characteristics such as quality, genuineness and taste, it is extremely important to have detailed information about the amount of sugar in specific foods. As far as the analysis of simple carbohydrates is concerned, there is a growing interest in the development of new analytical methods for their simultaneous quantification, both in raw materials and in finished products.

Fruit juices and nectars are an important part of the human diet and have become very popular due to their multiple benefits on human health. Their adulteration not only has a clear negative impact on the consumer, but also on the productive sector, since authentic products must compete with the less expensive and adulterated ones. The most common practices of adulteration of fruit juices are dilution with water, the addition of sugars, and the addition of matrices derived from fruit such as pulp, or mixing with cheaper fruit juices. In addition, compounds such as flavorings, dyes, artificial sweeteners, stabilizers and preservatives may also be added. The relative concentration of sucrose, glucose and fructose falls within a consolidated range in different types of fruit juices.

The knowledge of the qualitative and quantitative distribution of sugars in fruit juices is important to establish their authenticity, and also to assess their quality and verify any microbiological alterations during storage [J. Agric. Food Chem. 2015, 63, 2639- 2646]. Glucose and fructose are the main components of honey, and the sophistications with cheap syrups from different sources (typically corn syrup, invert sugar syrup, fructose syrup and rice syrup) are very difficult to detect. In fact, the syrups contain mixtures of fructose, glucose, maltose, maltotriose, dextrins and other sugars in varying proportions depending on the production process, and the distribution profile of these carbohydrates can be easily manipulated to mimic that of honey. However, honey contains as many as 25 different oligosaccharides that have been related to its botanical and/or geographical origin [Food Chemistry 1 998, 63 (4), 549-562]. Their quantity, compared to glucose and fructose, can be used in principle to detect sophistication with low-cost syrups.

For these reasons, the knowledge of the amount of carbohydrates in different types of food products and ingredients is essential to determine their properties and other important characteristics such as taste, degree of ripeness, authenticity, preservation. A fundamental step when analyzing carbohydrates in food matrices with standard methods (such as HPLC and GC), is their separation from the rest of the main components, such as lipids and proteins, which can interfere with their determination and precise quantification. Once isolated, carbohydrates can be analyzed directly or subjected to some additional treatments to facilitate their subsequent analysis. Currently accepted methods for the quantification of sugars and oligosaccharides are based on separation techniques (possibly preceded by derivatization) with different detection schemes.

Nowadays, existing techniques show a high inter-laboratory variability for the quantification of sugars.

In fact, sugars do not have chromophores and, sharing very similar structures, the usual chromatographic separation techniques are not always efficient.

The International Honey Commission recommends an HPLC method to determine fructose, glucose, sucrose, maltose and turanose in honey using a refractive index detector. The inter-laboratory variability of this method is acceptable for glucose and fructose (1 .5 - 3.2%), but is very high for the less abundant sucrose (1 1 .4%) and does not improve much using a Pulsed Amperometric Detector (PAD, 6.8 - 1 2.5%) (http://www.bee-hexagon.net/en/network.htm). The method can also be used to quantify other saccharides such as melezitose, erlose, isomaltose, raffinose, as originally described by Bogdanov and Baumann (Apidologie, 1 997 (extra issue), 1 - 59). Since uncertainty increases with decreasing analyte concentration, minority sugars are quantified with even lower statistical confidence.

High Performance Anionic Exchange Chromatography (H PAEC) with PAD detection represents an improvement over the HPLC (Food Carbohydrates: Chem., Phys. Prop. And Appl., 2005, 2, 68-1 04). It provides a rapid analysis for different sugars, but is influenced by interfering substances such as lipids and proteins, which must be removed beforehand.

The monosaccharides can be quantified using gas chromatography (GC) provided they are derivatized as alditol acetates (in the case of neutral sugars) or as a trimethylsilyl derivative (in the case of acidic sugars). In both cases two reactions are needed, which implies a large margin of sample loss.

The content of simple sugars can also be determined by exploiting enzymatic reactions, if an appropriate method is available to monitor the progression of the reaction. The enzymatic methods are very specific, usually quick and suitable even at low concentrations of sugar. For example, enzymes can be used to quantify oligosaccharides through chromatographic determination (HPAEC-PAD) of sugar content before and after treatment [J. Anal. Chem., 2000, 368, 739-758].

High-performance or low-pressure chromatography with dimensional exclusion can also be used to separate mixtures of oligosaccharides according to their size [J. Biochem. Biophys. Methods, 2003, 56, 253-264].

FT-I R spectroscopy was used to analyze the sugar composition in fruit juices. In this case, an ATR (attenuated total reflectance) accessory is commonly used and calibration sets containing sugar mixtures are needed to develop a PLS model for data fitting [J. Agric. Food Chem., 2002, 50, 31 04-31 1 1 ].

The detection of adulterated honey with sugar syrup is complex and many efforts have been made to address the problem by analyzing carbohydrates.

Currently, the identification of the presence of exogenous sugars in honey is based on ¹³ C/ ¹² C isotopic ratio measurements with the analytical technique I RMS (Isotopic Ratio Mass Spectrometry). However, this analytical method is limited because it allows the identification of C4 (cane, maize) type of metabolism sugars, but not of C3 type (sugar, wheat, beet, and other) metabolism sugars, the latter having an isotopic ratio ¹³ C/ ¹² C not dissimilar to that of the sugars naturally present in honey. For milk and its derivatives, it is important to quantify the lactose and the derivatives of this sugar. The lactose level, for example, must be controlled in foods labeled "lactose free" for people intolerant to that sugar. The level of lactulose, on the other hand, is considered an indicator of milk changes due to heat treatment.

It is therefore evident there exists the need for an efficient, reliable and reproducible analytical method for identifying and quantifying carbohydrates in complex mixtures such as food matrices.

For all these reasons, there is a growing interest in the development of new analytical methods for the simultaneous quantification of carbohydrates both in raw materials and in finished products. Nuclear Magnetic Resonance (NMR) has shown its applicability to the analysis of simple and complex carbohydrates in food matrices without sample pretreatment. New methods have been recently developed using 1 H NMR spectroscopy to simultaneously quantify different sugars in honey, fruit juices and wine (FoodScreener® with NMR developed by Bruker, which is based on the acquisition of the specific spectroscopic fingerprint of each individual sample. These profiles are compared to a large database of authentic food samples using a multivariate statistical approach).

Although these methods are very useful for screening, they lack precision, they are limited to a few compounds, and their quantification thresholds are still too high. A different field from food matrices, but similar in terms of difficulties of analysis, is represented by biofluids such as urine, whey, saliva and breast milk, easily accessible and usable for the diagnosis of diseases and for the monitoring of drug therapy in clinical studies. For example, preliminary studies have indicated that NMR can easily detect increased levels of ketone bodies, such as 3D-hydroxybutyrate, acetacetate, acetone and glucose in the urine of diabetic patients [A. D. Maher, J. C Lindon, and J. K Nicholson Fut. Med. Chem., 2009, 1 (4), 737]. In the metabolomic context, the advantage of the NMR lies not only in the simplicity of sample preparation, but also in the fact that experiments (pulse sequences) can be adapted to the observation of specific classes of molecules with different physical properties (e.g. diffusion coefficient and relaxation times). However, the problem of spectral crowding still remains, which is generally, addressed using high magnetic fields or particular pulse sequences such as J-resolved spectroscopy. [Lindon, J. C, Nicholson, J. K., Holmes, E. and Everett, J. R. Concepts Magn. Reson., 2000, 1 2: 289-320].

In this context, where extended spin systems, such as in carbohydrates, exist, selective 1 D TOCSY (Total Correlation Spectroscopy) experiments are among the most immediate choices to isolate the signals of each chemical species in a complex mixture, and represent a promising approach. [Anal Chem, 2005, 77 (8), 2455-63]. In fact, the TOCSY signals can provide a good quantification of the species sought even when these are found in concentrations 1 000 times lower than those of the main components of the matrix. However, the relative intensities of the TOCSY signals are strongly influenced by the scalar coupling constants and by the relaxation times of the spin system, and these parameters must be taken into account. Furthermore, the strong spectral overlap in complex mixtures (see for example the anomalous region of a honey solution) requires much more selectivity than that obtained with standard shaped pulses.

One problem in the use of methods such as the selective TOCSY is that, in order to detect the individual molecular species within complex mixtures, this method can induce the simultaneous excitation of several molecular spin systems, especially in cases where the frequencies of target resonances are very close to each other. When this happens, problems may occur with the "purity" of the individual TOCSY signals observed and/or with their attribution to specific spin systems. To overcome this problem in US2007/0055456, TOCSY selective spectroscopy was proposed alongside a statistical method based on the Pearson correlation index, which provides a test for the purity of the single TOCSY peaks and allows the assignment of the peaks to a specific spin system.

The work J. Magn. Reson. 170 (2004) 97-1 03, (DOI : 1 0.1 01 6 / j.jmr.2004.06.004) describes a gradient-assisted chemical shift filter (ge-CSSF) for the in phase excitation of multiplets generated by superimposed 1 H signals. The ge-CSSF produces a rapid separation of the signals for the compounds with different spin- spin relaxation times and chemical shift differences down to only 1 -2 Hz. The aim of the present invention is to provide an analytical method which exploits 1 H-NMR 1 D TOCSY and provides an optimal resolution of the signal and consequently an accurate quantification of predetermined substances (for example carbohydrates) present in a complex matrix, regardless of whether they are major or minor components.

DEFINITIONS AND ABBREVIATIONS NMR = Nuclear Magnetic Resonance

1 H NMR = proton NMR spectroscopy

chemical shift: it is a number from which relates to the resonance frequency of a nucleus in different types of spectrometer

CSSF = Chemical Shift Selective Filter

TOCSY = Total Correlation Spectroscopy

FI D = Free Induction Decay: the signal revealed by the spectrometer, and from which the NMR spectrum is obtained

RF = Radio Frequency

1 D = one-dimensional (referred to an NMR spectrum)

SUMMARY OF THE INVENTION

The present invention solves the aforesaid problems by means of a method of identification and quantification of a predetermined substance in a complex mixture probably containing said substance, said method comprising:

- subjecting a sample of said complex mixture to the acquisition of 1 H-NMR 1 D CSSF TOCSY module at a selected frequency characteristic of the spin system of the substance;

- quantification of that substance, if present in the mixture, by comparison of the value of the integrated signal intensity, calculated in a determined frequency range selected for that substance, with the values of a calibration curve previously constructed through acquisitions ¹ H-NMR 1 D CSSF- TOCSY at selected frequencies characteristics of the spin system of that substance in a series of standard solutions in which that substance was present at known concentrations, said standard solutions consisting of artificial complex mixtures similar to that to be analyzed.

The present invention fits well within the framework of the "Societal Challenges" of the Horizon 2020 strategy, with particular reference to the theme "Food security, sustainable agriculture and forestry, marine and maritime and inland water research, and the abeconomy". A main objective of this challenge is to ensure a supply of safe, healthy and high quality food products. The method of the present invention represents a new and efficient analytical tool for identifying and quantifying simple carbohydrates in food, and food ingredients, which can lead to a better control of possible adulterations.

Furthermore, a more detailed understanding of food composition can help to study and understand the positive and negative effects of simple carbohydrates on human health.

The study of the levels of these metabolites or their derivatives in biofluids is of great importance in medical diagnostics.

DETAILED DESCRI PTION OF THE INVENTION

The CSSF according to the invention is assisted by magnetic field gradients for the in-phase excitation of the selected resonances [J. Magn. Reson., 2004, 1 70, 97- 103]. With reference to the problem of the TOCSY signal purity, the CSSF selective excitation scheme employed in the present invention solves the problem of the simultaneous excitation of several signals (unless these are exactly superimposed) and therefore eliminates the need for a statistical analysis of data robustness. In the case of residual spectral overlap, the following aspects can be modulated: soft pulses, TOCSY mixing time, CSSF bandwidth. Preferably, the acquisition according to the method of the invention is preceded by a pre-calibration sequence, which induces in the sample the same type of thermal perturbation introduced by the TOCSY module.

An important parameter in the resolution of very similar chemical shift signals is the duration of the filter, whose length is inversely proportional to the selected band. Preferably, the filter duration is set between 50 and 250 ms. The substances identifiable by the method of the invention must have a molecular structure bearing extended spin systems. Preferably for the purposes of the present invention the substances that can be the object of identification and quantification are selected in the group consisting of carbohydrates, carbohydrate derivatives, ketones, organic acids and their derivatives, lipids, amino acids and their derivatives; more preferably carbohydrates.

Table 1 shows the excitation frequencies and the range of signals to be integrated into the CSSF-TCOSY spectrum for a series of sugars typically found in honey.

The method of the present invention is preferably applied to food matrices with a high carbohydrate content such as honey, alcoholic beverages (wine, beer, liqueurs), fruit juices, jams, milk (cow's and mother's milk) and derivatives, rice, potatoes, cereals and flours, tomatoes, legumes.

It can also be applied to biofluids such as urine, plasma, saliva and sweat.

The analysis of solid matrices requires a minimum treatment of the matrix to obtain a sample suitable for the 1 H-NMR 1 D CSSF TOCSY.

If the complex mixture to be analyzed is a honey, the sample to be analyzed according to the invention is prepared in such a way that the total quantity of sugar is 1 5-25% w/w with respect to the total weight of the sample, preferably 20% w/w, and pH (or better, pD) is buffered in the 4.00-5.00 interval.

The buffer is preferably a phosphate buffer.

If the complex mixture to be analyzed is a honey, the sample, before dilution in D2O, must be subjected to inactivation of the enzymes naturally present in honey.

Preferably, the inactivation takes place by heat treatment, preferably by heating to

40-50 °C for about 2 days.

If the complex mixture to be analyzed is a drink (fruit juice, beer, wine) the sample does not need any pre-treatment or dilution ; it is preferable to perform a filtering to obtain a homogeneous solution.

If the complex mixture to be analyzed is urine or saliva, the sample does not need any pre-treatment or dilution, it is preferable to centrifuge and analyze the supernatant.

If the complex mixture is milk, it can be studied raw, but it is preferable to study the whey obtained by acidifying and eliminating the rennet by centrifugation. For each complex mixture, calibration curves are built using the absolute area for each analyte. In general, at least 5 standard solutions are needed with increasing amounts of analytes (e.g. sugars) to obtain a calibration curve. The calibration curve is obtained by reporting the absolute integrals of the selected peak for each analyte against its concentration (see for example Table 3 and Figures 4-5). The amount of analyte (e.g. sugar) in unknown solutions (standard or real sample) will then be obtained using a linear regression.

According to the present invention, standard solutions are artificial complex mixtures mimicking (as far as possible) a typical complex mixture of the matrix to be analysed. Ideally, a standard solution should reproduce any possible matrix effect that may alter the observable quantities on the analyte by the combined effect of all components of the sample other than the analyte itself. In the experimental practice, standard solutions are prepared in such way as to reproduce the main physico- chemical characteristics of the natural matrix (e.g. pH, ionic strength, ecc). As an example, a practicable standard for honey must account for the extremely high concentration of fructose and glucose in natural honey (see par. 7 of the experimental section). Similarly, a practicable standard solution for milk may be prepared by dissolving the standard analytes directly into whey (in place of water). Along the same line, a practicable standard solution for wine is an ethanol/water mixture, whereas a good standard solution for dilute fruit juices generally consist in water only.

The present invention can be better understood in the light of the following examples.

BRIEF DESCRIPTION OF THE FIGURES

FIG. 1 - On the left, conventional 1 D spectrum of buffered standards. On the right, respective CSSF-TCOSY spectra obtained by centering the selective pulse at the optimal frequency for each sugar.

FIG. 2 - A) Black line: expanded region of the 1 D spectrum (4.9-5.5 ppm) of a chestnut honey diluted in a buffer solution at pH 4.40. Black arrows: the arrows, positioned on the spectrum of the real sample, indicate the exact resonant frequency of the signal of the anomeric proton. B). Overlaying CSSF-TOCSY spectra of sugars in buffer solution (gray traces) and chestnut honey (black traces). The gray rectangles indicate the signals used for quantification.

FIG. 3 - A) Black line: expanded region of the 1 D spectrum (4.7-3.1 ppm) of an acacia honey diluted in a buffer solution at pH 4.40. Black arrows: the arrows, positioned on the spectrum of the real sample, indicate the exact resonance frequency of the signal of the anomeric proton. B) Superposition of the CSSF- TOCSY spectra of sugars in buffer solution (gray traces) and in acacia honey (black tracks). The gray rectangles indicate the signals used for quantification.

FIG. 4 - Upper Graph : determination of the amount of maltulose in chestnut honey through the calibration curve. Graph below: method of standard additions. In this case, the concentration is obtained from the intercept of the line. In black the sugar signal to be quantified. The value obtained by both methods is exactly the same and is shown in the gray circle.

FIG. 5 - Graph above: determination of the amount of maltulose in acacia honey through the calibration curve. Graph below: method of standard additions. In this case the concentration is obtained from the intercept of the line. In black the sugar signal to be quantified. The value obtained by both methods is exactly the same and is shown in the gray circle.

FIG. 6 - Effect of the filter/spin relaxation length on the intensity of the CSSF TOCSY signals. Continuous black line: CSSF TOCSY of trehalose obtained with a filter with a maximum duration of 50 ms; dotted black line: CSSF TOCSY of trehalose obtained with a filter with maximum duration of 250 ms.. The decrease in intensity in the case of the longer filter is due to spin relaxation.

FIG 7 - Scheme of the precalibration pulse sequence for TOCSY-CSSF experiments.

FIG. 8 - a: 1 H NMR spectra of a fruit juice. Upper trace: untreated fruit Juice. Lower trace: fruit juice of the same sample with addition of standard sorbitol, b: Expanded regions of 1 H NMR spectra where sorbitol protons resonance, c: CSSF-TOCSY spectra of standard sorbitol in water (upper line), of Apple juice spiked with standard sorbitol (middle line) and of untreated Apple juice (lower line), d: CSSF-TOCSY spectra of Pineapple juice spiked with standard sorbitol (upper line) and of untreated pineapple juice (lower line). The oval inset highlights the presence of sorbitol. FIG. 9 - A: 1 H NMR spectrum of serum extracted from lactose-free milk. B. CSSF- TOCSY spectrum of the same sample spiked with standard lactose. C. CSSF- TOCSY spectrum of standard lactose in water.

EXPERIMENTAL PART

1. Standard sugars.

D-glucose > 99.5%, D (+) Mannose, D (-) fructose > 99%, D (+) Turanose > 98%, Hexlasma > 94%, Isomaltotriose, D (+) Melibiose > 99.0%, D (+), Raffinose pentahydrate > 98.0%, D— igluconic acid sodium salt, Palatinose hydrate > 99%, L- Ramnose monohydrate minimum 99%, Saccharose > 99.5%, D (+) Maltose monohydrate minimum 98%, Melezitose > 99.0%, Trehalose dihydrate (Certified Reference Materials), Maltulose > 98.0%, Nigerose > 94.0%, D-Panose > 97%, Maltotriose, 98% Isomaltose, Gentiobiose > 85%, 1 Kestose > 98.0%. Supplied by the company SIGMA-ALDRICH.

2. Buffer solution for NMR (NMR buffer)

It is known that pH and dilution of samples have an effect on the integrated intensity of NMR signals. Even the chemical shifts of some compounds are sensitive to pH, which must therefore be controlled. However, carbohydrates not containing acidic or basic groups demonstrate a reproducibility of the chemical shift even for small variations in pH. In the case of honey, the samples must be diluted to reduce viscosity and to facilitate their handling. It is interesting to note that preliminary 1 H NMR tests indicate that quantitative results in honey are influenced by the high amount of inverted sugar present, which changes the thermodynamic activity of the aqueous medium and alters the relaxation times of the dissolved species due to the high viscosity. The samples to be analyzed are prepared in such a way that the total amount of sugar is 20% w/w: this represents a good compromise between sample viscosity and instrumental sensitivity.

The buffer solution for NMR analysis was prepared by dissolving 2.55g KH2PO4 and 2.45mg NaN ₂ in 50ml D2O (d-99.97%) and regulating the pD at 4.40 with H3PO4. 3. Spectra acquisition.

All NMR experiments were acquired on a Bruker Avance I I I spectrometer operating at 500.1 3 MHz of proton frequency and equipped with a "z-gradient broadband inverse" probe (BBI).

The probe temperature has been set at 298.1 K, but can be varied as long as it remains constant during the acquisition of all the necessary spectra. An accurate control of temperature and pH gives a high reproducibility on the resonance frequencies of all sugars. On this premise, it is possible to keep the parameters of acquisition unaltered for each sample, using an automatic acquisition procedure of the spectra assisted by an auto sampler. The stability of the instrument has been demonstrated by calculating the variation of reference signals between spectra of a standard solution (see next paragraph) acquired at different times over a period of eight months. The standard deviation of the integral found is small and falls within the range 0.2% - 1 %

Spectrum acquisition parameters. The acquisition time for minor sugars is 7 min with the following acquisition parameters: number of scans 8, number of cycles 14, 8K points, relaxation time 2s, spectral window 6000Hz. For glucose and fructose, the acquisition times are 2 minutes.

Spectrum processing parameters. All the spectra were processed with an ACDLab 201 2 using an automatic procedure with the following parameters: zero filling 32k, exponential window function with 0.5Hz line broadening. The signals selected for each analyte were automatically integrated after having fixed the appropriate interval centered on the selected resonances. Absolute values were considered. Selective TOCSY experiments

In order to resolve the largest possible number of signals, the method of the present invention employs a CSSF assisted by magnetic field gradients for the phase excitation of the selected resonances [J. Magn. Reson., 2004, 1 70, 97-1 03]. This signal is then used as the source for a subsequent magnetization transfer along the spin system of the selected sugars. The CSSF is based on the constructive addition of the on-resonance signals, while off-resonance magnetization is canceled due to destructive interference caused by the evolution of the chemical shift. This is obtained by summing up several FI Ds acquired with a gradually increased period of evolution of the chemical shift. CSSFs are able to produce a separation of signals with relative differences of chemical shift up to 1 -2 Hz. The additional use of magnetic field gradients and the filtering of the coherences to zero-assure that the obtained spectra are free of artifacts. The highly selective excitation, together with an isotropic mixing scheme, provides the subspecies of the desired spin systems. An important parameter in the resolution of very similar chemical shift signals is the duration of the filter, whose length is inversely proportional to the selected band. A filter with a total duration of about 50 ms isolates signals with relative distances of about 8 Hz; a filter with a duration of about 250 ms instead isolates signals with relative distances of about 2 Hz. An intrinsic disadvantage to the use of a long filter is the partial loss of signal by transverse relaxation (about 1 0% for a filter of 250 ms applied to a signal with T2 = 1 s, see Fig. 6). A collateral disadvantage is instead represented by its very narrow selection band, a characteristic which makes it less tolerant to possible errors in the positioning of the offset. Especially for this reason the filter with a shorter duration is always preferable, even in cases where the CSSF excitation profile is semi-selective. This provided that the TOCSY sub-probe resulting from said excitation exhibits some isolated (and therefore integrable) signals of the single species of interest.

A problem with the high selectivity of CSS filters is given by their extreme sensitivity to small offsets of the RF transmitter. In practice, the pulse train used in the TOCSY mixing slightly warms the sample by varying the resonance frequencies by a few hertz. This implies that a simple ¹ H NMR spectrum acquired under standard conditions cannot be used to select the filtering frequencies for the CSSF. To overcome this drawback, a pre-calibration sequence has been constructed that induces in the sample the same type of thermal perturbation introduced by the TOCSY module, incorporating the train of pulses before magnetization reaches equilibrium (Fig.7). The use of this calibration is preferable to reduce acquisition times (especially when working in automation), but in fact not essential. 4. Sample pretreatment

Before analyzing the various honeys, each sample was heated to about 40-50 ° C for 2 days in order to deactivate the enzymes. This operation is necessary because the enzymes naturally present in honey, once this is dissolved in water, can significantly reduce the concentration of some sugars (mainly sucrose, maltose and isomaltotriose, trehalose).

Apart from filtering to obtain a homogeneous solution, fruit juices are analyzed without any kind of manipulation and dilution.

5. Honey samples

NMR samples were prepared by dissolving about 240 mg of honey in the NMR buffer and adjusting the ratio honey (mg)/buffer (ml) to exactly 240 mg/ml. The pH was carefully adjusted to 4.40.

6. Sugar standard solutions in NMR buffer:

Each sugar was dissolved in the "NMR buffer" solution. Of each solution the conventional 1 D spectrum was acquired and the CSSF-TOCSY spectrum centered on the frequency chosen appropriately for each sugar. The choice of the excitation frequency was made after acquiring different TOCSY spectra of each standard, centered on the signals of the anomeric protons. A comparison of all these spectra allowed the identification of the optimal excitation frequency for each sugar in the TOCSY spectra. From the same analysis, the spectral regions were identified containing specific signals for each sugar that can be used for quantification. Figure 1 shows, on the left, all the conventional spectra and on the right the selective spectra containing specific signals (highlighted with the rectangle) chosen for quantification.

Table 1 shows the excitation frequencies and the range of signals to be integrated into the CSSF-TCOSY spectrum for each sugar. Table 1 - Excitation frequency (calculated as the difference in Hertz, A, from the frequency of the glucose proton β taken as a reference) and the integration interval of the 21 carbohydrates analyzed.

Tmax Integration region (ppm) sugar Δ (Hz)

(ms) (PP ^m )

Glucose 50 3.43 - 3.53

Glc

Fructose 431.55 50 3.76 - 3.86

Fru

eparation of "artificial" honeys for the calibration curves To build the calibration curves, 8 standard solutions were prepared (called "artificial honey" solutions, AHS) containing glucose and fructose and the minor sugars present in the honey on which all the analytical evaluation of the procedure was performed.

The concentration level chosen for each sugar corresponds to the average values reported in the literature, for that particular sugar in honey, in order to reproduce the natural matrix as closely as possible.

The 8 standard solutions contained a constant amount of glucose (74.00 mg/ml), and fructose (98.28 mg/ml) and variable concentrations of minor sugars, chosen in such a way as to always obtain the same total amount of sugars. An example of the procedure is shown in the diagram in Table 2. The pH was adjusted to 4.40. The eight concentration ranges, indicated by letters in Table 1 , were a = 0.80 - 1 .30 mg/ml, b = 1 .70-2.50 mg/ml; c = 2.77-3.90 mg/ml, d = 4.20-5.30 mg/ml, and e = 5.50- 6.50 mg/ml, f = 6.60-7.90 mg/ml, g = 8.20-9.80 mg/ml, h = 10.00-12.70 mg/ml. The calibration curves were obtained by reporting the absolute value, for each sugar, of the integral of the selected CSSF-TOCSY peaks, according to its concentration.

Table 2 - Scheme of concentrations of 8 minor sugars in 8 AHS

In the case of glucose and fructose, the calibration curves were built differently; in fact, solutions containing only glucose and fructose have been used with a different ratio of the two sugars, maintaining the total sum of the sugars always equal to about 172 mg/ml. This procedure was used to reproduce the same density and viscosity as a true honey solution. Calibration curves are used to estimate the concentration of sugars of unknown mixtures by including the absolute intensity (Va) of the total peaks selected in CSSF- TOCSY experiments in the equation. This value was converted into g/1 00 g of honey in order to compare the data with the literature.

Internal standard. To compensate for possible instrumental fluctuations, use can be made of an internal standard, i.e. a stable chemical species which is added both to the sample and to the standard solutions. In this case, the calibration curves are built by replacing the absolute value of integrated intensities Va for the signal of interest with the ratio Va/Vs ,where Vs is the integrated intensity for the signal of the internal standard in the corresponding one-pulse control spectrum. In this way, the proposed methodology becomes independent from the periodic check of the calibration curves.

8. Analytical performance of the method

The linearity and accuracy (sum of precision and accuracy), the limit of detection (LOD), the limit of quantification (LOQ) were determined on the AHS ("artificial" honey solutions). The statistical parameters were determined on 1 5 of the 21 standard sugars analyzed, which had sufficient quantities for validation.

9. Linearity, repeatability and accuracy

The linearity was tested from the correlation coefficient of the regression obtained by plotting the absolute intensity of the total peak of the analytes, obtained from the 8 AHS spectra, against their concentrations.

Table 3 shows the equations of the linear regressions and the corresponding correlation coefficients.

Table 3 - Linear regression parameters, Detection limit in g / 100g honey (LoD) and quantification limit in g/100g honey (LoQ) for the 15 sugars considered. $ standard error on the slope, * standard error on the intercept. ^a Values may change if the acquisition parameters and / or sample dilution are changed. Sugar Slope S _a Intercept Sb ^* R ² LoQ ^a LoD ^a

Glucose 3673 39.2 0.9957 0.09 0.028

Fructose 3330 35.2 0.9968 0.18 0.05

Ramnose 7882 75.07 0.037725 0.4980 0.9997 0.03 0.0091

Mannose 3760 44.0 -0.950 0.288 0.9995 0.06 0.018

Sucrose 4314 54.9 -0.192 0.384 0.9995 0.05 0.015

Palatinose 1710 63.5 -0.22 0.3444 0.9945 0.0509 0.015

Maltose 1160 13.4 -0.379 0.0938 0.9995 0.20 0.06

Maltulose 1700 23.2 -0.169 0.142 0.9991 0.10 0.03

Turanose 2490 18.1 -0.130 0.134 0.9997 0.17 0.05

Trehalose 5040 105 1.38 0.68 0.9978 0.05 0.015

Melibiose 1633 67.89 -0.470 0.395 0.9931 0.13 0.039

Isomaltotriose 1310 48.0 -0.323 0.214 0.9960 0.18 0.055

Raffinose 1000 13.5 0.458 0.0672 0.9996 0.21 0.064

Erlose 1826 21.4 0.175 0.0700 0.9997 0.08 0.024

Melezitose 1630 17.8 -0.0633 0.111 0.9993 0.13 0.038

The high values of the correlation coefficients (R2> 0.99) of the regressions confirm the existence of a linear correlation between the experimental values and the concentration. The repeatability and accuracy of the method were determined by further "AHS" containing the 15 sugars considered. The repeatability of the method was obtained by calculating the relative standard deviation (RSD%) of 9 different measurements of an integrated signal (three different preparations of the "artificial honey" solution and three different data acquisitions). The accuracy of the method was expressed by evaluating the agreement between the measured concentration (mc) and the nominal concentration (nc) of the test sugar, such as (mc-nc) ^* 100/nc (bias). Table 4 shows the values obtained.

Table 4 - Repeatability (% RSD) and accuracy (Bias%) in AHS

Very good repeatability values are obtained (% RSD <2.2%) along with an excellent agreement between the nominal and experimental concentration demonstrated by the low% bias values.

10. Detection and quantification limits.

The limits of detection (LOD) and quantification (LOQ) have been defined as the concentration of the analyte which corresponds to a signal to noise ratio equal to 3 and 10 times the average level of instrumental noise, respectively. The signal-to- noise ratio (S/N) was estimated on the signal of maximum intensity in the multiplet under examination. The LOD and LOQ limits were determined by plotting the S/N ratio against the concentrations of the peaks selected for each sugar in artificial honey. The data are shown in Table 3.

1 1 . Special cases: The example of sucrose and maltose. Because of their structural analogy, the sucrose disaccharide (Glc ( 1 -2β) Fru) has a spectrum whose resonances are significantly overlapped with those of the trisaccharide eriose (Glc (cc1 -4) Glc (cc1 -2β) Fru). However, in the spectrum of eriose it is possible to isolate a quantifiable signal according to the procedure described above. The quantification of sucrose, on the other hand, is indirectly obtained by subtracting the eriose contribution from the selected signal. Another analogous case is that of maltose (Glc (cc1 -4) -Glc) whose resonances are superimposed on those of maltotriose (Glc (cc1 -4) Glc (cc1 -4) Glc). Similarly, a specific signal is identified for maltotriose whose contribution is subtracted from that of maltose.

12. Application of the method to samples of authentic honeys.

Figure 2 and Figure 3 illustrate the results obtained on two samples of real honey. In box A of both figures two expanded zones of the conventional 1 D spectrum are reported, where the anomeric protons of the sugars resonate. Figure 2 refers to chestnut honey, while Figure 3 refers to an acacia honey. In boxes B there is a comparison between CSSF-TOCSY spectra obtained from buffer standard solutions (gray traces) and CSSF-TOCSY spectra obtained from honey (black traces). As evident from this comparison, the trace obtained from the real samples corresponds perfectly to that obtained from the standard solutions. This means that, despite the spectral complexity present in the conventional honey spectra, due to the overlap of the anomeric protons of the isomers of all the sugars present, the optimized sequence is able to select the sugar to be quantified.

-The case of palatinose and isomaltotriose highlights the specificity of the method. As evident, the palatinose (Figure 3 box B) and the isomaltotriose (Figure 2 panel B) resonate in an overlapping region where other sugars resonate, but despite this the pulse sequence allows us to isolate their spin system without any artifact.

-The raffinose case further supports the potential of the whole procedure. At the frequency of the raffinose, in the chestnut honey, the one-dimensional spectrum displays a doublet that may easily be attributed to this sugar. However, the raffinose spin system in the CSSF-TOCSY is absent. To verify this data and to exclude errors in the excitation frequency, a further spectrum was acquired after adding the standard raffinose to the chestnut sample (gray spectrum in box B of figure 2). This data confirms the absence of raffinose, the exact frequency of excitation and highlights the fact that with this experiment we avoid any assignment errors that can instead result by looking at the one-dimensional spectrum. Moreover, this observation justifies the usefulness of the CSSF-TOCSY also for isolated or concentrated signals that may hide overlaps which hamper the accuracy of the method. Table 5 shows the quantities of glucose and fructose, the two predominant sugars in honey, determined on three honey samples with the procedure according to the present invention and with the use of one-dimensional as proposed in literature.

Table 5 - Glucose and Fructose values determined with the conventional method and with the method according to the invention

With the literature procedure the value is always larger than that determined with the CSSF-TOCSY, which indicates the higher specificity and exactness of the method proposed here.

-The filter used in the CSSF-TOCSY is very selective, however in the honey matrix, the most complex matrix in terms of carbohydrate content, some sugars resonate at a distance of a few Hertz. For example, trehalose and mannose have a resonance frequency difference as low as 3 Hz. As a consequence, the signals of the interfering sugar are present in both the selective spectra. It is interesting to note that the co- presence of the two spin systems does not preclude the quantification of these two sugars, having completely resolved spectra (Figure 1 ).

-Finally, we demonstrate that the CSSF-TOCSY works very well also for minor sugars partially overlapped with very concentrated sugars: the maltose resonances are partially overlapped to those of glucose, which is about 40 times more concentrated. As shown in Figure 2B, the CSFF-TOCSY spectrum of maltose in honey (black trace in Figure 2B) is not affected by the presence of glucose and has the same profile as the standard maltose in buffer solution (gray trace in Figure 2B). 13. Analytical recovery

The accuracy of the entire analytical procedure was evaluated on three honey samples by recovery, a parameter that allows to determine the losses of analyte during the experimental procedure.

Recovery was determined using the standard addition method. The stock solution contained 9 minority sugars under examination (6 mg/ml) dissolved in NMR buffer. A typical series of standard additions consists of seven different concentration levels for each sugar. The seven solutions were prepared by dissolving exactly 240 mg of honey in different volumes of stock solution, to reach increasing concentrations of sugar, and by adjusting the final volume to 1 ml with the NMR buffer. The levels of concentration of the selected sugars are: 0.24, 0.66, 1 .02, 1 .46, 1 .82, 2.15 and 2.60 g / 100 g honey. The results obtained with the standard additions were compared with those obtained from the calibration curves (table 6). Acacia honey and chestnut honey and honeydew honey have been considered in the case of raffinose.

The percentage ratio of the two values returns the value of the recovery. Figures 4 and 5 illustrate the employed methodology.

Table 6 - Content (as g / 100g honey) of the minor sugars present with a concentration larger than the limit of quantification and determined with the method of standard additions (SA) and calibration curves (CC) and relative recovery (R.%) ^* sugar determined on a honeydew sample.

Acacia Chestnut

Sugars Cone. AS Cone. CC R. % Cone. AS Cone. CC R. %

Sucrose 1.35210.046 1.3210.02 9814 <LoQ

Maltose 1.8710.10 1.9010.03 10116 0.6610.04 0.67010.012 101110

Maltulose 0.8710.02 0.86110.006 98.712.4 3.3110.06 3.25610.023 98 1 4

Palatinose 0.42310.017 0.43910.005 10414 1.0610.07 1.08810.012 103 17

Turanose 1.9910.05 2.0110.04 10113 2.0410.06 1.9910.04 97 1 3

Trehalose <LoD 0.08310.02 0.074210.0006 89 1 16

Mannose <LoQ 0.0710.02 0.085010.0014 121 1 34

Melezitose <LoQ 0.75310.003 0.7610.01 100.911.4

Raffinose ^* 0.3210.03 0.3110.01 9614 14. Identification and quantification of sorbitol in fruit and fruit juices

The method according to the invention allows a rapid quantification of sorbitol in fruit juice whose spectrum is largely superimposed to the resonances of other major sugars (Fig. 8 panel a and b). For this purpose, other NMR techniques (bidimensional, j-resolved) are not useful and traditional methods (HPLC) require a large quantity of sample not always available. On the contrary, CSSF-TOCSY spectra provide clear evidence of the presence of sorbitol and allow its quantification in fruit juice. As an example, Fig. 8 panel c and d report CSSF-TOCSY subspectra acquired on apple juice and pineapple juice respectively. The oval inset highlights the presence of sorbitol in apple juice and the absence of the same sugar in pineapple juice.

15. Correction of instrumental fluctuations with an internal reference. The addition of an internal standard to the sample allows to compensate for possible instrumental fluctuations. To this aim, the absolute value of integrated intensities for the signal of interest (Va) is replaced with the ratio Va/Vs where Vs is the integrated intensity for the signal of a standard. In this way, the method of the invention becomes independent from periodic check of the calibration curves.

16. Residual lactose in dairy products "lactose free".

The method of the invention allows a rapid quantification of lactose in especially in lactose free derivatives obtained upon addition of lactases. Besides being comparably time-consuming, none of the current traditional methods appears to provide a rapid screening of this sugar. The essential drawback of the majority of existing analytical methods for lactose determination in milk is that many of them cannot detect low lactose concentration (below 0.5 gkg ^-1 ). On the contrary, CSSF- TOCSY spectra provide a straightforward methodology to quantify the amount of residual lactose with limit of detection lower than 0.1 gkg ^-1 . See Fig. 9.