Introduction

Arsenic (As) is a well-known toxic element naturally present in the Earth’s crust [1]. Several studies have shown the different responses of microorganisms exposed to this metal, such as the demethylation process in fungi [2], contamination in food [3], and bioaccumulation, transformation, and release in Microcystis [4]. Arsenic is considered to be one of the most toxic environmental substances. Both natural and anthropogenic sources contribute to the presence of As in environmental media [5].

Cyanobacteria are the only known prokaryotes that perform oxygen-evolving photosynthesis, commonly found in diverse environments including marine and freshwater, as well as in terrestrial habitats [6]. They are responsible for about 50 % of earthly photosynthesis, and they are the main contributors of the global oxygen cycle. Metals, such as arsenic, are known to exercise a negative influence on cyanobacteria photosynthesis [7, 8]. Cyanobacteria photosynthetic processes are known to occur in cyanobacteria thylakoid membrane [9]. Lipids of cyanobacteria have an important function in the thylakoid membrane structural components, and consequently, they are vital for their life processes [10]. The lipid composition of cyanobacteria thylakoid membrane is mostly formed by galactolipid species of monogalactosyldiacylglycerol (MGDG) and digalactosyldiacylglycerol (DGDG). These lipids are crucial for the stability and activity of photosynthesis systems [11]. A decrease in their concentration has been related to a reduction of the photosynthetic activity, changes in chloroplast ultrastructure, and growth arrest [12].

Liquid chromatography coupled to mass spectrometry (LC-MS) is currently being employed in lipid analysis due to its high resolution and sensitivity, enabling the simultaneous identification of multiple constituents in the same analyzed sample [13]. However, this analytical methodology, when applied in complex lipidomic studies, generates a huge amount of data in the analysis of a single biological sample [14]. A recently introduced strategy of selection of mass spectrometry regions of interest (ROIs) is proposed to handle the extremely large size of these LC-MS data sets, using affordable computer resources, without mass accuracy loss [15]. The ROI-based method significantly decreases the size of LC-MS data sets, allowing simultaneous analysis of multiple data sets in the study of the same system under different experimental conditions.

Multivariate curve resolution alternating least squares (MCR-ALS) is a chemometric tool applied to resolve multicomponent responses from unknown unresolved mixtures [16–19]. MCR-ALS has been applied to the resolution of elution/concentration and mass spectra profiles for the different components existing in complex cell lipid mixtures analyzed by chromatographic methods [14]. Recently, it has also been applied successfully to resolve complex omics profiling problems, such as in lipidomics [18, 19] and metabolomics [20] studies.

The aim of this work was to evaluate the changes operating on lipidomic profiles of cyanobacteria (Anabaena and Planktothrix agardhii) when they were treated with As(III). As2O3 was selected as a source of As(III) since it is the main precursor of most of As compounds [21]. To accomplish this objective, a recently introduced data analysis strategy which combines the ROI concept in LC-MS together with MCR-ALS was applied. Changing lipids were identified and preliminary biochemical interpretations of the results are discussed.

Materials and methods

Chemicals and solvents

As2O3 was purchased from Sigma. Analytical grade methanol and chloroform were purchased from Merck and Carlo Erba, respectively. HPLC Gradient Grade acetonitrile was purchased from Fischer Chemicals. Lipid standards were obtained from Avanti Polar Lipids.

Cyanobacteria culture and As(III) exposure

Anabaena and P. agardhii were both obtained from Culture Collection of Algae and Protozoa (CCAP1403/21 and CCAP1459/11, respectively; SAMS Limited) and were grown aseptically in 500-mL Erlenmeyer flasks containing BG-11 medium (with NaNO3 as a source of nitrogen) [22] at 25 °C under continuous fluorescent white light with an intensity of 15 μmol photons m−2 s-1, until an optical density of 0.1–0.2 at 750 nm (OD750) was reached. Then, the cultures were transferred to 50-mL Falcon flasks with different concentrations (0, 1, and 2 mM) of As(III); cells were maintained in the mentioned conditions for 7 days. The choice of these As(III) concentrations is based on a previous toxicity test on cyanobacteria (data not shown) in which the concentration 1 mM started to reduce normal cell growth.

Estimation of protein content

Protein content was measured using Pierce™ BCA Protein Assay Kit (Thermo Scientific). Bovine serum albumin was used as a protein standard. In a 96-well plate, 20 μL of every sample was mixed with 100 μL of the working solution (BCA reagents A and B, 1:50) and then incubated at 37 °C for 1 h. Absorbance was measured at 562 nm (Synergy 2 Multi-Mode Reader, Biotek Instruments, Inc.).

Microscopy and chlorophyll spectra

Cyanobacteria were visualized through a Nikon eclipse 90i microscope using a ×60 oil objective, and images were captured with a Nikon DS-Ri1 digital camera. Chlorophyll spectra were measured in a Synergy 2 Multi-Mode Reader (Biotek Instruments, Inc.) UV–Visible spectrophotometer in the range of 300–800 nm.

Lipid extraction

For every sample, 250 mL of cyanobacteria cultured in 500-mL Erlenmeyer flasks was used. Cyanobacteria cultures were centrifuged (5000 rpm, 20 min, 4 °C) and the supernatants were discarded to recover cell pellets. The extraction has been performed as previously described [19]. Briefly, 100 μL of Mili-Q water was added to the cyanobacteria pellets and the suspension was transferred to borosilicate glass tubes. This was followed by the addition of 500 μL of chloroform and 250 μL of methanol in each tube. This mixture was fortified with different internal standards of lipids (C18(plasmalogen)-18:1 phosphatidylcholine, C18(plasmalogen)-18:1 phosphatidylethanolamine, C16(plasmalogen) lysophosphatidylcholine, 1,2,3-17:0 triacylglycerol, 1,3-17:0 D5 diacylglycerol, 17:0 monoacylglycerol, 17:0 cholesteryl ester, 16:0 D31-18:1 phosphatidic acid, 16:0 D31-18:1 phosphatidylcholine, 16:0 D31-18:1 phosphatidylethanolamine, 16:0 D31-18:1 phosphatidylserine, 16:0 D31-18:1 phosphatidylglycerol, 16:0 D31-18:1 phosphatidylinositol, 17:1 lysophosphatidylcholine, 17:1 lysophosphatidylglycerol, 17:1 lysophosphatidylserine, 17:1 lysophosphatidic acid, 17:1 lysophosphatidylcholine, 17:1 lysophosphatidylinositol), 200 pmol each (10 μL of 20 μM stock solutions in absolute ethanol). Next, samples were vortexed and sonicated until they appeared dispersed. In order to increase the efficiency of the extraction, samples were left under agitation overnight in a water bath at 48 °C, as described previously for other lipid extractions [23]. Samples were then cooled and evaporated under N2 stream and subsequently resuspended in 500 μL of methanol. The resulting suspension was then transferred to 1.5-mL Eppendorf tubes and evaporated again. Then, samples were resuspended in 150 μL of methanol and were centrifuged (10,000 rpm, 3 min) to discard all precipitates from the samples. Finally, 130 μL of each sample supernatant was taken and transferred to UPLC vials for LC injection.

Liquid chromatography mass spectrometry

LC-MS analysis was performed using a Waters Acquity UPLC system connected to a Waters LCT Premier orthogonal accelerated time-of-flight mass spectrometer (Waters) operated in positive ion electrospray ionization mode. Full scan spectra from 50 to 1800 Da were acquired, and acquisition of every MS spectrum was taken approximately 0.2 s.

Mass accuracy and reproducibility were maintained by using an independent reference spray via the LockSpray interference. The analytical column was a 100 × 2.1-mm inner diameter, 1.7-μm C8 Acquity UPLC bridged ethylene hybrid (Waters). The two mobile phases were: phase A, H2O 2 mM ammonium formate, and phase B, MeOH 1 mM ammonium formate. Both contained 0.2 % of formic acid. The flow rate was 0.3 mL min−1. Tthe gradient of A/B solvents started at 20:80 and changed to 10:90 in 3 min, from 3 to 6 min remained at 10:90, changed to 1:99 in 6 min until minute 15, remained 1:99 until minute 18, and finally returned to the initial conditions until minute 20. The column was held at 30 °C.

Chemometric data analysis

Raw data files obtained from the LC-MS analysis of every sample were transformed to CDF format by Databridge function of MassLynx™ V. 4.1 software. LC-MS data were uploaded into the MATLAB workspace environment (version R2012b, The Mathworks Inc.) using the functions mzcdfread and mzcdf2peaks of the Bioinformatics Toolbox (The Mathworks Inc.). LC-MS data were arranged and aligned according to their m/z values in a square regular data matrix, with retention times in the rows and selected m/z values in the columns. Different strategies were adapted to build such a data matrix. We have recently proposed a new protocol based on the selection of the ROIs [15]. This strategy allows for the selection of m/z traces whose intensity signals are higher than preselected threshold values (not noise or spurious measurements), and it takes advantage of the sparsity of the measured MS data and of their mass accuracy. Using this approach, full-scan chromatographic MS matrices with the highest experimental mass accuracy are obtained with low storage requirements. A relatively low number (200–500) of high-resolution MS (intensity, m/z) signals were finally taken into account, with more than 100-fold computer storage reduction (Fig. 1).

Fig. 1
figure 1

Example of MS-ROI data matrix augmentation and application of MCR-ALS to the simultaneous analysis of eight individual MS-ROI data matrices: four control samples and four treated samples. See [15] and Eq. 2

Full size image

Every MS-ROI data matrix was then normalized taking into account the areas of the internal standards added and the protein content measured for each sample. This normalization was done by multiplying each matrix by a factor obtained as follows: (200 × mg protein)/mean area of standards, where 200 refers to the picomoles of standards added in the lipid extraction. After normalization, individual MS-ROI matrices obtained in the analysis of every Anabaena and P. agardhii samples were used to build four different augmented matrices. In Fig. 1, the exemplary construction of a MS-ROI augmented matrix was given. Since MS-ROI individual data matrices have different number of ROI m/z values, a preliminary step of ROI rearrangement to consider all those ROI m/z values (common and not common) with significant traces (retention times where MS intensity values are higher than the threshold values at the considered ROI m/z value) is performed. In case an individual matrix does not have a particular significant trace at this ROI m/z value, low random intensity values at the noise level are assigned. See Gorrochategui et al. [15] for more details about MS-ROI augmented matrix building. As shown in Fig. 1, the MS-ROI augmented data matrix D aug is built with four control samples and four treated samples, and the final dimensions of this augmented matrix are the total number of retention times considered in the eight chromatographic runs times the number of finally considered m/z ROI values. The first MS-ROI augmented matrix was built with the four non-treated (control) and the four 1-mM As(III)-treated Anabaena samples. The second MS-ROI augmented matrix had the four non-treated (control) and the four 2-mM As(III)-treated Anabaena samples. The third MS-ROI augmented matrix had the four non-treated (control) and the four 1-mM As(III)-treated P. agardhii samples. Finally, the fourth MS-ROI augmented matrix was built with the four non-treated (control) and the four 2-mM As(III)-treated P. agardhii samples. In order to perform this matrix augmentation column-wise, the columns of every ROI data matrix should have the same m/z ROI values. See Gorrochategui et al. [15] for details on how to implement this in practice.

In this work, the MCR-ALS method [16, 17] has been used for the analysis of the augmented ROI data matrices described above. Since the MCR-ALS method has been described elsewhere [15–17], only a brief explanation is given below. This method is based on the fulfillment of a bilinear measurement model, which assumes that the measured MS spectra along the chromatographic separation are the sum of the MS spectra of the lipids of the investigated system (plus some background, solvent, and instrumental noise contributions) weighted by their relative concentrations. This bilinear model can be described by the simple data matrix equation

$$ \mathbf{D}=\mathbf{C}{\mathbf{S}}^{\mathbf{T}}+\mathbf{E} $$
(1)

In the case of a single LC-MS-ROI data matrix, D contains the MS intensity signals at the selected m/z ROI values (columns) and retention times (rows). C has the elution or concentration profiles of resolved components, and S T has the resolved pure MS-ROI spectra of these components. E has the non-explained variance and experimental error. MCR-ALS solves Eq. 1 by an Alternating Least Squares algorithm which calculates iteratively the concentration, C, and the pure MS-ROI spectra, S T, matrices fitting optimally the experimental data matrix, D. MCR-ALS optimization requires the initial proposal of a number of components and of an initial estimate of either the C or S T matrix. A sufficiently large number of components (for instance, 80 or more) can be initially selected to explain most of the data variance (e.g., more than 90 % of the total data variance) and to properly describe the experimental data features (chromatographic peaks). This number of components can also include some components that are not strictly related to the lipid signals but to the solvent, background, or other detector and instrumental signal contributions. Initial estimates of either the C or S T matrix are obtained from the more distinct rows (MS spectra) or columns (chromatograms) of the data matrix. These are good initial estimates for MCR-ALS due to the high selectivity of the LC-MS data. Once the number of components and initial estimates of either C or S T are provided, during the ALS optimization, various constraints are applied to model appropriately the shapes of the C and S T profiles [24]. For ROI LC-MS data, non-negativity was applied to ensure the concentration and spectral profiles to be positive. Also, during the ALS optimization, the MS spectra were normalized to have all maximum intensity equal to 1, to avoid scale indeterminacies (intensity ambiguity) and to give relative quantitative information in the concentration profiles. Since the MS-ROI spectra are sparse and very selective (a lot of zero values), rotation ambiguities [25] are practically not present in this case. Fitting MCR-ALS results can be performed by using the explained data variance calculated as: R 2(%) = ∑d 2 ij /∑d 2 ij  × 100, where i = 1,…,I (number of data matrix rows, retention values) and j = 1,…,J (number of data matrix columns, number of m/z ROI values), d ij is the MS-ROI individual intensity value in the experimental matrix D, and d* ij is the corresponding value calculated by MCR-ALS using the optimal C and S T matrices finally obtained, as has been shown in previous works [16, 17].

The MCR-ALS bilinear model can be extended to the simultaneous analysis of multiple individual data matrices using the column-wise augmentation strategy described in Fig. 1. In this particular case, the simultaneous analysis of multiple MS-ROI data matrices coming from the same type of cyanobacteria culture (either Anabaena or P. agardhii) under different conditions (control samples and As(III)-treated samples at two concentrations, 1 and 2 mM) was performed using the appropriate MS-ROI column-wise augmented data matrix (see Fig. 1). Equation 1 for MCR analysis of a single data matrix is now extended to Eq. 2 for MCR analysis of the new column-wise MS-ROI augmented data matrices [16, 17].

$$ \left(\begin{array}{l}{\mathbf{D}}_1\\ {}{\mathbf{D}}_2\\ {}{\mathbf{D}}_3\\ {}\dots \\ {}{\mathbf{D}}_{\mathbf{K}}\end{array}\right)=\left(\begin{array}{l}{\mathbf{C}}_1\\ {}{\mathbf{C}}_2\\ {}{\mathbf{C}}_3\\ {}\dots \\ {}{\mathbf{C}}_{\mathbf{k}}\end{array}\right){\mathbf{S}}^{\mathbf{T}}+\left(\begin{array}{l}{\mathbf{E}}_1\\ {}{\mathbf{E}}_2\\ {}{\mathbf{E}}_3\\ {}\dots \\ {}{\mathbf{E}}_{\mathbf{k}}\end{array}\right)={\mathbf{C}}_{\mathrm{aug}}{\mathbf{S}}^{\mathbf{T}}+{\mathbf{E}}_{\mathrm{aug}} $$
(2)

In this equation, the column-wise augmented MS-ROI data matrix, D aug, is decomposed by the bilinear model into the product of a column-wise augmented matrix, C aug, having the elution profiles of the resolved components in every individual MS-ROI data matrix, by the MS spectra matrix, S T, having the pure MS-ROI spectra of these components. E aug is the corresponding matrix which will have the model unexplained variance (for the considered number of components) and experimental error. It is important to remark here that the proposed MCR-ALS data analysis strategy does not imply any shift nor shape assumption nor correction of the chromatographic elution profiles of the same component in the different chromatographic runs. It only needs that the m/z values match among the different MS spectra included in D or D aug and that every component can be defined by a unique pure MS spectrum. This data analysis procedure has high flexibility and resolution power, it takes profit of the high resolution of the MS detector, and it allows for the lack of reproducibility of chromatographic separation conditions.

Once the MCR-ALS results are obtained, the C aug and S T matrices are examined in detail to extract qualitative (identification) and quantitative information and to allow the interpretation of the investigated effects (by As(III) treatment). The elution/concentration profiles resolved by MCR-ALS for every component are in the columns of the C aug matrix for all chromatographic runs simultaneously analyzed. Comparison of the areas of these elution profiles provides the relative quantitative information about a particular resolved component in the different analyzed samples. Comparison of the areas between the control and As(III)-treated samples allows for the statistical assessment of the produced effects at a particular significance level (p values below 0.05). Only those components showing statistically significant differences between their concentrations (peak areas) in the control and As(III)-treated samples are further considered for their chemical compound identification. This identification is performed looking for the MCR-ALS MS-ROI-resolved spectrum corresponding to the same component in the S T matrix. The m/z ROI values of these spectra have full-scan MS mass accuracy and are used for compound identification using the Human Metabolome Database (http://www.hmdb.ca) and Lipid Maps (http://www.lipidmaps.org) online databases. Lipids are assigned from m/z values associated to smaller delta values (difference between query mass and adduct mass), and they were compared with those found in previous works.

Results

Morphological changes and chlorophyll content of cyanobacteria exposed to arsenic

First, morphological changes of cyanobacteria (Anabaena and P. agardhii) were assessed after 7 days of arsenic exposure. As depicted in the optical microscope images of Fig. 2, cyanobacteria filaments showed a visible decrease of green coloration along with the increasing concentration of As(III). At 2 mM As(III), P. agardhii species presented considerable fading of their green pigmentation while Anabaena species showed complete discoloration.

Fig. 2
figure 2

Light microscope pictures of cyanobacteria filaments under As(III) treatment. Anabaena (a) and P. agardhii (b) cyanobacteria species incubated for 7 days with 0, 1, and 2 mM of As(III) in Blue-Green medium

Full size image

Since green pigmentation in cyanobacteria is associated with the composition of photosynthetic pigments like chlorophyll [6], changes in this green pigmentation were first analyzed by UV–Visible spectrometry (Fig. 3).

Fig. 3
figure 3

Chlorophyll absorption spectra of As(III)-treated cyanobacteria. Anabaena (a) and P. agardhii (b) exposed for 7 days to 0, 1, and 2 mM As(III) (brown, blue, and yellow lines, respectively)

Full size image

The absorbance spectra of treated cyanobacteria were recorded. The results showed that absorption bands between 400 and 500 nm were practically eliminated after arsenic treatment, especially in the case of Anabaena (Fig. 3A). Considering that the main peak of chlorophyll is around 450 nm [26], the results suggested that the decrease of green coloration was related to chlorophyll degradation. An important decline of chlorophyll under arsenic exposure has already been described in previous works [4]. Rocchetta et al. [27] also showed that algal cells exposed to heavy metals such as copper and chromium suffered from multiple biochemical alterations affecting the photosynthesis. In this work, in order to study the changes in lipid composition due to As(III) treatment, an exhaustive untargeted lipidomic study was performed.

Chemometric analysis of LC-MS data of cyanobacteria exposed to As(III)

The effects of different concentrations (0, 1, and 2 mM) of arsenic after 7 days of exposure on the two cyanobacteria (Anabaena and P. agardhii) cultures were investigated by LC-MS in positive ion mode. LC-MS raw data were first normalized (using internal standards and total protein content) and compressed according to the ROI strategy (see “Materials and methods”). As mentioned, the use of the ROI strategy allowed substantial reduction of computer storage and processing time without losing mass accuracy [15].

In Fig. 4, the positive ion mode full-scan LC-MS chromatograms (after selection of the ROIs) are compared for samples extracted from Anabaena cultures without any As(III) treatment (Fig. 4A) to those obtained after 7 days of exposure with 1 mM As(III) (Fig. 4B). Distinct features are clearly noticed in these chromatograms already showing the effects of metal treatment.

Fig. 4
figure 4

LC-MS-ROI chromatogram of a single sample (from one single-data matrix). Anabaena control (a) and 1-mM As(III)-treated (b) samples. In the X-axis are the number of retention times in the analysis of one sample (giving one data matrix with 637 rows or retention times; see Fig. 1)

Full size image

In order to investigate the effects on lipid concentration changes in both cyanobacteria cultures due to the As(III) treatment at 1 and 2 mM, four different MS-ROI column-wise augmented data matrices were built. Each new column-wise augmented data matrix contained eight MS-ROI individual data matrices, the first four being the control replicates and the last four the treatment replicates, as explained in “Materials and methods.” In Fig. 5, the first MS-ROI augmented matrix for Anabaena at 1-mM As(III) treatment is shown as an example.

Fig. 5
figure 5

LC-MS-ROI chromatograms of multiple samples (from one column-wise augmented data matrix). On the left are four Anabaena control samples (section marked on pink slashed square). On the right are four 1-mM As(III)-treated samples (section marked on green slashed square). In the X-axis are the total number of retention times in the analysis of the eight samples (giving one column-wise augmented data matrix with 8 × 637 = 5096 rows or retention times; see Fig. 1)

Full size image

The eight chromatographic sections distinguished correspond to the eight investigated samples. As shown in the picture, plots of the LC-MS-ROI data for the control samples (four regions on the left) had higher intensity signals than the plots of the four 1-mM As(III)-treated samples (four regions on the right), suggesting an important loss of lipid species concentrations due to the As(III) treatment.

The augmented LC-MS-ROI data matrices of four control and four treated cyanobacteria samples were then analyzed by MCR-ALS (see Fig. 1), using non-negativity constraints, to resolve elution and spectral profiles of the sample lipid constituents. MS-resolved pure spectra (S T) were also normalized to have the same equal intensity, transferring all relative quantitative information to elution profiles (C). MCR-ALS was performed using a large enough number of components (80) to explain enough data variance and to assure the resolution of a sufficiently large number of elution and MS spectra profiles with reasonable chemical shapes. In the four cases, f, the explained variances were between 93 and 95 %. A small number of the MCR-ALS-resolved components could be related to solvent and background contributions, whereas all other had reasonable chromatographic and spectral shapes and features. Only these ones were investigated in more detail, and their concentration changes between the control and As(III)-treated samples were estimated.

Once the elution profiles corresponding to the same component in the control and treated samples of the same augmented MS-ROI matrix were resolved, they were further analyzed to assess the possible concentration changes due to As(III) treatment. As explained in “Materials and methods,” this was performed by comparison of the peak areas and/or heights of these two types of samples. A statistical Student’s t test was used in every case to assess for the statistical difference between the peak areas/heights of the elution profiles of the same component in the control and treated samples. As an example, the MCR-ALS-resolved elution profiles and mass spectra for two components that presented significant changes between controls and treatments are represented in Fig. 6 for both Anabaena 1-mM and P. agardhii 1-mM augmented matrices. On one hand, in Anabaena 1-mM samples on the left of Fig. 6, component 79 exhibited an important decrease in the treated samples (four sections on the right), as shown in Fig. 6A, with an m/z ROI value of 800.6217 (Fig. 6B), which was further identified as MGDG(36:2). In the case of P. agardhii 1-mM samples on the right of Fig. 6, component 21 presented a similar decreasing profile in the treated samples (Fig. 6E). With an m/z ROI value of 551.4229 (Fig. 6F), this molecule was further identified as 3-cis-hydroxy-carotene. Bar plots of the areas and the heights of the resolved peaks in both examples represented in both cases (Fig. 6C, D, G, H) also confirmed the changes produced by As(III) treatment.

Fig. 6
figure 6

Examples of results obtained in the MCR-ALS simultaneous analysis of multiple LC-MS-ROI chromatograms (from the MCR-ALS analysis of column-wise augmented data matrices having four control samples and four 1-mM As(III)-treated samples; see Eq. 2). Results are given for component 79 (left) in the analysis of Anabaena sample and for component 21 (right) in the analysis of P. agardhii sample. a, c MCR-ALS-resolved elution profiles (C aug in Eq. 2). b, f MCR-ALS-resolved MS pure spectra (S T in Eq. 2) with the m/z ROI values at their intensity maxima. c, g Bar plots of the integrated peaks areas of each replicate. d, h Bar plots of the peaks heights of each replicate

Full size image

In Tables 1 and 2, lipids showing significant changes according to the t test (at the p < 0.05 significance level) are listed with their retention times, measured m/z, explained variance (R 2), identified compounds, adduct, calculated m/z, mass error (in parts per million), area and height fold changes, and p values. These lipids are given and summarized for Anabaena and P. agardhii cyanobacteria cultures treated with 1 and 2 mM As(III), respectively. Most of the listed lipids are associated with cyanobacteria membranes and with pigments related to the photosynthesis process. Interestingly, at 1-mM As(III) treatment (Table 1), a 4- to 5-fold decrease of several MGDG species has been detected for both cyanobacteria species. These galactolipids are associated with cyanobacteria membrane and are known to be involved in thylakoid membrane packing and to provide a mechanism for the separation of the components of the photosystem within the thylakoid membrane [12]. In addition, as mentioned before, we observed a significant loss in the concentrations of pigments related to the photosynthetic system in 1-mM Anabaena samples, such as chlorophyll a, and its sub-product, pheophytin a. Also, we found a large 4-fold decrease of some carotene compounds such as 3-cis-hydroxy-carotene and carotene-3,3′-dione. Carotenes have key roles in light absorption for use in the photosynthesis process, as well as a protective role of chlorophyll from photodamage [28]. In P. agardhii, we also found a significant 5-fold decrease of 3-cis-hydroxy-carotene, as has been shown in the example of Fig. 6. Another interesting feature observed is the decrease of triacylglycerol (51:0), only found in Anabaena samples.

Table 1 Summary of molecules that showed significant changes between untreated and 1-mM As(III)-treated samples for both Anabaena and P. agardhii using the proposed LC-MS-ROI-MCR-ALS methodology
Full size table
Table 2 Summary of molecules that showed significant changes between untreated and 2-mM As(III)-treated samples for both Anabaena and P. agardhii using the proposed UPLC-MS-ROI-MCR-ALS methodology
Full size table

Similar results were obtained for 2-mM As(III) treatment (Table 2). In this case, the decrease of MGDGs that showed a reduction in their levels at 1 mM was more intense in both cyanobacteria. Also, a new MGDG appeared reduced at this concentration. In Anabaena samples, the decline of pheophytin and carotene concentrations was more pronounced (more than 10-fold). For P. agardhii samples, pheophytin a appeared reduced in the treated samples, and the reduction of 3-cis-hydroxy-carotene levels observed at 1 mM was more severe (30-fold). The area fold changes of the more significant features at 1- and 2-mM As(III) treatments for both cyanobacteria species are compared and summarized in Fig. 7.

Fig. 7
figure 7

Areas fold change representation of the common changing lipids observed at 1 and 2 mM of As(III) treatment for the two cyanobacteria species. Common changes between the two As(III) concentrations for Anabaena (a) and P. agardhii (b). The statistical significance of each bar is represented with asterisks: *p < 0.05, **p < 0.01, ***p < 0.005

Full size image

Discussion

A LC-MS-ROI-MCR-ALS methodology has been proposed to perform the untargeted lipidomic analysis of two cyanobacteria species (Anabaena and P. agardhii) after 7 days of As(III) exposure at 1 and 2 mM. The application of the ROI strategy allowed the compression of the large amount of data from full-scan LC-MS data without any loss of information. Comparison of the elution profiles resolved by MCR-ALS provided information about which lipids more significantly changed their concentration after cyanobacteria exposure to different doses of As(III). The results of these analyses revealed strong affectations on the concentrations of some chemical constituents involved in photosynthesis, which is the essential survival process of cyanobacteria. Although the changes were not identical for both species of cyanobacteria, the families of compounds affected were similar and followed the same concentration reduction pattern. Among the compounds involved in As(II) toxicity, the most significant changes were on the concentration levels of some MGDG galactolipid species. These compounds are the most abundant lipids in all photosynthetic tissues, including those of higher plants, algae, and cyanobacteria. They are directly linked to cyanobacteria membrane, with two main functions: to allow efficient packing of the thylakoid membrane and to separate the components of the photosystem within the thylakoid membrane [10]. In a previous study, it has been shown that a low ratio of MGDG to DGDG is crucial to maintain the stability and proper behavior of protein membranes [11]. In addition to galactolipids, a significant drop in the concentration of photosynthetic pigments was observed. On one hand, in Anabaena, chlorophyll was almost completely suppressed, as well as its sub-product, pheophytin a. The concentration loss of pheophytin a was also observed in the case of P. agardhii at 2 mM of As(III) treatment. On the other hand, the concentrations of some carotene compounds such as 3-cis-hydroxy-carotene and carotene-3,3′-dione were importantly reduced. Whereas chlorophyll a has a key role in absorbing energy from light and acting as a primary electron donor in the electron transport chain of photosynthesis, carotenes behave as allied compounds to chlorophyll in the photosynthetic process as they are able to absorb light energy and transfer it to chlorophylls and also to protect chlorophyll from photodamage. This loss of pigmentation is in agreement with the observed discoloration of cyanobacteria and with the alterations of the UV spectra of the chlorophyll of both species (the results suggest that chlorophyll a is degraded to pheophytin a, which in turn is also reduced). The strong reduction of pigment concentration levels together with the reduction of galactolipid concentration levels indicated that As(III) exposure induced a collapse of the photosynthetic process.

Altogether, a general disruption of the photosynthetic process has been detected after As(III) treatment in both cyanobacteria species, which reveals the strong toxicity induced by this metal, present in a variety of contamination sources.

Conclusions

The combination of a non-target methodology based on the use of the regions of interest (ROI) and multivariate curve resolution analysis on LC-MS full-scan data has been shown to be a powerful approach for the investigation of major lipidomic changes in cyanobacteria cultures when exposed to the highly toxic metal As(III). Lipid composition changes on Anabaena and P. agardhii cyanobacteria cultures when exposed for 7 days to As(III) showed that these changes were mainly associated with changes in the thylakoid cyanobacteria membrane composition (MGDG) and with photosynthetic compounds such as chlorophyll a and some carotenoids. Overall, the results of this study suggested that the presence of increasing concentrations of As(III) induced the collapse of the whole cyanobacteria photosynthetic process.