The response to sulfate deficiency of plants and freshwater green algae has been extensively analysed by system biology approaches. By contrast, seawater sulfate concentration is high and very little is known about the sulfur metabolism of marine organisms.
Here, we used a combination of metabolite analysis and transcriptomics to analyse the response of the marine microalga Emiliania huxleyi as it acclimated to sulfate limitation.
Lowering sulfate availability in artificial seawater from 25 to 5 mM resulted in significant reduction in growth and intracellular concentrations of dimethylsulfoniopropionate and glutathione. Sulfate-limited E. huxleyi cells showed increased sulfate uptake but sulfate reduction to sulfite did not seem to be regulated. Sulfate limitation in E. huxleyi affected expression of 1718 genes. The vast majority of these genes were upregulated, including genes involved in carbohydrate and lipid metabolism, and genes involved in the general stress response.
The acclimation response of E. huxleyi to sulfate deficiency shows several similarities to the well-described responses of Arabidopsis and Chlamydomonas, but also has many unique features. This dataset shows that even though E. huxleyi is adapted to constitutively high sulfate concentration, it retains the ability to re-program its gene expression in response to reduced sulfate availability.
Sulfur is essential for the growth of all living organisms. Photosynthetic organisms cover their sulfur requirements by taking up and assimilating inorganic sulfate (reviewed in Takahashi et al., 2011a). In terrestrial and freshwater ecosystems, sulfate is often in low concentration and limits growth; therefore, specific responses to sulfate deficiency have evolved in various taxa. Plants and freshwater algae respond to sulfate deficiency primarily by increasing sulfate uptake and assimilation capacity (Yildiz et al., 1994; Clarkson et al., 1999; Hoefgen & Nikiforova, 2008), while, for example, the green alga Chlamydomonas reinhardtii is also capable of inducing arylsulfatases needed for utilization of alternative sulfur sources (de Hostos et al., 1988). Sustained sulfate deficiency leads to decrease in photosynthetic activity and to reprogramming the plant metabolism to ensure sufficient resources are available for seed production (Hoefgen & Nikiforova, 2008).
By contrast, sulfate is plentiful for the diverse sulfate-reducing organisms in the sea, because its concentrations in seawater reach 25–28 mM. The high sulfate concentration in the marine environment seems to have been an evolutionary driver in the expansion of modern phytoplankton groups (Ratti et al., 2011). In accordance with the high availability of sulfate in seawater, many phytoplankton species synthesize and accumulate large amounts of the sulfur-containing metabolite, dimethylsulfoniopropionate (DMSP). Many roles have been proposed for this compound in phytoplankton, from a simple osmolyte or a sink for excess electrons to a signal molecule for biotic interactions in marine environment (Stefels, 2000; Steinke et al., 2006). It is also a precursor of dimethylsulfide (DMS), an atmospheric gas with a great impact on the geochemical sulfur cycle and, possibly, global climates (reviewed in Giordano et al., 2005; Quinn & Bates, 2011).
Because of the contrasting sulfate concentrations in the marine (25–28 mM) and soil/freshwater (10–50 μM) environments, sulfate has never been considered as a factor limiting productivity in the oceans. However, we hypothesized that phytoplankton is well adapted to the high sulfate availability such that a reduction in its availability would affect growth. The intriguing question arising from this hypothesis is: Do marine phytoplankton retain the ability to respond to sulfate limitation found in plants and freshwater algae? Recent progress in the genomics of marine phytoplankton (Tirichine & Bowler, 2011) paved a way for addressing this question at an overall transcriptome level. Among the phytoplankton species available for such analysis, the coccolithophore Emiliania huxleyi seems to be the most suitable model because of the large intracellular concentrations of DMSP, which accumulates to 50–242 mM in different strains of E. huxleyi (Steinke et al., 1998).
Here we show that lowering sulfate availability to concentrations below those that E. huxleyi encounters in the natural environment indeed reduces growth and DMSP concentration. Transcriptomics analysis on E. huxleyi as it adjusted to sulfate deficiency revealed that some acclimation responses to this stress condition are conserved among E. huxleyi, Chlamydomona, and Arabidopsis, but many are unique to this marine microalga.
Materials and Methods
Algal material and growth conditions
Emiliania huxleyi CCMP1516 was obtained from the Provasoli-Guillard National Center for Marine Algae and Microbiota (NCMA, Bigelow, ME, USA). The alga was grown in axenic batch culture in 250-ml conical flasks with 150 ml enriched artificial seawater medium (ESAW) (Berges et al., 2001) in a growth chamber (MLR 351; Sanyo; Loughborough, UK) at 15°C under a light : dark cycle of 14 h : 10 h and an irradiance of 180 μmol m−2 s−1. Culture flasks were gently stirred by hand on a daily basis. Based on frequent microscopy observation, this strain does not appear to produce coccoliths under these culture conditions. ESAW medium contains 25 mM Na2SO4 and 363 mM NaCl; for the sulfate deficiency studies, the sulfate was replaced by NaCl to keep the ionic strength constant (i.e. each mmol Na2SO4 was substituted by 3 mmol NaCl). Three independent cultures per treatment were inoculated with the same volume of a control culture (50 μl, c. 50 000 cells), which was grown in ESAW medium to the mid-logarithmic phase of growth. Samples for analysis were always taken c. 2 h into the light phase. For the transcriptome analysis the samples were taken 6 d after inoculation, in the middle of the exponential phase, to ensure the cells were still in the process of acclimation.
Determination of E. huxleyi growth
Cell density (cells ml−1) and cell volume (μm3 ml−1) were determined using a Coulter multisizer (Beckman Multisizer 3, High Wycombe, UK) with a 100 μm aperture tube. Measurements were carried out with 100-μl culture samples diluted to 10 ml with 0.2 μm filtered seawater. The growth rates were calculated as [loge(N2) – loge(N1)]/(d2−d1), where Ni is the number of cells at day di. The ratio of variable to maximum chlorophyll fluorescence (Fv/Fm), which gives an estimate of PS II efficiency, was measured with a Walz Phyto-Pam phytoplankton analyser (Heinz Walz GmbH, Effeltrich, Germany) after 30 min dark adaptation.
Determination of DMSP
Emiliania huxleyi particulate DMSP (DMSPp) concentration was determined using headspace gas chromatography (Steinke et al., 2000). Two to three ml of culture was gently filtered through 25 mm Whatman GF/F filters (nominal pore size 0.7 μm) using a hand-operated vacuum pump. The filters were placed in 4-ml vials containing 3 ml of 0.5 M NaOH and the vials were immediately sealed with a screw thread cap and Teflon coated septum to render than gas-tight. To ensure complete cold alkali hydrolysis of DMSP to DMS, vials were incubated for at least 24 h in darkness at room temperature. They were equilibrated for 1 h at 30°C before analysis using a gas chromatograph with a flame photometric detector (Shimadzu 2010, Milton Keynes, UK) and a 30 m × 0.53 mm CP SIL 5CB column (Varian, Wokingham, UK). For analysis, 50 μl of headspace gas was withdrawn using an autosampler (MPS 2, Gerstel, Mülheim, Germany) using a 100-μl gas-tight syringe and injected into the GC (Steinke et al., 2000). DMSP concentration was determined using a calibration curve for known quantities of DMSP (linear between 0.1 and 25 μM) treated by the same alkali hydrolysis procedure.
Adenosine 5'-phosphosulfate reductase activity
Adenosine 5'-phosphosulfate reductase (APR) activity was determined as the production of [35S]sulfite, assayed as acid volatile radioactivity formed in the presence of [35S]APS and dithioerythritol (Koprivova et al., 2008). Ten mL culture aliquots were centrifuged (10 min, 10 000 g), supernatants removed and pellets re-suspended in 1.5 ml of culture medium and re-centrifuged (5 min, 10 000 g). The cells were disrupted by sonication on ice in 500 μl extraction buffer (50 mM Na/KPO4 pH 8; 30 mM Na2SO3; 0.5 mM AMP, 10 mM dithioerythritol). The extracts were centrifuged (30 s, 1000 g) to remove cell debris and 20 μl was used for APR measurement. Protein concentrations were determined with a protein assay kit (Bio-Rad), using bovine serum albumin as the standard.
HPLC analysis of low molecular weight thiols
Thiols were extracted from cells filtered from 15 to 25-ml culture aliquots using hot methanesulphonic acid (Dupont et al., 2004). Total cysteine and glutathione were analysed following the method of Koprivova et al. (2008).
In order to measure the sulfate uptake, E. huxleyi cultures were grown in 500-ml conical flasks with 250 ml ESAW medium containing 25 mM (control) or 5 mM sulfate. Fifty millilitres of control or 100 ml of sulfate-deficient cultures were filtered onto 47-mm diameter 1.2 μm filters (Millipore), and washed with 200 ml sulfate-free medium to remove sulfate. The cells were re-suspended in 50-ml tubes with 10 ml ESAW medium containing 25 or 5 mM sulfate. The cell density and volume was determined for each tube. [35S]sulfate was added to a specific activity of 192 kBq ml−1 and the cells were incubated for 60 min in the light. The cells were collected by filtration, washed twice with 100 ml S-free medium and placed into 20-ml scintillation vials. To dissolve the filters and disrupt the cells, 5 ml of tissue solubiliser (Solene®-350, PerkinElmer, Cambridge, UK) was added and the vials were kept overnight at room temperature. The next day, 10 ml of scintillation cocktail Optisafe 3 (Perkin Elmer) was added and [35S] radioactivity was determined by scintillation counting (Wallac 1409, Perkin Elmer).
RNA isolation and expression analysis
Total RNA was isolated by standard phenol/chlorophorm extraction and LiCl precipitation. For quantitative RT-PCR (qPCR) first-strand cDNA was synthesized from 1 μg of total RNA using a QuantiTect Reverse Transcription Kit (Qiagen), which includes a DNAse step. The qPCR reactions were performed in duplicate for three independent samples using gene-specific primers (Table S7) as described in Lee et al. (2011). All quantifications were normalized to the ACTIN gene (gene ID 74049).
For sequencing, total RNA from three independent sulfate-limited and control cultures was repurified using an RNeasy Plant isolation kit (Qiagen) with on column DNAse removal of contaminating DNA. To prepare Illumina RNAseq libraries poly-A RNA was isolated from 5 μg of total RNA and the mRNA was fragmented to an average size of 100 bp using the manufacturer's instructions (Illumina mRNAseq Kit, Illumina, Cambridge, UK). First strand cDNA synthesis used Superscript III reverse transcriptase (Invitrogen, Paisley, UK) and 3 μg random hexamer primers (Illumina). Second strand cDNA and RNAseq libraries were prepared according to the manufacturer's instructions (Illumina). Following a PCR amplification and size selection, the mRNA libraries were sequenced using the Illumina HiSeq 2000 platform to generate paired end 50 bp reads.
All sequenced reads were quality controlled (QC) by removing reads containing ‘N's and those with read-lengths shorter or longer than 50 nucleotides. The QC reads for each biological replicate were separately aligned to the E. huxleyi reference transcriptome at JGI (Emihu1_best_transcripts.fasta from http://genome.jgi-psf.org/Emihu1/Emihu1.download.html) using TopHat (Trapnell et al., 2009, 2012). Further analyses were performed with the Cufflinks tools suite (Trapnell et al., 2010). Differential expression between the control and sulfate-deficient cultures was calculated by Cuffdiff using the FPKM (Fragments Per Kilobase of exon per Million fragments mapped) normalization, false discovery rate of 5%, and Cuffdiff-min-alignment-count parameter of 622.
The sequenced reads were then aligned to the E. huxleyi reference genome (Emihu1_scaffolds.fasta) using TopHat to examine the number of reads that mapped to the genome but not the transcriptome. Differential expression was then calculated using the Tophat-Cufflinks-Cuffcompare-Cuffdiff pipeline. To compare our transcripts to the predicted E. huxleyi transcripts, the Cuffdiff transcript expression file (providing the assembled transcripts) was compared to that of the Emihu_1_best_genes.gff annotation file from JGI. Reads that did not map to the E. huxleyi reference genome were de novo assembled using OasesOptimizer. The resulting transcript assembly was used as a reference and the unmapped reads were analysed for differential expression. To obtain insight into the identity of the novel transcripts, they were subjected to BLAST analysis (for details of the procedures see Supporting Information Methods S1).
For comparison with other organisms, transcriptomics data from A. thaliana grown for 8 d on normal sulfate supply and transferred on sulfate-limited conditions for 6 d (Nikiforova et al., 2003) and Chlamydomonas exposed to sulfate starvation for 6 h (González-Ballester et al., 2010) were used. In these studies A. thaliana and especially Chlamydomonas would still have been acclimating to sulfate limitation, thereby representing a good comparison with our data for E. huxleyi.
Results and Discussion
E. huxleyi growth is limited by low sulfate concentration
Growth of plants and green algae can be limited by sulfate, therefore, we tested whether sulfate availability alters E. huxleyi growth. Batch cultures of E. huxleyi were grown at sulfate concentrations of 25 mM (control), 10, 5 and 1 mM, and cell density and volumes were monitored (Fig. 1a,b). Lowering the sulfate concentration to 10 mM had no effect on growth rate. Interestingly, the cell volume reduced by c. 60% over the time course of the experiment with 10 and 25 mM sulfate; similar decreases in cell volume have been shown for other strains of E. huxleyi (Van Rijssel & Gieskes, 2002). Further reduction to 5 mM sulfate reduced the growth rate in the exponential phase by c. 50% from 0.62 in the controls to 0.30 d−1 and the cells were also c. two-fold larger than those grown at higher sulfate concentrations (Fig. 1a,b). This sulfate concentration is three orders of magnitude higher than the concentration that limits the growth of plants; indeed, plants grow normally even at 5 μM sulfate (Hawkesford & De Kok, 2006), while freshwater algae adapt to environments with sulfate concentrations in the 0.01–1 mM range (Giordano et al., 2005). At 1 mM sulfate growth was very slow (0.05 d−1; Fig. 1a) and cell volume was c. six-fold larger volume than control cells by day 13 (Fig. 1b). The efficiency of PS II was not substantially affected by low sulfate conditions apart from an initial drop in Fv/Fm in the 1 mM cultures that within 2 d recovered to control concentrations (Fig. S1).
These results are consistent with a recent report by Ratti et al. (2011) showing significantly slower growth rates for E. huxleyi (strain PML92/11) at 5 and 1 mM sulfate compared to 10 mM and higher sulfate concentrations. The same was true for the dinoflagellate Protoceratium reticulatum (PRA0206), but not for the green alga Tetraselmis suecica (PCC 305) or the marine cyanobacterium Synechococcus sp. (UTEX LB2380) (Ratti et al., 2011). Thus, it seems that increase in sulfate concentration in seawater was one of the major evolutionary drivers for the success of chlorophyll a+c phytoplankton, including the coccolithophores (Ratti et al., 2011). It is interesting to note that the E. huxleyi cultures grew normally at 10 mM sulfate, as this was the concentration in water in which these algae evolved (Ratti et al., 2011). Also the growth of the halophilic green alga Dunaliella salina was limited by sulfate concentrations < 0.1 mM (Giordano et al., 2000). The cessation of cell division and increase in cell size observed in E. huxleyi at 1 mM sulfate was similar to the effects of sulfur deprivation on Chlamydomonas (Melis et al., 2000; Zhang et al., 2002), but may also be a response to uncoupling between growth rate and division rate.
Decreased sulfate availability also affected intracellular DMSP concentration (Fig. 1c). Under control conditions intracellular DMSP remained at a stable and high concentration (c. 260 mM) throughout the experiment, whereas cells growing at lower sulfate concentrations showed a notable reduction in DMSP concentration. DMSP decreased until day 8, and then remained relatively stable at c. 75%, 40% and 20% of control concentrations for cultures containing 10, 5 and 1 mM sulfate, respectively. Interestingly, at 10 mM sulfate the DMSP concentration was altered but growth was unaffected, whereas cultivation at 5 and 1 mM sulfate decreased both growth and DMSP accumulation. Surprisingly, Ratti et al. (2011) did not observe a significant change in DMSP content in E. huxleyi strain PML92/11 between sulfate concentrations of 5 and 20 mM. However, the maximal intracellular DMSP concentration in strain PML92/11 reached only c. 100 mM (Ratti et al., 2011), which is similar to the concentration found in E. huxleyi CCMP 1516 grown at 1 mM sulfate (Fig. 1). Given that a five-fold reduction in sulfate concentration in the medium resulted in reduced growth rate and DMSP accumulation, 5 mM sulfate was chosen as the sulfate deficiency treatment for all subsequent experiments. It has to be noted, however, that our experiments used sulfate deficiency as a tool rather than to mimic actual environmental conditions.
Next we tested whether the observed changes in growth were reversible and thereby connected to sulfate availability. E. huxleyi was grown in 25 and 5 mM sulfate media for 8 d, that is, late exponential phase. Half of the sulfate-deficient cultures were supplemented with sulfate to restore the concentration to the control 25 mM level. The sulfate addition rapidly increased the specific growth rate from 0.15 to 0.32 d−1 and after 3 d the cell numbers in the supplemented cultures were only 10% lower than in the controls and double that of the sulfate-deficient cultures (Fig. 2a). This increase in cell numbers was accompanied by a decrease in cell volume which reached control levels after 48 h (Fig. 2b). Sulfate restoration also enhanced DMSP accumulation and after 48 h DMSP concentrations were indistinguishable from those of the control cells (Fig. 2c). One of the markers for sulfate deficiency in plants is the induction of APR, and therefore we determined the activity of this enzyme in the three treatments. Unlike intracellular DMSP concentration, APR activity was not significantly different in the different cultures throughout the experiment. The activity decreased in all the cultures at 24 h and thereafter, but this was probably connected with the end of the exponential phase of growth (Fig. 2d).
Regulation of E. huxleyi sulfur metabolism by sulfate deficiency
Having determined that a decrease in sulfate concentration from 25 to 5 mM limits growth and DMSP accumulation in E. huxleyi, we examined the acclimation response of the algae to these conditions. Cellular cysteine content decreased by c. 60% during culture growth, but no difference was detected between sulfate-deficient and control cultures (Fig. 3a). However, as with DMSP, glutathione (GSH) concentration was c. two-fold lower in sulfate-deficient cells than in the controls after 6 and 10 d of cultivation (Fig. 3b). A reduction in GSH concentration is a typical consequence of sulfate deficiency in plants (Hirai et al., 2003; Nikiforova et al., 2003). In plants and Chlamydomonas, another typical response to sulfate starvation is an increase in sulfate uptake capacity (Pootakham et al., 2010). We tested E. huxleyi cultures grown at 5 and 25 mM sulfate for [35S]sulfate uptake at both concentrations and found that the 5 mM-grown cells showed an c. three-fold higher uptake of sulfate at both concentrations compared to control cells grown at 25 mM sulfate (Fig. 4). Thus, in common with plants, E. huxleyi reacts to sulfate deficiency by increasing sulfate uptake.
General transcriptome analysis of E. huxleyi response to sulfate deficiency
In order to enable comparison of the acclimation responses to sulfate deficiency of E. huxleyi with other organisms, we analysed the transcriptomes of sulfate-deficient and control E. huxleyi cells. Total RNA was isolated from three independent sulfate-deficient and control cultures. At the time of sampling the average growth rates were 0.67 and 0.46 d−1 in the control and sulfate-deficient cultures, respectively. The control cultures achieved five generations from the beginning of the treatment, whereas the sulfate-deficient cultures achieved four, reaching average intracellular DMSP concentrations of 226 and 103 mM, respectively. The RNA was subjected to RNA sequencing using the Illumina platform at The GenePool Edinburgh. The resulting 50 bp reads for each biological replicate were aligned separately to the E. huxleyi CCMP1516 reference transcriptome (Emihu1_best_transcripts.fasta) based on the Joint Genome Institute (JGI) (http://genome.jgi-psf.org/Emihu1/Emihu1.home.html). The E. huxleyi genome assembly predicted 39 125 genes, compared with predicted 10 402 gene models in Phaeodactylum tricornutum (Bowler et al., 2008) and 11 390 models in Thalassiosira pseudonana (Armbrust et al., 2004). The much larger number of gene models is due to the presence of diploid alleles for many genes.
From a total of 58 871 530 reads (after quality control), 37 008 141 mapped to the reference transcriptome, leaving 21 863 389 (37.1%) reads unmapped (Fig. 5, Table 1). From the 39 125 predicted E. huxleyi transcripts in the JGI reference transcriptome, 16 729 transcripts were identified using the Cufflinks tool suite. The expression levels were normalized using the FPKM as a measure of expression strength. Differential expression was then calculated for each locus shared between the control (ehux25) and sulfate-deficient (ehux5) samples using a false discovery rate of 5%. Amongst the shared Cufflinks transcripts 278 were found to be differentially expressed (Table S1) using the Cuffdiff min-alignment-count parameter as a threshold (see Methods S1). Of these transcripts 224 were upregulated and 54 were downregulated in the sulfate-deficient cultures. In addition, 29 E. huxleyi transcripts were only expressed in the ehux25 dataset and 1029 E. huxleyi transcripts were expressed solely in ehux5 (Table S2).
Table 1. Numbers of RNAseq reads mapping to transcriptome and genome
No. reads in dataset (after QC)
No. reads unmapped to transcriptome
% reads unmapped to transcriptome
No. reads unmapped to genome
% reads unmapped to genome
Given the large number of reads unmapped to the transcriptome, we mapped all reads to the E. huxleyi reference genome (Emihu1_scaffolds.fasta) to identify previously undiscovered transcripts. Using this approach, we found that 48 367 452 reads mapped to the reference genome, leaving 10 504 078 (17.78%) of the reads unmapped (Table 1). Through this mapping we identified 20 416 transcribed genomic loci (Fig. 5) and compared them to the 39 125 predicted transcripts. The total transcript space for our models was 42 331 395 nucleotides with an average transcript length of 2073 bp (compared to 67 258 384 bp in Emihu1_best_transcripts, with an average transcript-length of 1719 bp). Of our assembled transcripts, 15 680 463 nucleotides had no overlap with the E. huxleyi predicted transcripts (i.e. they were ‘missing’ from the E. huxleyi annotated gene models). This assembly resulted in 11 576 novel transcripts (not overlapping any of the predicted E. huxleyi transcripts) and 8840 transcripts that overlapped with the annotated transcripts. The amount of overlap between our newly assembled transcripts and the predicted gene models at the JGI site was 46.15%. Qualitatively, this is consistent with analysis of general gene expression in P. tricornutum by EST sequencing which identified an additional 1968 transcripts not predicted as gene models (Maheswari et al., 2010), and the 3470 unpredicted transcripts in T. pseudonana tiling array data (Mock et al., 2008). The large number of unannotated transcripts is, however, surprising and may reflect the difficulties of the gene prediction software regarding the high GC content of E. huxleyi genome. Thus, the E. huxleyi genome and transcriptome are much larger than those of the two sequenced diatom species. The calculation of differential expression of these transcripts determined by Cuffdiff resulted in identification of 325 loci, of which 254 were upregulated and 71 downregulated (Table S3). Among the upregulated and downregulated transcripts that mapped to genome scaffolds 121 and 31 transcripts, respectively, did not overlap with any gene models (Table S3).
The 10 504 078 reads that would not map to the reference genome (3467 678 paired-end reads and 3568 722 single reads) were assembled de novo. Using the Cufflinks tool suite, 7712 expressed transcripts were identified. Differential expression was then calculated for each locus shared between ehux25 and ehux5. From this we found 335 transcripts that were significantly differentially expressed: 298 upregulated and 37 downregulated in ehux5 (Table S4). The identity of these new transcripts was investigated by BLAST. Some of these unmapped transcripts correspond to chloroplast and mitochondrial transcripts that are not included in the scaffolds; however, the majority of the most highly upregulated genes had no homology with any known sequences. Altogether, the analysis identified 1718 differentially regulated transcripts (325 from genome alignment, 1058 present in one condition only, 335 novel transcripts), 1253 of them upregulated.
The expression pattern obtained by RNA sequencing was verified by quantitative RT-PCR (qPCR) for seven differentially regulated genes with gene IDs: 450514, 452597, 315901, 229382, 456731 (upregulated), and 454260 and 432295 (downregulated). For all these genes the difference in transcript abundance according to RNA sequencing data agreed well with the qPCR results (Fig. S2).
Functional categories of genes affected by sulfate limitation in E. huxleyi
In order to identify functional categories, the transcripts were annotated using the Superfamily database v1.73 and this enabled description of 14 249 genes for which domain assignment was available. To obtain information on biological processes affected by sulfate deficiency the genes were further characterized by KEGG and KOG annotation (Ogata et al., 1999; Tatusov et al., 2003). Fig. 6 shows the distribution of KOG functional categories amongst the transcripts upregulated by sulfate deficiency. The most prominent functional groups were ‘Signal transduction mechanisms’ and ‘Post-translational modification, protein turnover, chaperones’.
Iterative Gene Analysis (iGA) gave deeper insight into the biological processes affected by sulfate deficiency. Using the KEGG pathway classification 2723 genes were annotated and assigned to 90 metabolic pathways. The iGA identified 23 upregulated and 30 downregulated KEGG pathways in the transcriptome of sulfate-deficient cells (Table S5). Amongst the upregulated pathways four can be assigned to carbohydrate metabolism (ascorbate and aldarate, butanoate, and pyruvate metabolism, and citrate cycle) and 5 pathways to lipid metabolism (fatty acid, bile acid, sphingoglycolipid, prostaglandin and leukotriene, and glycerolipid metabolism). By contrast, the downregulated gene set had no single prominent group, but is rather a representative cross-section of various KEGG classes. Among the KEGG pathways affected by sulfate deficiency, several linked to sulfur metabolism were found. Glutathione metabolism was found in the upregulated pathways, whereas methionine, thiamine and serine metabolic pathways were downregulated. Interestingly, degradation pathways for a range of xenobiotics were downregulated under sulfate deficiency, likely reflecting the reduced availability of GSH.
The iGA analysis was also performed using KOG annotation of 9725 transcripts clustered into 2254 KOG functional groups. Table S5 shows the 40 and 23 KOG groups enriched amongst the up- and downregulated transcripts, respectively. The KOG groups uniquely detected in the upregulated list included 10 clusters assigned to the ‘Cytoskeleton’ class, five clusters to ‘Inorganic ion transport and metabolism’ and single clusters assigned to ‘Energy production and conversion’, ‘Nuclear structure’ and ‘Function unknown’. By contrast, the groups appeared more evenly distributed throughout the downregulated list. Importantly, sulfate deficiency resulted in an enrichment of the ‘Sulfate/bicarbonate/oxalate exchanger SAT-1 and related transporters (SLC26 family)’ class, driven mainly by strong induction of two sulfate transporter transcripts (441761, 453061), which agrees well with the increased sulfate uptake capacity of sulfate-deficient E. huxleyi cells (Fig. 4).
Transcriptional regulation of enzymes of sulfur metabolism
Given the known effects of sulfate deficiency on sulfate uptake and assimilation in various organisms, we compared the expression of genes involved in these processes in sulfate-deficient and control E. huxleyi cells. It was shown previously that the sulfate assimilation pathway in E. huxleyi is organized in a similar way to plants (Kopriva et al., 2009) and the corresponding genes have been identified (Table S6).
Our BLAST analysis identified 16 putative sulfate transporters (STR) in the haploid E. huxleyi genome (Table S6, Fig. S3). The transporters fall broadly into four groups, the plant/fungi/animal SCL26 type, the SAC1/SLT Na+/ co-transporter family, the SLC13 family, and several other transporters characterized by the sulfate transporter and anti-sigma antagonist (STAS) domain (Takahashi et al., 2011b). In accordance with the increased sulfate uptake (Fig. 4), transcripts of three putative sulfate transporters STR1 (protein ID 363809), STR2 (441761), and STR3 (453061) increased in sulfate-limited cells 5-, 10- and 15- fold, respectively (Fig. 7, Table S6). These genes encode transporters of the SLC26 family and contain the characteristic STAS domain (Takahashi et al., 2011b), and therefore probably represent real sulfate transporters. In addition, one gene (STR13; 230466) from the group of STAS-containing genes and an SLC13 family gene (STR16; 443760) have been upregulated. This agrees well with the induction of sulfate transporters in Chlamydomonas (González-Ballester et al., 2010) and Arabidopsis (Maruyama-Nakashita et al., 2006) and is shown for comparison in Fig. 7. However, it is highly probable that not all the 16 STR genes encode genuine sulfate transporters. The SLC13 transporters often function as Na+ di- or tricarboxylate carriers and may not participate in sulfate uptake in E. huxleyi. The genes of the SAC1/SLT family are only distantly related to their Chlamydomonas counterparts and their role in sulfate transport has yet to be established, particularly as they were not regulated by sulfate deficiency in E. huxleyi, whereas the Chlamydomonas SLT1 and SLT2 are upregulated by sulfate starvation (Pootakham et al., 2010).
A common response to sulfate deficiency is the increase in sulfate reducing capacity by upregulation of APR in plants (Nikiforova et al., 2003), or ATP sulfurylase and sulfite reductase in Chlamydomonas (Ravina et al., 2002). Surprisingly, however, in E. huxleyi neither APR activity nor transcript levels for APR, ATPS and SiR were affected by sulfate deficiency (Fig. 7). Interestingly, while APR activity in E. huxleyi was not regulated by changes in sulfate concentrations, it was c. 10-fold higher than APR activity typically measured in Arabidopsis or Chlamydomonas (Ravina et al., 2002; Vauclare et al., 2002). APR activity and the general capacity to reduce sulfate might thus be high enough in E. huxleyi, such that further increases would be meaningless.
The transcript levels of genes involved in cysteine synthesis, OASTL4 (452198), OASTL6 (445218), OASTL7 (430252), OASTL8 (440100) and OASTL10 (442172), isoforms of OAS thiollyase, and SAT3 (248485) and SAT4 (234967) of serine acetyltransferase (SAT), increased under sulfate deficiency. Thus, in all three organisms, E. huxleyi, A. thaliana and C. reinhardtii, at least one isoform of SAT and OAS thiollyase were induced by sulfate deficiency (Fig. 7). In plants SAT has an important role in controlling the sulfate assimilation pathway: its overexpression increases the content of sulfur-containing metabolites and strong inhibition leads to growth reduction (Blaszczyk et al., 1999; Haas et al., 2008). The increase in SAT expression in E. huxleyi may facilitate cysteine synthesis when the substrate concentration is strongly diminished.
The regulation, or the lack thereof, of STR1, STR2, STR3, ATPS1, APR, SAT3 and OASTL6 was confirmed by qPCR (Fig. S2, Table S6).
GSH metabolism was found among the KEGG pathways upregulated by sulfur limitation (Table S5). This was mainly because of a strong upregulation of four genes encoding GSH-transferases (233986, 224152, 349113, 442908) and a GSH peroxidase (433534), which are connected with oxidative stress rather than GSH metabolism and reflect the general upregulation of stress-related genes by sulfur deficiency. However, two isoforms of GSH synthetase (51736, 121060) were also strongly, c. six-fold, upregulated as well as a minor isoform of γ-glutamylcysteine synthetase (113513) the first enzyme in GSH synthesis. This contrasts with no transcriptional regulation of GSH synthesis in Arabidopsis and Chlamydomonas. However, in Arabidopsis the γ-glutamylcysteine synthetase, which has much higher control over GSH synthesis, is regulated post-translationally by the redox state (Hicks et al., 2007). The genes for this enzyme in E. huxleyi have a different evolutionary origin, being more similar to animal genes than plant ones, and therefore the pathway may be regulated differently.
In E. huxleyi, methionine is not only an essential amino acid for protein synthesis, but also a precursor for DMSP synthesis. Interestingly, Met metabolism was among the KEGG pathways downregulated by sulfur limitation. Indeed, the genes for two components of S-adenosylmethionine (SAM) cycle, SAM synthase and S-adenosylhomocysteine hydrolase were significantly downregulated, by 15% and 40%, respectively (Table S6). On the other hand, homocysteine S-methyltransferase and cobalamin-independent methionine synthase, catalysing the last step of Met synthesis, were induced by sulfur deficiency. This response is similar to Chlamydomonas rather than to Arabidopsis, because in the green alga the genes of SAM cycle were downregulated while they were upregulated by S deficiency in Arabidopsis (Nikiforova et al., 2003; González-Ballester et al., 2010). On the other hand, Met metabolism is coordinately downregulated in Chlamydomonas (González-Ballester et al., 2010), whereas some genes for Met synthesis were upregulated in E. huxleyi. The genes of the alternative pathway of Met recycling, the Yang cycle, have been found in E. huxleyi, but expressed to very low levels, particularly compared to the SAM cycle (Table S6), so this pathway may not play a very important role. Interestingly, all five genes involved in the SAM cycle belong amongst the 60 most highly expressed genes in control cultures of E. huxleyi, as judged from the FPKM values, pointing to a very high importance of this pathway for the alga, connected most probably with high need for methylation, including DMSP synthesis.
In sulfate-deficient Arabidopsis sulfolipid content is strongly reduced and the genes for sulfolipid synthesis are downregulated (Nikiforova et al., 2003). By contrast, the genes are upregulated in C. reinhardtii (González-Ballester et al., 2010), reflecting the much larger sulfolipid pool in this alga, where sulfolipids are actively degraded as a source of sulfur for protein synthesis (Sugimoto et al., 2007, 2010). In E. huxleyi sulfolipid synthesis genes were not differentially regulated, suggesting that sulfolipid turnover is not affected as in Chlamydomonas. This might reflect the importance of sulfolipids for marine organisms adapted to low phosphate availability (Van Mooy et al., 2006) on the one hand, and the presence of a large sulfur pool in DMSP suitable for sulfur recycling during sulfate limitation, on the other.
Chlamydomonas responds to sulfate deficiency by induction of extracellular sulfatases that allow utilization of organic sulfates (de Hostos et al., 1988). These enzymes are not present in higher plants, but in E. huxleyi three transcripts for arylsulfatases (95583, 107777, 433677) were found only in the transcriptome of the sulfate-deficient cells (Table S2) suggesting a similar mechanism for sulfur scavenging.
While the response of gene expression and metabolite accumulation to sulfate deficiency has been well described, much less is known about the molecular mechanisms of sulfate-sensing and signalling. In Arabidopsis, the SLIM1 transcription factor is responsible for the upregulation of sulfate transporter genes (Maruyama-Nakashita et al., 2006), whereas in Chlamydomonas the SAC1, Na+/ transporter seems to be the sensor of sulfate status, and SNRK2.1 and SNRK2.2 (SAC3) are essential for the transcriptional response (Davies et al., 1999; González-Ballester et al., 2008). In the E. huxleyi genome there are several genes belonging to the same family as SAC1. However, SAC1 itself is not regulated by sulfate starvation in Chlamydomonas (González-Ballester et al., 2010), and the same is true for all the E. huxleyi genes of the SAC1/SLT group of transporters. Similarly, there are > 90 genes with similarity to SAC3 in the E. huxleyi genome so it is impossible to assign a similar function to any of them. No protein homologous to SLIM1 is encoded in E. huxleyi genome.
General response to sulfate deficiency
The fundamental difference in the response to sulfate deficiency in E. huxleyi compared to Arabidopsis and Chlamydomonas is the ratio between upregulated and downregulated genes. The general response to prolonged sulfate deficiency in plants, equivalent to the late acclimation phase of the E. huxleyi cultures, is a slowing down of metabolism and shortening of the life cycle (Hoefgen & Nikiforova, 2008). Accordingly, in multiple microarray experiments significantly more transcripts were repressed by sulfate starvation rather than induced (Hirai et al., 2003; Maruyama-Nakashita et al., 2003; Nikiforova et al., 2003). The same was true for Chlamydomonas, where > two-fold more transcripts were downregulated by sulfate deficiency than upregulated (González-Ballester et al., 2010), and for D. salina where sulfate deficiency resulted in decreased Rubisco accumulation and PEP carboxylase and nitrate reductase activities (Giordano et al., 2000). By contrast, in E. huxleyi 1029 transcripts were present only in sulfate-deficient cells compared to 29 that were found only in the controls. Also among those transcripts detected in both conditions, > four-fold more genes were upregulated than downregulated. The upregulated transcripts include many that reflect general stress including GSH transferases (see ‘Glutathione metabolism’ above), flavodoxin (ID 68288), protein disulphide isomerase (443239, 447219), or immunophillin (435425) (Table S1), all of which are also induced in sulfate-deficient Chlamydomonas (González-Ballester et al., 2010). Among other stress-related genes upregulated in sulfur-limited E. huxleyi, two encode GSH peroxidase (433534, 67177; Table S6). GSH peroxidase is particularly interesting, because in many organisms this enzyme contains selenocysteine (Forstrom et al., 1978). Emiliania huxleyi also synthesizes selenoproteins, but it is unique in primarily taking up selenite and not selenate (Araie & Shiraiwa, 2009). The two confirmed selenoproteins of E. huxleyi (443239, 417208) were also upregulated by sulfur deficiency. Transcripts encoding several selenoproteins, and two Se binding proteins, accumulated in sulfur deprived Chlamydomonas cells while a gene encoding Se-binding protein was upregulated in Arabidopsis (Nikiforova et al., 2003; González-Ballester et al., 2010). This regulation in Chlamydomonas and Arabidopsis may be a response to increased uptake of selenate, caused by upregulation of sulfate transporters, which are capable of uptake of selenate, but not selenite. Thus, the driver for upregulation of the selenoproteins in E. huxleyi is more probably their function in stress response and not a sink for Se.
The decline in transcript levels for genes associated with photosynthetic electron transport, chlorophyll biosynthesis, and light harvesting observed in Arabidopsis and Chlamydomonas (Nikiforova et al., 2003; González-Ballester et al., 2010) were not observed in E. huxleyi. In accordance with this, our Fv/Fm data showing no effect of the limitation on Photosystem II quantum yield indicated that the substantial decrease in photosynthesis and chlorophyll synthesis observed in sulfate-deficient plants and green algae (Wykoff et al., 1998; Giordano et al., 2000; Maruyama-Nakashita et al., 2003; González-Ballester et al., 2010) might possibly not be so dramatic in E. huxleyi. Photosynthesis, however, has a high demand for reduced sulfur to ensure synthesis of proteins and co-enzymes. It is possible that during the acclimation response to sulfate limitation E. huxleyi uses sulfur re-allocated from the large DMSP pool and so does not need to reduce their synthesis and limit photosynthesis.
The level of intracellular DMSP decreased concurrently with decreasing sulfate in the E. huxleyi cultures. Rather than a simple decrease in concentration due to reduced sulfate availability, this might be an active process to redirect sulfur from DMSP into other metabolic processes. This is corroborated by upregulation of genes involved in the synthesis of alternative osmolytes proline and glycine betaine, pyrroline-5-carboxylate reductase (protein ID 349043) and betaine-aldehyde dehydrogenase (437142, 417844), respectively. Unfortunately, the genes involved in DMSP synthesis have not yet been unequivocally identified, so it is impossible to establish whether the decrease in DMSP is caused by the downregulation of its synthesis. Lyon et al. (2011) proposed four enzymes to catalyse DMSP synthesis based on their regulation by salinity in the diatom Fragilariopsis cylindrus. E. huxleyi homologues of two of these genes, aminotransferase (456646, 369841) and diaminopimelate decarboxylase (438904), were highly expressed as expected for a major pathway, but not significantly (q > 0.05) regulated. The other two genes, S-adenosylmethionine methyltransferase (464166, 254918) and NADPH reductase (100136, 106956, 120452), were either not expressed at all or expressed to a very low level and are thus very unlikely to participate in DMSP synthesis in this alga. However, in line with the sulfur redistribution hypothesis, (Stefels, 2000), a significant upregulation of two genes encoding proteins with IDs 459683 and 470487 was observed. These genes, annotated as Class III acyl CoA transferases, are homologues of the bacterial DddD (DMSP-dependent DMS production) genes involved in DMSP degradation to DMS and 3-hydroxypropionate (Todd et al., 2007, 2010). DMSP synthesis is not only a large pool for sulfur, but also a significant sink for carbon. One of the proposed functions of DMSP is an overflow metabolite allowing safe dissipation of excess energy and reducing power (Stefels, 2000). In sulfate-deficient E. huxleyi cells transcript levels increased for genes involved in the citric acid cycle –succinyl-CoA synthetase (417649), succinate dehydrogenase (432409) and citrate synthase (467883) – and in fatty acid biosynthesis – acetyl-CoA carboxylase (455280), beta-ketoacyl-ACP reductase (433820), or acyl-CoA dehydrogenase (437926). It is thus possible to speculate that carbon that cannot be used for DMSP synthesis might be redirected into synthesis of fatty acids, as was seen previously for diatoms subjected to nitrogen deficiency (Hockin et al., 2012).
Another process specific to E. huxleyi, which may be relevant for its response to sulfate starvation, is calcification. Emiliania huxleyi can respond to the needs to dissipate excess energy, for example, during high light intensities, by increasing the degree of calcification (Paasche, 2001; Xu & Gao, 2012). Such similar physiological roles of calcification and DMSP metabolism may explain the overall higher intracellular DMSP concentration and its notably greater decrease under low sulfate concentrations in naked cells of E. huxleyi CCMP 1516 from this study compared to the calcifying strain PML92/11 from Ratti et al. (2011). A series of experiments using calcifying strains subjected to various concentrations of sulfate, calcium and irradiance intensity would shed more light on the link between DMSP metabolism and calcification.
In conclusion, we have shown that, despite being adapted to high sulfate concentrations in seawater, the marine microalga Emiliania huxleyi still retains the genetic program to respond to artificial sulfate deficiency. Whereas the upregulation of sulfate uptake and cysteine synthesis in E. huxleyi is in common with plants and freshwater algae, the general response is significantly different. Instead of slowing down photosynthesis and primary metabolism E. huxleyi responds to sulfate deficiency by upregulation of genes involved in carbohydrate and fatty acid synthesis, and appears to redirect sulfur and carbon from DMSP into these alternative metabolite pools. Whether this type of response to sulfate deficiency is a specific feature of E. huxleyi or is common among diverse marine algae taxa remains to be elucidated.
M.B. was supported by a University of East Anglia (UEA) Zuckerman PhD Studentship and T.B. was supported by a Natural Environment Research Council (NERC) UK SOLAS Knowledge Transfer grant (NE/E001696/1). S.K.'s research is supported by BB/J004561/1 grant from BBSRC and the John Innes Foundation. G.M. was funded through a UK NERC Advanced Fellowship (NE/B501039/1). The RNAseq was supported by NERC Biomolecular Analysis Facility grant MGF317 and the JIC and UEA Earth and Life Systems Alliance. We are grateful to the US Department of Energy Joint Genome Institute and the scientific community who produced the E. huxleyi 1516 genome sequence. We thank Urmi Trivedi, GenePool Edinburgh, for initial bioinformatics analysis of RNAseq data. We are also grateful to Gareth Lee and Rob Utting at UEA for technical support.