A snapshot view on the ethylene biosynthesis and signaling pathway in Arabidopsis
Autocontrol of ethylene biosynthesis
Mechanistic control of ethylene signal components
The auxin–ethylene circle
Tissue- and cell-type-specific regulation of ethylene
Cellular basis of ethylene effects on growth
The vegetative development of plants is strongly dependent on the action of phytohormones. For over a century, the effects of ethylene on plants have been studied, illustrating the profound impact of this gaseous hormone on plant growth, development and stress responses. Ethylene signaling is under tight self-control at various levels. Feedback regulation occurs on both biosynthesis and signaling. For its role in developmental processes, ethylene has a close and reciprocal relation with auxin, another major determinant of plant architecture. Here, we discuss, in view of novel findings mainly in the reference plant Arabidopsis, how ethylene is distributed and perceived throughout the plant at the organ, tissue and cellular levels, and reflect on how plants benefit from the complex interaction of ethylene and auxin, determining their shape. Furthermore, we elaborate on the implications of recent discoveries on the control of ethylene signaling.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Ethylene is a gaseous plant hormone that affects multiple aspects of plant development, although it is best known as the ripening hormone (Abeles et al., 1992). Physiological effects are detectable at ambient levels as low as 0.1 μl l−1. The first discovery of ethylene effects dates back over a century (Neljubow, 1901). Neljubow demonstrated that ethylene causes horizontal growth of pea seedlings, inhibition of elongation and radial swelling (‘triple response’). Conclusive evidence that ethylene is produced by plants was presented by Gane in 1934. Although playing a primary role in ripening, abscission and senescence, many aspects of vegetative growth are also influenced by ethylene (Smalle & Van Der Straeten, 1997). First, it has a positive effect on the germination of several plant species. Furthermore, it influences growth in both darkness and light. In ethylene-treated etiolated Arabidopsis seedlings, the ‘triple response’ is seen as an inhibition of elongation and radial swelling of the hypocotyl, and an exaggerated apical hook. In general, primary and lateral root elongation is inhibited, whereas root hair development is stimulated. By contrast, in hypocotyls of light-grown seedlings, as in stems of some semi-aquatic species, and during shade avoidance, ethylene can stimulate elongation. Although leaf emergence is enhanced by ethylene, leaf expansion is mostly inhibited. Ethylene production is tightly regulated by internal signals during development and in response to environmental stimuli, both biotic and abiotic.
The chemical properties of ethylene are crucial to an understanding of its biology (Abeles et al., 1992). The diffusion coefficient in air is c. 10 000 times that in water. Furthermore, it is c. 14 times more soluble in lipids than in water. Like other olefins, ethylene binds to metals such as Cu(I). Copper ions serve as a cofactor in high-affinity ethylene binding to its receptors.
II. A snapshot view on the ethylene biosynthesis and signaling pathway in Arabidopsis
The biological precursor of the gaseous hormone ethylene is methionine, which is converted to S-adenosylmethionine (SAM) by SAM synthetase. SAM is the substrate of 1-aminocyclopropane-1-carboxylic acid (ACC) synthase (ACS), which forms ACC, the direct precursor of ethylene. ACC is further oxidized by ACC oxidase (ACO) to ethylene, with CO2 and cyanide as by-products. In most of the species investigated, including Arabidopsis, ACS and ACO are members of large and small multigene families, respectively (De Paepe & Van Der Straeten, 2005).
The ethylene signaling pathway has been elucidated mainly by genetic analysis (Fig. 1). In Arabidopsis, ethylene is perceived by a family of five receptors that are located at the Golgi and endoplasmic reticulum (ER) membranes (Dong et al., 2010). The receptors are negative regulators of the signaling pathway, inactivated on ethylene binding. They have been shown to physically interact with CONSTITUTIVE TRIPLE RESPONSE 1 (CTR1), a Raf kinase homolog (Clark et al., 1998; Gao et al., 2003), and with ETHYLENE INSENSITIVE 2 (EIN2), an N-RAMP metal transporter-like protein (Bisson et al., 2009) (Fig. 1). Genetic analyses have placed EIN2 downstream of CTR1, which itself works downstream of the receptors. In contrast with CTR1, EIN2 is an ER-anchored protein (Bisson et al., 2009). Ctr1 and ein2, both loss-of-function mutants in single gene families, have opposite phenotypes, identifying them as negative and positive regulators of ethylene signaling, respectively. In vivo interaction studies have shown that, in the absence of ethylene, CTR1 is tightly associated with the receptor complex (Clark et al., 1998), blocking signal transfer to EIN2 (Fig. 1). This may occur by the prevention of receptor interaction with EIN2 (Bisson & Groth, 2010). When ethylene binds to the ETHYLENE RESISTANT 1 (ETR1) receptor, CTR1 may change conformation and allow the interaction of EIN2 with the kinase domain of ETR1. Thus, the receptor, CTR1 and EIN2 work as constituents of a multicomponent high-molecular-weight complex, detected on purification of receptors (Chen et al., 2010). ETR1 has also been shown to bind with a regulator, REVERSION TO ETHYLENE SENSITIVITY (RTE1) (Rivarola et al., 2009; Dong et al., 2010).
The signal is further passed on to the nuclear components, in a mechanistically, as yet unidentified, manner. EIN3 and EIN3-like (EIL) transcription factors are the first players in a dual transcriptional cascade, initiating the expression of ethylene response factors (ERFs), which themselves function as ethylene responsive element binding proteins (EREBPs) that stimulate the transcription of target genes (Fig. 1).
Plants have developed mechanisms to control and fine tune hormone production. These mechanisms can be inhibitory, in order to prevent the accumulation of supra-optimal levels, or stimulatory, to enhance or speed up a response. How does a plant decide to proceed along either one of these options? Recent research has shown that distinct mechanisms regulate a different subset of ethylene biosynthesis genes, typical for either feedback inhibition or autocatalytic effects.
Enhanced hormone signaling often leads to a reduction in biosynthesis, termed auto-inhibition or negative feedback. Conversely, mutations resulting in hormone insensitivity lead to a positive feedback on biosynthesis. This is the case for gibberellins (GAs) (Martin et al., 1996), brassinosteroids (BRs) (Li et al., 2001) and cytokinins (Glover et al., 2008). Likewise, ethylene insensitive mutants of the ethylene receptors or EIN2 typically show high levels of ethylene production, whereas ctr1-1 mutants exhibit lower ethylene production levels than wild-type plants (Thain et al., 2004). However, ein3eil1 double mutants, despite having strong insensitive phenotypes similar to ein2 mutants (An et al., 2010), do not overproduce ethylene when grown in long days (Table 1). By contrast, when grown in continuous darkness, they produce more ethylene than the wild-type; yet, the detected levels are below those measured in etr1 and ein2 (Table 1). This suggests that a branch parallel to the EIN3/EIL1 transcriptional cascade controls the negative feedback, and that this mechanism is dependent on light conditions. Keeping ethylene levels low is especially important during vegetative development, in order to allow an optimal expansion of the root and the shoot – processes that are strongly dependent on the light regime.
Table 1. Ethylene production in Arabidopsis thaliana wild-type Columbia plants and in ethylene insensitive mutants grown for 7 d on half-strength Murashige–Skoog medium in a photoperiod of 16 h light : 8 h darkness or in continuous darkness
Average ethylene production
Average ethylene production
Light-grown plants (pl per plant per h)
Dark-grown plants (pl per plant per h)
Accumulation was performed for 24 h starting on day 7. Measurements were performed as described in Ellison et al. (2011). Data presented are from at least three independent biological repeats (sealed container with c. 100 seedlings). Data are average ± SD. The ethylene production rate is expressed as pl ethylene produced per plant per hour.
1.3 ± 0.4
0.3 ± 0.1
6.9 ± 0.8
2.6 ± 0.5
7.6 ± 0.2
3.4 ± 0.8
1.9 ± 0.7
0.4 ± 0.1
0.9 ± 0.2
0.3 ± 0,1
1.9 ± 0.3
1.0 ± 0.1
2.3 ± 0.5
0.5 ± 0.1
Several ACC synthases coordinately generate the basal ethylene levels in Arabidopsis (Tsuchisaka et al., 2009). Octuple loss-of-function mutants in ACC synthases have levels of ethylene production 10 times lower than those of the wild-type (Tsuchisaka et al., 2009). In higher order (pentuple and hexuple) mutants, ACS5 and ACS7 are induced and are thus probably involved in the feedback regulation of ethylene production (Fig. 2). In addition, ACS8 is under negative control of ethylene signaling, as transcripts of this isozyme accumulate in ein2-1 mutants (Thain et al., 2004). Interestingly, ACS5, ACS7 and ACS8 belong to the phylogenetic group B (type II) of ACC synthases, and can homo- or heterodimerize (Tsuchisaka & Theologis, 2004a). Homodimerization could contribute to the autoregulation. For instance, ACS8 homodimers are highly active (Tsuchisaka & Theologis, 2004a), but their abundance is expected to be relatively reduced in the presence of high ethylene levels (Thain et al., 2004) (Fig. 2). Hence, less active heterodimers may form and reduce ethylene synthesis. In addition, the type II ACC synthases (ACS4,5,9) are also under specific post-translational control by ETHYLENE OVERPRODUCER 1 (ETO1) and ETO-like (EOL) proteins, which mark them for proteasome-mediated degradation (Christians et al., 2009). Feedback inhibition of ethylene signaling through the up-regulation of ETO1 activity is therefore an additional possible level of control (Fig. 2).
Autocatalytic or feedforward effects on ethylene production are associated with stress situations (Vriezen et al., 2003; An et al., 2010) and developmental processes, such as ripening, shedding, wilting and senescence. At the molecular level, the stimulation of ethylene production can be achieved through the elevation of transcript levels or the stabilization of key biosynthetic enzymes. The control of transcript levels has been observed during tomato fruit ripening, where specific ACS and ACO genes are induced at different stages of ripening. For instance, LeACS4 is the main enzyme associated with autocatalytic ethylene production (Cara & Giovannoni, 2008). In Arabidopsis, ACS6 is transcriptionally repressed in higher order ACS mutants, which could relate to feedforward regulation (Tsuchisaka et al., 2009). Indeed, ACS6 is especially expressed in stress situations, in which autocatalysis is important. Recent findings have unveiled part of the mechanism of ethylene production under stress. Mitogen-activated protein (MAP) kinases MPK3 and MPK6 are associated with a variety of stresses. They are under the control of other MAPK kinases, such as MKK2, MKK3, MKK4, MKK5, MKK7 and MKK9 (Menke et al., 2004; Takahashi et al., 2007; Xu J et al., 2008; Yoo et al., 2008; Cho et al., 2009; Beckers et al., 2009) (Fig. 2). MPK3 and MPK6 enhance the stability of ACC synthases (Joo et al., 2008; Fig. 2), a function mainly associated with pathogen infection (Han et al., 2010) and senescence (Zhou et al., 2009). Again, a phylogenetic distinction can be seen. MPK3 and MPK6 cause stabilization of the type I (class A) ACC synthases ACS2 and ACS6, which are mainly responsible for Botrytis-induced ethylene production, whereas type II (class B) ACS5 and ACS9 are not involved in Botrytis-induced ethylene production (Han et al., 2010). However, although ACS6 is strongly associated with leaf senescence, it is not the only ACS involved. Indeed, an octuple mutant combination of ACS, resulting in minute ethylene production, is necessary to inhibit plant senescence, whereas lower order mutants exhibit wild-type senescence (Tsuchisaka et al., 2009). This is reminiscent of the suppression of tomato fruit ripening by antisense RNA, which requires almost complete inhibition of ethylene production (Oeller et al., 1991).
Although there is agreement on the role of MKK9 and MPKs in the control of ethylene biosynthesis (Joo et al., 2008; Xu J et al., 2008; An et al., 2010), their involvement in ethylene signaling has been subject to debate (Yoo et al., 2008; Hahn & Harter, 2009; An et al., 2010; Zhao & Guo, 2011). Studies based on the analysis of mkk9 loss-of-function mutants and the overexpression of MKK9 locked in an active state (MKK9a) yielded opposite results (Yoo et al., 2008; An et al., 2010). Mkk9-1 seedlings are less sensitive than the wild-type to lower ACC concentrations, whereas mkk9-5 seedlings are not. A constitutively active MKK9a was capable of generating short hypocotyls in an etr1 mutant background, whereas this was not the case in an ein2 background (Yoo et al., 2008). However, in a study using a dexamethasone inducible system driving MKK9a, the phenotype in the etr1 background could not be confirmed, suggesting that MKK9 is not an essential part of ethylene signaling (An et al., 2010). Nevertheless, it is striking that MKK9 reacts to ethylene by relocation to the nucleus (Yoo et al., 2008). In view of its involvement in leaf senescence (Zhou et al., 2009), this relocation may be one of the initial events that controls autocatalytic ethylene production during senescence.
In summary, negative feedback is particularly active during vegetative growth, whereas autocatalytic ethylene production is stress related. They rely on the regulation of a different subset of ACC synthases, which, in the case of autocatalysis, is controlled by a MAPK cascade.
IV. Mechanistic control of ethylene signal components
1. At the transcript level
Of the family of five receptors in Arabidopsis, three are inducible by ethylene within 2 h of treatment (ETHYLENE RESPONSE SENSOR 1 (ERS1), ERS2 and ETR2; Binder et al., 2004b). This rapid auto-inducibility has a clear biological function. When ethylene is present, it can bind the receptors and switch them to an inactive state, relieving the repression on the pathway, thus leading to growth inhibition. When ethylene is removed, the growth rate recovers more rapidly than can be reconciled with the dissociation time of the receptors. Hence, it is hypothesized that, once exogenous ethylene has been removed or the concentration lowered below threshold values, the de novo synthesis of receptors in the active state (devoid of ethylene) contributes to rapid growth recovery by suppressing ethylene signaling. The newly synthesized receptors are thought to interfere with higher order ethylene-bound inactive receptor complexes, and consequently allow the restoration of receptor activity which negatively regulates the ethylene pathway, a mechanism of re-sensitization. Given the velocity of induction after ethylene treatment found for ERS1, ERS2 and ETR2, it is probable that they are involved in primary responses. Other primary response genes include RTE1, EBF2 and the biosynthesis gene ACO2 (Raz & Ecker, 1999; Potuschak et al., 2003; De Paepe et al., 2005; Resnick et al., 2006). Some of these genes have been used as general reporters for the ethylene response, but it should be kept in mind that they may show tissue or cell-type specificity of expression, and thus may not always be suitable. In support of this hypothesis, an in silico analysis of ethylene-related gene expression in roots indicated that different cell types have different transcript levels of signaling components (Dugardeyn et al., 2008). In contrast with transiently induced primary response genes, ERS1, ERS2 and ACO2 remain induced after 24 h of exposure, which could allow plants to adapt to the ethylene signal, and reach a persisting ‘on’ state (Zhong & Burns, 2003). In theory, the transcript level of these genes in a wild-type, untreated root could thus represent the ethylene signaling state in a particular cell type. As mentioned before, the caveat of such an interpretation is that these genes can also be under the control of an expression program determined by cell identity (independent of ethylene) or, in untreated seedlings, the threshold necessary for their induction may not be reached. An extensive database of microarray-based transcript levels in fluorescence assisted cell sorting (FACS)-sorted cell types from untreated Arabidopsis roots has been developed (Brady et al., 2007). A prediction of the ethylene signal, based on the analysis of expression patterns in the root tip, is presented in Fig. 3. A primary requirement is the presence of the unique ethylene signaling stimulating component EIN2. This is the case for many cell types, but the highest level appears in tier1 of the columella (referred to as columella 1 in the database). Transcripts of the downstream component EIN3 and the ethylene inducible genes ETR2, ERS2 and ACS6 are also highly represented in these cells compared with the surrounding tissues, suggesting a strong ethylene signal. Hence, the expression patterns in the first tier of columella cells may reflect a strong ethylene signaling state.
As EIN3-type transcription factors have a central role in ethylene signaling, their targets are obviously ethylene regulated, albeit not uniquely controlled by ethylene. ETHYLENE RESPONSE FACTOR 1 (ERF1) is such a target gene of EIN3, belonging to the large family of ETHYLENE RESPONSIVE ELEMENT BINDING PROTEINS/APETALA2 (EREBP/AP2) domain genes. However, only c. 10% of this family appears to be ethylene regulated at the transcript level, and only a few have been shown to be involved in ethylene-regulated growth control, some with a positive and others with a negative regulatory function (Alonso et al., 2003; Nakano et al., 2006; Hess et al., 2011; Yang et al., 2011). A small subfamily of four genes, called ETHYLENE DNA BINDING FACTOR (EDF1–4), plays an overlapping role in the ethylene response and is involved in the regulation of elongation growth. Although single edf mutants do not display any phenotype in a triple response assay, quadruple edf mutants are slightly insensitive to treatments with ACC, showing longer hypocotyls in these conditions. However, because of this partial insensitivity, the EDFs are thought to control only one branch of the EIN3/EIL1-regulated elongation growth (Alonso et al., 2003). Undoubtedly, among the plethora of targets, the variation in expression patterns of ERFs helps to regulate specific ethylene responses in given tissues or cell types.
2. At the post-transcriptional level
Regulation of receptor function The ethylene receptors require copper ions as a cofactor to bind ethylene (Rodriguez et al., 1999; Binder et al., 2010). RESPONSIVE TO ANTAGONIST 1 (RAN1) functions as a metal transporter that delivers copper to the receptors (Hirayama et al., 1999; Woeste & Kieber, 2000; Binder et al., 2010). These receptors repress ethylene signaling; the binding of ethylene leads to the inactivation of receptors and, consequently, derepression and activation of ethylene responses. Intuitively, one might assume that RAN1 will be necessary to generate the ethylene response. Yet, the fact that the strong loss-of-function allele ran1–3 has a constitutive triple response phenotype (Woeste & Kieber, 2000) suggests that the copper cofactor is also important for the conformational characteristics of the receptor in the absence of ethylene. Indeed, exogenous addition of copper can partially rescue the constitutive ethylene response phenotype of ran1–3 mutants. The amount of copper-containing receptors will thus determine how strong is the repression of the pathway. Treatment of seedlings with copper chelators induces a stronger constitutive ethylene response in weak ran1 mutants at low concentrations and in the wild-type at high concentrations; a response which is not visible in ethylene insensitive mutants (Binder et al., 2010). This suggests that the lack of copper in soils may trigger the ethylene response.
In addition to copper, the ETR1 binding protein RTE1 has been proposed to modulate the conformation of the ETR1 receptor protein (Resnick et al., 2008). Double mutant analysis using rte1, in combination with various dominant insensitive etr1 alleles, revealed the ethylene binding domain of ETR1 as a possible target for RTE1 function.
Regulation of the abundance of the positive response regulators EIN2 and EIN3 by the 26S proteasome The control of protein accumulation by the 26S proteasome has been demonstrated for many components in plant hormone signaling (Santner & Estelle, 2009). For the ethylene signaling pathway, two crucial steps are known to be strongly regulated by ubiquitin-mediated degradation: the membrane-associated EIN2 and the transcription factors EIN3 and EIL1. The stability of EIN2 is regulated by the F-box proteins ETP1 and ETP2, the levels of which are down-regulated by ethylene (Qiao et al., 2009). As a consequence, the accumulation of the EIN2 protein is sustained by ethylene.
The level of EIN3 is regulated by E3 ligases carrying the F-box proteins EBF1 or EBF2. It has been demonstrated that ethylene causes degradation of the EBF proteins, through an as yet uncharacterized E3 ligase in an EIN2-dependent, but EIN3/EIL1-independent, manner (An et al., 2010). This removal of EBF proteins consequently leads to an accumulation of EIN3 protein, which stimulates downstream ethylene responses. Subsequently, EBF2 transcription is induced as a direct target of EIN3 (De Paepe et al., 2005; Konishi & Yanagisawa, 2008), resulting in dampening of the ethylene signal, allowing the plant to resume normal growth. This is presumably by the breakdown of EIN3, as EIN3-overexpressing plants show the same defect of delayed growth recovery as ebf2 mutants (Binder et al., 2007) (Fig. 4). By contrast, EBF1 transcript does not accumulate on ethylene treatment (De Paepe et al., 2004). Hence, EBF2 is the only control point that attenuates the ethylene response as part of a positive–negative transcriptional–post-transcriptional feedback loop driven by ethylene (Fig. 4). Ebf1 mutants have a slightly stunted phenotype in the absence of or in low levels (< 0.025 ppm) of ethylene, whereas ebf2 mutants display ethylene hypersensitivity only at higher hormone concentrations (Guo & Ecker, 2003; Gagne et al., 2004). The oversensitivity of ebf2 loss-of-function mutants leads to extreme dwarfism in seedlings treated with ethylene. This phenotype is strongly reminiscent of the enhanced ethylene response (eer) mutants. Although the EBF levels have not been checked in eer mutants, a link with the EBF substrate EIN3 has been demonstrated, as the TFIID interacting transcription factor EER4 interacts with EIN3 (Robles et al., 2007). As eer4 and ebf2 mutants share a phenotype of enhanced ethylene response, EER4 could thus be part of the feedback loop that steers EBF2 transcription (Fig. 4).
V. The auxin–ethylene circle
Auxin profoundly changes the shape of the plant, often in a very similar manner to ethylene (Dugardeyn & Van Der Straeten, 2008). However, the mechanisms by which these pathways influence each other await further elucidation.
1. Auxin–ethylene: reciprocal interactions in the root
Ethylene affects the root architecture by regulating bending, lateral root initiation, hair formation and elongation (Smalle & Van Der Straeten, 1997). The inhibition of primary root growth by ethylene is partly a result of hormonal cross-talk. First, ethylene augments the stability of DELLA proteins, which negatively regulate the GA response (Achard et al., 2003). In addition, exogenous ethylene and auxin treatment both lead to a reduction in root elongation, whereas ethylene can cause the accumulation of auxin to levels supraoptimal for growth, in a highly entangled cooperative manner (Ruzicka et al., 2007). A vast number of observations have indicated interactions between these hormones. Panoramic gene expression analysis has revealed that many ethylene-regulated genes are dependent on the level of auxin, which promotes or attenuates the ethylene effect, and that expression changes can be triggered by ethylene-induced auxin activity (Stepanova et al., 2007). By screening for ethylene insensitive mutants, several auxin-related factors have been found: the auxin biosynthesis genes WEI2/ASA1 (WEAKLY ETHYLENE INSENSITIVE 2/ANTHRANILATE SYNTHASE ALFA1) and WEI7/ASB1 (WEAKLY ETHYLENE INSENSITIVE 7/ANTHRANILATE SYNTHASE BETA1), and the auxin transport genes EIR1/PIN2 (ETHYLENE INSENSITIVE ROOT1/PIN-FORMED2) and AUX1 (Roman et al., 1995; Ruzicka et al., 2007; Swarup et al., 2007). In addition, axr3-1/iaa7 auxin signaling mutants are insensitive to both auxin and ethylene. This underlines the high degree of auxin dependence of ethylene phenotypes. By contrast, ethylene insensitive mutants are auxin sensitive with respect to root elongation, albeit that most of the ACC synthases are auxin inducible (Tsuchisaka & Theologis, 2004b).This indicates that ethylene can work downstream of auxin, or vice versa, suggesting a regulatory loop (Fig. 5a). The expression of auxin biosynthesis genes WEI2/ASA1, WEI7/ASB1, WEI8/TRYPTOPHAN AMINOTRANSFERASE 1 (TAA1) and TRYPTOPHANAMINOTRANSFERASE RELATED 2 (TAR2) is stimulated by ethylene (Stepanova et al., 2005; He et al., 2011). Furthermore, ASA1 and ASB1 genes are highly expressed in tier1 of the columella of untreated plants (Fig. 3). Hence, ethylene-controlled expression of auxin biosynthesis is probably a way in which endogenous ethylene regulates root growth. Interestingly, tier1 columella cells also harbor the main determinants for lateral auxin distribution in Arabidopsis, the PIN3 and PIN4 proteins (Friml et al., 2002a,b). Hence, control of auxin levels by ethylene in these cells will determine how much auxin is directed towards the lateral root cap and eventually regulates the effect on cell inhibition in the elongation zone (Ruzicka et al., 2007). In addition, the auxin transporters AUX1 and PIN2 are up-regulated by ethylene in root tips (Stepanova et al., 2005; Ruzicka et al., 2007) (Fig. 5a). The last two proteins determine the auxin flow by stimulating basipetal auxin transport through the epidermis and lateral root cap towards the root elongation zone (Marchant et al., 1999; Ruzicka et al., 2007). Thus, ethylene regulates both the synthesis and distribution of auxins.
An approach in which cell-type-specific complementation leads to the suppression of insensitivity for ethylene-mediated root growth inhibition, caused by the aux1 mutation, led to the conclusion that AUX1 activity in the lateral root cap is necessary and sufficient to support the ability of ethylene to inhibit cell elongation (Swarup et al., 2007). AUX1 expression is increased in the presence of ethylene in the lateral root cap (Ruzicka et al., 2007). It is not clear whether ethylene regulates AUX1 directly and in the same cells, or whether the ethylene signal is perceived in other cell types and a response is generated that leads to the effects on the auxin transport machinery, for instance via increased levels of auxin. In this respect, it should be considered that anthranilate synthase and tryptophan aminotransferase genes are also up-regulated by ethylene in the root tip, in cell types other than the lateral root cap (Stepanova et al., 2005; He et al., 2011). However, if ethylene regulation of auxin transport in the lateral root cap is direct, this would imply that ethylene perception in these same cells is sufficient to generate the ethylene response. In support of this idea, the removal of root caps in maize eliminates the inhibition of elongation (Hahn et al., 2008). Alternatively, ethylene-induced differences in auxin metabolism (for instance, by up-regulating synthesis in the root tip (Ruzicka et al., 2007), or by enhancing the transport from shoot to root (Lewis et al., 2011), causes autoregulation of auxin transport towards the root elongation zone, by an as yet unidentified mechanism. In addition, dominant interference with the auxin signal by expression of the gain-of-function mutant axr3-1 gene suggests that, in addition to the epidermis, the cortex and endodermis also use the downstream auxin signal for ethylene-inhibited root growth. This implies that the function of ethylene may be limited to increasing auxin biosynthesis in the tip and enhancing its transport towards the lateral root cap (Swarup et al., 2007).
As stated above for elongation growth and tropic bending, auxin and ethylene act in a cooperative manner. However, during the control of lateral root initiation, these two hormones work as antagonists: auxin stimulates and ethylene reduces the formation of lateral root primordia (Lewis et al., 2011). Ethylene causes draining of auxin towards the apex by stimulating auxin transport through the action of the AUX1 and LAX3 influx carriers and PIN3 and PIN7 efflux regulators. This leaves the mature root zone with a smaller amount of auxin and, consequently, fewer lateral root primordia (Lewis et al., 2011).
In summary, the role of ethylene in the regulation of root architecture relies on the spatial regulation of auxin accumulation, which eventually leads to a reduction in root elongation and lateral root development.
2. Hypocotyl, stem and petiole elongation: is everything auxin as well?
As is the case for roots (Le et al., 2001), growth inhibition of hypocotyls by ethylene is a rapid response. Root cells stop elongating within 10 min, and etiolated seedlings are inhibited within 15 min after the start of exposure to ethylene, suggesting the presence of components essential for the generation of the ethylene response (Le et al., 2001; Binder et al., 2004b, 2007). Interestingly, very similar kinetics for auxin-controlled growth have been observed. This is concomitant with changes in global gene expression within 5–15 min of application (Abel et al., 1994) and in protein levels within 15–30 min after auxin application (Oeller & Theologis, 1995). Stimulatory effects of auxin on light-grown hypocotyls are visible 10–15 min after application (Christian & Luthen, 2000). Assuming that exogenous auxin requires transport to the location at which it affects growth, a similar mechanism by which ethylene regulates endogenous auxin distribution in roots may be present. Yet, the apparent similar velocity of response may preclude the interdependence of both hormonal pathways and suggest that effects can be attributed to either auxin or ethylene. Kinetic studies using different mutants and marker lines will be needed to shed light on this issue. However, the rapid growth response to auxin is not regulated by the TRANSPORT INHIBITOR RESISTANT 1/ AUXIN-RELATED F-BOX (TIR1/AFB) system (Schenck et al., 2010), which could make it difficult to find a suitable set of auxin mutants for the analysis of the rapid ethylene response.
In contrast with root growth inhibition, ethylene-induced reduction of hypocotyl extension in etiolated plants is biphasic, with a first rapid decrease in growth rate, followed by a first plateau and a slower phase. The first plateau does not occur in ebf1 single mutants (Binder et al., 2007). The first phase of growth inhibition is extremely sensitive to ethylene, down to concentrations as low as 1.5 ppb. However, during such low-dose exposure, the growth rate recovers to the nontreatment level, indicating adaptation. Moreover, pretreatment with low levels of ethylene desensitizes plants to higher levels of ethylene, as long as growth recovery from the initial treatment is not complete (Binder et al., 2004a). Desensitization has been attributed to the de novo synthesis of receptors and the signaling feedback through EBF2 (Binder et al., 2004a,b). At this point, it is not known which mechanism determines the switch between the dark-induced inhibition and light-induced stimulation of growth by ethylene. In this context, it may be of interest to determine the short-term growth responses in the light, as has been performed for etiolated seedling growth and hyponasty in Arabidopsis (Binder et al., 2004a; van Zanten et al., 2009). Light shows a stronger restriction than ethylene of hypocotyl growth, as etiolated ctr1 or wild-type hypocotyls are longer than those of light-grown wild-type seedlings (Vandenbussche et al., 2007). Furthermore, it has been shown that the elongation is dependent on an active GA pathway which is not a direct target for the ethylene signal output. Ethylene is expected to work as a parallel signal, possibly with auxins or BRs acting downstream (Vandenbussche et al., 2003a, 2007; De Grauwe et al., 2005). In this respect, the strong ethylene dependence of many of the responses to auxin may indicate the necessity for an auxin optimum regulated by ethylene. High auxin concentrations induced by ethylene may be supraoptimal for growth, in hypocotyls as well as in roots (Ruzicka et al., 2007; Vandenbussche et al., 2010; Zadnikova et al., 2010).
The regulation of differential growth in shoots, in the majority of cases, although being centered around auxin, often involves ethylene as well (Guo et al., 2008; Chaabouni et al., 2009). A well-documented example is the regulation of apical hook development of young seedlings grown in darkness. This is a case of auxin-controlled differential growth, enhanced by ethylene. As mentioned above, the triple response in etiolated seedlings, comprising stunted growth, hypocotyl/stem thickening and loss of geotropism, is visible in Arabidopsis as an exaggerated apical hook. These ethylene-regulated developmental changes are indispensable for seedling survival. The apical hook is necessary to protect the apical meristem and cotyledons when the seedling pushes through mulch or soil (Darwin & Darwin, 1881). The importance of ethylene in this process is clear from ethylene insensitive mutants which experience difficulties emerging from soil (Harpham et al., 1991). Endogenous ethylene may thus serve as a signal for thigmomorphogenesis to preserve the apical hook structure.
During apical hook development in Arabidopsis, ethylene controls auxin distribution through both auxin influx and efflux. Ethylene can enhance the expression of the auxin influx carrier AUX1 in the hook (Fig. 5b). The expression of AUX1 is stronger at the concave side, contributing to the formation of an auxin maximum (Vandenbussche et al., 2010). In addition, the flow of auxin through the hypocotyl from cotyledons to base is inhibited by apically positioned PIN3 (at the cellular level) in the presence of ethylene (Zadnikova et al., 2010). Together with increased expression of the auxin biosynthesis gene TAR2, and concomitantly increased auxin levels, these factors control the auxin content in the apical hook on ethylene exposure. As a result, the auxin signal is fine tuned by ethylene, which is visible at the level of AUX/IAA gene expression. A strong induction of IAA3/SHY2 and IAA13 genes is observed at the concave side of the apical hook in the presence of ethylene (Zadnikova et al., 2010; Fig. 5b). Likewise, in tomato, SlIAA3 gene expression increases at the concave side on ethylene treatment (Chaabouni et al., 2009). Antisense SlIAA3 plants have enhanced ethylene responses related to differential growth: apical hook exaggeration and – typical of tomato plants – petiole epinasty (Chaabouni et al., 2009). In Arabidopsis, IAA3/SHY2 is a probable candidate key factor for control of de-etiolation, including hook opening.
It is remarkable that ethylene uses similar mechanisms in both root elongation and apical hook development to exert an effect: ethylene relies on enhanced auxin production and the regulation of auxin transport to steer auxin maxima and minima. In addition, in both root elongation and formation of the apical hook, the auxin response depends on downstream GAs. Auxin coming from the shoot enhances DELLA breakdown in the roots (Fu & Harberd, 2003). Although the auxin efflux inhibitor 1-naphthylphthalamic acid (NPA) is extremely potent in stimulating the opening of the apical hook – even in the presence of ethylene (Zadnikova et al., 2010) – an enhanced GA signal provoked by the absence of the main elongation inhibiting DELLA proteins reduces this effect (Achard et al., 2003). This places the output of the GA pathway downstream of auxins, with DELLA function necessary at the convex side to stop elongation and cause consequent hook opening.
VI. Tissue- and cell-type-specific regulation of ethylene
Hormonal responses differ among and within a tissue; this also implies that not all cell types are equally important in generating the response. For example, it has been shown that the BR responses in the shoot epidermis are sufficient for shoot expansion (Savaldi-Goldstein et al., 2007) and that BR perception in the root epidermis controls root meristem size (Hacham et al., 2011). Similarly, the presence of an auxin signal in the epidermis and GA signaling in the endodermis of the root are essential for their effect (Swarup et al., 2005; Ubeda-Tomas et al., 2008).
Indirectly, the ethylene signaling state can be revealed using EIN3-binding site (EBS)-driven reporter gene expression (Stepanova et al., 2007). EBS carries the EIN3 binding site from the EIN3 target ERF1 combined with a minimal 35S promoter, and was fused to the coding sequence of a reporter gene (luciferase or glucuronidase). A drawback of this ‘biosensor’ is that it is based on one of the many targets of a single member of the EIN3 family, which may exclude tissues in which EIN3 is not expressed or inactive. Moreover, specialization of EIN3 and EIL1 towards seedling and adult plant responses, respectively, may also limit the more general use of such a biosensor (An et al., 2010). Therefore, despite being potential indicators of the ethylene signal, EBS reporters do not necessarily convey where ethylene is produced or accumulates.
Although it is generally accepted that all plant organs are capable of synthesizing ethylene from ACC, the production depends on the tissue type and the developmental stage (Abeles et al., 1992; Dugardeyn et al., 2008). Differences in the expression levels of ethylene biosynthesis genes may be markers for differential ethylene production within an organ. For instance, the ACS multigene family, encoding nine polypeptides in Arabidopsis, has a strong member-specific spatial regulation (Tsuchisaka & Theologis, 2004b). Promoter::GUS fusions of these genes revealed that all members, except ACS9, are expressed in seedlings, albeit with overlapping patterns. For example, in light-grown seedlings, expression is very general in cotyledons (eight genes), whereas only two genes are expressed in hypocotyls. This suggests a tissue-specific diversity of ethylene production (Tsuchisaka & Theologis, 2004b). In addition, in roots of maize, ZmACS genes are differentially expressed in the lateral root cap and outer (ZmACS6) and inner (ZmACS2/7) cortex layers. Cortical expression of ZmACS2/6/7 was observed in cell division, elongation and differentiation zones. By contrast, ZmACO expression was detected in the root cap, protophloem and the companion cells associated with metaphloem sieve elements (Gallie et al., 2009). Likewise, in Arabidopsis, in silico analysis indicates that the expression of the known ACO genes does not cover all cell types of the root (Dugardeyn et al., 2008); however, at least one member of the ACS family is expressed in each cell type. If the currently described ACO genes are the only ones functional in the multigene family, and provided that the described expression patterns reflect enzyme activity, this would imply that, for ethylene formation, ACC needs to be transported to cells that are capable of oxidizing the precursor to ethylene. Transport studies are limited to long-distance transport of ACC along the apical–basal axis of the root and shoot (Bradford & Yang, 1980; Fuhrer & Fuhrerfries, 1985; Finlayson et al., 1991), or in stamens (Kiss & Koning, 1989). Nevertheless, as ACC is readily soluble in an aqueous environment, lateral cell-to-cell transport through the symplast could also occur. The advantage of ethylene being a small, gaseous and lipophilic hormone is that it does not need a specific transport mechanism to spread in plants. In the cells in which it is produced, it is probably trapped in the membranes. Diffusion to surrounding tissues will depend on the tissue type. Based on a partition coefficient of 1/10 000 in the aqueous/gas phase, ethylene will essentially be released in the gas phase wherever there is an air barrier. Thus, aerenchyma can serve as a fast transportation route. Moreover, diffusion through the epidermis releases the hormone to the exterior of the plant, and allows it to serve as a long-distance signal to neighboring plants. As gas exchange essentially occurs through stomata, it can be assumed that ethylene is also released or taken up through the stomata. Interestingly, ethylene induces stomatal closure, indicating specific regulation within guard cells and an interaction with the abscisic acid pathway (Desikan et al., 2006). Hence, the production of ethylene as a stress hormone in appropriate amounts is likely to help in the survival of adverse conditions, such as drought or air pollution (Wilkinson & Davies, 2009).
Current analytical methods do not allow the measurement of ethylene production at the cell-type level within an organ or tissue type. Hence, it is not known which cell types actually produce or sense ethylene. ACO activity may be indicative, but the production of ethylene is obviously only possible if ACC is effectively present, as illustrated above. For analyses at the cellular resolution, the hormone should either be measured in situ by a biosensor system (Sadanandom & Napier, 2010), or highly sensitive mass spectrometry should be applied to determine the hormone in minute amounts of cell material collected after FACS-based separation of green fluorescent protein (GFP)-marked cell types (Petersson et al., 2009). A high-resolution map of auxin distribution in the root apex has been established based on the latter method, and could also be established for ACC (Petersson et al., 2009). By contrast, measuring ethylene release from a small group of cells is an enormous technical challenge. Ethylene should either be trapped during the cell isolation process or derivatized, after which the product could be followed up in situ. However, none of the currently available indicators for ethylene seem to be valid for such an approach, as molecules known to react with alkenes, such as KmNO4 or Br2, lack specificity. In addition, the collection of sufficient amounts of material to exceed the detection limit of currently used devices is another major bottleneck. Alternatively, a biosensor system with electrochemical detection could be developed based on ACO activity, as was performed for cytokinin measurements based on cytokinin oxidase (CKX) activity (Kowalska et al., 2011).
VII. Cellular basis of ethylene effects on growth
1. Growth effects of ethylene are not restricted to cell expansion
Ethylene contributes to the plasticity of growth and proves to be a signal that can lead to rapid and multilevel-controlled responses. A reduction in ethylene is an important signal for stress relief and indicates that expansion growth can be resumed. The inhibition of cell expansion is a well-described ethylene response at both the shoot and root levels (Kieber et al., 1993; Rodrigues-Pousada et al., 1993; Roman et al., 1995). However, effects of ethylene on growth are not restricted to cell expansion. Both ethylene biosynthesis and signaling are important in regulating the meristem size (Thomann et al., 2009), and ethylene also controls the balance between proliferation and quiescence of stem cells in the root (Ortega-Martinez et al., 2007). Furthermore, it has been shown that enhanced activation of the ethylene signal coincides with osmotic stress-induced cell cycle arrest in leaf primordia (Skirycz et al., 2011). This response was independent of EIN3, and could indicate an alternative pathway that parallels the canonical signaling pathway from EIN2 downward. EER5 defines such a pathway (Christians & Larsen, 2007). EER5 regulates ethylene signaling independent of EIN3 and physically interacts with EIN2. Could EER5 be involved in cell cycle arrest? Both EIN2 and EER5 interact with subunits of the COP9 signalosome (CSN) (Christians & Larsen, 2007). In view of the positive control of CSN on cell cycle progression, this is a promising path to explore (Dohmann et al., 2008).
Although osmotic stress causes an increase in endoreduplication in leaf primordia, this is not controlled by ethylene (Skirycz et al., 2011). These findings are in contrast with the situation in hypocotyls, in which ethylene enhances endoreduplication (Dan et al., 2003; Saibo et al., 2003). In both hypocotyls and leaf primordia, cell division is negatively affected by ethylene (Dan et al., 2003; Skirycz et al., 2011), with the exception of ethylene induction of cell division leading to the formation of stomata (Saibo et al., 2003). It is interesting that endoreduplication is differentially affected in these tissues. Hence, cell identity provides additional information to link or uncouple ethylene signaling from the mitotic cell cycle and endocycle. It can be speculated that cell volume may be an essential parameter. Enhanced endoreduplication may relate to radial expansion that occurs in the presence of ethylene, as shown, for instance, in hypocotyls (Saibo et al., 2003).
2. Cell wall dependence of ethylene effects
From the previous discussion, it is clear that ethylene largely works by influencing auxin transport and synthesis. A question that remains to be answered in this regard is how ethylene can exert its effect on cell size in such an extremely rapid manner (10 min in roots). It is very probable that the gene and/or protein changes caused by ethylene, followed by the auxin signaling cascade, cannot occur sufficiently rapidly to generate this growth response. Hence, an auxin-independent mechanism could account for the fast response, whereas consolidation and more profound inhibition may require a genomic response. In this section, we propose a hypothesis for the direct effects of ethylene on cell wall dynamics.
Recent electrophysiological experiments involving proton-specific vibrating probes have demonstrated that ethylene (or its precursor ACC) is capable of switching the plasma membrane H+-ATPases from the active to the inactive state, resulting in alkalinization of the root surroundings, especially at the start of the elongating zone. Furthermore, activation of the proton pumps by fusicoccin recovers part of the ACC-induced reduction in cell elongation, suggesting that a more acidic pH of the cell wall is needed for cell wall expansion (Staal et al., 2011).
During elongation, the cell wall is a center of activity (Cosgrove, 2005; Nishitani & Vissenberg, 2007). Ethylene-induced alkalinization of the cell wall, by inactivating the plasma membrane proton pumps, is expected to especially influence expansin activity, as it is known to require an acidic pH for activity. Xyloglucan endotransglucosylase/hydrolases (XTHs), however, have many isozymes with different pH activity profiles (Maris et al., 2011), and an in situ fluorescent assay (Vissenberg et al., 2000) has shown that the grafting activity (xyloglucan endotransglucosylase, XET) in the elongation zone is unchanged after ACC addition (Verbelen et al., 2008), suggesting that ethylene affects multiple aspects of cell wall dynamics.
In addition to altering cell wall loosening capacities, peroxidase-mediated cross-linking of cell wall polymers can also account for part of the elongation arrest (Knox, 1995). De Cnodder et al. (2005) have provided evidence in favor of this hypothesis. As peroxidases can be present in the cell wall, even in elongating cells, ethylene or ACC could activate these enzymes without the need for a genomic response. Activation could simply rely on the production of reactive oxygen species (ROS), necessary as cofactors in the cross-linking event (De Cnodder et al., 2005). All of this evidence strongly points towards the cross-linking event being responsible for the fast response observed after ACC addition. In addition, as proposed by Seifert & Blaukopf (2010), it is expected that plant cells can sense cell wall polymer structure as well as its mechanical performance. Different kinases have been proposed to fulfill the role of ‘membrane integrity sensor’, and it is tempting to propose that, on ACC treatment, lack of wall loosening and increased cross-linking events might initiate such signaling cascades.
3. Regulation of anisotropic cell expansion
From the above discussion on cell elongation arrest, it could be hypothesized that root bending following a gravitropic stimulus may be based on similar control mechanisms, as it is also caused by the inhibition of cell elongation, albeit differential on the upper and lower sides of the root, mediated by the asymmetric redistribution of auxin. During this bending response, auxin only needs to be perceived in the elongating epidermal cells (Swarup et al., 2005), implying that the inhibition of epidermal elongation alone is sufficient to induce a top-down graded elongation arrest of underlying cells, without conspicuous bulging of internal tissues. This strongly points to the controlling role of the epidermis in elongation, especially of underlying tissues. In this respect, it has recently been described that skewing of epidermal cells would automatically lead to cell shape changes of underlying tissues, simply by geometrical constraints (Weizbauer et al., 2011). It is therefore puzzling that auxin perception in the epidermal cells alone, leading to decreased epidermal cell lengths (Swarup et al., 2005), is not sufficient to cause the dramatic elongation arrest seen by ACC treatment. This implies that the expansion of underlying tissues also regulates the extent of epidermal expansion. The fact that elongation apparently needs to be altered in all tissues simultaneously to result in maximal growth arrest might suggest that ethylene needs to be received in more cell types than the epidermis. Cell-type-specific alteration of ethylene sensitivity will need to be investigated to provide an answer to this question.
Another striking feature of the ACC response is that, after prolonged exposure, radial expansion is seen in cells of the treated root elongation zones. This is reminiscent of the root phenotype after blocking the gibberellic acid response in the endodermal cells by a dominant negative mutant DELLA protein (gai; Ubeda-Tomas et al., 2008). As a result of the endodermal anisotropic growth arrest coupled to the observation that plant cells cannot slide past one another (Sinnott & Bloch, 1939), surrounding cortical and epidermal cells lose anisotropic growth and start to bulge out. Bulging and buckling of cells is also frequently seen in roots with cellulose defects, whether after treatment with cellulose synthesis inhibitors (Hogetsu et al., 1974) or in mutants (Arioli et al., 1998). Indeed, it has been shown in fei1fei2 mutants that ACC can be linked to changes in cell wall synthesis, although the exact mechanism remains elusive (Xu SL et al., 2008). Double mutants of two leucine-rich receptor-like kinases FEI1 and FEI2 showed short radially swollen roots in the presence of sucrose, accompanied by a striking defect in cellulose biosynthesis. The expansion and cellulose synthesis defect was reverted in fei1fei2 roots by the inhibition of ACO, strongly suggesting a link between ACC and cell wall synthesis. Xu SL et al. (2008) proposed that ACC, rather than ethylene itself, acts as a signaling molecule to regulate cell expansion, a hypothesis that has been confirmed recently (Tsang et al., 2011). This also stresses the fact that the interpretation of responses elicited by ACC should not be simply extrapolated to ethylene-related processes. Furthermore, mutants such as cev1, affected in cellulose synthesis, and sabre mutants, showing a swollen root phenotype, can be partially rescued by blocking ethylene action (Aeschbacher et al., 1995; Ellis et al., 2002). The above-mentioned evidence could lead to the wrong idea that cellulose and microtubule arrangements, guiding cellulose synthases (Paredez et al., 2006), are the only factors limiting radial expansion. Additional factors seem to be required (Wiedemeier et al., 2002), and the differential gene expression caused, for example, by the gai mutation in endodermal cells (Swarup et al., 2007) could identify other key regulators of anisotropic cell growth.
As is the case for the control of root elongation, cell wall mechanics are probably key processes in the ethylene regulation of shoot growth. When grown in the dark, hypocotyl cells first deposit thick cell walls, after which they become thinned during subsequent elongation. This is especially visible in the outer epidermal wall, reaching values of two to ten times thicker than the inner walls, before becoming thinned (Derbyshire et al., 2007). Nevertheless, the more inner (epidermal, cortical and endodermal) walls also become thinner during expansion, albeit to a lesser extent (Refregier et al., 2004; Derbyshire et al., 2007). Interference with the deposition of this initially thick wall, for example by cellulose synthase inhibitors, results in the loss of elongation when applied before the start of elongation, but has no effect on subsequent elongation when the thick wall has been laid down (Refregier et al., 2004). One could question whether the normal procedure of thick wall deposition, followed by thinning during elongation, is present in constitutive ethylene response mutants showing dwarfed phenotypes, and whether this phenomenon is present in the root as well. Interference with cellulose deposition often results in swelling of the cells, a feature that is also observed in dwarfed ethylene mutants, albeit to a lesser extent. However, ethylene insensitive mutants have longer hypocotyls. Is this the result of the deposition of an initially even thicker outer epidermal cell wall, of continued thinning, or of another mechanism, for example changes in pectin metabolism (Pelletier et al., 2010)? Cellulose deposition is needed before the actual elongation starts, but it remains to be seen whether growth phenotypes change when etiolated wild-type plants are treated with ethylene at different time points (before or 48 h after stratification). A kinetic approach to growth responses and cell wall thickness can provide an idea of the effect of ethylene on cellulose synthesis or additional factors. Overall, it is clear that the correlation between ethylene signaling, cell wall synthesis, thickness, composition and mechanics remains to be uncovered.
In addition to wall thickness, cellulose orientations in the outer and inner epidermal cell walls are different. The outer wall shows a predominant random arrangement of cellulose microfibrils, restricting radial expansion of the hypocotyl. Only in a short time window, the orientation becomes transverse in a few cells lying as a belt around the organ. This transiently allows elongation of the epidermal cells. Inner epidermal walls and the walls of the inner tissues always show a transverse arrangement (Crowell et al., 2011). Whether this belt with transiently transverse cellulose microfibrils is also present in roots and whether it disappears more rapidly or slowly in different ethylene mutants are presently unknown, but this could clarify the role of ethylene at the level of wall architecture and enhance our knowledge on its mechanism of action in the control of cell elongation.
The authors acknowledge the Fund for Scientific Research-Flanders (G.0524.07 and G.0298.09), Ghent University and the Belgian Science Policy Office (IUAP P6/33) for financial support. F.V. is a postdoctoral researcher of the Fund for Scientific Research-Flanders (FWO).