Overexpression of AtHb1is targeted to seed and increases seed weight

We generated transgenic Arabidopsis plants expressing the endogenous AtHb1 under the control of the seed-specific LeB4 promoter [19]. Northern blot analysis of siliques from homozygous T3 plants demonstrated significant AtHb1 expression, whereas in WT plants the endogenous AtHb1 expression was not detectable under standard conditions (Figure 1A; for additional transgenic lines see below). RT-PCR analysis showed that, overexpression of AtHb1 under the control of the LeB4 promoter was restricted to siliques/seeds in the transgenic plants (minor expression in roots; Figure 1B). Comparison of manually isolated seeds with whole siliques (including seeds) revealed that LeB4-driven expression is mainly localized in seeds in agreement with previous results [19]. To avoid any stress-induced artefacts that might be induced by dissection of seeds from the siliques, whole siliques were used for further studies

Overexpression of AtHb1reduces the endogenous level of nitric oxide in seeds

A qualitative fluorescence assay with diaminofluoresceine-2-diacetate (DAF-2DA) was used for detection of endogenous NO in WT and AtHb1 embryos under standard and hypoxic stress conditions.

To induce moderate hypoxic stress in the seeds, intact plants were treated with artificial air mixes containing only 10.5 kPa oxygen (corresponding to half atmospheric oxygen saturation) for one hour. Seeds of WT plants showed a slight induction of AtHb1 expression under these conditions (Figure 1C), but its expression level was still much lower than in the transgenic plants. Microarray results confirmed the higher abundance of AtHb1 mRNA in transgenics under hypoxia (>3-fold, Figure 2A, marked by asterisk).

Under standard growth conditions, NO was not detectable in the embryos of either WT or AtHb1 plants using the fluorescence assay. Possibly, the steady state level of NO was below the detection limit of the assay. Under moderate hypoxia, WT showed a clear fluorescence signal (in green), while AtHb1 overexpressors did not (Figure 1D). This indicated strongly decreased NO levels in the latter. Thus, the transgenic approach resulted in lower levels of NO. The induction of AtHb1 expression (Figure 1C) and enhanced NO emission (Figure 1D) in WT further indicated that the moderate stress treatment was sufficient to induce hypoxia in seeds.

Experimental set up for microarray analysis

To assess changes in gene expression in seeds/siliques due to AtHb1 overexpression in detail, we focused on line L1-1, which showed the strongest transgene expression. Six other independent transgenic lines were involved in further studies (see below).

WT and transgenic plants were exposed to moderate hypoxia (10.5 kPa) or normoxia (21 kPa; control) for one hour. Three biological replicates were used for hybridization to Affymetrix ATH1 arrays. A cluster dendrogram of transcript signal intensities from the 12 arrays showed a high reproducibility of the biological replicates from each data set (genotype+treatment), and indicated a greater influence of the genotype than the treatment on transcriptional profiles (Additional file 1A). Transcript analysis by qRT-PCR showed a high correlation (R² = 0.83) with the microarray data, confirming the reliability of the data (Additional file 1B).

We compared the transcriptome of WT and AtHb1 siliques/seeds under control and hypoxic conditions, as well as the hypoxic responses in each genotype. Differentially expressed genes were extracted from the data base by applying the following cutoffs: a fold-change of >2 and a p-value of <0.05. A total of 1,010 genes were identified as differentially expressed in all of the comparisons. Differentially expressed genes were grouped into eight clusters (Additional file 2 and 3), classified into functional groups using the MapMan bin code [20] and ordered by pathways. The heat map display in Figure 2 gives a detailed view of the altered pathways (also listed in Additional file 4).

To confirm that microarray data of L1-1 are reproducible in further transgenic lines, we analyzed the expression of selected genes in six other AtHb1-overexpressing lines by qRT-PCR (Figure 3). A set of transcripts that have been shown in the microarray analysis to be upregulated by AtHb1 overexpression was selected for qRT-PCR analysis. All of the transgenic lines exhibited an enhanced expression of the genes from representative signaling, redox and metabolic pathways compared to the WT, indicating similar expression profiles due to AtHb1 overexpression in independent transgenic lines.

AtHb1overexpression induces stress-related regulatory pathways under non-stress conditions

Genes implicated in signaling pathways, like receptor kinases, wall-associated kinase 1 (WAK1, At1g21250) and MAPK kinase 9 (At1g73500) were also upregulated compared to WT. WAK1 is a transmembrane protein containing a cytoplasmic Ser/Thr kinase domain and an extracellular domain bound to the pectin fraction of cell walls [21], thus enabling communication between cell wall and cytoplasm. Phosphorylation via WAKs has been shown to play a pivotal role in cell wall metabolism [22], which was significantly altered by AtHb1 overexpression. WAK1 expression is induced by SA treatment [23], thus, higher expression of WAK1 and two S-adenosyl-L-methionine:carboxyl methyltransferases indicates an involvement of SA signaling in the regulatory networks controlled by AtHb1. In addition, the expression of 11 transcripts encoding receptor kinases, such as transmembrane kinase RLK5 and other leucine-rich repeat family proteins as well as Ser/Thr kinases, revealed the presence of different signaling pathways. Interestingly, RLK7 (At1g09970) has recently been shown to be involved in the control of seed germination and tolerance to oxidative stress [24]. Using genetic approaches the authors provided evidence for a positive correlation of RLK7 expression and enhanced tolerance against H2O2.

Transcripts encoding proteins involved in redox homeostasis, such as manganese superoxide dismutase (MnSOD, At3g56350) and two glutathione-S-transferases, were upregulated in AtHb1 overexpressors. This was accompanied by higher expression of defence-related proteins, i.e. dehydrins and major latex proteins (MLP-related) (Figure 2A).

Ubiquitin-mediated proteolysis is essential for plant development and responses to environmental stimuli [25]. AtHb1 induced the expression of three RING finger E3 ligases of the C3CH4-type (At4g14365, At2g27940, At1g30860) and two F-box proteins (SKP1/At2g45950 and kelch repeat/At1g80440) (Additional file 6). RING finger ligases and E3 ligases from the SKp1, F-box (SCF) complex play an essential role in auxin metabolism by degrading AUX/IAA proteins, and thereby regulating concentrations of IAA [25]. This is probably linked to downregulation of auxin transport and signaling in AtHb1 plants.

AtHb1overexpression in seeds alters expression of genes involved in primary metabolism

AtHb1 overexpression induces various changes in transcripts related to carbohydrate, cell wall, N- and lipid metabolism, as well as potentially associated transporter gene activities and photosynthesis. As deduced from GO analysis of transcript data, the cell wall was the most affected cellular compartment in AtHb1 seeds showing a clear underrepresentation (Additional file 5). Other decreased biological processes are linked to cell wall biogenesis and modification. This is illustrated by the concurrent downregulation of more than 30 cell wall-related genes encoding cellulose synthases, arabinogalactan-proteins (AGPs), pectinesterases, expansins, xyloglucan-xyloglucosyl transferases and polygalacturonases (see MapMan visualization, Additional file 7). This indicates a strong repression of cell wall synthesis, cell wall modification, pectin degradation, cell expansion and cell wall turnover. Two transcripts (At1g70290, At2g18700) encoding class II trehalose-6-P synthase/phosphatase (TPS8, TPS11) were preferentially expressed in AtHb1 plants. These transcripts are also potentially linked to cell wall metabolism, as it was found that perturbation of trehalose metabolism in embryos of the tps1 mutant leads to changes in cell wall composition and thickness [26]. Lipid metabolism also showed transcriptional alterations; fatty acid elongation and desaturation were activated but transcripts involved in squalene and steroid metabolism were repressed. In addition, transcripts for malate synthase and isocitrate lyase (key enzymes of the glyoxylate pathway) were upregulated in AtHb1 seeds. Furthermore, transcripts encoding the 4Fe-4S cluster protein of photosystem I and key enzymes of the photorespiratory pathway (glycolate oxidase/GOX, At3g14415; serine hydroxymethyltransferase 4/SHMT4, At4g13890) were downregulated.

Nitrogen metabolism appears to be affected in AtHb1 seeds based on the downregulation of nitrate reductase 2 (NIA2, At1g37130) and nitrite reductase 1 (NiR1, At2g15620). Several transcripts involved in amino acid metabolism differed significantly between transgenic and WT (S-adenosylmethionine synthetase, S-adenosyl-L-homocysteinase, asparaginase, cystine lyase, delta-1-pyrroline-5-carboxylate synthetase).

Several transporter gene activities were commonly downregulated in AtHb1 seeds, namely those involved in sugar, amino acid and oligopeptide transport (POT). Most of these are proton-coupled transporters. In addition, five genes from different subgroups of the aquaporin family were downregulated. These genes play a role in nutrient flow and/or are implicated in remobilization [27, 28].

Changed gene interactions due to AtHb1overexpression point to alterations in cell wall metabolism

To infer gene-to-gene interactions we used the MRNET approach which extracts statistical dependencies between genes [29]. The reconstructed network of gene interference for the top 20 genes that are differentially expressed between WT and AtHb1 overexpressing seeds under control conditions showed clear differences (Additional file 8). In WT, the gene encoding fasciclin-like arabinogalactan protein 13 (FLA13; At5g44130) was the central hub. AGPs, such as FLA13, play a role in plant cell elongation/cell wall biogenesis, and are assumed to act as signal molecules [30]. Proteins containing fasciclin domains have also been shown to function as adhesion molecules in a broad spectrum of organisms [31]. There were multiple interactions of this hub with genes encoding proteins localized to the cell wall (e.g. xyloglucan:xyloglucosyl transferase, xyloglucan endotransglycosylase 3 (XTR3), proline-rich protein 2 (ATPRP2) and acid phosphatase class B family protein) or otherwise involved in extracellular matrix modifications (e.g. midchain alkane hydroxylase, which is involved in cuticular wax biosynthesis; [32]). Most of the genes are implicated in stress-responses and related to hormone (ABA, GA) action. Overexpression of AtHb1 directly or indirectly perturbed the strong multiple interactions of the hub gene FLA13, shifting the main regulatory point to ATPRP2. It has been shown, that ATPRP2 is one of the key genes involved in cell specification [33]. Cell specification in the embryo might be coupled to maturation processes, which are characterized by high storage- but extremely low mitotic-activity. Downregulated expression of ATPRP2 (and associated genes) in AtHb1 plants might therefore indicate decelerated cell specification and thus, an extented growth phase.

Evaluation of adaptive stress responses in wild type seeds

Most of the adaptive responses in WT seeds have also been described for shoots and roots of Arabidopsis plants. Mustroph et al. [18] identified a core set of 49 translated hypoxia-induced mRNAs in 21 different Arabidopsis cell populations. From this core set, 35 genes (~70%) were also found to be upregulated in seeds, indicating similar adaptation strategies to hypoxia regardless of tissue/organ identity. The possible induction of the glyoxylate cycle in combination with lipid degradation (phospholipase C, phosphodiesterase) was not observed in other Arabidopsis tissues and might therefore be seed-specific. The induction of the glyoxylate cycle could represent an alternative mechanism to generate sugars and sustain energy supply under unfavourable conditions in seeds. Interestingly, malate synthase and isocitrate lyase are also enhanced in carbon-starved cucumber cotyledons [34]. The higher expression of genes involved in sugar, amino acid, oligopeptide and general nutrient (aquaporins) transport in WT (column 2 in Figure 2B) and the significantly reduced sucrose concentrations (see below) indicates nutrient, particularly sugar, depletion in WT upon hypoxia.

In general, WT seeds showed a strong transcriptional and metabolic response to moderate hypoxia. Metabolism and signaling of hormones (ABA, ethylene, JA, SA and GA) which are described to be important triggers in response to oxidative stress [15, 16] are strongly induced in seeds. Activation of specific transcription factors and signaling pathways nicely illustrates a cross-talk of hormone action and regulatory pathways, particular for ethylene. Upregulation of MAPKK9, MAPK3 (At3g45640) accompanied by activation of ACC oxidase1 (At2g19590) as well as ten members of the AP2/EREBP family represents an example how signaling cascades are linked together in adaptive stress responses. Experiments with maize suspension cultures showed a correlation of varying class-1 hemoglobin levels and changed NO concentrations with ethylene formation [35]. Enhanced ethylene biosynthesis under hypoxia is linked to lower hemoglobin expression, coinciding with the stronger induction of ethylene synthesis and signaling in the WT compared to the AtHb1 plants in our experiments. Beside the strong activation of several WRKY transcription factors and MYB44 (At5g67300), transcripts related to redox regulation were clearly induced. Rising concentrations of H2O2 in WT upon hypoxia correlate with transcriptional activation of several ROS generating/scavenging enzymes coinciding with other studies [36, 37]. The upregulation of several class II TPS genes and the reduction of trehalose-6-P (T6P) levels was part of the hypoxic response in WT (two of them are also induced in transgenics under control conditions). Interestingly, T6P metabolism was identified as being part of a hypoxic response that is conserved in some pro- and eukaryotes [38]. T6P may be involved in coordination of carbon partitioning between primary metabolism and cell wall synthesis [39]. Therefore, altered expression of TPS genes - together with changes in cell wall metabolism - accentuates the possible role of T6P metabolism in regulation of carbon partitioning. In general, the alterations in regulatory and metabolite pathways provide a framework of seed-specific responses to hypoxia.

AtHb1overexpression attenuates transcriptional stress responses

Under hypoxic stress treatment, a significantly lower number of transcripts exhibited altered expression in AtHb1 compared to WT (254 and 413 genes, respectively). Consequently, the stress response observed in AtHb1 was much reduced, especially in regulatory/signaling pathways, but also for specific pathways in primary metabolism. Transcriptional alterations in WT upon hypoxia partly shared a commonality with those induced by AtHb1 overexpression under control conditions, or with transcripts additionally induced in AtHb1-overexpressing plants after hypoxia (Figure 4). The moderate hypoxic response in seeds of transgenic plants, in combination with genes induced by AtHb1 overexpression that have been shown to be implicated in the WT hypoxia response, points to a kind of "pre-adaptation" to oxidative stress. Among the differences between the two genotypes in their hypoxic responses, several biological processes stand out, namely, stress-related signaling, redox pathways and primary/energy metabolism (Figure 2, Additional file 4). These differences are discussed in detail below.

First, hypoxia induced stress-related signaling and redox pathways in WT. GO analysis for functional assignments of upregulated genes showed strong overrepresentation of responses to abiotic/biotic stress and other biological processes related to stress responses, especially responses to ABA and JA. Evaluation of promoter motifs within the 5'-flanking regions of hypoxia-induced genes revealed that W-box, ABRE, DREB, G-box, MYC2, MYCATERD1, GADOWNAT, Z-box, I-box and Evening Element motifs were significantly overrepresented. This finding is significant because almost all of these recognition sites have been implicated in hormone signaling (ABA, ethylene) and in general stress responses. In addition to these changes in hormone signaling pathways, transcripts directly involved in biosynthesis of ABA, ethylene, JA and SA were commonly upregulated in WT. In contrast genes related to SA, GA and ABA metabolism were not induced by hypoxia in AtHb1 plants. In fact, a strong repression of ABA synthesis/signaling was evident from the down regulation of NCED4 and several ABA-responsive genes, among them ATEM6 and AtHVA22b (which were already induced under control conditions by AtHb1 overexpression). In addition, ABRE binding site motifs were enriched in the set of downregulated genes in AtHb1 plants after hypoxia (Table 3). Another striking difference between the genotypes is the opposite regulation of transcripts encoding the gibberellin regulated proteins 2 and 3 (GASA 2, 3); they are highly upregulated in the WT after hypoxic treatment whereas a strong repression was observed in transgenic seeds. Calcium signaling seems to play a role in the hypoxic response of WT, as indicated by the upregulation of six transcripts encoding calmodulins and calmodulin binding proteins, accompanied by an induction of calcium dependent protein kinase and the plastidic Ca ²⁺ -ATPase1 (ACA1, At1g27770). The transcriptional activation of calmodulins which are the primary calcium receptors in plant cells and calcium binding proteins, could serve as substrate for phosphorylation by calcium dependent protein kinases, then activating transcription factors by phosphorylation. Altogether this points to existing calcium dependent signaling pathways in the hypoxia response in wild type seeds, which were not observed in AtHb1 overexpressors.

The second major difference between AtHb1-overexpressing plants and WT concerned primary and energy metabolism. Hypoxia induced multiple changes in transcripts related to these processes in WT, but only moderate changes in AtHb1 plants. For example, in WT we encountered a clear induction of glycolysis and fermentation (FBP aldolase, PFK, PDC1, ADH1) as well as strongly induced nitrogen assimilation as suggested by preferential expression of NIA2 and NiR1. In WT, cell wall metabolism was downregulated as evidenced by repression of six transcripts encoding pectinesterases and four encoding polygalacturonases, indicating that cell wall metabolism is one of the key processes affected by hypoxia. Induction of carbonic anhydrases and genes implicated in lipid degradation and the glyoxylate cycle (malate synthase, isocitrate lyase) was apparent in the WT response but not in AtHb1 plants. The activity of transporter genes is directly linked to primary metabolism. The strong induction of genes encoding proline transporter, POT as well as TIP1.2 and TIP3.2 is also restricted to the hypoxia response in WT and might reflect a higher demand for remobilizing storage compounds and thus, indicating nutrient depletion in WT. The alterations observed in the transgenic plants were restricted to upregulation of glycolysis/fermentation (PFK, PDC1, ADH1) and a few transcripts related to cell wall degradation.

AtHb1plants show less pronounced metabolic adjustment under transient hypoxia

The steady state level of amino acids, sugars, metabolic intermediates and H2O2 were measured in seeds/siliques of both genotypes under control and hypoxic conditions. Under control conditions, the levels of phosphoglycerate and ADP-glucose (starch precursor) were higher in WT versus AtHb1 plants, while sucrose and UDP-glucose (cell wall precursor), showed elevated levels in AtHb1 plants (Figure 5, values are given in Additional file 9). Remarkably, the levels of many metabolites changed after hypoxic treatment in WT but were barely altered in AtHb1 plants. In WT plants only, the levels of T6P and sucrose dropped significantly, while pyruvate increased (indicative of enhanced glycolytic flux and/or a partial block of the TCA cycle). Altogether, the metabolite profiles of the two genotypes illustrated a strong metabolic adjustment in WT in response to moderate hypoxia, whereas in AtHb1 only marginal changes were detected. This differential response was clearly visualized using principal component analysis (PCA; insert in Figure 5). Transcript data hinted at shifts in ROS metabolism in transgenic plants and in the hypoxic response of WT. Measurements of H2O2 levels in both genotypes under control and hypoxic conditions are consistent with transcriptional activities of H2O2 generating and scavenging enzymes. Higher concentrations in AtHb1 seeds/siliques compared to WT under control conditions (Figure 6A) correlate with preferential expression of MnSOD1 and glutathione-S-transferases. Upon hypoxia, H2O2 levels in WT increased but were unchanged in AtHb1 seeds. Activation of respiratory burst oxidase homologue D, MnSOD1, redoxins, three glutathionine-S-transferases and alternative oxidase 1D (AOX1D, At1g32350) in WT indicates an enhanced ROS metabolism under hypoxia.

Overexpression of AtHb1promotes respiration and maintains the energy status under transient hypoxia

To investigate changes in energy metabolism we measured the respiratory activity of developing seeds. Under control conditions respiration rates were similar in both genotypes (1.7 ± 0.2 pmol/µg embryo min). However, under hypoxia, respiration in AtHb1 plants (line 1-1, 1.05 ± 0.14 pmol/µg min) was about 40% higher than in WT (0.73 ± 0.13 pmol/µg min) pointing to a higher energy supply in the former. Indeed, both the adenylate energy status (AEC = (ATP+0.5ADP)/(ATP+ADP+AMP)) and total ATP levels were elevated in AtHb1 versus WT under hypoxia (Figure 6B-C).

Direct comparison of microarray data from the two genotypes under hypoxic conditions identified only the gamma-subunit of the chloroplast ATPase to be significantly upregulated in AtHb1 seeds. Screening our dataset for other differentially expressed transcripts involved in electron transport chain/ATP synthesis, we found five other transcripts, encoding ATP synthase, NADH dehydrogenase, NADH:ubiquinone oxidoreductase, cytochrome C oxidoreductase subunit 5c (COX 5C), with a tendency to higher expression in AtHb1 seeds under hypoxia (fold-changes between 1.4 and 1.64 and p-values < 0.05). These transcripts were found by qRT-PCR analysis to be nearly doubled in the AtHb1-overexpressing plants compared to WT (Figure 6D). Altogether, our data suggest that AtHb1 overexpression enables the seed to respire at higher rates especially under hypoxia, thereby increasing the ATP supply.

AtHb1overexpression induces stress-related signaling pathways and limits energy-consuming pathways

Under control conditions, AtHb1 overexpression activated several stress-related hormonal and signaling pathways. The fact that hormones and other components of signal transduction cascades work downstream of AtHb1 suggests that AtHb1 represents a high ranking signaling component with broad impact on regulatory networks. Most prominent was the induction of ABA synthesis/signaling, and the general repression of auxin transport/signaling. Evidence for induced ethylene and SA signaling came from induced MAPKK and WAK1-mediated signaling routes.

ROS formation also seems to be part of the AtHb1 signaling cascade as transcripts involved in formation and detoxification of H2O2 were clearly upregulated. Higher H2O2 levels in AtHb1 seeds confirmed the transcriptional activities. The pronounced upregulation of these stress-related signaling pathways might act in combination to "pre-adapt" the seeds to hypoxic stress. A role for plant non-symbiotic hemoglobins in redox regulation by improving the antioxidant status was previously hinted at by studies of alfalfa root cultures overexpressing a non-symbiotic hemoglobin [40]. Hb1-overexpressing lines revealed increased ascorbate levels as well as higher activity of enzymes involved in ROS removal. An enhanced oxidative stress tolerance during seed germination of Arabidopsis was induced by seed-specific overexpression of antioxidant genes [41]. Overexpression of MnSOD and/or combination with other genes encoding antioxidant enzymes during seed development and germination increased tocopherol contents and antioxidant capacities in mature seeds indicating beneficial effects of activated redox-related pathways on oxidative stress tolerance.

Alterations in transcriptional networks were accompanied by changes in primary metabolism. Cellulose synthesis, deposition of pectin fragments, incorporation of arabinose-derived sugars and glycosyl-transferring reactions all require energy and use activated nucleotide sugars. Thus, cell wall metabolism is clearly dependent on the energy and carbon status of cells. The decrease in transcripts related to cell wall metabolism in AtHb1 plants was the most prominent finding. The analysis of gene-to-gene interactions (MRNET approach) indicates that AtHb1-mediated downregulation of the hub gene FLA13 is of central importance for the proposed changes in cell wall metabolism. Its downregulation might eventually affect cell elongation, energy usage and carbon partitioning. Downregulation of cell wall metabolism might represent a major strategy to reduce energy (as well as carbon) consumption. Higher concentrations of UDP-glucose (precursor for cell wall synthesis) and sucrose support this idea.

Consistent with such energy saving adjustments is the transcriptional repression of proton-coupled transporters and photorespiration. Both require energy in the form of ATP, and thus, their repression implies a reduction in energy consumption. Another striking feature was the downregulation of NIA2 and NiR1 by AtHb1 overexpression under control conditions. While this might indicate lower nitrate assimilation (which imposes a high energy demand), the level of free amino acids was not reduced in transgenic seeds but rather elevated. The shift in NIA2/NiR1 expression could also be linked to NO signalling, because NIA can produce NO from nitrite [42–44]. High NO concentrations correlate with NIA activation and high nitrite levels [45, 46]. Genetic studies using the nia1nia2 double mutant indicate that NIA is a major enzymatic source of NO formation in plants [47]. Subsequently, the coordinated downregulation of NIA2/NiR1 due to AtHb1 overexpression could prevent the accumulation of nitrite and subsequent NO formation. This would contribute to the lower steady state NO levels in the transgenics (beside the NO scavenging function of AtHb1).

Transcripts encoding key enzymes of photorespiration (GOX, SHMT4) were downregulated by AtHb1 overexpression. Photorespiration results in a net loss of fixed carbon and energy. The apparent repression of this pathway is a further indication for the energy-saving mode of metabolism. The preferential expression of the β-carbonic anhydrase1 in WT might also be related, as this enzyme is known to control CO2 availability to Rubisco and thereby regulate photorespiration [48, 49].

Overall, alterations in the metabolism of AtHb1-overexpressing seeds point to an energy-saving mode of metabolism.

NO formation and signaling pathways are repressed by AtHb1overexpression resulting in improved respiration under stress

AtHb1-overexpressing seeds showed a much attenuated hypoxic response, with only some of the characteristic pathways being induced under hypoxia (e.g. enhancement of ethylene signaling, JA metabolism, redox-related transcripts and MYB transcription factors). Of particular note is the repression of the ABA response in the AtHb1 overexpressors, which contrasts with the strong induction observed in WT plants. Major differences were also obvious in calcium-dependent and GA-mediated signaling pathways. Both seem to play a much less significant role when compared to WT (e.g. GASA2/GASA3 showed the opposite responses in the two genotypes). Similarly, at the metabolite level, only minor alterations were apparent in response to hypoxia (in contrast to WT).

Another major difference in the hypoxic response of the two genotypes was the reduction of NO levels in AtHb1-overexpressing seeds. This agrees with previous findings [4, 50] and could be attributable to AtHb1-mediated degradation of NO [7] and/or the restriction of NO formation via transcriptional downregulation of NIA2/NiR1. As AtHb1 overexpression represses NIA2 and NiR1 activity under control conditions and especially after hypoxia treatment it could be concluded that NO formation is strictly prevented by the reduction of NO precursors (e.g. nitrite). Studies from Wang et al. [51] provided evidence that NIA2 is responsible for stress-induced NO formation in Arabidopsis roots. They demonstrated that NIA2 is phosphorylated by MAPK6 leading to an increase of NR activity and subsequently NO formation. MAPK3 also interacted with NIA1 and 2 in the yeast two-hybrid system implying a role for activation of NIA activity. The transcriptional upregulation of MAPK3 and NIA2 in WT seeds after hypoxia is in agreement with this finding. Possibly MAPK3 represents a seed-specific transducer of environmental stimuli whereas MAPK6 is predominantly involved in NO biosynthesis in roots. Assuming that overexpression of AtHb1 lowered levels of NO in planta, the present approach enabled us to discriminate between the more general hypoxia response and the target genes specifically induced by higher NO levels in WT. The direct comparison of the transcriptome of both genotypes under hypoxic conditions (Figure 2, column 2) revealed differences which might be specifically attributed to NO signaling. Calcium signaling is linked to NO signaling pathways [52, 53] and possibly directly involved in the regulation of hemoglobin expression [54]. NO induces a rapid increase in calcium concentrations [55, 56], and vice versa [53]. This relationship was found in transgenic plants, where both NO levels and calcium-dependent signaling were lowered compared to WT. Hints for a crosstalk of NO and GA signaling came from studies with isolated aleurone cells of Arabidopsis. Bethke et al. [57] showed that NO works upstream of GA in a signaling pathway, supporting our results that GA is possibly linked to higher NO levels in WT. NO-responsive genes in Arabidopsis were identified by microarray analyses using the synthetic NO donors SNP and NOR-3 [58, 59]. Among them genes involved in calcium signaling (calmodulins, calcium binding proteins), sugar and peptide transporters as well as glycosyltransferases which are preferentially expressed in the WT under hypoxic conditions. Based on our genetic approach we can separate these transcripts from transcripts of stress-related pathways (which are part of the hypoxia response without NO synthesis/accumulation).

According to our working hypothesis, lower NO levels in AtHb1-overexpressing seeds were expected to stimulate respiration because NO inhibits cytochrome C oxidase [60, 61]. In fact, seeds of the transgenic plants retained respiratory activity as well as higher expression of COX 5C transcripts under hypoxia, whereas the WT switched to a "stress" mode. Congruently, there was a preferential expression of other genes related to electron transport chain/ATP synthesis in AtHb1 plants. Combined with repression of energy-demanding processes (e.g. cell wall metabolism) this eventually leads to an improved energy status of cells in AtHb1-overexpressing seeds.

Generation of transgenic plants, growth conditions and treatment

The coding region of AtHb1 (At2g16060) was PCR-amplified (F-GGATCCGAGGTTGTGAAATATTATGGAG and R-GGATCCTAGGATTTTGGAATGCACACTA BamHI sites underlined) using a full-length AtHb1 clone (kindly provided by P. Geigenberger, LMU Munich, Germany). After subcloning into the pCR4-TOPO vector, AtHb1 was introduced into the modified binary vector pBAR between the LeB4 promoter [19] and OCS terminator. After sequencing, the construct was mobilized in Agrobacterium tumefaciens EHA105 and used for transformation of Arabidopsis thaliana Col-0 plants by floral dipping [62]. Homozygous plants were selected on phytagar plates with ½ Murashige and Skoog medium [63] supplemented with phosphinothrycin (50 µg ml-1) and characterized by Southern blot analysis. Plants were grown at 22°C under a 16/8-h photoperiod, with a relative air humidity of 60% and an approximate light intensity of 100-150 µmol photons m^-2 second^-1.

Hypoxic and normoxic treatments were carried out with transgenic (T3) and WT plants 45 days after germination (DAG) corresponding to the mid phase of maturation ~11/12 days after pollination. Plants were aerated with a gas mixture containing 10.5% O2 (composed of a 1:1 mixture of ambient gas and N2) or ambient gas containing 21% O2 for control samples in darkness. After one hour, plants were decapitated and immediately frozen in liquid N2. About 70-80 siliques of the same developmental stage were dissected in liquid nitrogen and pooled for one biological replicate. Both hypoxic and control treatment runs were repeated twice to provide biologically replicated samples. From the pool of biological replicates sample material was used for microarray and metabolite analyses.

Northern blot and RT-PCR analysis

Isolation of total RNA from siliques/seeds was performed according to Heim et al. [64]. For northern blot analysis, 10 µg total RNA were blotted on nylon membrane (Hybond-N+, Amersham) and hybridized with a [³²P]-labelled 635-bp fragment of Arabidopsis AtHb1 cDNA. A 25S rDNA fragment was used as loading control.

For cDNA synthesis, isolated total RNA was treated with RNAse free TURBO DNase (Ambion) and 1 µg RNA was reverse transcribed using oligo(dT) primer and SuperScript III reverse tanscriptase (Invitrogen, Karlsruhe, Germany). Gene-specific primers for AtHb1 were used in the PCR reactions.

RNA preparation and microarray hybridization

Total RNA was isolated from intact siliques using a GENTRA kit (Biozym, Germany) according to the manufacturer's instructions. RNA was further purified using an RNeasy Kit (Qiagen) and subjected to DNAse digestion (Qiagen). Total RNA was quantified using a NanoDrop ND-1000 UV-Vis spectrophotometer (Nanodrop Technology) and RNA quality was assessed using an Agilent 2100 Bioanalyzer (Agilent Technology). Three independent biological replicates of each genotype (WT, AtHb1) and treatment (hypoxia, control) were hybridized to Affymetrix ATH1 Arabidopsis GeneChips (n = 12). Preparation of labelled cRNA and hybridization of oligonucleotide chips was performed at the Deutsches Ressourcenzentrum für Genomforschung (Germany).

Data analysis

Data were processed with the Affymetrix MicroArray Suite software package (MAS 5.0) and the resulting CEL files were analyzed using Bioconductor packages (http://www.bioconductor.org/) in R (http://cran.at.r-project.org/). Data were normalized using the Robust Multi-array Average (RMA) method [65]. Analysis of differentially expressed genes in the different comparisons was performed with the LIMMA package using the RMA normalized expression values [66]. The Benjamini and Hochberg method was selected to adjust p-values for multiple testing and to determine false discovery rates (FDRs) [67]. Genes were deemed to be differentially expressed only when (1) calculated p-value was < 0.05, (2) mean of the signal log2 ratio was > 1, and (3) signal intensities of probe sets from at least two of the three biological replicates were designated as "present" calls in the PMA analysis. Genes differentially expressed in all of the comparisons (i.e. in at least one of the four comparisons) were used as data sets for the subsequent clustering and gene category analyses.

K-means clustering was performed by means of the TMeV software package using log2 signal ratio data. The MapMan visualization tool was used for functional characterization of differentially expressed genes. Enrichment analysis of Gene Ontology (GO) terms for differentially expressed genes was performed as in Horan et al. [68]. For identification of conserved motifs in the promoters of differentially expressed genes the online tool Athena (http://www.bioinformatics2.wsu.edu/cgi-bin/Athena/cgi/analysis_select.pl) was used with the default settings.

All microarray data from this study have been deposited in NCBI Gene Expression Omnibus (accession number GSE23846).

Reconstruction of the gene regulatory network

Inferring regulatory networks from microarray data was done based on the information theoretic approach MRNET (package minet Bioconductor/R) using the top 20 of differentially expressed genes (given in Additional file 10). MRNET is based on the maximum relevance/minimum redundancy algorithm. The algorithm starts with computing the pairwise mutual information (MI) between all gene pairs. The resulting MI matrix is then manipulated to identify regulatory relationships and to reduce the number of false positives.

Quantitative Real-Time PCR

RNA preparations from microarray experiments were used for cDNA synthesis (see above). The Power SYBR Green PCR mastermix was used to perform reactions in an ABI 7900 HT Real-Time PCR system (Applied Biosystems, CA, USA). Data were analyzed using SDS 2.2.1 software (Applied Biosystems). Five replicate measurements were conducted for each gene. Expression values were normalized with transcript levels of the actin 2 gene (At3g18780) and calculated as an arithmetic mean of the replicates. Dissociation curves confirmed the presence of a single amplicon in each PCR reaction. Log2 fold-changes were calculated after Livak and Schmittgen [69]. Efficiencies of PCR reactions were determined using LinRegPCR software (http://www.gene-quantification.de/download.html). A list containing primers for the tested genes is given in Additional file 11.

Fluorescence detection assay for nitric oxide in embryos

Analysis of NO levels was done using DAF-2DA fluorescence detection [70]. Freshly isolated Arabidopsis embryos were incubated in 1 ml buffer solution containing: 50 mM sucrose, 10 mM KCL, 0.1 mM CaCl2, 10 mM MES-Tris (pH 5.6) and 50 µM DAF-2DA (Calbiochem, Germany). The buffer was aerated with 15 µM oxygen. After 1 h incubation, embryos were rinsed with fresh buffer to remove excess fluorophore. Fluorescence was analyzed using a laser scanning confocal microscope (510 Meta, Carl Zeiss, Jena, Germany).

Respiratory oxygen uptake

About 100 Arabidopsis seeds were incubated in 2 ml buffer (100 mM sucrose, ¼ MS-medium, 10 mM MES-NaOH, pH 6.35). Gas tight closed vessels equipped with an oxygen sensor SP-PSt3 and connected to a Fibox 3 oxygen meter (PreSens Sensing GmbH, Regensburg, Germany) were used. Oxygen concentration in the samples was registered during a time period of 3 min. From recorded data the respiration rate of seeds was calculated by linear regression.

Determination of metabolic intermediates, storage products and seed weight

Sugar-phosphates, nucleotide sugars and organic acids were extracted in chloroform/methanol (3:7 v/v) and measured by anion-exchange chromatography linked to tandem mass spectrometry [71]. For amino acid measurements 10 mg of powdered, frozen material was extracted in ethanol (80%, v/v), supplemented with 25 nmol norvaline as internal standard. Collected supernatants were vacuum-dried and resuspended in 250 µl water. Derivatization and separation of amino acids was performed according to Thiel et al. [72]. H2O2 was quantified using the Amplex Red Hydrogen Peroxide/Peroxidase Assay Kit (A22188; Molecular Probes, Invitrogen GmbH, Darmstadt, Germany) according to the manufacturer's instructions. Adenine nucleotides were measured as in Rolletschek et al. [73].

Average weight and number of mature seeds was determined in 4 independent batches of plants. In each batch, we used 5 individual plants per genotype, and counted the number of siliques per plant and the number of seeds per siliques (n = 10). From this we counted the total number of seeds per plant. Average seed weight was analysed in three generations (T3-T5) using an electronic microbalance (M2P, Sartorius, Göttingen, Germany). Total lipid of mature seeds was analyzed as fatty acid methyl esters by gas chromatography [74]. Total nitrogen and total carbon content were measured by elemental analysis (Vario EL3, Elementaranalysesysteme, Hanau, Germany).