Freshly matured seeds usually exhibit primary dormancy, a trait defined as the failure of viable seeds to germinate under favourable conditions [8]. Seed dormancy plays a crucial role in the survival of plant species, but is also important for agricultural practice to prevent pre-harvest sprouting under cool, high humidity conditions [24]. Primary dormancy can be released by either cold stratification, which is a low-temperature treatment of imbibed seeds, or by an extended period of dry seed storage (after-ripening; AR) [9].

The transition from dormancy to germination is a critical step in the life cycle of plants [25]. The plant hormone abscisic acid (ABA) has long been known to play a major role in the establishment and maintenance of seed dormancy and the inhibition of seed germination, whereas gibberellins (GAs) and several other hormones, including brassinosteroids, ethylene, and cytokinins, have been shown to promote seed germination [38]. However, it is especially the balance between ABA and GA that controls the decision to germinate or not [19]. Mutations in genes regulating ABA levels or -sensitivity result in a reduced degree of seed dormancy [34, 35]. Whereas GA biosynthesis or sensing mutants result in a block of germination [23, 31, 33]. This hormonal control is also integrated with the seed’s responses to environmental conditions, such as light [45], temperature [55, 60] and nutrients [40].

Recent advances in gene expression analysis using microarrays allow genome-wide expression studies to characterize seed dormancy and germination [10, 20, 21, 26, 27, 36, 41, 52]. Carrera et al. [12] used a targeted transcriptomics approach in imbibed non-dormant mutants (aba1 and abi1) compared to wild-type seeds that were or were not after-ripened. They concluded that, in Arabidopsis, after-ripening and dormancy are controlled by genetically separate pathways, and that ABA only affects the induction and maintenance of dormancy in imbibed seeds, but not after-ripening. The work also showed that application of exogenous ABA to after-ripened seeds does not mimic dormant seed states with respect to gene expression profiles. Recently it was shown that seed dormancy maintenance in the imbibed state was mainly controlled at the transcriptional level [3] and that transcriptional differences between dormant and non-dormant seed become visible already at early imbibition [17, 48].

Despite extensive research on dormancy and germination, many questions about the molecular mechanisms controlling these traits remain unanswered, likely due to its genetic complexity and the large environmental effects which are characteristic of these quantitative traits. Employing whole-genome scans for quantitative trait loci (QTL) is a common approach to identify genes involved in complex phenotypes. Particular attention in this method is given to the role of natural variation in the regulation of traits related to plant adaptation. Natural variation has been used to identify loci that control seed dormancy in nature. QTL analyses on six Recombinant Inbred Line (RIL) populations have identified eleven DELAY OF GERMINATION (DOG) QTL of which nine have been confirmed by near isogenic lines (NILs). The different DOG loci affect dormancy mainly by distinct genetic pathways as was concluded from the absence of strong epistatic interactions in the QTL analysis. This finding was confirmed by transcriptome analyses in freshly harvested dry seeds of the main DOGNILs, these lines showed distinct expression patterns compared to their genetic background Landsberg erecta (Ler). The genes identified in the different DOGNILs represent largely different gene ontology profiles [6].

Here we aim at identifying genes that are required for dormancy maintenance and germination of imbibed seeds. Moreover, we focus on what is in common between the different pathways. For this the transcriptome of freshly harvested (dormant) and after-ripened (AR; non-dormant) 24-h imbibed seeds of the same set of DOGNILs and Ler was investigated. We have identified sets of 46 and 25 genes that were up-regulated in seeds of all dormant (D-up) and all after-ripened (AR-up) DOGNILs and Ler, respectively. We have investigated their role in seed performance by analysing knock-out (KO) mutants in these genes. With seed performance we refer to the capacity of seeds to germinate under various environmental conditions. Traits that contribute to seed performance are seed dormancy, seed longevity (as estimated in an accelerated aging test) and germination under stress conditions, such as high salt, osmotic stress and ABA treatment [32]. In this study we have characterised several genes affecting seed performance.

In our search for novel players in the regulation of Arabidopsis seed dormancy we employed a comparative transcriptomics approach for 24 h imbibed dormant and after-ripened DOGNILs and Ler seeds. The same genotypes were earlier used to investigate the transcriptome of dormant dry seeds [6], which revealed that seed dormancy in the DOGNILs is mainly controlled by different additive genetic and molecular pathways. In dry seeds hardly any differences are found between dormant and after-ripened seeds, however as soon as seeds are being exposed to water, differences in the transcriptomes are evident. Based on these results we hypothesize that dormancy induction in the DOGNILs during seed maturation, for which dry seeds are the readout, is largely regulated by distinct molecular pathways, however dormancy maintenance during seed imbibition and the start of germination are likely very conserved processes. This conservation allowed us to identify a robust set of genes which are expressed at 24 h imbibed dormant and AR seeds. The genes identified depend a lot on the time-points chosen. From our earlier work we know that already at early imbibition (3 h after the start of imbibition) the first differences between dormant and after-ripened seeds can be identified [17]. However, we also know that most of changes in gene expression are related to seed rehydration itself and that those changes are similar between dormant and after-ripened seeds. To exclusively identify differences that are related to dormancy maintenance and germination we have chosen to investigate the transcriptome at 24 h after imbibition. This robustness of the identified genes was confirmed by comparison with previously published expression analysis that were performed with seeds of the Cvi accession at a range of physiological states [10, 17]. Of the dormancy and AR-up genes 63% and 32%, respectively, overlapped with genes that were identified by Cadman et al. [10] (Additional file 1: Table S1). Moreover, the D-up genes are also clearly higher expressed in dormant Cvi seeds as compared to AR Cvi seeds, as well as that the AR-up genes are on average higher expressed in AR seeds when compared to dormant seeds (Additional file 2: Figure S2). Some of the identified genes had previously been shown to play a role during germination and/or priming in several plant species. Among these are GA3OX2 (AT1G80340) a key gene in the gibberellin biosynthesis pathway, PIF6, involved in the phytochrome signalling pathway and ASG5 which is involved in protein and amino acid phosphorylation. The identification of these known dormancy mutants was the incentive to investigate the other dormancy and after-ripening specific genes. We took a reverse genetics approach by using T-DNA insertion lines for the differentially expressed genes and, indeed, we identified genes that had not been related to seed dormancy or germination before. Out of our target list of 66 genes, eight do not currently have any confirmed knock-out line available. This is consistent with a recent report that 12% of Arabidopsis genes do not have insertion lines available in previously generated collections [44]. The fact that a majority of the mutants showed near-wild type dormancy phenotypes, can be explained in several ways. The location of the T-DNA insertion may be decisive, e.g. whether in an intron, an exon or in untranscribed regions, such as promoters. Also, T-DNA–induced mutations do not always result in highly effective mutagenesis. Insertion in the protein-coding region of a gene generates a knockout in 86% of the cases and only 41% of the cases if the insertion is in front of the start codon [56]. Furthermore, gene redundancy may mask any phenotypic difference in plants in which the expression of only one homologue is disrupted [28]. In addition, in our experiments we used mutants with a Columbia-0 background that normally has low dormancy, which consequently does not allow the visualisation of small effects towards a decreasing dormancy level. Seed dormancy can be regulated by either inhibitory or promoting gene expression, considering the fact that in D-up genes we found both mutants that are less dormant (Fig. 3; KO11, 14 and 17) and more dormant (KO5, 16, 19, 23, 25 and 30). These examples demonstrate the inability to predict phenotypes based on expression pattern alone.

Many seed performance characteristics (i.e. seed desiccation tolerance, seed longevity and seed dormancy) are acquired during seed maturation. If genes affect seed maturation in general, it is likely that pleiotropic effects occur. In our study, some of the mutants showed a phenotype for more than one germination trait. The aah mutant for example displayed a phenotype for all investigated seed traits, except longevity. This enzyme degrades allantoate which is required to recycle purine-ring nitrogen in plants. The aah T-DNA mutant is unable to grow on allantoin as sole nitrogen source [57]. Furthermore, it is well known that conditions favouring nitrate accumulation in mother plant may lead to lower seed dormancy levels [1]. Since AAH is a key gene in the purine pathway [57], we speculate that defects in this gene block the pathway and, hence, availability of ammonium, resulting in increased primary dormancy and also affecting other seed performance traits of the mutant. Atnadh-me1, asg5 and ufp mutants affected both dormancy and longevity and one additional trait. It is known that seed longevity can be a pleotropic effect of genes that regulate other traits, such as seed maturation [53], response to temperature [37], oxidative stress [14] and dormancy [7, 39]. Previous studies, using mutant analysis, have shown that the seed dormancy mutants dog1 and rdo4 also have a reduced seed longevity phenotype [7, 39].

Loss of dormancy is expressed as opening of the germination window (permissive range of environments) [19]. It is because of this that germination under stress (i.e. salt or osmotic stress) often correlates with initial seed dormancy levels. We revealed two cases for which reduced dormancy indeed coincided with reduced sensitivity to salt stress (upf (KO20) and puf (KO41)). For some mutants, nbrsfp (KO1), upf (KO20) and aah (KO30), germination patterns on both NaCl and mannitol correlated positively, probably because both treatments confer osmotic stress. Two of these mutants (nbrsfp and aah) displayed enhanced sensitivity to salinity and osmotic stress. The NAD(P)-binding Rossmann-fold superfamily protein has oxidoreductase activity, binding, catalytic activity and, based on TAIR annotation, it is located in the endomembrane system. Upf was the only mutant to be clearly more insensitive to both salt and osmotic stress, which indicates that this mutant is primarily osmotolerant. Furthermore, for some of the salt-tolerant mutants RmlC-like cupins super family protein (KO32), unknown protein KO41 and axy8 (KO47) the germination rates on mannitol were similar to that of wild type. Likely, for the salt-tolerant lines, genes were mutated whose products are elements of stress signalling and inhibit germination under conditions of saline stress. The salt sensitive glx2 mutant was more tolerant to the application of ABA. The glyoxalase pathway consists of the two enzymes GLX1 and GLX2 and has a vital role in chemical detoxification. In Arabidopsis thaliana, GLX2 is required during abiotic stress, as was concluded from the higher sensitive of glx2-1to salt stress and anoxia seeds compared to wild type seeds. Moreover, GLX2-1-OE seeds are more resistant to anoxic stress than wild type [18].

Interestingly, axy8 (KO47) showed both a higher tolerance to ABA and to salt. AXY8 encodes an α-fucosidase acting on hemicellulose xyloglucan (XyG) that occurs in the primary cell wall of all vascular plants. Due to its high levels in elongating tissues [11] and structural alterations during cell elongation [46], XyG has been proposed to be a major player in extension growth [15]. This was confirmed by the induction of genes involved in XyG metabolism during cell elongation and upon the addition of the growth hormone auxin [50]. Overall these findings emphasize the importance of cell wall remodelling in the germination process, especially in response to stress conditions.

In both the dormancy and after-ripened sets we identified many genes encoding enzymes. This result might be linked to the fact that we looked at 24-h imbibed seeds. At this stage most of the cells in the embryo are potentially metabolically active. This also activates hydrolytic and synthetic enzymes and growth hormones to mobilize nutrients and synthesize ingredients for growth. These include the genes encoding ABA- and GA-biosynthesis- and -deactivation enzymes that play critical roles in determining the ABA-GA balance in seeds, and hence, dormancy and germination [42, 58].

Among the identified genes are also nucleotide binding proteins and transcription factors, such as members of the heat stress transcription factor family (HSFA1E, KO15), members of the ERF transcription factor family, transcription factor-related protein (KO5), TRANSCRIPTION FACTOR HOMOLOGOUS TO ABI5 and PIF6 (KO19). Interestingly all are found in the D-up state and for the ones that mutants were analysed they showed either more dormancy (transcription factor-related protein and pif6) or more sensitivity to ABA (hsfa1a).

We identified seed dormancy and germination phenotypes for genes that had not been associated with seed dormancy before. We tested only one T-DNA allele per line which may not be a definitive prove that the insertional mutation causes the observed phenotype, as many as 50% of the lines may contain additional inserts at unknown loci [22, 44]. We identified germination related phenotypes for nearly 50% of the investigated genes, which is far higher compared to what can be expected from a random selection of genes. It is also far higher compared to the genotypes that we identified for genes identified in earlier transcriptome analyses. Nevertheless, it is possible that a second locus may cause the phenotype of interest, or may alter the phenotypic effect of a knockout mutation. This work therefor represented an inventory of genes that are likely involved in the control of seed dormancy or germination. However, in depth studies are required to reveal the molecular mechanism by which these genes affect these important seed traits.