Greek Graviera Cheese Assessment through Elemental Metabolomics—Implications for Authentication, Safety and Nutrition

This study presents the comprehensive elemental profile of Greek Graviera (Gruyère) cheeses. In total, 105 samples from nine different geographic regions produced from sheep, goat and cow milk and their mixtures were assessed. Elemental signatures of 61 elements were investigated for determination of geographic origin and milk type. Regional and milk type classification through Linear Discriminant Analysis was successful for almost all cases, while a less optimistic cross validation exercise presented lower classification rates. That points to further research using a much larger sample set, increasing confidence for cheese authentication utilizing also bioinformatics tools under development. This is the first study reporting signatures of 61 elements in dairy products including all sixteen rare earth elements and all seven precious metals. Safety and quality were assessed regarding toxic and nutritive elements. According to both EU and USA regulations and directives, Graviera is a nutritional source for trace and macro elements with low levels of toxic elements.


Introduction
Food authentication's importance is increasing due to the consumers' interest in accurate food labeling, forcing producers and retailers to follow. Regulatory authorities are interested in analytical methods for food authenticity to support law enforcement [1,2]. Dairy products play a central role in a nutritious and balanced diet and their consumption has been associated with several health benefits due to their high contents of protein, essential fatty acids and minerals [3].
"Graviera" (Gruyère) is a hard type cheese, holding the second place, after "Feta", in the Greek dairy production volume. It is produced mainly from a mixture of sheep and goat milk although Gravieras can be found made of solely sheep, or goat or cow or a mixture of all three kinds of milk.
Most of the 700 Greek cheese-making dairies are small-medium size, collecting milk over a radius of about 30 km, although very few can collect milk over substantially longer distances. The majority of dairy sheep and goat farms are also of small to medium size, grazing their animals near the farm [4]. Most of the Gravieras are commercialized with a geographical denomination, but only three of them are registered under the Protected Designation of Origin (PDO) EU scheme, namely: "Graviera Agrafon",
Sixty-one elements were used as predictor variables to develop a method for assigning Graviera origin to nine geographic regions. The classification table (Table 5) shows that one sample from the Epirus region (sample 15, Table S1, from Arta) and another from Thessaly (sample 98, Table S1, from Larisa) were misclassified as coming from Central Greece. However, both Arta and Larisa are adjacent to the Central Greece region. That could provide an adequate explanation for the misclassification as goats and sheep movement between adjacent regions could modify their elemental content and subsequently the milk they produce. However, after further investigation we found that the two dairies in Arta and Larisa, to fulfill their needs, purchased milk from Central Greece, Amfiloxia and Lamia, respectively. The third misclassified sample was from Macedonia region (sample 45, Table S1, from Grevena, prepared from cow milk) classified as South Aegean. Cows in the Macedonia region obtain a large portion of their feed from grass that grows locally that is rich in REEs and aluminum [27] (aluminum ores are usually accompanied by REEs [28]). This is reflected in the content of REEs and aluminum found in Gravieras produced in that region. Sample 45 shows less than 50% REEs and aluminum content compared to all other samples from this region. This points to a different feeding scheme using mostly imported feeds as in the South Aegean. Through a cross validation exercise using the leave-one-out approach, the above results are quite optimistic and the classification rate was just 32.7%. This warrants further research enhancing the sample bank with much more samples resulting from different production periods.
Elemental metabolomics has potential for detecting production method (feed with pasture vs imported/dried feeds). This needs further research with feeding experiments. A useful aspect of elemental metabolomics applied to dairy products could be a bioinformatics tool to detect the feeding scheme utilizing soil composition analysis. Details on the tool/algorithm can be extracted from the discussion above on the Macedonia region's cows and soil.
The most significant predictor variables are the rare earths Ce, Er, Eu, Ho, La, Sm, Tm, Yb, the actinide Th, the precious metals Pt, Re, Ru, the ultra trace elements Hf, Nb, Sb, W, the trace elements Ag, Al, As, B, Ba, Cd, Co, Cr, Cs, Fe, Ga, Hf, Mo, Nb, Ni, Pb, Sb, Se, Sr, V, W, Zn and the macro elements Ca, Mg and P The classification was more successful using the comprehensive elemental signature as proposed by elemental metabolomics [19]. This result on cheeses is in contrast to previous studies on authentication of game meat [25], wines [24] and split-peas [23], where specific groups of elements such as REEs were sufficient.
Levels of REEs were higher in Crete cheeses, probably reflecting the vegetation and soil composition [21]. This is most pronounced for the light REEs (LREEs) Pr and Nd (Table 1). This is in accordance with previous findings that Crete is enriched in LREEs, due to monazite and allanite ores [29]. It is interesting to note that another couple of LREEs, Eu & Sm, were enriched, by three times, in cheese from Central Greece, pointing to further authentication markers. These findings about Eu and Sm need further research such as a check of soil composition differences and the influence of different flora grown there.
Usually all rare earths in different materials follow the same pattern, i.e., they are all enriched or depleted: fava Santorini's [26]; Italian milk [30]; mushrooms substrates [31]. However, this pattern is differentiated by genetic factors as seen in two different mushroom species [31]. The production method could also differentiate the pattern as seen in game and farmedrabbits [25].
North Aegean cheeses showed much higher levels of Rb, Cs (alkali metals) and Sr, Ba (alkaline earth metals) in agreement with previous studies [9][10][11][12]15,18,[32][33][34], where alkali and alkaline earth metals, especially Rb, Cs, Sr and Ba, were proven reliable cheese authenticity markers. Another interesting result shown in Table 2 is that the precious metals Au, Pd & Ru were found in higher amounts in Thessaly's cheeses. Geological data [35] explain the increased content of precious metals transported from the Pindus mountain range by the Pinios River to Thessaly. This is in accordance with the view that precious metals are potential authenticity markers [22].     Our data are in line with previous findings. Camin et al. also stated [12] that levels of Cu, Mo, Ni, Fe, Mn, Ga and Se showed significant differences between grated hard cheeses. Osorio et al. [15] found that Ag, Ba, Ca, K, Mg, Mn, P and Sr presented different profiles for different Halloumi cheese production locations and also highlighted the potential of Sr for traceability information from soil as it cannot be added from cheese making equipment. All elements commended by Osorio in addition to Ag presented significant differences between regions in our work. Korenovska and Suhaj [11], working with Slovakian, Polish, and Romanian Bryndza cheeses, found also that Cr, Hg, Mn and V along alkali and alkaline earth metals were the best elemental indicators.
In agreement with our results, Pillonel et al. [18] working with Emmental cheese support the view that the elemental profile allows the discrimination of close regions of production, where the distances are in the order of a few tenths of kilometers up to 150 km. It should be noted that the distance range for adjacent regions in our study is between 10 to 150 km. This is the first study reporting precious metals, rare earth and ultratrace element assessment for dairy product authentication.

Milk Type
The Gravieras used in this study were manufactured from mixed sheep and goat milk (78), sheep (10), goat (8) and cow (8) and one from sheep, goat and cow milk. The comprehensive elemental fingerprint (Table S2) of 61 elements was used for the classification according to the milk type. Best markers of milk type were: Bi, Cr, Fe, Mn, Ni, Se, Sr, Zn, Mg and P (Table S2). The classification table (Table 6), shows that all 78 cheeses from sheep and goat milk were correctly categorized. Our data did not contain information on the % percentage of goat milk used. This explains the misclassification of two out of the ten sheep milk Graviera samples into the sheep and goat class. This group is not so well defined and homogenous as even the same producer uses different percentages of goat milk, according to its availability. Only one cow Graviera sample was misclassified into the sheep + goat group. This Graviera was from the Macedonia region (sample 41, Table S1, from Grevena). Through a cross validation exercise using the leave-one-out approach, the above results are quite optimistic and the classification rate was only 50.5%. Previous attempts to classify cheeses according to the milk type, in comparison to our study, were restricted to small elemental fingerprints. Fresno et al. [34] were the first to report differences in P, K, Mg, Zn, Fe and Mn, concerning various ripened and unripe Spanish cheeses. Necemer et al. [32] found that, for Slovenian cheeses, the best milk-type indicators were Ca, Br, Zn and Sr. Our study is the first concerning the determination of cheese milk type using elemental metabolomics.

Contribution to Total Diet-Safety Aspects and Nutritional Value
Regarding toxic elements such as, Cd, Pb, Sn and Sb, the examined samples presented low values for most of them. In more detail, Pb levels were determined to range from 19.8 µg kg −1 (North Aegean) to 38.3 µg kg −1 (Central Greece), i.e., similar to the Khozam et al. study [36] on Lebanese cheese (32.4 µg kg −1 wet weight), but noticeably lower than that reported by Vural et al. study [37] for south-eastern Anatolia-Turkey cheese (4600-7700 µg kg −1 wet weight), the Lante et al. study [38] for Crescenza and Squacquerone cheeses (600 µg kg −1 fresh weight) and the Mendil et al. study [39] for Turkish cheeses (110-960 µg kg −1 wet weight).
The Food and Agriculture Organization/World Health Organization Joint Expert Committee on Food Additives (JECFA) has established a Provisional Tolerable Weekly Intake (PTWI) for several toxic elements and especially heavy metals. According to the Hellenic Statistical Authority, the daily intake of cheese in the Greek population is 94.7 g/person. For an adult person (e.g., a man of 75 kg body weight), the percentage intake of each toxic element is provided in Table 7. The highest intake is observed for As (12.8%) while the lowest is for Sn (0.0007%), reflecting the high allowed limit for Sn. Column 4 shows the % intake according to the Reasonable Daily Intake of cheese based on the Canadian Food Inspection Agency, 57 g cheese consumption per day. Here, it must be mentioned that reasonable intake has been estimated considering the food habits of Canadians. As regards for these calculations, a mean value of all analyzed Graviera cheese samples was taken into account for each element. Concerning nutrition, trace amounts of Fe, Mn, Mo, Zn, Co, Ni, Cr, Se, Cu, Si, I, and F are necessary for proper human health, apart from H, C, N, O, Na, K, S, Cl, Mg, Ca, and P which are required in relatively large quantities in a diet. There is also a group of elements called ultra-trace minerals, including V, Sn, Ni, As, and B, that are being investigated for possible biological function but currently do not have clearly defined biochemical roles [19]. Thus, in order to prevent nutrient deficiencies, but also to reduce the risk of chronic diseases such as osteoporosis, cancer and cardiovascular disease, scientific food committees around the world have established specific limits for each element intake with values adapted to different population groups (children, adolescents, pregnant women or older people). The European Commission has established Nutritive Reference Values for adults, according to Regulation (EU) No 1169/2011 (25 October 2011) that are presented in Table 8. WHO/FAO and USDA (United States Department of Agriculture) have also established Recommended Dietary Allowances (RDA) indicating the amount of an individual nutrient that people need for good health depending on their age and gender. As shown in Table 8, % Ca intake from Graviera was sufficient and ranged from 54% to 111%, while P ranged from 56% to 94%. Moreover, the Ca-P ratio was 1.4:1, so consumption of Graviera cheese is one the most convenient ways for proper intake of both minerals through the diet. High dietary Ca-P ratios play important role in bone health [41]. The % zinc intake ranged from 16% to 31%. Iron, Cr and Mo % intakes ranged from 6.2% to 35%, 74% to 185% and 12% to 22%, respectively. Regarding Mg, Cu, Mn and Se the % intake is less significant. These results highlight Graviera cheese as good source of trace and macro elements, especially for Ca, P, Zn, Cr, Fe and Mo.
Compared with other studies like the Moreno-Rojas et al. study of different cheese types [10], nutritive elements were found at similar levels. Camin et al. [12] found lower Fe and higher Se and Mo in various European hard cheeses such as PDO Parmigiano Reggiano. Suhaj et al. [33] determined Cr in lower levels, Mo and Ca in slightly lower levels and Mn in slightly higher levels in some European Emmental and Edam hard cheeses than the Graviera samples in our study.

Instrumentation and Reagents
Chemicals used were nitric acid (Suprapur ® , 65% w/v, Merck, Darmstadt, Germany), hydrogen peroxide (Suprapur ® , 30% w/v, Merck, Darmstadt, Germany), ICP internal standards of Ge and In and ICP-MS certified multi-element standards (all from Inorganic Ventures, NJ, USA). Ultrapure water with a resistance of 18.2 MΩ cm −1 obtained from a MilliQ plus system (Millipore, Saint Quentin Yvelines, France) was used in all procedures.
Elemental content was determined using a Perkin Elmer (SCIEX, Toronto, ON, Canada) 9000 Series ICP-MS. Inductively coupled plasma mass spectroscopy is predominantly used in authentication studies due to its capability for rapid ultra-trace level multi-element determinations [42].

Sample Collection, Preparation and Digestion
One hundred and five Graviera cheese samples were used for the purposes of this study. The geographical origin of the samples is reported in Table S1 and depicted in the map shown in Figure 1. Most of the samples were collected from small-medium dairies and the rest from the respective local markets. The sampling strategy excluded large dairies that are able to collect milk from different Greek regions and bulk it in their premises for the production of their own trade mark. Samples were taken from a lot and after grinding, they were preserved in a freezer (−32 • C) before analysis.
Sample digestion was performed with a microwave-assisted digestion system (CEM, Mars X-Press, Matthews, NC, USA). Approximately 0.50 g of cheese was weighted in an analytical balance in a polypropylene tube. Then, 4.0 mL of HNO 3 was added to pre-digest samples for 30 min. The resulting cheese suspension was transferred quantitatively, with the use of 4.0 mL HNO 3 and 2.0 mL H 2 O 2 to the microwave digestion PTFE vessel. The samples were heated in the microwave accelerated digestion system according the following program: the power was ramped during 20 min from 100 to 1200 W and held for 15 min. The temperature reached a maximum of 200 • C and followed by a cool-down cycle for 15 min. PTFE vessels were sealed throughout the aforementioned cycle to avoid volatilization losses. Although all samples were completely brought to solution, to disregard any small particle passing optical inspection entering the ICP-MS, solutions were filtered with polyester disposable syringe filters 0.20 µm/ 15 mm (Chromafil, Macherey-Nagel, Düren, Germany). Before injection in the ICP-MS, sample solutions were diluted, as required, with ultrapure water.
Molecules 2019, 24, Firstpage-Lastpage; doi: FOR PEER REVIEW www.mdpi.com/journal/molecules geographical origin of the samples is reported in Table S1 and depicted in the map shown in Figure  1. Most of the samples were collected from small-medium dairies and the rest from the respective local markets. The sampling strategy excluded large dairies that are able to collect milk from different Greek regions and bulk it in their premises for the production of their own trade mark. Samples were taken from a lot and after grinding, they were preserved in a freezer (−32 °C) before analysis.

ICP-MS Analysis
The studied elements assessed were: Limits of quantification for all were lower than those determined in the samples (Table S3). Operating conditions of the ICP-MS were as follows: nebulizer gas flow of 0.75 L min −1 , ICP RF power of 950 W, lens voltage of 7 V, pulse stage voltage of 950 V and sample uptake rate of 26 rpm. Calibration curves ranges were from 1 ng kg −1 to 1000 µg kg −1 for rare earths, precious metals and ultra-trace elements, while from 0.01 µg kg −1 to 10 mg kg −1 for trace and macro elements. Indium was used as internal standard for rare earths, precious and ultra-trace elements, while germanium was used for trace and macro elements. In detail, the daily analytical procedure is: Steps to be repeated after 4 h: Standard reference materials

Calibration and Quality Assurance
To assess the accuracy of the process the following standard reference materials were obtained from the European Commission, Joint Research Center, institute for reference materials and measurements IRMM, Belgium and the National Institute of Standards & Technology (NIST), USA:
The standard reference materials were subjected the same analytical process: Digested three different times, each digestate measured in triplicate (Table 9). Recoveries were in the range 67-121% for all elements other than Se. In order to overcome Ar 2+ interferences we measured Se 82.

Statistical Analysis
Statistical analysis was performed using SPSS software (IBM, Armonk, NY, USA) for the descriptive statistics and cross validation and Statgraphics Centurion XV software (Statpoint technologies, Warrenton, VA, USA) in order to analyze the data using statistical models and predictive analyses (Linear Discriminant Analysis).

Conclusions
We present results from 61 elements in cheese for the first time with implications in food authentication, safety and nutrition. Further work is in progress for data selection, increasing confidence for food authentication using bioinformatics tools under development. This is the first study reporting signatures of 61 elements including rare earth elements and all the precious metals in cheese. We highlight the application of elemental metabolomics to human nutrition assessing both nutritive and toxic elements. The results demonstrate that elemental metabolomics could be potentially used for discrimination of cheeses produced in different geographical zones and milk type. The method needs further improvement by bioinformatics tools to automate data cleaning done manually.
In comparison to molecular analysis, elemental metabolomics is simple and convenient. The first step is accurately weighing samples in capped polypropylene tubes to analyze when convenient, when adequate samples are collected and when instrumentation is available. There are no requirements concerning temperature, time, or any other storage condition. The only requirement is creation of comprehensive elemental metabolome databases for food authentication, quality and safety. Elemental metabolomics are becoming more affordable by lowering the ICP-MS purchasing cost and increasing capabilities concerning interferences [19]. We envisage open access elemental databases for improvement of human nutrition and health.
Supplementary Materials: The following are available online, Table S1: Sample description, Table S2: Mean value and SEM, Standard Error of the Mean (number of samples) of the elements for all milk types. The results are expressed in µg kg −1 except for the macro elements, which are expressed in g kg −1 . Table S3. Mass of quantification, limits of Detection (LoD), limits of Quantification (LoQ) (µg kg −1 ) and coefficient of determination.