• Open Access

New insights into Chlamydomonas reinhardtii hydrogen production processes by combined microarray/RNA-seq transcriptomics


Correspondence (fax +49-521-10612290;

email Olaf.kruse@uni-bielefeld.de)


Hydrogen production with Chlamydomonas reinhardtii induced by sulphur starvation is a multiphase process while the cell internal metabolism is completely remodelled. The first cellular response is characterized by induction of genes with regulatory functions, followed by a total remodelling of the metabolism to provide reduction equivalents for cellular processes. We were able to characterize all major processes that provide energy and reduction equivalents during hydrogen production. Furthermore, C. reinhardtii showed a strong transcript increase for gene models responsible for stress response and detoxification of oxygen radicals. Finally, we were able to determine potential bottlenecks and target genes for manipulation to increase hydrogen production or to prolong the hydrogen production phase. The investigation of transcriptomic changes during the time course of hydrogen production in C. reinhardtii with microarrays and RNA-seq revealed new insights into the regulation and remodelling of the cell internal metabolism. Both methods showed a good correlation. The microarray platform can be used as a reliable standard tool for routine gene expression analysis. RNA-seq additionally allowed a detailed time-dependent study of gene expression and determination of new genes involved in the hydrogen production process.


Research in hydrogen production in Chlamydomonas reinhardtii is of specific interest, because alternative renewable energy sources are highly desired and needed in the near future. It is known that C. reinhardtii produces molecular hydrogen under anaerobic conditions, a phenomenon that can be induced by sulphur deprivation (Melis and Happe, 2001). Under such conditions, C. reinhardtii cells remodel their internal metabolism and use hydrogen production as a valve system to prevent over-reduction of the chloroplast stroma. During the last decade, intensive research has been conducted on aspects of the cellular metabolism connected to the H2 production process. These investigations include cellular processes such as fermentation (Catalanotti et al., 2012; Hemschemeier et al., 2008b; Mus et al., 2007; Philipps et al., 2011), photosynthesis and respiration (Chochois et al., 2009, 2010; Desplats et al., 2009; Esper et al., 2006; Kruse et al., 2005; Lecler et al., 2011; Tolleter et al., 2011; Torzillo et al., 2009) and the regulation of the hydrogenase HydA (Happe and Kaminski, 2002; Happe et al., 2002). Several new findings have demonstrated remarkable reorganization capacities of C. reinhardtii, especially the compensation of knocked-out or down-regulated genes related to fermentation processes (Catalanotti et al., 2012; Grossman et al., 2011; Magneschi et al., 2012; Philipps et al., 2011). Additionally, hydrogen production has been investigated through several systems biology approaches regarding changes in the metabolome (Doebbe et al., 2010; Matthew et al., 2009), in the proteome (Chen et al., 2010) and in the transcriptome (Mus et al., 2007; Nguyen et al., 2008) of C. reinhardtii to understand cellular adaptation in detail and to identify new targets or bottlenecks to improve hydrogen yield. A particular research focus was set on the cellular adaptation to sulphur stress induced anaerobiosis, resulting in the induction of distinct genes responsible for sulphate transport and assimilation, accompanied by a repression of gene expression of the vast majority of genes related to photosynthetic processes (Mus et al., 2007; Nguyen et al., 2011). These gene adaptation mechanisms are reported to be followed by the over-expression of gene products required for starch and lipid synthesis (Doebbe et al., 2010; Matthew et al., 2009) and in acclimatization/modulation processes that include changes in the amino acid composition of certain target proteins (Doebbe et al., 2010). The hydrogen production phase is characterized by the expressed and active hydrogenase, which is competing under anaerobic conditions for electrons derived from photosynthesis (Chochois et al., 2009, 2010; Doebbe et al., 2010) with fermentative-related enzymes, such as alcohol dehydrogenase (Hemschemeier and Happe, 2011; Hemschemeier et al., 2008a,b; Philipps et al., 2011).

To improve hydrogen production in C. reinhardtii, specific knowledge of the bottlenecks in all pathways and processes is required. Therefore, new, detailed analyses of the transcriptome and proteome under different conditions are necessary to elucidate these limitations and identify potential targets for improvement. Rapid progress in genome annotation of C. reinhardtii resulted in the design of a new microarray platform for advanced and general transcriptome analyses (Toepel et al., 2011). Chlamydomonas reinhardtii full-genome microarrays enable us to determine expression level variations of ~11 000 gene models. New annotations, however, predicted nearly 20 000 gene models, and RNA-seq data have been used to provide new insights into transcriptomic changes during nitrogen (Miller et al., 2010), sulphur (Boyle et al., 2012; Gonzalez-Ballester et al., 2010) or carbon dioxide (Fang et al., 2012) limitation. In contrast to RNA-seq, microarrays are relatively inexpensive, reliable systems for use on routine basis and have the potential to give a rapid overview of variations in transcript levels. RNA-seq platforms have many advantages in comparison with microarray technologies, such as higher gene coverage and increased sensitivity for differential gene expression. Additionally, predictions of new gene models and splicing variations can be realized by RNA-seq as well as the detection and characterization of mutation sites (Smith et al., 2008). However, data obtained from RNA-seq require stringent examination, and reproducibility of results is often low. As a typical consequence, an overestimation of highly abundant genes and a length-dependent amplification have been reported by this method. Full data analysis and data normalization for RNA-seq experiments is not yet standardized (Liu et al., 2011; Malone and Oliver, 2011; Oshlack et al., 2010; Wang et al., 2011). Here, we compare expression levels of induced/repressed transcripts generated by both methods to further characterize this process and identify new potential bottlenecks of hydrogen production in C. reinhardtii. However, full-data analysis and data normalization is not standardized yet (Ramskold et al., 2012). To validate RNA-seq data and to confirm the usability of microarrays, we intended to compare expression level of induced/repressed transcripts of C. reinhardtii during hydrogen production with both methods to further characterize this process and to identify new potential targets to reduce limitations by bottlenecks.


Hydrogen production in a mutant defective in non-photochemical quenching

Within this project, the C. reinhardtii strain npq4 (Niyogi et al., 1997) and its corresponding wild-type 4A+ were used to identify key elements of hydrogen production processes. Npq4 was used as a reference strain for these comparative analyses to evaluate the accuracy of both RNA-seq and microarray analyses. Npq4 is a mutant defective in non-photochemical quenching, which was suggested to yield in clear differences regarding its transcriptome upon stress induction by sulphur depletion and anaerobiosis (Bonente et al., 2011; Peers et al., 2009). Hydrogen production from C. reinhardtii induced by anaerobiosis via sulphur depletion is a multiphase process, and Npq4 demonstrated a typical course of hydrogen production; a lag phase of 24 h until hydrogen production began, followed by a ~72-h production phase with a total production of around 120–150 mL H2 per litre culture (Figure 1). This total hydrogen production rate in the mutant was only marginally higher when compared to the corresponding wild-type 4A+ (90–110 mL H2). PAM fluorescence measurements revealed that effective photosynthetic quantum yields decreased in both strains from 0.5 to 0.05 within the first 24 h followed by consistent photosynthetic efficiency reflected by quantum yields of ~0.01 for the remaining duration of the experiment (Figure 1).

Figure 1.

Hydrogen production (▼) and effective quantum yield (■) during hydrogen production of Chlamydomonas reinhardtii strains npq4 (grey) and 4A+ (black).

Comparison of microarray and RNA-seq to identify differential gene expression

Microarray and RNA-seq analyses were performed to identify and determine changes of transcript levels for distinct genes over the entire period of hydrogen production in npq4. The subsequent analysis of microarray data included the expression pattern for ~10 000 gene models (Toepel et al., 2011). We used a gene expression threshold with a cut-off of twofold for both up- or down-regulation and could identify 603 gene models with a down-regulation and 635 gene models with an up-regulation (see Figure 2 solid box: 635 genes red; 603 genes green). Quantitative RT-qPCR of several control genes were in accordance with our microarray data (see Supporting information) and are also in agreement with previous studies (Nguyen et al., 2011) performed with a different first-generation microarray platform (Eberhard et al., 2006). It should be noted that the microarray platform used in our study included almost 4000 new gene models including a large number of unknown gene models (Toepel et al., 2011). As a consequence, the majority of differentially expressed genes observed in this study are unknown or not yet fully characterized. However, we could confirm up-regulation of most of the genes identified in a previous study (Nguyen et al., 2011), for example, genes coding for proteins involved in sulphur metabolism/catabolism as well as transcripts related to lipid and starch metabolism (see Table 1). Additionally, the up-regulation of genes related to the pentose phosphate cycle and for genes related to fermentation processes could be confirmed. One major result, the gene expression of the isocitrate lyase (Icl), the key enzyme for the TCA cycle shunt, could be confirmed with our experiments. We were also able to verify the down-regulation of the majority of photosynthetic genes (PS I; PS II and ATPase) and increased transcript levels of Lhcbm9 and Lhcsr1 (Table 1), two distinct genes related to light harvesting and energy quenching (determined with RT-PCR from (Nguyen et al., 2011). Additionally, as we used Npq4, a mutant deficient in Lhcsr3.1 and Lhcsr3.2, we could control the specificity of the read alignment. Within both methods, RNA-seq and microarray analyses, no Lhcsr3.1 and Lhcsr3.2 reads could be identified in data sets derived from npq4 cells during hydrogen production. In contrast, comparative wild-type analyses showed a strong gene expression of Lhcsr3 genes, clearly demonstrating the reliability of both methods (Nguyen et al., 2011; Toepel et al., 2011).

Table 1. Differential gene expression for Chlamydomonas reinhardtii grown under hydrogen-producing conditions, determined with microarray and RNA-seq. Shown are the log2 ratios for both methods as mean values over the time course of the experiments
LocusNameDescriptionLog2 ratio microarrayLog 2 ratio RNA seq
Sulphur-related gene models
Cre12.g517150APR3APS reductase 31.71.9
Cre06.g273750SUA1Chloroplast sulphate transporter2.72.1
Cre03.g160400SAC1Sulphur acclimation protein−0.6−1.7
Cre01.g021200CGL47Control protein0.41.8
Cre01.g012150PMSR3Peptide methionine sulphoxide reductase 31.61.7
Cre06.g257650PMSR4Peptide methionine sulphoxide reductase 42.45.9
Cre03.g210200SOXSulphite oxidase2.23.0
Cre16.g693150SIR1Sulphite reductase3.16.3
Cre08.g365700SIR2Sulphite reductase0.61.9
Cre03.g210200SOXSulphite oxidase2.23
Cre12.g502600SLT1Sulphate transporter5.65.4
Cre10.g445000SLT2Sulphate transporter3.86.8
Cre02.g138950SULTR4;1Sulphate transporter 4.1−2.6−3.4
Cre13.g573250STR16Sulphate transporter−2.3−5.6
Cre13.g607050STR1Sulphate transporter0.74.4
Cre13.g597450 Sulphate transporter (glutaredoxin)1.32.9
Cre16.g656400SQD1Sulfoquinovosyldiacylglycerol 10.83.3
Cre16.g689150SQD2Sulfoquinovosyldiacylglycerol 20.0n.d
Cre01.g038550SQD2Sulfoquinovosyldiacylglycerol 21.81.7
Cre14.g615000MSRB2Sulphate transporter0.23.4
Cre02.g097900AAT5Aspartate aminotransferase 51.62.7
Cre07.g319400ACD1d-cysteine desulfhydrase0.23.2
Cre01.g012150ATMSRA3Peptide methionine sulphoxide reductase 31.61.7
Cre01.g036750 S-adenosyl-l-methionine-dependent methyltransferase1.83
Cre06.g288550ECP 76Extracelullar protein3.27.9
Cre12.g556000ECP 88Extracelullar protein3.27.9
Photosynthetic-related gene models
Cre06.g284200Lhcbm9PS II light harvesting complex5.37.6
Cre03.g156900Lhcbm5PS II light harvesting complex−0.3−5.6
Cre06.g283950Lhcbm4PS II light harvesting complex−1.4−3.3
Cre12.g548400Lhcbm2PS II light harvesting complex1.10.1
Cre12.g548950Lhcbm7PS II light harvesting complex−3.9−3.1
Cre23.g766250Lhcbm1PS II light harvesting complex−1.3−2.3
Cre56.g791050PSAD-2PS I subunit D-2−1.7−3.2
Cre10.g420350PSAE-2PS I subunit E-2−1.6−4.0
Cre09.g412100PSAFPS I subunit F−1.5−2.1
Cre12.g560950PSAGPS I subunit G−1.5−4.3
Cre07.g330250PSAH-1PS I subunit H-1−1.5−5.7
Cre17.g724300PSAKPS I subunit K−1.3−6.4
Cre12.g486300PSALPS I subunit l−2.0−4.5
Cre02.g082500PSANPS I reaction center PSI-N−1.3−3.1
Cre27.g775100PSAPPS I subunit P4.14.8
Cre07.g334550PSAOPS I subunit Pn.d.−10.0
Cre02.g124700MDB2Nac2 factorn.d.−2.8
Cre05.g243800PSB27PS II family protein−0.6−0.6
Cre08.g372450PSBQPS II subunit Q−0.6−3.3
Cre06.g261000PSBRPS II subunit R1.41.3
Cre11.g475250PSBWPS II reaction center W1.2−0.4
Cre02.g082750PSBXPS II subunit X−1.8−5.2
Cre02.g132800PSBO1PS II oxygen-evolving complex 1−1.5−5.0
Cre16.g650100PETNCytochrome b6f complex PetN−2.7−6.6
Cre18.g744400PETCCytochrome b6f complex PetC−2.7−4.3
Cre03.g182551PETE1Electron transporter; plastocyanin 1−3.0−5.0
Cre14.g626700PETFCytochrome b6f complex PetF−1.3−7.5
Cre12.g546150PETMCytochrome b6f complex PetM−2.1−6.2
Cre06.g283050LHCA1PS I light harvesting complex−2.6−2.9
Cre12.g508750LHCA6PS I light harvesting complex−2.7−3.4
Cre10.g454750LHCA3PS I light harvesting complex0.81.2
Cre18.g749750LHCA3PS I light harvesting complex−2.3−4.8
Cre06.g272650LHCA5PS I light harvesting complex−2.4−1.5
Cre07.g344950LHCA5PS I light harvesting complex−2.4−4.0
Cre16.g687900LHCA5PS I light harvesting complex−2.8−5.2
Cre10.g425900LHCA4PS I light harvesting complex−2.9−3.0
Cre13.g598900LHCA4PS I light harvesting complex−2.3−3.9
Cre08.g365900LHCSR1Chlorophyll A/B binding protein 16.98.2
Cre08.g367400LHCSR3Chlorophyll A/B binding protein 30.00.0
Cre08.g367500LHCSR2Chlorophyll A/B binding protein 20.1−0.1
Cre03.g148750CLH1Chlorophyllase 12.51.3
Cre10.g423500HO3Haeme oxygenase 30.52.7
Cre13.g600650 Pheophorbide a oxygenasen.d.2.4
Cre02.g120100RBCS1ARUBISCO small chain 1A−2.4−2.4
Cre02.g120150RBCS2RUBISCO small chain 2−2.5−3.6
Cre27.g774300 RUBISCO methyltransferase−2.0−4.2
Cre02.g129750 RUBISCO methyltransferase2.2n.d.
Cre03.g186450 RUBISCO methyltransferase−0.2−2.4
Cre04.g229300RCARUBISCO activase−1.3−0.7
Cre08.g368700 RUBISCO methyltransferase0.2−0.8
Cre12.g503800 RUBISCO methyltransferase−0.5−2.3
Cre16.g661350RMT1RUBISCO large subunit methyltransferase2.6n.d.
Carbohydrate-related gene models
Cre08.g385500AMA1Alpha-amylase 1n.d.1.6
Cre08.g362450AMA2Alpha-amylase 21.01.7
Cre19.g755050ISA3Isoamylase 3n.d.1.4
Cre03.g185250SS2Starch synthase 2n.d.2.4
Cre17.g721500 Granula bound starch synthasen.d.2.6
Cre10.g444700SBE2.2Starch branching enzyme 2.2−0.7−1.5
Cre11.g476650 Starch debranching enzymen.d.1.7
Cre07.g336950PHSStarch phosphorylase1.62.2
Cre12.g552200PHSStarch phosphorylase0.41.9
Cre03.g175400PGI1Glucose-6-phosphate isomerase0.41.3
Cre12.g553250PFK5Phosphofructokinase 52.22.8
Cre01.g029300TPI1Triosephosphate isomerase 11.43.5
Cre12.g485150GAPCP-1Glyceraldehyde-3-phosphate dehydrogenase 1P3.94.5
Cre22.g763250PGk1Phosphoglycerate kinasen.d.2.0
Cre06.g272050PGM 1Phosphoglycerate mutase, 2,3-bisphosphoglycerate-independent1.42.7
Cre01.g057900PYK 3Pyruvate kinase 30.42.3
Cre12.g533550PYK1Pyruvate kinase 1−1.0−3.1
Cre02.g141400PCK1Phosphoenolpyruvate carboxykinase 1−1.43.0
Cre05.g241750PDKPyruvate dehydrogenase kinase0.82.0
Cre06.g282800ICLIsocitrate lyasen.d.1.9
Cre08.g378150G6PD3Glucose-6-phosphate dehydrogenase 33.24.2
Cre12.g526800GND16-phosphogluconate dehydrogenase2.34.9
Fermentation-related gene models
Cre20.g758200ADH1Alcohol dehydrogenase2.25.3
Cre01.g044800PFL1Pyruvate formate lyase−2.00.5
Cre09.g396700ACK1Acetate kinase0.62.7
Cre01.g057800MLSMalate synthasen.d.3.5
Cre11.g473950PFRPyruvate ferredoxin reductasen.d.1.7
Cre10.g423250MDH2Malate dehydrogenase 2n.d.4.0
Cre03.g199800HYDA1Ferredoxin hydrogenasen.d.0.5
Cre09.g396600HYDA2Ferredoxin hydrogenase1.03.5
Cre06.g296700 Hydrogenase assembly factorn.d.2.1
Cre06.g296750HYDEFHydrogenase assembly factorn.d.3.0
Lipid metabolism-related gene models
Cre17.g711150FAD2Fatty acid desaturase 23.01.7
Cre23.g765700ACSAcetyl-CoA synthetase2.31.9
Cre05.g248150 Phospholipid/glycerol acyltransferase0.1n.d.
Cre09.g392300 Acyl-CoA N-acyltransferase1.10.4
Cre07.g312400DGK1Diacylglycerol kinase12.7n.d.
Cre10.g422850 Lipase0.32.8
Cre02.g127300 Lipase3.11.9
Cre07.g322900 Lipase−0.11.6
Cre02.g126050 Lipase−4.3−3.6
Cre02.g121200DGTT2Diacylglycerol acyltransferasen.d.2.5
Cre06.g299050DGTT3Diacylglycerol acyltransferase−0.3−1.6
Cre02.g106400PDAT1Phospholipid:diacylglycerol acyltransferase0.20.5
Cre07.g325550DGK4Diacylglycerol kinase 40.5n.d.
Cre17.g707300 Phospholipid/glycerol acyltransferase−0.8−1.3
Cre06.g268200TGD1Trigalactosyldiacylglycerol 11.63.8
Cre16.g694400TGD2Trigalactosyldiacylglycerol 2n.d.1.9
Proteases and protein kinases
Cre10.g459650 Ubiquitin-protein ligase−1.5−2.7
Cre12.g546650UBC7Ubiquitin carrier protein 7−1.9−1.8
Cre12.g521450NCLPP7Nuclear-encoded CLP protease P73.52.5
Cre10.g432150 Protein kinase−2.8−6.2
Cre12.g505250CPK24Calcium-dependent protein kinase 24−1.4−3.2
Cre10.g459650KEGProtein kinase; ubiquitin-protein ligase−1.5−2.7
Cre09.g413400 Protein kinase−2.6−2.4
Cre33.g782700 Protein kinase−3.0−2.2
Cre01.g001200 Protein kinase−2.2−1.9
Cre17.g698550 Protein kinase−1.4−1.9
Cre10.g457700CPK2Kinase cdpk isoform 2−1.6−1.7
Cre12.g549750CKL2Casein kinase-like 2−1.6−1.7
Cre03.g201900 Phosphatidylinositol-4-phosphate 5-kinase−2.9−1.6
Cre10.g466650CPK20Kinase 20−3.0−1.6
Cre16.g654300 Nucleoside diphosphate kinase−1.8−1.5
Cre02.g092150 Protein kinase1.61.5
Cre04.g223200MPK9MAP kinase 91.61.9
Cre02.g126650 Protein kinase2.02.7
Cre03.g173800ATSOS4Carbohydrate kinase1.62.7
Cre06.g255350 Hydroxyethylthiazole kinase2.06.1
Stress related gene models
Cre02.g077100GSH1Glutamate-cysteine ligase0.30.4
Cre17.g708800GSH2Glutathione synthetase 22.82.3
Cre03.g154950 Glutathione transferase1.03.0
Cre16.g688550GST1Glutathione transferase1.56.7
Cre12.g559800 Glutathione transferase1.61.6
Cre01.g064400 Glutathione transferase1.14.9
Cre10.g458450GPX5Glutathione peroxidase 50.72.3
Cre02.g078300GPX6Glutathione peroxidase 61.12.9
Cre02.g139650APX3Ascorbate peroxidase 31.31.6
Cre05.g233900APX4Ascorbate peroxidase 40.40.4
Cre01.g045700APX5Ascorbate peroxidase 50.52.4
Cre06.g285150APX6Ascorbate peroxidase 61.00.5
Cre02.g087700SAPXStromal ascorbate peroxidase2.1−1.6
Cre17.g712100MDAR1Monodehydroascorbate reductase 1−0.7−1.7
Cre06.g271200MDAR4Monodehydroascorbate reductase 4−1.5−3.2
Cre10.g456750DHAR2Dehydroascorbate reductase 20.30.3
Cre01.g044700DHAR3Dehydroascorbate reductase 301.8
Cre10.g456050FQR1Flavodoxin-like quinone reductase 11.32.2
Cre10.g456000FQR2Flavodoxin-like quinone reductase 22.11.8
Cre10.g456100FQR3Flavodoxin-like quinone reductase 33.32.5
Cre03.g167150 Flavin-binding monooxygenase2.62.1
Cre10.g466700 2-oxoglutarate (2OG) and Fe(II)-dependent oxygenase3.25.9
Cre01.g053000 NAD-dependent glycerol-3-phosphate dehydrogenase−1.4−2.4
Cre10.g421700 NAD-dependent glycerol-3-phosphate dehydrogenase0.52.9
Cre11.g472700PGPS2Phosphatidylglycerolphosphate synthase 22.90.3
Cre07.g346800 FAD-dependent oxidoreductase2.73.6
Cre12.g493500 FAD-dependent oxidoreductase2.72.3
Cre16.g671450 FAD-dependent oxidoreductase3.7n.d.
Cre02.g139200 FAD/NAD(P)-oxidoreductase1.8n.d.
Cre09.g395950AOX1Alternative oxidasen.d.1.4
Cre07.g350750PTO1Alternative oxidase1.22.2
Cre03.g172500PTO2Alternative oxidase1.70.9
Cre09.g417150CAT2Catalase 2n.d.1.6
Cre02.g096150MSD1Manganese superoxide dismutase 11.21.5
Cre10.g436050FSD1Fe superoxide dismutase 1n.d.2.8
Cre11.g477200 Isoflavone reductase like protein4.97.7
Cre07.g355500 Oxidoreductase3.62.5
Cre01.g057750 Thioredoxin3.24.1
Cre07.g315100 Thioredoxin2.43.4
Gene of undefined function
Cre02.g094250 Mitochondrial substrate carrier−3.7−1.0
Cre28.g776600 Mitochondrial substrate carriern.d.−2.6
Cre16.g650800TIM13Mitochondrial translocase 13n.d.−2.4
Cre06.g278750 Mitochondrial substrate carriern.d.−2.6
Cre01.g063200ACP1Acyl carrier protein 11.01.6
Cre13.g577100ACP1Acyl carrier protein 1−0.9−2.5
Cre14.g621650 S-malonyltransferase−3.1−3.7
Cre02.g144800NAGS2N-acetyl-l-glutamate synthase 2−0.1−2.4
Cre10.g434800 Stress-inducible protein2.61.7
Cre12.g495850 d-beta-Hydroxybutyrat–Dehydrogenasen.d.2.5
Cre07.g343050  3.75.4
Cre11.g471864  n.d.10.7
Cre56.g791150  3.44.6
Cre03.g177250  n.d.7.3
Cre07.g349350  1.72.9
Cre03.g192350  n.d.10.7
Cre02.g113400 Unknown conserved protein−3.1−2.6
Cre17.g741850 RNA binding protein−1.8−3.9
Cre02.g1146002-Cys Prx B2-cysteine peroxiredoxin B−1.4−1.6
Cre02.g085300  −1.3−4.3
Cre02.g087250  n.d.−2.3
Cre03.g153450  n.d.−1.5
Cre13.g570850  −2.4−4.0
Cre07.g352850  −0.4−2.1
Cre03.g164000TEF7 −3.2−6.0
Nutrient transport-related gene models
Cre02.g111050AMT1;3Ammonium transporter 1;3−0.4−3.2
Cre06.g284150AMT1;3Ammonium transporter 1;3n.d.−1.8
Cre26.g773300PHT2;1Phosphate transporter 2;1−1.8−3.5
Cre02.g144750PHT2;1Phosphate transporter 2;1−2.9−3.3
Cre26.g773350PHT2;1Phosphate transporter 2;1−4.4−2.6
Cre16.g686750PHT6Phosphate transporter 1;6−4.9−4.8
Cre16.g655200PHT2;1Phosphate transporter 2;1−1.0−2.1
Cre04.g221700 Cytochrome c oxidase, subunit III0.64.2
Cre03.g154350 Cytochrome oxidase 22.02.9
Cre16.g651050 Cytochrome c1.96.2
Cre01.g049500 Cytochrome oxidase 21.82.2
Cre06.g304350COX6BCytochrome C oxidase 6B1.13.6
Cre05.g232850COX17Cytochrome c oxidase 17n.d.2.0
Cre12.g516350COX10Cytochrome c oxidase 10n.d1.8
Cre01.g051900 Ubiquinol-cytochrome C reductase iron-sulphur subunit0.91.3
Cre01.g055550 Cytochrome c oxidase assembly proteinn.d.2.8
Figure 2.

Overview of differential expressed gene models in Chlamydomonas reinhardtii during hydrogen production determined with microarray (solid line; total: 1238) and RNA-seq (dotted line: total: 1598). Red dots represent up-regulated genes and green dots down-regulated genes.

In a second step, we used the microarray data for comparison with RNA-seq-determined transcript abundance levels. RNA-seq analysis demonstrated at least 52 million reads per sample with around 18 million unique reads. The Chlamydomonas genome (phytozome 4.3) was applied for mapping, and we calculated the gene expression for ~13 000 gene models (based on an estimation of total ~18 000 gene models), which equates coverage of ~70% with a correlation factor of 0.8–0.88 between all samples (except the control). Estimations of log2-fold changes out of the RPKM data were performed. For our analysis, we included only differentially expressed genes with an average twofold down-regulation (730) or up-regulation (868) compared to the control. Intriguingly for 260 gene models we identified reads only in the control sample, 15 of these gene models showed high transcript level just detectable under normal growth conditions, for example, PsaO (Cre07.g334550), Cah1 (Cre04.g223100), one triose phosphate transporter (Tpt2; Cre06.g263850), one porphobilinogen deaminase (Cre16.g663900), one RNA polymerase (Cre02.g086750) and several unknown genes (see Supporting information). In contrast, 280 newly induced genes during hydrogen production were counted, including many sulphur-related genes, stress-related genes, cytochromes and many unknown genes.

Comparison between both transcriptomic data sets was performed to determine whether the two independent techniques provided similar gene expression data and whether both methods display the same transcript expression pattern over the period of time. In Figure 3, a direct comparison for all differentially expressed genes (with log2-fold changes) derived from microarray analysis and from RNA-seq was plotted. We found overall the same tendency in transcript abundance for 1290 gene models and for 786, an identical gene regulation above the threshold level. There is a good correlation between both data sets, however, with differences in the dynamic range of changes in expression level (postulated previously by Gonzalez-Ballester et al., 2010). This enabled us to gain more information from RNA-seq data regarding time-dependent gene expression during the experiments.

Figure 3.

Comparison of differential genes expression in terms of transcript fold changes between microarray and RNA-seq data derived from hydrogen-producing Chlamydomonas reinhardtii cells. Plotted are mean values (log2 ratios) over all time points.

As mentioned above, ~63% (786 of 1238) of twofold differentially expressed genes derived from microarray analysis could be confirmed by identical expression profiles when RNA-seq was applied (see Figure 2). From these 786 gene models, 395 genes were down-regulated whereas 391 genes showed an up-regulation in both methods. Overall, ~45% of the differential expressed genes (320 of 786 genes) are of unknown function. The remaining gene models (55%) are related to lipid and starch metabolism (27 genes), sulphur metabolism (ten genes), photosynthesis (14 genes) and general stress response (five genes). In addition, numerous genes associated with flagella assembly and cell cycles were identified as being mainly down-regulated.

Time-resolved gene expression analysis during hydrogen production

To improve the time-resolved analysis of our transcript data, we used expression levels of all transcripts and clustered all genes according to their expression pattern. Clustering of differential expressed genes resulted in formation of distinct groups. In Figures 4 and 5, the clusters for both methods are presented. Microarray data suggested that the majority of genes are highly induced or repressed already 24 h after sulphur deprivation with no further changes during the rest of the experiment. In contrast, RNA-seq data seemed to be more sensitive, as we could distinguish between genes induced after 48 h, but also a considerable amount of genes where the expression profile changed in the later phase of the experiment. Of particular note was that we succeeded to identify a group of genes which showed identical transcript expression profile in both methods.

Figure 4.

Time-resolved cluster analysis of microarray-based transcript expression level (log2 ratio) for Chlamydomonas reinhardtii during hydrogen production (clusters were generated with GENESIS software, Pearson un-centred distance). Red represents up-regulated genes and green down-regulated genes. Gene models, log2 ratios and corresponding clusters are summarized in the Supporting information.

Figure 5.

Time-resolved cluster analysis of RNA-seq-based transcript expression level (log2 ratio) for Chlamydomonas reinhardtii during hydrogen production (cluster were generated with GENESIS software, Pearson's un-centred distance). Red represents up-regulated genes and green down-regulated genes. Gene models, log2 ratios and corresponding clusters are summarized in Supporting information.

Strongly up-regulated genes were found in RNA-seq clusters A-D (446 genes), and 629 genes from microarray data are shown in clusters C and D. In such clusters, early and consistent up-regulated genes were found including sulphur-related genes (21 gene models, e.g., Sua1, Slt1, Sir1, Ecp61 (Cre09.g409300), Ecp76 and Ecp88) and stress response genes (5 gene models), reductases (seven), genes encoding peptidases (five) and several cytochromes (five). With Lhcbm9 and Lhcsr1 and one starch phosphorylase enzyme (Cre07.g336950), three photosynthesis-related genes were also detected. An increased gene expression was determined for one chlorophyllase and some other chlorophyll-degrading genes.

In clusters H and I (RNA-seq) and clusters F and G (microarray), we identified genes that were strongly down-regulated including the majority of photosynthetic genes, for example, both small RUBISCO subunits (Rbcs1 and 2), PsbO, PsbQ, PsaG, PsaH, PsaK, PsaE, PsaL, Lhca1, Lhca5, Lhca7, Lhcbm1, Lhcbm5 and Lhcbm7. Other photosynthetic genes also demonstrated a decreased gene expression, like plastocyanin (PetE, Pcy1; Cre03.g182551) and several components of the cytochrome b6f complex (PetC, PetF, PetM, PetN) also showed a dramatic decreased gene expression. Just a few transcripts could be aligned for each subunit of the complex.

Specific early up-regulated genes were found by RNA-seq (clusters A, B and C); interestingly, this group includes enzymes like the mitochondrial pyruvate dehydrogenase kinase (Pdk3; Cre05.g241750), inhibitor of the citrate cycle starting enzyme), while the pyruvate dehydrogenase was down-regulated much later.

Differential gene expression of sulphur deprivation-related gene models

It is known that C. reinhardtii cells respond first to the lack of sulphur by increased sulphur assimilation (see Table 1 for related genes) and redistribution of internal sulphur. In our experiments, genes encoding the acetylglutamate kinase (Agk1; Cre01.g015000) and aspartate aminotransferase (Ast5) were induced, both initial key enzymes for recycling of amino acids. D-cysteine desulphhydrase (Acd1), an enzyme that produces sulphides from cysteine, was also highly abundant, thus providing a cell internal source of sulphur. Another enzyme responsible for redistributing intracellular sulphur by using methionine sulphoxide as substrate is the methionine sulphoxide reductase (Msrb2). The corresponding transcript was also highly up-regulated in the late phase of the experiment.

From these data, we could confirm an earlier observed shift within the cellular amino acid composition, with a decreased amount of cysteine and increased alanine concentration during hydrogen production (Doebbe et al., 2010). The amino acids composition can be also affected by N-acetyl-l-glutamate synthase activity (e.g. Nags2; Cre02.g144800), which is known to promote stress tolerance (Kalamaki et al., 2009). In our experiments, this gene was shown to be down-regulated, which could have increased sensitivity to environmental changes. Consequently, this gene would be a suitable target for genetic engineering in C. reinhardtii to construct more robust phototrophic strains.

Gene expression of lipid- and carbohydrate-related gene models

Another response to stress induced by sulphur deprivation and anaerobiosis is the accumulation of storage compounds like starch and lipid (Doebbe et al., 2010; Matthew et al., 2009). Lipid-related genes were up-regulated in regard to biosynthesis; for example, acetyl-coA synthetase (Acs; Cre23.g765700) and several unspecific acyltransferases (Cre05.g248150 and Cre09.g392300) demonstrated an increased transcript abundance (Table 1). We found a strong increase in transcripts involved in trigalactosyldiacylglycerol transport (Tgd1 and Tgd2) expression level, both enzymes involved in lipid metabolism, with a proposed transport function (Li et al., 2012; Lu et al., 2007; Roston et al., 2012; Wang et al., 2012; Xu et al., 2010). Our results further confirmed previous data (Doebbe et al., 2010) that showed that storage of lipids and starch occurs during the shift from aerobiosis to anaerobiosis. We could not confirm the up-regulation of nitrogen-induced lipid-related proteins determined by Boyle et al. (2012), such as Pdat1 (Cre02.g106400), Dgat1 (Cre01.g045900) and Dgtt1 (Cre12.g557750). However, we found that Dgtt2 (Cre02.g121200) is induced during the time course of our experiment; therefore, a nutrient-specific induction of Dgtt genes can be assumed. Genes involved in biosynthesis of fatty acids were not identified as being differentially regulated during hydrogen production (e.g. Kas1; Cre22.g765250, Kas2; Cre07.g335300). However, fatty acid desaturases (Cre17.g711150, Cre01.g037700, Cre16.g672900, Cre13.g590500), creating double carbon bonds, were strongly up-regulated during hydrogen production phase. Accumulation of unsaturated fatty acids was found in Chlamydomonas under nitrogen starvation and sulphur starvation (La Russa et al., 2012; Msanne et al., 2012). Further genes related to lipid metabolism were up-regulated in the late phase of the experiment, for example, lipases (Cre10.g422850, Cre07.g322900).

In regard to starch metabolism, we detected two starch synthase genes with increased expression level (Cre03.g185250, Cre17.g721500). Genes responsible for starch degradation like isoamylase 3 (Cre19.g755050), two alpha amylases (Cre08.g385500; Cre08.g362450), two phosphorylases (Cre07.g336950, Cre12.g552200) and starch debranching (Cre11.g476650) also showed an increased expression level during hydrogen production. Additionally, our data demonstrate the induction of glycolysis-related genes during hydrogen-producing conditions. Almost all enzymes of this pathway could be identified (see Table 1 and Figure 6). Most notably, the maximum in average gene expression was observed during peak hydrogen production (Figure 6).

Figure 6.

RKPM values calculated from RNA-seq data for all glycolysis-related genes during hydrogen production in Chlamydomonas reinhardtii. The dotted grey line determines the tendency of gene expression of all transcript models over time.

One of the major issues for the survival of the cell under (anaerobic) hydrogen-producing conditions is the maintenance of a balanced ATP/NADP ratio. Genes such as NAD+ kinases (Nadk2) (Takahara et al., 2010) were observed to be expressed particularly under these conditions, which are capable of increasing NADP supply. The imbalanced ATP ratio could be reduced by activity of apyrases (e.g. Cre06.g273500) that were also induced in our experiments. A major issue for the cell is the recycling of the NADPH, a challenge targeted by the pentose phosphate pathway as well as by the hydrogenase activity and the fermentation processes. All genes of the pentose phosphate pathway and a few fermentative-related genes like alcohol dehydrogenase (Adh1; Cre20.g758200), pyruvate formate lyase (Pfl1; Cre01.g044800, just RNA-seq) and acetate kinase (Ack1; Cre09.g396700) showed increased gene expression (see Mus et al. (2007)). These enzymes are the main source for reduction equivalents for the cell under anaerobic stress conditions. Other enzymes that could also provide reduction equivalents are malic enzymes (e.g. Cre06.g268750, Cre14.g629750, Cre14.g629700, Cre06.g251400) and isocitrate dehydrogenases (e.g. Idh3; Cre04.g214500), with all of these gene models showing an increased gene expression during the course of hydrogen production.

Differential gene expression of stress-related gene models

One of the major findings in our data was the strong up-regulation of genes responding to oxidative stress and detoxification (Table 1). Up-regulation of reactive oxygen species (ROS) scavengers such as L-ascorbate peroxidases (Apx3 and Apx5) and the thylacoidal peroxidase (SapX) was detected with RNA-seq and microarrays (Table 1). Such enzymes could be an indication of an induced glutathione–ascorbate cycle, and several studies investigated this cycle (Dietz, 2003, 2010; Maruta et al., 2010; Nagy et al., 2012; Sano et al., 2001; Shigeoka et al., 2002; Steenvoorden and van Henegouwen, 1997; Takeda et al., 1997, 2000; Urzica et al., 2012). However, the mono-dehydroascorbate reductases were not up-regulated (Mdar1 and Mdar4). Furthermore, the last enzyme in the cycle, the glutathione reductase, was not differentially expressed in our experiments, in contrast to several dehydroascorbate reductases (Dhar2 and Dhar3); glutathione peroxidases and transferases are also known to be involved in the detoxification process of ROS. ROS stress response regulation is predicted for Sor1 (Cre07.g321550), a DNA-binding protein, which is up-regulated in our experiment and found by Fischer et al. (2012). Detoxification of ROS is predicted for flavodoxin quinone reductases (Fqr1-3), which were also up-regulated during this experiment. Additionally, we found numerous genes up-regulated with a potential NADP or FAD oxidoreductase activity (e.g. Cre07.g346800, Cre16.g671450, Cre02.g139200), also known to be potential ROS detoxicants. The oxidative stress is also targeted by the terminal oxygenases (Pto1 and Pto2), which are involved in carotenoid synthesis (Carol and Kuntz, 2001). Additionally, the alternative oxidase Aox1, which is known to be induced under ROS stress in higher plants and involved in nitrogen metabolism in Chlamydomonas (Baurain et al., 2003), was observed to be up-regulated. The function for this enzyme during hydrogen production is not known, however, is most likely stress related. In the high hydrogen production mutant Stm6 (Kruse et al., 2005), which is highly light sensitive and also partly defective in efficient NPQ (Nguyen et al., 2011), a sharp increase in Aox expression levels was also observed. Increased transcript levels of other genes involved in detoxification such as catalase 2 (Cre09.g417150) and two superoxide dismutases (Cre10.g436050, Cre02.g096150) genes were also identified.

Differential gene expression of unknown gene models during peak hydrogen production

Of special interest was the group of unknown differentially expressed genes during hydrogen production, which could be successfully determined within both methods (Table 1). Examples for highly up-regulated genes found in both data sets with putative but not fully defined function and with a high potential as targets for detailed investigations are A glutathione transferase (Cre12.g559800), an isoflavone reductase like protein (Cre11.g477200), two oxidoreductases (Cre07.g355500, Cre12.g493500) and several thioredoxins (Cre01.g057750, Cre05.g248500, Cre07.g315100, Cre14.g624150). These enzymes are examples for redox stress-induced genes and therefore potential targets to improve oxidative stress tolerance in C. reinhardtii. Furthermore, a potential lipase (Cre02.g127300) and a fatty acid desaturase (Cre17.g711150) could be identified as being highly up-regulated, which could be of interest in regard to the lipid metabolism in C. reinhardtii.

Gene models with potential transport function showed a strong increase in transcript abundance (see Table 1), for example, Cre16.g656150, Cre10.g445000, Cre01.g061650 and Cre03.g166050. Furthermore, gene models related to phosphate uptake and few carbonic anhydrases [Cah1; Cre09.g405750, Cah7; Cre13.g607350, Cah9; Cre05.g243300 and Cag1; Cre12.g516450 (mitochondrial)] also showed an increased transcript level; however, the function of Cah7 and Cah9 during hydrogen production is unclear, because gene expression could be hardly detected so far (Moroney et al., 2011).

Examples for strong down-regulated genes that could be of interest are mitochondrial carrier proteins (Cre02.g094250, Cre28.g776600), one NAD-dependent glycerol-3-phosphate dehydrogenase (Cre01.g053000) and one acyl carrier protein (Cre14.g621650). Differential gene expression of such enzymes indicates potential problems in metabolite transport and communication between compartments in C. reinhardtii. We found several other mitochondrial genes, like succinate dehydrogenase or NADH dehydrogenase, which were not significantly affected in our experiment, while gene models encoding subunits of the cytochrome oxidase were strongly up-regulated at all time points. Interestingly, Cre09.g388150 encoding a mitochondrial translation factor (Mrpl36) was highly up-regulated at the late time points. Normally, this factor is involved in translation of mitochondrial proteins (Piao et al., 2009; Prestele et al., 2009).

Finally, we compared gene expression for unknown genes under hydrogen production and sulphur stress (Gonzalez-Ballester et al., 2010; Toepel et al., 2011) and determined several genes that could be identified with numerous reads in all data sets (see Table 1; sulphur-related genes).

Differential gene expression of genes coding for regulatory elements

A time-resolved analysis of RNA-seq data (RPKM values for all differential expressed gene models, plotted in Figure 7) lead to the hypothesis that C. reinhardtii first responds to sulphur stress by induction of transport systems, as described earlier (Gonzalez-Ballester et al., 2011; Toepel et al., 2011). However, before the cells reorganize its metabolism, gene expression of specific regulator genes changes. Our data support this hypothesis by demonstrating an early down-regulation of genes (Figure 7a) that are involved in functional assembly of multiprotein complexes, for example the mRNA maturation factor Nac2 (Mbd2; Cre02.g124700). Nac2 stabilizes the RNA of the photosystem II subunit PsbD and is a key element for the synthesis of photosystem II (Ossenbuhl and Nickelsen, 2000). Other examples for induction of transcription control factors are Tda1, which promotes the AtpA translation (Cre08.g358350, down-regulated), Mca1 promoting PetA maturation (Cre08.g358250, no change in gene expression), Tca1 promoting PetA translation (Cre09.g415500, down-regulated), Mrl1 promoting RbcL translation (Cre06.g298300, down-regulated) and Mbb1 promoting the maturation of PsbB (Cre09.g416200, no change in gene expression). These results clearly demonstrate that control of translation is a first response to stress in C. reinhardtii (Eberhard et al., 2011). Proteins involved in regulation of the RUBISCO (Rmt1) and potential methyltransferases like Cre27.g774300, Cre08.g368700 and Cre12.g503800 are also enzymes that regulate protein activity and showed a differential gene expression before complete pathways are remodelled. Flagellar-related genes and genes promoting cell cycle are down-regulated (see Supporting information), indicating a typical stress response and a switch from homoeostasis and cell maintenance to cell survival.

Figure 7.

Grouping of gene models according to the time point of highest transcript level during hydrogen production in Chlamydomonas reinhardtii. Plotted are the RKPM values (determined with RNA-seq) of all differential expressed gene models at each time point (a: maximum after 0 h, b: maximum after 48 h, c: maximum after 72 h and d: maximum after 96 h); the dotted grey lines represent the tendency of all gene expression of models during the time course.

Induction of cell death-related gene models

The hydrogen production phase is characterized by gene expression for genes responsible for stress resistance and NADPH recycling. The final step is the activation of genes inducing apoptosis and cell death. Late up-regulated genes (Figure 7d) are genes related to a specific group of genes and to the cessation of numerous intracellular processes. The end of hydrogen production in C. reinhardtii was reached after 96-h cultivation under sulphur deprivation. At this time, the entire cell system starts to degrade. Cell apoptosis could be initiated by differential expression of proteins like DAD1 (Defender against death, Cre02.g108400) or APAF1 (apoptotic protease-activating factor 1, found as protein but not annotated in Chlamydomonas) (Moharikar et al., 2007).

Activating proteins involved in protein degradation like ubiquitins (Ubq1, Cre18.g750000, Ubq7; Cre13.g563600), ubiquitin protein ligases and ubiquitin protein proteases (at least five genes are up-regulated late during the experiments, e.g. Cre06.g266350, Cre12.g533750, Cre08.g364550 (Figure 7d)); furthermore, autophagy-inducing genes were also up-regulated like Apg8 (Cre16.g689650) (Perez-Perez et al., 2010). Finally, late induced proteases (Cre16.g663350, Clpp2; Cre12.g521450, Cre06.g274700, Deg11; Cre12.g498500, Deg1; Cre02.g088400, Cep1; Cre09.g407700, Cep2; Cre05.g247800) lead to cell death (see Figure 8). The up-regulation of such enzymes as final step under sulphur starvation was also described earlier (Gonzalez-Ballester et al., 2010).

Figure 8.

RKPM values calculated from RNA-seq data of several ubiquitins and proteases during hydrogen production in Chlamydomonas reinhardtii. The dotted grey line determines the tendency of gene expression of all transcript models over time.


Hydrogen production in C. reinhardtii has been a major target of research, and during the last years, intensive systematic analysis of the hydrogen production metabolism resulted in a remarkable increase in knowledge regarding the process in general (Eroglu and Melis, 2011; Kruse et al., 2005; Nguyen et al., 2011). As a consequence, the understanding how cells remodel their metabolism and their ability to compensate mutations of specific genes (knock out, RNAi, etc.) dramatically increased. Strong examples are the role of fermentative-related genes (Catalanotti et al., 2012; Grossman et al., 2011; Magneschi et al., 2012) and neutral lipids during sulphur deprivation (La Russa et al., 2012). To date, hydrogen production rates are still below viable relevance to the alternative fuel production and must be improved. The H2 production phase is still too short, and prolongation of the process is required to increase system yields. A technical solution in this regard was recently provided (Lehr et al., 2012) by inducing hydrogen production phase by controlled and limited microsupply of sulphur. The data presented here demonstrate the possibility to use transcriptome analysis techniques to determine gene targets whose expression levels are crucial for efficient hydrogen production. Our results clearly show that both methods, RNA-seq and microarray, were able to provide complimentary data to further elucidate regulation of genes in C. reinhardtii during hydrogen production and completed our picture regarding differential gene expression for adaptation to sulphur stress, anaerobiosis and induction of semi-constant hydrogen production. The sulphur deprivation–related results are in very good accordance with previously published data (Gonzalez-Ballester et al., 2010; Toepel et al., 2011). Indeed, the expression of sulphur-related transport systems, redistribution of internal sulphur sources and gene expression responsible for accumulation of storage, for example, lipids and starch compounds, were verified with our experiments. We could additionally identify gene expression for distinct genes in regard to oxidative stress, photosynthesis and general metabolism. Furthermore, the larger dynamic range of the RNA-seq technique, as earlier already postulated by Gonzalez-Ballester et al. (2010), was confirmed here. It is noteworthy to mention that RNA-seq enabled for an improved time course-dependent analysis of gene expression, while the microarray data could only be used to identify relative transcript abundance variation that remained for most of the genes during course of the experiment.

The results furthermore clearly demonstrate that hydrogen production forces C. reinhardtii cells to deal with two major stressors, ROS and maintenance of intracellular energy balances. This is reflected by expression of genes responsible for transport systems and for remodelling metabolic pathways necessary to access internal sulphur sources. ROS accelerate the process of cell destruction and must be efficiently removed from the system, a strategy that is probably targeted by over-expression of genes coding protection-related proteins. The consequence of a better control of cellular answers to ROS is the maintenance of a minimum of photosystem II activity and photosynthetic electron transport, so that electron supply to the hydrogenase can be prolonged. Consequently, genes for targeted genetic manipulation are members of the glutathione–ascorbate cycle (such as ascorbate peroxidases) and genes coding for superoxide dismutases were shown to be up-regulated in the late phase of hydrogen production (Figure 8).

A prolonged hydrogen production as a result of partial PS II protection, for example, by efficient xanthophyll cycle, was already described by others (Scoma et al., 2012; Torzillo et al., 2009). Furthermore, the strong down-regulation of the whole cytochrome b6f complex implicates a potential bottleneck in electron transport in this photosynthetic multiprotein subunit. A stabilization of this complex could therefore also be a target of future studies. In this regard, an increased consumption of reduction equivalents by such enzymes could be problematic; however, a stable electron transport rate is crucial for biotechnological applications targeting to improve hydrogen production in C. reinhardtii. One of the main goals in future projects should be the generation of C. reinhardtii strains with stable Cyt-b6f- and PS I systems under stress conditions, even with low electron transport capacities. Philipps et al. (2012) demonstrated in this regard the effect of limited electron flow towards the hydrogenase in nitrogen-deprived C. reinhardtii cultures, with the consequence of low hydrogen production rates.

In addition, cell internal energy distribution has to be directed towards the hydrogenase, which can be realized by promoting efficient storage breakdown, minimizing competition reactions and providing reduction equivalents to the enzyme. Providing hydrogenase with electrons and protons is essential for productivity; therefore, NADPH and/or ferredoxin recycling is crucial (Winkler et al., 2010). Our data indicate that the cells most likely try to balance the ATP/NADPH ratio by induction of several alternative pathways and enzymes. However, a reduced activity of transport mechanisms and a disturbed interorganelle communication between chloroplast and mitochondria most likely reduce the functionality of the cellular system during the hydrogen production process. In addition, an increased competition between fermentation processes and hydrogenases for electrons reduces the capacity for hydrogen production.

Lipid degradation is strongly reduced under anaerobic conditions; therefore, accumulation of neutral lipids has to be avoided to increase H2 production rates as the lipid storage sink will be hardly available as a substrate for hydrogenase (see also (Doebbe et al., 2010; Miller et al., 2010; Philipps et al., 2012)). It has been also previously shown how accumulation of starch and lipids influences hydrogen production capacities in C. reinhardtii under different nutrient starvation conditions (Philipps et al., 2012). Within this work, several genes were identified to be involved in accumulation/degradation of polar lipids, being also a potential target for manipulation, with the goal to reduce energy storage as lipids. Difficulties of direct manipulation of potential key enzymes of the lipid metabolism were however demonstrated previously (La Russa et al., 2012), thus demonstrating metabolomic complexity and remodelling capacity of a cellular metabolism; therefore, construction of a complete metabolomic network is essential for a precise prediction of gene manipulation effects. One example for metabolomic complexity is the group of malic enzymes, proteins which are not directly related to lipid chemistry, but are known to affect lipid accumulation. Over-expression of malic enzymes increased lipid concentration in fungi; however, at the same time, these enzymes produce reduction equivalents, useful for hydrogen production in C. reinhardtii (Zhang et al., 2007). Malic enzymes, the pentose phosphate pathway and isocitrate dehydrogenases are main sources for metabolomic reduction equivalents during hydrogen production, and stabilization of function for these enzymes could also extend hydrogen production. However, prior to such engineering approaches, precise full energy balance analyses of potential future biorefineries are needed, because a high lipid-containing biomass (C. reinhardtii after hydrogen production) could also be useful in follow-up processes in which the remaining biomass is used to produce biomethane via fermentation (Mussgnug et al., 2010). The rate of gene expression of distinct lipases, as shown here (Table 1), is not sufficient for lipid degradation during hydrogen production, because the lack of oxygen prevents beta oxidation. Crucial for storage compound breakdown is also the activity of the respiratory chain in the mitochondria. Our data indicate that the transport of metabolites between the compartments is strongly affected during the time course of the experiments. The importance of a functioning proton pumping in C. reinhardtii during hydrogen production has been already described (Lecler et al., 2011). Therefore, it is feasible to suggest that stabilization of metabolite transport and enhanced degradation of starch and a reduced lipid accumulation should increase the hydrogen production phase significantly.

In addition, this study identifies a number of unknown genes that are highly up- or down-regulated during hydrogen production. The function of most of these genes can only be assumed, and characterization needs to be carried out in future studies to reveal the possible function during hydrogen production.

Our data furthermore demonstrated that mutation of lhcsr3 in Npq4 did only slightly improve hydrogen production in Chlamydomonas most likely due to the fact that stress-induced damage of photosynthesis during of sulphur deprivation phase was not increased compared to the corresponding parental strain. However, as it is known that NPQ is an important factor in regard to hydrogen production (Kruse et al., 2005; Nguyen et al., 2011), further detailed analysis of the different LHCSR isoforms and especially LHCSR1 (gene expression in npq4 was identical to wild type) is necessary to unravel the function of these enzymes during hydrogen production.

The comparison of results derived from both methods, microarray and RNA-seq, showed a good accordance. The higher dynamic range and capacity to detect unknown genes however clearly promotes the usage of transcript sequencing. Nevertheless, there are some critical factors in regard of RNA-seq data: The calculation of fold changes of stress-induced genes, where no reads could be determined in the control sample, is problematic. Furthermore, one has to determine whether the gene of interest was not expressed in the control sample or whether the sequences were not amplified. Furthermore, RNA-seq still could underestimate low abundant genes, for example, fermentative pathways genes. And finally, alignment and differentiation of reads for transcripts with high similarity, for example, LHCBM proteins or histones, is difficult and has to be analysed for each gene in detail by RT-Q-PCR. For the LHCBM proteins, we could just use a small amount of sequences specific to each isoform (for detailed explanations see Fang et al. (2012)). On the other hand, microarrays provide an easy-to-use platform with high reproducibility and analysis can be performed in a standardized way. Therefore, a combined application of both methods could be considered as a powerful strategy to achieve profound and deep insights into distinct cellular metabolic decisions.

Experimental procedure

Cultures, growth and H2 production conditions

As reference strains, we used the wild-type 4A+ and the non-photochemical quenching mutant npq4, which was generated by insertional mutagenesis, resulting in a knockout of the LHCSR3.1 and LHCSR3.2 genes (Niyogi et al., 1997; Peers et al., 2009). Cultures were grown mixotrophically in standard TAP media (Harris, 1989) until early stationary phase in constant light (50 μE/m2/s) at 25–30 °C, harvested via centrifugation and resuspended after washing (three times) in TAP minus S media (Melis and Happe, 2001). For H2 production measurements, cells were cultivated for 96 h in constant light (200 μE/m2/s1) in sealed 300 mL bioreactors. Hydrogen production was measured as volume determination of gas production, and gas quality was measured by gas chromatography as earlier described (Doebbe et al., 2010). Photosynthetic quantum yields were monitored with a MINI-PAM (Walz, Germany) (Nguyen et al., 2011).

Sample collection, RNA preparation and transcriptomics

Samples of C. reinhardtii from bioreactor-cultivated cultures (300 mL) were collected at 24, 48, 72 and 96 h after sulphur deprivation (T1, T2, T3 and T4). We excluded the time point T1 (24 h) from analysis and comparison, because a high amplification of GC-rich reads in RNA-seq was detected and adapted the labelling: 48 h (T1), 72 h (T2) and 96 h (T3) for both methods. Reference samples (T0) were harvested from cultures of the corresponding strain prior to sulphur deprivation. Samples were immediately centrifuged at 8300 g for 2 min at room temperature, cell pellets were immediately lysed with RNA lysis buffer and RNA was isolated as previously described (Nguyen et al., 2011). RNA samples were pooled from three independent experiments and used for microarray (three technical replicates) and RNA-seq analyses. RNA quality was determined with an Agilent© Bioanalyzer (Agilent Inc., Santa Clara, CA). Chlamydomonas reinhardtii microarray slides (Agilent© 4 × 44 k, no: 024664) were used for the transcript analyses (see Toepel et al. (2011) for details). RNA labelling (Quick RNA amplification and labelling kit; Agilent©) and microarray hybridization (16 h at 60 °C) were carried out according to the supplied manual. Microarrays were washed after hybridization according to the Agilent© manual, dried in a centrifuge and scanned with a 5-μm resolution in an Agilent© DNA microarray scanner. Data extraction was achieved using the feature extraction software (; Agilent©), and data were normalized and analysed using the software EMMA2 (see Toepel et al. (2011)). We used a robust normalization method (lowess), performed significance tests within all experiments and considered only those probes showing a significant change in their expression (P-values smaller than 0.05). To further limit our result set, we included only genes that demonstrated at least a twofold up- or down-regulation. RNA-seq was performed as described by (Illumina Inc., San Diego, CA). The cDNA libraries were assembled according to Illumina's RNA-seq protocol, loaded and sequenced as 36-mers as single reads. Raw sequence files were aligned against Chlamydomonas genome (phytozome 4.3) with a new developed program, SARUMAN (Blom et al., 2011), which allows rapid and precise alignment. We allowed a 6% error in the alignments and analysed the best position for each read. Without losing all other positions, only unique reads per gene model were used for the determination of the expression of a gene. Expression estimates were obtained for each individual run in units of RPKMs (reads per kilobase of mappable transcript length per million mapped reads) after normalization by the number of aligned reads and map-able transcripts (Boyle et al., 2012). Based on the normalized RPKM values, we estimated log2 ratios for the gene models. To calculate newly induced genes, characterized by zero reads in the control samples, we used average minimal expression level as control value. All transcriptomic data were clustered and visualized with the GENESIS software (Sturn et al., 2002) using Pearson's un-centred distance. Significance test was performed with EdgeR software (http://bioconductor.org/packages/2.10/bioc/html/edgeR.html).

Validation of RNA-seq and microarray data was performed for reference genes with quantitative RT-PCR as described (Nguyen et al., 2011).

Microarray data can be accessed at NCBI (Geo): GSE41728.

RNA-seq data can be accessed at ArrayExpress: E-MTAB-1329.


The authors thank the Federal Ministry of Science [BMBF ForSys Partner grant contract 0315265A] and the EU/Energy FP7 project SOLAR-H2 [contract 212508], for financial support and Prof. Krishna Niyogi from the University of California for providing the npq4 and 4A+ mutants. Patrick May was funded by the German Federal Ministry of Education and Research, Systems Biology Research Initiative “GoFORSYS”. We thank Christian Rückert (Bielefeld University) for the help with the RNA-seq experiments.