Seed maturation: Simplification of control networks in plants

Plant Science 252 (2016) 335–346

Contents lists available at ScienceDirect

Plant Science journal homepage: www.elsevier.com/locate/plantsci

Review article

Seed maturation: Simplification of control networks in plants Martine Devic ∗ , Thomas Roscoe Régulations Epigénétiques et Développement de la Graine, ERL 3500 CNRS-IRD UMR DIADE, Centre IRD de Montpellier, 911 avenue Agropolis BP64501, 34394, Montpellier, France

a r t i c l e

i n f o

Article history: Received 25 March 2016 Received in revised form 5 August 2016 Accepted 21 August 2016 Available online 22 August 2016 Keywords: Seed maturation AFL multigene family Redundancy Simplification Minimal control network

a b s t r a c t Networks controlling developmental or metabolic processes in plants are often complex as a consequence of the duplication and specialisation of the regulatory genes as well as the numerous levels of transcriptional and post-transcriptional controls added during evolution. Networks serve to accommodate multicellular complexity and increase robustness to environmental changes. Mathematical simplification by regrouping genes or pathways in a limited number of hubs has facilitated the construction of models for complex traits. In a complementary approach, a biological simplification can be achieved by using genetic modification to understand the core and singular ancestral function of the network, which is likely to be more prevalent within the plant kingdom rather than specific to a species. With this viewpoint, we review examples of simplification successfully undertaken in yeast and other organisms. A strategy of progressive complementation of single, double and triple mutants of seed maturation confirmed the fundamental role of the AFL sub-family of B3 transcription factors as master regulators of seed maturation, illustrating that biological simplification of complex networks could be more widely applied in plants. Defining minimal control networks will facilitate evolutionary comparisons of regulatory processes and the identification of an essential gene set for synthetic biology. © 2016 The Authors. Published by Elsevier Ireland Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Contents 1. 2. 3. 4. 5.

6. 7. 8. 9.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336 Seed development and maturation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336 AFL proteins in the plant kingdom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336 The AFL family of regulators − redundancy, specificity and regulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338 Approaches to simplify gene regulatory networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340 5.1. Simplification of a biological network by mathematical modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340 5.1.1. Flowering time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340 5.1.2. Seed maturation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341 5.2. Simplification of a control network by biological approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341 5.2.1. Dissection of the process in simple organisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341 5.2.2. Simplification by functional characterisation in a heterologous system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341 The example of human p53 regulation in S. cerevisiae . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341 The example of OLEOSIN1 in Physcomitrella patens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341 5.2.3. Simplification of the process in more complex organisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342 Learning from the example of the cell cycle control in S. pombe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342 Towards the elucidation of the fundamental roles of AFL proteins during seed maturation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342 Benefits of network simplification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343 Generalisation of the concept of “minimal control network” in plants and evolutionary considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344 Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344

∗ Corresponding author. E-mail address: [email protected] (M. Devic). http://dx.doi.org/10.1016/j.plantsci.2016.08.012 0168-9452/© 2016 The Authors. Published by Elsevier Ireland Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/bync-nd/4.0/).

336

M. Devic, T. Roscoe / Plant Science 252 (2016) 335–346

Appendix A. Supplementary data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344

1. Introduction Seeds are an important component of human and animal nutrition supplying calories in the form of starch, sugar, and oil together with amino acids, vitamins and microelements. The synthesis of the seed reserves occurs after pattern formation of the embryo during maturation, the second phase of seed development. In the final phase of embryogenesis, desiccation tolerance is acquired and dormancy is established. An understanding of the control of the seed maturation phase is essential to improve seed quality traits. The biochemical pathways leading to the production of seed reserves in Arabidopsis have been genetically analysed. Several master regulators of seed maturation have been identified. Despite this insight, there is no simple model of the control of initiation and the progression through maturation. This is partly due to the redundancy among the master regulators and the multiple levels of regulation added during evolution. It is necessary to define the core components of the system rather than to describe its complexity in order to understand the gene regulatory network controlling seed maturation. A minimal control network may therefore represent the extant equivalent of an ancestral regulatory gene. Here, we highlight the necessity for simplification and the approaches to define minimal control networks.

2. Seed development and maturation Seeds are complex structures, which have arisen in the Spermatophytes (seed bearing plants) more than 300 million years ago following whole genome duplication [1]. Gymnosperm seeds are composed of a diploid embryo (one maternal haploid set of chromosomes = 1 m and one paternal = 1 p) and a nourishing female gametophyte (2 m). Angiosperm seeds are the products of a double fertilisation and are composed of the seed coat, a diploid maternal tissue (2 m), the nutritive endosperm (often triploid 2m + 1 p) and the diploid embryo (1 m + 1 p), reviewed in [2]. The emergence of the seed trait has conferred advantages of protection of the reproductive structures and of dissemination as dry quiescent material. “Orthodox” Seeds are defined by their tolerance to desiccation (they can lose up to 90% of water during seed ripening). The embryo enters into a quiescent state and falls into dormancy concomitantly with the acquisition of desiccation tolerance. “Recalcitrant” seeds are less tolerant or are intolerant to desiccation, are less quiescent and cannot be well conserved, but they have the advantage of rapid germination. All seeds accumulate carbohydrate, lipid and protein reserves predominantly in the embryo of Gymnosperms and of exalbuminous Angiosperm seeds or in the endosperm of albuminous Angiosperm seeds. The formation of the embryo pattern, the accumulation of reserves, the acquisition of tolerance to desiccation and the entry into dormancy are the main programmes of seed development and are important agronomic traits defining seed quality. Collectively, these developmental processes comprise seed maturation. Irrespective of the qualitative differences listed above, all seeds essentially follow a common phase of seed maturation. Genetic analyses in Arabidopsis have led to the identification of four master regulators of seed maturation: ABSCISSIC ACID INSENSITIVE3 (ABI3) [3], FUSCA3 (FUS3) [4], LEAFY COTYLEDON2 (LEC2) [5] and LEAFY COTYLEDON1 (LEC1) [6]. Whilst LEC1 is a subunit of the CCAAT binding complex (HAP3 or NF-YB9), a general eukaryotic transcriptional regulatory complex, ABI3, FUS3 and LEC2 (collectively named AFL) are transcription factors with a plant specific B3

A. Phylogram of proteins with B3 domains ABI3 FUS3 LEC2

LAV

HSI2 HSL1 HSL2 RAV1 At3g25730 RAV2 At1g25560 At1g51120 At1g50680 NGA4 NGA2 NGA3 At2g36080 At5g06250 At3g11580 MONOPTEROS ARF3 ARF16 ARF10

RAV REM

ARF

B. Phylogram of the B3 domain of LAV proteins ABI3 FUS3

LEC2

52% HSL1 HSI2

35%

HSL2

68%

Fig. 1. Proteins possessing a B3 DNA binding domain in Arabidopsis. A. Subset of B3 domain proteins. The Arabidopsis genome contains 118 genes encoding proteins possessing a B3 DNA binding domain. The amino acid sequences of a selection of 22 proteins containing a B3 DNA binding domain were used to construct a phylogram using ClustalW and ClustalTree (ebi.ac.uk). These proteins can be divided into four major families. The LAV (LEC-ABI3-VAL) family contains the three AFL and the three HSI/VAL with the additional EAR (ERF-associated amphiphilic repression) domain. The RAV (Related to ABI3/VP1) family has 13 members, first identified by their homology to VP1. Certain RAV proteins possess an additional AP2 domain. The ARF (Auxin Response Factor) family has 23 members. ARF1, the founder member, was identified as binding upstream several auxin response genes. The REM family (REproductive Meristem) has 76 members that can be subdivided into 6 subgroups. Most of the REM proteins contain more than one B3 DNA binding domain. The graphic illustrates that the nearest neighbour proteins to the AFL are the HSI/VAL proteins. Members of the three other families, ARF, RAV and REM, constitute separate clades. B. Relatedness of the amino acid sequences of the B3 domain of AFL and HSI/VAL transcription factors. The numbers indicate the percentage of amino acid identity of the DNA binding domain only.

DNA binding domain (Fig. 1). The AFL regulators influence most aspects of seed maturation. For example, the abi3 and fus3 mutants are intolerant to desiccation, fus3 is not dormant (viviparous: can germinate on the mother plant) and abi3, fus3 and lec2 seeds contain less storage protein and lipid reserves and more starch. Homologues of AFL proteins have been identified in the genomes of all seed plants sequenced to date and in some moss and algae. It is hypothesised that their function in seed maturation is conserved among Spermatophytes [7–9]. Defining the core roles of AFL in the model plant Arabidopsis will provide insight as to regulation of the maturation phase of embryogenesis in cultivated crops. 3. AFL proteins in the plant kingdom Among the 118 proteins in Arabidopsis that contain a B3 DNA binding domain, ABI3, FUS3 and LEC2 have the most conserved B3 domain and are more similar to each other than to any other B3 protein [10] (Fig. 1). Analysis of individual AFL loss of function phenotypes has demonstrated the existence of a strong redundancy in their mode of action together with a limited specificity, that is, specialisation of function [11].

M. Devic, T. Roscoe / Plant Science 252 (2016) 335–346

337

Table 1 Number of AFL orthologues in the plant kingdom. Species

ABI3

FUS3

LEC2

Chlorophytes Chlamydomonas reinhardtii Volvox carteri

ZmAFL4

ZmAFL5,6

Other

1 ancestral AFL/VAL 1 ancestral AFL/VAL

Total

References

1 1

[18] [18]

Bryophytes Physcomitrella patens

3

0

0

0

0

2

5

[15,18,21]

Lycopodiophytes Selaginella moellendorffii

2

0

0

0

0

2

4

[15]

Amborellales Amborella trichopoda

1

0

0

0

0

2

3

Gymnosperms picea abies Chamaecyparis nootkatensis

1 1

ND ND

ND ND

ND ND

ND ND

ND ND

ND ND

[15] [91]

Monocots Brachypodium distachyon Brachypodium stacei Hordeum vulgare sorghum bicolor Oryza sativa Zea mays

1 1 1 1 1 1

1 1 1 1 1 1

0 0 0 0 0 0

1 1 1 1 2 1

1 1 1 1 1 2

0 1 0 0 0 0

4 5 4 4 5 5

[41,42] [15,92] [10,15,16,40] [15,17]

Dicotyledons Arabidopsis thaliana Glycine max Medicago truncatula Populus trichocarpa Manihot esculenta Solanum lycopersicum Solanum tuberosum Theobroma cacao Vitis vinifera

1 2 1 1 2 2 2 1 1

1 2 1 2 2 1 1 1 1

1 1 1 2 1 0 0 1 0

0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 1

3 5 3 5 5 3 3 3 3

[15] [15] [93,94] [15] [15] [95]

ND: not determined.

Although homologues of AFL genes are found in the genomes of seed plants, their presence and number varies among species. The resources of GRAMENE [12], Phytozome [13], Inparanoid [14], data from [15] and additional searches and phylogenetic tree analyses (Figs. 2 and 3, Supplementary Fig. 1) were used to produce a list of AFL orthologues in selected plant species whose genome has been entirely sequenced (Table 1). Since few genomes have been specifically examined for AFL orthologues, these data represent the result of a global search for orthologous genes. Furthermore, functional characterisation has been limited (referenced in Table 1) and frequently restricted to an individual AFL in a single species. Despite this caveat, the data in Table 1 reveal the ubiquity of AFL in seed plants and of their presence in lower plants as well. Homologues of ABI3, FUS3 and LEC2 are found in soybean, where there are two ABI3, two FUS3 and one LEC2. Orthologues of LEC2 have not been unambiguously identified in tomato (Solanum esculentum), potato (Solanum tuberosum) and grape (Vitis vinifera). Homologues of ABI3 and FUS3 have been identified in monocots species but not of LEC2. Instead, additional classes of B3 transcription factors, for example IDEF1 (orthologue of ZmAFL5) initially identified in rice [16] and ZmAFL4 in maize, are present. The total number of AFL-like is four in barley, five in both rice and maize. Furthermore, since Arabidopsis and most dicotyledonous seeds do not accumulate a large amount of starch in contrast to cereals, it is of interest that the maize ZmAFL4 protein, associated with starch accumulation in the endosperm [17], has orthologues in all monocots examined. To date three genomes of gymnosperms has been sequenced but none have been integrated yet into Gramene, Phytozome and Inparanoid databases. Based on the work of [15] and searches on EST databases, it was possible to identify ABI3 homologues in pine species but no other AFL. The genome of Amborella, a basal angiosperm, contains one ABI3 gene and two additional AFL genes, one of which resembles LEC2. In conclusion, homologues of ABI3 are unambiguously

recognised and identified in all flowering plants but orthologies with LEC2 and to a lesser extent with FUS3 are less evident between monocots and dicots. No AFL was identified in the genome of the picoalga Ostreococcus lucimarinus, but some AFL homologues were found in non-seed plants (Table 1. Chlamydomonas, volox, Physcomitrella and Selaginalla). ABI3/HSI, present as a single copy in green alga, is thought to be the founder member of the entire B3 DNA binding domain superfamily (Fig. 1) [18]. The phylogenetic tree of the ABI3/VP1 proteins (Fig. 2) groups closely the Amborella and gymnosperms proteins, away from the dicot and monocot clades. The non seed-plant proteins are positioned closer to the monocots than to Amborella proteins. ABI3 has two main roles during seed development: seed maturation involving the binding to the RY cis-element (a function common to each AFL) and activation of LEA (Late Embryogenesis Abundant) genes to acquire desiccation tolerance. This latter role is more specific to ABI3. RY elements are over-represented in the promoters of seed reserve biosynthetic genes (maturation), but not in the promoters of LEA genes [19]. This explains why a single AFL can compensate the absence of the other members for reserve accumulation but is inefficient for late embryogenesis traits [11]. The functionality of ABI3 orthologues has been tested in Physcomitrella [20,21]. The three ABI3/VP1-like (PpABI3A, PpABI3B, PpABI3C) genes in the moss have been shown to activate LEA genes in presence of abscisic acid (ABA) and to be necessary for drought tolerance. When expressed in abi3 seeds, PpABI3A can activate the expression of storage proteins and oleosin 2 genes, induce chlorophyll breakdown (seed degreening), and partially restore ABA sensitivity at germination. Unexpectedly, the transformed seeds are still desiccation intolerant [21]. An explanation for an incomplete complementation was that the interaction of ABI5 (bZIP) is weaker with PpABI3A than

338

M. Devic, T. Roscoe / Plant Science 252 (2016) 335–346

Fig. 2. Phylogenetic relationships of ABI3 proteins. The protein sequences (supplementary Fig. 1) were analysed by Clustal OMEGA (http://www.ebi.ac.uk/Tools/msa/clustalo/). The multiple alignment was used to build a phylogenetic tree at Clustal phylogeny (http://www.ebi.ac.uk/Tools/ phylogeny/clustalw2 phylogeny/). The cladogram is presented as a Neighbourjoining tree without distance corrections. The monocots and dicots proteins were clearly separated, with the non-seed plant proteins close to monocots. Amborella ABI3 was grouped with the two examples of gymnosperm ABI3. Chlamy: Chlamydomonas reinhardtii; Volvox: Volvox carteri; Physco: Physcomitrella patens; Sel: Selaginella moellendorffii; Ambo: Amborella trichopoda; Picea: picea abies; Cyp: Chamaecyparis nootkatensis; Brdi: Brachypodium distachyon; Brast: Brachypodium stacei; Hv: Hordeum vulgare; Sorbi: sorghum bicolor; Os: Oryza sativa; Zm: Zea mays; At: Arabidopsis thaliana; Soybn:Glycine max; Medtr: Medicago truncatula; POPTR: Populus trichocarpa; Manes: Manihot esculenta; Solyc: Solanum lycopersicum; Soltub: Solanum tuberosum; Thecc: Theobroma cacao; Vitis: Vitis vinifera.

with AtABI3. At present, the functional identities of the additional AFL in Physcomitrella and Selaginella remain inconclusive. 4. The AFL family of regulators − redundancy, specificity and regulation Common and specific functions of individual AFL were assessed in Arabidopsis in the context of reserve accumulation [11]. Specificity clearly exists as shown by their distinct gene expression profiles during seed development, differences in the level of control exerted over fatty acid modification, triacylglycerol and storage protein accumulation, mediated in part through their binding affinities for promoters of seed reserve-related genes. However, strong redundancy is evident since single AFL mutants continue to accumulate significant lipid and protein reserves. Storage protein and oleosin genes are common targets of ABI3 and FUS3 as established by chromatin immunoprecipitation [22,23] and of LEC2 [24]. In addition, when an individual AFL is ectopically over-expressed, the minor specificities are no longer evident and therefore, the AFL exhibit an equivalent function [11]. In this context, the unique functional AFL has lost most of its regulation and interaction with the two other AFL and additional regulators. This last point raises the question of the importance of the AFL network and its regulation in reserve accumulation. Furthermore, since the number of AFL is not constant among the plant genomes, the architecture of the network maybe variable according to the species. The AFL network controls reserve synthesis either directly by activating genes encoding enzymes of fatty acid and storage protein synthesis or indirectly through activation of secondary transcrip-

Fig. 3. Relationships among FUS3/LEC2/ZmAFL2/ZmAFL4-6 proteins. The B3 domain proteins that were not ABI3-like were analysed and represented as in Fig. 2. Globally, the FUS3/LEC2 proteins of dicots are separated from the distinct ZmAFL2, 4–6 proteins of monocots. The proteins from non-seed plants are grouped together and close to the dicot proteins.

tion factors such as WRI1. Regulation of the AFL network in the developing Arabidopsis seed is currently understood as a transcriptional activation at the transition from pattern formation to maturation and principally an epigenetic repression at the transition from seed to germination and seedling development [25,26 and references therein]. Robust control of the AFL network is based on auto- and mutual regulation and FUS3 is prominent within the network mediating activation and repression during seed development via feedback loops (Fig. 4). Activation of the AFL network is effected by diverse factors acting early in seed development. Overexpression of AGAMOUS15 (AGL15) up-regulates LEC1 and LEC2 and subsequently ABI3, FUS3 and LEC2 have been confirmed as direct targets of AGL15 [27]. MYB118 has been identified as an activator of AFL at the vegetative to embryonic transition where over-expression results in an increased expression of LEC1 and accumulation of storage lipid in vegetative tissues [28]. Expression of CHR5 during early embryo development leads to activation of AFL. By associating with the promoters of ABI3 and FUS3, CHR5 establishes an active chromatin state [29]. The timing of activation of maturation-related genes is critical. Mutations in DCL1 confirm a role for miRNAs in the induction of the maturation program. Indirect evidence suggests that during embryo patterning, miR156 acting on its target genes

M. Devic, T. Roscoe / Plant Science 252 (2016) 335–346

339

Fig. 4. Regulation of the AFL network during embryo development and transition to vegetative growth in Arabidopsis. The figure illustrates transcriptional and post-transcriptional control over AFL activation at the transition from embryo morphogenesis to maturation and epigenetic repression of the network at germination. Black arrows indicate auto and mutual control and the graphical representation of factors does not imply a hierarchy of genetic interactions. Green arrows indicate activation of the network components or potential activation (broken green arrows), blue arrows indicate feedback control and broken grey arrows indicate activation of reserve-related maturation genes. T-bars indicate repression or silencing. WRI1 is a tissue specific enhancer of fatty acid synthesis. Target genes of the AFL network are boxed. = WRINKLED1, TAG = Triacylglycerol.

SPL10 and SPL11, prevent the precocious expression of maturation genes including LEC2 and FUS3 [30,31]. At the morphogenesis-tomaturation transition SPL10 and SPL11 transcription is enhanced and repression is overcome. The morphogenesis-to-maturation transition may also be controlled by miR166 since among its targets PHAVOLUTA (PHV) and PHABULOSA (PHB), the latter was confirmed as a direct activator of LEC2 [32]. FUS3 is involved in feedback regulation of the activation network since AGL15, MYB118 and miRNA156 were shown to be direct targets [22]. The transition to vegetative development requires that the AFL network is efficiently repressed during germination to facilitate the transcriptional reprogramming necessary for seedling development. The AFL network is subject to a strong epigenetic control mediated by H3K27me3 transcriptional silencing. VAL/HSI B3 factors are closely related to the AFL factors and bind competitively to the RY cis-element. Interaction with co-repressors is proposed to lead to the recruitment of a histone deacetylase complex that targets genes containing RY motifs and active chromatin marks to transcriptionally silence chromatin [25]. Repression of the AFL network in vegetative tissues is also controlled by Polycomb Repressive Complexes (PRC) acting in concert with chromatin re-modeling proteins. PRC2 trimethylates Histone H3 such that target genes, including the AFL are H3K27me3 modified and hence repressed. PRC1 interacts with proteins that effect a switch in chromatin state where erasure of active H3K4me3 marks and replacement by repressive H3K27me3 marks repress seed developmental genes at the vegetative transition [33]. PRC1 also acts in concert with PRC2, recognising H3K27me3 marks to induce histone 2A monoubiquitination to further stabilise repression of target loci including AFL. The chromatin remodeller PICKLE is also involved in the repression of the AFL network mediated by cooperation with PRC2 at loci bearing repressive marks [34]. Additional proteins contribute to repression, for example ASIL1, which obstructs access of AFL to the RY-element [35] and SCL15 required for regulation of a chromatin remodelling factor to repress embryonic traits during

vegetative growth [36]. FUS3 also has a role in controlling repression of the AFL network by binding to VAL1, to PICKLE and the RING1b sub-unit of PRC1 in negative feedback loops. Despite the elaborate epigenetic, transcriptional and posttranscriptional control of the AFL network during seed development, it is remarkable that this complex regulation can be overcome by the ectopic expression of any individual AFL. The ability of a single AFL to induce reserve synthesis, (LEC2, [37]) supports our argument that redundancy exists among the AFL and is based on the presence of a B3 DNA binding domain common to each factor. That the production of reserve lipids in Arabidopsis seedlings following induction of FUS3 expression is independent of induction of LEC1, LEC1-L, LEC2 or ABI3 [38] clearly illustrates this point. Seed development between monocot and dicot species differs with respect to the formation of storage organs (scutellum, cotyledon, endosperm) and the partitioning of starch, protein and lipid reserves. Information concerning AFL control of reserve accumulation in monocots is fragmented and their intra- and interregulation, largely unknown. Studies have focussed on the late functions of VP1 (VIVIPAROUS1, orthologous to At-ABI3) in relation with ABA, dormancy, germination, LEA activation in maize [39], rice [40] and barley [41]. We examined evidence for AFL redundancy on seed reserves in monocots. Co-expression analyses have shown a relative conservation of the ABI3, FUS3 and LEC1 networks (including storage protein and carbohydrate metabolism), but not of LEC2 between monocots and dicots [7]. This comparison was based on systems biology studies mainly in wild type seed development. Functional studies on each member of the network and detailed description of their mode of action are necessary. In maize, a comprehensive study [17] has described the structure of the ZmAFL family. The ZmAFL3 (VP1) and ZmAFL4 factors each transactivate a maize oleosin2 promoter in a moss heterologous system and may act in synergy. However, the activation by ZmAFL4 may not be biologically relevant since ZmAFL4 is expressed in the endosperm and oleosin2 in the embryo. Thus, certain ZmAFL members retain

340

M. Devic, T. Roscoe / Plant Science 252 (2016) 335–346

a common ability to activate the accumulation of distinct types of seed reserves during the maturation phase, but they have also gained specialisation by their spatio-temporal expression pattern and sequence divergence, analogous to the Arabidopsis AFL. To test the full extent of their redundancy in vivo will require multiple mutant combinations or RNAi with limited specificity since several members of the maize AFL family are expressed in the same tissues. Remarkably, monocots often have one or two extra AFL belonging to the FUS3/LEC2 type in comparison to dicots (Table 1, Fig. 3). Since ZmAFL4 is expressed in the endosperm and found in all the monocots examined, the additional member(s) may function to regulate reserve accumulation in this starch accumulating tissue. AFL proteins in monocots have retained a similar mode of action with respect to the control of seed reserve synthesis. The presence of the RY cis-element has been described in the promoters of genes encoding reserve products in different monocots (in barley, ␤-hordein storage protein and trypsin inhibitor BTI-CMe [42], in Setaria pF128, a seed specific promoter for storage protein [43]). HvFUS3 has been shown to be involved seed storage protein accumulation in the endosperm and in the embryo [42]. Furthermore, ABI3 has been shown to activate storage protein genes in gymnosperms [44]. Examination of the AFL within the plant kingdom leads us to propose that the core function of the AFL is conserved among Spermatophyte species, but that their gene networks (the way the global function is distributed among the AFL members and their regulation) is not conserved. For example, a mutation in VP1 in maize induces precocious germination, whereas in Arabidopsis, a lesion in FUS3 results in more severe vivipary than is evident in abi3 seeds [45]. The redundancy inherent within the AFL obscures the fundamental and universal role of the AFL family in the control of seed reserves and therefore the regulatory network should be simplified to its minimal component: a single protein possessing a common AFL functionality (Fig. 5). We return to this concept in the following section. 5. Approaches to simplify gene regulatory networks 5.1. Simplification of a biological network by mathematical modelling A simplification of the system components is necessary to understand the basic mechanism that governs a developmental and metabolic pathway. This may be achieved by mathematical modelling using simplification or reduction, by regrouping genes or proteins or enzymatic reactions into kernels while preserving the core structure [46]. We illustrate this principle by the following examples. 5.1.1. Flowering time Plant species use different pathways in response to environmental cues allowing them to flower in the correct season. Key traits such as irreversibility and robustness to fluctuating signals have been conserved and involve the key regulators of the floral transition (perception of signals and floral meristem specification). These include the FLOWERING LOCUS T (FT) function and the interactions with FLOWERING LOCUS D, APETALA1, LEAFY and TERMINAL FLOWER1, which are all conserved in diverse species including tomato, rice and Arabidopsis. The major dynamic properties of the floral transition were captured in minimal regulatory networks of core components using mathematical modelling and simplification [47,48]. Since many hundreds of genes affect flowering time, this was achieved by describing only the major activities of groups of genes, referred to as “regulatory hubs” that represent one or more genes and proteins. Simplified models lack the finer details of the

Fig. 5. Genetic simplification leading to Minimal Control Network construction. Seed maturation A: During normal seed development, a complex auto- and inter-transcriptional network between the three AFL (ABI3, FUS3 and LEC2) controls the initiation and the progression of the different processes of the seed maturation phase [96]. This network controls the synthesis and the accumulation of protein and lipid reserves in the seed. B: In plants harbouring mutations in AFL genes, transformation with a single constitutively expressed AFL transgene restores reserve synthesis [11]. This activation is independent of the identity of the AFL member and of its regulation and is mediated through the action of the conserved B3 DNA binding domain. That reserves accumulate in a double mutant transformed with a single AFL indicates the existence of a threshold level for AFL function necessary for activation. This ‘Unique AFL’ system constitutes a Plant Minimal Network for seed reserve accumulation. Mitotic cell cycle (based on [69]) C: In a normal cell, the mitotic cell cycle of fission yeast relies on the association of a Cdk (Cdc2) with various cyclins (Cdc13, Puc1, Cig1, Cig2). The level of activity of a Cdk-cyclin couple (red gradient) regulated by several factors triggers the phase transition G1-S, G2-M. D: In yeast cell where the Cdk and the 4 cyclins have been deleted and replaced by a protein fusion between Cdc13 and Cdc2, the oscillation of Cdk activity is still occurring and is sufficient for the normal progression and the direction of the core cell cycle. The interaction of Cdk with specific cyclins at each cell cycle phase transition is not essential. This system represents the Minimal Control Network of the cell cycle independent of the main regulators.

biological system, yet they provide an understanding of the overall system behaviour and hence represent the core structure underlying the floral transition by simple feed-forward loops. Molecular genetic studies of flowering time in response to environmental clues have also been performed in rice, a short-day plant. Comparisons of the regulatory network in Arabidopsis, a long-day plant, and in rice, have shown that the core flowering pathway (minimal control network) is conserved and that other pathways have evolved for novel functions, increasing the diversity of flowering behaviours [49–51]. In addition, Arabidopsis and rice do not use the same number of genes to perform these core functions

M. Devic, T. Roscoe / Plant Science 252 (2016) 335–346

(e.g. one florigen in Arabidopsis FT, [52], two in rice Hd3a and RFT1 [53,54]). Although the core network integrating the environmental cues is conserved, its consequent mechanisms have diverged during evolution to produce sometimes opposite flowering responses (long-day for Arabidopsis and short-day for rice [49]). This has allowed rice to flower even during long-day conditions using a unique regulatory pathway [55]. Furthermore, flowering in Arabidopsis is controlled by a small number of large-effect genes, whereas in maize, it is controlled by many additive small-effect quantitative trait loci [56]. Interestingly, but as yet insufficiently documented, some AFL play a role in flowering time in Arabidopsis (for AtABI3 [57]) and in rice for (OsLFL1 = FUS3 [58]). Thus, the manner in which the core function is distributed with redundancy and/or specificity among regulatory genes and the presence or absence of gene families, is variable and probably represents the basis for the fine-tuning of the core network and permits robustness and adaptation for various growth conditions. 5.1.2. Seed maturation Similarly, it is important to determine the core and conserved functions of the AFL during seed maturation and more specifically, of each subset of functions including embryonic identity, reserve accumulation, dormancy, desiccation tolerance and germination. Several representations of OMICs data for seed development in individual species and for cross-species comparisons are available (PaNet http://aranet.mpimp-golm.mpg.de/ [59]). They allow the combined representations of metabolic pathways and fluxes and gene expression data. PlaNet was applied to compare the co-expression networks during seed maturation in monocots and dicots species [7] and showed the relative conservation of the ABI3/FUS3 and LEC1 networks but not of LEC2 as discussed above. These representations demonstrate the complexity of the gene regulatory networks but do not facilitate the extraction the core functions. Such representations have however, facilitated multiple species comparisons and led to the identification of yield related genes for crop improvement [7]. Simplified models, such as the ones developed for the regulation of the flowering time, separating the distinct functions of the AFL master regulators, are necessary. RIMAS (Regulatory Interaction Maps of Arabidopsis Seed Development, http://rimas.ipk-gatersleben.de/ [60]) proposes a simplified representation based on the design of electronic circuits System Biology Graphical Notation (SBGN). The advantage of RIMAS is to have dissected the different functions of the AFL into separate networks (maturation gene control, hormone metabolism, epigenetic control). Integrating the redundancy and threshold level of the three AFL for the initiation of seed reserves and updating the actual database, may constitute a basis to represent the minimal control network of AFL. 5.2. Simplification of a control network by biological approaches Key genes are unique in a simplified network, although they may represent several genes and proteins that have been combined into an “activity hub”. Empirical observation is the foundation for the construction of the mathematical models and their simplification in groups or modules. Intuitively, it is easier to build and test a model in plants in which the key regulatory genes are encoded by a single gene rather than by a gene family. This simplification may be achieved in three ways: 5.2.1. Dissection of the process in simple organisms The effects of auxin in the liverwort Marchantia [61] and the effects of abscisic acid in the bryophyte moss Physcomitrella patens [21] are examples. However this approach may be precluded or of limited value for processes specific to plants such as flowering time

341

where only subsets of the pathways maybe studied. In addition, simple organisms may have unanticipated complexity, by example Physcomitrella patens has three ABI3-like genes (Table 1).

5.2.2. Simplification by functional characterisation in a heterologous system The presence of gene families, the multiple functions of certain regulatory proteins and the multiple levels of control (epigenetic, transcriptional and post-transcriptional) together exacerbate the difficulty of determining the core components and structure of a network. An alternative and/or additional approach to simplify a network would be to isolate part of its components into an heterologous system, which contains some of the orthologous components, but not all. This approach allows the study of one part of the functionality of a pathway or of the regulation of a key gene, independently of the other components.

The example of human p53 regulation in S. cerevisiae p53 is conserved from worms to humans. It functions as a sensor of DNA damage and triggers cell cycle arrest (life) or apoptosis (death). A complex protein network, mostly involved in post-translational modifications, tightly regulates its activity. As a consequence, it is difficult to understand the regulatory principles in the p53 signalling network. The minimal requirements for functionally relevant p53 post-translational modifications were studied by expressing the human p53 together with its best-characterised modifier Mdm2 in budding yeast to circumvent this complexity [62]. Budding yeast does not contain p53 nor Mdm2 homologues and thus is an appropriate tool to study p53-Mdm2 interaction in isolation from its other regulators, still within a cellular context. This p53-Mdm2 module was sufficient to faithfully recapitulate key aspects of p53 regulation of higher eukaryotes. A similar approach could be taken with the AFL by isolating part of the network, particularly to address post-translational controls.

The example of OLEOSIN1 in Physcomitrella patens Whilst some higher plant-specific developmental processes cannot be directly transposed to moss, Physcomitrella patens remains an interesting model system to explore plant functions [63,64] and to dissect cellular processes including transcription [65], as well as hormonal and signalling pathways [21,66]. The development of a heterologous system in Physcomitrella for the analysis of transcription factor–DNA interactions has been validated with the BANYULS promoter (BANYULS encodes an enzyme involved in proanthocyanidin synthesis) and its main regulators [65]. This system has been used to dissect the AFL networks and transcription complexes activating the promoter of an oil body protein gene (OLEOSIN1) [67]. ABI3, LEC2 and LEC1 act in synergy and LEC2, LEC1 (NF-YB9) and NF-YC2 can form a ternary complex. FUS3 was not found to participate in this synergy, although in isolation, it activates the OLEOSIN1 promoter. Since Physcomitrella has three ABI3-like genes and two additional AFL-like genes, it is not an entirely neutral system for characterising AFL factors. The Physcomitrella system differs from a yeast one-hybrid system since it requires additional plant proteins for activation of the reporter gene. The yeast system only requires the binding of the plant transcription factor to its cognate cis element, the complete transcription machinery is independent of the identity of the plant protein. In contrast, in Physcomitrella, OLEOSIN1 activation probably utilises a moss bZIP to interact with Arabidopsis AFL raising the question as to whether sequence divergence influences the efficiency of activation as in [21]. However, the Physcomitrella system has the advantage over yeast that a greater number of proteins can be added simultaneously and Arabidopsis bZIP factors similar

342

M. Devic, T. Roscoe / Plant Science 252 (2016) 335–346

Fig. 6. Benefits of biological simplifications. Approaches to determine the core and essential functions of a cell and/or a multicellular organism. The concept of mathematically grouping of genes under a single one is similar to the biological simplification of the AFL or cyclin-CDK minimal control network. Isolating part of the network into an heterologous system facilitates the identification of essential interactions. Taken together, the core architectures deduced from mathematical modelling, the core functions from the biological simplification, the essential interactions of the network and the minimal gene set will aid minimal cell design with optimised metabolism.

to those described in [68] may be added to further customise the Physcomitrella system. 5.2.3. Simplification of the process in more complex organisms A first approximation would be to reduce the gene family to a single regulatory entity to construct the model. The model then could be applied to more complex species. A prerequisite would be to determine the extent of the redundancy within the gene family in planta by the study of the phenotype of single and multiple mutants. Learning from the example of the cell cycle control in S. pombe The control of the cell cycle in fission yeast [69] serves as a conceptual framework to develop a simplified plant control network for the dissection of seed maturation in Arabidopsis. The eukaryotic cell cycle uses the function of enzymes of the cyclin-dependent protein kinase (Cdk) family, which interact with specific regulatory subunits, the cyclins, to initiate the S phase (DNA replication) and the M phase (mitosis). Fission yeast possesses six cyclins and a single Cdk. Research has provided an accurate description of the molecular mechanisms controlling genome duplication and cell division [70]. Still knowledge of the complexity of cell cycle regulation has made it difficult to elucidate its basic determinants. A synthetic approach that generated a minimal cell cycle control

system was used to investigate the core of cell cycle regulation. Previous results [71] demonstrated that a single B cyclin can substitute for G1 cyclins and regulate S-phase and mitosis by the oscillation of Cdk activity. A fusion protein between a single Cdk and a single cyclin is sufficient to control the two main transitions of the cell cycle (S/G1, G2/M) during mitosis or meiosis of yeast cells depleted of all the other endogenous cyclins [69,72]. Furthermore, cells operating with this minimal module (lacking much of the known regulation) have no abnormal phenotype. Thus, compositionally distinct complexes are not absolutely required for mitotic and meiotic cell cycle progression. This surprising simplicity of the core cell cycle network challenges a number of paradigms in cell cycle control (Fig. 5). 6. Towards the elucidation of the fundamental roles of AFL proteins during seed maturation An objective of research into seed development is to identify the essential functions and requirements of the AFL in establishing embryonic identity and reserve accumulation. Conceptually, our work on the AFL genes [11] is a first step to the biological simplification of the seed maturation network (Fig. 5). We systematically analysed combinations of afl mutant phenotypes by

M. Devic, T. Roscoe / Plant Science 252 (2016) 335–346

complementation and confirmed the extent of functional redundancy among the AFL regulators. A unique ectopically expressed AFL essentially complements the seed maturation defects of reserve accumulation and morphology of afl embryos (the bending and the shape of the cotyledons and of the axis), but not the tolerance to desiccation and dormancy. This study further indicated the existence of a threshold necessary for function within the AFL pool. Furthermore, since no individual AFL was able to suppress the tolerance to desiccation, mid- and late-maturation programs were uncoupled. Such a minimal regulatory architecture should be valid for the control of reserve accumulation in most flowering plants. From this simple basic model, it is possible to incrementally increase the complexity. The Arabidopsis genome possesses three HSI/VAL genes in addition to the three AFL genes [73,74]. They also have highly redundant functions. The HSI/VAL proteins contain a B3 DNA binding domain very similar to that of the AFL (Fig. 1), which recognise similar target promoters. However, in contrast to the AFL, they act mainly as repressors mediated through their EAR (ERFassociated amphiphilic repression) domain [75]. These six genes are derived from a common ancestral gene. HSI/VAL proteins clearly repress the action of AFL during germination but their conflictual roles are unknown during seed development [76]. Reduction to a simplified AFL and HSI/VAL network comprising a single AFL and a single HSI/VAL would facilitate understanding of the mechanism underlying the AFL-HIS/VAL antagonism during seed development. Orthologues of HIS/VAL genes also exist in the genome of other plant species. However, rice has five AFL and two HSI/VAL genes [10], and so again, the network and its members are not conserved. An understanding of the interaction between a “generic AFL” and a “generic HSI/VAL” will be applicable to most plant species. A biological reduction to a single AFL is also conceptually similar to mathematical modular (top-down) control analysis: the complexity of the system is reduced by grouping genes or proteins or reactions and reactants into large modules connected by a small number of intermediates (Fig. 6). We believe that biological reduction will lead to a simplified mathematical modelling for the phase transition from embryogenesis to maturation. This biological reduction also showed that, despite the degree of redundancy that exists among the AFL, their common sequence (essentially the B3 DNA binding domain), is sufficient for the activation of lipid and protein reserve accumulation, but not for processes occurring during the late maturation programme. Thus the simplification obtained with a single AFL only applies to mid-maturation phase. In the absence of experimentation, this was suspected but not demonstrated. Through this approach, we obtained experimental evidence for the separation of seed maturation into distinct modules. Combining the study of simplified biological material with the establishment of mathematical models is expected to facilitate the identification of the core mechanism of a metabolic or developmental process or pathway. For example, a computational model of the molecular interactions of the minimal control network of cell cycle in fission yeast based on the Cdk-cyclin fusion protein and on the notion of quantitative CdK activity was built and challenged by experimentation [77]. The model supports the “quantitative” ¨ ¨ regulation of the cell cycle in contrast to the qualitative model of the interactions of Cdk with specific cyclins. It also provided new insights into the regulatory effect of the inhibitory phosphorylation of Cdk.

7. Benefits of network simplification The use of the simplified fission yeast possessing the minimal control network for cell cycle has allowed the establishment of the quantitative model of Cdk. The oscillation of activity with two thresholds (low at G1/S and high at G2/M) is sufficient to inde-

343

pendently regulate the two cell cycle phases and does not require specific cyclins. In addition, most canonical regulations by Wee1, Mik1 and Cdc25 have dispensable roles for the minimal cell cycle. This minimal system described in fission yeast, reveals the core control of the “generic” eukaryotic cell cycle [72]. The discovery that the oscillation of Cdk activity acts as the primary organiser of the cell cycle would not have been possible (or at least, greatly difficult) if the diversity and complexity of the numerous cyclins would have been taken into account. The construction of the minimal control network has revealed the essential components of the cell cycle control. The validated reductionist yeast model of p53-Mdm2 interaction allows further dissection of networks and the study of their dynamics, by the progressive addition of components to this minimal network. This novel approach maybe expected to lead to the elucidation of the core mechanisms of p53 regulation and allow testing of the strategies to counteract p53 malfunctions.

8. Generalisation of the concept of “minimal control network” in plants and evolutionary considerations Plants are excellent candidates for applying biological reductionism to understand the regulation of complex processes since their genomes encode large and numerous gene families [78]. The availability of several collections of T-DNA insertional mutants in Arabidopsis has initially facilitated the functional study of individual members of these families, evidencing their degree of redundancy. Subsequently, double and multiple combinations of mutants have helped to discover the role of a family or closely related members acting in specific pathways. For example, the complete mutant set of the FLOWERING LOCUS T/TERMINAL FLOWER 1 family has been studied (sextuple mutant, [79]). To introduce in these complex mutant backgrounds a single functional member of the family, ubiquitously and over-expressed will also help to identify specific and redundant functions. For example, actin is a highly conserved ubiquitous protein, which is essential for cellular processes. Arabidopsis contains eight actin genes that are grouped into vegetative and reproductive classes, functionally distinct. Mutations in any of the three vegetative actins revealed only mild phenotypes. When a unique vegetative actin is over-expressed in a double mutant, normal plants were recovered [80]. We believe that this type of approach could be more widely used to identify the minimal gene set and/or minimal control networks. Reducing the complexity of a network to a single gene will facilitate the study of its interaction with other partially redundant networks, which in turn, can be simplified. Once a minimal control network is established in a model plant, it may be more easily extended to more complex species. The minimal control network intuitively represents the ancestral network. The growth phenotype of the yeast lacking all the cyclins and genes but expressing the single Cdk-cyclin fusion is comparable to that of the wild type yeast with an efficiency of 80–90%. In a way, this model argues against the dogma that specific cyclins are required for specific cell cycle phase transition. However, this simple protein fusion is probably not sufficiently robust for cell survival in nature in competition with wild type yeast. The ancestral eukaryote may have possessed a single Cdk-cyclin complex with one Cdk and one cyclin. Extra Cdks and cyclins are necessary to fine-tune the regulation and the progression through the cell cycle. Selection for additional regulatory components in the modern yeast cell has established a requirement for these factors [72] such that the core process and mechanism are now obscured. By analogy, it can be hypothesised that a unique AFL gene in an ancestral plant gave rise to multiple genes in modern plants. Arabidopsis expressing a single functional AFL represents this

344

M. Devic, T. Roscoe / Plant Science 252 (2016) 335–346

progenitor and allows access to the core functions of AFL in the control of the seed maturation phase. Another important aspect of the identification of Plant Minimal Control Networks is the possibility to extend them to complex organisms in order to facilitate evolutionary-developmental biological comparisons. Gene regulatory networks can be used as evolutionary characters to compare organisms [81]. Resolving the Gene Regulatory Networks that underlie developmental processes in several species would enable comparison of these networks and identify similarities and differences in their components and interactions. We anticipate that minimal control network could constitute a “building block” for evolutionary developmental comparisons of regulatory processes among diverse plant species to determine ancestral relationships and provide insight as to how regulatory processes evolve. Defining plant minimal control network can constitute an approach to attaining one of the goals in cell biology: to identify the universal minimal gene set (essential genes) required to sustain life. Searches for essential genes have been performed in prokaryotes and in multicellular eukaryotes [82], including Arabidopsis [83,84]. This has facilitated the creation of synthetic prokaryotic cells containing a minimal genome [85]. Initially these synthetic cells survived only under defined growth conditions. When challenged by the environment and in competition with other prokaryotes, the requirement for additional genes for fitness is evident. The fine-tuning and the robustness of the networks require many more genes, conserved in most living organisms, rather than the bare essential [86]. This is no longer the case for the synthetic prokaryotes. Several synthetic prokaryotes have been constructed with growth comparable to their original strains. Many synthetic pathways have been introduced into microorganisms. Frequently the rate-limiting steps in the pathways are strengthened and the unnecessary pathways, which use intermediate metabolites or produce competing by-products, are removed to increase yield [87]. The achievement of a minimal yeast cell is within reach. Can this be achieved in plant cells and eventually in multicellular plants? Microalgae are used as cell factories to produce high-value products (e.g. terpenoids, [88]). Synthetic biology is also used in plant biology with ambitious goals such as engineering C4 rice plants [89] or introducing nitrogen fixation in non-leguminous plants [90]. Information on minimal gene set and minimal control network will facilitate the creation of modified or synthetic plant cells by optimising genome reduction or editing.

9. Conclusions The Arabidopsis AFL can be studied as a unique entity in a minimal control network to understand the core process. This biological reduction (a) revealed that individual AFL share partially redundant functions and (b) permited the dissociation and subsequent independent study of distinct developmental phases occurring during seed maturation. By combining the understanding of the plant minimal control network determined in a model plant, with the specialised functions and interacting networks of each component of the network in a given species, it will be possible to reconstitute the core process at its origin with its fine-tuning in the modern plants.

Acknowledgements This perspective was inspired by the presentation “Understanding Cell Cycle Control” of Sir Paul Nurse. This work was supported in part by the French Agence National de la Recherche, Grant ANR10-BLAN-1238 (CERES) and Grant ANR-11-ISV7-0002 (SYNERGY).

Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.plantsci.2016.08. 012.

References [1] Y. Jiao, et al., Ancestral polyploidy in seed plants and angiosperms, Nature 473 (2011) 97–100. [2] C. Baroux, C. Spillane, U. Grossniklaus, Evolutionary origins of the endosperm in flowering plants, Genome Biol. 3 (2002) (reviews 1026). [3] J. Giraudat, B.M. Hauge, C. Valon, J. Smalle, F. Parcy, H.M. Goodman, Isolation of the Arabidopsis ABI3 gene by positional cloning, Plant Cell 4 (1992) 1251–1261. [4] H. Luerssen, V. Kirik, P. Herrmann, S. Miséra, FUSCA3 encodes a protein with a conserved VP1/AB13-like B3 domain which is of functional importance for the regulation of seed maturation in Arabidopsis thaliana, Plant J. 15 (1998) 755–764. [5] S.L. Stone, et al., LEAFY COTYLEDON2 encodes a B3 domain transcription factor that induces embryo development, Proc. Natl. Acad. Sci. U. S. A. 98 (2001) 11806–11811. [6] T. Lotan, et al., Arabidopsis LEAFY COTYLEDON1 is sufficient to induce embryo development in vegetative cells, Cell 93 (1998) 1195–1205. [7] N. Sreenivasulu, U. Wobus, Seed-development programs: a systems biology-based comparison between dicots and monocots, Annu. Rev. Plant Biol. 64 (2013) 189–217. [8] P. Agarwal, S. Kapoor, A.K. Tyagi, Transcription factors regulating the progression of monocot and dicot seed development, Bioessays 33 (2011) 189–202. [9] Schallau, et al., Phylogenetic footprints in fern spore- and seed-specific gene promoters, Plant J. 53 (2008) 414–424. [10] K. Swaminathan, K. Peterson, T. Jack, The plant B3 superfamily, Trends Plant Sci. 13 (2008) 647–655. [11] T.J. Roscoe, et al., Complementation of seed maturation phenotypes by ectopic expression of ABSCISIC ACID INSENSITIVE3, FUSCA3 and LEAFY COTYLEDON2 in Arabidopsis, Plant Cell Physiol. 54 (2015) 1215–1228. [12] M.K. Tello-Ruiz, et al., Gramene 2016: comparative plant genomics and pathway resources, Nucleic Acids Res. 44 (2016) D1133–D1140. [13] D.M. Goodstein, et al., Phytozome: a comparative platform for green plant genomics, Nucleic Acids Res. 40 (2012) D1178–D1186. [14] E.L. Sonnhammer, G. Östlund, InParanoid 8: orthology analysis between 273 proteomes mostly eukaryotic, Nucleic Acids Res. 43 (2015) D234–D239. [15] Y. Li, et al., Stepwise origin and functional diversification of the AFL subfamily B3 genes during land plant evolution, J. Bioinform. Comput. Biol. 8 (2010) 33–45. [16] T. Kobayashi, et al., The transcription factor IDEF1 regulates the response to and tolerance of iron deficiency in plants, Proc. Natl. Acad. Sci. U. S. A. 104 (48) (2007) 19150–19155. [17] Grimault, et al., Role of B3 domain transcription factors of the AFL family in maize kernel filling, Plant Sci. 236 (2015) 116–125. [18] E.A. Romanel, et al., Evolution of the B3 DNA binding superfamily: new insights into REM family gene diversification, PLoS One 4 (2009) e5791. [19] G. Guerriero, et al., The RY/Sph element mediates transcriptional repression of maturation genes from late maturation to early seedling growth, New Phytol. 184 (2009) 552–565. [20] Yotsui, et al., ABSCISIC ACID INSENSITIVE3 regulates abscisic acid-responsive gene expression with the nuclear factor Y complex through the ACTT-core element in Physcomitrella patens, New Phytol. 199 (2013) 101–109. [21] H.H. Marella, Y. Sakata, R.S. Quatrano, Characterization and functional analysis of ABSCISIC ACID INSENSITIVE3-like genes from Physcomitrella patens, Plant J. 46 (2006) 1032–1044. [22] F. Wang, S.E. Perry, Identification of direct targets of FUSCA3, a key regulator of Arabidopsis seed development, Plant Physiol. 161 (2013) 1251–1264. [23] G. Mönke, et al., Toward the identification and regulation of the Arabidopsis thaliana ABI3 regulon, Nucleic Acids Res. 40 (2012) 8240–8254. [24] S.A. Braybrook, et al., Genes directly regulated by LEAFY COTYLEDON2 provide insight into the control of embryo maturation and somatic embryogenesis, Proc. Natl. Acad. Sci. U. S. A. 103 (2006) 3468–3473. [25] H. Jia, H.M. Suzuki, D.R. McCarty, Regulation of the seed to seedling developmental phase transition by the LAFL and VAL transcription factor networks, Rev. Dev. Biol. 3 (2014) 135–145. [26] C. Fatihi, D. Boulard, S. Bouyer, B. Baud, L. Dubreucq, Deciphering Lepiniec and modifying LAFL transcriptional regulatory network in seed for improving yield and quality of storage compounds, Plant Sci. 250 (2016) 198–204. [27] Y. Zheng, N. Ren, H. Wang, A.J. Stromberg, S.E. Perry, Global identification of targets of the Arabidopsis MADS domain protein AGAMOUS-Like15, Plant Cell 21 (2009) 2563–2577. [28] X. Wang, Q.W. Niu, C. Teng, C. Li, J. Mu, N.H. Chua, J. Zuo, Overexpression of PGA37/MYB118 and MYB115 promotes vegetative-to-embryonic transition in Arabidopsis, Cell Res. 19 (2009) 224–235. [29] Y. Shen, M. Devic, L. Lepiniec, D.X. Zhou, Chromodomain, Helicase and DNA-binding CHD1 protein CHR5, are involved in establishing active

M. Devic, T. Roscoe / Plant Science 252 (2016) 335–346

[30]

[31]

[32]

[33]

[34]

[35] [36]

[37]

[38]

[39]

[40] [41]

[42]

[43]

[44]

[45]

[46] [47] [48] [49] [50]

[51]

[52] [53] [54]

[55]

[56] [57]

[58] [59] [60]

chromatin state of seed maturation genes, Plant Biotechnol. J. 13 (2015) 811–820. M.D. Nodine, D.P. Bartel, MicroRNAs prevent precocious gene expression and enable pattern formation during plant embryogenesis, Genes Dev. 24 (2010) 2678–2692. M.R. Willmann, A.J. Mehalick, R.L. Packer, P.D. Jenik, MicroRNAs regulate the timing of embryo maturation in Arabidopsis, Plant Physiol. 155 (2011) 1871–1884. X. Tang, et al., MicroRNA-mediated repression of the seed maturation program during vegetative development in Arabidopsis, PLoS Genet. 8 (2012) e1003091. A.M. Molitor, Z. Bu, Y. Yu, W.H. Shen, Arabidopsis AL PHD-PRC1 complexes promote seed germination through H3K4me3-to-H3K27me3 chromatin state switch in repression of seed developmental genes, PLoS Genet. 10 (2014) e1004091. H. Zhang, B. Bishop, W. Ringenberg, W.M. Muir, J. Ogas, The CHD3 remodeler PICKLE associates with genes enriched for trimethylation of histone H3 lysine 27, Plant Physiol. 159 (2012) 418–432. M.J. Gao, et al., Repression of seed maturation genes by a trihelix transcriptional repressor in Arabidopsis seedlings, Plant Cell. 21 (2009) 54–71. M.J. Gao, et al., SCARECROW-LIKE15 interacts with HISTONE DEACETYLASE19 and is essential for repressing the seed maturation programme, Nat. Commun. 6 (2015) 7243. M. Santos Mendoza, B. Dubreucq, M. Miquel, M. Caboche, L. Lepiniec, LEAFY COTYLEDON 2 activation is sufficient to trigger the accumulation of oil and seed specific mRNAs in Arabidopsis leaves, FEBS Lett. 29 (2005) 4666–4670. M. Zhang, X. Cao, Q. Jia, J. Ohlrogge, FUSCA3 activates triacylglycerol accumulation in Arabidopsis seedlings and tobacco BY2 cells, Plant J. (2016), http://dx.doi.org/10.1111/tpj.13233. D.R. McCarty, T. Hattori, C.B. Carson, V. Vasil, M. Lazar, I.K. Vasil, The Viviparous-1 developmental gene of maize encodes a novel transcriptional activator, Cell 66 (1991) 895–905. T. Hattori, T. Terada, S.T. Hamasuna, Sequence and functional analyses of the rice gene homologous to the maize Vp1, Plant Mol. Biol. 24 (1994) 805–810. Z. Abraham, et al., A developmental switch of gene expression in the barley seed mediated by HvVP1 (Viviparous-1) and HvGAMYB interactions, Plant Physiol. 170 (2016) 2146–2158. M.A. Moreno-Risueno, N. González, I. Díaz, F. Parcy, P. Carbonero, J. Vicente-Carbajosa, FUSCA3 from barley unveils a common transcriptional regulation of seed-specific genes between cereals and Arabidopsis, Plant J. 53 (2008) 882–894. Y. Pan, X. Ma, H. Liang, Q. Zhao, D. Zhu, J. Yu, Spatial and temporal activity of the foxtail millet (Setaria italica) seed-specific promoter pF128, Planta 241 (2015) 57–67. Y. Zeng, N. Raimondi, A.R. Kermode, Role of an ABI3 homologue in dormancy maintenance of yellow-cedar seeds and in the activation of storage protein and Em gene promoters, Plant Mol. Biol. 51 (2003) 39–49. M. Keith, N.G. Kraml, P. Dengler, McCourt, fusca3: a heterochronic mutation affecting late embryo development in Arabidopsis, Plant Cell 6 (1994) 589–600. J.R. Kim, et al., Reduction of complex signaling networks to a representative kernel, Sci. Signal. 4 (2011) ra35. K.E. Jaeger, et al., Interlocking feedback loops govern the dynamic behavior of the floral transition in Arabidopsis, Plant Cell 25 (2013) 820–833. N. Pullen, et al., Simple network motifs can capture key characteristics of the floral transition in Arabidopsis, Plant Signal. Behav. 8 (2013) e26149. Z. Milec, M. Valárik, J. Bartoˇs, J. Safáˇr, Can a late bloomer become an early bird? Tools for flowering time adjustment, Biotechnol. Adv. 32 (2014) 200–214. R. Shrestha, J. Gómez-Ariza, V. Brambilla, F. Fornara, Molecular control of seasonal flowering in rice, arabidopsis and temperate cereals, Ann. Bot. 114 (2014) 1445–1458. C. Sun, D. Chen, J. Fang, P. Wang, X. Deng, C. Chu, Understanding the genetic and epigenetic architecture in complex network of rice flowering pathways, Protein Cell 12 (2014) 889–898. Corbesier, et al., FT protein movement contributes to long-distance signaling in floral induction of Arabidopsis, Science 316 (2007) 1030–1033. R. Komiya, A. Ikegami, S. Tamaki, S. Yokoi, K. Shimamoto, Hd3a and RFT1 are essential for flowering in rice, Development 135 (2008) 767–774. H. Tsuji, K. Taoka, K. Shimamoto, Regulation of flowering in rice: two florigen genes, a complex gene network, and natural variation, Curr. Opin. Plant Biol. 14 (2011) 45–52. H. Tsuji, K. Taoka, K. Shimamoto, Florigen in rice: complex gene network for florigen transcription, florigen activation complex, and multiple functions, Curr. Opin. Plant Biol. 16 (2013) 228–235. E.S. Buckler, et al., The genetic architecture of maize flowering time, Science 325 (2009) 714–718. S. Kurup, H.D. Jones, M.J. Holdsworth, Interactions of the developmental regulator ABI3 with proteins identified from developing Arabidopsis seeds, Plant J. 21 (2000) 143–155. L.T. Peng, et al., Overexpression of transcription factor OsLFL1 delays flowering time in Oryza sativa, J. Plant Physiol. 165 (2008) 876–885. Mutwil, et al., PlaNet: combined sequence and expression comparisons across plant networks derived from seven species, Plant Cell 23 (2011) 895–910. A. Junker, A. Hartmann, F. Schreiber, H. Bäumlein, An engineer’s view on regulation of seed development, Trends Plant Sci. 15 (6) (2010) 303–307.

345

[61] E. Flores-Sandoval, D.M. Eklund, J.L. Bowman, A Simple auxin transcriptional response system regulates multiple morphogenetic processes in the Liverwort Marchantia polymorpha, PLoS Genet. 11 (2015) e1005207. [62] B. Di Ventura, et al., Reconstitution of Mdm2-dependent post-translational modifications of p53 in yeast, PLoS One 3 (2008) e1507. [63] D.J. Cove, et al., The moss Physcomitrella patens: a novel model system for plant development and genomic studies, Cold Spring Harb. Protoc. 2 (2009) (pdb emo115). [64] M.J. Prigge, M. Bezanilla, Evolutionary crossroads in developmental biology: Physcomitrella patens, Development 137 (2010) 3535–3543. [65] J. Thevenin, et al., A new system for fast and quantitative analysis of heterologous gene expression in plants, New Phytol. 193 (2012) 504–512. [66] P.F. Perroud, et al., Defective Kernel 1 (DEK1) is required for three-dimensional growth in Physcomitrella patens, New Phytol. 203 (2014) 794–804. [67] S. Baud, et al., Deciphering the molecular mechanisms underpinning the transcriptional control of gene expression by L-AFL proteins in Arabidopsis seed, Plant Physiol. (2016), http://dx.doi.org/10.1104/pp.16.00034. [68] A. Yamamoto, Y. Kagaya, R. Toyoshima, M. Kagaya, S. Takeda, T. Hattori, Arabidopsis NF-YB subunits LEC1 and LEC1-LIKE activate transcription by interacting with seed-specific ABRE-binding factors, Plant J. 58 (2009) 843–856. [69] D. Coudreuse, P. Nurse, Driving the cell cycle with a minimal CDK control network, Nature 468 (2010) 1074–1079. [70] C.J. McInerny, Cell cycle regulated gene expression in yeasts, Adv. Genet. 73 (2011) 51–85. [71] D.L. Fisher, P. Nurse, A single fission yeast mitotic cyclin B p34cdc2 kinase promotes both S-phase and mitosis in the absence of G1 cyclins, EMBO J. 15 (1996) 850–860. [72] P. Gutierrez-Escribano, P. Nurse, A single cyclin-CDK complex is sufficient for both mitotic and meiotic progression in fission yeast, Nat. Commun. 6 (2015) 6871. [73] H. Tsukagoshi, A. Morikami, K. Nakamura, Two B3 domain transcriptional repressors prevent sugar-inducible expression of seed maturation genes in Arabidopsis seedlings, Proc. Natl. Acad. Sci. U. S. A. 104 (2007) 2543–2547. [74] M. Suzuki, H.H. Wang, D.R. McCarty, Repression of the LEAFY COTYLEDON 1/B3 regulatory network in plant embryo development by VP1/ABSCISIC ACID INSENSITIVE 3-LIKE B3 genes, Plant Physiol. 143 (2007) 902–911. [75] Ohta, et al., Repression domains of class II ERF transcriptional repressors share an essential motif for active repression, Plant Cell 13 (2001) 1959–1968. [76] H. Jia, D.R. McCarty, M. Suzuki, Distinct roles of LAFL network genes in promoting the embryonic seedling fate in the absence of VAL repression, Plant Physiol. 163 (2013) 1293–1305. [77] C. Gerard, et al., Cell cycle control by a minimal Cdk network, PLoS Comput. Biol. 11 (2015) e1004056. [78] Y.L. Guo, Gene family evolution in green plants with emphasis on the origination and evolution of Arabidopsis thaliana genes, Plant J. 73 (2013) 941–951. [79] W. Kim, et al., Generation and analysis of a complete mutant set for the Arabidopsis FT/TFL1 family shows specific effects on thermo-sensitive flowering regulation, J. Exp. Bot. 64 (2013) 1715–1729. [80] M.K. Kandasamy, E.C. McKinney, R.B. Meagher, A single vegetative actin isovariant overexpressed under the control of multiple regulatory sequences is sufficient for normal Arabidopsis development, Plant Cell 21 (2009) 701–718. [81] A.H. Fischer, J. Smith, Evo-devo in the era of gene regulatory networks, Integr. Comp. Biol. 52 (2012) 842–849. [82] R. Zhang, Y. Lin, DEG 5.0: a database of essential genes in both prokaryotes and eukaryotes, Nucleic Acids Res. 37 (2009) D455–458. [83] M. Devic, The importance of being essential: EMBRYO-DEFECTIVE genes in Arabidopsis, C.R. Biol. 331 (2008) 726–736. [84] D. Meinke, et al., Identifying essential genes in Arabidopsis thaliana, Trends Plant Sci. 13 (2008) 483–491. [85] M. Juhas, L. Eberl, J.I. Glass, Essence of life: essential genes of minimal genomes, Trends Cell Biol. 21 (2011) 562–568. [86] A.K. Ramani, et al., The majority of animal genes are required for wild-type fitness, Cell 148 (2012) 792–802. [87] D. Choe, et al., Minimal genome: worthwhile or worthless efforts toward being smaller? Biotechnol. J. 11 (2016) 199–211. [88] D. Gangl, et al., Biotechnological exploitation of microalgae, J. Exp. Bot. 66 (2015) 6975–6990. [89] R.T. Furbank, Walking the C4 pathway: past, present, and future, J. Exp. Bot. (2016) (pii: erw161). [90] F. Mus, et al., Symbiotic nitrogen fixation and challenges to extending it to non-legumes, Appl. Environ. Microbiol. (2016) 01055–01056 (pii: AEM.). [91] Y. Zeng, Y.A.R. Kermode, A gymnosperm ABI3 gene functions in a severe abscisic acid-insensitive mutant of Arabidopsis (abi3-6) to restore the wild-type phenotype and demonstrates a strong synergistic effect with sugar in the inhibition of post-germinative growth, Plant Mol. Biol. 56 (5) (2004) 31–46. [92] F. Carrari, et al., Cloning and expression of a sorghum gene with homology to maize vp1: Its potential involvement in pre-harvest sprouting resistance, Plant Mol. Biol. 45 (6) (2001) 631–640. [93] J. Verdier, et al., A regulatory network-based approach dissects late maturation processes related to the acquisition of desiccation tolerance and longevity of Medicago truncatula seeds, Plant Physiol. 163 (2) (2013) 57–74.

346

M. Devic, T. Roscoe / Plant Science 252 (2016) 335–346

[94] J. Verdier, et al., Gene expression profiling of M. truncatula transcription factors identifies putative regulators of grain legume seed filling, Plant Mol. Biol 67 (6) (2008) 567–580. [95] G.W. Bassel, R.T. Mullen, J.D. Bewley, ABI3 expression ceases following, but not during, germination of tomato and Arabidopsis seeds, J. Exp. Bot. 57 (6) (2006) 1291–1297.

[96] To, et al., A network of local and redundant gene regulation governs Arabidopsis seed maturation, Plant Cell 18 (2006) 1642–1651.