Polar Lipids of Commercial Ulva spp. of Different Origins: Profiling and Relevance for Seaweed Valorization

Macroalgae of the genus Ulva have long been used as human food. Local environmental conditions, among other factors, can have an impact on their nutrient and phytochemical composition, as well as on the value of the seaweed for food and non-food applications. This study is the first to initiate a comparison between commercial Ulva spp. from different European origins, France (FR, wild-harvested Ulva spp.), and Portugal (PT, farm-raised Ulva rigida), in terms of proximate composition, esterified fatty acids (FA), and polar lipids. The ash content was higher in PT samples, while FR samples had higher levels of proteins, lipids, and carbohydrates and other compounds. The profile of esterified FA, as well as FA-containing polar lipids at the class and species levels were also significantly different. The FR samples showed about three-fold higher amount of n-3 polyunsaturated FA, while PT samples showed two-fold higher content of monounsaturated FA. Quantification of glycolipids and phospholipids revealed, respectively, two-fold and three-fold higher levels in PT samples. Despite the differences found, the polar lipids identified in both batches included some lipid species with recognized bioactivity, valuing Ulva biomass with functional properties, increasing their added value, and promoting new applications, namely in nutraceutical and food markets.


Introduction
Given the negative environmental impact of current food systems, there is an urgent need to promote healthy and sustainable diets with low environmental impacts [1]. In this context, along with the fast-growing world population driving up global food demand, marine macroalgae (commonly known as seaweed) are of growing interest, as natural and sustainable sources of essential nutrients and health-promoting compounds [2,3].
Macroalgae contain a wide range of valuable compounds, such as polysaccharides, lipids, proteins, phenolic compounds, and pigments [2,4]. The composition of these compounds is influenced by natural fluctuations in environmental parameters (e.g., temperature, light, salinity, and availability of nutrients), as well as by biotic interactions [4][5][6][7].
profiling esterified FA. The evaluation of the variations between PT and FR samples was performed using multivariate and univariate statistical analysis.

Proximate Composition Analysis
The moisture content of freeze-dried and milled biomass was determined by drying samples (250 mg × 5 replicates) at 105 • C for 15 h, which were then placed in a muffle furnace at 575 • C for a duration of 6 h for ash determination [39]. Elemental analysis (C, H, N, and S) (2 mg × 5 replicates) was performed on a Leco TruspecMicro CHNS 630-200-200 elemental analyzer using a combustion furnace temperature of 1075 • C and afterburner temperature of 850 • C. Two nitrogen-protein conversion factors, 6.25 and 5.0, were used to calculate the protein content [40]. Lipid content was determined by gravimetry of polar lipid extracts (Section 2.3). The content of carbohydrates and other compounds was estimated by difference, subtracting the percentage of ash, proteins (N × 6.25), and lipids from 100%. All contents were expressed as a percentage of the dry weight (%DW), excluding moisture (% of freeze-dried sample weight).

Extraction of Polar Lipids
Extracts containing polar lipids were obtained from Ulva spp. samples using the Bligh and Dyer method [41], with slight modifications [29,39]. Dried samples (250 mg × 5 replicates) were homogenized by adding methanol (2.5 mL) and chloroform (1.25 mL), vortexing for 2 min, and applying sonication for 1 min, followed by incubation in ice with shaking for 2.5 h. The samples were then centrifuged for 10 min at 2000 rpm. After the recovery of the organic phase, the biomass residue was re-extracted with methanol (2 mL) and chloroform (1 mL). The combined organic phase was washed by adding Milli-Q water (2.3 mL), followed by centrifugation as above. The aqueous phase was washed with chloroform (2 mL), whereas the biomass residue was re-extracted again with methanol (2 mL) and chloroform (1 mL). The final organic phase (denominated as polar lipid extract) was dried to dryness, resuspended in chloroform, transferred to amber vials (previously weighed), dried again under nitrogen, weighed, and stored at −20 • C before analysis.

GC-MS Analysis of Esterified Fatty Acids
To obtain fatty acid methyl ester (FAME) derivatives by base-catalyzed transmethylation [42], dried lipid extracts (30 µg) were mixed with 1 mL of an internal standard solution (1.0 µg mL −1 of methyl nonadecanoate in n-hexane) and 200 µL of a KOH solution (2 mol L −1 in methanol). After vortexing for 2 min, 2 mL of a NaCl solution (10 mg mL −1 in water) were added and the sample was centrifuged for 5 min at 2000 rpm. A volume of 600 µL of the organic (upper) phase containing FAME derivatives was recovered and dried to dryness under nitrogen. FAME were then dissolved in 60 µL of n-hexane and 2 µL of this solution were used for gas chromatography-mass spectrometry (GC-MS) analysis (6890N Network GC system and 5973 Network Mass Selective Detector; Agilent Technologies, Santa Clara, CA, USA). The GC system was equipped with a DB-FFAP (J&W Scientific, Folsom, CA, USA) capillary column (30 m × 0.32 mm internal diameter, 0.25 µm film thickness) and used helium (1.4 mL min −1 ) as carrier gas, with injector temperature of 220 • C, detector temperature of 280 • C, and oven temperature of 80 • C (3 min) → 25 • C min −1 → 160 • C → 2 • C min −1 → 210 • C → 30 • C min −1 → 250 • C (10 min). The MS system was operated with electron impact ionization at 70 eV in full scan mode, scanning m/z 50-550 in a 1 s cycle. The FAME derivatives were identified by comparing their retention times and mass spectra with those of a commercial mixture of FAME standards (Supelco 37 Component FAME Mix, Sigma-Aldrich, Darmstadt, Germany) and by comparing with the available MS spectra in the Wiley 275 library and The Lipid Web library [43], the latter containing all FA found in the Ulva samples. For FA quantification, calibration curves were obtained under the same instrumental conditions from different dilutions of the FAME mixture (dilution factor of 10 to 400), all containing the same amount of internal standard (3.7 µg of methyl nonadecanoate). The FA in the lipid extracts were quantified using the calibration curve of the respective standard, except for those in which there was no standard in the FAME mixture. In these cases, the FA with the closest number of carbons and double bonds were used. Thus, the quantification of 16:4n-3, 18:4n-3, and 20:4n-3 was performed using the calibration curve of 20:4n-6.

Quantification of Glycolipids and Phospholipids
Glycolipids were quantified by measuring the sugar content using the orcinol method [44]. For that, 100 µg of each dried lipid extract in 2 mL of orcinol solution (0.2% in 70% H 2 SO 4 ) were incubated at 80 • C for 20 min. After cooling to room temperature, the absorbance was read at 505 nm (Multiskan GO, Thermo Scientific, Hudson, NH, USA). For determination of the sugar content in each lipid extract, a calibration curve was obtained by performing the same reaction with glucose standards (up to 40 µg) prepared from an aqueous glucose solution (5 mg mL −1 ). The content of glycolipids was estimated by multiplying the sugar content by 2.8 [45].
Phospholipids quantification was achieved by measuring the phosphorous amount using the molybdovanadate method [46], as done in our previous work [44]. The lipid hydrolysis was performed by adding 70% perchloric acid (125 µL) to each dried lipid extract (50 µg in a glass tube washed with 5% nitric acid), followed by incubation on a heating block at 180 • C for 1 h. After cooling to room temperature, Milli-Q water (825 µL), ammonium molybdate (2.5 g 100 mL −1 in Milli-Q water; 125 µL), and ascorbic acid (0.1 g mL −1 in Milli-Q water; 125 µL) were added to each sample, vortexing after each addition. Samples were placed in a water bath at 100 • C for 10 min and then allowed to cool in a cold-water bath. Phosphate standards containing up to 2 µg of phosphorus (P) were prepared from a sodium dihydrogen phosphate dihydrate solution (100 µg mL −1 of P), using the same procedure as samples, excluding the heating on the block heater. After reading the absorbance at 797 nm on the microplate UV-Vis spectrophotometer, the content of phospholipids was estimated by multiplying the amount of P by 25 [47].

Analysis of Polar Lipids by HILIC-MS and MS/MS
Polar lipids were analyzed by hydrophilic interaction liquid chromatography-mass spectrometry (HILIC-MS) (Ultimate 3000 Dionex HPLC system and Q-Exactive ® hybrid quadrupole Orbitrap ® mass spectrometer; Thermo Fisher Scientific, Waltham, MA, USA). The autosampler chamber was kept at 5 • C and the volume of the sample loop was 20 µL.  (17:0), 0.02 µg of lysophosphatidylcholine (19:0), 0.02 µg of ceramide (17:0/d18:1), 0.04 µg of dimyristoyl phosphatidylserine, 0.08 µg of tetramyristoyl cardiolipin, 0.08 µg of dipalmitoyl phosphatidylinositol, and 0.08 µg of dimyristoyl phosphatidic acid. The mass spectrometer was operated simultaneously in positive (3.0 kV) and negative (−2.7 kV) modes. The capillary temperature was 250 • C, sheath gas flow 15 U, and auxiliary gas flow 5 U (arbitrary units). In MS/MS experiments, a resolution of 17,500 and AGC target of 1 × 10 5 were used. The acquisition cycles consisted of one full scan MS and ten data-dependent MS/MS scans, repeated continuously with the dynamic exclusion of 60 s and an intensity threshold of 2 × 10 4 . MS data were acquired with a high resolution of 70,000 and automatic gain control (AGC) target of 1 × 10 6 . MS/MS data were recorded with a resolution of 17,500, AGC target of 1 × 10 5 , and a normalized collision energy ranging between 25, 30, and 35 eV. Data acquisition and visualization were performed using the Xcalibur data system (V3.3, Thermo Fisher Scientific, Waltham, MA, USA).
In the analysis of LC-MS data using MZmine 2.32 software [48], only peaks with an intensity higher than 1 × 10 4 were considered. The validated peaks were within the typical retention time of the respective lipid class, with a mass error of less than 5 ppm. The  [39,49,50]. For normalization of the LC-MS data, the integrated peak areas of the extracted ion chromatograms (XIC) were exported from MZmine software, and the peak area of each species was then divided by the peak area of the lipid standard with the closest retention time. The resulting matrix (.xls) with normalized areas of all species was used to study the origin-driven variations at the level of lipid species. To study variations at the level of lipid classes, a new matrix was created considering the sum of normalized areas of all lipid species within each class.

Statistical Analysis
The statistical analysis was done using R version 4.0.3 [51] in Rstudio version 1.3.1 [52]. The R package Metaboanalyst was used for imputation of missing values and log transformation of the data [53]. Principal component analysis (PCA) was done using R built-in function and R package pcaMethods [54]. Mann-Whitney U test was performed using R built-in functions with the Benjamin-Hochberg correction for the false discovery rate (FDR, q-value) [55]. Heatmaps were created from autoscaled data with the R package pheatmap [56] using Euclidean (clustering distance) and ward.D (clustering method). All statistics, graphics and boxplots were created using the R packages ggplot2 [57], plyr [58], dplyr [59], tidyr [60], and ggrepel [61].

Proximate Composition of Ulva spp. Biomass
The freeze-dried samples of Ulva spp. from France (FR) and Portugal (PT, Ulva rigida in this case) were analyzed for the determination of moisture and ash, proteins, and lipids contents, while carbohydrates and other compounds were estimated by difference (Table 1). The mean moisture content (expressed as a percentage of freeze-dried sample weight) in Ulva spp. was 8.77 ± 0.25% for FR and 4.45 ± 0.47% for PT (q < 0.05). The mean ash content (% of dry weight, %DW) of Ulva spp. of PT origin was 32.50 ± 0.34 %DW, while that of FR origin was 13.56 ± 0.52 %DW (q < 0.05).
The protein content was estimated from the determination of elemental nitrogen (Table 1) using two nitrogen-protein conversion factors: 6.25, a commonly used factor, and the factor 5, proposed as an appropriate factor for seaweed [40]. Using the factor 6.25, the mean protein content of Ulva spp. of FR origin was 15.59 ± 0.09 %DW, while of U. rigida of PT origin was 11.13 ± 0.31 %DW. Using factor 5, the mean protein content was 12.47 ± 0.08 %DW for FR and 8.90 ± 0.25 %DW for PT samples. Independently of the nitrogen-to-protein conversion factor used, the protein content was significantly higher in the FR samples than in the PT samples (q < 0.05).
The mean of the lipid content estimated by gravimetry of the polar lipid extracts was 2.23 ± 0.10 %DW for the FR samples and 1.14 ± 0.12 %DW for the PT samples. The lipid content was significantly higher in the FR samples than in PT samples (q < 0.01).
The content of carbohydrates and other compounds (estimated by difference) of Ulva spp. of FR samples (68.60 ± 0.56 %DW) was also significantly higher than that of the PT samples (55.23 ± 0.32 %DW) (q < 0.05).

Esterified Fatty Acid Profile
The profile of esterified fatty acids (FA) present in Ulva spp. samples from two origins (FR and PT) were determined by GC-MS (Table 2 and Supplementary Figure S1). A total of 14 FA were identified and quantified in both groups, including saturated (SFA), monounsaturated (MUFA), and polyunsaturated (PUFA) fatty acids. Table 2. Profile of esterified fatty acids present in Ulva spp. samples from France (FR) and Portugal (PT), determined by gas chromatography-mass spectrometry (GC-MS) and expressed in mg g −1 of polar lipid extract (ext) 1 . Multivariate principal component analysis (PCA) of the FA dataset showed that the experimental groups (FR and PT) were well discriminated in a two-dimensional score plot with the first two dimensions (dim), expressing 97.1% of the total variance (91.4% of dim1 and 5.7% of dim2). The two groups were separated along dim1 with FR samples located at positive values and PT samples at negative values (Figure 1a). The most relevant variables explaining the variability of dim1 (contributions > 10%) were PUFA 18:2n-6 (41.7%), 16:4n-3 (15.5%), 18:4n-3 (15.1%), and 22:5n-3 (11.5%) (Supplementary Figure S2), which were increased in the FR group.
Univariate analysis was performed using the FA dataset to test for significant differences between the two groups. The Mann-Whitney test revealed significant differences in the content of all FA except SFA 16:0 ( Table 2). In addition, the total FA content in the respective polar lipid extract was significantly higher in the FR samples (210.75 ± 24.07 mg g −1 ext) than in PT (164.84 ± 11.67 mg g −1 ext) (q < 0.05).
The clustering of all identified FA was visualized in a two-dimensional hierarchical clustering heatmap (Figure 1b). The FA were divided into two main clusters. From top to bottom, the first cluster included all identified MUFA (18:1, 16:1n-7, and 16:1n-9) and 2 SFA (14:0 and 22:0), which were significantly more abundant in the PT samples. The second cluster included all identified PUFA and SFA 18:0, which were significantly increased in the FR samples (Table 1) The clustering of all identified FA was visualized in a two-dimensional hierarchical clustering heatmap (Figure 1b). The FA were divided into two main clusters. From top to bottom, the first cluster included all identified MUFA (18:1, 16:1n-7, and 16:1n-9) and 2 SFA (14:0 and 22:0), which were significantly more abundant in the PT samples. The second cluster included all identified PUFA and SFA 18:0, which were significantly increased in the FR samples ( Table 1).
The atherogenicity index (AI) and thrombogenicity index (TI), proposed as indicators of the propensity of the diet or food to influence the incidence of coronary heart disease (CHD) [37], were calculated from FA profiles. The values of AI and TI were significantly lower in the FR samples, showing a higher content of PUFA. In contrast, the n-6/n-3 ratio was significantly lower in the PT samples.

Polar Lipidome Profile
The glycolipid and phospholipid contents in polar lipid extracts were estimated by colourimetric methods. The glycolipids represented 464.18 ± 52.45 mg g −1 of the PT sample extract and 233.33 ± 11.41 mg g −1 of the FR sample extract. The phospholipids represented 89.41 ± 8.88 mg g −1 of the PT extract and 27.24 ± 12.50 mg g −1 of the FR extract. Glycolipids and phospholipids were significantly higher in PT samples (q < 0.01).
A detailed analysis of the polar lipidome of Ulva spp. was performed using liquid chromatography coupled with high-resolution mass spectrometry (LC-MS). Considering the two groups (FR and PT), a total of 156 species of polar lipids of distinct molecular weight were identified and semi-quantified (Supplementary Tables S1-S3 Table 3 and Supplementary Figures S3 and S4). The atherogenicity index (AI) and thrombogenicity index (TI), proposed as indicators of the propensity of the diet or food to influence the incidence of coronary heart disease (CHD) [37], were calculated from FA profiles. The values of AI and TI were significantly lower in the FR samples, showing a higher content of PUFA. In contrast, the n-6/n-3 ratio was significantly lower in the PT samples.

Polar Lipidome Profile
The glycolipid and phospholipid contents in polar lipid extracts were estimated by colourimetric methods. The glycolipids represented 464.18 ± 52.45 mg g −1 of the PT sample extract and 233.33 ± 11.41 mg g −1 of the FR sample extract. The phospholipids represented 89.41 ± 8.88 mg g −1 of the PT extract and 27.24 ± 12.50 mg g −1 of the FR extract. Glycolipids and phospholipids were significantly higher in PT samples (q < 0.01).
A detailed analysis of the polar lipidome of Ulva spp. was performed using liquid chromatography coupled with high-resolution mass spectrometry (LC-MS). Considering the two groups (FR and PT), a total of 156 species of polar lipids of distinct molecular weight were identified and semi-quantified (Supplementary Tables S1-S3 Table 3 and Supplementary Figures S3 and S4).
All the lipid species identified in the two Ulva spp. were semi-quantified by the corresponding area of the chromatographic peak, divided by the area of the peak of an internal standard. To study the variations in lipid classes, the matrix with the normalized areas of all lipid species was used to calculate the abundance of each lipid class by adding the normalized areas of the lipid species of the same class.  The PCA scores plot of the lipid classes showed that the two groups (PT and FR) were separated in the first two dimensions, and the model captured 97.7% of the total variance in the dataset (93.1% dim1 and 4.6% dim2) (Figure 2a). The two groups were discriminated along dim1, the main discriminating component, with FR samples located at negative values and PT samples located at positive values ( Figure 2a). The most relevant contributors to dim1 (contributions > 7%) included PI (35.5%), SQMG (19.1%), SQDG (13.6%), and LPC (7.3%) (Supplementary Figure S5).
Univariate analysis by Mann-Whitney test showed that 9 of the 16 lipid classes (56.3%) were significantly different between the PT and FR samples (q < 0.05). DGTS, SQDG, SQMG, LPC, PC, PE, LPI, PG, and PI were increased in the PT samples.
Hierarchical clustering of the lipid class dataset showed the classes divided into two main clusters (dendrogram to the left of the heatmap in Figure 2b). From top to bottom, the first cluster included MGMG, DGDG, and MGDG. The second group consisted of the 9 classes that showed statistically significant differences between the experimental groups and 4 other classes (DGMG, MGTS, LPE, and LPG), all increased in the PT group.
Regarding the lipid species, a distinct number of lipid species was detected between the experimental groups (147 detected in the FR samples and 154 in the PT samples) ( Table 3)  Regarding the lipid species, a distinct number of lipid species was detected between the experimental groups (147 detected in the FR samples and 154 in the PT samples) ( Table  3) (Supplementary Tables S1-S3).
Upon a first visualization of the abundance of lipid species within each class, considerable differences were identified in the abundance of several species, inclusive with the major lipid species distinct by class (Table 3 and Supplementary Figures S6-S8). These variations in lipid species were also assessed using multivariate and univariate statistical analysis.
PCA analysis of the lipid species dataset showed that the two groups, the FR and PT samples, were discriminated in a two-dimensional scores plot, expressing 98.7% of the total variance in the dataset (94.9% of dim1 and 3.8% of dim2) (Figure 3a). The most relevant contributors to dim 1 (contributions > 4%) included exclusive species of each group: the galactolipid species MGDG (32:0) and MGMG (22:5), increased in FR samples, and the sulfolipid species SQDG (28:0), SQDG (30:1), and SQMG (16:1), increased in PT samples (Supplementary Figure S9). Upon a first visualization of the abundance of lipid species within each class, considerable differences were identified in the abundance of several species, inclusive with the major lipid species distinct by class (Table 3 and Supplementary Figures S6-S8). These variations in lipid species were also assessed using multivariate and univariate statistical analysis.
Univariate analysis by Mann-Whitney test showed that 128 of 156 lipid species (82.05%) were significantly different between PT and FR samples (q < 0.05). A twodimensional hierarchical clustering heatmap was created using the top 25 lipid species with the lowest q-values (Figure 3b). The first level of the upper dendrogram showed the samples clustered into two groups (FR and PT), as observed in the PCA scores plot (Figure 3a). The clustering of the lipid species regarding their similarity in the changes of lipid expression (dendrogram to the left of Figure 3b) showed that they were separated into two main clusters, one comprising three species of neutral glycolipids, DGDG (32:0), MGDG (32:0), and MGMG (22:5), which were more abundant in FR samples. Of these three species, MGDG (32:0) and MGMG (22:5) were only detected in FR samples. The second cluster consisted of 22 lipid species (1 MGMG, 5 SQDG, 2 DGTS, 2 MGTS, 1 DGDG, 3 PI, 4 SQMG, 1 PC, 1 PG, 1 DGMG, and 1 LPE), which were increased in PT samples. Of these 22 species, 10 species (45.45%) had a degree of unsaturation of 1, which is consistent with the higher amount of MUFA found in the PT samples (Table 2). Also, nine species (40.91%) were lysolipid species and eight species (36.36%) were only detected in PT samples. Univariate analysis by Mann-Whitney test showed that 128 of 156 lipid species (82.05%) were significantly different between PT and FR samples (q < 0.05). A two-dimensional hierarchical clustering heatmap was created using the top 25 lipid species with the lowest q-values (Figure 3b). The first level of the upper dendrogram showed the samples clustered into two groups (FR and PT), as observed in the PCA scores plot (Figure 3a). The clustering of the lipid species regarding their similarity in the changes of lipid expression (dendrogram to the left of Figure 3b) showed that they were separated into two main clusters, one comprising three species of neutral glycolipids, DGDG (32:0), MGDG (32:0), and MGMG (22:5), which were more abundant in FR samples. Of these three species, MGDG (32:0) and MGMG (22:5) were only detected in FR samples. The second cluster consisted of 22 lipid species (1 MGMG, 5 SQDG, 2 DGTS, 2 MGTS, 1 DGDG, 3 PI, 4 SQMG, 1 PC, 1 PG, 1 DGMG, and 1 LPE), which were increased in PT samples. Of these 22 species, 10 species (45.45%) had a degree of unsaturation of 1, which is consistent with the higher amount of MUFA found in the PT samples (Table 2). Also, nine species (40.91%) were lysolipid species and eight species (36.36%) were only detected in PT samples.

Discussion
Macroalgae (seaweed) are natural and sustainable sources of essential nutrients and health-promoting compounds [2,4]. Local environmental conditions and harvest time, among other factors, may have an impact on their composition [4,5,7] and, therefore, on their uses as raw material for applications in food and non-food industries. The present study is the first to initiate a comparison between commercial Ulva spp. from different European origins, France (FR, wild-harvested Ulva spp.) and Portugal (PT, IMTA-raised Ulva rigida), in terms of the proximate composition, esterified fatty acids (FA), and polar lipids, both at the level of the classes and lipid species.

Discussion
Macroalgae (seaweed) are natural and sustainable sources of essential nutrients and health-promoting compounds [2,4]. Local environmental conditions and harvest time, among other factors, may have an impact on their composition [4,5,7] and, therefore, on their uses as raw material for applications in food and non-food industries. The present study is the first to initiate a comparison between commercial Ulva spp. from different European origins, France (FR, wild-harvested Ulva spp.) and Portugal (PT, IMTA-raised Ulva rigida), in terms of the proximate composition, esterified fatty acids (FA), and polar lipids, both at the level of the classes and lipid species.
In this study, PT samples showed significantly higher ash content, while FR samples showed higher content of proteins, lipids, and carbohydrates and other compounds. Regarding the PT samples and comparing studies that used the same methodologies, slightly lower ash values have been previously reported for U. rigida from the same origin but harvested in different seasons/years [33,39]. Using the factor 6.25, the protein content obtained from the PT samples was an intermediate value compared to those reported for U. rigida from the same origin cultivated in summer 2018 [33] and autumn 2016 [39]. The lipid content obtained from the PT biomass was about two-fold lower than that previously reported for another batch of Ulva rigida produced in the same month [29]. A value of carbohydrates and other compounds (estimated by difference) similar to that obtained from PT samples has been reported for Portuguese U. rigida produced in autumn 2016 [39] but slightly lower than in summer 2018 [33]. In the case of PT samples, such value (55.23 ± 0.32 %DW) was also similar to the content of insoluble and soluble fibers (55)(56)(57) found in dried (temperature-controlled air-tunnel) and fresh (salted) Ulva rigida samples [71]. Regarding the FR samples, the lipid content obtained in this study (2.23 ± 0.10 %DW) was similar to that previously reported (even with variations in the lipid extraction procedures) for Ulva species collected in the same region (Brittany coast, France), namely U. rotundata collected in December 1991 (1.9 %DW) [62] and U. armoricana collected in June 2012 (2.6 %DW) [16].
The differences in the proximate composition of Ulva spp. biomass under study (FR and PT) may be related, although not exclusively, to different environmental conditions experienced by different geographic locations. In the case of other Portuguese U. rigida biomass farmed in the same location and analyzed for proximate composition using the same methodologies [33,39], the differences can also be related to different environmental conditions experienced by different seasons/years but also to different cultivation protocols. As observed in a laboratory experiment with U. pertusa [72], the lipid content increased at low temperature. The higher lipid content found in the FR samples may be related to the fact that they grew in cooler waters. The average water temperature in May in Bretagne (Saint-Malo and Brest, FR) was 12.6 • C (Source: www.seatemperature.org/europe accessed on 12 November 2020), while that recorded in May 2017 in seaweed tanks of ALGAplus (Aveiro, PT) was 17.5 • C [29]. In addition to temperature, the possible effect of other environmental factors, namely salinity, light exposure, and nutrient availability, on the lipid content cannot be excluded [72]. Probably, wild Ulva spp. (FR) also experienced stronger water currents than Ulva rigida cultivated in IMTA (PT). Ulva spp. from FR were subjected to tidal variations, while Ulva rigida cultivated under IMTA conditions (PT) grew in a more constant environment not limited by nutrients. However, the chosen cultivation strategy in these aquaculture systems also impacts nutritional composition, as the focus on the increase of the biomass yield is normally detrimental to nitrogen accumulation, and thus to the protein content [73]. On other hand, the possibility of having more than a single Ulva species cannot be excluded for wild-harvested biomass (FR), as well as of having species-specific variations [34,63]. More studies under controlled conditions are necessary to dissect the effect of each environmental factor but also of intrinsic factors (life cycle and phylogeny) and cultivation/processing protocols on the composition of Ulva spp., particularly in terms of lipid content and its detailed composition.
Despite the variations in proximate composition, the same esterified fatty acids (FA), as well as the same classes of polar lipids were identified from lipid extracts of Ulva spp. of different origins (FR and PT). However, the profiles of FA and polar lipids (the latter at level of the classes and lipid species) differed significantly.
Overall, the results of the FA analysis are in line with those presented in other studies (Supplementary Table S4), showing high levels of SFA 16:0 and C16 and C18 PUFA in Ulva spp. [8,16,29,30,32,34,62,67]. The amount of MUFA 18:1 was calculated from the sum of two peaks, which should correspond to n-7 and n-9, as shown in previous studies [30][31][32]34]. When comparing the FA profile obtained in this study with other previously published, it is noteworthy that most of the studies used different methodologies for lipid extraction and fatty acid analysis, even considering only the analysis of FAME derivatives (Supplementary Table S4). The differences reported for the FA profile within the same genus may also be related to different methods used, not excluding the influence of environmental and other factors mentioned above.
Comparing the experimental groups under study, the FR samples showed a higher content of PUFA, which are recognized protective dietary factors against coronary heart disease (CHD) [74]. In contrast, the n-6/n-3 ratio was significantly lower in the PT samples. However, for both groups, n-6/n-3 ratio was less than 1, which is considered beneficial for human health not only in reducing CHD [10] but also other pathologies such as cancer and associated complications [75,76].
Regarding the effect of temperature changes on algae, it is well accepted that the unsaturation degree increases at lower temperatures [77]. Such effect may have contributed to the higher PUFA content found in the FR samples grown in cooler water. However, other environmental factors (e.g., light, salinity, and nutrients) can modulate the FA profile [72]. Also, the influence of non-environmental factors in the FA profile, namely species-specific variations [34,63], cannot be excluded. On other hand, the differences in the composition of PUFA between FR and PT samples may lead to an altered production of oxylipins formed by enzymatic oxidation, which deserves to be explored in future studies. A previous work with Ulva species collected at various sampling sites in the lagoon Ria Formosa (Faro, Portugal) showed that the production of polyunsaturated aldehydes is dependent on the PUFA content [78]. This is important when selecting Ulva biomass for further applications, namely in the food industry, as this class of substances results in different flavors.
The classes of polar lipids identified in the two groups (FR and PT) are in agreement with those previously described for Portuguese Ulva rigida [29,33,39,79]. Among these classes showing significant variations, sulfolipids are the most recognized as bioactive compounds [11][12][13].
The total number of lipid species identified from FR samples (147) was lower than that identified in PT samples (154). SQDG (28:0) and SQDG (30:1), detected only in PT samples, were between site-specific lipid species previously identified in wild Ulva spp. from the Iberian Atlantic coast and farmed Ulva rigida from Aveiro (PT) with the potential to trace their geographic origin [33].
Despite the differences found in this work at level of the lipid species, it is of note that both batches (PT and FR) showed polar lipid species that were previously identified as contributing up to 50% similarity of the Ulva rigida cultivated in different seasons/years [79]. Effectively, a part of the lipid signature seems to remain conserved in Ulva species. Nonetheless, the strength of the results presented in this study is limited, namely by the single sampling. In further studies, it should be interesting to evaluate if the differences found at the level of lipid species between FR and PT samples remain in other samplings at different months and years. Additionally, more work is needed to understand the influence of the different environmental and non-environmental factors on the variations observed and then to assess the impact on potential applications.
Regarding the interest of polar lipids as health-promoting compounds, glycolipid species with previously reported bioactivity have been identified in Ulva spp. samples of FR and PT, namely SQDG (16:0/16:0) [12] and SQMG (16:0) [13] with antimicrobial activity and MGDG (18:4/16:4) [14], MGMG (16:3) [15], and MGMG (16:2) [15] with antiinflammatory activity. Of note, MGMG (16:3) was in the top 25 lipid species with the most significant variation between FR and PT samples. In contrast, no significant variation was observed in the abundance of MGMG (16:2). Although less studied in terms of bioactive properties compared to glycolipids, phospholipids and betaine lipids of algae have also been recognized as important bioactive compounds; namely, the PG and PC species of the red macroalga Palmaria palmata [80] and DGTS species isolated from the microalga Nannochloropsis granulata [81] have been identified as potential anti-inflammatory agents. Further studies are needed to explore the biological activity of Ulva spp. polar lipids and monitor the biomass value of Ulva species based on their location and time of harvesting but also on other factors as species-specific differences, cultivation conditions, and processing protocols. The presence of such bioactive polar lipids is important for the valorization of Ulva spp. as a functional food, especially in view of the global need for healthy and sustainable diets.

Conclusions
The commercial biomass of Ulva may differ considerably in terms of their composition at a given time. In this study, commercial Ulva spp. harvested in May of the same year from different European origins were compared: wild-harvested Ulva spp. from France (FR) and farm-raised Ulva rigida from Portugal (PT). As supported by statistical analysis, the PT samples showed higher ash content, while the FR samples revealed higher levels of proteins, lipids, and carbohydrates and other compounds. The profile of esterified fatty acids (FA), as well as of polar lipids containing esterified FA at the level of lipid classes and lipid species, were also significantly different. The polar lipid species that showed significant variations included potentially bioactive lipids, namely neutral glycolipids as MGMG (16:3) Improved knowledge on the composition and lipid variation of commercial Ulva spp. will allow us to optimize the best uses of these green macroalgae and valorize specific production sites. This is particularly relevant given the importance of Ulva spp. as a functional food, contributing to the promotion of a healthy and sustainable diet.