PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of annbotAboutAuthor GuidelinesEditorial BoardAnnals of Botany
 
Ann Bot. 2016 April; 117(5): 795–809.
Published online 2015 October 14. doi:  10.1093/aob/mcv151
PMCID: PMC4845800

Molecular mechanisms governing differential robustness of development and environmental responses in plants

Abstract

Background Robustness to genetic and environmental perturbation is a salient feature of multicellular organisms. Loss of developmental robustness can lead to severe phenotypic defects and fitness loss. However, perfect robustness, i.e. no variation at all, is evolutionarily unfit as organisms must be able to change phenotype to properly respond to changing environments and biotic challenges. Plasticity is the ability to adjust phenotypes predictably in response to specific environmental stimuli, which can be considered a transient shift allowing an organism to move from one robust phenotypic state to another. Plants, as sessile organisms that undergo continuous development, are particularly dependent on an exquisite fine-tuning of the processes that balance robustness and plasticity to maximize fitness.

Scope and Conclusions This paper reviews recently identified mechanisms, both systems-level and molecular, that modulate robustness, and discusses their implications for the optimization of plant fitness. Robustness in living systems arises from the structure of genetic networks, the specific molecular functions of the underlying genes, and their interactions. This very same network responsible for the robustness of specific developmental states also has to be built such that it enables plastic yet robust shifts in response to environmental changes. In plants, the interactions and functions of signal transduction pathways activated by phytohormones and the tendency for plants to tolerate whole-genome duplications, tandem gene duplication and hybridization are emerging as major regulators of robustness in development. Despite their obvious implications for plant evolution and plant breeding, the mechanistic underpinnings by which plants modulate precise levels of robustness, plasticity and evolvability in networks controlling different phenotypes are under-studied.

Keywords: Developmental robustness, species diversity, plasticity, capacitor, canalization, noise, network hubs, network motifs, redundancy, Hsp90, chromatin, rDNA, hormones

INTRODUCTION

Organisms constantly face changing environmental conditions, some of which are predictable and others are not. To cope and thrive in the face of changing environments, organisms have developed systems that enable phenotypic change or prevent it. At one extreme, there are some traits that exhibit little or no phenotypic change despite an environmental challenge; these traits display robustness or canalization (Waddington, 1961; Lempe et al., 2012). At the other extreme, there are traits that display a large degree of plasticity – programmed shifts in phenotype in response to different environments (Pigliucci, 2001), which are important for fitness. A population of genetically identical individuals exhibits little variation in a robust trait in the face of environmental change. In contrast, a plastic trait will exhibit a mean shift across genetically identical individuals in response to environmental change. Within each organism, traits range from robust to plastic, as determined by the underlying genetic network, to provide optimal fitness within changing environments (Waddington, 1961).

Across the broad array of traits required for fitness, organisms will differentially deploy robustness and plasticity to control phenotype. The specific selection pressures on an individual trait influence the degree to which this trait will display robustness or plasticity. When stabilizing selection acts on a trait, robustness is adaptive and plasticity is costly (Wagner et al., 1997; but see also Kawecki, 2000). One example is floral morphology, which displays strong stabilizing selection across angiosperms to maintain interactions with pollinators. In contrast, plasticity will be adaptive in environments with predictable changes in temperature, light or precipitation (Gomez-Mestre and Jovani, 2013). Finally, unpredictably fluctuating environments create selective regimes that favour stochastic plasticity, in which the phenotype unpredictably changes across or even within individuals (Clauss and Venable, 2000). Thus, environmental responses of all traits are likely tuned from robustness to programmed or stochastic plasticity depending upon the selective pressures and environments that influence that trait.

Although robustness and plasticity are mechanistically related, complete robustness in a trait precludes plasticity and vice versa. This absolute relationship, however, can be altered in a temporal fashion, such as in traits that display programmed plasticity, wherein they shift from one robust state to another (Fig. 1). For example, the transition to flowering in the annual plant Arabidopsis thaliana demonstrates a temporal shift from robust vegetative growth via a plastic transition period to robust reproductive growth (Swiezewski et al., 2009; Csorba et al., 2014). Once the transition to flowering has occurred, the trait is robust and, at least in A. thaliana, irreversible. Other traits, however, display reversible plasticity between alternating robust states. In plants, the shade avoidance response shows reversible plasticity wherein plants can alter growth in response to varying light exposures or shade (Devlin et al., 1999). Reversible plasticity is also displayed by the repeated seasonal transitions between vegetative and reproductive growth in perennial plants. As demonstrated with these examples, traits are shaped by a blend of robustness and plasticity acting at different times.

Fig. 1.
Changing environments induce phenotype-specific responses. Phenotypic responses to changing environments range from plastic to robust. Environmental changes that occur predictably lead to programmed responses. Shown here is the transition between vegetative ...

Where a trait lies on the robustness-to-plasticity spectrum is determined by the underlying genetics that controls the trait and how this genetic network responds to changing environments to optimize fitness. Within genetic networks, there are at least two components that influence a trait’s position on the robust-to-plastic spectrum: redundancy among genes and topology of gene interactions. Redundancy influences both phenotypic robustness and plasticity. For example, gene duplication can provide robustness through redundant function and can promote plasticity through sub-functionalization of gene copies. Similarly, network topology also influences both phenotypic robustness and plasticity. Network topology can be expressed as gene connectivity and the pattern of wiring among genes. Both the connectivity of a given gene and its position within the wiring of the genetic network will influence a gene’s role in robustness and plasticity. In most of the existing literature, the modulation of robustness and plasticity has been traced to specific molecular mechanisms and genes, some of which we briefly review here, and we suggest further candidates. Here, we argue that the effects of these genes and mechanisms are fundamentally acting through their effects on the underlying genetic (and molecular) networks.

The fact that whole-organism traits, such as flowering, undergo transition from robust to plastic states implies that this transition must be wired within existing genetic (and molecular) networks. In other words, the properties of the underlying network controlling growth must be able to maintain robust vegetative growth but simultaneously permit the plastic transition to flowering. Here, we review how recent research is illuminating the mechanistic underpinnings of the complex relationship between robustness and plasticity.

First, we consider the influence of morphological and genetic redundancy on plasticity and robustness, particularly considering whole-genome and single-gene duplications. Second, we discuss how features of gene regulatory networks, such as connectivity and specific network motifs affect, phenotypic plasticity and robustness. In addition to such system-level features, specific molecular mechanisms have been associated with tuning environmental responses. We discuss the role of the protein chaperone Hsp90, chromatin-modifying enzymes and ribosomal DNA (rDNA) copy number variation in influencing phenotypic plasticity and robustness.

REDUNDANCY AS A SYSTEMS-LEVEL FEATURE GOVERNING ROBUSTNESS IN PLANTS

One systems-level feature through which robustness arises is redundancy at both the morphological and the genome level. Redundancy is the presence of nearly identical duplicate parts wherein each duplicate suffices to replace the function of the other copy. Thus, redundancy allows a system to function fully even when a duplicate is lost. At the morphological level, redundancy manifests through continuous development, in which multiple copies of organs, such as branches and leaves, are produced as defined by a plant’s developmental programme and as enabled by environmental conditions.

At the genome level, plants show an unusual capacity to tolerate whole-genome duplication and hybridization – and thus redundant gene copies – presumably due to RNA-directed DNA methylation (RdDM), a plant-specific mechanism of gene silencing that prevents complications due to increased gene dosage (Ha et al., 2009b). The functional redundancy created by these events likely contributes to plant robustness. In this section, we discuss recent evidence that plants indeed utilize redundancy at the morphological and genetic levels to increase robustness.

Morphological redundancy in plant development promotes trait robustness

Plants are composed of repeating units, such as individual leaves, flowers, branches and roots. Unlike most animals, plants replace damaged units with new ones through continuous development, creating morphological robustness. Others have described plants as ‘populations of organ modules’ and suggested that the extreme morphological redundancy observed in plants compensates for their lack of mobility (Harper, 1980).

The trait leaf venation is a key example of robustness arising from morphological redundancy. Veins provide mechanical support to leaves as well as structures for the transfer of water and solutes (Roth-Nebelsick et al., 2001). Major or primary veins branch into minor or secondary veins and so forth, creating a branching network of veins. In angiosperms, these veins form a reticulated network of criss-crossing lines (Roth-Nebelsick et al., 2001). This reticulated network creates robustness by providing multiple alternative pathways by which solutes can move if veins are damaged by herbivore feeding or strong winds (Roth-Nebelsick et al., 2001). Does the generation of reticulate networks with high redundancy and robustness come at a cost? In other words, can an argument be made that the robustness of the vein network is selected? Price and Weitz (2014) modelled the costs of increased redundancy in reticulated vein patterns across 324 angiosperm species. On average, the length of the network only increased by 6.3 % when reticulated networks were compared with branched ones. Although this length increase may be an evolutionarily important cost, it is also correlated with increased photosynthetic rate (Price and Weitz, 2014). Thus, the formation of these highly redundant networks is likely cost-neutral or driven by increased photosynthetic rate. The presence of reticulated venation networks has been suggested to be a driver of the success of angiosperms (Brodribb and Feild, 2010), possibly due to both increased photosynthetic rate and increased robustness.

The reticulated nature of leaf venation is robust, not plastic. In contrast, components that underlie leaf venation, such as vein diameter and vein density, are plastic (Sack and Scoffoni, 2013). For example, the minor veins in leaves under high irradiance show a greater diameter than those in shaded leaves; the greater diameter of minor veins is also observed in leaves exposed to drought (Dorantes and Sánchez, 2006; Brodribb and Jordan, 2011). Vein density is also plastic; sun-exposed leaves tend to show higher density of minor veins compared with shaded leaves within trees (Sack and Frole, 2006). Thus, in order to maximize solute flow and photosynthesis, sun-exposed leaves develop additional redundant minor veins. It is unknown whether the genes generating the vein network control both its reticulated venation pattern and its plasticity or whether the two are controlled by separate molecular entities. As increased vein density increases network robustness, we speculate that the same genes may govern both.

Genome duplication/redundancy generates robustness while simultaneously generating the raw material for wiring networks for trait plasticity

Whole-genome duplication (WGD) and hybridization within plants are well-studied mechanisms by which plants can simultaneously generate robustness as well as increase the potential for fine-tuning development and environmental responsiveness and evolving new traits (Rieseberg and Carney, 1998; Fawcett et al., 2009).

After WGD or hybridization of closely related species, all genes are present in duplicate. Over time, many copies are lost and revert to their prior copy number. In some cases, the presence of a paralogue relieves the selection pressure acting on the original gene copy. This relief of selective pressure can result in the emergence of a novel function in one of the paralogues (neo-functionalization). In other cases, both gene copies become specialized in different aspects of the original function (sub-functionalization) (Conant and Wolfe, 2008). Both WGD and hybridization escape the challenge of dosage compensation that individual gene duplication events face. At their core, WGD and hybridization enable the building of the intricate genetic networks that allow both robustness and plasticity.

Many genes that are maintained in duplicate post-WGD continue to share some function. This redundancy in function is a well-documented source of robustness in all organisms, including plants. Indeed, in plants and many other organisms, single-gene deletions of duplicated genes have smaller effects on morphology than deletions of singletons (Hanada et al., 2009). The same tendency was observed for metabolite levels (Hanada et al., 2011). Thus, duplicated genes provide trait robustness in the presence of mutations. However, it should be noted that the extent of the functional redundancy provided by gene duplicates is rarely studied in much detail. Members of a closely related gene family may compensate for the loss of one gene under laboratory conditions or even most conditions, yet defects may arise in the context of a particular environment or a particular genomic background. Consequently, the concept of redundancy depends critically on the level of detail studied or the precision with which the phenotype was measured.

As paralogues diverge from one another in function, the robustness they provide for each other’s loss is reduced. For example, the well-studied BES1 and BZR1 transcription factors (TFs) of the brassinosteroid signalling pathway are frequently assumed to function redundantly (Krizek, 2009; Ye et al., 2012). This assumption arises from their high homology (88 % amino acid identity) and the fact that these TFs have been typically studied with individual dominant mutants. These mutations prevent degradation and convert the affected protein, either BES1 or BZR1, into the major effector of brassinosteroid signalling, obscuring subtle differences between the TFs. Indeed, close examination of BES1 and BZR1 reveals that the two TFs have different protein interaction partners, different splice forms and different patterns of expression, as well as different mutant phenotypes (Wang et al., 2002; Yin et al., 2002; Lachowiec et al., 2013; Jiang et al., 2015).

Paralogues are known to diverge in expression patterns in different subsets of tissues, developmental stages or environments (Diss et al., 2014). We argue that plasticity in gene expression patterns emerges because the presence of duplicated genes enables expression patterns to be specifically associated with each copy. The diversity in gene expression across closely related species is greater among duplicated genes than singletons (Gu et al., 2004; Ha et al., 2009a), and genes with more paralogues diverge in expression to an even greater degree (Ha et al., 2009a). For example, in cotton over 99 % of gene duplicates have diverged in tissue-specific expression patterns over the last 60 My (Renny-Byfield et al., 2014). These gene duplicates often complement one another’s expression pattern, suggesting that gene expression divergence contributed to the maintenance of these duplicated genes (Renny-Byfield et al., 2014).

Divergence in expression patterns among duplicate genes likely promotes plasticity in gene expression in response to changing environments. In A. thaliana, gene duplicates tend to be upregulated when exposed to biotic and abiotic to stresses, unlike singletons (Ha et al., 2007). Divergence in gene expression between duplicates was higher in response to abiotic and biotic stresses when compared with expression divergence in different tissues (Ha et al., 2007). In one specific example, the two copies of the alcohol dehydrogenase gene (Adh) in cotton diverged in their expression responses to cold, dark and submersion (Liu and Adams, 2007). Thus, we suggest that the transient robustness provided by redundant gene duplicates increases regulatory precision in response to specific environmental stimuli and enables appropriate plastic responses.

It is undisputed that WGD and hybridization of closely related species provide raw material for evolution, specifically the emergence of novelty. Plant defence metabolism or secondary metabolism is a key model system in which the role of WGD in generating trait plasticity and novelty is being studied (Chae et al., 2014). The continuous battle between pests and plants leads to continuous evolutionary pressure to evolve novel defence metabolites (Ehrlich and Raven, 1964). This Red Queen evolutionary scenario has led to the production of over 200 000 secondary plant compounds across angiosperms (Hartmann, 2007). Studies of the reference plant A. thaliana generated evidence supporting the role of both WGD and gene duplications in creating new glucosinolates, which are common defence compounds against herbivores (Kliebenstein, 2008; Hofberger et al., 2013). Duplicate genes for enzymes within the glucosinolate pathway show a higher level of retention in comparison with genes involved in primary metabolism (Kliebenstein, 2008; Hofberger et al., 2013). The redundancy produced by duplication has led to a number of instances in which one paralogue has evolved a brand-new enzymatic function or even a novel TF function leading to the generation of a new defence compound (Kliebenstein et al., 2001; Kroymann et al., 2003; Hansen et al., 2007, 2008; Li et al., 2008; Sønderby et al., 2010). Thus, WGD facilitates plasticity, here as the programmed response to herbivores, while also contributing to robustness.

Like WGD, hybridization of related species doubles genome content. Unlike WGD, hybridization combines two different genotypes, which permits the emergence of transgressive phenotypes. As a result, hybrids may be more robust to novel environments than progenitor lines. For example, the hybrid sunflower Helianthus paradoxus is significantly more salt-tolerant in laboratory experiments than either of its progenitor lines, H. annuus and H. petiolaris. The increased capacity of H. paradoxus hybrids to tolerate high-salt environments allows them to occupy salt marsh habitats (Welch and Rieseberg, 2002). Thus, in this case, increased robustness to environmental stress allows hybrids to colonize a new environment.

NETWORK WIRING AND ROBUSTNESS AND PLASTICITY

In addition to redundancy, the topology and connectivity of a network can dramatically influence robustness and plasticity. Connectivity describes the number of nodes with which a given node interacts. In terms of genetic networks, this concept is often applied to TF networks, by enumerating the input edges and output edges (targets) of TFs. In addition to simply counting the number of interactions per node, as we discuss in this section, the specific pattern of the connections or network wiring is critical for robustness and plasticity. For example, some TFs, like HY5, the master regulatory of photomorphogenesis, show many input and output edges despite changing environments. In contrast, some TFs show dramatic changes in both input and output edges, and yet others only show change in either input or output edges in response to stimuli (Sullivan et al., 2014). These patterns give clues as to the importance of a particular TF in a studied process, as well as to its possible role in robustness or plasticity. Intuitively, a factor that vastly changes connectivity in response to a given stimulus is of great importance for the ensuing plastic response. An example would be HSFA2, the known ‘amplifier’ of the heat-shock response, which gets more connected upon heat shock (Sullivan et al., 2014). In contrast, TFs that drastically change in the number of input edges while output edges do not change, are implicated as possible noise buffers. In fact, such detailed analyses of network topology can detect changes in regulatory feedback loops, including autoregulatory ones, along a time course. Feedback loops are associated with robust molecular phenotypes (Hornstein and Shomron, 2006).

Connectivity in gene networks is associated with robustness

Connectivity in gene networks follows an exponential distribution, with most genes having few interactors and few genes – the so-called network hubs – having many. This pattern is associated with robustness because the likelihood of a disrupting a hub is small, and disruptions elsewhere are buffered through the modular structure of genetic networks (Albert et al., 2000; Lehner et al., 2006; Levy and Siegal, 2008; Fu et al., 2009). However, high connectivity is not the only possible definition of a hub. We prefer the more specific definition of hubs as crucial nodes for information flow in which disruption produces significant phenotypic consequences and loss of robustness (Table 1). Such genes/molecules must not necessarily be highly connected within a network module; rather they could touch many modules across the entire network (Whitacre, 2012).

Table 1.
Definition of some terms

How are hubs identified? Traditionally, network hubs have been identified through genetic analysis as those genes that show outsize epistasis with many other genes. Examples include the chaperone Hsp90 in yeast, plants, flies, nematodes and fish (Rutherford and Lindquist, 1998; Queitsch et al., 2002; Yeyati et al., 2007; Jarosz and Lindquist, 2010; Casanueva et al., 2012), in addition to chromatin remodelling proteins in nematodes (Lehner et al., 2006), developmental regulators such as ERECTA in plants (Hall et al., 2007) and key molecules in microRNA pathways in both plants and animals (Cassidy et al., 2013; Schmiedel et al., 2015). Notably, Hsp90 is an example of a hub that is not part of a specific network module or pathway, but rather peripherally but essentially affects many, if not most, processes in living cells (Whitacre, 2012). Chromatin remodelling fits into this category too as it globally modulates gene expression.

Similar to identifying hubs in genetic networks, hubs can be defined as proteins that engage in many more functionally relevant protein interactions than others (Arabidopsis Interactome Mapping Consortium, 2011). Surely, most components of the protein degradation machinery would cause large phenotypic change when disrupted. Hsp90, a well-known player in trait robustness and plasticity (Sangster and Queitsch, 2005; Lempe et al., 2012), acts at the level of protein interactions together with its army of co-chaperones that define substrate specificity. Another example is the protein ELF3, which appears to serve as a scaffold for potentially many other proteins (Nusinow et al., 2011). Through its various interactions, ELF3 contributes to circadian regulation (Nusinow et al., 2011; Anwer et al., 2014), regulation of growth and development (Zagotta et al., 1996; Liu et al., 2001) and temperature response (Box et al., 2014; Mizuno et al., 2014; Nieto et al., 2015). By mapping gene expression noise, ELF3 was also identified as a major player in the robustness of gene expression and whole-organism traits (Jimenez-Gomez et al., 2011).

The importance of protein complexes and their changing connectivity for robustness is also illustrated by recent work on floral homeotic genes. The identity of floral organs is determined by the concentration of homeotic TF proteins and their complexes as described in the classic ABC model (Bowman et al., 1991) and its descendant, the ABCDE model (Theißen, 2001). The B-genes DEF and GLO (also known as PISTILLATA and APETALA 3 in other species) are primarily expressed in the second and third floral whorls, defining petal and stamen identity. Unlike the other MADS-box homeotic TFs, DEF and GLO exclusively heterodimerize with one another in derived angiosperms; their connectivity is reduced in comparison with other related TFs. Disrupting either gene suffices for B loss-of-function phenotypes. However, in basal angiosperms they can also homodimerize, thereby increasing DEF and GLO connectivity. Early angiosperms tend to have floral organs with weaker identity boundaries. Melzer and co-authors (2014) argue that the defined floral forms and sharp organ identify boundaries of derived angiosperms may be due to the obligate heterodimerization of the B-class proteins. Here, it appears that reduced rather than increased connectivity among DEF and GLO TF proteins results in more robust floral organ formation. It remains to be seen whether these same genes play a role in the plastic shifts of floral shape as seen in the plant species with environmentally determined monoecious flower formation.

In addition to genetic and protein interaction networks, much effort has been devoted to building expression-based networks (Ma et al., 2013, 2015) or networks based on TF occupancy at promoters. Using the latter approach, Sullivan et al. (2014) studied the dynamics of TF networks in A. thaliana seedlings transitioning from dark to light growth conditions and seedlings exposed to heat shock (Sullivan et al., 2014). Although many of their findings are validated by prior classic studies, the functional relevance of the many novel connections that are described has yet to be explored. As discussed above, when connectivity is divided between input and output edges for each TF, stark differences arise for TFs with comparable overall connectivity across conditions. These detailed patterns hold information about potential roles of particular TFs in robustness. For example, some TFs, such as EIN3 and MYC2, drastically change the number of input edges across conditions while their output edges remain more or less stable. This pattern suggests that such factors may act as environmental noise buffers across conditions because, despite vastly increased inputs, the outputs of these TFs remain stable.

However, it is non-trivial to predict the influence of TF hubs based on their connectivity. Due to the hierarchical nature of regulatory networks, regulatory effects of a TF can be multiplied or dampened across the regulatory network (Taylor-Teeples et al., 2014). For example, let us assume a TF directly interacts with only a few other TFs; however, these secondary TFs have a large number of targets. In this scenario, disruption of the upstream, less connected TFs may produce more severe phenotypic consequences than disrupting one of the secondary, highly connected TFs. The regulatory hierarchy may be particularly important in the regulation of metabolism (Taylor-Teeples et al., 2014). Support for the idea that direct connectivity is not highly predictive of phenotypic consequence was recently found by examining TF connectivity to genes in the biosynthetic pathway for the glucosinolate defensive metabolites in A. thaliana (Sønderby et al., 2010). The authors observed no relationship between the connectivity of a TF and the likelihood or magnitude of a corresponding TF mutation having a phenotypic effect on glucosinolate accumulation. It remains to be tested whether this conclusion holds when secondary connections are included in the network analysis.

Hub genes are known to be key targets of biotic factors, which may promote effective programmed plasticity responses. For example, plant-infecting viruses encode a small number of proteins but elicit broad transcriptional responses throughout a plant. A meta-analysis of transcriptional studies examining the effect of viral infection suggests that, across viral infections of A. thaliana, the most consistently responsive genes are hub genes (Rodrigo et al., 2012). Similar patterns were found for bacterial infections, which also preferentially affect hub genes (Mukhtar et al., 2011). A systematic test of A. thaliana proteins interacting with pathogen effector proteins from two bacteria that diverged 2 billion years ago discovered that a limited set of hub proteins interacts with both pathogen effectors (Mukhtar et al., 2011). Together, these studies suggest that diverse pathogens display convergent evolution such that they recurrently obtain the capacity to target hub genes within plants. Thus, in the case of biotic attackers, it appears that hubs are targeted and elicit broad, robust responses. These results create a centrally important chicken–egg question. Do the biotic attackers target hub genes to overcome plant defence? Or do plants utilize these hubs as trip-wires to detect biotic attackers and elicit a strong and robust defence? In the larger context of this review, are hubs generally maintaining robustness? Are hubs enabling both short-term plastic responses to environmental change and long-term evolutionary adaption? The available evidence suggests that both are true (Lehner et al., 2006; Fu et al., 2009).

Network motifs associated with robustness

In addition to connectivity, the topology of connections among genes determines the robustness of gene networks. Networks are composed of simple patterns or motifs of connections among nodes. Probably the best known motifs associated with robustness are feed-forward and feed-back loops (Alon, 2003; Hornstein and Shomron, 2006; Lempe et al., 2012; Payne and Wagner, 2014; Schmiedel et al., 2015). Feed-forward loops come in two flavours: coherent and incoherent.

Coherent feed-forward loops typically consist of a TF that regulates another TF or miRNA and they jointly regulate a target gene either positively or negatively (Fig. 2A). This network motif is often found in cell fate determination (Hornstein and Shomron, 2006) because its redundancy reinforces fate decisions. Examples have been described in mammalian cells (Xie and Cvekl, 2011) and flies (Cassidy et al., 2013). In plants, coherent feed-forward loops also contribute to the robustness of secondary cell wall synthesis. Prior studies attributed the robustness of secondary cell wall synthesis to the extensive functional redundancy among the contributing TFs (Carlsbecker et al., 2010). An in-depth survey of the TF network directing secondary cell wall synthesis demonstrates that TF redundancy arises from 96 feed-forward loops regulating the activity of TFs and enzymes driving this process (Taylor-Teeples et al., 2014).

Fig. 2.
Network motifs associated with robustness and plasticity. Several network motifs are associated with robust phenotypes. (A) Coherent feed-forward loops (left), such as those acting in secondary cell wall synthesis (right), maintain robust development ...

Incoherent feed-forward loops consist of a TF regulating another TF or miRNA. Together, they regulate a target in opposite ways, e.g. the first TF activates the target gene and the second TF or miRNA represses it (Fig. 2B). This motif has been implicated in dampening noise in gene and protein expression levels as it exists in regulation of both transcription and translation (Hornstein and Shomron, 2006; Schmiedel et al., 2015). In plants, incoherent feed-forward loops play a role in defence against biotic attacks. For example, an incoherent feed-forward loop exists in pattern-triggered immunity (PTI), which is driven by at least four signalling pathways: the jasmonate, ethylene, phytoalexin-deficient 4 and salicylic acid pathways (Tsuda et al., 2009; Kim et al., 2014). Using a multiple regression model, Kim and co-authors (2014) attributed robustness in PTI to an incoherent feed-forward loop in which biotic attacks induce both ethylene and jasmonate signalling, but the latter is repressed by ethylene signalling. Their argument is supported by a recent study demonstrating robust transcriptional responses to diverse Botrytis cinerea strains in wild-type A. thaliana plants compared with highly variable responses in jasmonate mutants (Rowe et al., 2008). In addition to providing robust responses to diverse biotic attacks, the jasmonate/ethylene incoherent feed-forward loop also prevents ectopic activation of PTI by damping noise (Kim et al., 2014). Both coherent and incoherent feed-forward loops containing miRNA were hypothesized in 2006 to provide robustness in development and reduce noise in gene expression and protein levels (Hornstein and Shomron, 2006). Recent studies have provided several specific examples in different organisms, although systematic analyses are lacking. Such analyses may be forthcoming, especially in the light of detailed occupancy-based TF networks in plants and other organisms and improved predictions for miRNA target sites. Like feed-forward loops, feedback loops have been known for decades to provide robustness in regulatory networks. This motif describes TFs, which regulate their own expression in autoregulatory feedback loops or their respective activator or repressor (Fig. 2C). Autoregulatory feedback loops play an important role in reinforcing cell fate or robust responses to the environment.

Another type of motif that promotes phenotypic robustness through redundancy is the biparallel motif. Biparallel motifs contain four nodes, in which one node regulates two others, and these two downstream factors jointly regulate a single target node (Fig. 2D). A variant of the biparallel motif regulates the termination of floral stem cells, a process critical for plant reproduction. POWERDRESS, a newly identified gene, regulates floral determinacy by promoting the transcription or Pol II occupancy at miR172 and CRC, both of which regulate the floral determinacy gene WUSCHEL. Notably, miR172 and CRC regulate floral determinacy at different times, with mir172 acting in early floral development and CRC acting later (Yumul et al., 2013). Using two functionally redundant pathways to control WUSCHEL expression likely contributes to robustness in floral and shoot determinacy (Laux et al., 1996).

Robust morphological pattern formation allows plants to maintain structure and harvest resources such as light. The positioning of lateral organs, phyllotaxis, is stereotypical for most plants and arises from the periodic initiation of organ primordia at the shoot apical meristem (SAM). The robustness of phyllotaxis appears to depend on a network motif in which one regulator promotes and a second regulator represses organ formation (Fig. 2E). The accumulation of the plant hormone auxin promotes the initiation of organ primordia (Reinhardt et al., 2000), and polar transport of auxin out of cells creates an inhibitory field that precludes cells from forming organ primordia (Reinhardt et al., 2003). Deterministic models of phyllotaxis show that interaction of inhibitory fields among cells composing the SAM explains the spatial and temporal initiation of organ primordia (Mirabet et al., 2012). Mirabet and co-authors (2012) instead modelled the transport of auxin among cells at the SAM as a stochastic process. In these stochastic models, irregularities in phyllotaxis arise, such as the simultaneous initiation of two primordia or reversals in spiral handedness. The same irregularities are observed in phyllotaxis mutants, suggesting that the stochastic model misses parameters. In Drosophila melanogaster embryonic development, two inhibitory fields interact to increase robustness (Okabe-Oho et al., 2009). Inclusion of a second inhibitory field in the model decreased irregularities. The dual field model traced irregularities to particular mis-functions in the SAM, such as the incorrect ranking of primordium age, to cause reversals in spiral handedness.

Indeed, 2 years later, cytokinin signalling emerged as the unknown secondary inhibitory field required for robust phyllotaxis (Besnard et al., 2014). Cytokinins were already known to regulate SAM size, and cytokinin signalling mutants also alter the shape of the SAM (Leibfried et al., 2005). Expression studies implicated AHP6, which is a cytokinin signalling inhibitor, as a strong candidate. Strikingly, mutants in AHP6 exhibit spiral handedness reversals. Closer inspection of ahp6 SAMs reveals simultaneous initiation of multiple primordia, indicating that AHP6 directs the temporal initiation of organ formation. Like auxin, AHP6 was present in a gradient among cells, centred on primordia. Thus, robust phyllotaxis arises through the activity of two inhibitory fields formed by auxin depletion and AHP6 accumulation, which together tightly control the temporal and spatial initiation of organ primordia (Besnard et al., 2014).

Network topology and motif rewiring controls robustness across environments

As network topologies are maintained or altered in response to changing environments, the same is observed for network motifs. The wiring between genes and targets can be altered to reinforce the programmed, plastic shift to another robust phenotypic state. For example, Sullivan et al. (2014) showed several examples of gain and loss of autoregulatory loops throughout photomorphogenesis (Sullivan et al., 2014). One example is EARLY PHYTOCHROME RESPONSIVE 1 (EPR1, REVILLE7), which is one of the first genes to respond to light exposure via phytochrome A and B networks. Upon exposure to light, EPR1 expression is induced and a negative autoregulatory loop emerges, tightly controlling expression (Kuno et al., 2003). Another distinct example is MYC2 (JAI1, JIN1 for jasmonate insensitivity), which is both an inhibitor of seedling photomorphogenesis and a key biotic defence regulator (Dombrecht et al., 2007; Shin et al., 2012; Schweizer et al., 2013). In response to light, MYC2 loses its positive autoregulation and generally loses connectivity. Thus, network topology is readily rewired in response to changing environments.

This plastic wiring also controls robustness within the shade avoidance response, in which connections between key morphological genes are re-wired. The shade avoidance response describes the stereotypical developmental switch that occurs when plants are in competition for light (grown in the shade of other plants) compared with when they are grown in open conditions. During high-light conditions, the underlying network is wired to produce high levels of auxin that lead to robust growth even in the presence of decreased auxin sensitivity (Hersch et al., 2014). In low light, however, the network is re-wired to produce less auxin due to the upregulation of the negative regulator of auxin production HFR1 and concomitantly increases in auxin sensitivity, possibly through increased expression of auxin co-receptors such as AFB1. Comparing high- and low-light networks by modelling demonstrates that the auxin response during high light is more robust (Hersch et al., 2014). In contrast, in low-light conditions plants show a more acute response to auxin, which is enabled by decreased lower robustness. Thus, plants modulate network topology and motifs to shift phenotype between robustness and plasticity in certain environments.

Temporal staggering of signalling produces robust responses

In response to pathogens, plants activate signalling cascades to reduce harm and overwhelm invaders (reviewed in Cui et al., 2015). As a first line of defence, plants recognize general pathogen features through activity of cell surface proteins and mount PTI. However, many pathogens have evolved to circumvent the innate immune response. Thus, plants utilize a back-up system in which specific pathogens elicit an immune response, known as effector-triggered immunity (ETI). In ETI, the innate immune response is raised and amplified, through signalling cascades triggered by the defence hormone salicylic acid (SA), among others. The SA signalling pathway induces the hypersensitive response, which leads to localized cell death at the site of infection, as well as systemic acquired resistance in uninfected parts of the plant. Of course, pathogens have evolved mechanisms to circumvent ETI by neutralizing factors that act in the SA pathway. It is clear that robustness in these mechanisms is critical to withstanding bacterial attack. Tsuda et al. (2009) report that downstream responses of the SA pathway are also activated in an SA-independent fashion. They trace this SA-independent activation to the delayed expression and sustained activity of the kinase MPK3. This is one of the kinases activated by the detection of pathogens in ETI, but activation of its expression is delayed relative to other SA signalling components. Therefore, robustness in ETI, despite SA signalling circumvention, emerges from the temporal staggering of SA signalling, specifically the sustained activity of MAPK3 to form a coherent feed-forward loop. It would be interesting to explore whether the robustness of ETI signalling is primarily due to the staggered, temporal component or the mere redundancy between the SA and MAPK3 signal elicitation, or both.

THOUGHTS ON ESTABLISHED MOLECULAR MECHANISMS THAT PROVIDE ROBUSTNESS AND ENABLE PLASTICITY AND POSSIBLE CANDIDATE MECHANISMS

In several species, robustness across diverse traits has been traced to the function of genes that perform crucial cellular functions and are involved in cellular homeostasis (Queitsch et al., 2002; Levy and Siegal, 2008; Lempe et al., 2012; Bauer et al., 2015). Examples include genes encoding certain protein chaperones (Rutherford and Lindquist, 1998; Queitsch et al., 2002; Yeyati et al., 2007), genes functioning in proper genome maintenance and genes modifying expression of many others (Lehner et al., 2006; Levy and Siegal, 2008). Importantly, such genes may affect diverse specific regulatory networks and motifs without being an integral component of these networks (Whitacre, 2012). We discuss specific examples below.

Hsp90-dependent robustness emerges from the maintenance of genetic networks

Hsp90 is a highly connected and evolutionarily conserved protein chaperone that facilitates the maturation of certain proteins (clients) and keeps them poised for activation. Hsp90 clients are often conserved hub proteins that comprise about 10–20 % of the cellular proteome (Zhao et al., 2005; McClellan et al., 2007; Jarosz and Lindquist, 2010; Gopinath et al., 2014). Hsp90 clients, among them many kinases, TFs and receptors, are conformationally flexible, which is the property that the chaperone is thought to recognize (Taipale et al., 2010). We argue that the well-documented role of Hsp90 in maintaining phenotypic robustness (Rutherford and Lindquist, 1998; Queitsch et al., 2002; Sangster et al., 2007; Yeyati et al., 2007; Jarosz and Lindquist, 2010) is fundamentally an epistasis phenomenon on a network scale. The interaction of Hsp90 with its many clients maintains robustness, yet the chaperone’s property of enabling signal transduction proteins to perceive and transduce signals is also of fundamental importance for phenotypic plasticity, in particular programmed plasticity. Environmental stresses such as increased temperature, drought, oxidative stress and altered conductivity compromise protein folding and increase demands for Hsp90 (Gerspacher et al., 2009; Cid et al., 2010; Park et al., 2010; Pratt et al., 2010; Sekimoto et al., 2010; Stensløkken et al., 2010). Under these conditions, Hsp90 clients will be less functional and the genetic network will be perturbed at many nodes. Hence, it is not surprising that Hsp90 function is environmentally responsive. Hsp90 levels are strongly upregulated in conditions that compromise protein folding to bolster protein homeostasis (Gerspacher et al., 2009; Cid et al., 2010; Park et al., 2010; Pratt et al., 2010; Sekimoto et al., 2010; Stensløkken et al., 2010).

As has been widely reported and reviewed (Rutherford and Lindquist, 1998; Queitsch et al., 2002; Sangster and Queitsch, 2005; Sangster et al., 2008; Jarosz and Lindquist, 2010; Jarosz et al., 2010), Hsp90 also maintains phenotypic robustness in the presence of genetic variation, a phenomenon that gave rise to the term ‘phenotypic capacitor’ (Rutherford and Lindquist, 1998). We view this aspect of Hsp90 function as an extension of the above-described epistasis phenomenon (gene × genome), in which the chaperone epistatically suppresses the phenotypic consequences of polymorphisms in either its clients or their direct and indirect interactors (which may include regulatory sequences) (Jarosz and Lindquist, 2010).

Because perturbing Hsp90 function impairs its numerous clients, it greatly alters network connectivity and wiring (Lempe et al., 2012). Across the human proteome, 9200 of the 70 000 known protein interactions are lost across five cancer cell lines when Hsp90 function is inhibited with potent and specific inhibitors (Echeverria and Picard, 2014). Lost connections are due to aggregation and degradation of misfolded Hsp90 clients as well as the depletion of protein complexes that contain Hsp90 clients. Clearly, this loss of connectivity will lead to network fragmentation (Echeverria and Picard, 2014). A similar loss of connectivity is observed for occupancy-based TF networks upon heat shock (Sullivan et al., 2014).

Perturbing Hsp90 in isogenic plant populations gives rise to a multitude of abnormal phenotypes (Queitsch et al., 2002). It remains unclear whether this diversity of phenotypes arises from the chaperone’s interactions with many specific clients or is due to generally increased noise in the genetic network causing random developmental errors. We speculate that the former, simpler explanation might be true; the published data are consistent with both scenarios. Most developmental decisions, many of which are governed by diverse Hsp90 clients, are thought to adhere to the two-forked checkpoint type, resulting in two possible fates. Let us assume an isogenic population that has to pass through many such Hsp90-dependent forks to complete development. To simplify matters further, all clients have an equal chance of being perturbed when Hsp90 function is inhibited, in which case the alternative fate is produced. Each individual decision at an Hsp90 fork is stochastic; yet, the available developmental paths and spectrum of possible phenotypes is in principle predictable. This simple model would produce the observed wide array of Hsp90-dependent phenotypes among isogenic seedlings (Queitsch et al., 2002), ranging from seeds failing germination to fully developed seedlings with specific organ defects. Applied to a simple quantitative trait like stem length, the significantly wider distribution of stem length in response to Hsp90 perturbation would arise as the overlap of two distributions representing a short and a long stem fate. In fact, there is increasing evidence from other systems that wide distributions can arise as the overlap of two or more sub-distributions. For example, differences in transcriptional bursts among mammalian cells have been recently linked with their cell volumes, presumably to keep transcript concentrations stable (Padovan-Merhar et al., 2015). Returning to Hsp90, let us now assume an isogenic population in which a particular client is sensitized to Hsp90 perturbation by genetic variation. Due the sensitized client, this fork decision is now strongly skewed towards a particular developmental path and phenotype upon Hsp90 perturbation. Indeed, strong genotype-specific phenotypes are observed in plants for both morphological traits and stem length (Queitsch et al., 2002). In short, we propose that Hsp90 perturbation does not simply increase random noise in genetic networks, but rather that the multitude of Hsp90 clients governing different fate decisions generates this impression. An empirical test of this hypothesis is lacking, but could be accomplished by focusing on a specific fate decision or the simple stem-length trait and profiling the molecular properties of phenotypic extremes.

Regulation of chromatin modulates phenotypic robustness

Mechanisms of chromatin regulation have the potential to affect phenotypic robustness because of the centrality of chromatin in regulating gene expression (Tirosh et al., 2010). Proteins involved in DNA methylation, chromatin remodelling and histone modifications are prime candidates for modulators of phenotypic robustness.

Chromatin modifications are proposed to be fast-changing, potentially heritable sources of phenotypic variation. For example, DNA methylation is thought be important in fluctuating environments, because though it is heritable, it is also readily reversible, unlike DNA sequence variation. However, a recent study of genome-wide DNA methylation patterns for almost 100 generations does not show any acceleration in the accumulation of epimutations compared with mutations among plants, even among plants grown in diverse environments (Hagmann et al., 2015). Despite the fact that certain epialleles do influence phenotypes (Weigel and Colot, 2012) and plasticity (Zhang et al., 2013; Kooke et al., 2015), there is little evidence for epimutations as a source of fast-acting heritable phenotypic change (Schmitz et al., 2011; Li et al., 2014a, b; Hagmann et al., 2015). This result supports the older notion that changes in DNA methylation primarily reinforce tissue- and cell-type-specific expression patterns and hence the robustness of cell fate decisions.

Chromatin remodelling genes maintain phenotypic robustness.

Ensuring that chromatin marks and nucleosomes are properly placed is crucial for trait robustness and plasticity. Like Hsp90, regulators of nucleosome positioning, are associated with maintaining robust phenotypes and enabling plasticity. In worms (Lehner et al., 2006), yeast (Levy and Siegal, 2008) and now in plants, SWI/SNF ATPase chromatin remodellers specifically have been identified as regulators of phenotypic robustness. In isogenic A. thaliana, over-expression of the SWI/SNF-type ATPase CHR23 results in increased variability in root length and leaf size in long-day but not short-day conditions (Folta et al., 2014). Long-day conditions typically promote more rapid growth, which may sensitize plants to robustness loss. However, CHR23 loss-of-function mutants show wild-type variance in root length and leaf size, suggesting that CHR23 dosage matters for phenotypic robustness. CHR23 has a close paralogue with many overlapping, essential functions; the double mutant is lethal. We speculate that the right balance of both chromatin-remodelling proteins generates phenotypic robustness. One may imagine that the CHR23 paralogue compensates for loss of CHR23; yet CHR23 overexpression overrides this cross-regulation and perturbs robustness. Notably, the increased phenotypic variation upon CHR23 overexpression correlates with increased variation of expression in genes associated with environmental stress (Folta et al., 2014).

Histone modifications promote robust plastic responses.

In plants, trimethylation of lysine 27 on histone H3 (H3K27me3) is associated with robust silencing of gene expression, which is critical for the transition between vegetative and reproductive growth. Certain A. thaliana strains require cold exposure to induce flowering. This cold exposure silences the expression of FLC, which encodes a potent flowering repressor, due to the accumulation of H3K27me3 marks. The initiation of FLC repression is aided by the long non-coding RNA COOLAIR (Swiezewski et al., 2009). Polycomb-group proteins, which are enzymes with histone methylase activity, also contribute to H3K27me3 at FLC (Angel et al., 2011). In general, Polycomb-group proteins silence gene expression and interact antagonistically with Trithorax-group proteins, which maintain active gene expression (Kennison, 1995). The non-coding RNA COOLAIR and Polycomb-group proteins work independently to generate a threshold of H3K27me3 levels that silences FLC expression and hence triggers flowering (Csorba et al., 2014).

A recent study argues that FLC H3K27me3 silencing translates into a robust transition to flowering due to the utilization of a digital, rather than an analogue, mechanism to monitor temperature (Angel et al., 2015). In the analogue (continuously varying) mechanism model, the probability of adding H3K27me3 modifications at FLC increases with increased cold exposure and that of removal of H3K27me3 modifications decreases. In the digital mechanism model (all or nothing), the number of cells containing enough H3K27me3 modifications to reach the threshold to silence FLC increases irreversibly with increased exposure to cold. The authors found that the latter model provided a better fit for available chromatin immunoprecipitation data for H3K27me3 and FLC expression. Under the digital model, intermittent periods of higher temperature during cold exposure should not affect the length of cold exposure necessary for transition to flowering. In contrast, the analogue model predicts that intermittent warm periods should lengthen the cold exposure necessary to initiate flowering. Testing both models experimentally, the authors found that the timing of flowering was not lengthened by intermittent short periods of higher temperature. The plant’s utilization of a digital all-or-nothing mechanism contributes to the robust transition between vegetative and reproductive growth and its irreversible nature, at least in the annual A. thaliana (Angel et al., 2015). How precisely COOLAIR and Polycomb-group proteins contribute to this digital mechanism of adding H3K27me3 modifications remains to be determined, but their role in FLC silencing makes both likely candidates for the underlying molecular mechanism.

Ribosomal DNA copy number variation as a putative source of phenotypic robustness akin to Hsp90

Ribosomal RNA (rRNA) is encoded in large repetitive DNA arrays, which are highly mutable and prone to copy number variation. Recent studies in humans, mice, worms and plants have uncovered vast variation in estimated rDNA copy number among individuals (Long et al., 2013; Thompson et al., 2013; Gibbons et al., 2014). Across 97 tested humans, rDNA copy number variation is associated with global gene expression changes and altered abundance of mitochondrial DNA (Gibbons et al., 2014). As one may expect, increased rDNA dosage is associated with increased expression of genes related to rDNA expression and function, such as those encoding ribonucleoproteins, various nucleolar proteins and many chromatin-remodelling proteins. Deletions of rDNA arrays in Drosophila melanogaster also modulates global gene expression (Paredes et al., 2011). It is unknown whether rDNA copy number variation affects phenotype primarily through altered rRNA expression or through replication conflicts arising from large arrays of repetitive DNA. The latter mechanism is supported by recent evidence, largely in bacteria and yeast (Merrikh et al., 2011, 2012; Kwan et al., 2013). The former would minimize interest in rDNA as a robustness mechanism in plants because plant rDNA arrays are largely silenced by RdDM (Douet et al., 2009; Benoit et al., 2013). In the absence of this mechanistic knowledge, the observed vast variation in rDNA copy number and its association with global gene expression and altered chromatin remodelling raise the question of whether rDNA dosage affects phenotypic robustness.

rDNA is intricately linked with cellular physiology, stress response and senescence. The encoded rRNAs are essential structural and catalytic components of ribosomes, the protein factories of every living cell (Kobayashi, 2011, 2014). Ribosome biogenesis is by far the most energy-consuming cellular process: rRNA accounts for ~80 % of total transcription, and transcription of genes encoding ribosomal proteins accounts for ~50 % of total Pol II-dependent transcription (Rudra and Warner, 2004; Gibbons et al., 2014). Thousands of large and small ribosome subunits are synthesized every minute in growing cells.

The rDNA arrays are organized within the nucleolus, in which ribosome assembly takes place (Olson et al., 2000). In the last decade, the nucleolus has emerged as a key player in sensing cellular stress and is considered the central hub for coordinating the stress response and maintaining protein homeostasis (Boulon et al., 2010). Many molecular chaperones, including Hsp90, are involved in nucleolar functions (Bański et al., 2010; Boulon et al., 2010). Given the extraordinary importance of rDNA and ribosome biogenesis, it is not surprising that a host of human diseases is associated with mutations in rDNA expression, rRNA processing and ribosome biogenesis (Freed et al., 2010). Several of these diseases show large differences in expressivity among carriers of the same mutation (Freed et al., 2010), possibly because of underlying, but uncharacterized, differences in rDNA dosage, which results in differential robustness to these mutations or aggravating environmental perturbations (Gibbons et al., 2015).

Although rDNA is traditionally under-studied in multicellular organisms, studies in yeast provide ample support for the importance of rDNA dosage in organismal fitness and lifespan (Kobayashi et al., 1998; Kobayashi, 2011, 2014; Kwan et al., 2013). Yeast cells with artificially increased or decreased rDNA copy number rapidly revert to wild-type rDNA copy number, suggesting strong selection (Kobayashi et al., 1998). The stability and replication of rDNA affect yeast replicative lifespan (Kobayashi et al., 1998; Kobayashi, 2011, 2014; Kwan et al., 2013). Together, the cumulative evidence of prior studies calls for a systematic and accurate assessment of rDNA copy number and its effect on phenotypic robustness.

CONCLUSIONS

Plants are remarkable – they accommodate environmental changes both predictable and unpredictable every day without the benefit of mobility and a centralized response system and while saddled with the task of continuous development. Although the fundamental principles governing robustness of development and environmental responses are conserved between animals and plants, important details differ, such as the well-documented ability of plants to tolerate WGD and hybridization. In the absence of systematic studies of network topology and motifs, we can only speculate whether strategies to ensure robustness and plasticity differ only quantitatively or also qualitatively. At a superficial glance, gene families involved in environmental responses, such as the heat shock TFs and various chaperones, have considerably expanded in plants compared with animals. Similar considerations apply among plants; one may expect different strategies between perennials and annuals, with the latter primarily investing in large numbers of offspring rather than continued robustness over often long periods of time. Lastly, even among annuals life history and ecological niche will influence where traits will fall on the robustness-to-plasticity spectrum.

As this review demonstrates, there are plenty of specific examples that highlight different aspects of how network architecture contributes to robustness and plasticity. Efforts to systematically probe genetic architecture in any organism are rare, and possibly even rarer in plant research with its limited resources. As a community we may want to move away from our standard strategy of throwing out outliers, as commonly done in plant pathology or circadian research. These outliers will hold essential information about altered wiring of underlying pathways and hence illuminate the network architecture maintaining trait robustness. In our quest to understand the universal principles governing trait architecture, its robustness, plasticity and ultimately evolvability, we should not lose sight of discovering mechanistic underpinnings. Outliers may hold clues to mechanisms; advanced computational and experimental approaches such as single-cell-derived networks hold promise for their discovery. Lastly, our new-found prowess in mechanistic and systems-based experimental approaches will not yield universal principles, assuming these exist, unless we apply them in the field and include diverse species in our studies.

ACKNOWLEDGEMENTS

This work was supported by the National Institutes of Health (DP2OD008371 to C.Q.), the National Science Foundation (DBI grant 0820580 to D.J.K., MCB grant 1330337 to D.J.K.), the USDA National Institute of Food and Agriculture (Hatch project number CA-D-PLS-7033-H to D.J.K.) and by the Danish National Research Foundation (DNRF99 to D.J.K.).

LITERATURE CITED

  • Albert R, Jeong H, Barabasi A-L. 2000. Error and attack tolerance of complex networks. Nature 406: 378–382. [PubMed]
  • Alon U. 2003. Biological networks: the tinkerer as an engineer. Science 301: 1866–1867. [PubMed]
  • Angel A, Song J, Dean C, Howard M. 2011. A Polycomb-based switch underlying quantitative epigenetic memory. Nature 476: 105–108. [PubMed]
  • Angel A, Song J, Yang H, Questa JI, Dean C, Howard M. 2015. Vernalizing cold is registered digitally at FLC. Proceedings of the National Academy of Sciences of the USA 112: 4146–4151. [PubMed]
  • Anwer MU, Boikoglou E, Herrero E, et al. 2014. Natural variation reveals that intracellular distribution of ELF3 protein is associated with function in the circadian clock. eLife. doi: 10.7554/eLife.02206. [PMC free article] [PubMed]
  • Arabidopsis Interactome Mapping Consortium. 2011. Evidence for network evolution in an arabidopsis interactome map. Science 333: 601–607. [PMC free article] [PubMed]
  • Auffray JC, Debat V, Alibert P. 1999. Shape asymmetry and developmental stability. On growth and form: spatio-temporal pattern formation in biology . Chichester: Wiley, 309–324.
  • Bański P, Kodiha M, Stochaj U. 2010. Chaperones and multitasking proteins in the nucleolus: networking together for survival? Trends in Biochemical Sciences 35: 361–367. [PubMed]
  • Bauer CR, Li S, Siegal ML. 2015. Essential gene disruptions reveal complex relationships between phenotypic robustness, pleiotropy, and fitness. Molecular Systems Biology 11, doi:10.15252/msb.20145264. [PMC free article] [PubMed]
  • Benoit M, Layat E, Tourmente S, Probst A V. 2013. Heterochromatin dynamics during developmental transitions in Arabidopsis—a focus on ribosomal DNA loci. Gene 526: 39–45. [PubMed]
  • Besnard F, Refahi Y, Morin V, et al. 2014. Cytokinin signalling inhibitory fields provide robustness to phyllotaxis. Nature 505: 417–421. [PubMed]
  • Boulon S, Westman BJ, Hutten S, Boisvert F-M, Lamond AI. 2010. The nucleolus under stress. Molecular Cell 40: 216–227. [PMC free article] [PubMed]
  • Bowman JL, Smyth DR, Meyerowitz EM. 1991. Genetic interactions among floral homeotic genes of Arabidopsis. Development 112: 1–20. [PubMed]
  • Box MS, Huang BE, Domijan M, et al. 2014. ELF3 controls thermoresponsive growth in Arabidopsis. Current Biology 25: 194–199. [PubMed]
  • Brodribb TJ, Feild TS. 2010. Leaf hydraulic evolution led a surge in leaf photosynthetic capacity during early angiosperm diversification. Ecology Letters 13: 175–183. [PubMed]
  • Brodribb TJ, Jordan GJ. 2011. Water supply and demand remain balanced during leaf acclimation of Nothofagus cunninghamii trees. New Phytologist 192: 437–448. [PubMed]
  • Carlsbecker A, Lee J-Y, Roberts CJ, Dettmer J, et al. 2010. Cell signalling by microRNA165/6 directs gene dose-dependent root cell fate. Nature 465: 316–321. [PMC free article] [PubMed]
  • Casanueva MO, Burga A, Lehner B. 2012. Fitness trade-offs and environmentally induced mutation buffering in isogenic C . elegans. Science 335: 82–85. [PubMed]
  • Cassidy JJ, Jha AR, Posadas DM, et al. 2013. miR-9a minimizes the phenotypic impact of genomic diversity by buffering a transcription factor. Cell 155: 1556–1567. [PMC free article] [PubMed]
  • Chae L, Kim T, Nilo-Poyanco R, Rhee SY. 2014. Genomic signatures of specialized metabolism in plants. Science 344: 510–513. [PubMed]
  • Cid C, Garcia-Descalzo L, Casado-Lafuente V, Amils R, Aguilera A. 2010. Proteomic analysis of the response of an acidophilic strain of Chlamydomonas sp. (Chlorophyta) to natural metal-rich water. Proteomics 10: 2026–2036. [PubMed]
  • Clauss MJ, Venable DL. 2000. Seed germination in desert annuals: an empirical test of adaptive bet hedging. American Naturalist 155: 168–186. [PubMed]
  • Conant GC, Wolfe KH. 2008. Turning a hobby into a job: how duplicated genes find new functions. Nature Reviews. Genetics 9: 938–950. [PubMed]
  • Csorba T, Questa JI, Sun Q, Dean C. 2014. Antisense COOLAIR mediates the coordinated switching of chromatin states at FLC during vernalization. Proceedings of the National Academy of Sciences of the USA 111: 16160–16165. [PubMed]
  • Cui H, Tsuda K, Parker JE. 2015. Effector-triggered immunity: from pathogen perception to robust defense. Annual Review of Plant Biology 66: 487–511. [PubMed]
  • Debat V, David P. 2001. Mapping phenotypes: canalization, plasticity and developmental stability. Trends in Ecology & Evolution 16: 555–561.
  • Devlin PF, Robson PRH, Patel SR, Goosey L, Sharrock RA, Whitelam GC. 1999. Phytochrome D acts in the shade-avoidance syndrome in Arabidopsis by controlling elongation growth and flowering time. Plant Physiology 119: 909–916. [PubMed]
  • Diss G, Ascencio D, DeLuna A, Landry CR. 2014. Molecular mechanisms of paralogous compensation and the robustness of cellular networks. Journal of Experimental Zoology. Part B, Molecular and Developmental Evolution 322: 488–499. [PubMed]
  • Dombrecht B, Xue GP, Sprague SJ, et al. 2007. MYC2 differentially modulates diverse jasmonate-dependent functions in Arabidopsis. Plant Cell 19: 2225–2245. [PubMed]
  • Dorantes AR, Sánchez JLM. 2006. Biochemical changes and development of leaf tracheary elements of a desert plant Fouquieria splendens ssp. breviflora exposed to water stress. Physiologia Plantarum 127: 650–657.
  • Douet J, Tutois S, Tourmente S. 2009. A Pol V-mediated silencing, independent of RNA-directed DNA methylation, applies to 5S rDNA. PLoS Genet 5: e1000690. [PMC free article] [PubMed]
  • Echeverria PC, Picard D. 2014. A global view of the proteome perturbations by Hsp90 inhibitors. In: Houry W, editor. ed. The Molecular Chaperones Interaction Networks in Protein Folding and Degradation. New York: Springer, 133–149.
  • Ehrlich PR, Raven PH. 1964. Butterflies and plants: a study in coevolution. Evolution 18: 586–608.
  • Fawcett JA, Maere S, Van de Peer Y. 2009. Plants with double genomes might have had a better chance to survive the Cretaceous–Tertiary extinction event. Proceedings of the National Academy of Sciences of the USA 106: 5737–5742. [PubMed]
  • Folta A, Severing EI, Krauskopf J, et al. 2014. Over-expression of Arabidopsis AtCHR23 chromatin remodeling ATPase results in increased variability of growth and gene expression. BMC Plant Biology 14: 76. [PMC free article] [PubMed]
  • Freed EF, Bleichert F, Dutca LM, Baserga SJ. 2010. When ribosomes go bad: diseases of ribosome biogenesis. Molecular BioSystems 6: 481–493. [PMC free article] [PubMed]
  • Fu J, Keurentjes JJ, Bouwmeester H, et al. 2009. System-wide molecular evidence for phenotypic buffering in Arabidopsis. Nature Genetics 41: 166–167. [PubMed]
  • Gerspacher C, Scheuber U, Schiera G, Proia P, Gygax D, Di Liegro I. 2009. The effect of cadmium on brain cells in culture. International Journal of Molecular Medicine 24: 311–318. [PubMed]
  • Gibbons JG, Branco AT, Yu S, Lemos B. 2014. Ribosomal DNA copy number is coupled with gene expression variation and mitochondrial abundance in humans. Nature Communications 5, doi:10.1038/ncomms5850. [PubMed]
  • Gibbons JG, Branco AT, Godinho SA, Yu S, Lemos B. 2015. Concerted copy number variation balances ribosomal DNA dosage in human and mouse genomes. Proceedings of the National Academy of Sciences of the USA 112: 2485–2490. [PubMed]
  • Gibson G. 2009. Decanalization and the origin of complex disease. Nature Reviews. Genetics 10: 134–140. [PubMed]
  • Gomez-Mestre I, Jovani R. 2013. A heuristic model on the role of plasticity in adaptive evolution: plasticity increases adaptation, population viability and genetic variation. Proceedings of the Royal Society of London. Series B: Biological Sciences 280: 20131869. [PMC free article] [PubMed]
  • Gopinath RK, You S-T, Chien K-Y, et al. 2014. The Hsp90-dependent proteome is conserved and enriched for hub proteins with high levels of protein–protein connectivity. Genome Biology and Evolution 6: 2851–2865. [PMC free article] [PubMed]
  • Gu Z, Rifkin SA, White KP, Li W-H. 2004. Duplicate genes increase gene expression diversity within and between species. Nature Genetics 36: 577–579. [PubMed]
  • Ha M, Li W-H, Chen ZJ. 2007. External factors accelerate expression divergence between duplicate genes. Trends in Genetics 23: 162–166. [PMC free article] [PubMed]
  • Ha M, Kim E-D, Chen ZJ. 2009a. Duplicate genes increase expression diversity in closely related species and allopolyploids. Proceedings of the National Academy of Sciences of the USA 106: 2295–2300. [PubMed]
  • Ha M, Lu J, Tian L, et al. 2009b. Small RNAs serve as a genetic buffer against genomic shock in Arabidopsis interspecific hybrids and allopolyploids. Proceedings of the National Academy of Sciences of the USA 106: 17835–17840. [PubMed]
  • Hagmann J, Becker C, Müller J, et al. 2015. Century-scale methylome stability in a recently diverged Arabidopsis thaliana lineage. PLoS Genetics 11: e1004920. [PMC free article] [PubMed]
  • Hall MC, Dworkin I, Ungerer MC, Purugganan M. 2007. Genetics of microenvironmental canalization in Arabidopsis thaliana. Proceedings of the National Academy of Sciences of the USA 104: 13717–13722. [PubMed]
  • Hanada K, Kuromori T, Myouga F, Toyoda T, Li W-H, Shinozaki K. 2009. Evolutionary persistence of functional compensation by duplicate genes in Arabidopsis. Genome Biology and Evolution 1: 409–414. [PMC free article] [PubMed]
  • Hanada K, Sawada Y, Kuromori T, et al. 2011. Functional compensation of primary and secondary metabolites by duplicate genes in Arabidopsis thaliana. Molecular Biology and Evolution 28: 377–382. [PMC free article] [PubMed]
  • Hansen BG, Kliebenstein DJ, Halkier BA. 2007. Identification of a flavin - monooxygenase as the S - oxygenating enzyme in aliphatic glucosinolate biosynthesis in Arabidopsis. Plant Journal 50: 902–910. [PubMed]
  • Hansen BG, Kerwin RE, Ober JA, et al. 2008. A novel 2-oxoacid-dependent dioxygenase involved in the formation of the goiterogenic 2-hydroxybut-3-enyl glucosinolate and generalist insect resistance in Arabidopsis. Plant Physiology 148: 2096–2108. [PubMed]
  • Harper JL. 1980. Plant demography and ecological theory. Oikos 35: 244–253.
  • Hartmann T. 2007. From waste products to ecochemicals: fifty years research of plant secondary metabolism. Phytochemistry 68: 2831–2846. [PubMed]
  • Hersch M, Lorrain S, de Wit M, et al. 2014. Light intensity modulates the regulatory network of the shade avoidance response in Arabidopsis. Proceedings of the National Academy of Sciences of the USA 111: 6515–6520. [PubMed]
  • Hofberger JA, Lyons E, Edger PP, Pires JC, Schranz ME. 2013. Whole genome and tandem duplicate retention facilitated glucosinolate pathway diversification in the mustard family. Genome Biology and Evolution 5: 2155–2173. [PMC free article] [PubMed]
  • Hornstein E, Shomron N. 2006. Canalization of development by microRNAs. Nature Genetics 38: S20–S24. [PubMed]
  • Jarosz D, Lindquist S. 2010. Hsp90 and environmental stress transform the adaptive value of natural genetic variation. Science 330: 1820–1824. [PMC free article] [PubMed]
  • Jarosz D, Taipale M, Lindquist SL. 2010. Protein homeostasis and the phenotypic manifestation of genetic diversity: principles and mechanisms. Annual Review of Genetics 44: 189–216. [PubMed]
  • Jiang J, Zhang C, Wang X. 2015. A recently evolved isoform of the transcription factor BES1 promotes brassinosteroid signaling and development in Arabidopsis thaliana. Plant Cell 27: 361–374. [PubMed]
  • Jimenez-Gomez JM, Corwin JA, Joseph B, Maloof JN, Kliebenstein DJ. 2011. Genomic analysis of QTLs and genes altering natural variation in stochastic noise. PLoS Genetics 7: e1002295. [PMC free article] [PubMed]
  • Kawecki TJ. 2000. The evolution of genetic canalization under fluctuating selection. Evolution 54: 1–12. [PubMed]
  • Kennison JA. 1995. The Polycomb and trithorax group proteins of Drosophila: trans-regulators of homeotic gene function. Annual Review of Genetics 29: 289–303. [PubMed]
  • Kim Y, Tsuda K, Igarashi D, et al. 2014. Mechanisms underlying robustness and tunability in a plant immune signaling network. Cell Host & Microbe 15: 84–94. [PMC free article] [PubMed]
  • Kliebenstein DJ. 2008. A role for gene duplication and natural variation of gene expression in the evolution of metabolism. PLoS One 3: e1838. [PMC free article] [PubMed]
  • Kliebenstein D, Lambrix V, Reichelt M, Gershenzon J, Mitchell-Olds T. 2001. Gene duplication and the diversification of secondary metabolism: side chain modification of glucosinolates in Arabidopsis thaliana. Plant Cell 13: 681–693. [PubMed]
  • Kobayashi T. 2011. Regulation of ribosomal RNA gene copy number and its role in modulating genome integrity and evolutionary adaptability in yeast. Cellular and Molecular Life Sciences 68: 1395–1403. [PMC free article] [PubMed]
  • Kobayashi T. 2014. Ribosomal RNA gene repeats, their stability and cellular senescence. Proceedings of the Japan Academy. Series B, Physical and Biological Sciences 90: 119. [PMC free article] [PubMed]
  • Kobayashi T, Heck DJ, Nomura M, Horiuchi T. 1998. Expansion and contraction of ribosomal DNA repeats in Saccharomyces cerevisiae: requirement of replication fork blocking (Fob1) protein and the role of RNA polymerase I. Genes & Development 12: 3821–3830. [PubMed]
  • Kooke R, Johannes F, Wardenaar R, et al. 2015. Epigenetic basis of morphological variation and phenotypic plasticity in Arabidopsis thaliana. Plant Cell 27: 337–348. [PubMed]
  • Krizek BA. 2009. Making bigger plants: key regulators of final organ size. Current Opinion in Plant Biology 12: 17–22. [PubMed]
  • Kroymann J, Donnerhacke S, Schnabelrauch D, Mitchell-Olds T. 2003. Evolutionary dynamics of an Arabidopsis insect resistance quantitative trait locus. Proceedings of the National Academy of Sciences of the USA 100: 14587–14592. [PubMed]
  • Kuno N, Møller SG, Shinomura T, Xu X, Chua N-H, Furuya M. 2003. The novel MYB protein EARLY-PHYTOCHROME-RESPONSIVE1 is a component of a slave circadian oscillator in Arabidopsis. Plant Cell 15: 2476–2488. [PubMed]
  • Kwan EX, Foss EJ, Tsuchiyama S, et al. 2013. A natural polymorphism in rDNA replication origins links origin activation with calorie restriction and lifespan. PLoS Genetics 9: e1003329. [PMC free article] [PubMed]
  • Lachowiec J, Lemus T, Thomas JH, Murphy PJM, Nemhauser JL, Queitsch C. 2013. The protein chaperone HSP90 can facilitate the divergence of gene duplicates. Genetics 193: 1269–1277. [PubMed]
  • Laux T, Mayer KF, Berger J, Jurgens G. 1996. The WUSCHEL gene is required for shoot and floral meristem integrity in Arabidopsis. Development 122: 87–96. [PubMed]
  • Lehner B, Crombie C, Tischler J, Fortunato A, Fraser AG. 2006. Systematic mapping of genetic interactions in Caenorhabditis elegans identifies common modifiers of diverse signaling pathways. Nature Genetics 38: 896–903. [PubMed]
  • Leibfried A, To JPC, Busch W, et al. 2005. WUSCHEL controls meristem function by direct regulation of cytokinin-inducible response regulators. Nature 438: 1172–1175. [PubMed]
  • Lempe J, Lachowiec J, Sullivan AM, Queitsch C. 2012. Molecular mechanisms of robustness in plants. Current Opinion in Plant Biology 16: 62–69. [PMC free article] [PubMed]
  • Levy SF, Siegal ML. 2008. Network hubs buffer environmental variation in Saccharomyces cerevisiae. PLoS Biology 6: e264. [PubMed]
  • Li Q, Eichten SR, Hermanson PJ, Springer NM. 2014a. Inheritance patterns and stability of DNA methylation variation in maize near-isogenic lines. Genetics 196: 667–676. [PubMed]
  • Li Q, Eichten SR, Hermanson PJ, et al. 2014b. Genetic perturbation of the maize methylome. Plant Cell 26: 4602–4616. [PubMed]
  • Li J, Hansen BG, Ober JA, Kliebenstein DJ, Halkier BA. 2008. Subclade of flavin-monooxygenases involved in aliphatic glucosinolate biosynthesis. Plant Physiology 148: 1721–1733. [PubMed]
  • Liu Z, Adams KL. 2007. Expression partitioning between genes duplicated by polyploidy under abiotic stress and during organ development. Current Biology 17: 1669–1674. [PubMed]
  • Liu XL, Covington MF, Fankhauser C, Chory J, Wagner DR. 2001. ELF3 encodes a circadian clock-regulated nuclear protein that functions in an Arabidopsis PHYB signal transduction pathway. Plant Cell 13: 1293–1304. [PubMed]
  • Long Q, Rabanal FA, Meng D, et al. 2013. Massive genomic variation and strong selection in Arabidopsis thaliana lines from Sweden. Nature Genetics 45: 884–890. [PMC free article] [PubMed]
  • Ma S, Shah S, Bohnert HJ, Snyder M, Dinesh-Kumar SP. 2013. Incorporating motif analysis into gene co-expression networks reveals novel modular expression pattern and new signaling pathways. PLoS Genetics 9: e1003840. [PMC free article] [PubMed]
  • Ma S, Bohnert HJ, Dinesh-Kumar SP. 2015. AtGGM2014, an Arabidopsis gene co-expression network for functional studies. Science China. Life Sciences 58: 276–286. [PubMed]
  • McClellan AJ, Xia Y, Deutschbauer AM, Davis RW, Gerstein M, Frydman J. 2007. Diverse cellular functions of the Hsp90 molecular chaperone uncovered using systems approaches. Cell 131: 121–135. [PubMed]
  • Melzer R, Härter A, Rümpler F, et al. 2014. DEF-and GLO-like proteins may have lost most of their interaction partners during angiosperm evolution. Annals of Botany 114: 1431–1443 [PMC free article] [PubMed]
  • Merrikh H, Machón C, Grainger WH, Grossman AD, Soultanas P. 2011. Co-directional replication-transcription conflicts lead to replication restart. Nature 470: 554–557. [PMC free article] [PubMed]
  • Merrikh H, Zhang Y, Grossman AD, Wang JD. 2012. Replication–transcription conflicts in bacteria. Nature Reviews. Microbiology 10: 449–458. [PMC free article] [PubMed]
  • Mirabet V, Besnard F, Vernoux T, Boudaoud A. 2012. Noise and robustness in phyllotaxis. PLoS Computational Biology 8: e1002389. [PMC free article] [PubMed]
  • Mizuno T, Nomoto Y, Oka H, et al. 2014. Ambient temperature signal feeds into the circadian clock transcriptional circuitry through the EC night-time repressor in Arabidopsis thaliana. Plant and Cell Physiology 55: 958–976. [PubMed]
  • Mukhtar MS, Carvunis A-R, Dreze M, et al. 2011. Independently evolved virulence effectors converge onto hubs in a plant immune system network. Science 333: 596–601. [PMC free article] [PubMed]
  • Nieto C, López-Salmerón V, Davière J-M, Prat S. 2015. ELF3-PIF4 interaction regulates plant growth independently of the evening complex. Current Biology 25: 187–193. [PubMed]
  • Nusinow DA, Helfer A, Hamilton EE, et al. 2011. The ELF4-ELF3-LUX complex links the circadian clock to diurnal control of hypocotyl growth. Nature 475: 398–402. [PMC free article] [PubMed]
  • Okabe-Oho Y, Murakami H, Oho S, Sasai M. 2009. Stable, precise, and reproducible patterning of bicoid and hunchback molecules in the early Drosophila embryo. PLoS Computational Biology 5: e1000486. [PMC free article] [PubMed]
  • Olson MOJ, Dundr M, Szebeni A. 2000. The nucleolus: an old factory with unexpected capabilities. Trends in Cell Biology 10: 189–196. [PubMed]
  • Padovan-Merhar O, Nair GP, Biaesch AG, et al. 2015. Single mammalian cells compensate for differences in cellular volume and DNA copy number through independent global transcriptional mechanisms. Molecular Cell 58: 339–352. [PMC free article] [PubMed]
  • Paredes S, Branco AT, Hartl DL, Maggert KA, Lemos B. 2011. Ribosomal DNA deletions modulate genome-wide gene expression: “rDNA-sensitive” genes and natural variation. PLoS Genetics 7: e1001376. [PMC free article] [PubMed]
  • Park K, Park J, Kim J, Kwak I., -S 2010. Biological and molecular responses of Chironomus riparius (Diptera, Chironomidae) to herbicide 2,4-D (2, 4-dichlorophenoxyacetic acid). Comparative Biochemistry and Physiology Part C: Toxicology & Pharmacology 151: 439–446. [PubMed]
  • Payne JL, Wagner A. 2014. The robustness and evolvability of transcription factor binding sites. Science 343: 875–877. [PubMed]
  • Pigliucci M. 2001. Phenotypic plasticity: beyond nature and nurture. Baltimore: JHU Press.
  • Pratt WB, Morishima Y, Peng H-M, Osawa Y. 2010. Proposal for a role of the Hsp90/Hsp70-based chaperone machinery in making triage decisions when proteins undergo oxidative and toxic damage. Experimental Biology and Medicine 235: 278–289. [PMC free article] [PubMed]
  • Price CA, Weitz JS. 2014. Costs and benefits of reticulate leaf venation. BMC Plant Biology 14: 234. [PMC free article] [PubMed]
  • Queitsch C, Sangster T, Lindquist S. 2002. Hsp90 as a capacitor of phenotypic variation. Nature 417: 618–624. [PubMed]
  • Reinhardt D, Mandel T, Kuhlemeier C. 2000. Auxin regulates the initiation and radial position of plant lateral organs. Plant Cell 12: 507–518. [PubMed]
  • Reinhardt D, Pesce E-R, Stieger P, et al. 2003. Regulation of phyllotaxis by polar auxin transport. Nature 426: 255–260. [PubMed]
  • Renny-Byfield S, Gallagher JP, Grover CE, et al. 2014. Ancient gene duplicates in Gossypium (cotton) exhibit near-complete expression divergence. Genome Biology and Evolution 6: 559–571. [PMC free article] [PubMed]
  • Rieseberg LH, Carney SE. 1998. Plant hybridization. New Phytologist 140: 599–624.
  • Rodrigo G, Carrera J, Ruiz-Ferrer V, et al. 2012. A meta-analysis reveals the commonalities and differences in Arabidopsis thaliana response to different viral pathogens. PLoS One 7: e40526. [PMC free article] [PubMed]
  • Roth-Nebelsick A, Uhl D, Mosbrugger V, Kerp H. 2001. Evolution and function of leaf venation architecture: a review. Annals of Botany 87: 553–566.
  • Rowe HC, Hansen BG, Halkier BA, Kliebenstein DJ. 2008. Biochemical networks and epistasis shape the Arabidopsis thaliana metabolome. Plant Cell 20: 1199–1216. [PubMed]
  • Rudra D, Warner JR. 2004. What better measure than ribosome synthesis? Genes & Development 18: 2431–2436. [PubMed]
  • Rutherford SL, Lindquist SL. 1998. Hsp90 as a capacitor for morphological evolution. Nature 396: 336–342. [PubMed]
  • Sack L, Frole K. 2006. Leaf structural diversity is related to hydraulic capacity in tropical rain forest trees. Ecology 87: 483–491. [PubMed]
  • Sack L, Scoffoni C. 2013. Leaf venation: structure, function, development, evolution, ecology and applications in the past, present and future. New Phytologist 198: 983–1000. [PubMed]
  • Sangster TA, Queitsch C. 2005. The HSP90 chaperone complex, an emerging force in plant development and phenotypic plasticity. Current Opinion in Plant Biology 8: 86–92. [PubMed]
  • Sangster TA, Bahrami A, Wilczek A, et al. 2007. Phenotypic diversity and altered environmental plasticity in Arabidopsis thaliana with reduced Hsp90 levels. PLoS One 2: e648. [PMC free article] [PubMed]
  • Sangster T, Salathia N, Lee HNN, et al. 2008. HSP90-buffered genetic variation is common in Arabidopsis thaliana. Proceedings of the National Academy of Sciences of the USA 105: 2969–2974. [PubMed]
  • Schmiedel JM, Klemm SL, Zheng Y, et al. 2015. MicroRNA control of protein expression noise. Science 348: 128–132. [PubMed]
  • Schmitz RJ, Schultz MD, Lewsey MG, et al. 2011. Transgenerational epigenetic instability is a source of novel methylation variants. Science 334: 369–373. [PMC free article] [PubMed]
  • Schweizer F, Fernández-Calvo P, Zander M, et al. 2013. Arabidopsis basic helix-loop-helix transcription factors MYC2, MYC3, and MYC4 regulate glucosinolate biosynthesis, insect performance, and feeding behavior. Plant Cell 25: 3117–3132. [PubMed]
  • Sekimoto T, Oda T, Pozo FM, et al. 2010. The molecular chaperone Hsp90 regulates accumulation of DNA polymerase η at replication stalling sites in UV-irradiated cells. Molecular Cell 37: 79–89. [PubMed]
  • Shin J, Heidrich K, Sanchez-Villarreal A, Parker JE, Davis SJ. 2012. TIME FOR COFFEE represses accumulation of the MYC2 transcription factor to provide time-of-day regulation of jasmonate signaling in Arabidopsis. Plant Cell 24: 2470–2482. [PubMed]
  • Sønderby IE, Burow M, Rowe HC, Kliebenstein DJ, Halkier BA. 2010. A complex interplay of three R2R3 MYB transcription factors determines the profile of aliphatic glucosinolates in Arabidopsis. Plant Physiology 153: 348–363. [PubMed]
  • Stensløkken K-O, Ellefsen S, Larsen HK, Vaage J, Nilsson GE. 2010. Expression of heat shock proteins in anoxic crucian carp (Carassius carassius): support for cold as a preparatory cue for anoxia. American Journal of Physiology. Regulatory, Integrative and Comparative Physiology 298: R1499–R1508. [PubMed]
  • Sullivan AM, Arsovski AA, Lempe J, et al. 2014. Mapping and dynamics of regulatory DNA and transcription factor networks in A . thaliana. Cell Reports 8: 2015–2030. [PubMed]
  • Swiezewski S, Liu F, Magusin A, Dean C. 2009. Cold-induced silencing by long antisense transcripts of an Arabidopsis Polycomb target. Nature 462: 799–802. [PubMed]
  • Taipale M, Jarosz D, Lindquist S. 2010. HSP90 at the hub of protein homeostasis: emerging mechanistic insights. Nature Reviews. Molecular Cell Biology 11: 515–528. [PubMed]
  • Taylor-Teeples M, Lin L, de Lucas M, et al. 2014. An Arabidopsis gene regulatory network for secondary cell wall synthesis. Nature 517: 571–575. [PMC free article] [PubMed]
  • Theißen G. 2001. Development of floral organ identity: stories from the MADS house. Current Opinion in Plant Biology 4: 75–85. [PubMed]
  • Thompson O, Edgley M, Strasbourger P, et al. 2013. The million mutation project: a new approach to genetics in Caenorhabditis elegans. Genome Research 23: 1749–1762. [PubMed]
  • Tirosh I, Reikhav S, Sigal N, Assia Y, Barkai N. 2010. Chromatin regulators as capacitors of interspecies variations in gene expression. Molecular Systems Biology 6: 435. [PMC free article] [PubMed]
  • Tsuda K, Sato M, Stoddard T, Glazebrook J, Katagiri F. 2009. Network properties of robust immunity in plants. PLoS Genetics 5: e1000772. [PMC free article] [PubMed]
  • Waddington CH. 1942. Canalization of development and the inheritance of acquired characters. Nature 150: 563–565.
  • Waddington CH. 1961. Genetic assimilation. Advances in Genetics 10: 257–293. [PubMed]
  • Wagner GP, Booth G, Bagheri-Chaichian H. 1997. A population genetic theory of canalization. Evolution 51: 329–347.
  • Wang, Nakano T, Gendron J, et al. 2002. Nuclear-localized BZR1 mediates brassinosteroid-induced growth and feedback suppression of brassinosteroid biosynthesis. Developmental Cell 2: 505–513. [PubMed]
  • Weigel D, Colot V. 2012. Epialleles in plant evolution. Genome Biology 13: 249. [PMC free article] [PubMed]
  • Welch ME, Rieseberg LH. 2002. Habitat divergence between a homoploid hybrid sunflower species, Helianthus paradoxus (Asteraceae), and its progenitors. American Journal of Botany 89: 472–478. [PubMed]
  • Whitacre JM. 2012. Biological robustness: paradigms, mechanisms, and systems principles. Frontiers in Genetics 3: 67. [PMC free article] [PubMed]
  • Xie Q, Cvekl A. 2011. The orchestration of mammalian tissue morphogenesis through a series of coherent feed-forward loops. Journal of Biological Chemistry 286: 43259–43271. [PMC free article] [PubMed]
  • Ye H, Li L, Guo H, Yin Y. 2012. MYBL2 is a substrate of GSK3-like kinase BIN2 and acts as a corepressor of BES1 in brassinosteroid signaling pathway in Arabidopsis. Proceedings of the National Academy of Sciences of the USA 109: 20142–20147. [PubMed]
  • Yeyati PLL, Bancewicz RMM, Maule J, Van Heyningen V. 2007. Hsp90 selectively modulates phenotype in vertebrate development. PLoS Genet 3: e43. [PMC free article] [PubMed]
  • Yin, Wang, Mora-Garcia S, Li J, et al. 2002. BES1 accumulates in the nucleus in response to brassinosteroids to regulate gene expression and promote stem elongation. Cell 109: 181–191. [PubMed]
  • Yumul, RE, Kim YJ, Liu X, Wang R, Ding J, Xiao L, Chen XY. 2013. POWERDRESS and diversified expression of the MIR172 gene family bolster the floral stem cell network. PLoS Genetics 9: e1003218 doi:10.1371/journal.pgen.1003218. [PMC free article] [PubMed]
  • Zagotta MT, Hicks KA, Jacobs CI, Young JC, Hangarter RP, Meeks-Wagner DR. 1996. The Arabidopsis ELF3 gene regulates vegetative photomorphogenesis and the photoperiodic induction of flowering. Plant Journal 10: 691–702. [PubMed]
  • Zhang Y, Fischer M, Colot V, Bossdorf O. 2013. Epigenetic variation creates potential for evolution of plant phenotypic plasticity. New Phytologist 197: 314–322. [PubMed]
  • Zhao R, Davey M, Hsu YC, et al. 2005. Navigating the chaperone network: an integrative map of physical and genetic interactions mediated by the hsp90 chaperone. Cell 120: 715–727. [PubMed]

Articles from Annals of Botany are provided here courtesy of Oxford University Press