Investigating mixotrophic metabolism in the model diatom Phaeodactylum tricornutum

Abstract

Diatoms are prominent marine microalgae, interesting not only from an ecological point of view, but also for their possible use in biotechnology applications. They can be cultivated in phototrophic conditions, using sunlight as the sole energy source. Some diatoms, however, can also grow in a mixotrophic mode, wherein both light and external reduced carbon contribute to biomass accumulation. In this study, we investigated the consequences of mixotrophy on the growth and metabolism of the pennate diatom Phaeodactylum tricornutum, using glycerol as the source of reduced carbon. Transcriptomics, metabolomics, metabolic modelling and physiological data combine to indicate that glycerol affects the central-carbon, carbon-storage and lipid metabolism of the diatom. In particular, provision of glycerol mimics typical responses of nitrogen limitation on lipid metabolism at the level of triacylglycerol accumulation and fatty acid composition. The presence of glycerol, despite provoking features reminiscent of nutrient limitation, neither diminishes photosynthetic activity nor cell growth, revealing essential aspects of the metabolic flexibility of these microalgae and suggesting possible biotechnological applications of mixotrophy.

This article is part of the themed issue ‘The peculiar carbon metabolism in diatoms'.

1. Introduction

Diatoms are unicellular eukaryotes responsible for about 20–25% of the global carbon dioxide fixation via photosynthesis (photoautotrophy). Resulting from a secondary endosymbiotic event in which a red alga was engulfed by a heterotrophic eukaryotic host, diatoms display a complex combination of genes and metabolic pathways acquired from endosymbiotic events and horizontal transfer with bacteria and viruses [1].

Ultimately, this chimeric metabolism is believed to be an essential component of the great evolutionary success [2,3] and the high biotechnological potential of these algae. Their potential industrial uses include the utilization of their silica shell for nanostructures [4], food applications [5] and production of triacylglycerols (TAGs) for biofuel under low-input conditions [6]. Given their metabolic flexibility, different growth modes can be induced in microalgae and diatoms in particular. The first is photoautotrophy, in which light energy directly fuels CO₂ conversion into reduced carbon via photosynthesis. This requires a photochemical conversion by the two photosystems (PSs), PSI and PSII, electron flow to generate reducing power (NADPH) and ATP, which are consumed for CO₂ uptake by Rubisco and the Calvin–Benson–Bassham (CBB) cycle.

The second mode is heterotrophy, in which algae grow in the absence of light by fermentation or respiration of exogenous sugars [7]. Among the diatom heterotrophs, two different categories can be recognized: (i) obligate heterotrophs (e.g. Nitzschia alba) that lack photosynthetic pigments and are thus not able to perform photosynthesis and (ii) facultative heterotrophs (e.g. Cyclotella cryptica) that can separately perform photosynthesis and respiration. C. cryptica is able to grow in the presence of glucose in the dark, but displays lower productivity than when operating in the photoautotrophic mode [8]. In fact, some microalgae are obligate photoautotrophs because of an inefficient uptake of carbon (reviewed in [9]). In keeping with this observation, it was shown that introduction of the gene encoding for the human glucose transporter (GLUT1) in P. tricornutum allowed the uptake of glucose in the dark, thereby improving biomass production [10]. A third mode of cultivation is mixotrophy, i.e. the growth in the presence of both light and organic carbon. This mode, which involves the utilization of respiration and photosynthesis simultaneously, is of particular interest to understand how the two major systems of energy metabolism harboured by plants and algae interact with one another. Various diatoms including P. tricornutum [11] and Navicula saprophila, and some species of Nitzschia [12] have been reported to grow mixotrophically, although with different efficiencies and substrate specificities. N. saprophila is able to grow in the phototrophic mode as well as in the presence of acetic acid, both in the presence or absence of light. However, the highest growth rate is observed when reduced carbon is added in the light (mixotrophy), roughly corresponding to the sum of the growth rates obtained in heterotrophy and phototrophy [12].

The model diatom P. tricornutum can grow on glycerol, acetate, glucose and fructose [11,13–15]. Recent results, showing that diatoms optimize their photosynthetic efficiency via constitutive energetic interactions between mitochondria and plastids [16], have provided a molecular interpretation for mixotrophy. Indeed, coupling of respiratory and photosynthetic activities via exchanges of NADPH and ATP provides a tight coordination of these two processes, which should optimize utilization of light and reduced carbon by mixotrophy. As the metabolic consequences of mixotrophy are still poorly studied, we combined metabolic modelling, transcriptomics, metabolomics and lipidomics approaches with physiological measurements to provide a detailed picture of the metabolic changes induced by this growth mode in an attempt to propose possible future applications of this trophic mode in biotechnology.

2. Methods

(a). Algal culture

(i). Strains and growth media

Axenic cultures of P. tricornutum (Pt1, CCAP 1055/3 [17]) were grown in a 250 ml flask in artificial seawater ESAW [18] supplemented with additional NaNO₃ and NaH₂PO₄ to reach a final concentration of 0.47 g l⁻¹ N and 0.03 g l⁻¹ P (this medium will be referred to as PHOT from now on). These elevated concentrations ensure that neither N nor P will be depleted during growth [19]. For N-depletion experiments, cells were shifted to an N-free medium (called here PHOT-N). Cells were grown in a chamber at 20°C, 40 µE m⁻² s⁻¹ irradiance with a 12 h light/12 h dark photoperiod and shaking at 100 rpm. For mixotrophic growth experiments, filter sterilized glycerol was added at a final concentration of 50 mM to both N-replete and N-deficient cultures, leading to the MIX and the MIX-N growth media, respectively.

To monitor algal growth, samples were taken daily (at the end of the light period) and growth was estimated by cell counting using a LUNA instrument (Logos Biosystems, USA). The initial inoculum was 0.5× 10⁶ and 2×10⁶ cells ml⁻¹ for N-replete and N-starved conditions, respectively. Cells were collected after 5 days of growth for the various -omics analyses.

(ii). Assessment of growth and respiration with Biolog plates

The effect of 190 different carbon sources on algal growth was screened to pinpoint possible candidates for mixotrophic cultivation of P. tricornutum using Biolog phenotype microarrays [20]. This microplate assay is based on the use of a 96-well plate containing pre-arrayed substrates such as carbohydrates, amino acids and carboxylic compounds. In this study, 2 × 10⁶ cells ml⁻¹ were resuspended in PHOT medium [19] and 160 µl of it was deposited into each well. Growth was followed daily and substrates that improved growth were selected from a triplicate experiment. A few select metabolites were scaled up for use in 100 ml flasks. A phototrophic control was grown in parallel.

(iii). Nitrogen and phosphate concentration

Nitrogen and phosphate concentration were determined in the supernatant using test strips (Reflectoquant nitrate and phosphate), via a RQflex reflectometer (Merck, Domsstadt, Germany). In the case of nitrogen, this approach was calibrated using a colorimetric assay kit to measure nitrite and nitrate (Sigma, USA), following the manufacturer's instructions. In the case of phosphate, the method was calibrated according to [21].

(b). Spectroscopy

(i). Chlorophyll fluorescence measurements

All the photosynthetic parameters were determined using a Speedzen MX fluorescence imaging set-up (JBeamBio, France) as described in [22]. For each sample, 3 × 200 µl of algal culture, at a cell concentration of 1–5 × 10⁶ cells ml⁻¹, were transferred into a 24-well plate. The maximum quantum yield of PSII (F_v/F_m = (F_m − F₀)/F_m) was determined after 15 min of dark incubation, where F_m and F₀ are the maximum and minimum fluorescence of dark-adapted cells, respectively. Non-photochemical quenching (NPQ) was calculated as (F_m − F_m′)/F_m′, where F_m′ and F_m are the maximum fluorescence of light-adapted and dark-adapted cells, respectively. Photosynthetic electron transfer rate (ETR) was calculated as 0.5 × I × Y(II), where 0.5 represents the fraction of light absorbed by PSII (half of the total incident light), I is the incident light intensity and Y(II) is the quantum yield of PSII in the light. The latter is defined as (F_m′ − F_ss)/F_m’, where F_ss is the fluorescence emission measured in the presence of the light [23].

(ii). Nile Red analysis

Accumulation of triacylglycerols was monitored by Nile Red (Sigma Aldrich) fluorescence staining as detailed in [19]. In brief, 40 µl of Nile Red dye (2.5 µg ml⁻¹ stock concentration, in 100% DMSO) was added to 160 µl cell suspension (1–5×10⁶ cells ml⁻¹) in a 96-well white microplate and mixed. After 20 min of incubation at room temperature in the dark, Nile Red fluorescence was measured (530/580 nm: excitation/emission). Data were then normalized per 10⁶ cells.

(c). Metabolite analysis

(i). Glycerol concentration

The glycerol concentration of 2 ml of filtered supernatant was evaluated by high performance liquid chromatography using a Shimadzu instrument equipped with a Hi-plex H+ (7.7 × 300 mm) Agilent column. The analysis was performed using as the mobile phase 5 mM H₂SO₄. The detection wavelength was set at 880 nm using an RI RID-10A detector with a flow rate of 0.6 ml min⁻¹ and a temperature of 60°C. Peak quantification was performed by comparison of a range of six standards.

(ii). Total lipid extraction

Total lipids were extracted according to [24]. About 20 mg of dried cells were homogenized with 1 ml of chloroform/methanol 2 : 1. The cells were lysed using a Tissue Lyser II (Qiagen) for 1 min at a frequency of 30 s⁻¹. The lysate was washed with 200 µl of 0.9% NaCl (w/v) and vortexed for some seconds in order to form the emulsion. The solution was centrifuged for 5 min at 60 000g to separate the two phases and the lower phase was placed in fresh pre-weighed glass tubes. The upper phase was washed with chloroform and subsequently lysis and centrifugation steps were repeated in order to recover more lipids. The wash with chloroform was repeated at least twice. The lower phases (containing lipids) were collected in glass tubes and evaporated under a nitrogen stream at 65°C. The glass tubes were subsequently reweighed to determine the lipids extracted as a percentage of dry cell mass.

(iii). Separation by thin layer chromatography, and analyses by gas chromatography–flame ionization detection and mass spectrometry

Glycerolipids were extracted from the lipid extract of P. tricornutum cells as described in [19]. To quantify the various classes of polar and non-polar glycerolipids, lipids were separated by TLC on glass-backed silica gel plates (Merck) using two distinct resolving systems [25]. To isolate non-polar lipids including TAG and free fatty acids, lipids were resolved by TLC run in one dimension with hexane:diethylether:acetic acid (70 : 30 : 1, v/v/v). Lipids were recovered from the silica powder after the addition of chloroform:methanol (1 : 2, v/v) with thorough mixing and collection of the chloroform phase [26]. Lipids were then dried under argon and either quantified by methanolysis and GC-FID or by MS.

(iv). Metabolite extraction and gas chromatography–mass spectrometry-based metabolite profiling

Metabolites were extracted with some modifications of the protocol described in [27]. Ten million cells were harvested on a Durapore-HV membrane filter disc 2.5 cm in diameter and with a pore size of 0.45 µm (Millipore, USA) by vacuum filtration. The filter with the cells was then transferred into a 1.5 ml tube and frozen in liquid nitrogen. Frozen samples were stored at −80°C until metabolite extraction. Metabolites were extracted by immersing the filter in 1 ml of 90% (v/v) methanol containing 0.1 µg ml⁻¹ U ¹³C sorbitol as an internal standard. The tubes were sonicated in a water bath-type sonicator for 1 min in ice cold water and then incubated at 4°C for 1 h with shaking. The remaining solution was centrifuged at 22 000g for 5 min at 4°C. A 50 µl aliquot of the supernatant was used for chlorophyll a determination, while a 900 µl aliquot was reduced to dryness using a SpeedVac vacuum concentrator (Thermo Fisher Scientific, USA). Dried samples were stored at −80°C after filling the tubes with argon gas. The metabolite profile was determined exactly as described in [27].

(v). Quantification of intracellular pyruvate

The concentration of pyruvate was evaluated by the fluorescence-based method using the pyruvate assay kit (Cayman Chemical, USA).

(d). Microarray analysis and statistics

(i). RNA extraction and gene expression analysis

Total RNA was extracted as described in [28]. Microarrays were performed in biological triplicates and analysis was carried out as described in [29]. Microarray data have been deposited on the Gene Expression Omnibus (https://www.ncbit.nlm.nih.gov/geo/) under accession number GSE91004.

(ii). Statistical analysis

Data from the three independent experiments were tested for statistical significance of the variations in gene expression using the t-test (calculated with the MeV 4.9 statistical software package) [30]. The three independent replicates were used to perform a one-class analysis using a p value of 0.01. A threshold of expression of absolute log₂ (fold change) value > ±0.75 was used to select genes differentially expressed between test and control conditions.

(e). Mathematical modelling

(i). Genome-scale metabolic model

A genome-scale metabolic model (GSM) of P. tricornutum was developed as described in [31], starting from the model of [32]. The current model consists of 449 reactions, 140 transporters and 355 metabolites, and comprises cytosolic, plastidial, mitochondrial and peroxisomal compartments. It can use NO₃, NH₄, SO₄, O₂, Pi and inorganic and/or organic carbon as input material for biomass production. It has been validated with respect to the laws of energy and mass conservation [33], and is able to produce all major biomass components (carbohydrate, lipid, amino acids, nucleotides, etc.) in phototrophic and mixotrophic conditions in experimentally observed proportions.

(ii). Flux balance analysis

The model was analysed using a modifification of flux balance analysis (FBA) [34,35], in which the underlying linear program is repeatedly solved while increasing the constraint representing the photon input flux [36]. For this study, flux in glycerol uptake was increased instead of photon input flux and, in addition, the proportions of individual biomass components were allowed to vary up to fivefold. The FBA formulation is as follows:

2.1

where the objective is to minimize the total flux in the system. Here, N is the stoichiometry matrix; ν is the rate vector; Nν = 0 defines the steady-state constraint; ν_μ defines the flux of photons (light intensity used in the experiment); ν_glycerol defines the flux of glycerol into the system, which was increased gradually; and defines the range of allowable fluxes in the biomass transporters. The lower bound is the experimentally observed proportion and the upper bound is arbitrarily five times higher. constrains the rate of cyclic photophosphorylation to be no higher than that of non-cyclic photophosphorylation; and sets the upper limit on the sum of the Rubisco carboxylase and oxygenase reactions. This can be regarded as proxy for overall limitation in the Calvin cycle. For this study, C = 0.8 mmol (g dry mass)⁻¹ h⁻¹ was based on the flux in Rubisco reactions obtained in phototrophic condition with the photon flux constrained to the experimental value.

Any solution to this equation thus generates a flux vector, ν, describing the individual fluxes of reactions in the system at steady-state growth with rates of production of biomass precursor defifined by the flux values in , the biomass transporters. To explore potential metabolic responses to increasing glycerol availability, equation (2.1) was solved repeatedly while increasing ν_glycerol to represent increasing glycerol uptake. For this aspect of the study, the allowed photon flux into the system, ν_μ, was constrained to be less than or equal to the light intensity used in the experiment, and the flux in the glycerol transporter was varied from 0.01 to 1.0 mmol (g dry mass)⁻¹ h⁻¹ (experimental glycerol consumption rate was estimated to be ≈0.1 mmol (g dry mass)⁻¹ h⁻¹ ).

3. Results

(a). Consequences of mixotrophic growth on biomass production in P. tricornutum cells

As a first step to investigate the effect of external reduced carbon sources, we measured growth and physiological properties of cells in Erlenmeyer flasks in cells of the pennate diatom P. tricornutum. Experiments were performed in PHOT medium to avoid nitrogen and phosphorous starvation during growth [19]. Glycerol was chosen as the respiratory substrate, as its consequences on metabolism have been already studied to some extent in this alga, and its low cost makes it a suitable substrate for possible future exploitation for biotechnology [11,37]. When tested in 50 ml Erlenmeyer flasks, glycerol enhanced biomass production by a factor of two as compared to growth on PHOT medium (figure 1a). Its effect was gradual, given the progressive consumption of this compound by the algae (figure 1b), and became clearly visible after 5 days of growth.

Figure 1.

Growth curves and nutrient consumption of Phaeodactylum tricornutum. (a) Growth curves of P. tricornutum cells in N-replete and N-starved conditions in the presence/absence of glycerol. Two different starting cell densities were used in the N-replete and N-starved conditions: 0.5 × 10⁶ cells ml⁻¹ and 2 × 10⁶ cells ml⁻¹, respectively. (b) Glycerol consumption of P. tricornutum cells in N-replete conditions. (c) Nitrate and (d) phosphate consumption kinetics in P. tricornutum cultures in the N-replete condition in the presence/absence of glycerol. Each result is the average of two biological replicates ± s.e.m. PHOT, light in the N-replete condition; PHOTO-N, light in the N-starved condition; MIX, light + glycerol in the N-replete condition; MIX-N, light + glycerol in the N-starved condition.

The enhanced growth capacity observed in glycerol-supplemented cells resulted in a much faster consumption of nitrogen (figure 1c) and phosphate (figure 1d), leading to a complete depletion by the end of the growth phase. Given that nutrient starvation (nitrogen starvation in particular) affects photosynthesis [38,39], we looked for possible effects of glycerol on photosynthetic parameters, measuring the quantum yield of PSII (F_v/F_m), the electron flow between PSII and PSI (ETR), and the photoprotective responses of the cells, as indicated by the NPQ parameter. We found (figure 2) that glycerol addition had only minor effects on photosynthesis not only at the beginning of the experiment (day 1), but also at day 5, i.e. after the consequences of this compound on growth become evident, as well as at the end of the growth experiment (day 10). As a corollary of this experiment, we confirmed the large decrease in photosynthesis in P. tricornutum cells upon transfer to N-starved conditions that we had previously observed in [19].

Figure 2.

Photosynthetic activity in P. tricornutum. (a) Photosynthetic efficiency represented as the F_v/F_m ratio; (b) non-photochemical quenching (NPQ) and (c) electron transport rate (ETR) of cells cultivated for 5 and 10 days in N-replete or N-starved conditions in both the phototrophic and mixotrophic mode. Each result is the average of two biological replicates ± s.e.m. PHOT, light in the N-replete condition; PHOT-N, light in the N-starved condition; MIX, light + glycerol in the N-replete condition; MIX-N, light + glycerol in the N-starved condition.

(b). Flux balance analysis of the P. tricornutum metabolic network

As the glycerol-mediated improvement of biomass productivity in P. tricornutum is not the result of a direct effect on photosynthesis, we explored its effect on other areas of cellular metabolism. For this purpose, FBA was performed on the GSM of P. tricornutum as described in §2(e). The number of reactions in the solution to equation (2.1) varied between 339 and 353 over the range of imposed glycerol uptake rates. The greatest variations were in the CBB cycle, tricarboxylic acid (TCA) cycle, oxidative pentose phosphate pathway (OPPP), mitochondrial electron transport chain (mETC), photorespiration, glycolysis, lipid synthesis, carbohydrate synthesis and lactate excretion. Flux in reactions associated with the CBB cycle, including that of Rubisco and CO₂ uptake, decreased with glycerol uptake. By contrast, flux in reactions associated with photorespiration, the TCA cycle, glycolysis and OPPP increased with glycerol uptake. Fluxes in lipid and carbohydrate synthesis reactions increased, leading to increased production of these classes of storage compounds. An increased excretion of lactate was also observed with glycerol uptake. The reactions that showed more than 1% variation in response to glycerol uptake form a connected subnetwork depicted in figure 3.

Figure 3.

The network composed of reactions exhibiting change in flux in response to increase in glycerol uptake. To understand the potential metabolic response in the presence of glycerol, flux balance analysis, with minimization of the sum of absolute fluxes as the objective function together with biomass production constraint, was performed repeatedly with gradual increase in glycerol uptake. This analysis was used to identify reactions with co-related response to change in glycerol uptake in the presence of light. Reactions r1–r20, Calvin cycle/glycolysis/gluconeogenesis; r21, Pyrdh, r22–r32; TCA cycle; r33–r35, C4 metabolism; r36–r40, Thr metabolism; r41–r45, glycerate metabolism; r46–r47, oxidative pentose phosphate pathway; r48–49, glycerol degradation; r50, RuBP oxygenase; r51, glycolate oxidase; r52, LHD I–V, etc.; AOX, alternative oxidase; L1–L2, light reactions. Reactions with increased flux are denoted by green, while those with reduced flux are denoted by red. Note: Flux in RuBP carboxylase decreases; however, flux in most of the reactions involved with the Calvin cycle increases, as they are also associated with other metabolic routes. External metabolites are denoted in the boxes; CO₂ and O₂ are considered to be exchanged between the organism and the medium. 2KG, 2-ketoglutarate; 2PG, 2-phospho-d-glycerate; 6PG, 6-phospho-gluconate; AcAld, acetaldehyde; AcCoA, acetyl-CoA; Acet, acetate; ADP, adenosine diphosphate; Ala, alanine; Asp, aspartate; ATP, adenosine triphosphate; BPGA, 1,3-diphosphate glycerate; Cit, citrate; Cyt_ox, cytochrome c oxidized; Cyt_red, cytochrome c reduced; dagat2d, diacyl glycerol acyltransferase 3 phosphate; DHA, docosahexaenoic acid; DHAP, dihydroxyacetone phosphate; E4P, erythrose-4-phosphate; EPA, eicosapentaenoic acid; F6P, fructose-6-phosphate; FA, fatty acid; fbac5, fructose bisphosphate aldolase; FBP, fructose bisphosphate; FBPaldol, FBP aldolase; FBPase, fructose 1,6-bisphosphatase; Fum, fumarate; G1P, glucose-1-phosphate; G6P, glucose-6-phosphate; G6Piso, G6P isomerase; GAP, glyceraldehyde-3-phosphate; Glt, glutamate; Gly, glycine; Glyc3P, glycerol-3-phosphate; Glyox, glyoxalate; gmd, GDP-mannose 4,6-dehydratase; IsoCit, isocitrate; Lacdh, lactate dehydrogenase; Leu, leucine; M1PDH, mannose-1-phosphate dehydrogenase; Mal, malate; NAD, nicotinamide adenine dinucleotide; NADP, nicotinamide adenine dinucleotide phosphate; OAA, oxaloacetic acid; OH-Pyr, 3-hydroxypyruvate; PEP, phosphoenol pyruvate; PGA, 3-phospho-d-glycerate; Pi, phosphate; ptd9, fatty acid desaturases 9; PUFA, polyunsaturated fatty acid; Pyr, pyruvate; Pyrdh, pyruvate dehydrogenase; Q, ubiquinone; QH2, ubiquinol; R5P, ribose-5-phosphate; Ru5P, ribulose-5-phosphate; Rubisco, RuBP carboxylase/oxygenase; RuBP, ribulose-1,5-bisphosphate; S7P, sedoheptulose-7-phosphate; SBP, sedoheptulose-1,7-bisphosphate; Ser, serine; Suc, succinate; SucAld, succinic semialdehyde; SucCoA, succinyl-S-CoA; TAG, triacylglycerol; Thr, threonine; TPI, triosephosphate isomerase; VAL, valine; X5P, xylulose-5-phosphate.

Organic and inorganic carbon were used when glycerol was assumed to be available. As shown in figure 3, CO₂ is fixed by the RuBP carboxylase reaction (r1 in figure 3) which drives the CCB cycle and thereby contributes towards biomass production. Glycerol is converted to glycerol-3-phosphate by glycerol kinase (r48). It can be used as the glycerol backbone for TAG synthesis and/or is further degraded to DHAP by glycerol-3-phosphate dehydrogenase (r49). DHAP can be used for xylulose-5-phosphate and ribose-5-phosphate production, which can be converted to ribulose-1,5-bisphosphate and enter the CCB cycle or photorespiration. It can also be used for pyruvate production via glycolysis or can be converted to fructose-6-phosphate and glucose-6-phosphate via reactions FBPaldol (r5), FBPase (r6) and G6Piso (r8), which may via subsequent reactions contribute towards an increase in carbohydrate production. Pyruvate can be fermented to lactate via Lacdh (r52) and/or it can be used for acetyl-CoA production by Pyrdh (r21). It can also be converted by pyruvate carboxylase (r34) to oxaloacetic acid, which can then enter the TCA cycle or be used via threonine metabolism (r36-r39).

The RuBP oxygenase reaction (r50) is active and glycolate produced as a consequence is subsequently metabolized to glyoxylate and thereafter used by the glyoxylate shunt of the TCA cycle. Glycine and serine produced during the process are metabolized to 3-phospho-d-glycerate via reactions of glycerate metabolism (r40–r45). Reactions involved with (internal) HCO₃⁻ fifixation, namely PEP carboxylase (r35) and pyruvate carboxylase (r34), as well as the TCA cycle and mETC, including the alternative oxidase, are also active.

(c). Metabolic and transcriptomic assessment of glycerol-mediated changes in P. tricornutum cells

To verify our model predictions, we experimentally assessed changes in metabolic pathways. Consistent with our prediction, previous results indicate that glycerol affects the cellular lipid content [37]. We confirmed this result by measuring TAG accumulation using Nile Red fluorescence (figure 4a,b). We found that, in nitrogen-rich medium, the presence of glycerol enhanced Nile Red fluorescence at day 5. The choice of day 5 to perform metabolic analyses is justified by the need to measure changes at the earliest stage of the glycerol response, to avoid artefacts related to cell ageing and/or consumption of other nutrients. Indeed, nutrient starvation, that of nitrogen in particular, largely increases the accumulation of TAGs in microalgae [40,41]. Thus, day 5, which represents the first data point where significant glycerol effects on growth are detected (figure 1), was the ideal time point to perform a detailed analysis.

Figure 4.

Glycerolipid production in P. tricornutum. (a) Epifluorescence images of cells, at their 5th day of growth, stained with Nile Red dye. (b) Neutral lipid content normalized per millions of cells determined by Nile Red staining at days 5 and 10. (c) Total lipid accumulation at day 5 in P. tricornutum expressed as a percentage of dry mass. (d) DAG and (e) TAG accumulation at day 5 in P. tricornutum normalized per mg dry cell mass. Each result is the average of two biological replicates ± s.e.m. PHOT: light in the N-replete condition; PHOT-N: light in the N-starved condition; MIX: light + glycerol in the N-replete condition; MIX-N: light + glycerol in the N-starved condition.

The finding that glycerol increases Nile Red fluorescence even in N-supplemented cells suggests that TAG accumulation is also increased by this compound in N-replete conditions. This conclusion was further substantiated by the quantification of the total lipid content (figure 4c), and of the DAG and TAG fractions (figure 4d–e) by mass spectrometry. We found that the TAG content was specifically increased by glycerol (figure 4e), in agreement with the Nile Red observations, in both nitrogen-replete and -starved cells. As expected, we also observed that nitrogen starvation not only induced a substantial TAG accumulation, but also enhanced the total lipid content. Conversely, glycerol addition did not modify this parameter significantly (figure 4c). In parallel to the changes in the TAG content, we found that the TAG fatty acid composition was modified by glycerol. In particular, C16 : 0 and C16 : 1 fatty acids were higher in glycerol-grown cells (electronic supplementary material, figure S1), suggesting that this compound induces de novo biosynthesis of fatty acids [19,25]. Glycerol also promoted accumulation of long chain fatty acids (i.e. 20 : 5) that usually correspond to lipids obtained from recycling of membrane lipids, as suggested by the decrease in phosphatidylcholine, the most abundant phospholipid species found in extraplastidic membranes of eukaryotic cells [42] (electronic supplementary material, figures S1 and S2, respectively). The increase in C20 : 5 has already been reported following N limitation in P. tricornutum. Thus, we conclude that glycerol addition affects lipid metabolism by mimicking most of the effects of nitrogen limitation [19], although clearly to a lesser extent. However, at variance with N limitation, the addition of glycerol had a positive effect on growth (figure 1a).

To better understand the effect of glycerol on cellular metabolism, we next compared the metabolite profiles of cells grown in the absence (PHOT) and presence (MIX) of this compound. While the levels of most of the metabolites analysed by GC-MS were not affected by glycerol addition, a few of them displayed significant changes. In particular, six metabolites (lactate, xylose, trehalose, docosahexaenoic acid (DHA) and mannitol) were increased by glycerol, while four (valine, alanine, guanidine and leucine) were decreased by this compound (figure 5). The changes of lactate, valine, alanine and leucine suggest that glycerol could affect the pyruvate hub (figure 6). Given that this metabolite could not reliably be measured via GC-MS, to test this possibility, we directly quantify the pyruvate content in the same cell extracts using a commercial kit (electronic supplementary material, figure S3). We found that glycerol addition increased the pyruvate content, suggesting a higher flux from glycerol to the pyruvate. The finding that pyruvate was increased probably provides a rationale for the increased respiratory activity in cells gown on glycerol, measured by a polarographic approach (electronic supplementary material, figure S4). However, the decrease in alanine, leucine and valine, all of which derive from pyruvate, was somewhat unexpected.

Figure 5.

Metabolomic analysis of P. tricornutum grown in the N-replete condition. The ratio log₂ MIX/PHOT > 0 (green spots) represents all the metabolites that were overexpressed in mixotrophy in the replete condition, while the ratio log₂ MIX/PHOT < 0 (red spots). Green points represent metabolites with increased concentrations, and red points those with decreased concentrations, under mixotrophic conditions. Each result is the average of six biological replicates.

Figure 6.

Hypothetical mode of action of glycerol supply on metabolism and mixotrophic growth of Phaeodactylum: the supply of glycerol can have an impact on (A) central carbon, (B) storage carbon and (C) lipid metabolism. Glycerol probably fuels the lower part of glycolysis, followed by the acetyl-CoA and TAG production and the upper part of gluconeogenesis, followed by carbohydrate production. Italic green type represents gene transcripts found to be upregulated in the presence of glycerol in microarray analysis. Gene ID numbers are also indicated. Bold green or red type represents metabolites that were detected in the current study by GC-MS analysis in cells grown in mixotrophy or phototrophy, respectively. Pyruvate was detected by a pyruvate assay kit (§2c(v)) and TAG in lipidomics analysis (§2c(iii)). Abbreviations: AcCoA, acetyl CoA; Glyc3P, glycerol-3-phosphate; DHAP, dihydroxyacetone phosphate; GAP, glyceraldehyde-3-phosphate; FA, fatty acid; Ala, alanine; Leu, leucine; Val, valine; F6P, fructose-6-phosphate; G6P, glucose-6-phosphate; EPA, eicosapentaenoic acid; DHA, docosahexaenoic acid; tpi, triose phosphate isomerase; fbac5, fructose bisphosphate aldolase; dagat2d, diacyl glycerol acyltransferase-3-phosphate; gmd, GDP-mannose 4,6-dehydratase; ptd9, fatty acid desaturase 9.

The complementary changes in trehalose and mannitol also point to a change in carbon storage metabolism (figure 6). The change in DHA is compatible with the modifications observed by lipidomic analysis. To corroborate the hypothesis that glycerol affects primary, storage carbon and lipid metabolism, we additionally performed a comparative transcriptomic analysis on the microarray using RNA from cultures grown for 5 days in the presence or absence of glycerol. Statistical analysis on the microarray data revealed 35 genes to be differentially expressed in the presence of glycerol (table 1). Notably, most of these genes encode proteins involved in lipid, amino acid and glycolytic metabolism (table 1), consistent with our conclusions concerning central carbon metabolism, lipid biosynthesis and storage derived from our metabolite analyses (figure 6).

Table 1.

Relevant genes regulated in mixotrophic growth. The table shows the genes selected by t-test analysis in microarray analysis that significantly changed (log₂ fold change > ±0.75, p value < 0.01) in transcript abundance between mixotrophic and phototrophic conditions. In the case of genes related to carbon metabolism (highlighted in green), subcellular localization prediction using ChloroP, TargetP [43] and ASAFind [44] is reported. Expression values are the average of three biological replicates ± s.d. Chloro, chloroplast localization; mito, mitochondrial localization; np, negative prediction.

gene id (version 2)	gene id (version 3)	gene expression (log2)	s.d.	annotation	pathway	chloro P/target P prediction	ASAfind prediction
Phatr2_49283	Phatr3_J49283	3.08	0.44	predicted protein
Phatr2_42882	Phatr3_J42882	2.34	0.52	predicted protein
Phatr2_28797	Phatr3_J28797	2.27	0.20	stearoyldesaturase (delta9desaturase)	EPA biosynthesis	others	np
Phatr2_33087	Phatr3_J33087	1.61	0.26	predicted protein
Phatr2_46822	Phatr3_J46822	1.58	0.14	P-loop containing nucleoside triposphate hydrolase
Phatr2_51289	Phatr3_J51289	1.58	0.30	FbaC5 (fructose-bisphosphate aldolase)	glycolysis	chloro	plastid, low confidence
Phatr2_30967	Phatr3_J30967	1.50	0.24	ketol acid reducto isomerase	super pathway leucine, valine	chloro	plastid, low confidence
Phatr2_50738	Phatr3_J50738	1.46	0.22	triosephosphate isomerase	glycolysis	chloro	plastid, low confidence
Phatr2_48920	Phatr3_J48920	1.67	0.01	fibrillarin
Phatr2_43961	Phatr3_J43961	1.44	0.22	alpha/beta hydrolase fold
Phatr2_31718	Phatr3_J31718	1.26	0.16	inosine-5′-monophosphate dehydrogenase	urate biosynthesis	others	np
Phatr2_44522	Phatr3_J44522	1.10	0.13	predicted protein
Phatr2_43703	Phatr3_J43703	1.04	0.14	oxoglutarate/iron-dependent dioxygenase
Phatr2_40998	Phatr3_J40998	1.03	0.06	zinc finger C2H2
Phatr2_31878	Phatr3_J31878	1.02	0.16	CAM2 (calmodulin)
Phatr2_45855	Phatr3_J45855	1.00	0.10	NTF2-like domain (PF07080)
Phatr2_25417	Phatr3_J25417	0.92	0.08	GDP-mannose 4,6-dehydratase 1	GDP-l-fucose biosynthesis I	others	np
Phatr2_9983	Phatr3_J9983	0.87	0.13	50S ribosomal protein L30e-like (PF01248)
Phatr2_45509	Phatr3_J45509	0.84	0.05	carbon-nitrogen hydrolase (PF00759)
Phatr2_43469	Phatr3_J43469	0.83	0.14	DGAT2D - diacylglycerol acyltransferase	Kennedy pathway	secretory	np
Phatr2_43024	Phatr3_J43024	0.77	0.04	tetratricopeptide repeat (PF13424)
Phatr2_32629	Phatr3_J32629	−0.75	0.08	predicted protein
Phatr2_46117	Phatr3_J46117	−0.77	0.10	predicted protein
Phatr2_15806	Phatr3_J15806	−0.78	0.09	PDS-like3, phytoene desaturase-like, phytoene dehydrogenase-like
Phatr2_47655	Phatr3_J47655	−0.79	0.06	endoribonuclease L-PSP/chorismate mutase-like (PF14588)
Phatr2_31876	Phatr3_J31876	−0.98	0.12	reverse transcriptase RNA-dependent DNA polymerase (PF07727)
Phatr2_48291	Phatr3_J48291	−1.02	0.10	EF-hand domain (PF13499)
Phatr2_1199	Phatr3_J1199	−1.04	0.15	leucine-rich repeat (PF13516)
Phatr2_48021	Phatr3_J48021	−1.04	0.15	predicted protein
Phatr2_2164	Phatr3_J2164	−1.12	0.27	d-xylose:proton symporter	xylose degradation	secretory	np
Phatr2_42568	Phatr3_EG02435	−1.15	0.18	helicase-associated (PF03457)
Phatr2_44192	Phatr3_EG02162	−1.23	0.19	predicted protein
Phatr2_40368	Phatr3_J40368	−1.30	0.12	predicted protein
Phatr2_38713	Phatr3_J38713	−1.70	0.26	ribonuclease H-like domain
Phatr2_46275	Phatr3_J46275	−2.29	0.25	HYP (FA desaturase type 1 domain)	unknown	others	np

Finally, the results obtained with glycerol lead us to explore whether other compounds could be used as substrates to boost biomass production in microalgae. To identify new substrates, we grew P. tricornutum cells in Biolog plates PM1 and PM2A, supplemented with 190 different carbon sources, for 6 days. This allowed monitoring of biomass productivity (via cell counting). We found that several compounds, not previously identified as mixotrophic substrates for P. tricornutum, enhanced growth of P. tricornutum cells (electronic supplementary material, figure S5). They include fumarate, aspartate, asparagine and serine. Despite the clear benefit of glycerol on the growth of P. tricornutum in flasks (figure 1a; electronic supplementary material, figure S5b,c), this compound did not improve growth at the microplate scale (electronic supplementary material, figure S5a). This may result from lower oxygen availability in the microwells circumventing the efficient respiration observed upon glycerol addition to liquid cultures (electronic supplementary material, figure S3). However, for the compounds which enhanced growth at the microplate scale, enhanced growth could be confirmed also in liquid cultures, confirming the utility of this approach in pinpointing biotechnologically relevant mixotrophic substrates.

4. Discussion

(a). Metabolic consequences of glycerol-mediated mixotrophic growth

The metabolic flexibility of diatoms has often been invoked to explain their evolutionary success. Indeed, these algae can grow in different modes: photosynthetically, simply converting the sunlight energy into reduced carbon via photosynthesis, or heterotrophically via sugar fermentation in the dark. The latter mode of growth is, however, not possible in P. tricornutum, which can use sugars in the dark only upon metabolic engineering [10]. However, P. tricornutum can grow mixotrophically, i.e. simultaneously using light and reduced carbon [11,37]. Here, we have investigated the metabolic consequences of glycerol addition on P. tricornutum cells, finding that this compound promotes higher growth and modifies the metabolism in terms of accumulation of carbon resources, including lipids and other carbon storage compounds. The results obtained from the modelling investigation allow identification of the pathways most likely to be involved in this process.

(b). Photorespiration

The glycerol metabolism of P. tricornutum was previously investigated using isotope labelling experiments with ¹³C-glycerol as the carbon source [45,46]. These studies highlighted that, under mixotrophic conditions, P. tricornutum cells mostly convert this compound into glycine and serine. A similar observation was made in our modelling analysis. As shown in figure 3, photorespiration is active and the glycolate produced by the RuBP oxygenase reaction can be metabolized to glycine and serine. The P. tricornutum metabolic network has confirmed the capability to produce glycine and serine via both photorespiratory and non-photorespiratory routes [47]. However, an interesting observation is that the model suggests the latter route is used under conditions of high glycerol availability. Previous analysis of this model [31] suggests that photorespiration becomes active as a result of supraoptimal light intensities. The increased photorespiratory flux is associated with increased fluxes towards glycine and serine.

In contrast, no significant changes were observed in the cellular levels of serine and glycine in our experimental analysis, despite the fact that the levels of several other amino acids displayed significant changes following the addition of glycerol. The relationship between the concentration of a metabolite and the fluxes of the reactions in which it is involved is complex and potentially counterintuitive. However, it is possible to say that if the activities of all reactions in a given pathway increase proportionately, then the fluxes of those reactions will increase by the same proportions while the concentrations of the intermediate metabolites will remain unaffected [48]. For instance, it is possible that the enhanced flux towards glycine and serine evaluated by others [45,49] and predicted by our model cannot lead to any measurable changes in the concentration of these amino acids, probably because of a tight coupling between production and consumption of these compounds. Consequently, there is no inherent inconsistency between the experimental and the model results reported here.

(c). Effect of glycerol on carbon metabolism

In contrast, other metabolic changes were pinpointed by our comparison of light- and light + carbon-driven growth, using complementary approaches (lipidomics, transcriptomics and metabolomics). After 5 days of growth, the majority of the differentially regulated genes encode proteins associated with carbon and lipid metabolism. This result suggests that glycerol induces a specific modification of cellular metabolism in order to cope with the increased availability of carbon. The extra energy flux provided by the glycerol to cells grown in non-limiting light conditions (figure 1a, 0–5 days) is possibly diverted to storage, as confirmed by Nile Red staining (figure 3a,b). Gene-encoding factors involved in cell regulatory processes and cell cycle progression were unaffected in cells grown under mixotrophic as opposed to phototrophic growth conditions. Given that glycerol has a clear effect on growth, we can speculate that these genes were differentially regulated in the first days following glycerol addition, but that their expression level had returned to normal by treatment day 5, which we investigated here. Most importantly, we found that the observed changes in central-carbon, carbon-storage and lipid metabolism in mixotrophically versus phototropically grown P. tricornutum cells are consistent with the modelling results.

In the modelling results, glycerol enters central carbon metabolism by its conversion into glycerol phosphate by glycerol kinase and thereafter into DHAP by the action of glycerol phosphate dehydrogenase. This is consistent with the observed effects of overexpression of two enzymes involved in glycolysis and gluconeogenesis, namely TPI (PHATR_50738) and FbaC5 (PHATR_51289) in the microarray analysis (table 1). The higher level of pyruvate (electronic supplementary material, figure S3) can be explained by the higher glycolytic flux. Modelling suggests that pyruvate can be fermented to lactate via LDH, act as a substrate for lipid synthesis or be converted via pyruvate carboxylase to OAA, which enters the TCA cycle. The latter route is supported by the experimentally observed increase in lactate concentration data and the increased respiratory rate.

The modelling results also suggest increased fluxes to storage carbohydrates in the presence of glycerol. This also is consistent with our experimental data, although we additionally observed an increase in trehalose and mannitol, and not just chrysolaminarin as predicted by the model analysis. This finding suggests that besides chrysolaminarin, which is the most abundant carbohydrate storage form in diatoms [50], P. tricornutum can accumulate, at least to some extent, other storage sugars when provided with external organic carbon. Consistent with this, P. tricornutum possesses a complete set of enzymes for the biosynthesis of trehalose and mannose and we observed that one of them was upregulated by glycerol (i.e. GDP-mannose 4,6-dehydratase, PHATR_25417). Moreover, we identified a putative mannitol dehydrogenase in the genome of Phaeodactylum (i.e. PHATH_30246) that is involved in the conversion of the mannose into mannitol. This gene, the expression of which is not modulated by glycerol, shows some homology (approx. 37.2%) with the mannose dehydrogenase enzyme M1PDH from the brown algae Ectocarpus siliculosus. Both the trehalose and mannitol pathways have already been identified in brown algae, where they appear to have been inherited from the red algal progenitor and via lateral gene transfer from Actinobacteria, respectively [51]. Thus, it seems reasonable to assume that a similar situation may be present in P. tricornutum. Alternatively, it is conceivable that the observed changes in mannitol and trehalose upon addition of glycerol represent a response to an increased osmotic pressure induced by this compound. Consistent with this idea, previous data from both prokaryotes and eukaryotes [52–54] have shown that both trehalose and mannitol are induced by osmotic stress.

Our experimental analysis also points to glycerol-mediated effects on lipid metabolism, consistent with previous reports [11]. Our modelling results suggests that glycerol can be metabolized to contribute to TAG biosynthesis by providing the glycerol backbone as well as providing substrates for fatty acid biosynthesis, producing the acyl groups required for TAG assembly (figure 6). These results are also corroborated by our microarray analysis, which reveals an upregulation of the fatty acid desaturase ptd9 (PHATR_ 28797) as well as the acetyl-transferase dagat2d (PHATR_ 43469) involved in the fatty acid and TAG biosynthesis, respectively. Moreover, our analysis of the fatty acid composition of TAGs in glycerol-treated cells suggests that their accumulation is due to de novo synthesis as well as to the degradation of some pre-existing membrane lipids. In general, TAGs can be produced by two main routes: (i) de novo synthesis of fatty acids directly incorporated into TAGs via the Kennedy pathway involving a diacylglycerol acyltransferase and de novo DAG and acetyl-CoA synthesis (e.g. [19,25]), or (ii) conversion of pre-existing polar glycerolipids [55]. TAGs generated by the first route contain newly generated fatty acids, i.e. high levels of C16 : 0 and C16 : 1. Conversely, TAGs obtained from recycling of membrane lipids contain fatty acids with a substantial proportion of elongated and polyunsaturated molecular species such as C20 : 5 fatty acids. Our data (electronic supplementary material, figure S1) indicate that glycerol-grown cells contain a higher amount of 16 : 0 and 16 : 1 fatty acids, consistent with the occurrence of de novo synthesis. However, glycerol also leads to an increased accumulation of C20 : 5 fatty acids, consistent with the occurrence of some membrane lipid turnover. This behaviour is very similar to that observed in P. tricornutum cells upon exposure to nitrogen starvation [19]. However, in contrast with N-limited cells, glycerol-supplemented cells do not display (i) any significant degradation of the most abundant thylakoid lipids, i.e. monogalactosyldiacylglycerol and digalactosyldiacylglycerol (electronic supplementary material, figure S2), or (ii) any loss of photosynthetic activity (figure 2). Thus, although the consequences of glycerol addition of lipids remind one of those of nitrogen starvation [19] in terms of TAG accumulation and changes in the fatty acid profile, this compound seems to act mainly at the level of lipid biosynthesis rather than degradation.

The model analysis highlights possible changes in flux distribution in the metabolic network and changes in biomass composition, again consistent with our experimental observations. However, FBA and related approaches do not fully explain changes in the steady-state concentration of internal metabolites. One of the routes for alanine synthesis is the reversible transamination of pyruvate and, following simple logic, one would expect to see an increase in alanine coincident with an increase in pyruvate. Similarly, pyruvate is also the carbon precursor of valine and leucine in the chloroplast, and an increase in these amino acids was also expected. A possible explanation for the observed decrease in valine and leucine concentration would be to assume a heterogeneous distribution of pyruvate in the cells. As fatty acid synthesis is boosted in the presence of glycerol and de novo fatty acid biosynthesis occurs in the chloroplast, it might be that while pyruvate increases globally at the cellular level (explaining the increased levels in electronic supplementary material, figure S3), its concentration could be lower in the chloroplast because of efficient consumption by the plastidial pyruvate dehydrogenase during fatty acid synthesis. An increased availability of glycerol-3-phosphate for lipid synthesis might also effectively ‘drain’ the fatty acid biosynthetic pathway. This could relieve product inhibition of pyruvate dehydrogenase, ultimately reinforcing pyruvate consumption in the chloroplast.

(d). Effect of glycerol on biomass productivity

Glycerol addition leads to a substantial increase in growth in Phaeodactylum (figure 1). However, a careful analysis of its effects on growth reveals that not all the glycerol provided to the cells is converted into biomass. While almost 1 gl⁻¹ of glycerol was consumed during algal cultivation (figure 1b), the cell number was increased from 5 × 10⁵ cells l⁻¹ to 2 × 10⁷ cells l⁻¹. Based on the relationship between cell number and dry biomass in our samples, this corresponds to around 0.6 g l⁻¹ of biomass increase. Thus, the increase in biomass cannot be entirely accounted for by the consumption of glycerol. The model results described above provide a potential resolution to this apparent contradiction: as the rate of glycerol uptake was increased, there was a decrease in the CBB cycle. Furthermore, the calculated fluxes in the mitochondria responded to increased glycerol uptake in a broadly opposite sense to those of the CBB cycle, leading to an increased production of CO₂. Both these responses resulted in a reduction in net CO₂ fixation in the model, and this provides a working hypothesis to explain the difference between the experimental consumption of glycerol and increase in biomass.

5. Conclusion

This work highlights the potential of mixotrophic growth for biotechnology applications. Indeed, mixotrophy can enhance biomass productivity while providing advantages in terms of lipid accumulation, which are normally seen under nutrient starvation. While glycerol induces some responses typical of nitrogen-limited cells with regard to TAG accumulation (figure 4) and fatty acid composition (electronic supplementary material, figure S1), no inhibition of the photosynthetic capacity of the cell is seen in the presence of this compound (figure 2).

By revealing the main metabolic pathways targeted by glycerol, our work also suggests possible targets for metabolic engineering. For instance, inhibition of the biosynthesis of storage carbohydrates could potentially divert carbon (derived from glycerol) towards TAG production. This was observed in our model analysis and has also already been reported in the case of the main sugar storage polymer, chrysolaminarin [56]. It is hoped that ongoing integration of our data into a mathematical model could reveal other possible targets to further increase algal production capabilities.

The observed capacity to use glycerol supplied in the medium suggests the existence of a glycerol transporter (most probably an aquaporin-type transporter). The identification and characterization of such a system would have relevant consequences for understanding its role in natural conditions (the ocean waters are often enriched in organic carbon due to the relatively high turnover of plankton components) and inspire biotechnological developments. The parallel use of engineered strains and new substrates for growth (electronic supplementary material, figure S5) could largely improve biomass productivity in mixotrophically grown microalgae.

Authors' contributions

V.V., A.E.F., D.S., J.J., A.R.F., A.F., M.P., E.M., J.P., A.L.M., M.P., G.C., D.P. and G.F. conceived and designed the experiments. V.V., A.E.F., D.D.B., M.C. and T.O. performed experiments. V.V., A.E.F., J.J., A.R.F., A.F., M.P., E.M., J.P., A.L.M., A.R.F., G.C., D.P. and G.F. analysed data. D.S. and M.P. performed the model simulation and wrote the modelling aspect of the manuscript. V.V., A.R.F., M.P. and G.F. wrote the manuscript and all the authors approved the final version of the manuscript prior to submission.

Competing interests

The authors have no competing interests.

Funding

V.V., D.S., M.P., D.P., A.F. and G.F. acknowledge the Marie Curie Initial Training Network Accliphot (FP7-PEPOPLE-2012-ITN; 316427). V.V., A.E.F., M.C., J.J., E.M., A.F., A.L.M., J.P., D.P. and G.F. acknowledge the Agence Nationale de la Recherche (ANR-12-BIME-0005, DiaDomOil). D.P. and G.F. acknowledge the Région Rhone-Alpes (Cible project). M.C., J.J., E.M., D.P. and G.F. acknowledge the CEA Bioénergies programme. E.M. was financed by Programme Investissement d'Avenir Oceanmics. G.F. acknowledges the CNRS Défi (CNRS 2013) and HFSP (HFSP0052). D.P. acknowledges GRAL Labex, ANR-10-LABX-49-01. A.F. acknowledges contribution by the Marie Curie Initial Training Network CALIPSO (ITN 2013 GA 607607).