Understanding how seed size is regulated in angiosperms is a key goal for plant science as seed size is an important component of overall seed yield. Angiosperm seeds comprise three clearly defined components, i.e. the embryo, endosperm and seed coat, with each having a distinct genetic composition which exerts different influences on seed development. Complex cross-talk and integration of signals from these different regions of the seed together determine its final size. The present review considers some of the major regulators of seed size, with a particular emphasis on the role of the seed coat in modulating endosperm proliferation and cellularization. The innermost layer of the seed coat, the endothelium, synthesizes flavonoids which are held to provide a defensive function against microbes, act as feeding deterrents, provide UV protection and to have a role in seed dormancy. A growing body of data suggests that flavonoids may also play a fundamental role in regulating communication between the seed coat and the endosperm. In the present review, we discuss how this may be achieved in the light of the fact that several flavonoids are known to be potent auxin transport regulators.
Seeds, either directly or indirectly, provide the majority of calories consumed by the human race . With the world population set to increase to approximately 9.3 billion by 2050, the need to increase crop production is of paramount importance . In this context, understanding how seed size is regulated at the genetic level is a key goal for plant scientists, as seed size is an important component of overall seed yield .
Seeds of flowering plants, the angiosperms, comprise three clearly defined components: the embryo, endosperm and seed coat. These have distinct genetic compositions which exert different influences on seed development. The embryo and endosperm are zygotic tissues and are the products of double fertilization, a key feature of angiosperm reproduction. Here, two sperm cells are delivered to the embryo sac (female gametophyte) where one fuses with the egg cell and the other with the central cell, leading to development of the embryo and endosperm respectively . In contrast with the embryo which is diploid, the endosperm is triploid, as the central cell typically contains two haploid polar nuclei from the female gametophyte and therefore has a two maternal to one paternal genomic ratio (2m:1p). Thus the development of the embryo and endosperm are under the control of both maternal and paternal genomes, whereas the seed coat is a sporophytic tissue and its development is regulated by the maternal genome only . Regulation of seed size therefore requires the co-ordination of growth and development of both zygotic and sporophytic maternal tissues .
In recent years, much progress has been made in uncovering genes acting in each seed compartment that ultimately affect seed size; however, it is becoming clear that factors acting in one tissue frequently affect the growth and development of others. Thus complex cross-talk and integration of signals from the different components of the seed must together determine its final size . In the following sections, we consider how both maternal and zygotic tissues influence seed size and discuss how interspecific and interploidy hybridizations can upset the delicately balanced programme of endosperm development. Finally, we discuss how mutants of the flavonoid biosynthesis pathway in the seed coat influence endosperm cellularization and suggest a linked role for auxin.
Determining seed size: roles for zygotic and maternal tissues
The seed coat is derived from two layers of tissue, the integuments, which constitute the outermost layers of the ovule. Fertilization of the ovule, and subsequent initiation of endosperm development, triggers a programme of growth and differentiation in the integuments to form the seed coat . Several maternal factors acting in the seed coat to influence seed size have been characterized, and these include TTG2 (TRANSPARENT TESTA GLABRA2), AP2 (APETALA2), ARF2 (AUXIN RESPONSE FACTOR2)/MNT (MEGAINTEGUMENTA), DA1, KLU (KLUH)/CYP78A5 (CYTOCHROME P450 78A5) and DA2 [3,5,6,8–10]. The transcription factors ARF2/MNT and AP2, along with ubiquitin pathway-related proteins DA1 and DA2, act to restrict seed size, with ARF2, DA1 and DA2 specifically restricting cell proliferation in the integuments. Conversely, TTG2 and KLU/CYP78A5 enhance seed size by regulating cell expansion and cell proliferation in the seed coat respectively. Importantly, these factors that act in maternal tissues affect development of the endosperm by influencing its proliferation and timing of cellularization, providing strong evidence of cross-talk between these two seed compartments.
Zygotic factors have also been identified that are important determinants of seed size, and among these are the IKU (HAIKU) pathway genes, IKU2, MINI3 (MINISEED3) and IKU1 which regulate the development of the early endosperm and expansion of the seed cavity [11,12]. IKU mutants have small seeds and endosperm proliferation is significantly reduced compared with wild-type. IKU2, a leucine-rich repeat kinase, is likely to be transcriptionally regulated by an IKU1–MINI3 complex . Importantly, although the IKU pathway acts specifically in zygotic tissues, IKU mutants also influence the seed coat, with cells of the integuments having smaller dimensions. These data again clearly point to the fact that regulation of seed size is governed by complex interactions between each seed component as proposed by Garcia et al. .
Overarching regulation of seed development, including size and shape, has been shown recently to be directed by BR (brassinosteroid), a steroid plant hormone. Importantly, BR was found to activate known positive regulators of seed size, for example the IKU pathway, and repress classic negative regulators of seed size such as AP2 and ARF2 .
Triploid block: a central role for the endosperm
Further clues to factors that regulate seed size come from an examination of seed development following intraspecific interploidy and interspecific hybridization . Despite the importance of hybridization in the evolutionary history of many plant taxa, crosses between plants frequently fail. The failure of such matings can occur for a host of reasons which can be classed broadly as pre-zygotic or post-zygotic hybridization barriers . Pre-zygotic barriers commonly affect the process of pollination so that the pollen grains ultimately fail to deliver their cargo of sperm to the embryo sac . Post-zygotic barriers, which are particularly relevant to this discussion, frequently relate to hybrid inviability through failure of seed development and weakness or sterility of the resulting hybrid [17–19]. Although several genetic mechanisms have been proposed to account for post-zygotic breeding barriers , failure of normal endosperm development is a common outcome of these unsuccessful matings . A large body of evidence now points to an imbalance of maternal and paternal genomic contributions in the endosperm being the primary cause of seed development failure in interploidy and interspecies hybridizations from a range of species [21–26].
Work by Lin [24,25] on maize (Zea mays) highlighted that deviations from the normal 2m:1p genomic ratio in endosperm frequently resulted in abnormal seed development. Thus extra copies of either the paternal or maternal genomes in the endosperm, as would occur in interploidy hybridizations, upset the normal developmental programme of the seed. An extension of this, the EBN (endosperm balance number) hypothesis, that accounts for failure of appropriate endosperm development in interspecies hybridizations of the same ploidy, has been proposed by Johnston et al. . The EBN hypothesis assigns a particular ‘genomic strength’ to maternally or paternally derived genomes from different species that equates to an effective ploidy. More recently, work on the model species Arabidopsis thaliana and maize has confirmed the crucial importance of the ratio of maternal and paternal genomic contributions in the endosperm to seed development and highlighted that the direction of interploidy crosses have profoundly differing effects on endosperm development [26,27].
These parent-of-origin-dependent defects in seed development result in changes to seed size; for example 2x×4x crosses (paternal excess) in Arabidopsis yield large seeds, whereas the reciprocal 4x×2x (maternal excess) cross yields small seeds (Figure 1). In Arabidopsis, seeds resulting from these crosses are generally viable despite aberrant endosperm development ; however, in maize, deviations from the normal 2m:1p ratio in endosperm are generally not tolerated and lead to seed abortion [25,27]. The failure of 2x×4x crosses is a common feature of matings in the plant kingdom and is known as the ‘triploid block’ because the resulting triploid embryos die as a consequence of abnormal endosperm development. The evolutionary implications of this effect cannot be underestimated for the triploid block is a very effective and immediate intrinsic post-zygotic breeding barrier between tetraploids and their diploid progenitors [14,28]. Polyploidy has played a fundamental role in the evolution of flowering plants, and thus reproductive isolation between plants of different ploidy in a population would facilitate speciation events.
Seed development in Arabidopsis thaliana following interploidy crosses
As we have seen, the endosperm component of the seed is genetically distinct from the embryo, originating from the fusion of the homodiploid central cell and a haploid sperm cell. Thus a ratio of two maternal to one paternal genomes (2m:1p) is established, which, if changed, for example through interploidy hybridization, results in aberrant development of the endosperm. These parent-of-origin effects, centred on the endosperm, are largely the result of differential genomic imprinting of maternal and paternal gametes [25–27,29]. Genomic imprinting occurs in both animal and plants and functions to silence genes in a gender-specific fashion. Many of the genes targeted by this epigenetic mechanism affect resource allocation to the developing embryo by maternal tissues and thus parental conflict is held to be one of the drivers behind its evolution [30,31]. In plants, genomic imprinting occurs principally in the endosperm , where maternally silenced genes encode growth promoters and paternally silenced genes encode growth inhibitors . Thus the large seeds that result from 2x×4x interploidy crosses in Arabidopsis are receiving an excess of paternally contributed growth-promoting genes which drive endosperm proliferation, whereas the reciprocal 4x×2x cross produces small seeds due to an excess of maternally active endosperm growth inhibitors. Importantly, as well as the endosperm funct-ioning as a nutritive tissue for the growing embryo, it also provides signals for co-ordinated development of the seed and hence plays a central role in its successful maturation [33,34]. A key event in the development of seeds is the point at which the endosperm cellularizes after an initial phase of free nuclear divisions without cytokinesis , and, as we shall see below, the timing of cellularization of this syncytial endosperm is a crucial determinant of seed size.
Endosperm cellularization and flavonoids
A combination of studies focused on identifying genes that function in the regulation of endosperm development and that are subject to genomic imprinting have together revealed a mechanism that explains the endosperm defects that result from interploidy crosses [28,36]. Central to this model is the FIS (FERTILIZATION-INDEPENDENT SEED) PcG (polycomb group) complex whose function is to restrict endosperm proliferation. The genes encoding this complex are subject to genomic imprinting, and expression is restricted to the maternal side where the complex represses (maternal) target gene expression via chromatin modification. Importantly, two key targets of the FIS complex are the PHE1 (PHERES1) and AGL62 (AGAMOUS-LIKE62) genes which act to increase endosperm size by delaying cellularization . In interploidy crosses, the FIS complex can lose control of appropriate regulation of endosperm development due to dysregulation of FIS target genes . This explains the seed phenotypes derived from paternal excess interploidy crosses which are characterized by overproliferation and late cellularization of the endosperm tissue, which frequently results in a concomitant arrest of embryo development and collapse of the seed [26,39].
Studies in Arabidopsis have uncovered variation in susceptibility to the triploid block with many ecotypes (e.g. C24 and Ler) being fairly tolerant of reciprocal 2x×4x interploidy crosses having more limited over- and under-proliferation of endosperm [14,39]. However crosses between 2x and 6x parents in these ecotypes display high levels of seed abortion characterized by extreme over- or under-proliferation of the endosperm, consistent with increased paternal or maternal contributions respectively, thus, in Arabidopsis, a ‘tetraploid block’ is a general feature . Interestingly, the Col (Columbia)-0 ecotype displays an asymmetric triploid block with 2x×4x (paternal excess) crosses producing very high frequencies of shrivelled seeds (75–90%), whereas the reciprocal cross produces small seeds more typical of those observed in Ler and C24 4x×2x crosses . This ‘Col syndrome’ is not, as yet, understood at the molecular level, but the reader is directed to the article by Scott et al.  for a full discussion of the potential mechanisms.
The identification of a potent maternal modifier, TTG2, of the Col syndrome that provides dramatic ‘rescue’ of 2x×4x crosses (with ttg2 as the seed parent) has highlighted the crucial role of the seed coat in regulating endosperm development and has also opened up the intriguing possibility that the flavonoid biosynthesis pathway may somehow titre an endosperm cellularization factor [14,39]. TTG encodes a maternally expressed WRKY transcription factor that has a role in regulating seed development through co-ordination of growth between the seed coat and zygotic tissues. ttg2 mutants have small seeds, reduced elongation of cells in the integuments and, in addition, are characterized by precocious cellularization of the endosperm . Mutant seeds are also pale in colour as they have a disrupted flavonoid biosynthesis pathway and therefore do not accumulate coloured flavonoid pigments in the endothelium, the innermost layer of the inner integument . TTG2 expression is principally localized to the endothelium in seeds where it acts as a positive regulator of this pathway . Could the flavonoid biosynthesis pathway therefore be responsible for precocious cellularization of endosperm and the concomitant reduction in seed size? Debeaujon et al.  surveyed a wide range of mutants defective in the flavonoid biosynthesis pathway and many had reduced seed size compared with wild-type, adding support to this thesis.
Two models have been proposed to explain how the ttg2 mutation can bring about early cellularization. The first, known as the ‘integument size-restriction model’, suggests that decreased integument cell expansion in ttg2 restricts the size of the embryo sac. This size reduction could thus raise the concentration of a cellularization factor to a point that triggers this developmental step earlier than in wild-type seeds . Although this is an attractive explanation, there appears to be no strong correlation between the volume of the embryo sac and ‘rescue’ of paternal excess resulting from interploidy crosses and thus an alternative mechanism seems likely . The second hypothesis, termed the ‘cellularization signalling model’, takes account of the disruption of flavonoid biosynthesis in the endothelial cell layer of the integuments. In this model, transport of a cellularization signal between the integuments and the endosperm occurs more readily in flavonoid biosynthetic mutants, triggering precocious endosperm cellularization . How transport is altered in this scenario is a matter of conjecture; however, it may simply relate to biophysical changes in the endothelium that alter its permeability to biomolecules or alternatively active processes may be involved (see the next section). Further evidence to support this model comes from interploidy crosses between a Ler seed parent carrying a mutation in the TT4 (TRANSPARENT TESTA 4) gene and a Col4x pollen parent where high frequencies of seed rescue were recorded  (Figure 2). Rescue was again characterized by a highly cellularized endosperm similar to that observed in ttg2×Col4x crosses. TT4 encodes the enzyme chalcone synthase which is the first committed step of the pathway.
The Arabidopsis ‘Columbia syndrome’ and maternal ‘rescue’ by a mutant of the flavonoid biosynthesis pathway
Although much data has now accumulated that points to a fundamental role for the endothelium in regulating endosperm cellularization, the observation by Debeaujon et al.  that ablation of this cell layer appears to have little effect on seed development is perplexing. By contrast, loss of the endothelium in petunia leads to degeneration of the endosperm . In physiological terms, complete absence of this flavonoid-producing cell layer is very different from a functional endothelium with altered flavonoid content and potentially altered transport capacity. Thus, until further data are available, this intriguing observation in Arabidopsis should be interpreted with caution.
A role for auxin transport?
Flavonoids are a diverse group of polyphenolic secondary plant compounds synthesized from phenylalanine and malonyl-CoA. Their roles are also varied, with some being colourful pigments that attract plant pollinators and others functioning as antimicrobial agents, feeding deterrents, signalling molecules and UV-screening compounds . Their biosynthesis has been extensively studied, and genes regulating the pathway and encoding its constituent enzymes have been largely characterized [44,45]. A simplified depiction of the pathway that operates in the Arabidopsis seed coat is shown in Figure 3. The major classes of flavonoids include the anthocyanins, flavonols, flavanols and the PAs (proanthocyanidins) or condensed tannins. In the endothelium of Arabidopsis seeds, anthocyanins and PAs accumulate in the vacuole with the PAs becoming oxidized in late seed development, giving seeds their brown colour .
The flavonoid biosynthesis pathway in the seed coat of Arabidopsis thaliana
A potential explanation for how endothelial flavonoids could modulate endosperm cellularization stems from the fact that, in addition to the roles outlined above, flavonoids are now recognized as potent auxin-transport regulators [46–49]. Auxin has a diverse suite of roles in plant growth and development, being important in the promotion of cell division, meristem maintenance, organogenesis and cell patterning . Thus, if flavonoids are directly involved in the mechanism that leads to the seed rescue phenotypes seen in ttg2 and tt4×Col4x crosses, a model that involves a known plant morphogen is attractive.
Evidence that auxin plays an important role in seed development comes from work on maize where rapid changes in auxin concentration have been measured in the endosperm . These changes in auxin distribution are mediated by the auxin efflux carrier ZmPIN1 (Zea mays PINFORMED1) in the endosperm and are defined during the cellularization phase . The exact mechanism whereby flavonoids exert their effects on auxin distribution is only partially understood, but seems to operate through both their direct binding to auxin transporters such as PGP (P-glycoprotein) 1, PGP4 and PGP19 and their indirect effect on PIN (PINFORMED) gene expression, although direct effects have not been ruled out [53–56]. Movement of auxin in plant tissues tends to be polarized and is directed by specific localization of carriers on the plasma membrane. In relation to this, the flavonol quercetin has been shown to promote asymmetric shifts in the localization of PIN1  and other flavonoids such as kaempferol and apigenin have also be shown to be active molecules in affecting polar auxin transport [46,49]. Importantly, mutations in the flavonoid biosynthesis pathway have been shown to have an impact on auxin transport, for instance tt4 plants have elevated basipetal transport of auxin and altered root responses to gravity and light as well as elevated transcription of PIN2 . Taken together, these data suggest that the important role of the endothelium and flavonoids in early seed development may well operate, at least in part, through the morphogenic properties of auxin.
Regulation of Fertilization and Early Seed Development: A Biochemical Society Focused Meeting held at the University of Bath, U.K., 11–13 September 2013. Organized and Edited by James Doughty (University of Bath, U.K.) and Thomas Dresselhaus (University of Regensburg, Germany).
AUXIN RESPONSE FACTOR2
CYTOCHROME P450 78A5
endosperm balance number
TRANSPARENT TESTA 4
TRANSPARENT TESTA GLABRA2
We thank Julia Watling for plant care and Dr Baoxiu Qi for excellent academic and technical assistance.
Financial support from the Ministry of Higher Education of Saudi Arabia (Ph.D. Scholarship awarded to M.A.) and the Biotechnology and Biological Sciences Research Council [grant number BBD0012341] is also gratefully acknowledged. We also thank Umm Al-Qura University, Saudi Arabia, for supporting M.A. in her scholarship bid.