Monitoring of pistachio (Pistacia Vera) ripening by high field nuclear magnetic resonance spectroscopy

Abstract The metabolic profiling of pistachio (Pistacia vera) aqueous extracts from two different cultivars, namely ‘Bianca’ and ‘Gloria’, was monitored over the months from May to September employing high field NMR spectroscopy. A large number of water-soluble metabolites were assigned by means of 1D and 2D NMR experiments. The change in the metabolic profiles monitored over time allowed the pistachio development to be investigated. Specific temporal trends of amino acids, sugars, organic acids and other metabolites were observed and analysed by multivariate Partial Least Squares (PLS) analysis. Statistical analysis showed that while in the period from May to September there were few differences between the two cultivars, the ripening rate was different.


Introduction
Pistachio tree (Pistacia vera L.) belongs to the Anacardiaceae family native of arid zones of Central and West Asia distributed throughout the Mediterranean basin. Its dry fruit is very popular and there are consistent evidences from epidemiologic and clinical studies of the beneficial effects of its consumption on risk of coronary heart disease, diabetes in women and on major and emerging cardiovascular risk factors (Ros 2010). Due to its commercial importance, several studies were carried out in order to better characterise this aliment ; Barreca et al. 2016).
Fruit ripening stage is a parameter of paramount importance for foodstuff processing since fruit molecular composition is heavily influenced by maturity. In fact, mature fruits show several differences compared to unripe ones, such as higher amounts of free carbohydrates and lower amounts of amino acids and organic acids (Capitani et al. 2010;; therefore, the harvest time could be adjusted taking into account the potential effect on consumer health of the aforementioned components. While this process is well documented for fresh fruits, ripening of shelled fruits is not as well studied. In fact, only in recent years (Zarei et al. 2014), the ripening of pistachio was studied, but the observed parameters were mainly physical ones (length, weight, etc.), total phenol amount and antioxidant activity. Due to the lack of a more detailed description of the metabolic events occurring, a deepened study of the changes in chemical composition during ripening is required.
Nuclear magnetic resonance (NMR) spectroscopy is one of the more reliable tools for this task. This technique is able to provide the identification and quantification of several molecular classes simultaneously even in complex matrices Consonni et al. 2013;Brasili et al. 2014;Praticò et al. 2014;, Diblasi et al. 2015.
In the present study, metabolic profile of pistachio aqueous extract was monitored over the period ranging from May to September by high resolution NMR spectroscopy. Several water-soluble metabolites were identified by means of 1D and 2D NMR experiments and their quantitative changes allowed to investigate the pistachio ripening. This study focused on pistachios belonging to two cultivars, 'Bianca' and 'Gloria' , grown in the area of Agrigento in Sicily (Italy). Since fruit ripening is influenced by many factors such as genetic factors, growth conditions, climate and soil composition, two cultivars with the same geographical origin were considered in order to correlate the observed differences only with the genetic factor.

Metabolic profile
The characterisation of pistachio metabolic profile as a function of ripening was carried out and the hydroalcoholic extracts were analysed by 1 H NMR spectroscopy (Figure 1 and Figure  S1). Resonance assignment and molecule identification were carried out by means of chemical shift values, signal multiplicity and area, as well as 2D homonuclear NMR experiment TOCSY and 2D heteronuclear experiments HSQC and HMBC (Figures S2,S3 and S4 respectively).
In addition to the aforementioned molecules, malic acid (MA) is present in pistachios in August and September, while in the last month of observation the presence of several other molecules was detected: ethyl-lactate (Et-LA), acetic acid (AA), citric acid (CA), raffinose (R), stachiose (St), melibiose (Mb), threalose (Tr) and trigonelline (Trig).
Thirty-four metabolites were identified, quantified and the assignments of their NMR resonances are reported in Table S1. Quantification of the metabolites was performed by signal integration and quantities were expressed in mg/g of nut by comparison of the relative integrals of their diagnostic resonances (Table S1) with the reference concentration and normalisation for the number of protons, while the quantitative analysis and the comparison between the two cultivars are reported in Figure S5.

Statistical analysis
In order to evaluate the presence of a common trend for pistachio ripening between the two cultivars, a Partial Least Square (PLS) analysis as a function of time was performed on each cultivar separately taking into account only those metabolites observable in every month. This kind of multivariate statistical analysis calculates Principal Components (PC) by means of a linear combination of the variables (i.e. the metabolites) in order to align the first of these components along the maximum variation of the parameter of interest (i.e. the  ripening time). The results are reported in Figures 2 and 3 for 'Bianca' and 'Gloria' cultivars, respectively. For both cultivars, a three component model was developed (R 2 = 099 and 096 for 'Bianca' and 'Gloria') showing a distinct separation of samples as a function of the time.
In greater detail, in both cultivars PC1 is associated with the evolution from May towards September while samples from June towards September are characterised by progressively higher values of PC2. The strong differentiation between May and the successive months could be attributed to the fact that in this stage the hull and the kernel of pistachios are not clearly discernible. This means that most of the metabolites observed in their spectra would be employed during the ripening process in order to fully develop the hull.
One difference between the two cultivars can be observed by analysing the grouping of the samples relative to the months of July, August and September. In particular, in 'Bianca' cultivar, the samples from June and July are closer between them than the analogues samples of 'Gloria' . Moreover, pistachios in August are closer to September ones in 'Bianca' cultivar, while in 'Gloria' samples the differences are more evident. It is important to remember that the amount of metabolites was expressed as mg/g of pistachio weight, and therefore does not take into account the total weight of the fruit, which in September is between three and four times the one in August.
A further information regarding the changes in the metabolic profiles as a function of ripening can be obtained from the analysis of PLS regression coefficients (left side of Figures  2 and 3 for 'Bianca' and 'Gloria' respectively). These coefficients, which take into account both PC1 and PC2, indicate the variables more important for the evolution of pistachios from May to September. Many molecules show the same trend for both cultivars: Ala and S increase while QA, ShA, G, Tyr and GAes decrease. These data are in good agreement with previous results for fresh fruits (Capitani et al. 2010), where a decrease of simple carbohydrates in favour of di-and trisaccharides is observed during ripening. Ala levels are also higher in more mature seeds since this amino acid is employed by the seedling as a ready source of nitrogen (Algéus 1949). Moreove,r molecules such as QA, ShA, GAes and aromatic amino acids are involved in the defence against pathogens, parasites and herbivores and are more abundant in the earlier stages of fruit development (Prusky et al. 1982). The fact that other molecules show a trend which is different between the two cultivars, such as branched amino acids, arginine and choline (statistically significant only for Bianca) could be ascribed to the intrinsic differences in the metabolic network of the two varieties.

Sample collection
Pistachios belonging to 'Bianca' and 'Gloria' cultivars were provided by the CRA-FRu-Centro di Ricerca per la Frutticoltura -Ministero Italiano per le Politiche Agricole Alimentari e Forestali (MIPAAF). The trees were 20 years old P. vera grafted on P. terebinthus and freely pollinated by P. vera males. The orchard was located in Sicily (Racalmuto, AG). Homogeneous grown condition was adopted for every sample. Nuts were collected monthly from the same tree from May to September (5 time points).

Sample preparation
Samples of four pistachio trees from both cultivars per each harvest time (a total of 8 nuts each month) were weighed and then extracted following a modified Bligh-Dyer protocol (Miccheli et al. 1991). Each kernel was grinded under liquid nitrogen and extracted by a cold mixture composed by chloroform, methanol and water in 2:2:1 proportion. Samples were stirred, kept at 277 K for one hour and then centrifuged for 20 min at 10,000-g at 277 K. The upper hydrosoluble phase and the lower lipophylic phase were carefully separated and dried. The hydroalcoholic phase was then resuspended in 600 μL of D 2 O containing 3-(trimethylsilyl)-propionic-2,2,3,3,-d 4 acid sodium salt (TSP, 2 mM) as internal chemical shift and concentration standard. All solvents and standard were purchased from Sigma Aldrich (St. Louis, u.S.A).

NMR experiments
Monodimensional 1 H NMR spectra were performed by collecting 64 scans for each sample on a Bruker Avance III 400 spectrometer (Bruker Spectrospin, Karlsruhe, Germany) operating at 9.4 T at 298 K acquiring the FIDs into 64 K points, the spectral width was 15 ppm. Presaturation was used to suppress the solvent resonance and the relaxation delay was set to achieve a 15 s total acquisition time.
TOCSY experiments were acquired with spectral width of 15 ppm, in both dimensions, a data matrix of 8 K × 256 points, mixing time of 120 ms and relaxation delay of 2 s. HSQC experiments were performed with a spectral width of 15 and 200 ppm for the proton and carbon, respectively, a data matrix of 8 K × 256 points, an average 1 J C-H of 145 Hz and recycle delay of 2 s. HMBC spectra were acquired with a spectral width of 15 and 250 ppm for the proton and carbon, respectively, a data matrix of 8 K × 256 points, long-range constant n J C-H of 8 Hz and recycle delay of 2 s.
Quantification of the metabolites was performed by signal integration and comparison with the reference integral at a known concentration. Due to the overcrowding of 1 H NMR spectra, only those signals which were not overlapped with other resonances were considered for integration.

Statistical analysis
Multivariate data analyses were performed with the unscrambler ver. 9.8 software (Camo Software AS, Oslo, Norway). The obtained quantities were collected in a data matrix where each row represented a sample and each column represented a metabolite. In order to avoid the effect due to the wide differences among the data, the variables were mean centred and divided by their standard deviations. PLS analysis was applied (Giuliani et al. 2004) and the results were validated through full cross-validation (leave-one-out) procedure and uncertainty test (Wold et al. 2009) that couples full cross validation to the Jack-knife principle (Efron 1987).

Conclusions
Pistachios from two Italian cultivars ailing from the renown production area of Agrigento (Sicily) were extracted according to a modified Bligh-Dyer protocol during the ripening process, and the NMR-based analysis of the hydroalcoholic extracts enabled to characterise their metabolic profile. Multivariate analysis shows a clear difference of the ripening processes occurring on August when the 'Gloria' cultivar samples show a spectral pattern similar to the September ones, while 'Bianca' metabolic profiles are still comparable with the ones observed in the month of June. This experimental evidence is consistent with agronomic knowledge according to which Gloria cultivar is an early one.