Determining the Authenticity of Spirulina Dietary Supplements Based on Stable Isotope and Elemental Composition

While the demand for Spirulina dietary supplements continues to grow, product inspection in terms of authenticity and safety remains limited. This study used the stable isotope ratios of light elements (C, N, S, H, and O) and the elemental composition to characterize Spirulina dietary supplements available on the Slovenian market. Forty-six samples were labelled as originating from the EU (1), non-EU (6), Hawaii (2), Italy (2), Japan (1), Portugal (2), Taiwan (3), India (4), and China (16), and nine products were without a declared origin. Stable isotope ratio median values were –23.9‰ (–26.0 to –21.8‰) for δ13C, 4.80‰ (1.30–8.02‰) for δ15N, 11.0‰ (6.79–12.7‰) for δ34S, –173‰ (– 190 to –158‰) for δ2H, and 17.2‰ (15.8–18.8‰) for δ18O. Multivariate statistical analyses achieved a reliable differentiation of Hawaiian, Italian, and Portuguese (100%) samples and a good separation of Chinese samples, while the separation of Indian and Taiwanese samples was less successful, but still notable. The study showed that differences in isotopic and elemental composition are indicative of sample origins, cultivation and processing methods, and environmental conditions such that, when combined, they provide a promising tool for determining the authenticity of Spirulina products.


Introduction
As demand for high-quality food supplements continues to grow, combined with a greater awareness of the importance of food quality and safety, consumers are prioritizing products with declared composition and geographical origin [1][2][3][4][5][6]. There is a limited understanding of nutritional composition across microalgal species, geographical regions, and seasons, all of which can substantially affect quality and safety value of products based on microalgae, which are available on the market, mostly in the form of nutritional supplements [7]. Interest in Spirulina dietary supplements is also proliferating due to its accepted nutritional properties, such as high protein, mineral, vitamin, pigment, and other beneficial phytochemicals content [5,8,9]. The high proportion of adulterated algal food products discovered in the past (more than 50%), resulting partly from the lack of quality control measures, reveals the severity of the fraud, which cheats the consumer and poses a potential health risk [10,11].
Microalgal products that are subjected to variations in quality due to unstable culturing conditions are likely to be deliberately adulterated [12]. Common adulterants in microalgae products are flour and mungbean powder, which contain significantly less protein compared to Spirulina and Chlorella and have lower production costs [5]. Environmental conditions such as climate change and variability that influence algal growth may affect potential biomarkers, but also the stable isotopic composition of sulfur, carbon, and nitrogen. These parameters could be used to verify the origin of microalgae and to and to raise awareness about quality and origin of these products among consumers. Thus, the objective of this study was to verify the quality and origin of all Spirulina dietary supplements sold commercially on the Slovenian market with combining carbon, nitrogen, sulfur and, for the first time, hydrogen and oxygen stable isotope ratios ( 13 C/ 12 C, 15 N/ 14 N, 34 S/ 32 S, 2 H/ 1 H, 18 O/ 16 O), with elemental composition, to verify the type of production and country of origin. To our knowledge, this is the first time that such an approach has been used to determine the authenticity of Spirulina products.

Sample Collection and Preparation
An attempt was made to collect all available Spirulina products on the Slovenian market. In total, 46 samples of Spirulina food supplements were gathered from physical and online stores over two months in 2018. The majority of samples (44 samples) contained only Spirulina spp., while two were mixed samples also containing other plant material (wheat grass and barley grass) or algae (Chlorella, Lithothamnium). Of the samples containing only Spirulina as an active ingredient, the majority (34 samples) were labelled as pure, and ten were declared to contain excipients. The samples were labelled as originating from Hawaii (n = 2), Italy (n = 2), Japan (n = 1), Portugal (n = 2), Taiwan (n = 3), India (n = 4), European Union (EU) (n = 1), non-EU (n = 6), China (n = 16), or were without declared origin (NS; n = 9). The samples were sold either dried in tablet, capsule, or powder form, or fresh (Table 1). No additional information was available regarding production and processing. Sample preparation included opening the capsules to extract the dry material, grinding the tablets to powder and, in the case of fresh samples, freeze-drying the contents followed by grinding to a powder. Once in powder form, all the samples were kept in sealed plastic containers in the dark at 4 • C.

Stable Isotope Ratio Analysis of Light Elements Using Isotope Ratio Mass Spectrometry
Measurements of the stable isotope ratios of light elements ( 2 H/ 1 H, 13 C/ 12 C, 15 N/ 14 N, 18 O/ 16 O, 34 S/ 32 S) were performed using Isotope Ratio Mass Spectrometry (IRMS) and are expressed in the δ-notation in ‰ according to Equation (1) [33]: where i stands for the highest, j stands for the lowest atomic mass number of the element E (H, C, N, O, S), and R is the isotope ratio between the heavier and the lighter isotope of the element ( 2 H/ 1 H, 13  For the stable isotope ratio analysis of light elements C, N, and S, powdered samples (4 mg) and tungsten oxide (WO 3 ) (4 mg) were weighed directly into tin capsules which were then sealed and placed into the autosampler of the elemental analyzer. The samples were prepared and analyzed in triplicate. Finally, the mean values were used. The 13 C/ 12 C, 15 [34]). The measurements' analytical precision was ±1‰ for δ 2 H and ±0.2‰ for δ 18 O.

Macro-Elemental Composition Analysis by X-ray Fluorescence Spectrometry
Analysis of Spirulina samples' macro-elemental composition was performed nondestructively by Energy Dispersive X-Ray Fluorescence Spectrometry (EDXRF) to determine the following elements (13): phosphorous (P), titanium (Ti), zinc (Zn), silicon (Si), bromine (Br), sulfur (S), chlorine (Cl), manganese (Mn), rubidium (Rb), strontium (Sr), potassium (K), calcium (Ca), and iron (Fe). Powdered samples were pressed into 0.5-1.0 g pellets for analysis using a pellet die and a hydraulic press. For fluorescence excitation disc radioisotope, excitation sources Cd-109 (20 mCi, Eckert and Ziegler, Berlin, Germany) and Fe-55 (25 mCi, Eckert and Ziegler, Berlin, Germany) were used. An EDXRF spectrometer with a PX5 digital pulse processor (Amptek, Bedford, MA, USA), an XR-100 SDD detector (Amptek, Bedford, MA, USA), and a PC-based, multichannel analyzer software package (DPPMCA) were used for the emitted fluorescence radiation detection. For light element analysis (Si, P, S, and Cl), the spectrometer operating in Fe-55 mode was equipped with a vacuum chamber, and for K, Ca, Ti, Mn, Fe, Zn, Br, Rb, and Sr analysis, measurements in Cd-109 mode were performed in the air. The energy resolution of the spectrometer was 125 eV at 5.9 keV. AXIL Spectral Analysis software was used to analyze the complex X-ray spectra. For quantification, the Quantitative Analysis of Environmental Samples (QAES) software developed in our laboratory was used [35,36]. Method validation was performed using 1573a (tomato leaves) and 1547 (peach leaves) NIST standard reference materials. The EDXRF analysis estimated uncertainty budget was 11% and was incorporated in the QAES software procedure.

Statistical Analysis
XLSTAT software (Addinsoft, Long Island, NY, USA, 2019) and SIMCA-P (version 17, Sartorius Stedim Biotech, Umeå, Sweden) were used for statistical analysis. Following basic statistical methods (maximum, minimum, median, and quartiles), multivariate statistical analyses methods, including Principal Component Analysis (PCA), Discriminant Analysis (DA), and Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA), were applied to identify further characteristic parameters for discrimination of samples based on their elemental and isotopic composition. Internal sevenfold cross-validation was used to determine the significant components of the models and thus minimize overfitting. The OPLS-DA study evaluated performance using the explained variation (R 2 X for PCA and R 2 Y for OPLS-DA) and predictive ability (Q 2 ). The OPLS-DA model prediction performance was also evaluated via specificity (true negatives) and sensitivity (true positives), calculated as described by Fiamegos et al. [37]. The accuracy (TP/TN)/(TP + FP + FN + TN) and F1 Score (F1 Score = 2 × (Recall × Precision) / Recall + Precision) of the model were calculated as described by Strojnik et al. [38]. Candidates for discriminant markers were selected by loading plots, which allow for visualization of the relationships between the formed groups and the variables and by the variable importance in the projection (VIP) values of the OPLS-DA models, where a value higher than one was considered the threshold.

Isotopic Composition of Spirulina Food Supplements from the Slovenian Market
While the detailed elemental composition of Spirulina dietary supplements from a nutritional point of view has already been presented [9], Spirulina stable isotopic profiles, including hydrogen and oxygen isotopic composition, were characterized here for the first time and, together with their macro-elemental composition, were used to verify the country of origin and authenticity of the samples. In the text, the data are presented as the median value (M) and interquartile range (IR, in parentheses) of elemental or isotopic composition. Results of the elemental composition are presented in Table S1, while stable isotope ratios of light elements C, N, S, O, and H (‰) are collected in Table 2. The content of macro-elements (>1 g/kg) in the commercial Spirulina supplements was as follows: K > P > S > Si > Cl > Ca, and of micro-elements (>1 mg/kg): Fe > Mn > Sr > Zn > Ti > Br > Rb. Among the macro-elements, potassium values in Spirulina samples ranged from 5.83 to 26.9 g/kg with a median value of 15.  (Table 1), which has δ 13 C values similar to C 4 plants (−17 to −9‰), while measured values in Spirulina samples are closer to those for C 3 plants (−40‰ to −20‰) [39]. The same could be assumed for the sample S33, where the presence of undeclared excipient with a higher δ 13 C values (such as corn maltodextrin) could explain the high measured δ 13 C value of -16.7‰. West et al. [17] found in their research that carbon values higher than −32‰ and lower than −29‰ are characteristic of crops grown in the shade or indoors, and crops with δ 13 C values higher than −29‰ were identified as grown outdoors. However, the classification of crops grown outdoors could also include crops grown indoors when good ventilation in the indoor environment was included. Therefore, the classification of an outdoor-grown crop includes open-grown crops cultivated both outside or inside a structure. Sample separation according to δ 13 C values is less reliable in our case, as the producers use several different synthetic or organic products to enrich the Spirulina growth medium. Therefore, the final products do not reflect the actual δ 13 C isotopic composition of the environment and microalgae, but we may assume that algae grown indoor would have lower δ 13 C values than −28‰. Our case samples were S1, S7, S34, S38, S40, S44, and S46.
The δ 15 N values spanned a broad range of values, i.e., -5.35 to 13.8‰ (M: 4.80‰, IR: 1.30-8.02‰), where the lowest values were again measured in Italian samples (S46, S44) and the highest in Hawaiian (S4, S26), Japanese (S1), and samples without declared origin (S37). A high variability in δ 15 N values can be attributed to using organic (manure, wastewater) and inorganic (synthetic) fertilizers. Since the nitrogen source in synthetic fertilizers is atmospheric N 2 , their δ 15 N value is around 0‰. In organic fertilizers, the δ 15 N values are higher, since they are primarily derived from animal waste [40]. Consequently, the crops fertilized with synthetic fertilizers obtain lower δ 15 N values than those fertilized with organic fertilizers. Additionally, a mixture of different fertilizers could be used, resulting in a relatively ambiguous nitrogen isotopic composition. Finally, variability in crop δ 15 N value could also result from using different amounts of fertilizer [17,40]. The field and laboratory study performed on algae also indicates that higher δ 15 N values (up to 11.1‰) are observed in algae exposed to organic manure compared to those exposed to synthetic inorganic fertilizers [41]. Another explanation for high δ 15 N values could be the use of a pool of NH 4 + enriched in 15 N. For instance, in Delaware estuary, the δ 15 N values of seston reached a maximum of +18‰ due to the fractionation during assimilation of NH 4 + ions [42].
The δ 34 S values ranged from -1.75 to 13.8‰ (M: 11.0‰, IR: 6.79-12.7‰). Here, the lowest values belonged to Indian (S7) and Italian (S44) samples, while the highest values were measured in Chinese (S11, S23, S25) and NS (S20, S42) samples. Studies on algae have shown little isotopic discrimination during the assimilation and reduction of sulfate. The isotopic composition of total sulfur in algae is depleted in 34 S by only 1-2‰ regarding the dissolved sulfate, which indicates that algae sulfate metabolism involves little or no isotope fractionation [43]. Therefore, δ 34 S values in algae will reflect those of meteoric water or water used in their growth medium and geology [44,45]. It is interesting to note that in the Hawaiian Islands, δ 34 S values of sulfates from volcanic ash and basalt-derived soils range from 6.3 to 15.4‰ [46] that are also in agreement with our data.
The lowest δ 2 H was determined in the NS (S39, S45), Chinese (S11, S27), and sample S5 (outside EU), while the highest values were found in NS samples with undeclared origins (S22, S37). The  The GMWL defines the ratio of the stable isotopes in natural meteoric waters (i.e., water derived from snow, rain, and other forms of precipitation) and is typically defined by the following equation: δ 2 H = 8.2 × δ 18 O + 10.8 [22]. Thus, our data indicate that hydrogen and oxygen isotopes in analyzed Spirulina samples originate mainly from local meteoric water and are only minimally affected by other processes such as metabolism.

Principal Component Analysis of All Spirulina Samples
Principal Component Analysis (PCA) was applied to identify trends and examine the distribution of variables in the analyzed samples. For this analysis 46 Spirulina commercial samples from the Slovenian market were obtained, and 18 analyzed parameters (macroelemental and isotopic composition data) were used (Figure 2). In the PCA score plot (Figure 2a), a grouping of the samples can be observed corresponding to different elemental and isotopic compositions of the included samples. Here, three outstanding groups represented by different variables can be identified. Information about the variables that con- The GMWL defines the ratio of the stable isotopes in natural meteoric waters (i.e., water derived from snow, rain, and other forms of precipitation) and is typically defined by the following equation: δ 2 H = 8.2 × δ 18 O + 10.8 [22]. Thus, our data indicate that hydrogen and oxygen isotopes in analyzed Spirulina samples originate mainly from local meteoric water and are only minimally affected by other processes such as metabolism.

Principal Component Analysis of All Spirulina Samples
Principal Component Analysis (PCA) was applied to identify trends and examine the distribution of variables in the analyzed samples. For this analysis 46 Spirulina commercial samples from the Slovenian market were obtained, and 18 analyzed parameters (macro-elemental and isotopic composition data) were used ( Figure 2). In the PCA score plot (Figure 2a), a grouping of the samples can be observed corresponding to different elemental and isotopic compositions of the included samples. Here, three outstanding groups represented by different variables can be identified. Information about the variables that contributed most to the grouping of the samples in the PCA is provided in the PCA variables loading plot (Figure 2b). , and S (S37: 3.88 and S22: 3.60 g/kg; M: 7.65 g/kg, IR: 7.14-8.26 g/kg). The lower content of these elements suggests a lower Spirulina content representing possible adulteration, since Spirulina typically contains high levels of P, K, Fe, and Zn [9,47,48]. Alternatively, a different growth medium could also result in different mineral compositions. For example, Michael et al. [49] showed a connection between using a poorer culturing medium (regarding elemental composition) and a lower mineral content in the final Spirulina product. However, given that S37 had the second highest δ 15 N value (13.3‰) among all samples, the use of organic fertilizers in its cultivation can be suggested. As this type of cultivation medium is rich in nutrients and has a positive effect on Spirulina mineral uptake [25][26][27], it is unlikely for the growth medium to be responsible for the poor mineral content in this sample. Additionally, the highest δ 18   7.14-8.26 g/kg). The lower content of these elements suggests a lower Spirulina content representing possible adulteration, since Spirulina typically contains high levels of P, K, Fe, and Zn [9,47,48]. Alternatively, a different growth medium could also result in different mineral compositions. For example, Michael et al. [49] showed a connection between using a poorer culturing medium (regarding elemental composition) and a lower mineral content in the final Spirulina product. However, given that S37 had the second highest δ 15 N value (13.3‰) among all samples, the use of organic fertilizers in its cultivation can be suggested. As this type of cultivation medium is rich in nutrients and has a positive effect on Spirulina mineral uptake [25][26][27], it is unlikely for the growth medium to be responsible for the poor mineral content in this sample. Additionally, the highest δ 18 O (S37: 25.8‰, S22: 27.2‰) and δ 2 H values (S37: -105‰, S22: -97.4‰) among all samples were observed for S37 and S22 (Figure 2b). High δ 18 O and δ 2 H values could be attributed to the proximity of the Spirulina culturing site to the equatorial region and the short distance from the sea, as hydrogen and oxygen isotopic composition are strongly latitude-and altitude-dependent [20,22].
Among the samples within the elliptical field in Figure 2a, samples S18, S9, and S44 in the upper right quadrant appear to form a separate group with distinct characteristics. Here, S18 and S44 are differentiated by their high Sr values (478 and 86.2 mg/kg, respectively; M: 23.0 mg/kg, IR: 13.1-30.8 mg/kg) and low δ 15 N values (0.77 and -3.92‰, respectively) compared to other samples (median value for δ 15 N: 4.80‰, IR: 1.30-8.02‰). Additionally, S18 also has the highest Ca content. High Ca and Sr values in S18 can be explained by the algae Lithothamnium in this sample. In contrast to Spirulina and Chlorella, which also make up this product and contain moderate Ca and Sr levels, Lithothamnium algae contain higher values of these elements [50,51]. In S44, the higher level of Sr could be attributed to contamination during the flaking process, as this is the only sample in a flake form. The flaking process's impact on contamination with certain elements has been shown several times in previous research, where metal contamination came from the enameled parts of the flaking rollers [52,53]. As δ 15 N values are primarily influenced by the nitrogen isotopic composition of the nitrogen source used during cultivation and internal transformations, they can indicate the type of fertilizer used, e.g., lower δ 15 N values, as are observed in S18 and S44, point to the use of inorganic fertilizers [17,40,54].
Additionally, S9 and S44 possess lower P content (S9: 5.06 and S44: 8.56 g/kg) than most samples (M: 10.9 g/kg, IR: 10.1-12.1 g/kg). This could be explained by the samples' lower amount of algal material since this is a mixed product, containing, in addition to Spirulina, Chlorella, barley and wheat grass. While algae (in this case, Spirulina and Chlorella) are rich in phosphorus, this is not true for cereal grasses, whose content is lower [47,48]. The lower P content in S44 (and partially also S9) could be due to the drying technique used in its production. It has been shown that different drying techniques reduce the P levels in the dried material compared to the fresh one [55]. As can be observed in Figure 2b, all the samples in this group (S9, S18, and S44) have somewhat higher oxygen and hydrogen stable isotope ratios in common, which range from 19. The Fe content in S26, S4, and S2 was 3.09, 3.48, and 3.29 g/kg, respectively, with a median of 0.69 g/kg (IR: 0.49-1.13 g/kg), while the Zn values in S26, S4, and S2 were 35.5, 52.7, and 43.6 mg/kg, respectively, and the median value for all samples was 15.1 mg/kg (IR: 10.3-21.5 mg/kg). The Cl content was 5.63 g/kg for S26, 5.77 g/kg for S4, and 3.07 g/kg for S2 (M: 2.02 g/kg, IR: 0.88-2.97 g/kg). The higher content of these elements in the Hawaiian and S2 samples may be due to their deliberate addition to the growth medium, which enables Spirulina to uptake and accumulate these elements. Spirulina elemental content has been previously shown to reflect that of the culturing medium [28,30,31]. Additionally, mineral addition to the Spirulina culturing medium to enhance its efficiency as a nutritional source of various minerals is not uncommon [28,30,31,56]. Additionally, the high content of elements Rb, Fe, Zn, and Cl could result from manure used as a fertilizer in the growth medium, as it has been shown that adding manure in crop cultivation results in elemental composition enhancement of the plants [57,58]. A deviation in δ 18  Similarly, as previously shown in samples S22 and S37, high δ 18 O and δ 2 H values can be attributed to the proximity of the production site to the equatorial region and sea and production at low altitudes [20,22]; all parameters are true for Hawaii. However, Figure 3 shows how samples S22 and S37 cannot be placed under Hawaiian samples due to differences in elemental composition. Contrarily, these samples appear to coincide with samples from non-EU regions or Asia, as the content of elements Cl, Fe, Zn, Br, and Rb (Figure 3b-f) is similar.  Another parameter, common to S4, S26, and S2 samples, is the high δ 15 N values. Measured values are 10.8‰, 13.8‰, and 8.81‰, respectively, with a median value for all samples of 4.80‰ (IR: 1.30-8.02‰). The nitrogen isotopic composition of the samples provides information about regional agricultural practices [20]. High δ 15 N values indicate the use of organic fertilizers, as the manure's stable isotopic composition has higher δ 15 N values than mineral fertilizers [59]. While Zarrouk's medium (Table S2) [60] is widely used as a standard medium for Spirulina production, manure (chicken, cow, pig) in Spirulina cultivation is also a common practice, as it represents a low-cost source of nitrogen and other necessary nutrients. The use of manure in Spirulina production results in good cellular growth and high pigment content while reducing production costs [25][26][27]. Additionally, the organic manures are enriched with microflora, which induces crops to uptake Another parameter, common to S4, S26, and S2 samples, is the high δ 15 N values. Measured values are 10.8‰, 13.8‰, and 8.81‰, respectively, with a median value for all samples of 4.80‰ (IR: 1.30-8.02‰). The nitrogen isotopic composition of the samples provides information about regional agricultural practices [20]. High δ 15 N values indicate the use of organic fertilizers, as the manure's stable isotopic composition has higher δ 15 N values than mineral fertilizers [59]. While Zarrouk's medium (Table S2) [60] is widely used as a standard medium for Spirulina production, manure (chicken, cow, pig) in Spirulina cultivation is also a common practice, as it represents a low-cost source of nitrogen and other necessary nutrients. The use of manure in Spirulina production results in good cellular growth and high pigment content while reducing production costs [25][26][27]. Additionally, the organic manures are enriched with microflora, which induces crops to uptake micronutrients [56]. The latter can be seen in our study in the high content of specific elements in S4, S26, and S2, as mentioned earlier.
Looking at these results, we can see a close connection between the Hawaiian samples, S4 and S26, and the sample without specified country of origin, S2.  (Figure 3c,d), indicating that it might also originate from Hawaii.
The clustering of the Italian samples (S44 and S46) using PCA analysis was unsuccessful (Figure 2a), despite having many common characteristics. One of the parameters separating them is the high Sr value in the S44, which was sold as flakes. As mentioned earlier, this observation is believed to be due to contamination arising during the flaking process. Unlike S44, S46 was obtained fresh and was subsequently lyophilized. Unlike the flaking process, lyophilization does not cause contamination and has little effect on mineral loss, except in the case of Mn [52,53,61]. Sample S44 has higher Ca content (3.45 g/kg), while the content in S46 is lower (0.46 g/kg). This finding could result from different processing, drying, and flaking techniques and undeclared excipients, as shown in previous research [55,62,63]. Additionally, the drying technique used can explain the higher level of Mn in S44 (84.9 mg/kg) than in S46 (32.9 mg/kg). As previously shown, freeze-drying can cause a decrease in the Mn content compared to oven drying [55]. The Zn content was also higher in the flaked sample (24.9 mg/kg) than in S46 (7.59 mg/kg), which is again believed to be a result of the flaking process [63] or selected drying method, as different drying techniques affect the Zn content differently [55]. Substantial variations in mineral profile have been observed between natural Spirulina and commercial products by Campanella et al. [62], which could result from various changing parameters introduced by the commercialization of the product, such as Spirulina biomass treatment, processing (washing, drying), packaging, and distribution.

Discriminant Analysis of Spirulina Samples
Discriminant analysis (DA) was performed using macro-elemental and isotopic composition data (Figure 4) to investigate the distribution of variables in more detail and reveal possible differences among the samples originating from China (n = 16), Hawaii (n = 2), India (n = 4), Italy (n = 2), Portugal (n = 2), and Taiwan (n = 3). The first two discriminant components (F1 and F2) account for 90.3% of the total variance. In the discriminant function score plot (Figure 4a), each cluster (centroid) is represented by a scatter plot. In the loadings plot (Figure 4b), they appear as vectors demonstrating a degree of association of the corresponding initial variable with the first two discriminant components. Red vectors indicate the most significant variables, and blue vectors represent the least significant variables for sample separation and clustering. Six groups of samples are identified in the DA score plot (Figure 4a).
A leave-one-out cross-validation (LOOCV) classified 82.8% of the samples correctly. The prediction ability was the highest for Hawaii, Italy, and Portugal (100%), and was the lowest for Taiwan (66.7%). The most critical variables for sample separation are Fe, Br, K, and P. DA analysis confirms our previous findings regarding distinct elemental and isotopic composition of Hawaiian samples. In addition, the separation of Italian and Portuguese samples was achieved using DA (Figure 4a).  A leave-one-out cross-validation (LOOCV) classified 82.8% of the samples correctly. The prediction ability was the highest for Hawaii, Italy, and Portugal (100%), and was the lowest for Taiwan (66.7%). The most critical variables for sample separation are Fe, Br, K, and P. DA analysis confirms our previous findings regarding distinct elemental and isotopic composition of Hawaiian samples. In addition, the separation of Italian and Portuguese samples was achieved using DA (Figure 4a).
Despite previously presented differences, the Italian samples have several similar characteristics which separate them from other samples in this study. For example, they possess the lowest content of Si (S44: 0.78 g/kg, S46: 0.68 g/kg; M: 5.06 g/kg, IR: 1.46-14.9 g/kg) and one of the lowest values of P (S44: 8.56 g/kg, S46: 6.64 g/kg; M: 10.9 g/kg, IR: 10.1-12.1 g/kg), the highest content of K (S44: 20.6 g/kg, S46: 26.9 g/kg; M: 15.2 g/kg, IR: 14.2-16.8 g/kg) among all samples, and a high Br content (S44: 8.00 mg/kg, S46: 7.07 mg/kg; M: 1.84 mg/kg, IR: 1.25-3.18 mg/kg). Similarly, as with samples from Hawaii, such an elemental composition could reflect the growth medium used, which appears to be specific for this production region. Moreover, S44 and S46 have the lowest δ 15 N values (-3.92 and -5.35‰, respectively), which points to the absence of organic fertilizers and the possible influence of natural processes such as nitrification, providing them with sufficiently specific δ 15 N composition that distinguishes these samples from others. The Italian samples also have one of the lowest δ 34 S values (S44: -0.61‰ and S46: 0.94‰; M: 11.0‰, IR: 6.79-12.7‰), which could be a result of combined meteoric water δ 34 S value and that of the added sulfur compounds in the Spirulina culturing medium. It could also indicate that Spirulina in S44 and S46 was cultivated in freshwater, since the δ 34 S values in seaweed are closer to seawater values, i.e., 17 to 21‰ [64,65].
Commercial Spirulina samples of Portuguese-declared origin appear to possess a distinct Mn and Br composition (Figure 4a  Similarly, as with samples from Hawaii, such an elemental composition could reflect the growth medium used, which appears to be specific for this production region. Moreover, S44 and S46 have the lowest δ 15 N values (-3.92 and -5.35‰, respectively), which points to the absence of organic fertilizers and the possible influence of natural processes such as nitrification, providing them with sufficiently specific δ 15 N composition that distinguishes these samples from others. The Italian samples also have one of the lowest δ 34 S values (S44: -0.61‰ and S46: 0.94‰; M: 11.0‰, IR: 6.79-12.7‰), which could be a result of combined meteoric water δ 34 S value and that of the added sulfur compounds in the Spirulina culturing medium. It could also indicate that Spirulina in S44 and S46 was cultivated in freshwater, since the δ 34 S values in seaweed are closer to seawater values, i.e., 17 to 21‰ [64,65].
Commercial Spirulina samples of Portuguese-declared origin appear to possess a distinct Mn and Br composition (Figure 4a,b). The Mn values for S29 and S30 are 192 and 195 mg/kg, respectively, with a median value for samples analyzed in DA (M All ) of 33.3 mg/kg (IR All : 28.0-38.2 mg/kg). The Br values in these samples are 2.71 mg/kg for S29 and 3.21 mg/kg for S30 (M All : 1.62 mg/kg, IR All : 1.04-3.21 mg/kg). Additionally, Portuguese samples have the highest Fe content, i.e., 1.14 (S29) and 1.12 g/kg (S30) (M All : 0.72 g/kg, IR All : 0.56-1.39 g/kg) and Si content (S29: 15.6 g/kg, S30: 15.1 g/kg; M All : 7.56 g/kg, IR All : 1.43-15.1 g/kg). Mn is the most important parameter for separating Portuguese and Chinese samples, together with δ 34 S values (S29: 7.34‰, S30: 7.01‰). The Mn, Br, and Fe levels in Spirulina products are highly dependent on their concentration in the growth medium; therefore, adding these minerals will result in increased levels in the microalgae. Moreover, the uptake of these elements is strongly affected by growth conditions, i.e., light intensity [28,30,31,66,67]. In this respect, Portuguese Spirulina product production might be specific. Regarding the Mn and Br content, using rich wastewater in these elements also increases their concentration in Spirulina [58]. High Si content results from silicon dioxide excipient's addition to the final product, which is evident from the declared product content ( Table 1). The δ 34 S composition of Portuguese Spirulina samples is in agreement with local δ 34 S results for rainwater (δ 34 S: 7.2‰) measured in neighboring Spain [68].
Further, OPLS-DA was applied to analyze the observed differences between six countries in the DA analysis and to investigate the goodness of fit (R 2 X) and prediction (Q 2 ) for the model. Obtained OPLS-DA resulted in five predictive and no orthogonal components (5 + 0), producing an R 2 X = 0.73, R 2 Y = 0.68, and Q 2 = 0.48. The F1 Score rate obtained by internal cross-validation was 86.2%, sensitivity was 96.2%, specificity was 57.1%, and accuracy was 87.9%. This model displayed high quality and goodness of fit and predictability (≥0.93) to differentiate among different countries, with the exception of Taiwan, where we obtained 0% predictability, which also supports our DA model. Moreover, OPLS-DA analysis for pairwise comparisons among all six countries ( Figure 5) was calculated, similarly as in the study performed by Potočnik et al. [69]. The separation between classes in the OPLS-DA score plots is evident. In Figure 5 [70][71][72]. Air pollution could, therefore, also be reflected in Spirulina products produced in polluted areas. The Sr values are also the highest among these samples, M Ch for Sr is 28.1 mg/kg (M All : 26.1, IR All : 15.2-31.1 mg/kg), which could be attributed to specific Spirulina processing techniques used [52,53]. In contrast, Chinese samples contain the lowest amount of Si (M Ch : 1.71 g/kg) among the samples included in the DA (M All : 7.56 g/kg, IR All : 1.43-15.1 g/kg), and the lowest δ 15 N value (M Ch : 2.32‰; M All : 3.04‰, IR All : 0.95 to 6.44‰). Lower Si content could partially be explained by the declared pure Spirulina composition of Chinese samples, while all Portuguese and half of the Indian samples contain Si as a part of the silicon dioxide excipient. Additionally, higher Si content in non-Chinese samples could come from various Spirulina growth medium mineral additions and Si in the water used for cultivation and different processing techniques. Different food processing techniques are known to reduce the Si content in the final product [73].
Results presented in Figure 5a-d confirm the results presented by PCA and DA analyses regarding distinct composition of Hawaiian, Italian, and Portuguese samples. Additional pairwise comparisons to confirm these findings are available in Figures S1 and S2.  . Differences in δ 13 C composition could be due to the addition of different nutrients to the growth medium, excipients to the final product, and cultivation conditions (open or closed system). Closed production systems enable better control over cultivation conditions, such as loss of CO 2 to the atmosphere, temperature, and pH, and possess lower carbon dioxide δ 13 C values [17,74]. There is a considerable difference in the δ 34 S values among Indian, Taiwanese, and Chinese samples as well-the δ 34 S values in Indian samples are substantially lower (5.7‰) than in Taiwanese samples (11.5‰) and Chinese samples (12.9‰). As Taiwan is an island, a higher influence of the sea than on the mainland (India, China) is expected. The δ 34 S values in rainwater decrease while moving inland and further from the sea. Therefore, higher δ 34 S values in rainwater and plants and algae are expected in regions closer to the sea [75]. In addition, Taiwan is a volcanic island, and therefore higher δ 34 S values can be also expected [46,76]. Higher δ 34 S values in Chinese samples, on the other hand, could be attributed, as mentioned previously, to high pollution levels in the area [70][71][72]. Separation of the Indian and Taiwanese samples from Chinese samples is also due to higher δ 15 N values (M Ch : 2.32‰, M Ind : 6.13‰, M Taiw : 6.22‰), which points to a specific fertilizing technique for these areas which possibly includes the use of organic fertilizers. Important parameters for distinction between Chinese, Taiwanese, and Indian samples are also δ 18  This could be attributed to the production of Taiwanese and Indian samples at lower latitudes, altitudes, or closer to the sea [20,22], similarly to Hawaiian samples. Another important variable that separates Spirulina samples from China, Taiwan, and India is the Si content, the median of which for Indian samples (M Ind ) is 13.8 g/kg, for Taiwanese samples (M Taiw ) is 7.94 g/kg, and for Chinese samples (M Ch ) is 1.71 g/kg (M All : 7.56, IR All : 1.43-15.1 g/kg), which can be explained by SiO 2 's addition to the final product in the case of Indian samples. Taiwanese and Chinese Spirulina, on the other hand, are declared as pure. As described above, there are various parameters affecting the Si content in these products, such as addition during cultivation, varying water Si concentration, and different processing techniques [73]. Differences were also found in Zn (M Taiw : 15.7 mg/kg, M Ch : 15.8 mg/kg, M Ind : 10.9 mg/kg), Fe (M Taiw : 0.66 g/kg, M Ch : 0.77 g/kg, M Ind : 0.46 g/kg), and K (M Taiw : 13.6 g/kg, M Ch : 15.6 g/kg, M Ind : 15.1 g/kg) values, which could be a result of mineral addition and use of different fertilizers in Spirulina growth medium [28,30,31,57,58].
According to the presented data, reliable and specific classification of samples S20, S21, S28, S39, S42, and S45 of undeclared origin (NS) is not possible. However, their elemental and isotopic compositions show similarity with Asian and declared non-EU samples ( Figure 3). The findings of this study show the importance of combining both elemental and isotopic values in verifying the country of origin and authenticity of Spirulina samples. However, the prediction ability and assessment of authenticity should be improved in the future by including a higher number of samples, as well as verified pure Spirulina samples.

Conclusions
Interest in Spirulina dietary supplements is growing among consumers due to vegetarianism, increasing malnutrition, and health awareness in the population. The high demand for Spirulina products and the challenging culturing conditions needed for producing highquality products make them a target for intentional adulteration and mislabeling [12,77]. This study has shown that combining stable isotope ratios of light elements (C, N, S, H and O) and elemental composition creates a promising tool for determining the authenticity of the commercial Spirulina dietary supplements, regarding their composition and geographical origin. Hydrogen and oxygen stable isotope ratios have been determined in Spirulina-based products for the first time in this study and show a correlation, as it occurs also in water, indicating that they originate mainly from local precipitation and that the influence of other parameters on their values is negligible.
A wide variability in the stable isotopic ratios and elemental composition among Spirulina samples of different declared origins was observed. Different statistical methods and reliable discrimination of Hawaiian, Italian, and Portuguese samples were also achieved, together with a good separation of Chinese samples. Discrimination between Taiwanese and Indian samples, however, was less successful but still notable. The parameters responsible for sample discrimination appear to be different culturing and processing techniques, environmental conditions (including pollution), and the geographical location of Spirulina production.
Additionally, this method shows promising results in exposing adulterated samples and samples mixed with other products and could be used in future studies for assessing product authenticity. A higher number of samples and more precise information on product culturing conditions, composition, and origin would result in more reliable results in Spirulina product authentication.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/foods12030562/s1, Table S1: Elemental composition of Spirulina dietary supplements from the Slovenian market; Table S2: Composition of Zarrouk's medium; Figure  S1 and S2: Pairwise comparisons between different declared countries of origin of Spirulina products.