Introduction

Plant-derived natural products have potential applications in medicinal chemistry, by means of chemotherapeutic, chemopreventive, antioxidant properties, protection against cancer, resistance towards coronary heart diseases, and various other biological functions (Newman and Cragg 2007; Yang et al. 2007). Therefore, the analytical description of plant extracts rich in polyphenols is of considerable interest for their classification and for understanding structure-activity relationships, which can guide the synthesis of analogue compounds with improved bio-availability, stability, nutritional quality, potency, and specificity (Quideau et al. 2011; Duthie et al. 2000; Ferguson 2001). Among foodstuff, dietary polyphenols from olive oil have been a large part of discussions, notably in the Mediterranean realm and its vast diversity of locally grown olives (Yang et al. 2007; Gutfinger 1981; Tsimidou et al. 1992; Aparicio and Luna 2002; Vinha et al. 2005; Haddada et al. 2007; Ocakoglu et al. 2009). Several classes of phenolic molecules (e.g., simple phenols, secoiridoids, lignans, and flavonoids) impart sensory and nutritional quality of olive oils, while providing multiple pharmacological properties (Bendini et al. 2007; Franco et al. 2014; Visioli et al. 1998). The total polyphenol content and the polyphenol profiling of an olive oil depend on many factors such as olive cultivar, geographical origin, the production, processing and storage techniques, and the olive oil age. While the total polyphenol content provides a quality gauge, identification of individual phenols can be useful as a fingerprint to characterize typical extra virgin olive oils. The most commonly used chemical and analytical methods applied to this purpose have been reviewed recently (Ignat et al. 2011).

As a technique of choice, NMR has been applied either on its own or as a chromatographic detector (LC/NMR) for the characterization of olive oils (Sacchi et al. 1998; Christophoridou and Dais 2009; Christophoridou et al. 2005; Mannina et al. 2001, 2002, 2010; Alonso-Salces et al. 2010; Fuentes De Mendoza et al. 2013; Bastoni et al. 2001; Longobardi et al. 2012; Camin et al. 2016; Mannina and Sobolev 2011). Direct NMR approaches, although technically easier to setup and perform than hyphenated techniques, can encounter bottlenecks in the study of complex mixtures, mainly stemming from spectral overlaps. By increasing the spectral resolution, multidimensional correlation NMR experiments significantly aid the spectral analyses of mixtures (Lin and Shapiro 1997). This has been demonstrated recently in the case of polyphenols (Charisiadis et al. 2017). For instance, NMR-based diffusion experiments (Johnson 1999; Nilsson et al. 2004; Reddy et al. 2013), alone or in combination with other NMR techniques, can greatly enhance the spectral simplification (Barjat et al. 1998; Viel and Caldarelli 2008). More recently, multiple-quantum (MQ) NMR, an experiment originally introduced at the beginning of 2D NMR (Munowitz and Pines 1986; Ernst and Bodenhausen 1990), has found new life for the structural analysis (Baishya et al. 2008; Manjunatha Reddy et al. 2009a, b) and the characterization of mixtures of small molecules (Dalvit and Bohlen 1997; Martineau et al. 2011, 2012) (Reddy et al. 2013, Guennec et al. 2015, Reddy and Caldarelli 2010, 2011, 2012; Piotto et al. 2011; Manjunatha Reddy and Caldarelli 2013; Manjunatha Reddy et al. 2016). In this latter case, the highest possible excitable MQ coherence (also referred to as maximum-quantum; MaxQ) order has shown intriguing potential for the speciation of mixtures, particularly of aromatic molecules. With respect to other popular 2D NMR methods, such as HSQC or TOCSY, the MaxQ NMR protocol is particularly apt to speciate multicomponent mixtures since it identifies in succession molecular fragments as a function of their number of protons. MaxQ NMR has been successfully applied for the characterization of aromatic region of the 1H NMR spectra in a mixture of mono- and polyaromatic hydrocarbons (Reddy and Caldarelli 2010), phenolic compounds in both a model mixture and in the polar part of an olive oil extract (Reddy and Caldarelli 2011) and environmentally relevant molecules (Reddy and Caldarelli 2012). The enhancement in spectral resolution of the MaxQ experiment stems from a dramatic reduction in the number of signals, which is due to two factors. Firstly, only molecular fragments of mutually coupled protons can produce MQ signals, thus providing a first filter (such as the commonly employed double-quantum filter for removing singlets). Then, the higher the observed MQ order, the fewer are the number of signals, with the limit case of a simple singlet for the MaxQ order. For example, in a network of five-coupled spins, there are a maximum of nearly 200 NMR signals that can be compressed into a singlet by observing the five-quantum order, a two-order magnitude reduction in the spectral complexity. This facilitates the identification of the molecules in the analyzed mixture. Since molecular fragments do not possess the same number of coupled protons, several orders must be recorded in principle to obtain a full analysis. For polyphenolics, for instance, MQ can vary between 1 and 5. MaxQ NMR can be employed, in principle, to any case where a detectable MaxQ coherence order can be excited. Even higher resolution could be achieved by including the MaxQ approach in 3D NMR experiments, and the advantage/cost ratio for going to these more complex techniques should be evaluated on a case-by-case basis (Manjunatha Reddy et al. 2016). Quantitative protocols were also established within the context of MaxQ, which requires the use of calibration curves, but their implementation was outside the scope of this paper (Reddy and Caldarelli 2012). Indeed, here we present a MaxQ NMR analysis for profiling polyphenols in extra virgin olive oils as a complement to total polyphenol content assessment. Although the polyphenols content and speciation may depend on the geographical origin, extraction methods, and processing and storage techniques, it was suggested that commonly encountered polyphenols such as lignans, ligstroside, and tyrosol derivatives are expected to be in relatively higher concentrations (Christophoridou et al. 2005) and thus dominate the MQ spectra as well. Since the NMR spectral parameters for these molecules exist in the literature, the MaxQ-NMR signals can be annotated accordingly, along the lines of a previously proof-of-principle work (Christophoridou and Dais 2009).

Materials and Methods

Samples

Three Italian extra virgin olive oils (batches I, II, and III) with different polyphenols content were provided by UNAPROL (Italian Consortium of Olive Oils). The polyphenol content as determined by UNAPROL according to a standard methodology (Montedoro et al. 1992) was of 595, 109, and 237 mg/kg of oil for batches I, II, and III, respectively.

The three olive oils were subjected to a liquid/liquid extraction procedure (Montedoro et al. 1993) using a binary solvent mixture, water/methanol, 20/80 v/v (Lesage-Meessen et al. 2001) 16 g of olive oil, in each case, was mixed with 16 mL of solvent and stirred in a vortex for about 20 min (three times) and centrifuged at 5000 rpm for about 10 min (three times), and the methanol fraction was collected. Methanol was evaporated under vacuum at 60 °C, and the residue was washed with 16 mL of hexane (three times). Hexane was then evaporated at 30 °C, and a solid residue was obtained. Three individual samples were prepared by dissolving 24.8, 15.0, and 14.6 mg of batches I, II, and III samples respectively in 500 μL of methanol-d.

NMR Spectroscopy

For all the extracts, the 1H (500 MHz) one-pulse and two-dimensional MaxQ–1Q correlation NMR experiments were performed using a Bruker Avance III 500 MHz NMR spectrometer equipped with a TXI cryogenic probe. For one-pulse experiments, 32 co-added transients were acquired using a relaxation delay of 4 s to an overall experimental time of 3 min. For MaxQ experiments, a three-pulse sequence, (π/2)-(τ/2)-(π/2)-(τ/2)-(π/2), was used to excite multiple-quantum transitions where τ was allowed to vary for optimal signal uniformity and sensitivity. Optimal τ values were found to be 220 ms for 3Q and 110 ms for 4Q excitation. No search for the optimal MQ excitation efficiency was performed (Köcher et al. 2016), for economy of time. Two-dimensional MaxQ spectra were acquired with 512 FIDs, each with 16 co-added transients, corresponding to a FID resolution of 9.76 × 0.625 Hz (p = 4) and 11.72 × 0.67 (p = 3), respectively along MaxQ and 1Q axes. Coherence selection was obtained through two half-sine-shaped pulsed field gradients placed before and after the reconversion pulse in a ratio of 1:3 and 1:4, respectively, for 3Q and 4Q coherences. For each sample batch, the total experimental time was about 16 h (8 h for each MaxQ-order). The MaxQ spectra were zero-filled to a 1024 × 2048 (1Q × MaxQ) data matrix and weighed with sine bell window functions. All spectra are displayed in magnitude mode. The MaxQ dimension was scaled by the p-quantum order in order to provide a homogeneous display along the MQ series. Spectral assignments were carried out based on the 1Q chemical shifts and on the multiplet structure.

Results and Discussion

Figure 1 illustrates one-pulse 1H spectra of three batches of olive oil extracts in methanol-d 4 plotted on top of each other for a comparison of aliphatic, aromatic, and aldehydes signals. Inspection of the aliphatic signals proves that, while the extracts are dramatically enriched in polyphenols, a few other molecules were present in low (and variable) content. Notice the spectral complexity in the aromatic spectral region, whereby signals from several hundreds of protons resonate in a spectral span of about 1 ppm, which makes it hard to distinguish the nature of polyphenols between the sample batches.

Fig. 1
figure 1

1H (500 MHz) one-pulse NMR spectra of three different olive oil extracts in methanol-d. Thirty-two co-added transients were acquired in each case to an overall experimental time of 3 min. The aromatic and aldehydes spectral regions are magnified for a better visual appreciation

Full size image

Figures 2, 3, and 4 illustrate the full and expanded views of MQ (p = 3–5) spectra recorded for the three olive oils, along with the numbering for the identified spin systems. The indirect dimension was plotted with a scale in which the axis is reduced by the p-quantum order, for numerical homogeneity and the sake of an easier comparison along the pQ series (Reddy and Caldarelli 2010). In this scale, the singlet corresponding to the maximum-quantum order appears at the weighed average of signals of the corresponding 1Q trace (i.e., the sum of the chemical shifts of the resonance times their integral, divided by the total integral). This is best exemplified by the peak at 6.99 ppm in the 5Q spectra of extracts of batches I and III. The numbering corresponds to the annotation as reported in Table 1, and the chemical structures (simple phenols, lignans, flavonols, and secoiridoids) are shown in Fig. 5. A visual comparison of the spectra in Fig. 2 suggests, to a first approximation, some qualitative variation in the speciation of the polyphenols content, which will be discussed more in details below. Note that increased noise level in the indirect dimension in the 4Q spectrum of the batch I sample can be linked to solid microparticles suspended in the sample that could not be filtered out.

Fig. 2
figure 2

1H (500 MHz) one pulse NMR spectra followed by two-dimensional 5Q–1Q, 4Q–1Q, and 3Q–1Q correlation NMR spectra. For multiple-quantum experiments, 16 transients were co-added for each of 512 t 1 increments corresponding to a total experimental time of 16 h using a relaxation delay of 2 s. The spectra are displayed in magnitude mode, and the MaxQ dimension is plotted with the reduced scale described in the text

Full size image
Fig. 3
figure 3

Expanded regions of 1H 4Q–1Q correlation NMR spectra showing the annotation of some of the molecular components. Relatively low intense peaks are spread up to 2 ppm (left) and the signals with higher intensities are resonating in a narrow spectral region in less than half a ppm (right-hand side, spectral intensity is downscaled to a factor 16). All spectra are displayed in magnitude mode, and the MaxQ dimension is plotted with the reduced scale described in the text. See Table 1 for the details of the annotation and Fig. 5 for the corresponding molecular structures

Full size image
Fig. 4
figure 4

Expanded regions of 1H 3Q–1Q correlation NMR spectra, depicting the identification of MaxQ signals of trisubstituted polyphenols. All spectra are displayed in magnitude mode, and the MaxQ dimension is plotted with a reduced scale. See Table 1 for spectral annotation and Fig. 5 for molecular structures

Full size image
Table 1 MaxQ NMR chemical shifts together with 1Q chemical shifts for a direct comparison of the polyphenols content in three Italian olive oils (batches I, II, III) are listed
Full size table
Fig. 5
figure 5

Chemical structures of polyphenols identified in olive oil extracts; simple phenols (2, 5, 6, 10, 13, 16, 18, 20, and 23), lignans (21 and 22), flavonols (12 and 24), and secoiridoids (3, 4, 7, 8, 17, and 19). Green-color circles represent the 1H coupled-spin networks.

Full size image

5Q–1Q NMR Correlation Spectra

Signals in these spectra arise only from monosubstituted benzene rings. Figure 2 shows that only one 5Q signal (compound 1) was observed, which could not be assigned to a known phenolic molecule on the basis of existing databases. This molecule was detected in batches I and III, but not in batch II. Qualitatively, the evolution of the intensity of this peak reflects the observed total concentration in polyphenolics of the three batches.

4Q–1Q NMR Correlation Spectra

Line shapes in these spectra are of two kinds, both possessing at least four coupled protons: singlets (MaxQ signals) in the indirect, MQ dimension for disubstituted benzene rings, and multiplets (non-MaxQ signals) for monosubstituted rings. In general, but not in all cases, non-MaxQ signals do not carry new information on the sample composition, as the associated molecule could be more easily identified in MQ spectra of a higher order. Expanded regions of the 4Q–1Q spectra of three different sample batches are given in Fig. 3. The right-hand side panels are zoomed out to a factor 16 for a better visual representation of spectral resolution. Fourteen molecular components were identified in the 4Q–1Q correlation spectrum (see Fig. 5 for chemical structures). Due to the presence of magnetically equivalent protons, 1, 4-disubstituted six-membered aromatic molecules have a relatively better signal to noise ratio in their 4Q spectra as compared to 1,2 and 1,3 disubstituted analogues (Manjunatha Reddy and Caldarelli 2013). It has to be noted that the relative signal intensities are also sensitive to the choice of the 4Q excitation duration in the pulse sequence (Manjunatha Reddy and Caldarelli 2013). The spectral resolution obtained for the 4Q spectra allows to distinguish seven MaxQ signals in less than 0.04 ppm (25 Hz) in the 4Q dimension and about half a ppm in the 1Q dimension. In other words, about 28 protons resonating in two distinct multiplets were identified. Neglecting the effect of intermolecular interactions on chemical shifts in the polyphenolic mixtures studied here, we assigned NMR resonances in this crowded region to tyrosol (2 and 6) and ligstroside (3, 4, 7, and 8) derivatives based on their 1Q chemical shifts in NMR databases obtained in previous studies (Christophoridou and Dais 2009; Christophoridou et al. 2005). Indeed, these commonly observed polyphenols have very similar chemical structures (Fig. 5), so that their NMR signals resonate in a narrow spectral region (6.7 to 7.1 ppm). 4-Hydroxyphenylacetic acid (5) was the last molecule identified in this spectral region. Seven more molecular components were found at relatively weaker intensities (Fig. 3, left column), three of which could be assigned to p-coumaric acid (10), apigenin (12), and 4-hydroxybenzoic acid (13). The remaining four molecular components could not be determined on the basis of the literature. Finally, a non-MaxQ signal (i.e., a multiplet in the pQ dimension) was observed in the 4Q spectrum of batch II, not related to any resonance measured in the 5Q spectrum. Nonetheless, it was possible to identify this compound on the basis of its 1Q signals to t-cinnamic acid, a monosubstituted phenolic molecule. The absence of its signal in the 5Q spectrum is most likely due to a non-optimal choice of the sequence parameters.

3Q–1Q Correlation NMR Spectra

In analogy to what was explained for the previous spectra, these spectra encompass singlets, MaxQ signals from trisubstituted benzenes and multiplets from di- and monosubstituted benzenes. Expansions of the 3Q–1Q correlation spectra are depicted in Fig. 4. Here, together with the 3Q signals, correlations appear from molecular fragments, with four and five spin networks already detected in the form of non-MaxQ signals, and can be easily recognized. Seven more molecular fragments were identified in the 3Q–1Q correlation spectra, namely, oleuropein aglycon (17), hydroxytyrosol (18), aldehydic form of oleuropein (19), hydroxytyrosol acetate (20), (+)-pinoresinol (21), (+)-1-acetoxypinoresinol (22), caffeic acid (23), and luteolin (24). This corresponds to an assignment of 24 proton chemical shifts in eight different molecules. Table 1 summarizes the qualitative polyphenols profiles of the olive oils studied here. As previously mentioned, the overall polyphenols content but even more its profile in terms of actual molecular species can depend on many different factors. In the case of the investigated olive oils, in batch I, which contains the highest total polyphenols content, 21 compounds were observed whereas there were 19 in batch II and 17 in batch III which has the lowest total phenol content. More specifically, 13 compounds were common to all three samples (2–8, 10, 12, 15, 21, and 22). Indeed, as described in previous studies, p-coumaric acid, apigenin, lignans, tyrosol, and ligstroside derivatives are major components in fresh olive oils. Concerning the compounds more specific to each batch, 4-hydroxybenzoic acid is present in batch I only whereas oleuropein aglycon, hydroxytyrosol acetate and an unknown structure (1) are present in batches I and II. In addition, signals of t-cinnamic acid and luteolin were found in batch II only. Specifically absent from the batch II sample were caffeic acid and three unknown structures (9, 11 and 14).

It can be concluded that MaxQ NMR analyses provide a fine characterization of the polyphenol content in extra virgin olive oils (EVOOs), as exemplified in three sample olive oils from Italy. Focusing on the signal of the six-membered aromatic rings, a series of spectra ranging from 5Q to 3Q allowed similarities and differences in polyphenol content to be determined.

Conclusions

To summarize, MaxQ NMR is capable of providing information on the qualitative and semi-quantitative polyphenol profile for olive oils. This approach is easy to implement, reasonably fast, and sensitive. Most importantly, it does not require a physical separation of the polyphenols. It can be anticipated that the information obtained by this technique could be applied to the study of olive oil aging or adulteration, by monitoring temporal charges of specific polyphenols as a function of chemical and biological processes. MaxQ NMR databases could be easily built for the further expansion of applicability of this method to probe polyphenols in various other mixtures of natural origin. Thus, application to the analysis of the polyphenol content in other foodstuff is a natural extension of the method.