Nutrient availability affects the polar lipidome of Halimione portulacoides leaves cultured in hydroponics.

Halophytes are increasingly regarded as suitable extractive species and co-products for coastal Integrated Multi-Trophic Aquaculture (IMTA) and studying their lipidome is a valid means towards their economic valorization. Halimione portulacoides (L.) Aellen edible leaves are rich in functional lipids with nutraceutical and pharmaceutical relevance and the present study aimed to investigate the extent to which its lipidome remains unchanged under a range of dissolved inorganic nitrogen (N) and phosphorus (P) concentrations typical of aquaculture effluents. Lipidomics analysis, done by hydrophilic interaction liquid chromatography coupled to high resolution mass spectrometry, identified 175 lipid species in the lipid extract of leaves: 140 phospholipids (PLs) and 35 glycolipids (GLs). Plants irrigated with a saline solution with 20-100 mg DIN-N L-1 and 3-15.5 mg DIP-P L-1 under a 1-week hydraulic retention time displayed a relatively stable lipidome. At lower concentrations (6 mg DIN-N L-1 and 0.8 mg DIP-P L-1), plants exhibited less PLs and GLs per unit of leaves dry weight and the GLs fraction of the lipidome changed significantly. This study reveals the importance of analyzing the lipidomic profile of halophytes under different nutritional regimens in order to establish nutrient-limitation thresholds and assure production conditions that deliver a final product with a consistent lipid profile.

bioactive properties have been related to GLs and PLs (e.g. anti-inflammatory and anticarcinogenic) as well as enhanced human cognitive functions and motor performance 28,[31][32][33][34][35][36] . Halophytes, contrarily to algae, have been particularly overlooked on that regard, and the existing lipid characterizations have been mostly limited to fatty acids, non-polar lipids and sterols 19,22,37,38 . To date, only one publication attempted to describe the polar lipidome of two edible halophyte species (Salicornia ramosissima J. Woods and Halimione portulacoides (L.) Aellen), using liquid-chromatography coupled with mass-spectrometry (LC-MS) 22 . Fully exploring the lipidome of halophytes is a major step towards their valorization as relevant cash-crops for both agriculture and aquaculture. Besides it is also important to take into consideration the potential variations in their lipidomic profile in response to changes in environmental and metabolic conditions 10,37,[39][40][41] . Understanding the circumstances and extent of those variations is essential to guarantee the supply of a consistent product when a stable lipid profile is a requisite.
The present study aimed to describe and assess potential shifts in the lipidome of sea purslane H. portulacoides leaves, grown hydroponically under different concentrations of dissolved inorganic nitrogen (DIN) and phosphorous (DIP). The concentrations used in this study aim to represent a wide range of possible values, as recorded in aquaculture effluents used in previous halophyte bioremediation studies under IMTA conditions [42][43][44][45][46] . To understand if contrasting concentrations of DIN and DIP affect the polar lipidome of H. portulacoides leaves, the present study tested the following null hypothesis (H 0 ): 'There are no significant changes in the polar lipidome of H. portulacoides cultivated in low-, medium-and high-input of DIN and DIP' . Lipid profile was evaluated by state of the art lipidomics analysis using HILIC coupled with mass spectrometry (MS) and tandem mass spectrometry (MS/MS), bioinformatic tools and statistical analysis.

Results
Total lipids, glycolipids and phospholipids quantification. Total lipid content was estimated by gravimetry and expressed as g 100 g −1 of dry weight (DW) (Fig. 1). Non-significant differences were detected between each treatment and the control (CT), with a tendency for increased lipid content in the leaves of H. portulacoides at higher concentrations of N and P in the solution: [N,P] low yielded 6,18 ± 0,99 g 100 g −1 of leaves dry weight (DW), followed by [N,P] med , with 7.96 ± 2.05 g 100 g −1 DW; and 9.23 ± 3.06 g 100 g −1 DW for [N,P] high . The total amount of lipid extract obtained from [N,P] high was similar to that recorded in the CT (9.44 ± 3.83 g 100 g −1 DW).
The levels of GLs and PLs present in the lipid extracts of the leaves of H. portulacoides were also estimated ( Fig. 2A), expressed as µg mg −1 of lipid extract. No significant differences were detected between treatments in neither GLs nor PLs contents. The overall average content of GLs was 455.87 ± 57.32 µg mg −1 of lipid extract and that of PLs was 175.49 ± 39.56 µg mg −1 of lipid extract.
Significant differences were recorded between treatment conditions regarding the concentration of GLs and PLs in the leaves of H. portulacoides, expressed as mg g −1 DW (Fig. 2B). The [N, P] low group had a significantly lower concentration of GLs (29.1 mg g −1 DW) than the CT (42.1 mg g −1 DW) and other treatments (38.2-39.1 mg g −1 DW), as well as a significantly lower concentration of PLs (9.7 mg g −1 DW) than the CT (17.5 mg g −1 DW) and other treatments (14.1-16.8 mg g −1 DW). The [N, P] med and [N, P] high groups did not differ from the CT in either type of lipids.
Lipidomic signature. A non-targeted lipidomics approach was used to evaluate the stability of the lipidome across treatment conditions. This approach provided a global profile of the polar lipid molecular species present in the extracts and potentially used as a lipid signature that characterizes states of N and P limitation and/or excess.
MS and MS/MS analysis allowed the accurate identification of 175 lipid species, namely 140 PLs (Table 1) and  35 GLs (Table 2), which were detected in all conditions. In a few cases, MS/MS spectra did not provide enough information to determine the fatty acyl composition, but the class was confirmed through the identification of the polar head and are therefore included in Table 1. No lipids were found to be unique to any one condition. The lipid classes identified were previously recorded in wild specimens 27 and are DGDGs, LPCs, LPEs, MGDGs, MGMG, PAs, PCs, PEs, PGs, PIs and SQDGs. The number of species identified per lipid class is represented in Fig. 3.
After raw data processing and species identification, the dataset was analyzed using chemometric statistical methods to extract and interpret data from a biologically relevant perspective, looking at changes in the lipidome in general and within specific lipid groups (GLs and PLs). A PCA analysis was applied to a matrix with all lipid species, to highlight possible changes in the total lipidome imposed by treatments, from which scores plot (Fig. 4A) and loadings plot (supplementary Figure S1-A available as Supplementary Material) of the two principal components were obtained. PCA did not differentiate treatment conditions and there was a higher degree of variability within the CT group (and, to a lesser extent, in [N,P] high ) compared with the other treatments.
Following this observation, PLS-DA was used to maximize the separation between conditions and the projection plot (Fig. 5A) revealed some degree of discrimination between [N,P] low and both CT and [N,P] high . There was no discrimination between CT and both [N,P] med and [N,P] high . The Variable Importance in Projection (VIP) scores were used to rank variables in terms of their importance in the projection of the PLS model, and the top 20 variables are presented in Fig. 5D PCA and PLS-DA were also applied to matrices composed of only PL or GL molecular species, in order to decipher if any one of these two major lipid groups were changing more markedly than the other. The PCA of PLs ( Fig. 4B; loadings in supplementary Figure S1-B) showed a very similar trend to the PCA plotted for the total lipidome data, suggesting no clear discrimination between treatment conditions. Similarly, the PLS-DA (Fig. 5B) also resembles the one obtained with the total lipidome matrix, with the most important species influencing the model also being PA 34:1, PA 36:3, PA 34:2 and PI 36:6 ( Fig. 5E).
The PCA of GLs (

Polar lipid changes in each class.
A univariate analysis was also performed regarding the average relative abundance of molecular species within each class of PLs and GLs. In the case of PLs, most of the significant differences in species relative abundance were observed between [N,P] low and either CT or [N,P] high or both conditions. These were observed in PCs (8 species), PEs (2), PGs (8), PIs (5), LPCs (2) and LPEs (2) (supplementary Figure S4). In PAs, 5 species were significantly different in terms of relative abundance between CT and the other  www.nature.com/scientificreports www.nature.com/scientificreports/ treatments. In the case of GLs, differences were mostly observed between [N,P] low and either CT or [N,P] high or both, in DGDGs (4 species), MGDGs (4) and SQDGs (5) (supplementary Figure S5

Discussion
The present study evaluated the polar lipidome signature of leaves from hydroponically grown H. portulacoides under different concentrations of N and P. The aim was to describe and reveal possible shifts in the lipidomic profile of its leaves under a wide range of DIN and DIP concentrations that represent possible IMTA contexts.   The total lipids extracted from H. portulacoides leaves ranged, on average, between 6.2-9.4 g 100 g −1 DW, with higher amounts extracted from the high-input treatments, including CT (Hoagland's solution). A recent study by Patel et al. (2019) 47 analyzed the total lipid content in the shoots (which can include leaves and other superior organs) of several halophytes and concluded that non-succulent halophytes (e.g. Sporobolus virginicus (L.) Kunth and Aeluropus lagopoides (L.) Trin. ex Thwaites) presented higher lipid content, between 5.5-7.2 g 100 g −1 fresh weight (FW), followed by shrubby halophytes (e.g. Atriplex nummularia Lindl. and Atriplex griffithii Moq.) with 2.6-2.8 g 100 g −1 FW, and succulents (e.g. Sesuvium portulacastrum (L.) L. and Salicornia brachiate Miq.), with 1.5-1.8 g 100 g −1 FW. Given H. portulacoides leaves have a moisture content of ~90%, the equivalent amount of lipid extract in fresh leaves ranges from 0.76-0.94 g 100 g −1 FW, which is much lower than the values mentioned above for other species. Nonetheless, other studies reported values (both in DW and FW) within the same order of magnitude as H. portulacoides, which contradict the values reported above. For instance, Salicornia bigelovii Torr. was reported to have 0.37 g of total lipids 100 g −1 FW 48 ; S. ramosissima, 1. PLs and GLs are two major lipid groups present in the total lipid extracts of halophyte leaves, carrying a wide array of fatty acids (FAs), from which α-linolenic acid (C18:3, n−3), palmitic acid (C16:0), linoleic acid (C18:2, n−6) and oleic acid (C18:1, n-9) are the most abundant 22,52,53 . In this study, H. portulacoides displayed a profusion of lipid species with C16 and C18 chains and some species exhibited polyunsaturated fatty acids with up to four double-bonds (e.g. MGDG 18:2/16:4; PG 16:0/18:4 and PG 18:2/18:4). Polyunsaturated FAs have been largely associated with beneficial health effects in humans and animals [54][55][56][57][58] and H. portulacoides leaves can be a good source for obtaining those FAs, given its relatively high lipid content compared with other halophytes.
The quantities of GLs and PLs in the lipid extract were comparable across treatments. The GLs constituted approximately 46% of the total extract, PLs constituted around 18% and, therefore, polar lipids constituted 64% of the total. When the concentrations were expressed in relation to the dry weight (DW) of leaves, GLs and PLs turned out to be significantly lower in [N,P] low treatment. Higher percentages of polar lipids were previously observed in other halophytes, especially GLs. For instance, the total lipidic extract of chloroplast-enriched portions from Salicornia perennans Willd. was reported to have 67% GLs and 31% PLs; Limonium gmelinii (Willd.) Kuntze, 60% GLs and 32% PLs; and Artemisia santonicum L., 80% GLs and 15% PLs 59 . Other halophytes displayed lower percentages of GLs, such as Halostachys caspica C.A.Mey. and Halocharis hispida (Schrenk) Bunge, with 22-29% GLs and 16-17% PLs in their extract 60 . In terms of the amounts of GLs and PLs per unit of leaves, a previous study reported values for several halophyte species to range between 5-47 mg GLs g −1 DW and 2-17 mg PLs g −1 DW 61 . Concerning the present study, H. portulacoides from the high-input treatments displayed values very similar to the upper-end of those ranges. Nonetheless, comparisons should be taken merely as an illustration of the range of possible concentrations in the edible portions of different species of halophytes.
Regarding the lipidome, the null hypothesis under test stated that no changes occur in the polar lipidome of H. portulacoides cultivated in low-, medium-and high-input of DIN and DIP. Since P is an important element of polar lipids, hydroponic conditions that would offer limited access of this element to the plant could promote alterations in the lipidome of leaves. A decrease in PLs in parallel with an increase in non-phosphorus GLs (e.g. SQDG and DGDG) and betaine lipids (in algae) was observed in plants and algae when exposed to conditions of P-limitation, as previously reported for Arabidopsis 62-65 , rice 66 , oat 67 , soybean 66,68 , periphyton 69 and Ulva 70 . From the PCA results it follows that the total lipidome signature of H. portulacoides leaves remained relatively www.nature.com/scientificreports www.nature.com/scientificreports/ unchanged across groups after long-term exposure to nutrient concentrations varying between 6-100 mg DIN-N L −1 and 0.8-15.5 mg DIP-P L −1 . However, a sequential overlap of treatment groups, from lowest to highest P-input ([N,P] low < [N,P] med < [N,P] high < CT), was evident in the PLS-DA projection. The [N,P] low group stood out as the group with the least amount of overlap with the other groups. When GLs were analyzed separately from PLs, both PCA and PLS-DA plots evidently discriminated the [N,P] low group from both the CT and [N,P] high groups. These results suggest that the GL profile of the leaves is changing according to the availability of P, but its effect on the total polar lipidome is masked by the PL profile which remains relatively stable across treatments. Therefore, at low-input concentrations of P, the leaves of H. portulacoides display a low degree of lipidome remodulation associated with significant changes in GLs. In plants, GLs are typically found in chloroplast thylakoids, being their major lipid constituents, but under P-limited conditions, GLs (particularly DGDG) can partially replace PLs in extra-plastidial membranes 71   The PLs that allowed some discrimination between groups were PA 34:1, PA 36:3 PA 34:2, PI 36:6 and PC 36:6. The PAs displayed higher intensity as the input of P increased meanwhile PI 36:6 and PC 36:6 displayed an opposite pattern. This is also evident from PAs relative abundances, as PA 34:1, PA 34:2 and PA 36:3 were more abundant in CT than the other groups, meanwhile C36:6 species were generally in lower abundance in CT. In plants, PAs are precursors of PL and GL synthesis and also function as signal molecules of environmental stress 79 . The marked differences in abundance of several PA species between CT and the other treatments could be related to the activation of different metabolic pathways mediated by the availability of P. Moreover, certain FA configurations were constantly associated with lipid species that displayed significant differences in relative abundance (e.g. C34:1, C34:2, C34:3 and C36:6). For instance, C34:3 displayed the highest abundance in [N,P] low across all classes of PLs (except PA) and GLs. In general, variations in relative abundance were observed most evidently between [N,P] low and both the CT and [N,P] high treatments, suggesting a possible metabolic adaptation from high-input to low-input conditions.
Following the observations discussed above, H. portulacoides was probably under some level of nutrient limitation under [N,P] low . Firstly, they exhibited less PLs and GLs per unit weight of leaves. Secondly, GLs were suffering some degree of remodulation. Thirdly, the relative abundances of certain species in each class changed as a function of N and/or P availability, as suggested by their gradual increase (or decrease) from low-input to high-input of N and/or P. Nonetheless, one could argue about the extent of nutrient limitation that H. portulacoides was potentially exposed to under the [N,P] low treatment, by looking at how other plants behaved in similar conditions. For instance, wild specimens of Arabidopsis thaliana (L.) Heynh, exposed to 0.03 mM P (similar to [N,P] low ) during 12 days followed by 4 days without P, were considered P-starved as they clearly exhibited significant decreases in shoot's PLs (PC, PE, PG and PS) and significant increases in shoot's GLs (MGDG, DGDG, SQDG) 80 . Some species of MGDG and DGDG were also found at markedly higher levels in soybean (Glycine max (L.) Merr.) leaves under P-limited conditions 68 . In this experiment, H. portulacoides did not exhibit such patent changes in the lipidome in the low-input conditions, which indicates that plants were not starved. Note, however, that H. portulacoides is a perennial plant and both A. thaliana and G. max are annual plants, and these different life history strategies might affect nutrient utilization and threshold conditions for nutrient-limitation 81,82 . Another important fact to consider is that the impact of P-limitation might not affect leaves homogeneously. For instance, in G. max under P-limitation, there seems to be a mechanism of P-remobilization from older leaves, where differences in the lipidome between limited and non-limited conditions were substantial, to younger leaves, where the lipidome profile between different conditions were very similar 68 . In the present study there was no control regarding leaves' age, as the lipidome was representative of the total pool of leaves from H. portulacoides.
Plants in general display a range of responses to low P, generally referred to as P-starvation responses, that aim to minimize the negative effects of its scarcity in plants (e.g. decreased growth, increased root/shoot ratio, increased root-hair density, increased carboxylate exudation, P-remobilization) [83][84][85][86][87] . Under the conditions of the present experiment, the extent to which H. portulacoides underwent a starvation response under [N,P] low that affected the polar lipids of its leaves was defined by a decrease in total GLs and PLs and some degree of lipid remodulation detected in the GLs pool. The availability of P was still high enough in the low-input treatment to maintain the PLs pool relatively unchanged.
Within an IMTA framework, it is fair to conclude that H. portulacoides is capable of maintaining a fully stable lipidome across a variety of N and P concentrations typical of aquaculture effluents, specifically 20-100 mg DIN-N L −1 and 3-15.5 mg DIP-P L −1 . At lower concentrations (e.g. [N,P] low values: 6 mg DIN-N L −1 and 0.8 mg DIP-P L −1 ) the lipidome of the leaves displays some changes, particularly regarding GLs, as well as generalized decrease in the quantity of polar lipids in the leaves. These changes suggest a metabolic adaptation to the lower nutrient conditions and could be indicative of nutrient limitation. Data on growth performance supports a scenario of nutrient-limited conditions in [N,P] low , as H. portulacoides exposed to those same concentration of N and P produced less biomass than those exposed to higher concentrations (Custódio et al., unpublished data).
Halimione portulacoides appears to be a good candidate for IMTA in terms productivity and nutrient-extraction 44,88 and has the potential to become a valuable co-product with uses in human nutrition 11 and for other applications 13,22,30 . A note should be made, however, about the possibility of halophytes accumulating undesired compounds if these are present in effluents, like metals 89,90 and chemicals used for therapy and prophylaxis in aquaculture 91 . This possibility must be taken into account when selecting halophytes for IMTA, since the accumulation of contaminants in edible plant organs can pose risks to human health 92 and species that do not accumulate or concentrate contaminants mostly in non-edible tissues will be more appropriate from a product-safety perspective. The same concern has been put forward regarding other extractive species (e.g. seaweeds) and changes in regulatory frameworks are necessary to promote safety of new products from IMTA 93,94 .
Determining which nutritional conditions can lead to nutrient-limitation scenarios is important information for future IMTA/halophyte producers, in order to guide nutrition strategies that guarantee a consistent end-product, especially under highly variable nutritional outputs which can occur in aquaculture activities. For researchers, this data can guide the establishment of reference nutritional concentrations for future studies targeting the production of H. portulacoides. Future lipidomic studies in H. portulacoides should also attempt to characterize and quantify seed oils, since these comprise a significant fraction of the aboveground biomass during the reproductive period of this species and could have valuable high-end applications, such as pharmaceuticals, biofuels, detergents, polymers and cosmetics.
Fully characterizing the diversity of lipid species across H. portulacoides tissues and how they change along the production cycle and environmental conditions is of tremendous importance for the commercial exploration of its lipids. This will allow for strategic choices to be made on how to produce it and manipulate its life cycle so to maximize the delivery of value-added compounds with commercial applications and consequently increase its economic value.  1″N 8°39′40.0″W) and 500 grafts with 4 nodes each were cut, put into polyethylene containers, irrigated with a modified Hoagland's solution and placed under natural sunlight and temperature to promote root development. The elemental composition of the modified Hoagland's solution was: 60 mg K L −1 , 56 mg N L −1 , 40 mg Ca L −1 , 16 mg Mg L −1 , 16 mg P L −1 , 1.12 mg Fe L-1, 0.34 mg Mo L-1, 0.28 mg B L −1 , 0.13 mg Zn L −1 , 0.11 mg Mn L −1 and 0.03 mg Cu L −1 . After three months, in July 2017, rooted plants were transferred and acclimated to indoor conditions for two weeks. In the second week, plants were progressively exposed to a target water salinity of 20 ppt, with increments of 5 every second day, prior to the beginning of the experiment.  Table S1 in the Supplementary Material available online. The solutions within each unit were continuously aerated with a small aerator to keep oxygen levels high and units were refilled with reverse-osmosis water as needed, to compensate for evapotranspiration. The treatment solutions were renewed weekly, as the retention time for nutrient extraction was set to one week. At the end of the growth trial, the leaves of individual plants were cut out, pooled by hydroponic unit and stored at −80 °C until further analysis. Water temperature and pH were measured regularly with a multi-parameter water quality meter and photosynthetically active radiation (PAR) was measured with a spherical micro quantum sensor (US-SQS/L, Heinz Walz GmbH, Pfullingen, Germany). Average values recorded at the end of each week, before renewal of treatment solutions, are presented as supplementary Table S2.
Leaves lipid extraction. Total lipids were extracted according to the method proposed by Bligh & Dyer (1959) 95 , modified for seaweeds and halophytes 22 . 3.75 mL of chloroform: methanol (1:2, v/v) were added to 100 mg of freeze-dried and grounded leaves followed by 2-minutes vortex stirring and 1-minute sonication. Samples were then incubated on an orbital shaker for 2.5 h, on ice. The homogenates were centrifuged at 2000 rpm for 10 -minutes. The chloroform: methanol extraction followed by centrifugation was repeated twice to improve extraction efficiency. After extraction, 2.3 mL of ultrapure water was added to each supernatant, stirred on the vortex and centrifuged at 2000 rpm for 10-minutes. Two liquid phases originate and the inferior organic phase, which contains the lipids, was recovered and dried under a stream of nitrogen gas. Each dried extract was dissolved in 600 µL of chloroform and transferred to dark vials. Lipid extracts were dried under nitrogen gas, weighed (for 'total lipid' calculation) and stored at −20 °C prior to LC − MS analysis.
Quantification of phospholipids. Quantification of the total phospholipid content was achieved by using the protocol by Bartlett and Lewis (1970) 96 . First, in glass tubes, 125 μl of percloric acid (70 % v/v) was added to dried lipid extracts and the mixtures were incubated during 40 minutes at 170 °C. In the meantime, standards were prepared, also in glass tubes, using 0.1 to 2 μg of phosphorous. After, 825 μl of ultrapure water, 125 μl of ammonium molybdate (2.5 % v/v) and 125 μl of ascorbic acid (10 % v/v) were added to each sample and standards. All tubes were then vortexed. Tubes were incubated in water bath at 100 °C for 10 minutes and transferred to ice to cool down. Absorbance of samples and standards were measured at 797 nm using a microplate reader (Multiskan GO, Thermo Scientific, Hudson, NH, USA).
Quantification of glycolipids. Quantification of the total glycolipid content was achieved using the orcinol assay, as done in our lab 25,97 . First, an orcinol solution (0.2 % v/v in 70 % sulfuric acid) was prepared and 1 mL was added to tubes with N 2 -dried lipid extract samples. Tubes were heated at 80 °C for 20 minutes and transferred to ice to cool down. Absorbance of samples and standards were measured at 505 nm using a microplate reader (Multiskan GO, Thermo Scientific, Hudson, NH, USA). The concentration of glucose was calculated by comparing the data with those of glucose standards (between 0-50 μg prepared from an aqueous solution containing 2 mg mL −1 of glucose and following the same procedure as experimental samples).  98 . Prior to analysis, the lipidomic dataset was normalized by dividing peak-intensity values of each molecular species with the peak-intensity of their respective internal standard (Supplementary Dataset 1). Secondly datasets were created for each lipid class, where the relative abundance of each molecular species was computed for each replicate (Supplementary Dataset 2).

Analysis of polar lipids by high resolution LC-MS and
Prior to the multivariate analysis, data normalization procedures -log-transformation followed by auto-scaling -were employed to decrease the influence of high-concentration metabolites and increase the statistical strength of low-concentration metabolites. Both unsupervised (Principal Components Analysis -PCA) and supervised (Partial Least Squares Discriminant Analysis -PLS-DA) methods were used.
The univariate analysis consisted on the analysis of variance, using the non-parametric Kruskal-Wallis test, of species i) peak-intensities and ii) relative abundance within each class. Post-hoc Dunn's test was used for pairwise comparisons and the Benjamini-Hochberg method was used to control for type-I errors 99 . Significant differences were assumed at a critical p-value <0,05.

Data availability
All data generated or analyzed during this study are included in this published article and its Supplementary Material files.