Many cyanobacteria are capable of utilizing light energy for nitrogen fixation. As a by-product of this nitrogenase mediated catalysis, hydrogen gas is produced. Several approaches to increase hydrogen production from cyanobacteria exist. Usually, these approaches are non-targeted. Here we exemplify how DNA-microarray based gene-expression analysis and bioinformatic visualization techniques can be used to analyze nitrogen and hydrogen metabolism from the filamentous, heterocyst forming cyanobacterium Nostoc PCC 7120. We analyzed the expression of 1249 genes from major metabolic categories under nitrogen fixing and non-nitrogen fixing growth. Of the selected genes, 494 show a more than 2-fold expression difference in the two conditions analyzed. Under nitrogen-fixing conditions 465 genes, mainly involved in energy metabolism, photosynthesis, respiration and nitrogen-fixation, were found to be stronger expressed, whereas only 29 genes showed a stronger expression under non-nitrogen fixing conditions. To help understanding probe hybridization, all expression data were correlated with potential target secondary structures and probe GC-content. For the first time the expression of high light-induced stress proteins (HLIP-family) is shown to be linked to the nitrogen availability.

Cyanobacteria play an increasing in biotechnology (Singh et al. 2005). One potential but not yet feasible application is the photobiological production of hydrogen gas (Wünschiers and Lindblad, 2003; Dutta et al. 2005; Prince and Kheshgi, 2005). Cyanobacterial hydrogen metabolism is closely linked to dinitrogen assimilation in nitrogen-fixation. All known nitrogenases evolve 1 to 7.5 mole hydrogen gas as they convert one mole dinitrogen to two mole ammonia. This apparent loss of energy is circumvented by uptake hydrogenases that allow the cyanobacteria to regain energy. Mutants which lack uptake hydrogenases and thus produce substantial amounts of hydrogen gas have successfully been engineered (Happe et al. 2000; Masukawa et al. 2002). The genetic and metabolic regulatory networking behind photobiological hydrogen production is only understood in fragments.

Heterocystous nitrogen-fixing cyanobacteria like Nostoc sp. strain PCC 7120 (Nostoc PCC 7120; formerly Anabaena PCC 7120) respond to the deprivation of combined nitrogen with morphological changes, i.e., the formation of heterocysts. These specialized non-dividing cells develop more or less equidistantly along the filaments with a ratio of about one heterocyst per 10 vegetative cells. In comparison to vegetative cells, the heterocyst is larger and more rounded. It provides an environment with low oxygen partial pressure since it lacks oxygen-evolving photosystem II activity and has a higher respiration rate (Böhme, 1998). Furthermore, it is surrounded by a thick glycolipidic cell wall that reduces the diffusion of oxygen. Heterocysts provide amino acids and receive carbohydrates from their neighboring vegetative cells. The differentiation process begins within a few hrs after combined-nitrogen deprivation, i.e. growth on dinitrogen as sole nitrogen source, and requires approximately 24 hrs to complete. Single components involved in this process have been described (Golden and Yoon, 1998; Adams, 2000; Wolk, 2000; Meeks and Elhai, 2002; Herrero et al. 2004).

A key protein in heterocyst formation is NtcA (Herrero et al. 2001; Herrero et al. 2004), which is the global nitrogen regulator that controls, e.g., the expression of genes essential for heterocyst development such as the ABC-type transporter genes devABC and the regulator genes hetR and patS. Among many other factors that have been demonstrated to be important for heterocyst development are enzymes that are involved in the formation of heterocyst-specific glycolipids and polysaccharides. The driving forces for these structural changes occurring during heterocyst formation are metabolic requirements: the dinitrogen fixing multi-enzyme complex nitrogenase requires a low oxygen partial pressure and a high supply of ATP and reduction equivalents (Dixon and Kahn, 2004). ATP may be generated by either cyclic photophosphorylation or oxidative phosphorylation while low-potential electrons may be generated from the degradation of carbohydrates produced during photosynthesis (Haselkorn and Buikema, 1992). Obviously, profound regulatory events coincide with growth on dinitrogen.

A powerful tool to study gene expression and its regulation is the DNA-microarray technique. Previously, we analyzed the expression of individual genes and operons of Nostoc spp. that are involved in nitrogen metabolism (Wünschiers and Lindblad, 2003). Here we describe an oligonucleotide based DNA-microarray expression analysis, where each gene is covered by up to 10 unique probes. Until now, only few oligonucleotide DNA-microarray based gene-expression analyses with heterocystous cyanobacteria have been reported (Ehira et al. 2005; Imashimizu et al. 2005; Ehira and Ohmori, 2006; Higo et al. 2006; Lechno-Yossef et al. 2006). Only two among these focus on the effect of nitrogen sources on gene expression. While Ehira and Ohmori (2006) follow the time course of nitrogen deprivation, ending at 24 hrs after removal of anorganic nitrogen from growth medium, Lechno-Yossef et al. (2006) looked only at the expression differences after 14 hrs of such a shift. In our experiment we look at the effect at continuous growth, i.e., fully shifted and equilibrated cultures.

Therefore we employed a novel, recently developed microarray technique where probe synthesis, hybridization, and signal detection take place in one device at strongly controlled physical conditions (Baum et al. 2003; Güimil et al. 2003). The expression data were used to (a) validate the technique employed and (b) obtain a global overview about the effect of growth on dinitrogen on Nostoc PCC 7120. For the latter, we set up a convenient data-processing pipeline based on a MySQL database and a web-based graphical user interface. This front-end allows users to visualize gene-expression data on KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway maps.

Strains and culture conditions

The cyanobacterium Nostoc sp. strain PCC 7120 (formerly Anabaena sp. strain PCC 7120) was grown on either dinitrogen (nitrogen fixing) or combined nitrogen (non-nitrogen fixing) in batch cultures. Non-nitrogen fixing conditions were obtained by growing cells in BG11₀ (Stanier et al. 1971) supplemented with 5 mM NH₄Cl and 10 mM HEPES (pH 7.5). Nitrogen-fixing conditions were obtained by growing cells in BG11₀. All cultures were grown in continuous white light (Thorn Polylux 4000 and Osram Warmton Warm Light 400 to 700 nm; 40 µEm^-2s^-1) at 26ºC with a magnetic stirrer being used to retain a homogeneous suspension. Cultures were harvested at mid-logarithmic growth phase.

RNA isolation

Total RNA was extracted from 100 ml cultures at A_730nm = 0.5. The cells were harvested by centrifugation at 6,000 x g for 10 min together with 50 ml crushed ice in a 250 ml centrifuge bottle. The cell pellet was frozen in liquid nitrogen and thawed on ice. The cells were suspended in resuspension buffer (0.3 M sucrose, 10 mM sodium acetate, pH 4.5), transferred to an Eppendorf tube and pelleted at 12,000 x g for 5 min. The pellet was suspended in 250 µl resuspension buffer with 75 µl 250 mM Na₂-EDTA, and the suspension was incubated on ice for 5 min. 375 µl lysis buffer (2% (w/v) SDS, 10 mM Na-acetate, pH 4.5) was added, followed by incubation at 65ºC for 3 min. 700 μl of 65ºC phenol was added to the lysed cells, followed by incubation at 65ºC for 3 min and then at -70ºC for 15 sec. The suspension was centrifuged at 12,000 x g for 5 min, the upper phase was collected and the hot phenol treatment was repeated twice, followed by an extraction with hot phenol:chloroform (1:1). 1/5 volume of 10 M LiCl and 2.5 volumes of 99.5% ethanol was added and the RNA was precipitated at -20ºC for 30 min. The pellet was washed with 80% ethanol, suspended in water and stored in aliquots at -70ºC. Quality and quantity of total RNA was analyzed with an Agilent 2100 Bioanalyser using the RNA 6000 Nano LabChip kit (Agilent Technologies, Boeblingen, Germany). If necessary, the RNA preparation was treated with 40 units of RNase-free DNase I (Amersham-Pharmacia Biotech, Uppsala, Sweden) according to the manufacturer's instructions. The average RNA yield per 100 ml culture at A_730nm = 0.5 was 50 µg.

Preparation of biotin labeled, fragmented cRNA

From 10 µg of total RNA, low molecular weight RNA, e.g., tRNA and 5S rRNA, were removed by size exclusion chromatography (MEGAclear kit, Ambion). To remove 16S and 23S rRNA, the MICROBExpress kit from Ambion was used. To increase the abundance of low expressed transcripts, the remaining RNA was linearly amplified by a modified Eberwine protocol (Eberwine et al. 1992) as follows. If not differently stated, all enzymes and chemicals were purchased from Invitrogene.

First strand synthesis. The pelleted RNA from the previous mRNA-enrichment steps was resuspended in 4.25 µl water and mixed with 1 µl of T7 random hexamers (0.5 µg/µl; 5'-GGC CAG TGA ATT GTA ATA CGA CTC ACT ATA GGG AGG CGG NNN NNN-3'). Following incubation at 70ºC for 10 min, 4ºC for 2 min and 23ºC for 5 min, 3.75 µl reaction mix (2 µl 5x first strand synthesis buffer (1 µl 0.1 M DTT, 0.5 µl 10 mM dNTP mix, 0.25 µl 40 U RNase OUT) and 200 U Superscript II polymerase) was added to the RNA/primer mix. First strand synthesis reaction was performed with the following temperature scheme: 37ºC for 20 min, 42ºC for 20 min, 50ºC for 15 min, 55ºC for 10 min and 65ºC for 15 min. After adding 0.5 µl RNase H the reaction mix was incubated for another 30 min at 37ºC and 2 min at 95ºC.

Second strand synthesis. The product of the first strand synthesis was mixed with 43.8 µl water and 15 µl 5x second strand synthesis buffer (20 U DNA-polymerase I, 1.5 µl 10 mM dNTP and 1 U RnaseH) and incubated for 2 hrs at 16ºC. After addition of 10 U T4 DNA-polymerase the reaction mix was first incubated at 16ºC for 15 min and then at 70ºC for 10 min.

Isolation of ds-cDNA. Double stranded cDNA was isolated from the product of second strand synthesis according to standard procedures (Maniatis et al. 1982).

In vitro transcription. The pelleted ds-cDNA was resuspended in 1.5 µl water. The MEGAscript T7 kit (Ambion) was used for in vitro transcription. In addition to the standard nucleotides, 3.75 µl 10 mM Bio-16-CTP (NEN) and 3.75 µl 75 mM Bio-11-UTP (Roche) were added to the reaction mix. This led to the formation of biotinylated cRNA.

cRNA-isolation. The RNeasy kit (Qiagen) was applied for cRNA-isolation. All steps were performed according to the manufacturer's instructions.

cRNA-fragmentation. For cRNA-fragmentation 15 µg cRNA was resuspended in 2.5 µl water and 2.5 µl 2 x fragmentation buffer (5 x stock: 200 mM Tris, 150 mM Mg-acetate, 500 mM K-acetate, pH 8.1). The reaction mix was incubated for 5 min at 94ºC. The fragmentation reaction was performed immediately prior to hybridization and checked by alkaline agarose electrophoresis.

Oligonucleotide probe selection

A unique Nostoc PCC 7120 probe set (as many 25-mer probes per open reading frame (ORF) as possible) was calculated based on the full genome sequence (retrieved online from CyanoBase: http://www.kazusa.or.jp/cyanobase/Anabaena/index.html) using a combination of sequence uniqueness criteria and rules for selection of oligonucleotides likely to hybridize with high specificity and sensitivity. The selection criteria were as described in Lockhart et al. (1996) with modifications for the longer probes used here (25-mers instead of 20-mers). If available, 10 unique probes per ORF were used in the experiments.

Light-activated in situ oligonucleotide synthesis was performed as described by Singh-Gasson et al. (1999) using a digital micromirror device, which is part of the geniom one device (febit biotech GmbH, Heidelberg/Germany). The synthesis was performed within the geniom one device on an activated three-dimensional reaction carrier consisting of a glass-silica-glass sandwich (DNA-processor). Four individually accessible microchannels (referred to as arrays), etched into the silica layer of the DNA-processor, were connected to the microfluidic system of the geniom device. Using standard DNA-synthesis reagents and 3'-phosphoramidites with a photolabile protecting group (Hasan et al. 1997; Beier and Hoheisel, 2000), oligonucleotides were synthesized in parallel in all four translucent arrays of one reaction carrier. Prior to synthesis, the glass surface was activated by coating with a silane-bound spacer.

Non-competitive hybridizations were performed with 7.5 µg fragmented cRNA (see above) in a final volume of 10 µl. The hybridization solution contained 100 mM MES (pH 6.6), 0.9 M NaCl, 20 mM EDTA, 0.01% (v/v) Tween 20, 0.1 mg/ml sonicated herring sperm DNA, and 0.5 mg/ml BSA. RNA-samples were heated in the hybridization solution to 95ºC for 3 min followed by 45ºC for 3 min before being placed in an array which had been prehybridized for 15 min with 1% (w/v) BSA in hybridization solution at room temperature. Hybridizations were carried out at 45ºC for 16 hrs. After removing the hybridization solutions, arrays were first washed with non-stringent buffer (0.005% (v/v) Triton X-100 in 6 x SSPE) for 20 min at 25ºC and subsequently with stringent buffer (0.005% (v/v) Triton X-100 in 0.5 x SSPE) for 20 min at 45ºC. After washing, the hybridized RNA was fluorescence-stained by incubating with 10 µg/ml streptavidin-phycoerythrin and 2 µg/µl BSA in 6 x SSPE at 25 for 1ºC 5 min. Unbound streptavidin-phycoerythrin was removed by washing with non-stringent buffer for 20 min at 25ºC. Two biological replicates of each condition have been hybridized.

The CCD-camera based fluorescence detection system, equipped with a Cy3 filter set, integrated into the geniom one automate was used. 36 pixels per spot were available for data analysis. Processing of raw data, including background correction, array to array normalization and determination of gene-expression levels, as well as calculation of expression differences were performed as described before (Zhou and Abagyan, 2002). All steps were carried out using the PROP-algorithm of the geniom application software which is based on the MOID-algorithm described by Zhou and Abagyan (2002). Background correction is based on probes with no corresponding mRNA-target and the average of the lowest 5% expressed genes. Data normalization is based on iteratively correcting the raw data on these genes. Significance levels for differentially expressed ORFs have been calculated using a two-sided t-test.

In order to correlate probe secondary structures and GC-contents with their expression value, the secondary structure of each ORF was calculated using the Vienna RNA Package Version 1.4 (Hofacker et al. 1994). The probes were aligned to their corresponding ORF and the potential probe structure extracted. The number of hybridized (stem duplexes) versus free (loops and bulges) probe nucleotides as well as the probe's GC-content were used in further analysis.

All gene-expression data obtained are saved in the Hydrogenase Database (HyDaBa). This relational database allows cross-linking of the expression data with the annotated genome data from NCBI and Cyanobase and pathway maps available from KEGG. The latter is achieved in real-time via a SOAP-interface. HyDaBa is based on a Apache Webserver, MySQL database and a front-end programmed in PHP. All data are publicly accessible via this web interface that can be accessed using “guest” and “hydaba06” as user name and password, respectively.

In the present study we analyzed the expression of 1249 selected genes from 16 metabolic categories (ca. 20% of the complete genome) of Nostoc PCC 7120 cultures under nitrogen fixing and non-nitrogen fixing conditions (Table 1). Therefore we applied a DNA-microarray based approach (Figure 1).

Preparation of the DNA-processor

Oligonucleotide synthesis, hybridization with target cRNA, and signal detection were performed with one single device, named geniom one (febit biotech GmbH, Heidelberg/Germany, Baum et al. (2003)). 25-mer oligonucleotide probes were synthesized in situ on the DNA-microarray surface. In order to obtain a broad picture of gene-expression differences between nitrogen fixing and non-nitrogen fixing Nostoc PCC 7120 cultures, 500 manually and 749 randomly selected target genes from all major metabolic categories were analyzed (Table 1; Kaneko et al. (2001)). This selection was based on the genome sequence and annotation available from the CyanoBase consortium. In order to ensure reproducibility of the microarray analysis, up to 10 unique 25-mer oligonucleotide probes per target ORF were distributed randomly over the DNA-processor. Due to their small size, 132 ORFs were represented by fewer than 10 unique probes. Of theses, 78 represent unknown and 15 hypothetical proteins, respectively. Of the remaining only 5 ORFs were represented by less than 4 unique probes.

General data analysis

Figure 2 shows a section of four arrays used in this analysis. Due to in situ probe synthesis with a digital micromirror device both the spot morphology and topology are extremely homogeneous. The use of one physical surface for all arrays and the fixed placement of the slide during all processing steps results in very low experimental variation. In order to visualize the signal-to-noise ratio the fluorescence-intensity ratio of either two RNA-samples from different growth conditions or from two RNA-samples from the same growth condition are plotted in double-logarithmic scales (Figure 3). It can be clearly seen that the comparison of two different metabolic states scatters much broader than the self-to-self comparison. The variance of the data is displayed by their respective Pearson correlation value r. Five ORFs in the self-to-self comparison show an unexpectedly large variation. In three cases (asr7152, alr7535 and alr7580) this can be explained by their low, close to threshold fluorescence-signal intensity.

It is often observed in oligonucleotide-based DNA-microarray experiments that probes directed against one single transcript show large hybridization level variations. The reason for this fluctuation remains still unknown and is one major reason for the necessity to calculate the expression value for each transcript from several unique probes (Kuo et al. 2002; Pozhitkov et al. 2006). We analyzed the influence of both GC-content and secondary structure formations on the hybridization signal.

Figure 4a shows the correlation between the number of guanine and cytosine nucleotides (GC-content) in the 25-mere probes and the corresponding hybridization signal. There is no probe with less than 3 or more than 17 GCs. In the range between 7 and 13 GCs there is a clear linear correlation between GC-content and hybridization signals. To many (more than 50%) or to few (less than 25%) GCs in the probe result in non-linear behavior. As can be seen from the number of observations given in the plot, less than 0.7% (328 out of 48208) of all unique probes are affected by this non-linear behavior. At the current stage it is not possible to draw conclusions from the extremes on both sides of Figure 4a because they are only represented by few data. In contrary to the GC-content, the predicted secondary structure (stems, loops and bulges) of the transcript has only little influence on the hybridization signal (Figure 4b). This was expected because (a) the target was chopped into smaller fragments prior to hybridization, and (b) the hybridization conditions are set such that no secondary structures should form in either the probe or the target.

The effect of individual probe hybridization signals is usually ignored in oligonucleotide-based DNA-microarray experiments. Instead, the average over all probes is used for each transcript. However, these effects have to be taken into account if only few probes are available for particular transcripts. Furthermore, we can conclude that a large portion of the hybridization signal variation is intrinsic to the probe sequence and can not be explained by currently known DNA-duplex formation physico-chemistry.

DNA-microarray experiments involve accumulation and management of large amounts of data. Apart from the experimental data, information from open access knowledge databases and sequence analysis are collected. To provide optimal accessibility to all data we set up a MySQL database on an Apache driven Internet server. The database holds both raw and processed data. Besides data management the database allows cross-connectivity of expression data with annotations from NCBI database, Cyanobase and KEGG (Figure 5) In order to access and query the database (which has been coined HyDaBa) a PHP-based and Internet-accessible front-end has been developed. This front-end helps to query the data, guides the user to define and store new queries, allows data up-and download, and can be easily extended due to its modular setup with template pages. The most important feature of HyDaBa constitutes the mapping of gene-expression data onto metabolic charts from the KEGG database (Figure 5). Technically, this has been achieved by using a SOAP interface (Kawashima et al. 2003). Equally important is the possibility to query for all data available for a given ORF. HyDaBa can be accessed at http://www.hydaba.uni-koeln.de/ using “guest” and “hydaba06” as user name and password, respectively.

Growth on dinitrogen as sole nitrogen source acts like a positive transcriptional switch in Nostoc PCC 7120. There is a much larger fraction of genes stronger expressed under nitrogen fixing than under non-nitrogen fixing conditions and only a minority of genes shows a decreased expression level. Only 17 annotated and 12 hypothetical ORFs exhibit a significantly higher expression under non-nitrogen fixing conditions (Table 2) whereas 281 annotated and 184 hypothetical ORFs are more strongly expressed under nitrogen fixing conditions. In Figure 6, these gene-expression differences are clustered according to the participation of the corresponding ORF in specific metabolic categories. The strongest expressed genes participate in photosynthesis and respiration (K). Closer analysis reveals that 21 of the 29 strongest expressed genes in this group belong to photosynthesis, 11 of which are structural proteins of phycobilisomes (Table 3). These findings clearly illustrate the extensive energy demand for nitrogen fixation. The cell expands its light harvesting complexes in order to direct more light energy to the photosystems and produce both more ATP and NADPH. The stronger expression of proteins involved in respiration underlines previous findings that the respiration rate is increased under nitrogen fixing conditions in cyanobacteria (Murry and Wolk, 1989). It is believed that this process supports the removal of oxygen which otherwise would inactivate the nitrogenase enzyme complex (Fay, 1992).

As a key global regulator, NtcA plays an important role in the expression of many genes involved in heterocyst differentiation and nitrogen assimilation. For the unicellular, non-differentiating cyanobacterium Synechococcus PCC 7942 it has been shown that the binding affinity of NtcA to its target DNA-sequence is elevated by 2-oxoglutarate (Vázquez-Bermúdez et al. 2002; Vázquez-Bermúdez et al. 2003). Thus, 2-oxoglutarate exerts a direct role on NtcA-mediated transcription activation. Furthermore, it plays a central role in sensing the nitrogen status, or rather the C/N-balance. Although synthesized in the heterocyt, 2-oxoglutarate cannot serve as a substrate for glutamate synthesis in heterocysts because the necessary ferredoxin glutamine 2-oxoglutarate amidotransferase is not expressed in heterocysts (Martín-Figueroa et al. 2000). This conversion is done in the vegetative cells. Thus, glutamate is imported, while 2-oxoglutarate and glutamine are exported to vegetative cells (Figure 7). We found key enzymes catalyzing the synthesis of 2-oxoglutarate, aconitase hydratase (2.8-times; EC 4.2.1.3) and isocitrate dehydrogenase (4.7-times; EC 1.1.1.42), respectively, being stronger expressed under nitrogen fixing condition (Figure 7). The hetC gene (alr2817), which encodes a putative ABC-transporter that is essential for heterocyst formation, has been shown to be a direct target of the transcriptional regulator NtcA (Muro-Pastor et al. 1999). Indeed we see a 5.8-time stronger expression under nitrogen fixing conditions. Table 4 shows expression differences for all known ORFs involved in heterocyst formation included in this study.

The conversion of dinitrogen to ammonia, catalyzed by the nitrogenase enzyme complex, is only the first step in a series of reactions that make nitrogen available to the cell. The nitrogenase enzyme complex provides two products that are metabolized, hydrogen gas and ammonia, respectively. The former is taken up by an uptake hydrogenase while the latter is incorporated to glutamate by the glutamine synthase yielding glutamine. Figure 7 gives an overview over the main pathways and enzyme complexes involved in nitrogen fixation. The nitrogenase consists of three subunits, the molybdenum-iron protein alpha chain (NifD, all1454), the molybdenum-iron protein beta chain (NifK, all1440), and the iron protein (NifH, all1455). We found nifK and nifH to be more than 10-times more strongly expressed under nitrogen fixing conditions. One ORF (nifH2, alr0874) of yet unknown function that is paralogous (92% similarity, 86% identity) to the iron protein (nifH, all1455) was found to be expressed at a very lower level and is only slightly stronger expressed under nitrogen fixing conditions (Figure 7). Thus, the gene product of nifH2 is probably not involved in the nitrogen fixation reaction. The glutamine synthase (glutamate-ammonia ligase) is encoded by glnA (alr2328). Northern blot studies in Nostoc PCC 7120 have shown that the glnA transcript is present in both nitrogen-fixing as well as non-nitrogen fixing cultures, but more abundant in the latter (Orr and Haselkorn, 1982). This is in accordance with our results.

Other genes

Phycobilisomes. Phycobilisomes are the major light-harvesting complexes of cyanobacteria. These are multiprotein assemblies that are functionally associated with photosystem II and constitute up to 50% of the total cellular protein. It has been shown previously that phycobilisomes serve as a nitrogen storage. Upon nitrogen starvation they can be completely degraded within two days. Phycobilisome degradation is thought to provide substrates for protein biosynthesis. As discussed above, we observed strong expression of major components involved in photosynthesis under nitrogen fixing conditions. Accordingly, all proteins included in this analysis and that constitute phycobilisomes are around 3-times more strongly expressed under nitrogen-fixing conditions.

The authors like to thank Dr. M. Baum and N. Rittner from febit for technical assistance with the microarray experiments and H. Eckes for initial work on the database front-end. R.W. likes to thank Prof. D. Tautz for his continuous support.