Glyphosate-Based Herbicide Toxicophenomics in Marine Diatoms: Impacts on Primary Production and Physiological Fitness

Featured Application: Application of non-invasive bio-optical techniques to evaluate the ecotoxicity of glyphosate-based pesticide in marine diatoms with confirmation by classical biochemical tools. Abstract: Glyphosate is the main active component of the commercial formulation Roundup ® , the most widely used chemical herbicide worldwide. However, its potential high toxicity to the environment and throughout trophic webs has come under increasing scrutiny. The present study aims to investigate the application of bio-optical techniques and their correlation to physiological and biochemical processes, including primary productivity, oxidative stress, energy balance, and alterations in pigment and lipid composition in Phaeodactylum tricornutum , a representative species of marine diatoms, using the case study of its response to the herbicide glyphosate-based Roundup ® formulation, at environmentally relevant concentrations. Cultures were exposed to the herbicide formulation representing e ﬀ ective glyphosate concentrations of 0, 10, 50, 100, 250, and 500 µ g L − 1 . Results showed inhibited. Pigment composition and fatty acid proﬁles proved to be e ﬃ cient biomarkers for the highest glyphosate-based herbicide concentrations, while bio-optical data separated controls from intermediate concentrations and high concentrations.

(DHA) (omega-3 [ω-3] class) [33]. Vertebrates cannot synthesize essential fatty acids and the ability to produce LC-PUFA is also limited; they do not obtain these through their diet [34,35]. Furthermore, LC-PUFA plays key roles in animal and human health, being major components of neurological tissues [36].
Photosynthesis has been shown to be impaired by glyphosate-based herbicides within wide ranges of concentrations, including in cordgrass Spartina densiflora from spray application at concentrations of 720-7200 g ha −1 [37], to direct contact of 10-80 mg L −1 in Anabaena fertilissima [38]. In both cases, there was inhibition of CO 2 assimilation and depletion of intermediates from the photosynthetic carbon reduction cycle due to deregulation of the shikimate pathway [39,40]. In such scenarios, non-invasive high-throughput bio-optical screening tools, such as Pulse Amplitude Modulated (PAM) fluorometry, emerge as invaluable techniques to evaluate ecotoxicity in photosynthetic organisms [12,41,42]. Through PAM it is possible to translate the fluorescence signals as proxies to the bioenergetics involved in photosynthesis in a non-destructive form [13,[41][42][43][44] to efficiently assess primary productivity [45,46] and the physiological effects in plant material at different concentrations of contaminants [12,41,42,44].
The half-life of glyphosate in water varies between 45 and 60 days [47] and phytoplankton have a fast response in the presence of contaminants [13,25]. Thus, it is of the utmost importance to evaluate the potential effects of glyphosate in the environment, namely on primary producers. The present study aimed to correlate bio-optical data with the ecotoxicological effects of the exposure of the glyphosate-based herbicide Roundup ® (i.e., currently found in estuarine and marine systems) [48][49][50][51], with the working hypothesis that, similarly to other photosynthetic organisms, this herbicide will have a negative impact on primary productivity, antioxidant enzyme activity, energy balance, pigment and fatty acid composition of the model diatom Phaeodactylum tricornutum, and its potential implications on estuarine or coastal marine ecosystems.

Experimental Setup
The model diatom P. tricornutum Bohlin (Bacillariophyceae) (IO 108-01, IPMA, ALISU-Algae Collection of the University of Lisbon, Lisbon, Portugal) was harvested from monoclonal cultures in 250 mL of f/2 medium [52] in culture flasks (Drechsel-type gas washing bottles) under controlled conditions for 4 days (18 ± 1 • C, under constant aeration and a 14 h light/10 h dark photoperiod). The growth chamber was programmed to simulate sunrise and sunset using a sinusoidal function with a light intensity at noon simulating a natural light environment (RGB 1:1:1, Maximum PAR 80 µmol photons m −2 s −1 , 14/10 h day/night rhythm). According to the Organization for Economic Cooperation and Development (OECD) guidelines for algae bioassays [53] and the recommended initial cell density for microalgae cells with similar size to P. tricornutum, initial cell concentration was approximately 2.7 × 10 5 cells mL −1 . Forty-eight hours after inoculation, cultures were exposed to 0, 10, 50, 100, 250, or 500 µg L −1 glyphosate for 48 h [24,41,46], obtained from the glyphosate-based herbicide "Roundup ® Pronto" containing 7.2 g L −1 of glyphosate. Since no studies have measured glyphosate in marine water, the herbicide concentrations were chosen based on the range of environmental concentrations found in agricultural water streams in relevant literature [49][50][51][52], as it was found that this chemical presents the same half-life in saltwater as it does in freshwater [10,11]. While other substances compose Roundup ® , glyphosate is the main active component. Exposure took place 48 h after inoculation to ensure that the experiment was performed during the cell exponential growth phase [24,41,46]. There were three replicates for each herbicide concentration from a total of 18 experimental units. To avoid contamination, the labware was washed with HNO 3 (20%) for 48 h, rinsed thoroughly with ultra-pure water and autoclaved. All culture manipulations were performed in a laminar airflow chamber using aseptic techniques.

Growth Rates and Cell Harvesting
During the exposure trials, daily cell counting of P. tricornutum subjected to different glyphosate-based herbicide concentrations was performed using a Neubauer improved counting chamber, with an Olympus BX50 (Tokyo, Japan) inverted microscope, at 400-times magnification. Culture growth was determined from the difference between initial and final logarithmic cell densities divided by the exposure period [54], expressed as the mean specific growth rate per day. Samples for photochemical and biochemical analysis were collected after 48 h of exposure to glyphosate-based herbicide (4 days after inoculation). Based on the glyphosate-based herbicide concentrations, we determined the No Observed Effect Concentration (NOEC) and the Lowest Observed Effect Concentration (LOEC) [55]. Furthermore, by application of a sigmoidal dose-response curve to the endpoint measurement for each exposure concentration, we also determined the Effective Concentration (EC) which inhibited growth by 10% (EC 10 ), 25% (EC 25 ), and 50% (EC 50 ) [55]. At the end of the exposure time and after the chlorophyll fluorescence measurements (see next section), samples of 30 mL of culture were centrifuged at 4000× g for 15 min at 4 • C (Sigma 2-16K, Sigma Laborzentrifugen GmbH, Germany). The supernatant was removed, and pellets were immediately frozen in liquid nitrogen and stored at −80 • C until analysis.

Bio-Optical Assessment through Chlorophyll a Pulse Amplitude Modulated (PAM) Fluorometry
Pulse amplitude modulated (PAM) chlorophyll fluorescence measurements were performed using a FluorPen FP100 (Photo System Instruments, Drasov, Czech Republic) on 15 min dark-adapted samples using a 1 mL cuvette. Culture cell density was assessed daily, using a non-actinic light to measure minimum chlorophyll fluorescence (F 0 ). Analysis of chlorophyll transient light curves (Kautsky plot) was carried out using the OJIP test according to [41]. Fluorometric analysis parameters and their description can be accessed in Table 1.

Pigment Analysis
Pigment extraction was performed according to methodologies from previous works [12,13,46]. Pure acetone was added to sample pellets and maintained in an ultra-sound cold bath for 2 min, ensuring total disaggregation of cell material, and kept in the dark at −20 • C for 24 h to prevent degradation. Samples were centrifuged for 15 min at 4000× g and 4 • C and the supernatants were scanned by a dual-beam spectrophotometer from 350 nm to 750 nm at 0.5 nm steps (Shimadzu UV-1603, Shimadzu Co., Kyoto, Japan). Using the SigmaPlot Software, the absorbance spectrum was introduced in the Gauss-Peak Spectra (GPS) fitting library. Pigment analysis was performed according to [56], allowing the detection of chlorophyll a and c, pheophytin a, β-carotene, fucoxanthin, diadinoxanthin (DD), and diatoxanthin (DT).

Antioxidant Enzyme Assays
The soluble protein fraction was extracted at 4 • C from cell pellets in 1 mL of 50 mM sodium phosphate buffer (pH 7.6) with 0.1 mM Na-EDTA and placed in an ultrasound bath for 1 min from the previously collected pellets. The homogenate was centrifuged at 10,000× g for 10 min at 4 • C to remove debris and the supernatant collected to a new tube. Protein concentration was determined according to [57]. Catalase (CAT) activity was measured according to [58], monitoring H 2 O 2 consumption and the consequent decrease in absorbance at 240 nm (ε = 39.4 mM −1 cm −1 ). The reaction mixture contained 50 mM of sodium phosphate buffer (pH 7.6), 0.1 mM of Na-EDTA, and 100 mM of H 2 O 2 with the reaction being started by the addition of 100 µL of extract. Ascorbate peroxidase (APX) was assayed according to [59]. The reaction mixture contained 50 mM of sodium phosphate buffer (pH 7.0), 5 µM of H 2 O 2 , and 0.25 µM L-ascorbate, and the reaction was also initiated with the addition of 100 µL of the extract. The activity was recorded as the decrease in absorbance at 290 nm and the amount of ascorbate oxidized calculated from the molar extinction coefficient (ε = 2.8 mM −1 cm −1 ). Superoxide dismutase (SOD) activity was assayed according to [60] by monitoring the reduction of pyrogallol at 325 nm. The reaction mixture contained 50 mM of sodium phosphate buffer (pH 7.0) and 0.24 mM of pyrogallol and ultra-pure water, with the reaction being started by the addition of 10 µL of extract. Control assays were done in the absence of substrate to evaluate the autoxidation of the substrates. All the assays were performed at 25 • C in a UV500 UV-Visible Spectrometer (Unicam, Waltham, MA, USA).

Lipid Peroxidation Analysis
Lipid peroxidation products were determined according to [61]. Sample pellets were homogenized briefly in 1.5 mL of 10% (v/v) Trichloroacetic acid (TCA), containing 0.4% (w/v) thiobarbituric acid (TBA) and placed in an ultrasound bath for 1 min. The reaction was conducted at 100 • C for 30 min; immediately after it was halted through placement in ice, and after centrifugation at 15,000× g for 10 min at 4 • C, 1 mL of the supernatant was collected and mixed with 1 mL of 0.4% TBA and incubated again under the same conditions. After again cooling down in ice and centrifuging, the absorbance at 532 nm and 600 nm of the supernatant was recorded by spectrophotometry. The concentration of malondialdehyde (MDA) was determined using the molar extinction coefficient (ε = 155 mM −1 cm −1 ).

Fatty Acid Profiles
The fatty acid analysis was performed according to [46] by direct trans-esterification of sample pellets, in freshly prepared methanol sulfuric acid (97.5:2.5, v/v), at 70 • C for 60 min, using the internal standard pentadecanoic acid (C15:0). Fatty acid methyl esters (FAMEs) were recovered using petroleum ether, dried with an N 2 flow, and re-suspended in an adequate amount of hexane. Through gas chromatography (Varian 430-GC gas chromatograph equipped with a hydrogen flame ionization detector set at 300 • C, Middelburg, The Netherlands), 1 µL of the FAME solution was analyzed, setting the injector temperature to 270 • C, with a split ratio of 50. The fused-silica capillary column (50 m × 0.25 mm; WCOT Fused Silica, CP-Sil 88 for FAME; Varian, Middelburg, The Netherlands) was maintained at a constant nitrogen flow of 2.0 mL min −1 and the oven set to 190 • C. Fatty acids identification was performed by comparison of retention times with standards (Sigma-Aldrich) and chromatograms analyzed by the peak surface method, using the Galaxy software. To determine the membrane saturation levels, the double bond index (DBI) was calculated according to [46]:

Energy Balance
Cell pellets were homogenized by ultrasonication on ice (3x 10s at A = 20%) using 1 mL of Milli-Q water. Aliquots were taken from each sample for the analysis of lipid, carbohydrate and protein contents, and electron transport system (ETS) activity. In all assays, ultrapure water was used as reaction blank. The spectrophotometric measurements were performed in triplicates, at 25 • C, using a synergy H1 Hybrid Multi-Mode microplate reader (Biotek ® Instrument, Winooski, VT, USA).

Energy Available
The energy available (Ea) was measured by determining the total protein, carbohydrate, and lipid contents and transforming the results into energetic equivalents (combustion energies: 17,500 mJ mg carbohydrates −1 , 24,000 mJ mg protein −1 , and 39,500 mJ mg lipid −1 ) [61]. Extraction and quantification of total lipids, proteins, and carbohydrates were performed according to [62,63] with minor modifications [64]. Cell pellets were resuspended in 50 mM sodium phosphate buffer (pH 7) containing 1 mM phenylmethylsulfonyl fluoride (PMSF). Cell disruption was performed with 0.42-0.6 mm glass beads (Sigma-Aldrich) for 15 min at 6.5 ms -1 (FastPrep-24, MP Biomedicals). Cell extract was centrifuged at 10,000× g for 20 min at 4 • C. The supernatant was stored at −80 • C until further analysis. Total protein content in the samples was determined using Bradford's method [56]. Total lipids were extracted by adding 250 µL of chloroform (spectrophotometric grade, Sigma-Aldrich), 250 µL of methanol (spectrophotometric grade, Sigma-Aldrich) and 125 µL Milli-Q water to 150 µL of the sample. After centrifugation at 1000× g for 5 min, the organic phase and interphase were removed and 500 µL of H 2 SO 4 was added to 100 µL of lipid extract and charred for 15 min at 200 • C. The mixture was cooled down to 20 • C, 1.5 mL of deionized water was added, and total lipid content was determined by measuring the absorbance at 375 nm and compared to a calibration curve using tripalmitin values as standard. Total carbohydrate content was determined by adding 50 µL of 15% TCA to the 150 µL of sample and subjected to −20 • C for 10 min. After centrifugation at 1000× g for 10 min, the total carbohydrate content of the supernatant fraction was quantified by adding 50 µL of 5% (v/v) phenol and 200 µL of 18 M H 2 SO 4 to 50 µL extract [65]. Following 30 min of incubation at 20 • C, absorbance was measured at 492 nm and compared to a calibration curve using glucose as standard.
The cellular energy consumption (Ec) was determined by using the ETS data (for each 2 µmol of INT-formazan formed, 1 µmol of O 2 was consumed in the ETS), transforming the calculated quantity of oxygen consumed into energetic equivalents by using the specific oxyenthalpic equivalents for an average lipid, protein, and carbohydrate mixture of 480 kJ mol O 2 −1 [62].

Cellular Energy Allocation
The cellular energy allocation (CEA), a methodological approach that integrates the energy available and energy consumption of an organism, was standardized to 10 6 cells and calculated based on measurements of lipid, carbohydrate, and protein content and ETS activity for each sample as follows [67]: where: Ec (energy consumption) = ETS activity mJ h −1 10 −6 cells (4)

Statistical Analysis
Each variable was evaluated through one-way ANOVA with Tukey's multiple comparisons test (GraphPad Prism 6.03 for Windows, GraphPad Software, San Diego, CA, USA), regarding differences among glyphosate-based herbicide concentrations. The data obtained from the Kautsky plots, pigment, and fatty acid profiles, were used as the basis for the construction of the respective resemblance matrixes based on the Euclidean distances between samples. To classify and separate the different treatment groups, statistical multivariate models based on the Kautsky plot, pigment composition, and fatty acid profile variable were generated using Canonical Analysis of Principal Coordinates (CAP), through the non-parametric multivariate analysis packages in Primer 6 software as described previously in other works [12,45,68,69].

Cell Growth Rates
Regarding cell density, after 48 h exposure P. tricornutum cultures were negatively affected, in particular by the two highest concentrations of glyphosate-based herbicide (250 and 500 µg L −1 ), with slight decreases in cell density observed only at lower concentrations (50 and 100 µg L −1 ) ( Figure 1A).

Cellular Energy Allocation
The cellular energy allocation (CEA), a methodological approach that integrates the energy available and energy consumption of an organism, was standardized to 10 6 cells and calculated based on measurements of lipid, carbohydrate, and protein content and ETS activity for each sample as follows [67]: where:

Statistical Analysis
Each variable was evaluated through one-way ANOVA with Tukey's multiple comparisons test (GraphPad Prism 6.03 for Windows, GraphPad Software, San Diego, CA, USA), regarding differences among glyphosate-based herbicide concentrations. The data obtained from the Kautsky plots, pigment, and fatty acid profiles, were used as the basis for the construction of the respective resemblance matrixes based on the Euclidean distances between samples. To classify and separate the different treatment groups, statistical multivariate models based on the Kautsky plot, pigment composition, and fatty acid profile variable were generated using Canonical Analysis of Principal Coordinates (CAP), through the non-parametric multivariate analysis packages in Primer 6 software as described previously in other works [12,45,68,69].

Cell Growth Rates
Regarding cell density, after 48 h exposure P. tricornutum cultures were negatively affected, in particular by the two highest concentrations of glyphosate-based herbicide (250 and 500 µ g L −1 ), with slight decreases in cell density observed only at lower concentrations (50 and 100 µ g L −1 ) ( Figure 1A). However, after 48 h exposure only decreases in specific growth rates, with lower divisions per day and higher doubling time, were observed at the highest glyphosate-based herbicide concentrations (>250 µ g L −1 ) ( Figure 1B). Therefore, a negative effect was observed in the two highest herbicide formulation concentrations regarding all the determined growth parameters. Based on the However, after 48 h exposure only decreases in specific growth rates, with lower divisions per day and higher doubling time, were observed at the highest glyphosate-based herbicide concentrations (>250 µg L −1 ) ( Figure 1B). Therefore, a negative effect was observed in the two highest herbicide formulation concentrations regarding all the determined growth parameters. Based on the glyphosate-based herbicide concentrations present in the environment surrounding the diatom population used in the experiment, we determined the respective NOEC (10 µg L −1 ), the LOEC (50 µg L −1 ), the EC 10 (15.4 µg L −1 ), the EC 25 (94.4 µg L −1 ), and the EC 50 (225.9 µg L −1 ).

Bio-Optical Assessment of Diatom Photochemistry
Apart from the lowest concentration (10 µ g L −1 ), the Kautsky plots showed decreasing fluorescence values with increasing herbicide quantities (Figure 2), particularly at the highest glyphosate-based herbicide concentrations (250 and 500 µ g L −1 ).   Figure S1). Regarding the oxidized quinone pool size, there was a decrease only in the highest glyphosate-based herbicide concentrations (250 and 500 µg L −1 ). However, there were no significant changes in the number of Q A redox turnovers until maximum fluorescence was reached (N), except for the highest concentration in which enhancement was observed. The same pattern occurred in the energy needed to close (reduce) all RCs (S M ). At the highest glyphosate-based herbicide exposure concentrations (250 and 500 µg L −1 ), a decrease in the probability of a PS II chlorophyll molecule functioning as an RC (γ RC ) was detected. However, no differences were observed in the Q A reduction rate (M 0 ). The analysis of the photochemical process from light-harvesting electronic transport, represented by the four main energy fluxes, demonstrates the effects of exposure of the different herbicide concentrations in P. tricornutum cultures (Figure 3). While the active OECs showed a decrease only under the highest glyphosate-based herbicide concentration (500 µg L −1 ), the P G , the grouping probability that correlates with the disconnection between the two PS II units, increased with the highest glyphosate-based herbicide concentration (250 and 500 µg L −1 ) (Figure 4).
Between PS II and PS I, photochemical processes showed a significant decrease in the contribution of light (TR 0 /DI 0 ) and dark (ψ 0 /1 − ψ 0 ) reactions of the photochemical cycle in the highest glyphosate-based herbicide concentration (250 and 500 µg L −1 ). A similar pattern was observed in the reaction center density within the PS II antenna chlorophyll bed (RC/ABS). On the other hand, at the PS I level there was a significant enhancement in the activity of this photosystem (δ R0 /1 − δ R0 ) in response to the highest glyphosate-based herbicide concentrations, also leading to an increase in the equilibrium constant for the redox reaction between both photosystems towards the PS II (ψ E0 /(1 − ψ E0 )). Furthermore, regarding PS I, intrinsic changes led, in turn, to an increase of the electron transport from PQH 2 to the reduction of the PS I end acceptors (RE 0 /RC).
In the rapid light curve (RLC)-derived parameters, the photosynthetic efficiency (α) and the maximum electron transport rate (ETR max ) only decreased with the highest glyphosate-based herbicide concentration (500 µg L −1 ), with no effect on photoinhibition (β) and light saturation (E k ), with only a slight reduction at 250 µg L −1 glyphosate-based herbicide exposure in the latter parameter (Supplementary Figure S2).  (H), contribution or partial performance due to the light reactions for primary photochemistry (TR 0 /DI 0 )), in Phaeodactylum tricornutum following a 48 h exposure to a glyphosate-based herbicide formulation in different concentrations (mean ± s.d., n = 3, different letters indicate significant differences at p < 0.05).
with only a slight reduction at 250 µ g L −1 glyphosate-based herbicide exposure in the latter parameter (Supplementary Figure S2).

Diatom Antioxidant Enzymes
The total protein content in P. tricornutum only showed a significant increase at the highest glyphosate-based herbicide concentration (500 µ g L −1 ) ( Figure 6A). Regarding the antioxidant enzymes CAT, APX, and SOD ( Figure 6B-D), only the latter showed statistical differences and only at the highest glyphosate-based herbicide concentration (500 µ g L −1 ), where activity decreased significantly.

Diatom Antioxidant Enzymes
The total protein content in P. tricornutum only showed a significant increase at the highest glyphosate-based herbicide concentration (500 µg L −1 ) ( Figure 6A). Regarding the antioxidant enzymes CAT, APX, and SOD ( Figure 6B-D), only the latter showed statistical differences and only at the highest glyphosate-based herbicide concentration (500 µg L −1 ), where activity decreased significantly.

Diatom Lipid Peroxidation and Fatty Acid Profile
Lipid peroxidation in P. tricornutum increased greatly after being subjected to the two highest glyphosate-based herbicide concentrations (250 and 500 µ g L −1 ) (Figure 7).

Diatom Lipid Peroxidation and Fatty Acid Profile
Lipid peroxidation in P. tricornutum increased greatly after being subjected to the two highest glyphosate-based herbicide concentrations (250 and 500 µg L −1 ) (Figure 7).  catalase (CAT, B), ascorbate peroxidase (APX, C), and superoxide dismutase (SOD, D) enzymatic activities in Phaeodactylum tricornutum following a 48 h exposure to a glyphosate-based herbicide formulation in different concentrations (mean ± s.d., n = 3, different letters indicate significant differences at p < 0.05).

Diatom Lipid Peroxidation and Fatty Acid Profile
Lipid peroxidation in P. tricornutum increased greatly after being subjected to the two highest glyphosate-based herbicide concentrations (250 and 500 µ g L −1 ) (Figure 7).  The total fatty acid content of P. tricornutum increased about two-fold after exposure to the two highest concentrations of the herbicide formulation for 48 h (Figure 8). The fatty acid (FA) profile was also affected by the same concentrations (Figure 9).
Appl. Sci. 2020, 10, x FOR PEER REVIEW 13 of 23 The total fatty acid content of P. tricornutum increased about two-fold after exposure to the two highest concentrations of the herbicide formulation for 48 h (Figure 8). The fatty acid (FA) profile was also affected by the same concentrations ( Figure 9).  The eicosapentaenoic acid (EPA) (20:5), an omega-3 fatty acid, showed a significant decrease at the highest glyphosate-based herbicide concentration, whilst triunsaturated hexadecatrienoic acid (16:3), exclusively present in plastidial lipids, increased (Figure 9).
While no significant differences were observed for FA major classes and derived ratios (Supplementary Figures S3 and S4), a tendency for an increase was observed in the omega 6/omega 3 ratio at the two highest concentrations of herbicide formulation (Supplementary Figure S5).

Energy Balance
Regarding the energy balance in P. tricornutum, the two highest glyphosate-based herbicide concentrations (250 and 500 µg L −1 ) triggered an increase in Ea (within proteins, lipids, and carbohydrates) and ETS, with CEA also higher in those concentrations and statistically different from control and the lowest glyphosate-based herbicide concentration (10 µg L −1 ) but not from intermediate ones (50 and 100 µg L −1 ) (Figure 10).
Appl. Sci. 2020, 10, x FOR PEER REVIEW 14 of 23 While no significant differences were observed for FA major classes and derived ratios (Supplementary Figures S3 and S4), a tendency for an increase was observed in the omega 6/omega 3 ratio at the two highest concentrations of herbicide formulation (Supplementary Figure S5).

Energy Balance
Regarding the energy balance in P. tricornutum, the two highest glyphosate-based herbicide concentrations (250 and 500 µ g L −1 ) triggered an increase in Ea (within proteins, lipids, and carbohydrates) and ETS, with CEA also higher in those concentrations and statistically different from control and the lowest glyphosate-based herbicide concentration (10 µ g L −1 ) but not from intermediate ones (50 and 100 µ g L −1 ) ( Figure 10).

Multivariate Classification
The multivariate CAP analysis using the Kautsky plot, the pigment composition, and the fatty acid profile of P. tricornutum cultures exposed to different glyphosate-based herbicide concentrations revealed two different classifications, clearly dependent on the class of biomarkers applied ( Figure  11).

Multivariate Classification
The multivariate CAP analysis using the Kautsky plot, the pigment composition, and the fatty acid profile of P. tricornutum cultures exposed to different glyphosate-based herbicide concentrations revealed two different classifications, clearly dependent on the class of biomarkers applied (Figure 11). The CAP plot based on the Kautsky plot allowed the separation in three groups: (1) Control, (2) intermediate glyphosate-based herbicide concentrations (10 to 100 µ g L −1 ), and (3) highest glyphosate-based herbicide concentrations (250 and 500 µ g L −1 ). On the other hand, the CAP analysis of both pigment composition and fatty acid profile produced similar group classifications, clustering in one group control and intermediate glyphosate-based herbicide concentrations (10 µ g L −1 to 100 µ g L −1 ) but separating into distinct groups the highest glyphosate-based herbicide concentrations (250 and 500 µ g L −1 ). The Kautsky plot CAP analysis produced the model with the highest overall classification efficiency (72.2%), followed by the pigment content (61.1%) and lastly the fatty acid profile (44.4%).

Discussion
The cell density of Phaeodactylum tricornutum decreased with exposure to increasing concentrations of glyphosate-based herbicide concentrations with an EC50 of 225.9 µ g L −1 . A comparable effect was observed with similar glyphosate concentrations in freshwater diatoms Gomphonema parvulum, Achnanthes minutissima, and Amphora veneta [19] and some cyanobacteria such as Planktothrix agardhii and Microcystis aeruginosa [70]. Besides the potential direct effect of this herbicide, it is also possible that, due to its chelating properties, the bioavailability of macro-and micronutrients (e.g., divalent cations including calcium, magnesium, manganese, and iron) is reduced, limiting growth [71]. Current environmental concentrations and the unrestrained use of glyphosate-based products pose a continued threat to diatom populations and likely bottom-up effects in trophic webs that need to be further explored. The effect of the glyphosate-based mixture

Discussion
The cell density of Phaeodactylum tricornutum decreased with exposure to increasing concentrations of glyphosate-based herbicide concentrations with an EC 50 of 225.9 µg L −1 . A comparable effect was observed with similar glyphosate concentrations in freshwater diatoms Gomphonema parvulum, Achnanthes minutissima, and Amphora veneta [19] and some cyanobacteria such as Planktothrix agardhii and Microcystis aeruginosa [70]. Besides the potential direct effect of this herbicide, it is also possible that, due to its chelating properties, the bioavailability of macro-and micronutrients (e.g., divalent cations including calcium, magnesium, manganese, and iron) is reduced, limiting growth [71]. Current environmental concentrations and the unrestrained use of glyphosate-based products pose a continued threat to diatom populations and likely bottom-up effects in trophic webs that need to be further explored. The effect of the glyphosate-based mixture on P. tricornutum was very clear in several photochemical-related variables and showed lower photosynthetic efficiency, particularly at the highest concentrations of the glyphosate-based herbicide. However, exposure to the highest concentration led to an increase in all pigments, particularly in chlorophyll content, in opposition to what has been described for higher plants [72] and is likely a compensatory mechanism by which the diatoms counteract the lower efficiency of the affected reaction centers. In other works [73][74][75][76][77], an increase in chlorophyll a concentration was observed in algae communities exposed to low concentrations of atrazine (10 to 32 µg L −1 ) but not at higher concentrations. Thus, this rise in chlorophyll content associated with the increase in fucoxanthin content may provide cells with the ability to counteract the decrease in the number of RC/CS, and therefore an increase in the efficiency of each RC available to trap photons occurs. In P. tricornutum, the PS II light-harvesting complex is composed of the fucoxanthin-chlorophyll protein (FCP) complex, which may deregulate the light-harvesting capacity, increasing both the ability to trap photons and the number of RC available for reduction [46]. However, there was an increase in photoprotective pigments, including β-carotene, diadinoxanthin, and diatoxanthin, which often allows the cells to increase the energy dissipation through non-photochemical processes, preventing possible photoinhibitory events. Moreover, a decrease in the contribution of the dark reactions from Q A − to plastocyanin reinforces the hypothesis that glyphosate at higher concentrations may have nefarious effects on the responses studied, mainly at the ETC level in diatoms [45].
Regarding the fatty acid analyses, the most striking result was the significant increase in the total lipid content, which was nearly twice that in control when cells were exposed to the two highest herbicide concentrations. These increases in microalgae lipid contents are commonly associated with stress conditions, such as nutrient limitation or temperature increase [16,46]. Additionally, exposure to these higher glyphosate-based herbicide concentrations also induced changes in the fatty acid profile of the diatom. It is interesting to note that concomitantly with the increase in the contents of photosynthetic pigments there was also a rise in the relative amounts of 16:3 present in plastidial lipids, namely monogalactosidiacylglycerol, one of the most abundant thylakoid lipids [46]. Changes in chloroplastidial FA composition can lead to alterations of the redox potential [76]. Quinones in the ETC are more efficient at the functional level, showing higher QA reduction rates, which follows the mildly positive effects observed at the PS I level. On the other hand, there was a decrease in the percentage of EPA, usually abundant in all P. tricornutum membranes, which likely translates into an impairment in membrane-dependent reactions, such as transmembrane transport and thylakoid and mitochondrial electron transport. This could explain the lower culture growth rates at the highest glyphosate-based herbicide concentrations. Moreover, since EPA is an omega-3 fatty acid with key roles in vertebrate biochemistry, the decrease in its contents (reflected also in the omega-6/omega-3 ratio) is likely to negatively affect the nutritional quality of the diatom, inevitably impacting all the trophic web [77]. Nevertheless, because P. tricornutum showed tolerance to the lowest glyphosate-based herbicide concentrations (up to 100 µg L −1 ) and is still able to remain alive in the highest concentrations, this can lead to biomagnification induced by glyphosate-based herbicides, leading to a possible vector of contamination and bioaccumulation. This should be addressed in future works due to the possibility of contamination within the trophic web.
If, on one hand, photosynthetic and photoprotective antioxidant pigments increased, which may be also partly explained by the lack of changes in CAT and APX activities, on the other hand, the decreased activity of SOD can lead to the accumulation of lethal superoxide anions and consequent oxidation of several biological components, such as proteins, lipids, and EPA fatty acid [78]. However, a decrease in the energy transduction at the ETC, from Q A to plastoquinone, originated a lethal situation of increased redox accumulation in the affected photosystems [79]. Therefore, the energy reaching the PS I was lower, although the effectiveness of the energy transport from PQH 2 to the PS I was not affected (RE 0 /RC), and even increased under the highest glyphosate-based herbicide concentrations.
Yet, a severe shift in the redox equilibrium between photosystems towards the PS II (ψ E0 /(1 − ψ E0 )) was still observed, decreasing the redox potential reaching the PS I, and thus impairing the light and dark reactions of photosynthesis and the renewal of substrates in the PS I [45].
Nevertheless, the available energy (Ea) increased in the highest glyphosate-based herbicide concentrations, particularly at 500 µg L −1 , probably due to an increase in lipid and protein concentrations. The increase in Ea has been observed in green algae, which accumulated energy reserves as a response to environmental stress [64,80,81]. A possible explanation could be related to the need for algae to counteract the increase in energy expenditure, as observed in the present study. Therefore, following a similar pattern as Ea, the energy consumption demonstrated by the mitochondrial ETS of P. tricornutum increased significantly at the highest concentrations, reaching a 300% increase in the highest concentration relative to control. This increase in energy consumption has been observed in other microalgae [65]. Nonetheless, the Ea was higher than the ETS, which led to a slight increase of the CEA in the highest concentrations but only allowing a differentiation of the highest concentrations from the control and the lowest concentrations. This increase in CEA may reflect a higher impairment in mitochondria functioning relative to chloroplast functioning that can ultimately impact the latter due to biophysical disturbances [82]. While these energy balances appear to indicate an increase in the energy available for growth or cell division, the highest glyphosate-based herbicide concentrations inhibited diatom culture growth. Based on this data, stress forces the energy to shift from photosynthesis into energy consumption. This may change photosynthetic electron pathways, leading to Ea downregulation and an increase in ETS to avoid carbon waste, as a strategy to handle an excess of metabolic energy. There are different metabolic pathways under which energy spillover can be managed, but current relevant knowledge is still very scarce [32].
The CAP analyses were very efficient in the assessment of the effects of the glyphosate-based herbicide on the photochemistry, pigment composition, and lipid metabolism of P. tricornutum, as well as on the potential of these metabolic features as toxicity biomarkers. However, only the use of bio-optical techniques (e.g., Kautsky plot) allowed an efficient identification of all the different glyphosate-based herbicide exposure treatments (control, low, intermediate, and high concentrations). As in previous studies, multivariate analysis efficiently classified groups subjected to different exposure levels, in comparison to the often-used univariate analysis [12,45,69]. While glyphosate is the main component of Roundup ® , other substances are also in its composition and it is of great interest in future works to assess synergies and/or antagonisms between the herbicide formulation versus glyphosate in its pure form, comparing the values to those measured in the habitats.

Conclusions
Glyphosate-based pesticides, particularly at high concentrations, have a clear effect on several metabolic pathways in marine diatoms (changes in pigment profile, photosynthetic impairment, and decreased antioxidant capacity). Even though photoprotective mechanisms were induced, membrane damage still occurred. These facts highlight the application of pulse amplitude modulated chlorophyll fluorescence bio-optical methods as a promising tool for inclusion in ecotoxicological studies with several advantages relative to the classical biochemical approaches, namely fast data acquisition, lower monetary costs, and a high number of variables obtained while ensuring a high degree of accuracy. Thus, bio-optical tools appear as a fast, inexpensive, and reliable method for toxicophenomic assessment of the impacts of this widespread pesticide in marine diatoms, with a clear dose-response pattern. Moreover, the high volume of data produced by these bio-optical techniques can be efficiently applied in multivariate analysis, providing a classification with a high degree of accuracy. With this in mind, PAM fluorometry appears to be a promising tool for inclusion in ecotoxicological studies aiming towards a possible use in long-range detection through satellite imagery in a global chemical pollution-monitoring network.

Acknowledgments:
The authors would like to thank Caroline Casey and Joana Roma for revising the English language, grammar, punctuation, and spelling of the manuscript.

Conflicts of Interest:
The authors declare no conflict of interest.