The SAM is the source of all cells that ultimately form the shoot, including the subset that ends up building the leaves. Generally, cells in the central zone (CZ) of the SAM divide at a relatively low rate and remain in an undifferentiated state, whereas cells at the peripheral zone (PZ) divide faster and differentiate into organs such as leaves, axillary nodes, and floral parts (Veit, 2004; Braybrook and Kuhlemeier, 2010). In dicots, the SAM consists of three layers L1, L2, and L3; epidermal (L1) and subepidermal (L2) layers are known as tunica and the inner layer (L3) is called the corpus (Satina et al., 1940).

From the cellular perspective, on-going (slow) division in the stem cell niche will cause cells to become displaced away from the quiescent center, where at some well-defined place they lose their stem cell fate and acquire the actively dividing state. This transition is controlled by the interplay of a regulatory loop involving the homeodomain transcription factor WUSCHEL (WUS) in the rib zone (RZ) and CLAVATA gene products (CLV1, CLV2, and CLV3) expressed in the CZ of the SAM (Brand et al., 2000; Schoof et al., 2000; Carles and Fletcher, 2003; Yadav and Reddy, 2011). The WUS and CLV based pathway operates through two mobile signals: CLV3 and a hypothetical WUS mediated signal (Figure 2). CLV3 encodes a small secreted ligand that is produced specifically in L1 and L2 cells, and moves into the underlying L3 cells where it binds with receptor like proteins CLV1 (LRR receptor kinase) and/or CLV2 (receptor-like protein), which in turn inhibit WUS activity (Clark, 2001; Carles and Fletcher, 2003). WUS activity in the L3 cells induces the production of a non-cell-autonomous signal that moves to the stem cells and activates the expression of CLV3 there (Haecker and Laux, 2001; Braybrook and Kuhlemeier, 2010). It was proposed that the L1 produced miR394 signal is necessary for spatial organization of the SAM. This mobile microRNA regulates WUS mediated stem cell maintenance by inhibition of F box protein LEAF CURLING RESPONSIVENESS (LCR; Knauer et al., 2013).

Upon mutation in WUS the stem cells precociously transit into the peripheral actively dividing zone, ultimately consuming the stem cell niche and thereby the meristem. Inversely, in clv1 and clv3 mutants WUS activity of SAM cells is maintained much longer, whereby the stem cell niche and consequently the SAM as a whole enlarges dramatically (Clark et al., 1993, 1995; Laux et al., 1996). Several mathematical models have focused on the WUS–CLV interaction, predicting to various degrees how their expression domains are modulated through mutation or misexpression (Jonsson et al., 2005; Nikolaev et al., 2007; Hohm et al., 2010). Recent experimental studies supported by mathematical modeling have shown that WUS movement is essential for direct transcriptional repression of the differentiation program (Yadav et al., 2013) as well as in restricting its own accumulation through activating its negative regulator CLV3 (Yadav et al., 2011).

It has been postulated that signaling by the plant hormone cytokinin (CK) regulates WUS expression via CLV-dependent and CLV-independent mechanisms (Gordon et al., 2009) to promote SAM growth and maintenance with WUS repressing the transcription of ARABIDOPSIS type-A RESPONSE REGULATORS (ARRs), which are the negative regulators of CK signaling (Leibfried et al., 2005; Sablowski, 2007; Figure 2). Indeed, mutations in CK receptors (Higuchi et al., 2004) and over-expression of the CK dehydrogenase gene family of Arabidopsis (AtCKX; Werner et al., 2003) reduce meristem size and leaf area, indicating a relation between the SAM and leaf size. It appears however that that the number of leaf founder cells is not an important determinant of the final leaf size. For instance, a meta-analysis across a wide range of cactus species indicates that the size of the SAM correlates closely to the number of leaves formed and has only minor implications for their ultimate size (Mauseth, 2004).

Once progenitor cells are outside the stem cell niche, they need to decide whether they will contribute to the main axis or will differentiate into lateral appendices such as leaf primordia. This decision is primarily governed by the accumulation of the plant hormone auxin its influx carrier [AUXIN RESISTANT (AUX1) and its PIN-FORMED1 (PIN1)] efflux transporter (Bayer et al., 2009; Guenot et al., 2012). The efflux carriers orient the transport of auxin toward neighboring cells with a higher auxin concentration, leading to the formation of accumulation patterns across the cell population. Several mathematical modeling studies (reviewed in De Vos et al., 2012) have simulated phyllotactic patterning based on feedback interactions between auxin and PIN distribution. Some models postulate that AUX1 creates auxin accumulation mainly in L1 layer cells, whereas PIN1 is initially localized in the protodermal (L1) layer cells and causes drainage of auxin toward the base of the shoot by inducing vascular strand differentiation in L2/3 layer cells of the SAM (Reinhardt et al., 2003; de Reuille et al., 2006; Figure 3).

Arabidopsis leaf differentiation from the apical meristem is abolished in the auxin biosynthetic triple mutant yuc1 yuc4 pin1 (Cheng et al., 2007), whereas longer plastochron and irregular vegetative growth occurs in the pin1 mutant (Guenot et al., 2012). Cells in the SAM, which serve as a stem cell population, differ from the cells in the leaf primordium. This distinction is controlled by complex and still poorly characterized regulatory networks in which the antagonistic relation between two families of transcription factors, KNOTTED-like homeobox (KNOX1) and ASYMMETRIC LEAF1/ROUGH SHEATH2/PHANTASTICA (ARP) proteins (Byrne et al., 2002; Hay and Tsiantis, 2006, 2010) plays a crucial role (Figure 3). KNOX is expressed in all meristem cells except those at the site of the organ initiation (Jackson et al., 1994; Long et al., 1996), whereas the (ARP family) AS1 mRNA is expressed in the primordia forming cells, but not in the meristem (Byrne et al., 2000). KNOX1 is required to maintain undifferentiated cells in the SAM (Scofield and Murray, 2006) and increases CK biosynthesis (Yanai et al., 2005), whereas ARP initiates differentiation in the leaf primordium (Byrne et al., 2002). High levels of auxin restrain CK biosynthesis by the repression of KNOX1 activity (Su et al., 2011; Figure 3). High auxin and AS1 expression also suppress the expression of the KNOX gene BREVIPEDICELLUS (BP), which is required for leaf initiation (Hay et al., 2006). In Arabidopsis, the KNOX1 gene SHOOTMERISTEMLESS (STM) acts antagonistically with ARP gene products to control the induction of leaf primordia. The loss of function mutant stm fails to produce a SAM, also preventing leaf formation (Long et al., 1996), whereas the as1 and as2 mutants have small and round leaves (Byrne et al., 2002).

It has been demonstrated that the KNOX proteins trigger CK biosynthesis through the activation of IPT7 (encodes the CK biosynthetic enzyme isopentenyl transferase) and repress the transcription of gibberellin (GA) biosynthetic genes that encode GA20-oxidases (Sakamoto et al., 2001; Jasinski et al., 2005; Figure 3). Thus, high CK and low GA maintain stem cell identity in SAM cells by preventing cell differentiation (Gordon et al., 2009; Veit, 2009).

After acquiring “leaf” identity, the cells in the primordium have to develop a polarity gradient along the dorso-ventral axis. Once the position of the leaf primordium is established, a further increase in cell proliferation rates stimulates primordium outgrowth from the SAM. If this growing primordium is removed by tangential incision, another primordium arises which is cylindrical and abaxialized (lacking a flat leaf blade). This highlights the importance of signals originating in the SAM and received by cells in the primordium to determine polarity. This so called Sussex signal is yet to be identified (Sussex, 1951). Waites and Hudson (1995) proposed that the dorsal and ventral sides of the leaf are specified in the early development of the leaf primordium, when it is still located within the SAM. They showed that the PHAN gene (encoding a MYB type transcription factor) in Antirrhinum majus is involved in ab/ad-axial leaf polarity. Subsequently the phabulosa-1 (phb-1d) mutant was characterized in Arabidopsis whose leaves were unable to develop a blade and were radially symmetrical (McConnell and Barton, 1998).

Our knowledge of the regulation of antagonistic transcription factors specifying upper and lower sides has greatly increased, but the molecular signals exchanged between cells on both sides of the primordium to create this polarity are yet to be identified. Adaxial domain identity is determined by the expression of PHABULOSA (PHB), PHAVOLUTA (PHV), and REVOLUTA (REV) genes, which encode class III homeodomain-leucine zipper (HD-ZIPIII) proteins (McConnell et al., 2001). The identity of cells in the abaxial domain depends on the expression of KANADI [KAN; which encodes a Golden2/Arabidopsis response-regulator/P starvation/acclimatization response (Psr1; GARP) transcription factor; Eshed et al., 2001; Kerstetter et al., 2001] and the YABBY gene family (Siegfried et al., 1999; Eshed et al., 2004). These two classes of genes produce signals that suppress each other’s expression: the expression of PHB/PHV/REV genes in cells located at the abaxial side is inhibited de by KAN and inversely KAN expression in abaxially located cells in inhibited by the activity of PHB/PHV/REV genes, providing a feedback communication between the two sides (Tsukaya, 2013b; Figure 4).

Two small RNAs, the 21-nucleotide microRNA (miR165/166) and the 24-nucleotide transacting small interfering RNA (ta-siRNA), ta-siR-ARF, are also involved in determining leaf polarity (Chitwood et al., 2007, 2009; Nogueira et al., 2007). ARGONAUTE1 (AGO1) affects the regulation of miR165/166, which stimulates the cleavage of HD-ZIPIIIs transcripts in cells located on the adaxial side (Kidner and Martienssen, 2004) whereas, AGO7/ZIPPY (ZIP) stabilizes ta-siR-ARF, which further targets the degradation of auxin-related transcription factors, ETTIN (ETT)/ARF3 and ARF4 on the abaxial side (Adenot et al., 2006; Hunter et al., 2006). FILAMENTOUS FLOWER/YABBY3 (FIL/YAB3), a member of the YABBY family, up-regulates KAN1 and ARF4, which establishes a positive feedback loop (Bonaccorso et al., 2012; Figure 4).

Differences in cell growth rates along the principal developmental axes are crucial in determining final leaf shape. In addition to specification of adaxial and abaxial side of the leaf, margin specific cell fate is induced in cells residing at the boundary between these two surfaces (McHale, 1993). In contrast to regulation of leaf blade outgrowth, the influence of ad/abaxial specific genes on marginal cells is yet to be explored. Recently, middle leaf domain specific WUS-RELATED HOMEOBOX (WOX) genes were reported, which affect leaf blade outgrowth and margin specific development (Figure 4). These transcription factors (WOX1 and PRESSED FLOWER (PRS), i.e., WOX3) are repressed by KAN. The loss of function of WOX1 and PRS causes instable organization of ad-abaxial polarity (Nakata and Okada, 2012; Nakata et al., 2012).

A relatively simple computational model supported by time-lapse data and clonal analysis by Kuchen et al. (2012) accounts for local differences in cell growth rates and direction driving organ level shape changes during early Arabidopsis leaf development. A central model assumption is a yet uncharacterized early tissue polarity system that deforms during growth. CUP-SHAPED COTYLEDON (CUC) genes are emerging as prime candidate organizers of tissue polarity (Hasson et al., 2011). Correspondence of cell polarity and PIN1 auxin transporter patterns points to their involvement in such an organizer-based model (Scarpella et al., 2006). Moreover, it was reported that the outgrowth of lobes at the leaf margin is specified by a local auxin maximum as a result of the polar distribution of the PIN1 transporter (Hay et al., 2006). The transcription factor CUC2 which is expressed at the leaf sinuses and is negatively regulated by miR164 (Nikovics et al., 2006), promotes generation of these PIN1-dependent auxin maxima, which was supported by computer simulations (Bilsborough et al., 2011). Recently, a homeodomain protein REDUCED COMPEXITY (RCO) was reported to enhance serration by repressing growth at the flanks of initiating leaves (Vlad et al., 2014). Interestingly, CUC2, PIN1, and TEOSINTE BRANCHED/CYCLOIDEA/PCF (TCP), and KNOX have been implicated in compound leaf development illustrating that this regulatory system has the capacity to take more extreme forms than observed in Arabidopsis leaves (Barkoulas et al., 2008; Hay and Tsiantis, 2010; Koyama et al., 2010). A better understanding of the regulatory mechanisms operating at the cell level during the early phases of leaf outgrowth will likely provide invaluable insights into how diverse leaf morphologies are established.

In contrast to the morphology of the leaf primordium, the final size and shape of the leaf differ widely among species. Differences in leaf outgrowth are often interpreted as the result of cell division producing a certain number of cells and subsequent cell expansion determining their mature size. However, this is an overly simplistic view. Firstly, the relationship between cell division and expansion is complex and the two processes can mutually compensate each other (Tsukaya, 2002; Beemster et al., 2003). A theoretical framework to understand this phenomenon was provided by Green (1976), who proposed that cell growth and partitioning (division sensu-strictu) are two processes that co-occur in proliferating cells, whereas in expanding cells cell-growth continues in absence of partitioning. Clearly, this framework allows for continued growth irrespective of inhibited or stimulated cell division activity, at least until cells get too small or too large to function normally. However, the view that cell growth in proliferating cells is equivalent to that in expanding cells is overly simplistic. It is clear that whereas dividing cells grow by increasing cytoplasmic volume, expanding cells primarily increase their internal volume by expanding their vacuolar volume.

Cytoplasmic growth is mainly based on macromolecular synthesis and therefore consumes a lot of energy. A crucial role in ensuring a sufficient supply of elementary building blocks is played by the Target of Rapamycin (TOR) pathway. TOR, a Serine/Threonine kinase of the phosphatidylinositol-3-kinase-related kinase (PIKK) family is an essential controller of cytoplasmic growth and metabolism in plant cells. It controls a range of cellular responses such as ribosome biogenesis, translational initiation, cell proliferation, cell expansion and autophagy (Zhang et al., 2013; Sablowski and Dornelas, 2014; Figure 5).

In yeast and animals, there are two TOR complexes: TORC1 and TORC2 whereas in plants there is only evidence for TORC1. TOR, Regulatory associated protein of TOR (RAPTOR), and Lethal with Sec13 protein 8 (LST8) are three key components of TORC1 in Arabidopsis (Moreau et al., 2012). In contrast to other eukaryotes, much less is known about TOR signaling in plants. Arabidopsis thaliana is insensitive to the drug rapamycin, which is extensively used to study TOR function in yeast and animal systems, which has formed a major obstacle to study TOR in plants (Menand et al., 2002; Anderson et al., 2005; Deprost et al., 2007; Sormani et al., 2007). However, a recent study shows that plants do respond to rapamycin at the concentration of 10 μM, which is 100 times more than the concentration used for yeast and mammals (Xiong and Sheen, 2012). Mutation in TOR is embryo-lethal in plants (Menand et al., 2002; Ren et al., 2011) and therefore, an alternative approach of inducible knockout for conditional inhibition of TOR has to be used for functional studies (Deprost et al., 2007; Caldana et al., 2013).

An Arabidopsis T-DNA insertion line which overexpresses AtTOR produces bigger leaves containing larger cells (Deprost et al., 2007). Accordingly, down-regulation of TOR by inducible artificial microRNA produced smaller leaves with fewer cells along with up-regulation of metabolic pathways like the Krebs cycle. These lines also accumulated storage products such as starch and lipids like triacylglycerides (TAGs) displaying reduced growth (Caldana et al., 2013). Mutation of LST8, a member of the TOR complex (TORC1) decreased plant size by producing fewer and smaller leaves alongside a higher amount of starch and amino acids and reduced sucrose concentrations. Moreover, down-regulation of genes involved in cell wall formation like expansins (EXPs) and CELLULOSE SYNTHASE-LIKE G3 in this mutant demonstrates a role for TOR in cell expansion (Moreau et al., 2012). The TORC1 component RAPTOR accumulates in dividing and expanding cells whereas mutation in AtRaptor1B slows down the leaf initiation. A double mutant of AtRaptor1A and AtRaptor1B exhibited normal seedling growth, but was unable to maintain post embryonic development (Anderson et al., 2005).

TORC1 promotes the phosphorylation of S6 kinase (S6K) and eIF4E-binding proteins (E-BP1), which controls translation and ribosome biogenesis (Krizek, 2009). Again, auxin is involved through the regulation of EBP1 protein stability in Arabidopsis (Horvath et al., 2006; Figure 5). TORC1 also regulates autophagocytosis which ensures synthesis, degradation and recycling of cellular components (Sablowski and Dornelas, 2014). Recent studies have revealed that the TOR pathway is an essential controller for cellular development, which regulates cell expansion and cell cycle simultaneously. The TOR pathway is directly connected to cell cycle regulation by mediating E2Fa phosphorylation and activation of DNA synthesis genes (Xiong et al., 2013) and regulates cell wall modification and degradation processes like senescence and autophagy (Caldana et al., 2013). Importantly, there is a link between the TOR pathway and nutrient status. TOR plays a central role in connecting growth related genes to glucose signaling (Xiong et al., 2013).

Given this central role in connecting growth regulation to nutritional status, the TOR pathway is a key regulatory hub in organ development. Although molecular insight of its functioning in animal systems is rapidly increasing, currently we are not aware of mechanistic models that include this knowledge, a void that we expect to be filled in the coming years.

In addition to cell volume increase by cytoplasmic growth, cell proliferation exponentially increases the number of cells in the developing leaf. In general cell growth needs to be sufficiently balanced by cell division for stable tissue growth (Sablowski and Dornelas, 2014). The cell division cycle is a unidirectional process, tightly regulated by a molecular mechanism that is largely conserved between all eukaryotes (Inze et al., 1999; Dewitte and Murray, 2003; Inze and De Veylder, 2006; Figure 6).

The plant cell cycle is controlled by the activity of complexes consisting of a cyclin-dependent kinase as the catalytic subunit and a cyclin as the regulatory subunit. A-type cyclin dependent kinase (CDKA) and D-type cyclin (CYCD) are central to the G1/S phase transition in which the cell activates DNA duplication. CDKA is a key protein to control cell division in A. thaliana, and is present at a constant level throughout the cell cycle (Porceddu et al., 2001; Joubes et al., 2004; Gaamouche et al., 2010). Overexpression of a dominant negative CDKA;1 of A. thaliana in tobacco plants inhibited cell division rate, resulting in the formation of fewer, but larger cells resulting in an overall reduction of leaf area (Hemerly et al., 1995). It has also been demonstrated that the CDKA;1 activity maintains the SAM cells in an undifferentiated state. Expressing a dominant negative allele of CDKA;1 in the STM domain of the shoot apex produces smaller leaves with a reduced number of epidermal cells (Gaamouche et al., 2010). Interestingly, irregularly shaped epidermal cells observed in a CDKA;1 dominant negative mutant expressed from the STM promoter point toward CDKA influencing cell wall and cytoskeleton properties (Borowska-Wykret et al., 2013). A triple cycd3;1–3 loss of function mutant in the Arabidopsis leaf shows a decreased cell number and reduced CK response (Dewitte et al., 2007). The plant hormones auxin, CK, brassinosteroid (BR), and GAs increase the level of CYCD, thereby activating CDKA (Riou-Khamlichi et al., 1999; Francis and Sorrell, 2001; Inze and De Veylder, 2006; Perrot-Rechenmann, 2010). The repression of ABP1 (AUXIN BINDING PROTEIN1) negatively influences transcript levels of CYCDs, which results in impaired cell division in the leaf (Braun et al., 2008), whereas BRs up-regulate the expression of CYCD3 and promote cell division through a mechanism that requires de novo protein synthesis (Hu et al., 2000). The BR biosynthesis mutant det2 (de-etiolated2 = cro1) and dwf1 (dwarf1 = cro2) produce fewer cells and a smaller leaf, which can be reversed by brassinolide application, indicating a dual role of BR in division and expansion (Nakaya et al., 2002).

The activity of CDKA/CYCD complexes is itself controlled by CDK activating kinases CDKD and CDKF coupled with CYCH, which activates the complex through a phosphorylation cascade. CDKF;1 was also found to activate CDKD;2 and CDKD;3 by T-loop phosphorylation (Umeda et al., 2005; Takatsuka et al., 2009). The active CDKA/CYCD complex triggers the dissociation of E2F/DP heterodimeric complex from RBR (retinoblastoma-related protein) through phosphorylation. Additionally, it initiates the destruction of E2Fc/DP/RBR transcriptional repressor complex by the Skp-Cullin1-F-Box (SCF) E3-ubiquitin protein ligase (Inze and De Veylder, 2006). The RBR protein regulates the activity of E2F transcription factors to control cell proliferation. It is an essential regulator of the cell cycle to coordinate between cell division, differentiation and homeostasis (Desvoyes et al., 2006; Borghi et al., 2010). Once the E2F/DP complex is separated from RBR, it initiates G1 to S transition by activating the transcription of genes required for DNA duplication. Furthermore, the E2F-like DEL transcription factors compete with E2F/DP for DNA binding sites (Figure 6).

Arabidopsis has three typical E2Fs (E2Fa, E2Fb, and E2Fc), two dimerization proteins (DPa and DPb) and three atypical E2Fs (E2F/DEL2, E2FE/DEL1, and E2Ff/DEL3; Mariconti et al., 2002). E2Fa and E2Fb stimulate entry into and progression of the S-phase and overexpression of these transcription factors leads to an enlarged phenotype due to enhanced cell proliferation (De Veylder et al., 2002; Sozzani et al., 2006). Auxin positively regulates E2Fb protein levels (Magyar et al., 2005). Additionally, the AXR1 transcript level was found to be high in an E2Fb overexpression line (Sozzani et al., 2006). On the other hand, E2Fc is a negative regulator of the S-phase where a decreased level leads to leaves and cotyledons with more but smaller cells (del Pozo et al., 2006). Auxin affects the E2Fc protein level in Arabidopsis. Mutation in the AXR1 gene leads to impaired modification of the CUL1 protein, a structural component of the E3–SCF complex, with the Ub-related protein RUB, and shows increased E2Fc protein levels (del Pozo et al., 2002). Atypical E2Fs/DELs are transcriptional regulators for endoploidization, which act independently of DPs and RBR. It is still unclear if they compete with typical E2Fs for binding sites or actively repress gene transcription (Vlieghe et al., 2005; Berckmans and De Veylder, 2009).

Activated E2Fs in Arabidopsis target the expression of genes involved in DNA repair and chromatin dynamics such as CDC6, CDT1, MCM3, ORC1 and ORC3, RNR, and PCNA. However, they also influence the expression of genes responsible for G2-M transition like CDKB1, MYB, and ANAPHASE-PROMOTING COMPLEX/CYCLOSOME (APC/C; de Jager et al., 2001; Ramirez-Parra et al., 2003; Boudolf et al., 2004b; Vandepoele et al., 2005; Lammens et al., 2008; Naouar et al., 2009). Replication origin sites are silenced in the G1 phase where CDC6 and CDT1 together with ORC allow the loading of MCM to the replication origins; hence promoting the complex for activation of the S phase. Later, the DBF–CDC7 complex phosphorylates ORC which then moves and exposes the replication initiation site for the replisome complex, allowing replication to start (Blow and Dutta, 2005; Francis, 2007). The plant hormone ABA negatively regulates the expression of the CDT1a gene (Castellano Mdel et al., 2004). Methyl jasmonate (MeJA) was also reported to affect initiation of DNA replication by inhibiting the pre-replication complex (Noir et al., 2013; Figure 6).

The Kip related proteins (KRPs) are direct inhibitors of CDK activity (ICKs). ICK1 inhibits the CDKA/CYCD complex in response to negative stimuli like abscisic acid (ABA; Wang et al., 1998; Van Leene et al., 2011). Kinematic analysis showed that the overexpression of KRP genes inhibits cell division rate, resulting in serrated leaves of reduced size with fewer but enlarged cells (De Veylder et al., 2001; Kang et al., 2007). Beemster et al. (2006) could simulate the effect on cell numbers using a computational model. Mutation of a single KRP gene does not have any effect on leaf growth, while down-regulation of multiple KRP genes increases final leaf area and cell proliferation (Cheng et al., 2013). The mechanism behind this enhanced leaf growth is yet to be explained. Auxin and CK activate CDKA and CYCD, while KRP4 transcription is down-regulated (Cho et al., 2010). It has been reported that auxin signaling is translated into modified KRP expression through PRZ1-mediated chromatin remodeling (Anzola et al., 2010). The mutual antagonistic effect of CDKA;1 and KRPs was highlighted in a model of the G1/S transition control in pollen (Zhao et al., 2012). The plant-specific F-box protein F-BOX LIKE 17 (FBL-17) is central in the proposed regulatory network, in particular by mediating the degradation of KRP1,3,4,6,7 (Kim et al., 2008). The plant-specific SIAMESE (SIM)/SIAMESE RELATED (SMR) family also inhibits CDK activity in a number of specific tissues, for instance the repression of the mitotic cycle in trichomes (Walker et al., 2000). GA is proposed to promote mitotic cycles by lowering expression of both KRP and SIM. CKIs act in a DELLA-dependent manner (Achard et al., 2009) and enhance expression of E2Fe/DEL1 (Claeys et al., 2012).

After DNA duplication, cells enter the G2 phase to prepare them for division through mitosis. CDKA and CDKB as well as CYCA, CYCB, and CYCD are involved in this process. The plant-specific B-type CDKs are subdivided into two groups: CDKB1 (with CDKB1;1 and CDKB1;2) and CDKB2 (with CDKB2;1 and CDKB2;2). CDKB1 accumulates from late S to M phase while CDKB2 is specifically expressed from G2 to M phase (Menges et al., 2005). Overexpression of a dominant negative CDKB1;1 causes early exit from the M phase, which increases the ploidy level in the leaf (Boudolf et al., 2004b). It has been reported that CDKB1;1 forms a functional complex with CYCA2;3 to trigger mitosis in Arabidopsis (Boudolf et al., 2009). Inhibition of CDKB2;1 and CDKB2;2 via the expression of an amiRNA leads to a dwarf phenotype with an abnormal SAM (Andersen et al., 2008). Jasmonates (JAs) cause G2 arrest in tobacco (Nicotiana tabacum) Bright Yellow-2 cell cultures by reducing CYCB1;1 and CDKB (Swiatek et al., 2004). Similarly, inhibition of mitotic phase genes by methyl JA causes G2 arrest in Arabidopsis cell cultures (Pauwels et al., 2008).

An additional level of regulation of CDKA/CYCD complex activity involves inactivation through phosphorylation by the WEE1 protein kinase. Overexpression of WEE1 inhibits plant growth by arresting cell division (De Schutter et al., 2007). However, in plants WEE1 is probably not a core cell cycle regulator, but rather a DNA damage checkpoint kinase (Dissmeyer et al., 2009).

Dissmeyer et al. (2009) have proposed and implemented alternative G2 phase modules in the form of mathematical models starting from an existing generic model. The primitive unicellular algae, Ostreococcus tauri, contain a bona fide CDC25, which antagonizes WEE1 phosphorylation (Inze and De Veylder, 2006). In Arabidopsis a small CDC25 like phosphatase can counteract the role of WEE1 kinase in vitro (Landrieu et al., 2004), but this CDC25-like protein has arsenate reductase activity and is most likely not involved in cell cycle regulation (Dissmeyer et al., 2010). The Arabidopsis genome therefore lacks a functional copy of the CDC25 gene, which means that generic cell cycle models (Novak and Tyson, 1993) need to be adapted to reflect the situation in plants.

Similar to the G1 phase, the CDK/CYC complex can be activated in G2 by a CDK-activating kinase pathway, involving CDKF and CDKD coupled with CYCH. The activated CDK/CYC complex promotes MYB repeat (MYB3R) transcription factors to bind with M phase Specific Activators (MAS) elements at the promoter region of the target genes. Afterward, MYB3R phosphorylation activates the expression of M phase specific genes such as KNOLLE, CDC20, CYCA, and CYCB and NACK1 (Berckmans and De Veylder, 2009; Figure 6).

There are two E3 ubiquitin ligase complexes involved in cell cycle control, SCF-PROTEIN and APC/C, which mark targets for degradation by the 26S proteasome. In analogy to the SCF complex playing an important role in G1 to S phase by degrading cell cycle inhibitors (KRPs; Fulop et al., 2005; Hershko, 2005; Ren et al., 2008) the APC/C is essential for the G2 to M transition. The exit from the M phase is regulated by the degradation of cyclins through ubiquitination by the APC/C in association with the activators CELL CYCLE SWITCH 52 (CCS52) and CDC20 (Fulop et al., 2005; Sullivan and Morgan, 2007; Marrocco et al., 2010).

Overexpression of APC10 promotes the cell division rate by degradation of CYCB1;1 which causes enlarged leaves (Eloy et al., 2011). Mutation of HOBBIT, a CDC27 subunit of the APC, in the SAM leads to accumulation of high levels of the auxin response inhibitor AXR3/IAA17, indicating that its activity would be involved in targeting AUX/IAA proteins for degradation (Blilou et al., 2002). CCS52 is an important regulator for controlling exit of mitosis. There are two classes of CCS52 in A. thaliana, CCS52A (CCS52A1 and CCS52A2) and CCS52B (Fulop et al., 2005). A-type CCS52 activators are typically expressed from late M to late S-G2 and regulate the onset of endoreduplication in leaves (Fulop et al., 2005; Lammens et al., 2008). CCS52B, like the APC/C activator CDC20, peaks from early G2 to M phase exit (Fulop et al., 2005; Kevei et al., 2011). Induced expression of CCS52B affects branching in trichomes where CCS52B^OE line forms four to five branches, while the wild type has only three branches (Engler et al., 2012). However, more study is needed in particular to understand role of CCS52B in leaf development. It has been suggested that complementary phase-dependent expression of A and B-type CCS52 activators enable a fine-tuned APC/C regulation during the cell cycle (Tarayre et al., 2004). Expression of the negative regulator of CCS52A1 activity, ULTRAVIOLET-B-INSENSITIVE 4 (UVI4) peaks at the G1-to-S transition (Heyman et al., 2011) and determines cell number and size of leaves, likely through stabilization of CYCA2;3 required for mitotic cell divisions (Imai et al., 2006; Boudolf et al., 2009; Heyman et al., 2011). Its homolog UVI-Like/OMISSION OF SECOND DIVISION 1 (OSD1) influences meiosis and does not directly impact leaf size. However, it is expressed during the mitotic cell cycle peaking at the G2-to-M transition, possibly preventing endomitosis (=incomplete mitosis; Heyman and De Veylder, 2012). Besides the mechanism of the inhibitory action of the UVI4 and OSD1 regulators, not much is known about the specific targets of the APC/C (Heyman and De Veylder, 2012). Apart from a recent study pointing to CK up-regulating CCS52A1 in the Arabidopsis root (Takahashi et al., 2013), not much is known about the role of hormone signaling on APC/C regulation. However, in stress-conditions GA likely modulates APC/C activity through DELLA dependent down-regulation of the UVI4 and DEL1 negative regulators (Claeys et al., 2012).

Despite of the high level of conservation of the core cell cycle machinery (Harashima et al., 2013), many plant-specific features exist. Plants are characterized for instance by a remarkably broad cyclin family with many species-specific isoforms (at least 49 in Arabidopsis, cf. Sablowski and Dornelas, 2014). The involvement of different orthologous of many core cell cycle genes goes together with a functional diversity which manifests itself in gene expression differences between species, developmental and environmental conditions. Indeed, Beemster et al. (2005) showed by means of a microarray study that the expression profile of roughly half of all cell cycle genes differed between roots and leaf primordia. de Almeida Engler et al. (2009) found generally high expression of core cell cycle genes in leaf primordia, in the lamina of young leaves and in vascular tissue of expanding leaves. A number of cases of developmental stage-specific expression are described in the relevant sections throughout this review. However, for many others the functional significance has yet to be clarified.

Generally cells go through a regular cell cycle with the S phase (DNA duplication) followed by the M phase (mitosis). Endoreduplication, endoreplication, endoploidization or, in short, the endocycle is the process whereby DNA replicates repeatedly without alternating divisions through mitosis, causing a high ploidy level in the cell.

To establish endoreduplication, the CDK activity essentially has to be kept low enough and several ways have been proposed to achieve this (Berckmans and De Veylder, 2009; De Veylder et al., 2011). For instance, CYCD3;1 which is specifically expressed in proliferating tissues, reduces endoploidization (Dewitte et al., 2003). CDKB1 activity is also essential for the G2 to M transition (Beemster et al., 2005). Overexpression of a dominant negative CDKB1;1 interferes with cell cycle progression causing G2 arrest (Boudolf et al., 2004a). CDKB1;1 forms an active complex with CYCA2;3 to suppress endoreplication in the leaf (Boudolf et al., 2009). Loss of CYCA2;3 function increases ploidy in mature leaves (Imai et al., 2006). The INCREASE LEVELS OF PLOIDY (ILP1) gene, which encodes a protein homologous to the C terminal region of mammalian GC binding factor, is proposed to be involved in transcriptional repression of A2-type cyclins (Yoshizumi et al., 2006). Expression of B-type cyclins, on the other hand, was repressed by decreased phosphorylation of three-repeat MYB proteins (MYB3Rs; Ito et al., 2001; De Veylder et al., 2011). The E3 ubiquitin ligase complex, APC/C coupled with CCS52 influences endocycle onset by controlling proteolytic degradation of G2-M specific cyclins like CYCB1;1 and CYCB1;2 (Kasili et al., 2010) as well as CYCA2;3 (Boudolf et al., 2009; Figure 7). CCS52A1 and CCS52A2 knockout plants have reduced DNA ploidy levels in leaves (Lammens et al., 2008; Kasili et al., 2010). The previously mentioned plant-specific CCS52A1 inhibitor UVI4 is likely involved in securing the G2-to-M transition and therefore preventing endocycle onset (Heyman and De Veylder, 2012). Cells with increased ploidy levels in osd1 cotyledons and the developmentally severely compromised uvi4 osd1 suggest some functional redundancy between UVI4 and its homolog OSD1 (Iwata et al., 2011; Cromer et al., 2012). Mutation in SAMBA, a plant specific subunit of the APC complex which probably activates A2-type cyclin degradation, induces enhanced endoreplication in Arabidopsis leaves (Eloy et al., 2012). Other factors such as the DP/E2F like transcription factor E2Fe/DEL1 are involved in controlling APC/C activity. Down-regulation of DEL1 triggers the expression of the CCS52A2 gene, forcing cells to enter endoreduplication (Lammens et al., 2008).

The plant specific cell cycle inhibitor SIM gene that encodes for a member of the SMR family also plays a role in endoreduplication (Walker et al., 2000; Churchman et al., 2006). A mutation in SIM causes repressed endoreplication leading instead to mitotic divisions in leaf trichomes. SIM interacts with CDKA;1 and D type CYCs (Churchman et al., 2006), and it was suggested that inhibition of CDKA:CYCD3 complexes might be the mechanism responsible for its role in endoreduplication onset (De Veylder et al., 2011). Indeed, CYCD3;1 overexpression inhibits endoreduplication in Arabidopsis leaves (Dewitte et al., 2003), whereas a cycd3 triple mutant displays premature onset of endoreduplication in young leaves (Dewitte et al., 2007). Other SIM family members, such as SMR1/LGO, might also promote polyploidization (Roeder et al., 2010). In fact, SMR1 and SMR2 interact with CDKB1;1 and its interactor CYCB2;4 associates with SMR11 (Van Leene et al., 2010).

Contrary to the emerging insights into endoreduplication onset, it is only poorly understood how the endocycle is sustained. It has been envisaged that the cell cycle inhibitor KRP controls CDKA activity by inhibiting the CDKA/CYCD complex to maintain the CDK oscillations needed for DNA replication in the endocycle. KRPs were reported to regulate mitosis and endoreplication in a dose dependent manner where low concentrations promote the endocycle while high levels cause cell cycle arrest (Verkest et al., 2005b). On the one hand, overexpression of KRP2/KRP5 in mitotically active cells inhibits cell division and enhances endoreplication (Verkest et al., 2005a; Jegu et al., 2013) while on the other hand its overexpression in postmitotic cells inhibits endocycle in Arabidopsis leaves (Schnittger et al., 2003). This implicates that KRPs are an important candidate for the regulation of rate and duration of endoreduplication in expanding leaf cells.

Arabidopsis leaves, cells enter into the endoreduplication process as a consequence of decreasing auxin concentrations. It has been observed that the mutants in auxin signaling, biosynthesis and transport show a rapid transition from mitotis to endocycle causing increased ploidy level in cotyledons (Ishida et al., 2010) but the detail of this mechanism is still not known. Ethylene and GAs are hypothesized to positively affect endoreduplication (Gendreau et al., 1999; Perazza et al., 1999; Swain et al., 2002). JAs were shown to inhibit cell proliferation as well as endoploidization in a COI1 (encoding an F-box protein which is a part of the SCF complex) dependent manner in Arabidopsis leaves (Noir et al., 2013). It also negatively regulates the expression of key determinants of DNA replication like CDC6A (Noir et al., 2013).

Despite of our increasing knowledge on the molecular mechanism of endoreduplication, its actual function remains ambiguous with proposed roles in promoting cell expansion, stress resistance or DNA damage protection (De Veylder et al., 2011). In any case, modulating CDK–CYC activity levels in various ways remains a central principle. This pertains to the ubiquitin dependent degradation of KRPs or the mechanism by which plant specific cell cycle inhibitors (SIM/SMR) as well as developmental and environmental signals influence the endocycle. Roodbarkelari et al. (2010) have modeled endocycle onset in Arabidopsis trichomes with KRP and the CULLIN4 ubiquitin ligase controlling G1/S and SIM and APC/C controlling G2/M transitions. Computational approaches to predict tissue distributions of cell ploidy combined with in vivo ploidy maps (Boudolf et al., 2004b) would provide powerful insights to better understand its regulation and relationship to cell expansion.

Leaf development involves two major phases. The first phase is dominated by proliferative activity and the second phase by cell expansion (Figure 8). There is a correlation between cell division activity and organ growth, so the timing of cell division has a large influence on the final leaf size (Korner et al., 1989; Meyerowitz, 1997; Gonzalez et al., 2012). As cell division ceases the cell continues expanding. This transition from division to expansion is manifested as a cell cycle arrest front which remains fixed at some position for a particular time period and then moves rapidly toward the base of the leaf blade (Andriankaja et al., 2012). Several regulators appear to control the transition from proliferation to expansion. Auxin plays an important role in the transition phase. It induces the expression of AUXIN-REGULATED GENE INVOLVED IN ORGAN SIZE (ARGOS) gene, encoding for an ER localized protein of unknown function (Hu et al., 2003). Overexpression and down-regulation of ARGOS increases and decreases leaf size, respectively. It regulates the action of a DNA-binding protein ANT (AINTEGUMENTA) and of CYCD3;1 (Hu et al., 2003; Figure 8). Loss of function of ANT blocks the increase in leaf growth in ARGOS overexpressing plants. The Arabidopsis ORGAN SIZE RELATED1 (ORS1) shares a conserved domain with ARGOS and ARGOS LIKE (ARL), and positively regulates cell division and expansion in the leaf (Hu et al., 2003; Feng et al., 2011). Like the ANT family proteins, the GROWTH REGULATING FACTOR (GRF) and TCP transcription factors are essential regulators of leaf growth (Figure 8). The Arabidopsis GRF family comprises nine members. Overexpression of AtGRF1 and AtGRF2 results in larger leaves whereas the grf1/2/3 triple mutant reduces leaf size, both as a result of alterations in cell size (Kim et al., 2003). Overexpression of GRF5 increased cell number with prolonged growth, whereas the grf5-1 mutant shows narrow leaves with reduced cell numbers. GIF1 (GRF-INTERACTING FACTOR1), also known as ANGUSTIFOLIA3 (AN3) interacts with GRF5 (Horiguchi et al., 2005). GIF1 overexpression increases leaf size with leaves having more cells, whereas its absence reduces cell proliferation (Lee et al., 2009).

miR396 negatively regulates six members of Arabidopsis GRF together with GIF1 (Liu et al., 2009; Figure 8). Interestingly, overexpression of miR396 in a mutant deficient for GRF1 reduces SAM size (Rodriguez et al., 2010). The miR396 targeted GRFs are also essential for leaf polarity (Wang et al., 2011). The TCP family of transcription factors regulates the expression of miR396 and miR319 (Palatnik et al., 2003; Rodriguez et al., 2010). A point mutation in the miR319 target site of TCP4 induces miR396 which in turn decreases GRF expression and results in smaller leaves (Figure 8). Similarly, overexpression of TPC4 decreases leaf size (Rodriguez et al., 2010). Transcription factors from the TCP family such as CINCINNATA (CIN) of Antirrhinum, LANCEOLATE (LA) of tomato and CIN-TCPs of A. thaliana control cell cycle arrest (Nath et al., 2003; Palatnik et al., 2003; Ori et al., 2007). In Arabidopsis, up-regulation of miR319 in the jaw-D mutant reduces the expression of TCP2, TCP3, TCP4, TCP10, and TCP24 producing large and wrinkled leaves (Palatnik et al., 2003). Down-regulation of single, double, and triple TPC genes resulted in proportional increase in leaf size and crinkliness (Schommer et al., 2008).

In addition to these transcription factors, other genes are also essential to promote the transition from division to expansion. The putative ubiquitin binding protein DA1 and the E3 ubiquitin ligase BIG BROTHER (BB) also known as ENHANCER OF DA1-1 (EOD1) controls organ size by restricting the duration of cell proliferation (Figure 8). In the da1-1 mutant, the production of a dominant negative protein negatively affects both DA1 and the DA1-related (DAR) protein and the overexpression of DA1 results in large leaves with increased cell numbers. ABA induces the expression of DA1, whereas the da1-1 mutant was less sensitive to ABA, implicating a role for ABA in determining final leaf size through control of mitotic exit (Li et al., 2008). Mediator complex subunit 25 (MED25 also known as PFT1), functions together with DA1 in controlling leaf growth by restricting cell proliferation (Figure 8). Overexpression of MED25 causes smaller leaves with reduced cell numbers and cell sizes, whereas a loss of function mutant enhances organ size with increased duration of cell proliferation and expansion (Xu and Li, 2011). Loss of function mutation of the RING-finger protein encoding BB leads to enlarged leaves and small changes in expression levels substantially alter organ size suggesting it controls cell division and leaf size in a dose dependent manner (Disch et al., 2006). The KLUH (KLU)/CYP78A5 gene, encoding for a cytochrome P450, required for generating a mobile growth signal distinct from the classical phytohormones, is also an essential regulator for leaf size control. Overexpression of KLU induces enlarged leaves having more cells whereas in the klu mutant premature arrest of cell proliferation causes smaller leaves (Anastasiou et al., 2007; Stransfeld et al., 2010). The SWP gene encodes a protein with similarities to subunits of the Mediator transcriptional regulatory complex of RNA polymerase II. It also plays a role in defining the period of cell proliferation. In the swp mutant leaf size was reduced due to less cells, which was partially compensated by an increase in final cell size (Autran et al., 2002; Figure 8).

Like auxin, sugar signaling controls leaf growth possibly via the ARGOS pathway (Hu et al., 2003; Wang and Ruan, 2013). BR also regulates leaf growth by controlling cell division and expansion. The BR deficient mutant constitutive photomorphogenesis and dwarfism (cpd) produces smaller leaves with fewer cells of reduced size (Zhiponova et al., 2013); however, the molecular mechanism controlling this process is yet to be clarified. The progressive general cell proliferation arrest front of epidermal and mesophyll cells is followed by a second cell cycle arrest front for dispersed meristematic cells (DMCs) that is controlled by the putative transcription factors PPD1 (PEAPOD1) and PPD2 (White, 2006).

The transition to the expansion phase is essentially dependent on the regulators of cell cycle arrest. Many factors have been implicated in the regulation of the cell division arrest front. An important question is how the spatiotemporal dynamics of the arrest front could be explained. Coordination through one or more gradients of (non-cell autonomous) growth regulators appears to be the most likely mechanism. However, Efroni et al. (2008) hypothesized a mechanism for organ differentiation through an internal self-advancing sequential maturation program, where rate and time of advancement is regulated by a cell-autonomous developmental clock (timed) program. CIN-TCPs would play a governing role in this mechanism for the leaf. Sufficient details for building simulation models of the transition phase still appear to be lacking in Arabidopsis. However, in the monocotyledonous maize leaf a clearer picture is arising, where a peak in the activity of GA is instrumental in regulating the spatial location of the transition (Nelissen et al., 2012). Possibly models that will be developed for this monocotyledonous system can be adapted to better understand the same process in the Arabidopsis leaf.

Cell expansion is an essential step in determining final leaf size that is governed by different mechanisms in each stage of cellular development. Expansion in meristematic cells is determined by both increases in cytoplasmic and nuclear volume whereas in differentiated tissues it is mainly determined by turgor-driven vacuolar enlargement that allows the accumulation of water and solutes (Wolf et al., 2012; Sablowski and Dornelas, 2014). Turgor driven cell expansion is the result of multiple steps like cell wall relaxation to accommodate water uptake, wall extension by turgor pressure, dehydration/cell wall stiffening and the accumulation of cell wall components (Cosgrove, 2005; Wolf et al., 2012; Figure 9).

In the plant cell wall cellulose microfibrils are associated through hemicellulose tethers to form the cellulose–hemicellulose network, which is embedded in a pectin matrix. Primarily, auxin or brassinolide induce activity of P-type plasma membrane proton ATPase (AHA) that causes acidification of the apoplast and in turn activates hydration and cell wall loosening by EXP proteins and xyloglucanendotransglucosylase/hydrolases (XTHs), xyloglucan endohydrolase (XEH), and xyloglucan endotransglucosylase (XET; Yokoyama and Nishitani, 2001; Rose et al., 2002; Caesar et al., 2011; Wolf et al., 2012). Dyson et al. (2012) developed a model of hemicellulose dynamics in an expanding cell wall, showing how the action of XTH and EXP family enzymes determine yield and extensibility of the wall as encapsulated by the classical Lockhart equation (Lockhart, 1965). The mechanism controlling cell wall swelling/hydration is yet to be clarified. Antisense and sense constructs of the Arabidopsis EXP10 gene produce smaller leaves with altered morphology and larger leaves having bigger cells, respectively (Cho and Cosgrove, 2000). Wall hydration allows cell wall extension through structural alterations. Cell wall relaxation stretches the plasma membrane which promotes opening of Ca²⁺ channels. The resulting increase in cytoplasmic calcium affects growth by inhibiting P-ATPases that cause alkalization of the apoplast and inhibition of EXP activity. It also activates NADPH-oxidase which promotes secretion of superoxide into the cell wall, which is further converted into hydrogen peroxide. These reactive oxygen species promote cross linking of cell wall components, which causes cell wall dehydration and strengthening. At the end, wall thickness is reinstated by biosynthesis of membrane lipids, cell wall components and proteins, and appropriate channelization of these materials to their final cellular destination (Wolf et al., 2012).

Auxin does not always promote cell expansion as its concentration has also been observed to fall during leaf expansion (Braun et al., 2008), suggesting a more complicated dose–response relation. The yucca and sur mutant of Arabidopsis have elevated auxin levels and smaller leaves (Boerjan et al., 1995; Zhao et al., 2001). Mutation in EXIGUA (EXI) genes, which encode for different subunits of cellulose synthase complex required for secondary cell wall biosynthesis, produces small leaves having defects in cell expansion (Rubio-Díaz et al., 2012). Growth anisotropy, the existence of directions with distinct growth properties is determined by the orientation of the stiff cellulose microfibrils, which in turn is controlled by the orientation of cortical microtubule (CMT) arrays guiding cellulose synthase (Paredez et al., 2006). Uyttewaal et al. (2012) showed by experimental and modeling approaches that the microtubule severing protein katanin mediates the response of cells to mechanical stress in the Arabidopsis SAM. The alignment between PIN1 polarity and microtubule orientation in the SAM indicates a tight biophysical coupling between morphogenesis and auxin transport as further corroborated by mathematical modeling (Heisler et al., 2010).

Cell ploidy level is strongly correlated with mature cell size in many plant species (Sugimoto-Shirasu and Roberts, 2003). Alteration in genes specific for G2–M transition affects the onset of endocycle with earlier onset typically leading to enhanced ploidy. A notable exception is down-regulation of Arabidopsis REGULATORY PARTICLE AAA-ATPASE (RPT2a), which encodes a subunit of the 26S proteasome that causes enlarged plant organs having less but bigger cells. DNA content was higher in some organs but not in all which suggests that the increase in organ size was not the result of endoploidization (Kurepa et al., 2009). Another exception is KRP2 overexpression in Arabidopsis, which does not alter timing of cell cycle exit, but induces fewer and enlarged cells in combination with lower endoploidy levels (De Veylder et al., 2001). A recent study of different Arabidopsis mutant and transgenic lines with altered cell sizes showed strong differences in the effect of a same doubling of nuclear ploidy levels, by tetraploidization, on mature cell size (Tsukaya, 2013a). This indicates that genetic factors strongly affect and complicate the general relationship between endoploidy and size, by thus far unknown mechanisms.

In the process of leaf development, cells have the ability to differentiate into distinct cell types such as guard cells, vascular tissue cells, and trichomes, enabling them to perform diverse specialized functions. All these cell types develop from undifferentiated proliferating cells in the young primordium under the control of regulatory pathways that are increasingly being elucidated.