Based on our previous work dealing with the involvement of ROS in seed physiology, we can provide an overview of the changes in ROS content throughout all the stages of the sunflower (Helianthus annuus) seed life, as depicted in Fig. 1 for hydrogen peroxide content. Compared to an initial value in the range of the nanomole/g DW found in dry mature dormant seeds, H₂O₂ content varies dramatically depending on the physiological state of the seed. This content is quite high at the beginning of seed development, probably because the moisture content is high enough to allow metabolic activities, but subsequently decreases during seed desiccation. Also, seed imbibition entails a large increase in H₂O₂ content. The basal production of hydrogen peroxide during seed imbibition, which also occurs in imbibed dormant seeds, mainly results from metabolism resumption as approx. 2–3% of the oxygen used by the mitochondria is converted into superoxide anion and hydrogen peroxide [7]. However, the ability of seeds to germinate seems to be linked to the accumulation of a critical level of H₂O₂, since this molecule is produced at a higher level in non-dormant imbibed seeds than in dormant imbibed seeds (Fig. 1). This suggests that there exists a differential regulation of ROS producing and scavenging mechanisms in these two types of seeds. Interestingly, H₂O₂ also accumulates during dry storage of seeds, which is, in a first step, associated with dormancy release during after-ripening [7], and then, in a second step, with aging during prolonged storage [8]. Therefore, hydrogen peroxide can in turn be either beneficial (alleviation of dormancy) or detrimental (aging) for germination, depending on its accumulation level within the embryonic cells.

Besides documenting the variability of ROS production at the various stages of seed life, our data also highlight that sources of ROS fluctuate substantially from the beginning of embryogenesis to the end of germination. Such a variability of the ROS production sources is related to the molecular mobility and to the cytoplasmic viscosity, which are known to govern the occurrence and rate of metabolic reactions [9]. They both vary dramatically in seed tissues from a glassy state that prevails in the dry tissues to a fully imbibed state with cells containing free water. In the dry state enzymes are probably not active and in this case ROS probably originate from non-enzymatic reactions such as lipid peroxidation [10] or Amadori and Maillard reactions [11]. In contrast, in hydrated seeds, potentially all the cell compartments may become sources of ROS. Thus, these molecules can be produced during the catabolism of lipids (glyoxysomes) and purines (peroxisomes), respiratory activity (mitochondria), electron transfer in photosystems (chloroplasts) or through the activity of NADPH oxidase (plasma membrane), amine oxidase and peroxidase (cell wall) or cytochrome P450 (cytosol) [3,12]. One should emphasizes that the variability in ROS sources makes the study of the regulation of ROS accumulation quite complex, especially knowing that antioxidant mechanisms have also to be considered [13]. Finally, for being able to play a role in cell functioning, ROS much reach their cellular targets. As for their production, the changes in cytoplasmic viscosity would also play a key role in the signaling vs deleterious roles of ROS. In the dry state of seeds ROS would tend to reach their targets close to their production sites. In contrast, the presence of free water in hydrated seeds would allow the ROS, or at least their long living forms such as H₂O₂, to travel within the seed to reach targets far from their production sites. ROS produced in the dry state may by this way become effective as cellular messenger only when the seeds become hydrated, i.e. during germination.

A signaling role of ROS in seed physiology has recently been documented, notably in the regulation of germination and dormancy [6,13]. Increased generation of ROS seems to be a common feature of the so-called germination sensu stricto phase (phase II of the full process of germination), which is the critical step of the process since it involves the activation of a regulatory system controlled by intrinsic (i.e., dormancy) and extrinsic (i.e., environmental conditions, such as temperature, oxygen and water availability) factors. Hence, hydrogen peroxide [14–20], nitric oxide (NO) [21,22], hydroxyl radicals [15] and superoxide radicals [14,15] have been shown to accumulate during seed germination in various species. Many hypotheses have been put forward for attributing a function to these compounds in the germination process, including their contribution to cell wall loosening during endosperm weakening, programmed cell death of aleurone layer of cereal grains or protection of the emerging seedling against pathogens [6]. At the cellular level, ROS are also known to regulate the cellular redox status, to cause the oxidation of proteins and to trigger specific gene expression.

We propose here that ROS also play a key role in the completion of germination and that they should be considered as messengers or transmitters of environmental cues during seed germination. The success of germination tightly depends on external factors such as temperature, light, oxygen and water availability. This implies that seeds must be endowed with internal sensors able to translate the environmental cues into the cellular mechanisms leading to germination. In whole plants it has been widely demonstrated that abiotic stresses such as heat and cold stress, UV or hypoxia cause the production of ROS [23]. We assume that germination in non-optimal conditions is a stressful situation, associated with ROS generation that in turn would prevent radicle emergence. Such a role of ROS would therefore be at the interface between signaling and deleterious effect. For example, for soybean (Glycine max) seeds that are chilling sensitive and hardly germinate at temperatures below 10 °C, Pozmik et al. [24] observed that prolonged imbibition at 5 °C inhibits germination and leads to oxidative stress and lipid peroxidation. Similar observations were made for sunflower seeds during their germination at 5 °C, a suboptimal temperature for this species (Bailly, unpublished data). Also, incubation of seeds at too high temperatures induces an oxidative stress in numerous species. For example, inhibition of germination of sunflower seeds above 40 °C was associated with oxidative stress and lipid peroxidation [25]. In natural conditions, seed germination can also be prevented in the presence of various chemicals present in the soil and arising from root exudation, leaching and volatilization, and from passive decomposition of plant residues. These allelochemicals often belong to the group of secondary metabolites [26]. We have shown that the phytotoxic effect of sunflower allelochemicals on germination of mustard (Sinapis alba) seeds was associated with an increase in the cell membrane permeability, lipid peroxidation, accumulation of H₂O₂ and alteration in the activities of antioxidant enzymes [27]. Seed priming, which consists in a controlled imbibition preventing radicle protrusion, is an interesting tool for studying the mechanisms associated with germination sensu stricto. It can also be considered as a stress since water is not provided in a sufficient amount to allow radicle protrusion through the seed coat. In sunflower, seed priming is associated with stimulation in antioxidant enzyme activities, mainly catalase and superoxide dismutase, which is very similar to what has been found in whole plants challenged by a drought stress [28].

However, under appropriate conditions of germination, ROS must be present in sufficient levels for allowing completion of this process. This is for example illustrated by the use of an NADPH oxidase inhibitor, diphenyleneiodonium, which decreases the levels of superoxide ion and hydrogen peroxide during imbibition and prevents the germination of rice (Oryza sativa) [29] and of warm-season C4-grasses [22]. But this is also clearly highlighted from studies on the mechanisms of seed dormancy. These mechanisms involve complex regulatory networks of gene regulation that are under the control of plant hormones [30–32]. With regards to the data dealing with ROS metabolism during dormancy imposition and release, Bailly [13] proposed that ROS could be a ubiquitous signal involved in dormancy alleviation and that they could facilitate the shift from a dormant to a non-dormant status in seeds. In many species, release of dormancy may occur during storage of seeds in dry conditions (after-ripening). This process is very intriguing because in such conditions water is probably not available for biochemical reactions. Oracz et al. [7] and El-Maarouf-Bouteau et al. [33] provided evidence for the existence of a tight correlation between sunflower seed dormancy alleviation in the dry state, accumulation of ROS and peroxidation products and irreversible protein oxidation (carbonylation) in cells of embryonic axes. These data suggested that ROS play a role of signal in dormancy alleviation, presumably allowing a shift from a metabolism characteristic of dormant seeds to a metabolism characteristic of non-dormant (after ripened) seeds. As mentioned above, in dry seed tissues the mechanisms responsible for ROS production are probably non-enzymatic. Yet, enzymatic reactions cannot be totally excluded since enzymes such as lipoxygenase can be active when water activity is as low as 0.4, i.e., at MC below 0.1 g/g DW for many species [34]. Furthermore, the presence of localized hydrated pockets within the dry seed is likely to exist [35]. There are also several lines of evidence showing that ROS play a role during germination of sunflower seeds, as related to their dormancy level. Thus, non-dormant embryonic axes accumulate more ROS than dormant ones during their imbibition. For example, cyanide treatment that alleviates sunflower seed dormancy leads to ROS generation, while treatment in the presence of the ROS generating compound methylviologen provokes dormant seed germination [7,33]. Finally, exogenous H₂O₂ stimulates the germination of dormant seeds of barley (Hordeum vulgare) [36–38], rice [39], apple (Malus domestica) [40] and Zinnia elegans [41]. Additional evidence on the putative role of ROS in dormancy control comes from studies on the interaction of these compounds with plant hormones as ABA, GA, ethylene, auxin or brassinosteroids, all of which being well known for playing a major role in seed dormancy and germination [31,42]. ROS could play a key role in the transduction of the hormone signal since the interaction between hormones and ROS in other developmentally controlled processes in plants has been reported (for review see [43]). The relationship between H₂O₂ and ABA is the most credible and documented. Thus, H₂O₂ has been shown to inactivate ABI1 and ABI2 type 2C protein phosphatases, two enzymes involved in ABA signaling [44,45]. Furthermore, treatment of dormant barley seeds with hydrogen peroxide results in a decrease in endogenous ABA level [36,38] and alleviation of apple embryo dormancy by cyanide induces an increase in H₂O₂ level and a decrease in ABA content [40]. Interaction of ROS with ethylene and GAs, two hormones that are involved in dormancy alleviation of seeds of various species, has also been documented. It appears that ethylene and ROS transduction pathways share common components, GA₃ modifying the redox status of aleurone proteins during germination and ROS stimulating the expression of genes involved in GA signaling (for review see [6]). Altogether these data suggest that ROS act directly, or as messengers of hormonal networks, as signaling molecules involved in the transition from a dormant to a non-dormant state. However, the putative mechanisms of action of ROS at the cellular level in the regulation of dormancy are far from being resolved. ROS have been shown to trigger protein oxidation during dormancy alleviation [7], but they can also control MAP kinase cascade activation, inhibit phosphatases, activate Ca²⁺ channels and Ca²⁺-binding proteins, modulate redox potential and gene expression.

The deleterious role of ROS in seeds is due to their high reactivity toward biomolecules, including proteins, sugars, lipids and nucleic acids. The shift from a signaling to a deleterious role is related to the accumulation of these ROS above a threshold level that leads to various cellular alterations and damage. Two physiological processes occurring in seeds appear as being tightly linked to the deleterious role of ROS, seed desiccation and seed aging.

Dehydration to a MC around 0.04–0.12 g H₂O g⁻¹ DW is the normal terminal event in the development of orthodox seeds, i.e., seeds that tolerate desiccation. In fact, the desiccation tolerance of these seeds, which is generally acquired during seed filling and lost during seed germination (after radicle protrusion), depends on their physiological state [46,47]. In contrast, other seeds named recalcitrant have a high MC and do not withstand intensive dehydration [46,47]. Oxidative processes and free radicals are usually considered to be involved in molecular and cellular damage induced by a wide range of stresses including dehydration [48,49]. In desiccation-sensitive organs such as recalcitrant seeds [50] and seedlings of maize (Zea mays) [51] and wheat (Triticum spp.) [52,53], water loss is associated with an accumulation of free radicals and lipid peroxidation, and substantial damage is then evident at the level of cell structure [47,49,54]. The free radical damage hypothesis of desiccation injury postulates that various protective mechanisms become progressively unable to protect the cell structures against ROS generation during water loss and subsequent rehydration [49]. In wheat seedlings for example the activity of the antioxidant enzymes, glutathione reductase (GR), superoxide dismutase (SOD) and catalase (CAT), increases during dehydration in both tolerant and sensitive tissues, but these activities are sustained on rehydration only in the tolerant tissues [52,53]. During development of orthodox seeds, acquisition of desiccation tolerance is associated with various biochemical, cellular and molecular events including synthesis of specific proteins, accumulation of oligosaccharides, and activation of antioxidant defense systems [46,47]. Among these mechanisms, the antioxidant systems seem to play a key role. During maturation-drying for example the embryo H₂O₂ content decreases in sunflower [55] or remains very low in wheat [56] probably due to a reduction of mitochondrial activity, the main source of H₂O₂ generation, and to an antioxidant defense system sufficiently efficient to control ROS accumulation during water loss. In addition, in sunflower, our results suggest that H₂O₂ may also play a role in the regulation of CAT gene expression and transduction pathway of the dehydration signal, probably via a mechanism close to that described by Polidoros and Scandalios [57], i.e., through ARE-like motifs. In bean (Phaseolus vulgaris), acquisition of drying tolerance is also related to a reorientation of the enzymatic antioxidant defense system, drying tolerant dried seeds displaying high CAT and GR activities and low SOD and APX (ascorbate peroxidase) activities while the opposite is observed in immature dried seeds [58].

ROS are also suspected to be involved in loss of seed viability since a long time [10,59] and the free radical hypothesis of aging remains popular and widely acknowledged. This indeed makes sense and the bases of seed aging are very similar to those found in other organisms. The theory postulates that during storage loss of viability results from a concomitant accumulation of ROS, associated damage, and a loss of the antioxidant mechanisms, thus causing uncontrolled oxidative damage during seed imbibition. This assumption has been supported by many studies carried out with a wide range of seed species. For example, in sunflower Bailly et al. [17,60,61] have shown that a loss of vigor and viability was mainly associated with a loss of catalase activity and therefore with an altered capacity of hydrogen peroxide detoxification in the early steps of seed imbibition leading to lipid peroxidation. However, for some other species, it appears, as for example for wheat [62], that loss of viability is not always associated with lipid peroxidation, either directed against lipid membranes or lipid reserves. In sunflower seeds, we have shown that the mechanism of loss of viability during aging depends on the seed moisture content. Below a threshold value of ca. 0.21 g H₂O g DW⁻¹ hydrogen peroxide accumulated rapidly within the seeds placed at 35 °C, but without causing lipid peroxidation. This suggests that, in this case, oxidative damage are directed toward the non-lipidic cellular fraction, i.e., presumably proteins or nucleic acids. At the opposite, when seed MC is higher than the threshold value mentioned above, loss of seed viability was clearly associated with lipid peroxidation. These data allow pointing out the plasticity of the mechanisms involved in seed loss viability. They may in part explain the controversy about the role of the oxidative processes in seed aging. It is likely that targets of ROS may vary as a function of seed MC, and we do recommend paying attention to the experimental conditions used to assess the mechanisms of seed aging.

Our knowledge about the roles of ROS in seed physiology allows us to present a comprehensive view of their mechanism of action, which brings a new view on the data presented in Fig. 1. We propose here the concept of an “oxidative window” for germination. A working scheme illustrating this concept is presented in Fig. 2. This window defines a critical level of ROS not to overcome, which would otherwise prevent germination, and a ROS threshold level, below which germination (radicle protrusion) cannot occur. Within this oxidative window ROS may play a role in cell signaling by interplaying with the hormone signaling pathways or by triggering the cellular events associated to ROS signalization, such as gene expression, calcium movements or control of redox status. In inappropriate environmental conditions or consequently to seed aging the amount of ROS in seed tissues exceeds the upper limit of the oxidative window, which leads to oxidative damage and inhibits germination or leads to abnormal seedlings. At the opposite, cellular signaling associated to the mechanism of seed germination is not triggered if the amount of ROS is too low, i.e., as in dormant seeds. Future challenge of studies dealing with ROS in seeds will have to take into account the diversity of the roles of these compounds. It will also be of particular interest to properly document the putative sources of these compounds, to identify their cellular targets and to determine if they are the signal linking environmental cues to hormone signalization.