Abstract Aquatic environments are heavily impacted by human activities including climate warming and the introduction of xenobiotics. Due to the application of silver nanoparticles as bactericidal agent the introduction of silver into the environment strongly has increased during the past years. Silver ions affect the primary metabolism of algae, in particular photosynthesis. Mixotrophic algae are an interesting test case as they do not exclusively rely on photosynthesis which may attenuate the harmful effect of silver. In order to study the effect of silver ions on mixotrophs, cultures of the chrysophyte Poterioochromonas malhamensis were treated in a replicate design in light and darkness with silver nitrate at a sub-lethal concentration. At five time points samples were taken for the identification and quantitation of proteins by mass spectrometry. In our analysis, relative quantitative protein mass spectrometry has shown to be a useful tool for functional analyses in conjunction with transcriptome reference sequences. A total of 3,952 proteins in 63 samples were identified and quantified, mapping to 4,829 transcripts of the sequenced and assembled transcriptome. Among them, 720 and 104 proteins performing various cellular functions were differentially expressed after eight days in light versus darkness and after three days of silver treatment, respectively. Specifically pathways of the energy and primary carbon metabolism were differentially affected by light and the utilization of expensive reactions hints to an energy surplus of P. malhamensis under light conditions. The excess energy is not invested in growth, but in the synthesis of storage metabolites. The effects of silver were less explicit, observable especially in the dark treatments where the light effect could not mask coinciding but weaker effects of silver. Photosynthesis, particularly the light harvesting complexes, and several sulphur containing enzymes were affected presumably due to a direct interference with the silver ions, mainly affecting energy supply. Introduction Protist predators are the primary agents of top-down control of bacteria and primary producers, and thus key organisms determining the fate and transport of organic matter in nature [[42]1]. Shifts in species composition of protist communities as well as changes in their functional ecology will therefore affect the overall architecture of aquatic food webs. Due to the immense diversity of protists and their small size the ecology of these organisms has, up to recently, largely been addressed on the level of broad functional groups. Protists, as e.g. small flagellates, comprise a large number of taxonomically quite different organisms with different growth and survival strategies ranging from mainly photosynthetic flagellates (e.g. Uroglena), that only take up very small numbers of bacteria, probably as vitamin source, to entirely heterotrophic flagellates (e.g. Spumella spp.). Between these extremes, there are organisms that utilize both photosynthesis and phagotrophy to different extents depending on the environmental conditions experienced (e.g. [[43]2, [44]3]). Mixotrophy is particularly widespread in chrysophytes (throughout the manuscript, this term refers to the chrysophytes sensu lato, comprising the Synurales). Further, chrysophytes are of special interest because their abundance may increase during phases of lake re-oligotrophication [[45]4, [46]5] and because they often dominate mixotrophic phytoplankton populations [[47]6]. Newly developed molecular methods now allow for deeper insights into the ecology of these organisms and with that into the function and the regulation of processes within the microbial food web. Light and food availability as well as stressors such as heavy metals can have a crucial impact on community structure and functioning [[48]7–[49]9]. Environmental stressors will become more and more important due to ongoing environmental changes. Pollution is among the most serious stressors affecting the composition of aquatic communities and the well-being of organisms [[50]10]. Among metal pollutants silver ions are one of the most toxic forms, and have thus been assigned to the highest toxicity class, together with cadmium, chromium (VI), copper and mercury [[51]11]. Its toxicity to a wide range of microorganisms combined with its low toxicity to humans led to the development of a wealth of silver-based products in many bactericidal applications [[52]12, [53]13] accounting, for example, for more than 1000 nanotechnology-based consumer products [[54]14]. Accordingly, silver is a widely distributed metal in the environment originating from its different forms of application as metal, salt and nanoparticle. The recent upward trend in production (estimated 500 t/a worldwide) [[55]15] and application resulted in an increasing release of silver nanoparticles (AgNP) as well as of ionic silver into the environment as can be seen from elevated levels of silver in the aquatic environment [[56]16–[57]19]. Continuous input of silver can cause severe changes in species composition and species succession in phytoplankton communities [[58]9, [59]20]. Ionic silver thus has the potential to disrupt their internal dynamics and feedbacks, to upset the microbial loop and to undermine the entire plankton community on which aquatic food webs are based. Silver has been shown to affect particularly metabolic pathways associated with photosynthesis and with that specifically phytoplankton [[60]9, [61]21]. A binding of silver ions to chlorophyll and a decreasing efficiency of photosynthesis with increasing silver exposure might be causative [[62]9]. The chrysophyte Poterioochromonas malhamensis is a widely used model organism in microbial ecology. P. malhamensis can utilise a wide selection of different organic food sources and can be considered a food source generalist [[63]8]. Different P. malhamensis isolates have been shown to assimilate the main part of their required carbon via phagotrophy or osmotrophy and only to a minor extent by photo—autotrophy [[64]8, [65]22–[66]27]. P. malhamensis does not grow but sustains cell metabolism with the help of light-derived energy when organic carbon sources are limited [[67]28]. When grown osmotrophically, i.e. in the absence of particulate food, the growth of the mixotrophic Poterioochromonas malhamensis is limited by the organic matter [[68]28]. In general, data so far indicate that with increasing importance of phototrophic nutrition (Spumella / Paraphysomonas → Poterioochromonas → Ochromonas → Dinobryon) there is a parallel decrease in maximum growth rate [[69]8]. This is expected, since the development of a photosynthetic apparatus (chloroplasts, enzymes) is considered to be much more energetically expensive than heterotrophic degradation of food (e.g. [[70]29]). The effect of pollutants, in particular of silver, may therefore depend on the nutritional mode, i.e. differ between dark and light treatments. In this study, we extend existing transcriptome and metatranscriptome studies on the physiological effect of silver by a controlled proteomic time course experiment. We address the effects of light and silver stress on the mixotrophic protist Poterioochromonas malhamensis using label-free quantitative mass spectrometry. The impact of light, near the light compensation point of P. malhamensis, and silver stress are assessed and discussed with regard to their metabolic implications. Since light is not essential for growth in P. malhamensis with enough organic carbon sources available, we hypothesize a possible damage to the photosystem due to silver should have a low impact on the primary metabolism. We firstly focus on the effect of light in general and secondly analyse the effect of silver with respect to these findings. Additionally, we discuss the potential of a proteomics approach for such an analysis. Materials and Methods Cultivation and sample preparation The flagellate Poterioochromonas malhamensis strain DS originates from the mesotrophic Lake Constance, Germany. In Essen, P. malhamensis is permanently cultivated as an axenic culture in organic NSY medium [[71]30]. This strain has been used in a variety of laboratory studies under the name Ochromonas sp. (e.g. [[72]26, [73]31]) and under the name Ochromonas sp. strain DS [[74]30, [75]32, [76]33]. All samples were taken under sterile conditions. Depending on cell density (300,000 to 1,000,000 cells/ml), we took 50 to 150 ml of culture to receive an equally large cell pellet for proteome analysis (centrifuged at 4,400 rcf, Rotina 35 R, Hettich Lab Technology). Pellets were washed three times in 75% ethanol ahead of further analyses. Experimental setup All P. malhamensis treatments were kept at 18°C and 35 µE m^-2 s^-1 light intensity in a 16/8h light-dark-cycle. For the experiments we chose a light intensity approximately corresponding to light compensation point, i.e. 35 µE m^-2 s^-1. The stock culture was incubated until reaching a cell density of 300,000 cells per ml (start value, SV), which corresponds to the early stationary phase. Thereupon, the stock culture was split up in 16 subcultures in 750 ml culture bottles. All cultures were kept at the same temperature for five days, but 8 cultures were kept in darkness and the other 8 in light. After five days samples were taken from the cultures for mass spectrometry (MS) measurements. This marked day 0, after which silver was added as a second stressor. To half of the cultures in dark and half of the cultures in light silver nitrate was given at an end concentration of 3 µg/l. Silver concentrations of 5 µg/l and below have been shown to affect both phytoplankton community composition and community gene expression profiles [[77]9]. We thus decided for 3 µg/l Ag^+ in order to induce a sub-lethal response to silver stress. Samples for MS measurement were taken 1, 3 and 5 days after silver application. This resulted in one start value measurement, in 8 replicates for light and dark at day 0 and for the time points 1, 3 and 5 in four replicates for the conditions light and silver (LS), light but no silver (LN), darkness with silver (DS) and darkness and no silver (DN) (see [78]Fig 1). Replicate IDs consist of treatment (LS, LN, DS, DN) + replicate (A, B, C, D) + sampling day (0, 1, 3, 5). If, in the following, all replicates of a treatment are considered, the letter for the replicate (A-D) will be skipped. Fig 1. Experimental setup of light and silver nitrate exposure of cultures. [79]Fig 1 [80]Open in a new tab The experiment was carried out in four different treatments, each treatment in four replicates (i.e. each blue box represents four replicates). First, light/no light was applied for 5 days. Then silver nitrate was added to two of the four treatments and samples were taken after 1, 3 and 5 days. Mass spectrometry Cultivation and sample preparation The frozen cell pellets were warmed to room temperature and taken up in 1 ml phosphate buffered saline (PBS). In order to break the cells, release the proteome and to shear the genomic DNA, the samples were lysed by ultrasonification in a Bioruptor® (Diagenode) for 5 min. The turbid solution was cleared by centrifugation (12,000 rcf, 4°C, 5 min). The supernatant was transferred to a fresh 1.5 ml Eppendorf tube. The protein concentration was determined by using a modified Bradford assay (RotiNanoquant; Roth) following the manufacturer’s instructions. Reduction/alkylation and tryptic digestion We first removed an aliquot corresponding to 50 μg total protein content from each sample and performed an acetone precipitation (ratio of sample to acetone = 1:4). The samples were vigorously stirred and stored over night at -20°C to completely precipitate all proteins. After centrifugation (18,000 rcf, 4°C, 5 min) the supernatant was aspirated and the pellet dried at room temperature for 10 min. The pellets were taken up in 25 μL 6 M urea in 100 mM ammonium bicarbonate (ABC) and extensively vortexed. The solutions were cleared by centrifugation and the supernatants were transferred to 96-well-plates (Eppendorf). The proteins were reduced by addition of dithiothreitol (DTT, 8 mM final) and incubation at 37°C for 30 min and secondly alkylated by addition of iodoacetamide (IAM, 16 mM final) and incubation at 30°C for 30 min. In order to quench the excess IAM, we added more DTT to a final concentration of 24 mM. After this treatment 500 ng Lys-C (1:100; Wako Laboratory Chemicals) was added and the samples were incubated for 3 h at 37°C. The samples were diluted to 25 mM ABC and 1.5 M urea. 800 ng sequencing grade Trypsin (1/60; Promega) were added and the samples were incubated over night at 37°C while shaking. On the next morning, the samples were acidified by adding formic acid solution (FA, final 0.5% v/v). Sample clean-up for liquid chromatography and mass spectrometry (LC-MS) Acidified tryptic digests were desalted on home-made C18 StageTips as described [[81]34]. On each 2 disc StageTip we loaded around 15 μg peptides (based on the initial protein concentration). After elution from the StageTips, samples were dried using a vacuum concentrator (Eppendorf) and the peptides were taken up in 10 μL 0.1% formic acid solution. LC-MS/MS experiments were performed on an Orbitrap Elite instrument (Thermo; [[82]35]) that was coupled to an EASY-nLC 1000 liquid chromatography (LC) system (Thermo). The LC was operated in the one-column mode. The analytical column was a fused silica capillary (75 μm × 38 cm) with an integrated PicoFrit emitter (New Objective) packed in-house with Reprosil-Pur 120 C18-AQ 1.9 μm resin (Dr. Maisch). The analytical column was encased by a column oven (Sonation) and attached to a nanospray flex ion source (Thermo). The column oven temperature was adjusted to 45°C during data acquisition and in all other modi at 30°C. The LC was equipped with two mobile phases: solvent A (0.1% formic acid, FA, in water) and solvent B (0.1% FA in acetonitrile, ACN). All solvents were of Ultra Performance Liquid Chromatography (UPLC) grade (Sigma). Peptides were directly loaded onto the analytical column with a maximum flow rate that would not exceed the set pressure limit of 980 bar (usually around 0.3 − 0.6 μL/min). Peptides were subsequently separated on the analytical column by running a 240 min gradient of solvent A and solvent B (start with 7% B; gradient 7% to 35% B for 220 min; gradient 35% to 100% B for 10 min and 100% B for 10 min) at a flow rate of 300 nl/min. The mass spectrometer was operated using Xcalibur software (v2.2 SP1.48). The mass spectrometer was set in the positive ion mode. Precursor ion scanning was performed in the Orbitrap analyzer (FTMS) in the scan range of m/z 300-1,800 and at a resolution of 60,000 with the internal lock mass option turned on (lock mass was 445.120025 m/z, polysiloxane) [[83]36]. Product ion spectra were recorded in a data dependent fashion in the ion trap (ITMS) in a variable scan range and at a rapid scan rate. The ionization potential (spray voltage) was set to 1.8 kV. Peptides were analysed using a repeating cycle consisting of a full precursor ion scan (1.0 × 10^6 ions or 30 ms) followed by 15 product ion scans (1.0 × 10^4 ions or 50 ms) where peptides are isolated based on their intensity in the full survey scan (threshold of 500 counts) for tandem mass spectrum (MS2) generation that permits peptide sequencing and identification. Collision-induced dissociation (CID) energy was set to 35% for the generation of MS2 spectra. During MS2 data acquisition dynamic ion exclusion was set to 60 s with a maximum list of excluded ions consisting of 500 members and a repeat count of one. Ion injection time prediction, preview mode for the FTMS, monoisotopic precursor selection and charge state screening were enabled. Only charge states higher than 1 were considered for fragmentation. Peptide and protein identification using MaxQuant RAW spectra were submitted to an Andromeda [[84]37] search in MaxQuant (v1.5.0.25) using the default settings [[85]38]. Label-free quantification and match-between-runs was activated [[86]39]. MS/MS spectra data were searched against the in-house generated peptide database of predicted candidate open reading frames from the assembled transcriptome of P. malhamensis (39,357 entries; NCBI BioSample ID: SAMEA3936180). All searches included a contaminants database (as implemented in MaxQuant, 267 sequences). The contaminants database contains known MS contaminants and was included to estimate the level of contamination. Andromeda searches allowed oxidation of methionine residues (16 Da) and acetylation of protein N-terminus (42 Da) as dynamic modification and the static modification of cysteine (57 Da, alkylation with iodoacetamide). Enzyme specificity was set to “Trypsin/P”. The instrument type in Andromeda searches was set to Orbitrap and the precursor mass tolerance was set to ±20 ppm (first search) and ±4.5 ppm (main search). The MS/MS match tolerance was set to ±0.5 Da. The peptide spectrum match false discovery rate (FDR) and the protein FDR were set to 0.01 (based on target-decoy approach). Minimum peptide length was 7 amino acids. Peptides were assembled into proteins and proteins sharing the same identified peptides were grouped together into one protein group. For label-free protein quantification (LFQ) unique and razor peptides were allowed. Modified peptides were allowed for quantification with a minimum score for modified peptides of 40. For convenience we will talk of proteins instead of protein groups hereafter, but it has to be noted that these might comprise several proteins with shared peptides. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [[87]78] partner repository ([88]https://www.ebi.ac.uk/pride/archive/) with the dataset identifier PXD005146. Analysis of identified proteins The assembled and annotated transcriptome of P. malhamensis was used to assign gene names, gene symbols, orthology and pathway information from the Kyoto Encyclopedia of Genes and Genomes (KEGG release 2014-06-23, [[89]40]) to the identified proteins. For the KEGG annotation an E-value threshold of 10^−5 was used. The analyses were performed only for proteins with annotated transcript matches in the reference transcriptome. For pathway analyses the KEGG Orthology information was used, which assigns cross-species orthology information to genes by a KEGG Orthology identifier (KO ID). These KO IDs allow the mapping of the proteins via the transcript annotation to KEGG reference pathways. The KEGG pathway analyses were restricted to the categories Metabolism, Genetic Information Processing, Environmental Information Processing, Cellular Processes and Organismal System. All analyses were performed and plots were generated with the R statistical framework (v3.2.4, [[90]41]) and ggplot2 (v2.1.0, [[91]42]). Presence-absence analysis A presence-absence analysis of proteins was performed to describe the proteomes and compare the coverage of KEGG pathways with those of the reference transcriptome. For pathway annotation the KEGG BRITE functional hierarchy system level A and B were used. Furthermore, this data was used to detect whether silver and light treatment switches the protein synthesis of specific proteins on or off. For protein presence-absence analysis proteins which could be identified with a minimum of two unique peptides were used. For each KEGG pathway the proteins identified by MS were counted. Statistical tests of differences between treatments were carried out on total KO ID occurrence counts and at pathway level using the generalized linear model (GLM) test of the R package edgeR (v3.12.0, [[92]43]). Differential expression analysis The LFQ intensities from the quantitative MS were used to perform expression analyses. Replicate 4 of the treatment dark and no silver at time point 3 (sample DND3) was removed due to a high deviation from the intensity values of the other replicates (see sample 48 (bottom) in [93]S1 Fig). For the other samples raw intensities were filtered to exclude hits for transcripts with missing values in more than one replicate and remaining missing values were imputed using k-nearest neighbour imputation from the R package impute (v1.42.0, [[94]44]) after normalization. The raw intensities were normalized by applying a variance stabilization method implemented in the R package limma (v3.26.9, [[95]45, [96]46]). Limma was also used to perform statistical testing of differential protein expression. For principal component analysis (PCA) and redundancy analysis (RDA), the R package vegan (v2.3-2, [[97]47]) was used. By the KO identifiers the significant differentially expressed proteins were assigned to KEGG reference pathways and an in-house R package was used to colour differentially expressed proteins in the KEGG pathway maps and conduct a pathway enrichment analysis using a hypergeometric test. To visualize the pathways for publication, we used Cytoscape (v3.3.0, [[98]48]) and the Cytoscape app KEGGscape (v0.7.0, [[99]49]). Results During the whole experiment, the strain P. malhamensis DS showed a more agitated movement and higher cell numbers in the light treatments. From a starting value of about 300,000 cells per ml, cell numbers in the treatments in light increased and reached a constant number of about 1,000,000 cells per ml (irrespective of silver application). The cell numbers of the dark treatments stayed at a value of about 300,000 cells per ml and even further decreased under silver application to a value of 200,000 cells per ml. Thus, in the dark treatments in contrast to the light treatments a difference in cell numbers with and without silver was visible. To understand these effects on the molecular level proteins were analysed and quantified with mass spectrometry, of which the results are described in the following. Presence-absence analysis Altogether, we could identify 3,952 protein matches by mass spectrometry using the assembled transcriptome of P. malhamensis as reference, which comprises in total 39,537 transcripts. The number of detected proteins is comparable or higher to other proteomic studies that use transcriptomes or genomes for the identification, yielding between 660 to 3,886 proteins [[100]50–[101]54]. Considering the different treatments, the number of identified proteins ranges on average between 3,582 proteins within treatment DN3 to 3,837 proteins within treatment SV. The filtered P. malhamensis transcriptome contains 14,726 transcripts with a KEGG annotation E-value ≤10^−5. We excluded all protein matches with a KEGG annotation E-value >10^−5, yielding from 4,128 KO IDs within treatment LN3 to 4,486 KO IDs within treatment SV. In total 925 KO IDs within treatment LN3 to 962 KO IDs within treatment SV were assigned to one or more pathways, resulting in total pathway matches from on average 1,924 within treatment DS1 to 2,011 within treatment SV ([102]Fig 2). At the level of KEGG BRITE functional hierarchy system A, on average 47% of the KO IDs were found in the category Metabolism followed by the categories Genetic Information Processing with 22%, Organismal System with 15% as well as Cellular Processes and Environmental Information Processing with 8% assignments. In comparison, the reference transcriptome had 1,599 unique KO IDs with 3,297 matches to the five KEGG categories: Metabolism (39%), Genetic Information Processing (23%), Organismal System (17%), Cellular Processes (11%) and Environmental Information Processing (10%), see [103]Fig 2. Thus, the proteome covers a considerable amount of the transcriptome with 59% of all identified KO IDs. Using mass spectrometry, Bai et al. [[104]53] identified about 40% of the genes in Aspergillus flavus, while [[105]52] show for the sperm cells of Caenorhabditis elegans that 1080 of 2511 (approx. 30%) protein annotations overlap with the transcriptome. A review on Apicomplexa proteome and transcriptome studies, estimates the proportion of the covered proteome in 6 studies to lie between 10 to 45% [[106]54]. Fig 2. Assignment of protein matches to KEGG pathways A. [107]Fig 2 [108]Open in a new tab Assignment of uniquely counted genes for each treatment to the KEGG BRITE functional hierarchy level A: Metabolism, Genetic Information Processing, Environmental Information Processing, Cellular Processes and Organismal System. The median over the replicates for each treatment is shown with standard deviation. Treatments are labelled according to the following scheme: light and silver (LS), light but no silver (LN), samples in darkness with silver (DS) and in darkness and no silver (DN). Treatment IDs consist of treatment (LS, LN, DS, DN) + sampling day (0, 1, 3, 5). At day 0 no silver was added yet, therefore LN0 = LS0 and DN0 = DS0, but they were already split up in 4 replicates each. SV represents the start value and Poterioochromonas DS the counts per pathway category for the transcriptome. Furthermore, we assigned uniquely counted KO IDs of the reference transcriptome and all treatments to 298 KEGG pathways (KEGG BRITE functional hierarchy level C) which were summarized to 31 pathway groups (KEGG BRITE functional hierarchy level B) within the five KEGG categories (KEGG BRITE functional hierarchy level A) to calculate the coverage of each pathway. On pathway group level the treatments of different sampling days had very similar coverages. Therefore, we exemplarily illustrated the treatments of day 3 ([109]Fig 3). For the pathway groups of the KEGG database, we found diverging coverages from those of the reference transcriptome. Many pathway groups showed an approximately 5–10% lower coverage than the corresponding ones of the reference transcriptome. Especially pathways included in the KEGG category Genetic Information Processing i.e. transcription and replication and repair as well as pathways of the group “Cell growth and death” of the KEGG category Cellular Processes had an approximately 25% lower coverage than the reference transcriptome. In contrast, the pathway groups “Metabolism of terpenoids and polyketides”, “Biosynthesis of other secondary metabolism” and “Xentobiotics biodegradation and metabolism” of the KEGG category Metabolism as well as the pathway group “Membrane transport” of the KEGG category Environmental Information Processing showed only a 2% decrease in coverage. In addition to the pathways of the previously named pathway groups we identified on pathway level similarly covered pathways like glycolysis, citrate cycle, biosynthesis and degradation of fatty acids, many amino acids and ketone bodies as well as photosynthesis, translation and regulatory systems like the proteasome. Fig 3. Assignment of protein matches to KEGG pathways B. [110]Fig 3 [111]Open in a new tab Assignment of uniquely counted genes for time point 3 to KEGG pathway groups of hierarchy level B. Colours of pathways on x-axis indicate KEGG hierarchy level A from [112]Fig 2. The median over the replicates for each treatment is shown with standard deviation. Treatments are labelled according to the following scheme: light and silver (LS), light but no silver (LN), samples in darkness with silver (DS) and in darkness and no silver (DN). Treatment IDs consist of treatment (LS, LN, DS, DN) + sampling day (3). SV represents the start value and P. malhamensis DS the counts per pathway category for the transcriptome. The statistical tests for each time point with a generalized linear model that tested for differences between all treatment combinations did not yield significant results, neither based on total gene counts nor pathway coverage. Expression analysis Raw LFQ intensities were measured with quantitative mass spectrometry for 63 samples and proteins mapping to 4,829 transcripts. After filtering of missing values, imputation and normalization (see [113]Methods) 2,297 proteins remained, which were subsequently analysed for differential protein expression. A principal component analysis (PCA) was used to visualize the amount of variance in the normalized LFQ intensities and the grouping of the samples due to differing intensities. The strongest effect can be seen on the first two axes of the PCA, which pronouncedly shows the difference between the light and dark treatments ([114]Fig 4A). Furthermore, a redundancy analysis (RDA) including, supplementary to the LFQ intensities, the factors day (1, 3, 5), light conditions (L, D) and silver treatment (S, N) was applied. A stepwise model reduction of the complete initial model resulted in an overall model with day and light conditions as significant factors (p-values ≤ 0.005, see [115]Fig 4B). Fig 4. Ordination plots. [116]Fig 4 [117]Open in a new tab (A) Principal component analysis (PCA) and (B) redundancy analysis (RDA) of normalized intensity matrix. Samples are depicted as numbers, different days are indicated by colours and symbols show samples kept under dark (circle) and light (triangle) conditions. The coloured background indicates the separation between samples in light (light grey) and dark (dark grey) conditions in the PCA. Arrows in (B) show significant factors of the reduced model. Samples were numbered for readability according to [118]S1 Table. The significant factors can be assessed using the RDA, but it is unclear which proteins are responsible for the differences. This can be analysed by testing for significant differences in protein intensities between the treatments. To assess differential expression a linear model was used and the effects listed in [119]Table 1 were tested for each time point. The strongest effect was visible on day 3. All tested contrasts showed the highest number of significant differentially expressed proteins (adj. p-value <0.01) at this time point. Table 1. Number of differentially expressed proteins at all time points. Effect Nr. Effect Explanation Day 1 Day 3 Day 5 I LN vs. DN Light effect, no silver 90 720 0 II LS vs. DS Light effect, silver 61 114 32 III LS vs. LN Silver effect, light 0 0 8 IV DS vs. DN Silver effect, dark 0 104 0 V (LS vs. LN) vs. (DS vs. DN) Difference of silver effect in light vs. dark 0 187 0 [120]Open in a new tab Although light caused the severest changes in expression, at time point 3 the effect of silver for the dark samples and the difference of the silver effect in light and dark was strong as well. A beta-uniform mixture model (R package BioNet v1.29.1, [[121]55]) fitted to the p-value distributions showed a strong deviation from the uniform-distribution (random noise) and a high signal component for the effects I, IV and V (data not shown). The pronounced differences of the silver treatment under light and dark conditions (V) show that the organism reacts in a distinct way, possibly due to an interference of silver with enzymes necessary for the dark or light reactions of photosynthesis. A large proportion of proteins was impaired by several treatments (see [122]Fig 5). E.g. 48 KO IDs annotated to affected proteins overlapped between the light effect without silver and the silver effect in darkness for an adjusted p-value cut-off of 0.01. We also found a significant correlation (p-value <0.001) in log fold-changes (logFC) for the top 100 most differentially expressed proteins for these two treatments. However, the log fold-changes of the light effect without silver were considerably higher. Fig 5. Differentially expressed KOs. [123]Fig 5 [124]Open in a new tab Number of significantly differentially expressed proteins (adjusted p-value <0.01) annotated to (A) assembled transcripts and to (B) KEGG Orthology identifiers per tested effect at time point 3. I-V represent tested effects according to [125]Table 1. All significantly differentially expressed proteins per test at time point 3 are provided in [126]S2 Table. Pathway enrichment analysis To identify significantly altered pathways, e.g. pathways that contain a large number of the significantly differentially expressed proteins for the light effect without silver (effect I), a pathway enrichment analysis was performed and enriched pathways were visually inspected (see [127]Fig 6 for the two most affected pathways). The enrichment analysis on KEGG metabolic pathways yield an increased expression of proteins involved in photosynthesis, carbon fixation in photosynthetic organisms, glycolysis/gluconeogenesis, N-glycan biosynthesis, inositol phosphate metabolism, riboflavin and methane metabolism for the effect I (light). A decreased expression of several proteins was present in parts of the tricarboxylic acid (TCA) cycle and oxidative phosphorylation. Other pathways without a clear directional change in expression of a majority of significant proteins include: ascorbate and aldarate metabolism, purine metabolism, glutathione metabolism, glycerophospolipid metabolism, glyoxylate and dicarboxylate metabolism, lysosome and synaptic vesicle cycle. Fig 6. Enriched pathways in light vs. dark without silver (time point 3). [128]Fig 6 [129]Open in a new tab Depicted are two KEGG metabolic pathways with a high enrichment in differentially expressed proteins. (A) shows the glycolysis/glyconeogenesis and (B) carbon fixation in photosynthetic organisms. Significantly differentially expressed proteins are coloured in red for a positive logFC and green for a negative logFC. Proteins in white were identified with quantitative MS, but did not show differential expression. In comparison, significantly differentially expressed proteins for the silver effect in darkness (IV) and also the interaction effect of silver and light (V) led to an increase in glycolysis/gluconeogenesis, parts of the TCA cycle, pyruvate metabolism, carbon fixation in photosynthetic organism and a decrease in oxidative phosporylation. Apparently, the silver treatment partially affects similar processes as the light vs. dark treatment, we will review these findings in more detail in the discussion. For a complete list of all significantly enriched pathways per test at time point 3 see [130]S3 Table. Discussion Potential of proteomics in microalgal ecophysiology Quantitative protein mass spectrometry has evolved into an indispensable and versatile examination method in molecular biology research. In environmental biology including freshwater ecosystem research, its regular application is, however, very limited, mainly due to a lack of established quantitative proteomics methodologies and reference databases. In the last years, mainly (meta)genomic and/or (meta)transcriptomic approaches were used to determine or monitor overall functional capacities or nutrient/multiple stressor-induced changes in freshwater microorganisms [[131]9, [132]56–[133]59]. Transcription levels reflect the regulation of genes but the effect on the proteome and thus on the functional level remains uncertain. In fact in most instances, changes in the protein repertoire of microorganisms in a community are indirectly inferred from transcriptomics experiments thus ignoring regulation on a post-transcriptional level. In a recent study, Xie et al. [[134]60] used in a similar approach the isobaric tags for relative and absolute quantitation (iTRAQ) technique to investigate the cellular response of the algae Phaeodactylum tricornutum to sub-lethal concentrations of aluminium. They found aluminium-induced toxic effects on photosynthesis leading to an increase in reactive oxygen species (ROS) and energy production. In contrast to our study, the sequenced and annotated genome of P. tricornutum was readily available to analyse the proteome data. Here we demonstrate that proteomics is a promising tool for ecophysiological research on microalgae if a transcriptome or genome is sequenced which can be used to construct a reference database. In our study, only 23 proteins (e.g. ATP synthase, ATPase, HSP70, actin) could be assigned to NCBI sequences in comparison to 3,952 hits on the sequenced transcriptome of P. malhamensis. This shows the up to now insufficiency of available database references when working with lesser