Evolution of leaf developmental mechanisms




  •  Summary 1

  • I. Introduction 1
  • II. Genetics of shoot development in simple-leafed model organisms 4
  • III.  Dissected leaves 11
  • IV.  Conclusions and future directions 13
  •  Acknowledgements 14

  •  References  14


Leaves are determinate organs produced by the shoot apical meristem. Land plants demonstrate a large range of variation in leaf form. Here we discuss evolution of leaf form in the context of our current understanding of leaf development, as this has emerged from molecular genetic studies in model organisms. We also discuss specific examples where parallel studies of development in different species have helped understanding how diversification of leaf form may occur in nature.

I. Introduction

Leaves are determinate organs that serve as the main photosynthetic structures of land plants. Leaves are borne by the stem and in most seed plants axillary buds develop at the stem–leaf junction. Although these attributes of leaves are fairly well conserved across seed plants, there are many cases in which one or more of them is absent. For example, plants belonging to the genera Guarea and Chisocheton (Meliaceae) retain prolonged growth capability as the leaf tip is able to produce new leaflets for more than four years (Steingraeber & Fisher, 1986; Fisher & Rutishauser, 1990) and modified leaves of cacti are not photosynthetic. These particular leaf forms represent only a small range of the striking degree in variation of leaf form that characterises the plant kingdom. Recent research is focused on understanding the molecular genetic basis of leaf development in model organisms such as the strap-shaped grass leaf of maize or the spoon-like leaf of Arabidopsis thaliana. Mechanisms that control generation of species-specific variations in leaf shape such as the ones described above are less understood. However, considerable interest now exists in the area of comparative development and this should help us to understand how evolution and development are intertwined to generate natural variation in leaf form.

Leaves originate from a group of pluripotent cells, termed the shoot apical meristem (SAM). The SAM is an indeterminate structure that resides in the growing tip of plants. Its complexity and architecture vary in different plant taxa; however, in all cases it generates lateral appendages. Therefore, understanding of meristem function is critical for understanding leaf development. Moreover, a wealth of both classic and recent data suggests that meristems and leaves are in constant communication, suggesting the interdependence of the two structures (Sussex, 1954; Sussex, 1955; Snow & Snow, 1959; Hanawa, 1961; Eshed et al., 2001; Reinhardt et al., 2003b).

In this review we briefly consider the various forms of vegetative meristems and leaves as they manifest themselves during land plant evolution. We subsequently discuss molecular genetic studies on leaf and meristem development in model organisms inasmuch as those provide a conceptual framework that can aid and inform evolutionary studies. Finally, we focus on specific examples where parallel genetic studies in different species have helped understand how modifications in developmental pathways defined in model genetic systems may relate to generation of distinct leaf forms. We will, however, not exhaustively discuss the mechanisms of leaf and meristem development as such; for this, the reader is referred to specific reviews on the subject.

A. Meristem evolution

1. One cell for all seasons  A comprehensive phylogeny is required to understand the evolutionary origin of complex traits present in higher plants. Current evidence suggests that Charales, a group of green algae, is sister to land plants (Kenrick & Crane, 1997b; Karol et al., 2001). Thus the common ancestor of land plants and their related algae was probably a relatively complex organism with a branched filamentous body. Moreover, similarities identified between Charales and land plants likely reflect homology rather than evolutionary convergence (Karol et al., 2001).Thus, the study of aquatic organisms like Chara spp. offers an exciting opportunity to unravel the evolution of developmental and genetic mechanisms that were recruited during the colonisation of terrestrial environments. Morphological studies described the body structure of Chara (Pickett-Heaps, 1967) and revealed that growth occurs at the tip, where a single domed cell divides perpendicularly to the longitudinal axes. This division generates a cell at the base of the tip which will divide again to form the nodal and internodal cells (Fig. 1). While the nodal cell expands, leading to the elongation of the main filament, the internodal cell divides further, periclinally generating lateral branches (Pickett-Heaps, 1967) (Fig. 1). Thus, even though simpler, this pattern of growth is reminiscent of the way the shoots of land plants develop.

Figure 1.

Morphology of the growing tip of Chara in longitudinal plan. Internodal (I) cells are surrounded by small bark cells (R cortical cells). Lateral branches (S) always originate at nodes (K). Figure reproduced with permission from Pickett-Heaps (1975).

2. Apical cell in seedless land plants  A single apical cell is also easily identifiable in the growing tip of mosses and ferns (Gifford, 1983) (Fig. 2a,d). In Dryopteris (fern), ablation of this cell by puncturing results in arrest of organ formation as soon as the remaining cells of meristem are used up (Wardlaw, 1949; Lyndon, 1998). Analogous experiments in the fern Osmunda showed that the apical cell can be replaced (Kuehnert & Miksche, 1964) such that lateral organs originate from the newly established apical cell. It is not clear whether the different results obtained in these studies are due to the use of different techniques in different species, but in any event these experiments identify the apical cell as the source of new cells for building the body of ferns (Lyndon, 1998).

Figure 2.

Schematic representation of longitudinal section through the SAM from various taxonomic groups. (a,d) Type 1 SAM (monoplex) of a fern where periclinal division never occurs. (b,c,e,f) Type 2 SAM of a Gymnosperm (b,e) or of Angiosperms (c,f) with a single tunica layer (c) or with two layers of tunica (f). In type 2 meristems, divisions of the initial cells can occur anticlinally to provide length, or periclinally to increase breadth. Figure reproduced with permission from Kwiatkowska (2004). Dashed lines in (a–c) represent division planes; arrows represent the directions in which the progeny of the initial cell is displaced during growth. (e,f) Representation of the pattern of cell walls within the meristem.

3. Shoot apical meristem of Angiosperms  Angiosperm meristems are more complex and can be divided into three distinct functional zones: the central zone (CZ), at the apex of the meristem; the peripheral zone (PZ) surrounding the CZ; and the rib zone (RZ), beneath the CZ. Lateral organs are produced from founder cells recruited from the PZ, whereas stem tissue is derived from cells recruited from the RZ. The central zone is capable of self-renewal, and CZ-derived cells replenish the PZ and RZ, as these zones lose cells that get incorporated into differentiating lateral organs and the stem. Therefore, CZ cells of the Angiosperm SAM meet the general definition of ‘stem cells’.

4. Patterns of cell division identify two main meristem types in land plants  Evolutionary relationships between the different types of meristems in vascular plants are not completely resolved. Functional studies offer some clues on this problem. In tomato, laser ablation of a small part of the SAM that includes the CZ does not result in inhibition of organogenesis (Reinhardt et al., 2003a). Rather, de novo formation of a new CZ occurs and hence organ formation is never compromised. This is reminiscent of the situation in the fern Osmunda and may indicate that the CZ, as a whole, is homologous to the apical cell. In this vein, it is tempting to speculate that functional similarity among the meristematic cell of Chara, the apical meristematic cell in seedless land plants and the CZ in higher plants may reflect evolutionary relatedness.

However, it is also possible that evolution of these different meristem types is polyphyletic and that apparent similarities are a result of convergent evolution. Based on the mode of cell division, Philipson (1990) recognised two types of meristems. In type 1 (refered to as monoplex in Newman, 1965), a single pyramidal apical cell points inwardly and divides following the planes of the inclined walls (Fig. 2a). This is found in bryophytes, ferns, Equisetum (Equisetopsida) and Selaginella (Lycopsida), in which the apical cell never divides periclinally, and hence doesn't directly contribute to the inner tissues of the apex. Type 2 meristems (simplex and duplex in Newman, 1965) are characterised by the presence of one or more layered initials of which the innermost divides periclinally and contributes directly to inner tissues. These are found in seed plants and lycopsids (Fig. 2b,c). It is possible that physical constrains may have a role in generating the different patterns of cell division observed in different meristem types. For example, it has been proposed that periclinal division of the superficial cell layer(s) (sometimes referred to as tunica) in type 2 meristems is inhibited by the pressure created as a result of growth of internal cells (collectively referred to as corpus) (Wegner, 2000).

The relationship between different types of meristem may be resolved by studying the degree of conservation of the genetic networks that control meristem function in these different groups.

B. The origin of leaves

Morphologists have described two types of leaves: microphylls and megaphylls. Microphylls are structures that lack complex venation patterns or parenchymatous tissues and are found in Lycophytes and Equisetum; however, in the latter group this trait is derived. In contrast, megaphylls are much larger leaves with complex venation typical of ferns and higher plants (Euphyllophytes). The evolution of these structures and the relationship between them has been a matter of debate for a long time. Based on cladistic analysis which also included leafless fossils, Kenrick & Crane (1997a) convincingly argue that microphylls and megaphylls evolved independently. They also propose that microphylls originated by sterilisation of the sporangium, a reproductive structure that consists of a mass of spores included in a thin capsule.

Megaphylls, on the other hand, are thought to have evolved from dichotomously branched structures that were characteristic of extinct members of the Euphyllophytina. According to Zimmerman's ‘telome’ theory (Zimmermann, 1938, 1952), megaphyll evolution involved three transformations. Firstly, the three-dimensional branching architecture changed to a planar branching system (planation); secondly, laminar outgrowth originated as a modification of the lateral branch (webbing); and finally, these webbed outgrowths fused to form a proper leaf lamina (fusion).

Cladistic analysis suggests that megaphylls in the Euphyllophytina are likely to have evolved independently in all major clades of this group (ferns, sphenopsids and seed plants) (Kenrick & Crane, 1997a). Therefore planation, webbing and fusion are steps of an evolutionary process that has been recruited multiple times during the evolutionary history of land plants. These modifications lead to a pinnately compound vegetative leaf which is the likely morphological state of basal seed plants (Kenrick & Crane, 1997a).

Environmental conditions may have had an important role in the evolution of land plant leaves as high levels of CO2 in the middle Paleozoic are thought to have delayed the spreading of megaphylls. Leaves of early land plants possessed low stomatal density, possibly owing to the high CO2 content in the atmosphere, and this may have imposed a tight physical limit on evaporative energy loss. It has been calculated that the increase of solar energy interception by leaf lamina expansion under these conditions could have resulted in an increase in leaf temperature, which ultimately would have led to high-temperature damage and collapse of photosynthetic productivity (Beerling et al., 2001). As CO2 levels dropped, however, it would have been possible for leaf enlargement and increase of stomatal density to occur. This was accompanied by decreases in temperature, thus lowering the risks of overheating (Osborne et al., 2004) and likely allowing establishment of megaphylls in multiple plant lineages.

II. Genetics of shoot development in simple-leafed model organisms

Extensive studies of mutants isolated in forward genetic screens has resulted in the identification of a number of key genes that control meristem and leaf development and the conceptualisation of genetic hierarchies in which these genes participate. Most of this work has been done in three model organisms: the eudicot Asterid Arabidopsis thaliana, the eudicot Rosid Antirrhinum majus and the monocot grass maize.

A. Genes involved in formation and maintenance of the meristem

1. The WUSCHEL/CLAVATA feedback loop  Apical meristems initiate new lateral organs throughout the plant's lifetime. This requires that cells recruited into lateral organs are constantly replenished such that the meristem maintains itself. The WUSCHEL (WUS) and CLAVATA (CLV) genes have been identified as key players in this process (Clark et al., 1993, 1996, 1997; Mayer et al., 1998; Fletcher et al., 1999; Brand et al., 2000). Loss-of-function mutations in CLV genes result in larger shoot meristems and increased floral organ number, indicating their role in maintaining meristem size (Clark et al., 1993, 1997). Conversely, mutations in WUS result in failure to maintain the meristem. CLV1 encodes a leucine-rich receptor-like kinase that physically interacts with the leucine-rich protein CLV2. WUS encodes a homeodomain transcription factor, which is expressed in a cell cluster low in the CZ, known as the organising centre. WUS induces stem cell identity in the overlying cells of the CZ, which are characterised by the expression of CLV3. A small secreted CLV3 protein, in turn, is thought to interact with the more widely expressed CLV1/CLV2 receptor complex to limit the area of WUS expression in the organising centre, thereby preventing accumulation of excess stem cells. This negative-feedback loop maintains an equilibrium state, and, consequently, a relatively constant cell number and size of the SAM.

2. Evolutionary conservation of CLV function  The CLAVATA signal transduction pathway appears conserved between monocots and eudicots. The fasciated ear2 (fea2) gene of maize is the likely orthologue of CLV2 and null mutations in this gene lead to larger inflorescence and flower meristems (Taguchi-Shiobara et al., 2001). Similarly, the FLORAL ORGAN NUMBER1 (FON1) gene of rice encodes for a leucine-rich repeat receptor-like kinase similar to CLV1 of Arabidopsis. FON1 is expressed in all aerial meristems but mutations in this gene affect only the size of floral meristems, suggesting that redundant genes may be functional in the SAM and inflorescence meristem (Suzaki et al., 2004). As changes in meristem size also affect phyllotaxy (Jackson & Hake, 1999; Giulini et al., 2004), these studies beg the question of whether modifications in the CLAVATA signal transduction pathway have a role in regulating organ position number or size in nature. For example, Mauseth (2004) showed that meristem size varies from 80 to 1500 µm between various cacti species. It would thus be interesting to know whether small changes in expression patterns of genes such as CLV and WUS in these species suggest a role in species-specific modifications in meristem size.

3. KNOX proteins  CLV and WUS genes are necessary for regulating stem cell number in the meristem, but maintenance of the meristem also requires the activity of class I KNOTTED-like homeobox (KNOX) genes. KNOX genes are members of a plant-specific clade of the Three Amino acid Loop Extension (TALE) superclass of homeobox genes. Knotted1 (kn1) was the first homeobox gene isolated in plants and was identified from gain-of-function mutation in maize that produced ‘knots’ or outgrowths of aberrantly differentiated tissue on the leaf (Vollbrecht et al., 1991). In some species, misexpression of KNOX genes is able to induce ectopic meristem development (Sinha et al., 1993; Lincoln et al., 1994; Schneeberger et al., 1995; Chuck et al., 1996; Sentoku et al., 2000). Loss-of-function mutations in kn1, and in its Arabidopsis homolog SHOOTMERISTEMLESS (STM), result in failure to maintain a SAM (Long et al., 1996; Vollbrecht et al., 2000). Taken together, these data suggest that KNOX genes maintain cells in an indeterminate state. Consistent with this idea, the down-regulation of KNOX expression in a group of founder cells in the meristem is an early marker of leaf initiation.

4. Factors acting downstream of KNOX proteins  To understand the function of KNOX transcription factors, it is essential to identify their target genes. A wealth of evidence suggests that KNOX proteins act at least in part via regulating activity of the growth regulators cytokinin and gibberellin (GA) (Li et al., 1992; Kusaba et al., 1998; Tanaka-Ueguchi et al., 1998; Ori et al., 1999; Hay et al., 2004) (Fig. 3). Experiments conducted in different species have shown that KNOX overexpressing plants display elevated cytokinin levels (Kusaba et al., 1998; Ori et al., 1999; Frugis et al., 2001), suggesting that KNOX proteins may induce cytokinin biosynthesis. In contrast, KNOX proteins repress GA biosynthesis. In tobacco, the KNOX protein NTH15 directly represses transcription of Ntc12, a gene encoding a GA20 oxidase, required for GA biosynthesis (Sakamoto et al., 2001). In Arabidopsis, genetic evidence demonstrates that GA acts antagonistically to KNOX in the meristem (Hay et al., 2002). Because GA promotes some aspects of cell differentiation like transverse division and longitudinal expansion, possibly via changing the orientation of cortical microtubules (Shibaoka, 1994), these observations are consistent with the idea that KNOX proteins act to repress differentiation. This notion is further supported by the recent finding that KNOX proteins may directly repress lignin biosynthetic genes, which can be considered as terminal differentiation genes in plants (Mele et al., 2003).

Figure 3.

Cartoons depicting some of the factors controlling SAM and leaf development in Arabidopsis. Arrows indicate positive regulation; T bars indicate negative regulation. KNOX genes (in white) are expressed throughout the SAM (grey oval) and are down-regulated in the incipient leaf primordium (P0, delimited by dotted line) and in developing leaves. KNOX genes may be down-regulated in P0 in response to an auxin gradient. STM negatively regulates AS1 and AS2 within the SAM, and its down-regulation in leaves allows AS1 and AS2 expression. In turn, AS1 and AS2 act as a heterodimer to negatively regulate BP1, KNAT2 and KNAT6. STM negatively regulates the GA biosynthetic gene GA20ox1, of which expression is consequently restricted to the leaves. KNOX genes may act positively on cytokinin levels, which are hence high within the SAM. PHABULOSA (PHB), PHAVOLUTA (PHV) and REVOLUTA (REV) are expressed in the SAM and throughout incipient leaf primordia, and in the adaxial domain of older leaf primordia. YABBY (YAB) and KANADI (KAN) genes are expressed in the abaxial domain of leaf primordia to promote abaxial fates and organ growth. YAB activity also contributes to exclude KNOX expression from leaves. Auxin is transported (grey arrows) through the outer layer of the shoot apex towards the SAM. Auxin flux is diverted away by the existing primordium (P1) but can reach the organogenic periphery on the left flank of the meristem where its accumulation promotes primordium initiation (P0).

5. How conserved is KNOX function amongst land plants?  KNOX-like genes are present in some algae (Acetabularia), mosses, ferns and Gymnosperms (Sano et al., 2005). This has led to the question of whether these genes are homologues of class I KNOX genes that are required for meristem function in Angiosperms and clearly distinguishable from their class II counterparts that are more broadly expressed and have not yet been ascribed functions. Reconstruction of KNOX gene phylogenies indicates that both mosses and ferns have clearly identifiable class I and class II KNOX genes, showing that the division of KNOX genes in these two classes happened early in land plant evolution or before. Denser sampling in green algae will be required to resolve the precise chronology of KNOX gene evolution.

The presence of class I KNOX orthologues across land plants leaves opens the question of what function they perform in different lineages. So far, hypotheses about this are based on gene expression analysis. Such data indicate that class I KNOX genes are expressed in meristematic tissues of ferns, Gymnosperms and Angiosperms, whereas class II genes appear to be ubiquitously expressed. In addition, class I genes are expressed in compound leaves of ferns in a similar fashion to compound leaves of higher plants (see Section III.A) (Bharathan et al., 1997; Hjortswang et al., 2002; Pham & Sinha, 2003; Sano et al., 2005). These data tentatively support a role for class I genes in meristem development in seedless plants; however, in the absence of functional data this remains speculative. A step towards this direction was reported by Sano et al. (2005), who showed that, in a similar fashion to Angiosperm genes, overexpression of class I, but not class II, KNOX genes isolated from the fern Ceratopteris richardii can induce a lobed leaf phenotype in Arabidopsis. This supports the hypothesis that KNOX gene function is conserved amongst land plants, but only the generation of loss-of-function mutants in relevant species will provide formal proof for this.

B. From a meristem to a leaf

Not only does the meristem have to maintain itself, but it also has to produce lateral organs such as leaves. Anatomical changes in meristem shape, such as swelling of the flank of the SAM, mark emergence of a leaf primordium. These changes are also associated with changes in cell division activity and gene expression.

1. Establishing founder cell identity  Cells from the PZ of the meristem that are recruited into incipient leaf primordia are called founder cells (P0 in Fig. 3). Founder cell number is variable amongst different species (Poethig & Sussex, 1985; Furner & Pumfrey, 1992; Irish & Sussex, 1992; Poethig & Szymkowiak, 1995; Weigel & Jurgens, 2002; Grandjean et al., 2004). An early marker for cell recruitment into a leaf primordium is down-regulation of KNOX expression in the founder cells (Jackson et al., 1994; Smith et al., 1992). KNOX down-regulation in leaves requires activity of ARP (ASYMMETRIC LEAVES1 [AS1]/ROUGHSHEATH2 [RS2]/PHANTASTICA [PHAN]) proteins, which belong to the MYB transcription factor family (Fig. 3). ARP loss-of-function mutations disrupt leaf development in a manner similar to KNOX misexpressing mutants or transgenic lines. In Arabidopsis, AS1 is likely to act as a heterodimer with the Lateral Organ Boundaries (LOB) protein AS2 to repress KNOX expression (Xu et al., 2003). Nevertheless, KNOX repression in P0 is maintained in rs2 (Schneeberger et al., 1998; Timmermans et al., 1999), as1 and as2 mutants (Ori et al., 2000) by an as yet unidentified mechanism.

2. ARP genes: conservation of function during evolution?  As KNOX down-regulation in leaf primordia requires activity of orthologous ARP genes, these appear to perform a conserved function. However, differences in arp loss-of-function phenotypes in different species have led to some uncertainty regarding their precise roles in leaf development. Indeed, Antirrhinum phan mutants were first studied because they display severe adaxial–abaxial (i.e. upper–lower, Fig. 4) polarity defects as they condition leaf abaxialisation, indicating that PHAN acts to promote adaxial identity (Waites & Hudson, 1995). Such phenotypes are not readily observed in rs2 or as1 mutants. However, recent studies demonstrated that as1 and as2 mutants in Ler background do display adaxial–abaxial polarity defects (Sun et al., 2002; Xu et al., 2003) and that AS2 transcripts are detected on the adaxial face of embryonic cotyledons (Iwakawa et al., 2002). These data indicate that AS1 and AS2 may promote adaxial identity, thus to an extent reconciling the as and phan mutant phenotypes.

Figure 4.

The three axes of asymmetry within a leaf. On the proximal–distal axis, the distal end is the leaf tip and the proximal end is attached to the stem. On the adaxial–abaxial axis, the adaxial side is adjacent to, and the abaxial side is distant from, the meristem. The medial–lateral (M–L) axis spans the leaf, from the mid vein to the edge of the blade.

Nevertheless, the degree to which as1 (or phan) phenotypes are attributable to KNOX misexpression is unclear. Indeed, loss of function of both the BREVIPEDICELLUS (BP, originally described as KNOTTED-LIKE in Arabidopsis thaliana 1: KNAT1) and KNAT2 genes that are misexpressed in as1 and as2 mutants fails to suppress leaf phenotypes in those mutants. This indicates that these mutant phenotypes are not a consequence of inappropriate KNOX expression. This, however, may be because KNAT6, a third class I KNOX gene, is also misexpressed in as mutant leaves. Thus, the simultaneous knocking out of BP1, KNAT2 and KNAT6 may be required to understand the contribution of KNOX repression to AS function.

Some insight into this problem came by studying tobacco plants where PHAN activity is knocked down (McHale & Koning, 2004). In these antisense NsPHAN plants, juvenile leaves display broad as1-like leaves, whereas adult ones are needle-like with bladeless petioles (McHale & Koning, 2004), resembling phan mutants of Antirrhinum (Waites & Hudson, 1995). Both juvenile and adult leaves display ectopic expression of the NTH20 KNOX gene. GA application (which is known to suppress KNOX misexpression phenotypes) suppressed juvenile leaf phenotypes, whereas it had no discernible effect on adult phenotype. These results suggest that in juvenile leaves, NsPHAN acts via KNOX repression to regulate leaf development. Adult phenotypes are not modified by GA, suggesting that these occur independently of changes in GA. The authors propose that phenotypes of antisense NsPHAN adult leaves reflect distal displacement of stem-like vascular patterning into leaf petioles. This work strengthens the view that KNOX repression is a developmentally important downstream function of ARP proteins. However, this evidence does not exclude a role for ARP proteins in processes unrelated to KNOX regulation.

The maize RS2 gene is able to fully complement the Arabidopsis as1 mutation (Theodoris et al., 2003), supporting the idea that the RS2/AS1 pathway is functionally conserved between monocots and eudicots. This leaves open the question of why maize rs2 mutants do not show polarity defects. One possible explanation for this is that only some ARP downstream targets (e.g. KNOX) are conserved between monocots and eudicots. An alternative explanation is that differences in loss-of-function phenotypes between maize and eudicots reflect differences in the manner by which these taxa elaborate leaf lamina (Tsiantis et al., 1999b).

Finally, a recent study indicates that the ARP function in down-regulating KNOX genes may have evolved early in vascular plant evolution, or before. An ARP gene from Selaginella (a Lycophyte) is expressed in a mutually exclusive pattern with KNOX in leaves and was able to rescue the Arabidopsis as1 mutation (Harrison et al., 2005). Because leaves originated independently in Lycophytes and Euphyllophytes (see Section I.B), these data also suggest that the KNOX–ARP interaction was recruited at least twice during plant evolution to drive leaf formation.

3. The role of auxin in primordium initiation  A role for the growth regulator auxin in leaf initiation was proposed a long time ago (Snow & Snow, 1937). Recent molecular genetic work supports this classical evidence. Mutation in the polar auxin transport (PAT) efflux carrier PINFORMED1 (PIN1) and treatment of Arabidopsis or tomato with PAT inhibitors led to a failure to initiate lateral organs. Conversely, application of the natural auxin IAA to the flank of the SAM leads to the formation of a new primordium (Reinhardt et al., 2000, 2003b; Vogler & Kuhlemeier, 2003). Furthermore, it has been shown that regulated distribution of auxin is involved in determining the position of leaf inception in addition to promoting organogenesis (Reinhardt et al., 2003b) (Fig. 3).

Interestingly, recent evidence suggests that auxin may be required for the down-regulation of KNOX expression in initiating leaf primordia (Scanlon, 2003) (Fig. 3), thus suggesting hitherto unidentified links between the KNOX developmental pathway and auxin (Zgurski et al., 2005).

Similarly, links between KNOX activity and auxin homeostasis are suggested by observations that KNOX misexpression mutants of maize show defects in PAT (Tsiantis et al., 1999a; Scanlon et al., 2002).

C. Establishing domains within the leaf

Once founder cells have been initiated, further growth involves initiation and elaboration of new axes of asymmetry. Organs emerging from the SAM typically have three axes of asymmetry: proximal–distal (from the base to the tip of the leaf), medial–lateral (from the midrib to the margin) and adaxial–abaxial (from the upper to the lower epidermis) (Fig. 4).

1. Elaboration of proximal–distal asymmetry  The maize leaf has been invaluable in understanding mechanisms controlling establishment and elaboration of the proximal–distal axis of leaves. This is because maize leaves, like those of many grasses, are clearly divided into the proximal sheath and the distal blade. The blade–sheath boundary is defined by a linear fringe of epidermal tissue termed the ligule flanked by two wedges of tissue called auricle (Fig. 5g). This simple structure lends itself to genetic analyses and has allowed isolation of many mutants that disrupt domain specification across the proximal–distal axis.

Figure 5.

Histological zonation of the leaf according to the upper/lower leaf zone hypothesis. (a,b) Scanning electron microscopy images showing the approximate position of the upper (U) and lower (L) domains in early stages of leaf development of Arabidopsis suecica (a) and maize (b). The asterisk marks the meristem dome. (c–e) Final products of activities of the upper and lower leaf zones in adult leaves of Arabidopsis suecica (c), tomato (d) and maize (e). (f) Close-up of the tip of a Sansevieria leaf showing the unifacial tip produced by the upper leaf zone. (g) Close-up of the encircled region of the maize leaf shown in (e). Scale bars: (a,b), 50 µm; (c–f), 1 cm. P, plastochron; St, stipule; Pt, petiole; Ptl, petiolule; Tl, terminal leaflet; Ll, leaflet; A, auricle; Lg, ligule; S, sheath; B, blade; Rc, rachis.

Both liguleless1 (lg1) and lg2 mutants affect the production of auricle or ligule at varying degrees during vegetative development (Becraft et al., 1990; Sylvester et al., 1990; Harper & Freeling, 1996; Walsh et al., 1998), and the double mutant lg1; lg2 shows a novel phenotype as neither ligule nor auricle are formed on any leaf (Harper & Freeling, 1996). Lg1 encodes a nuclear protein belonging to the squamosa-promoter binding proteins class (Moreno et al., 1997) that acts cell autonomously and is expressed in the ligular region of developing leaves. Lg2, on the other hand, encodes a putative transcription factor of the basic leucine zipper class (Walsh et al., 1998) that acts non-cell autonomously and is expressed in the meristem and developing ligule regions (Moreno et al., 1997; Walsh et al., 1998). Because LG1 acts cell autonomously, it may be responsible for the reception of signals directing formation of ligule and auricle, whereas the non-cell autonomously acting LG2 may restrict these signals to a thin line across the leaf (Becraft & Freeling, 1991; Harper & Freeling, 1996).

Liguleless functions are conserved amongst grasses.  In rice, two genes are known to affect ligule development: auricleless (aul) and liguleless (lg). The aul mutant displays an almost identical phenotype to maize lg2. The rice lg, the barley liguleless (lig, formally li) and the maize lg1 loci are collinear, and so they are likely to be orthologous. However, in the lg mutant of rice and li of barley, no ligule is formed on any leaf, in contrast to maize lg1, in which later leaves develop rudimentary ligule. This suggests a separation of functions for LG1 in maize such that a duplicate gene may play a role in ligule formation on upper leaves (Pratchett & Laurie, 1994; Harper & Freeling, 1996). The Arabidopsis genome does not appear to contain orthologues of Lg1, indicating that this function may be monocot or grass specific and raising fascinating questions on when LG1 evolved and what the ancestral function of genes that gave rise to LG1-like sequences was.

Ectopic expression of KNOX genes in leaves of maize affects proximal–distal patterning.  All gain-of-function KNOX misexpressing mutants of maize condition a transformation of blade into sheath. This means that a distal portion of the leaf resembles a more proximal one. One interpretation for this is that KNOX expression in leaves disrupts a predetermined maturation schedule of the leaf, according to which cells progress from founder cells to sheath to auricle/ligule to blade (Freeling, 1992). According to this model, transformations of blade to sheath are an indication that ‘relatively uncommitted cells exist in an otherwise older leaf’ (Freeling & Hake, 1985; Becraft & Freeling, 1994; Fowler & Freeling, 1996; Fowler et al., 1996; Muehlbauer et al., 1997; Muehlbauer et al., 1999). As predicted by this model, KNOX genes are normally expressed in the SAM and are excluded from leaf primordia.

Concrete support for these ideas comes from analysing the effects of KNOX misexpression at different times during leaf development. Muehlbauer et al. (1997) showed that ectopic expression of the KNOX gene Liguless3 early during leaf development led to sheath-like sectors, whereas later missexpression gives rise to ectopic auricle/ligule. However, the precise pathways that mediate effects of KNOX misexpression on the blade–sheath boundary are unknown.

2. Establishment of the medial–lateral domain: analysis of narrow sheath function supports the upper/lower leaf zone histogenesis model 

Medial–lateral asymmetry is evident from early stages of development of maize leaves. This is because incipient primordia surround the meristem with the mid vein being formed on one flank and leaf margins on the opposite flank defining the medial–lateral axis (Fig. 5b). Mutations in the duplicated redundantly acting narrow sheath (ns) genes cause the deletion of a lateral domain in maize leaves that includes leaf margins (Scanlon et al., 1996). Absence of ns activity results in failure to down-regulate KNOX genes in a domain of the founder cells which would normally contribute to the formation of margins. The resulting lack of marginal domains is responsible for the development of a narrow leaf (Scanlon et al., 1996).

Recently, the ns genes have been cloned and shown to be expressed in two foci at the lateral edges of leaf founder cells in the meristem, whereas their expression persists in the margins of lateral organ primordia (Nardmann et al., 2004). Although some questions remain regarding the precise localisation of NS protein in the LI vs L2 domains of the meristem, this information on NS localisation broadly supports analysis data indicating that NS acts to recruit the lateral domain of leaves within the meristem (Scanlon & Freeling, 1997). This suggests that different domains of the meristem may be recruited separately and begs the question of which factors recruit the central domain of the leaf.

Intriguingly, NS is highly similar to the Arabidopsis protein Pressed Flower (PRS) (Matsumoto & Okada, 2001; Nardmann et al., 2004) which encodes a WUSCHEL-like homeodomain protein required for lateral sepal development but not reported to have a role in leaf development. Inspired by their maize findings, Nardmann et al. (2004) were able to uncover a previously undescribed phenotype of the prs mutant, namely the absence of stipules at the base of the leaf. This observation is in good agreement with the leaf zonation and histogenesis model elaborated by comparative morphologists (Troll, 1955; Kaplan, 1973). According to this model, all Angiosperm leaves are divided into an upper and a lower leaf zone (Fig. 5). In eudicots, the upper zone gives rise to the lamina and petiole and the lower zone only yields the leaf base and stipules. In contrast, in many monocots the lower zone gives rise to the sheath and the larger part of the leaf blade, whereas the upper zone is limited to the forerunner tip, a rudimentary structure at the distal end of the leaf (Fig. 5f).

According to this model, mutations in genes like ns that act in the lateral domain of the lower leaf zone would be predicted to affect development of the large portion of the leaf in monocots (i.e. sheath and lower blade), whereas equivalent mutations in dicots should have much less obvious phenotypes that include effects on stipules, as those are the most prominent product on the lower leaf zone in such species. Comparative consideration of the ns and prs phenotype follows these predictions, thus supporting the idea that differential elaboration of the upper and lower leaf zones may underly generation of different leaf forms in maize and Arabidopsis.

3. Elaboration of adaxial–abaxial asymmetry 

Flattening of the leaf occurs early in primordium development (see Fig. 5a,b) This reflects the fact that leaves are initiated from the meristem as polarised structures with a clear adaxial–abaxial axis. This asymmetric development of leaves is functionally important in many species as the adaxial side can be specialised for light capture and the abaxial surface for gas exchange.

A connection between adaxial–abaxial polarity of lateral organs and polarity of the meristem was first suggested 50 years ago by surgical experiments in which the lateral organ primordia were separated from the apical meristem by incision (Sussex, 1954, 1955; Snow & Snow, 1959; Hanawa, 1961). Where separation precedes primordium formation, the isolated primordia develop into radially abaxialised organs. This indicates that the meristem could be the source of a signal required to promote adaxial fate and that, in absence of this signal, an abaxial fate occurs by default. Furthermore, these experiments suggest that establishment of adaxial–abaxial polarity is required for proper growth of the lamina.

The first extensively studied mutant with defects in adaxial–abaxial polarity was phantastica (phan) of Antirrhinum (Waites & Hudson, 1995; Waites et al., 1998), discussed in Section II.B in the context of regulation of KNOX expression. phan mutants develop leaves with variable loss of adaxial–abaxial asymmetry. Severely affected leaves are radial and completely abaxialised. These observations suggest a model whereby lamina outgrowth requires the juxtaposition of adaxial and abaxial domains. PHAN encodes a MYB transcription factor, which is expressed at the future site of leaf initiation and in leaf primordia. Throughout this period, PHAN expression is uniform along the adaxial–abaxial axis, indicating that PHAN does not itself provide adaxial–abaxial information. Rather, PHAN may interact with other proteins (such as the AS2 orthologue of snapdragon) that have spatially restricted expression pattern. The precise relationship of KNOX misexpression and polarity defects in phan mutants remains unclear.

Conversely, Phabulosa (Phb-1d), Phavoluta (Phv) and Revoluta (Rev) gain-of-function mutations in Arabidopsis lead to an adaxialisation of lateral organs (McConnell & Barton, 1998; McConnell et al., 2001; Emery et al., 2003). These genes encode homeodomain-leucine zipper-containing proteins (HD-ZIPIII). Analysis of loss of function in these genes indicates that they promote adaxial leaf identity and are required for meristem maintenance (Emery et al., 2003). In agreement with this, HD-ZIPIII genes are expressed in the SAM and throughout incipient leaf primordia, but later their expression becomes polarised to the adaxial side (McConnell et al., 2001; Otsuga et al., 2001; Emery et al., 2003; Juarez et al., 2004a,b) (Fig. 3). Recently, it has been shown that the restriction of HD-ZIPIII expression to the adaxial domain involves at least in part microRNA activity (Reinhart et al., 2002; Rhoades et al., 2002; Emery et al., 2003; Tang et al., 2003; Juarez et al., 2004a). Recent evidence suggests that miRNA-dependent PHB regulation is mediated by DNA methylation of the template chromosome (Bao et al., 2004).

Conservation of HD-ZIPIII function and regulation.  Analysis of the Rolled leaf mutants of maize shows that the role of HD-ZIPIIIs in adaxial specification is conserved between monocots and eudicots. Juarez et al. (2004a) demonstrate that dominant mutations in the miRNA complementary site of the maize orthologue of the revoluta gene, Rld1, also perturb organ polarity and condition misexpression of the Rld1 transcript. Nelson et al. (2002), who originally characterized the Rld phenotype, showed that Rld is not only a likely adaxialising factor but that disruption of Rld regulation perturbs signalling between the adaxial and abaxial domains of the maize leaf, thus highlighting the importance of cross-domain signalling in polar identity specification.

Floyd & Bowman (2004) showed that this miRNA-based regulation of HD-ZIPIII genes is conserved in all lineages of land plants, including bryophytes, lycopods, ferns and seed plants. This suggests that miRNA restriction of HD-ZIPIII expression may be a very ancient function recruited independently in different body parts to generate organ polarity (see Section II.C.3).

Specification of abaxial fate.  Down-regulation of the HD-ZIPIII genes is likely to allow expression of the KANADI (KAN) and YABBY (YAB) genes, which are redundantly required to establish abaxial identity (Sawa et al., 1999; Siegfried et al., 1999; Eshed et al., 2001; Kerstetter et al., 2001). KAN and YAB genes encode GARP and HMG transcription factors, respectively. Both YABBYs and KANADIs belong to small gene families, and KAN proteins are likely to repress adaxial–promoting HD-ZIPIIIs that repress them (Eshed et al., 2001, 2004). This mutually antagonistic relationship between KAN and HD-ZIPIII appears to drive lamina growth that is dependent on activity of the YABBY proteins. Importantly, ectopic expression of abaxial specifying factors, particularly KANADIs, has detrimental effects to meristem activity, suggesting that abaxial identity and meristem function are incompatible. This is further highlighted by the finding that YABBY activity contributes to exclusion of meristem-promoting KNOX expression from leaves (Kumaran et al., 2002) (Fig. 3).

Conservation and divergence of YABBY gene activity.   YABBY genes have also been analysed in other species, and this provides novel insights in the organisation of pathways that specify abaxial fate. In Antirrhinum, the YABBY gene GRAMINIFOLIA (GRAM) acts non-autonomously alongside its paralogue, PROLONGATA (PROL), to promote lamina growth via abaxial cell proliferation (Golz et al., 2004; Navarro et al., 2004). Abaxial-promoting GRAM activity is dispensable in the absence of adaxial fate specification, indicating that GRAM acts in the abaxial side of the leaf to repress adaxial identity, perhaps via repressing an HD-ZIPIII activity. GRAM and PROL also act together in a distinct developmental pathway that promotes adaxial cell fate. The multifaceted activity of GRAM highlights the complexity of the signalling pathways that operate to specify developmental identities in higher plants and further underlines the importance of signalling between the abaxial and adaxial sides of the leaf.

Analysis of YABBY genes in monocots had led to interesting surprises. In maize, two YABBY genes are expressed on the adaxial side of the leaf, contrary to the situation in Arabidopsis (Juarez et al., 2004b), leading the authors to suggest that, in maize, YABBY genes may direct lateral outgrowth rather than determine cell fate. Consistent with this, mutations in the rice YABBY gene DROOPING LEAF result in defects in midrib formation possibly due to reduced cell division (Yamaguchi et al., 2004).

The HD-ZIPIII/KANADI regulatory system also controls vascular polarity and may be ancient.  Interestingly, the polar differentiation system defined by antagonistically acting HD-ZIPIII and KANADI proteins does not only operate in leaves but also in vascular tissue. Gain-of-function mutations in HD-ZIPIII gene family members condition formation of radialised vascular bundles such that the xylem that normally develops internally (adaxially) in the bundle now surrounds the phloem, thus resulting in adaxialised bundles (McConnell et al., 2001). Conversely, plants mutant for three HD-ZIPIIIs (revoluta, phabulosa and phavoluta) develop vascular bundles where phloem surrounds the xylem and thus can be considered abaxialised (Emery et al., 2003). Additionally, plants that lack activity of three KANADI gene family members phenocopy the vasculature of HD-ZIPIII gain-of-function mutants by developing adaxialised bundles, thus indicating that KAN activity is required for correct elaboration of abaxial vascular identity.

Because vasculature evolved before leaves in land plants, Emery et al. (2003) have proposed that the HD-ZIPIII/KANADI regulatory system may have been independently recruited in leaves and vascular tissue to facilitate polar differentiation. This mirrors the already discussed relationship between HD-ZIPIII proteins and miRNAs (see Section II.C.3) and therefore raises the question of whether KANADIs and miRNAs act in the same pathway to repress HD-ZIPIIIs.

These results indicate that whereas HD-ZIPIII expression and function seem to be conserved between monocots and eudicots, YABBY expression patterns, and perhaps function, may have diverged between these lineages. Thus KAN and HD-ZIPIII proteins are likely to be intimately related to axial pattering of the leaf, whereas YABBY proteins may have primarily growth-promoting function. Comparative analysis of YABBY expression patterns and function in diverse species may help to test this.

D. Growth of the primordium by cell division

The precise role of cell division in regulating size and shape of leaf primordia is a matter of some controversy. This is because in some cases alteration of cell division is not correlated with change in leaf size or shape. For example, expression of a dominant-negative Arabidopsis CDKA in transgenic tobacco results in almost normal leaves (Hemerly et al., 1995). Nevertheless, in other cases, modulation of cell cycle gene activity leads to a disturbed leaf organogenesis, as illustrated by overexpression of CyclinD3 (Riou-Khamlichi et al., 1999; Dewitte et al., 2003) or CDK inhibitors (Wang et al., 2000; De Veylder et al., 2001; Jasinski et al., 2002). These observations suggest that cell cycle regulation and leaf developmental programs are intertwined. One factor that may link developmental patterning with cell division is the APETALA2-domain transcription factor AINTEGUMENTA (ANT). ANT regulates organ size by maintaining the meristematic competence of cells during organogenesis, in part by promoting the expression of CyclinD3 (Mizukami & Fischer, 2000). ANT itself may be regulated by the auxin inducible protein ARGOS (Hu et al., 2003). Therefore, ARGOS may transduce auxin signals to regulate organ growth via ANT. Additionally, a symmetrical distribution of auxin may be required for symmetrical leaf growth by inducing symmetrical cell division patterns (Zgurski et al., 2005).

Precise regulation of growth distribution in a leaf has a key role in controlling final leaf shape and size. Classical studies of leaf development in different species have shown that a front of cell cycle arrest moves gradually from the tip to the base of the leaf (Avery, 1933; Sylvester et al., 1990; Tsuge et al., 1996; Donnelly et al., 1999). Nath et al. (2003) showed that in the cincinnata (cin) mutant of Antirrhinum the progression of this front is delayed and has a strongly concave instead of a weakly convex shape, leading to a longer period of growth of the marginal regions compared to the medial regions. Thus cin plants display excess growth in the leaf margin, leading to leaf curvature (Nath et al., 2003). CIN encodes a TCP (TEOSINTE BRANCHED1 [TB1]/CYCLOIDEA [CYC]/PROLIFERATING CELL FACTOR [PCF]) transcription factor that is expressed in the actively dividing region of the lamina, possibly overlapping with the cell division arrest front during leaf growth. The authors suggest that CIN might make cells more sensitive to the arrest signal through regulation of cell cycle gene expression.

The Arabidopsis dominant mutant jaw-D displays a phenotype similar to cin caused by misexpression of a miRNA complementary to several Arabidopsis TCP-like genes. This results in decreased transcript levels of these genes and hence cin-like phenotypes (Palatnik et al., 2003). TCP genes with miRNA target sequences are found in a wide range of species, suggesting that miRNA-mediated control of leaf morphogenesis is conserved between plants with very different leaf forms.

The transcription factor JAGGED (JAG) is another important regulator of lateral organ growth (Dinneny et al., 2004; Ohno et al., 2004). Mutations in JAG condition serrated organs and affect most severely the distal regions of organs, resulting in jagged edges. JAG is expressed in the growing regions of lateral organs. Dinneny et al. (2004) proposed that JAG function is to slow the arrest of cell division in the distal region of organs. Interestingly, CIN and JAG seem to have opposite functions toward the cell cycle, suggesting that accurate exit from the cell cycle is critical for achieving final leaf shape and size. Investigation of the signalling pathways via which these transcription factors control cell cycle genes will be an important aspect of future research on leaf development.

III. Dissected leaves

So far, we have discussed leaf development in species with simple leaves where the leaf blade is entire. However, a very common form of leaf shape in nature is the dissected (or compound) form. In dissected leaves, the leaf blade is divided into distinct subunits, called leaflets. Leaflets emanate from the leaf rachis and they may or may not be borne by a short structure termed petiolule (Fig. 5d). The two main variants of dissected leaf morphology in Angiosperms are pinnate and palmate. In pinnately compound leaves, the leaflets are arranged along the rachis, which is elongated between the leaflets (Fig. 5d). Palmate leaves have no rachis elongation between the leaflets; therefore, the leaflets all converge at a common point at the distal end of the petiole. The manner in which leaflets are arranged on the rachis and differences in leaflet shape result in the variety of compound leaf architectures seen in nature.

Importantly, leaves are initiated from the meristem as entire structures in dissected leaf species, as their simple counterparts. The leaf primordium subsequently gives rise to leaflets via mechanisms which are not clearly understood. The main systems where the genetics of leaf dissection has been studied are tomato and pea, where a range of mutations affecting leaf patterning and the degree of leaf dissection have been isolated.

A. KNOX function and leaf dissection

The pinnately compound leaf of tomato (Fig. 5d) is characterised by a terminal leaflet and three to four pairs of lateral leaflets that are produced in a basipetal (from tip to base) sequence. The leaflets are lobed and lobes themselves develop largely acropetally (from base to tip) on each leaflet. Many studies have suggested functional links between KNOX expression in leaves and leaf dissection in tomato. In a pioneering study, Hareven et al. (1996) demonstrated that tomato leaves express KNOX genes, unlike most of their simple counterparts, and that KNOX overexpression in tomato results in super-dissected leaves. These data indicated that final leaf morphology is sensitive to the level of KNOX expression within the tomato leaf. Further support for this idea came from studying Mouse ears (Me) and Curl (Cu), two dominant mutations that cause inappropriate expression of the tomato STM orthologue Tkn2 (Chen et al., 1997; Parnis et al., 1997). Both Me and Cu show increased leaf dissection; however, leaf form is dramatically different, with Cu leaves also being massively compressed. These differences are likely to result from distinct modes of Tkn2 misexpression associated with the two alleles, thus reinforcing the idea that compound leaf form is very sensitive to perturbations of both the level and pattern of KNOX expression.

This raises the question of whether tomato leaves possess regulatory hierarchies that precisely define KNOX expression. Multiple recessive alleles at the CLAUSA locus condition increased leaflet number and elevated KNOX expression (Avivi et al., 2000), indicating that CLAUSA is required to define the correct KNOX expression pattern and hence the level of leaf dissection. Cloning of the CLAUSA gene and other genes in which mutations render similar phenotypes (http://zamir.sgn.cornell.edu/mutants/) will yield novel insights into how tomato leaf shape and KNOX expression are controlled.

The possible role of KNOX expression in leaves and leaf dissection across Angiosperms was investigated by Bharathan et al. (2002) using an antibody raised against the maize KN1 protein. The authors report that KNOX localisation in leaf primordia correlates well with dissected morphology early in development. Interestingly, however, it does not always correlate with final leaf morphology, such that certain species express KNOX genes in young leaf primordia that are dissected and yet present a simple final leaf form. This is likely to reflect secondary morphogenesis that in some cases can modify the dissected form presented by young leaf primordia. Mutational analyses in species that express KNOX genes in leaves will be required to determine the precise role of KNOX expression in controlling leaf shape in such taxa.

B. The KNOX/GA regulatory module and leaf dissection

As reduction in GA biosynthesis has been shown to partially mediate KNOX activity in the meristem of plants with simple leaves (see section II.A.4), the question arises as to whether it also does so in the compound leaves of tomato. This appears to be the case, as KNOX-mediated increases in leaf dissection are accompanied by repression of expression of the tomato GA biosynthetic gene Le20ox1 (Hay et al., 2002). Additionally, in tomato, both GA application and constitutive GA signalling result in decreased leaf dissection and antagonise KNOX-mediated increased leaf dissection (Hay et al., 2002). Thus the KNOX/GA regulatory module has a role in regulating the level of leaf dissection in tomato and perhaps other compound-leafed plants.

C. The role of ARP genes in control of dissected leaf form

As ARP proteins regulate KNOX expression across Angiosperms (see Section II.B.1,2), they are strong candidates for being regulators of KNOX expression and leaf complexity in plants with dissected leaves. Antisense reduction of LePHAN expression results in aberrant leaflet placement, suggesting that PHAN is required for correct elaboration of dissected leaf form. However, the precise manner by which this happens and the exact role of KNOX misexpression in elaborating the LePHAN antisense phenotype is unclear (Kim et al., 2003a; Kim et al., 2003b). Antisense LePHAN tomato plants demonstrate variable reduction in the size of the adaxial domain. When expression of LePHAN is detected along the whole adaxial face of leaf primordia, a pinnate leaf form develops, whereas confinement of LePHAN expression to the distal region of the leaf primordia results in peltately palmate compound leaves (Kim et al., 2003a). This correlation appears to hold when comparing expression patterns of PHAN in pinnately vs palmately compound leaf species. The authors suggest that the role of PHAN in controlling compound leaf form reflects its role in specifying the adaxial domain, and that the boundary between adaxial–abaxial domains is required not only for blade outgrowth but also for leaflet formation (Kim et al., 2003a; Champagne & Sinha, 2004). Interestingly, in tomato and many other dissected leaf species studied by Kim et al., PHAN appears to be coexpressed with KNOX genes in the meristem. The functional significance of this is unclear given that PHAN is a negative regulator of KNOX genes, and it is interesting that Pien et al. (2001) reported that PHAN expression in tomato is the same as in simple-leafed species, i.e. excluded from the meristem. The reasons for these discordant results are unclear and may indicate that PHAN expression in tomato is dynamic during development.

D. Other genes that control leaf dissection

The entire (e), trifoliate (tf), solanifolia (sf), potato leaf (c) and Lanceolate (La) mutations simplify the tomato leaf and hence define loci likely to regulate leaf dissection. The first four are recessive and, as such, define genes whose activity is required for dissected leaf form; La is dominant and therefore it is unclear how exactly the LA gene product may control leaf development. However, homozygote La plants show meristem defects, indicating that common La-sensitive activities may drive leaflet production at the rachis and leaf production at the meristem. Which activities those are is unknown; however, as La has the opposite effects on leaf dissection to KNOX overexpression (Avasarala et al., 1996; Hareven et al., 1996), it is tempting to speculate that the mutant La allele may antagonise KNOX function. Interestingly, in La, e and tf mutants, KN1 overexpression does not restore a wild-type compound architecture; rather, the basic architecture seen with these mutations is reiterated (Hareven et al., 1996; Parnis et al., 1997). Thus simple-leafed Lanceolate mutant becomes lobed in a manner similar to Arabidopsis plants that misexpress KNOX genes. This suggests that KNOX activity can only ramify an already established plan, and that it is not a simple loss of KN1-like function that causes reduced dissection in these mutants. Cloning of these genes will be important for understanding how leaf dissection is regulated.

E. KNOX-independent control of leaf dissection in pea

The relationship between KNOX expression in leaves and the generation of the dissected leaf form may not be a universal mechanism for compound leaf formation. In fact, in pea, which has compound leaves, KNOX expression is excluded from leaves (Hofer et al., 2001), revealing the existence of a KNOX-independent pathway for leaf dissection.

In contrast to tomato, leaflet primordia are produced in an acropetal sequence on the compound leaf primordium. Study of mutants with altered leaf complexity allowed the identification of genes involved in pea leaf development. For example, in unifoliata (uni) loss-of-function mutants, compound leaves are converted to simple (Hofer et al., 1997). UNI encodes a transcription factor and is the pea orthologue of Arabidopsis LEAFY (LFY) (Weigel et al., 1992) and Antirrhinum FLORICULA (FLO) (Coen et al., 1990) genes, which control floral development. UNI function may be required for leaves to maintain a transient phase of indeterminacy (i.e. ability to generate novel structures) that leads to leaflet initiation. A similar function is performed by STAMINA PISTILLOIDA (STP) gene, the pea orthologue of Arabidopsis UNUSUAL FLORAL ORGANS (UFO) and Antirrhinum FIMBRIATA (FIM) genes (Taylor et al., 2001). UNI and STP act synergistically, to promote leaflet formation consistent with the interaction observed between LFY and UFO in Arabidopsis, and between FLO and FIM in Antirrhinum (Ingram et al., 1995; Lee et al., 1997). Interestingly, the ectopic expression of UFO in Arabidopsis results in lobed leaves reminiscent of those observed with ectopic KNOX expression (Lee et al., 1997). Moreover, lobbing in 35S::UFO plants is dependent on the presence of LFY, highlighting the potential of the LFY/UFO functional doublet for generating leaf dissection.

Thus increased or prolonged state of indeterminacy is required for specifying dissected leaf form, but the molecular determinants of this prolonged indeterminacy may differ in different species. So far, two independent pathways have been associated with leaf compoundness, either the re-establishment of KNOX expression in leaves or the involvement of the UNI/STP genes. Interestingly, loss of function of the tomato FLO/LFY orthologue FALSIFLORA (FA) results in a slightly reduced number of intercalary leaflets (Molinero-Rosales et al., 1999), suggesting a role of FA in generating dissected leaf form. This result and the analysis of expression pattern of FLO/LFY orthologues in different compound species (Busch & Gleissberg, 2003) have suggested that the FLO/LFY pathway may function in compound leaf development in species other than pea (Champagne & Sinha, 2004). One possibility is that compound leaf development may in principle be regulated by a combination of KNOX and LFY/UFO genes. In pea, the role of KNOX genes may have been taken over by the UNI/STP doublet. Interestingly, these two pathways could be interrelated. For example, in Arabidopsis, UFO expression requires STM activity (Long & Barton, 1998), suggesting that UFO may mediate aspects of KNOX function required for leaf dissection.

KNOX expression is excluded from the pea leaf; however, dissected leaf form in this species is very sensitive to perturbations of KNOX expression. This was demonstrated by Tattersall et al. (2005), who showed that inappropriate KNOX expression resulting from mutations in the pea ARP gene CRISPA is associated with severe proximal–distal transformations including formation of ectopic stipules. Unlike tomato and other dissected leaf species (Kim et al., 2003a), CRISPA and the pea STM orthologue PSKn1 are expressed in mutually exclusive domains in the shoot apex (Tattersall et al., 2005), suggesting that PHAN expression in the meristem is not an inherent component of dissected leaf development programmes.

F. The concept of the blastozone and organogenic activity in leaves

In a similar fashion to compound leaves, simple leaves of some species can develop lobes, but the lamina remains a single unit because the sinuses between lobes never reach the mid vein. Nevertheless, in both lobed and dissected leaves, new axes of lateral growth can be identified. After an initial phase in which the primordium starts to develop, a second phase follows (organogenic phase) during which leaflets or lobes are initiated. This organogenic phase is absent in most leaves with entire margins.

Parts of the leaf margins that are competent at establishing new axes of growth and hence at producing lobes or leaflets are sometimes referred to as ‘lateral blastozones’ (Hagemann & Gleissberg, 1996). This term is preferable to the alternative ‘marginal meristems’ because the later is more appropriately used to denote areas with stem-cell-like properties that are capable of histogenic activity, such as the SAM and root apical meristem. Nevertheless the concept of blastozones denotes that only a portion of the total population of proliferating cells is destined to give leaflets; however, the presence of ‘marginal meristems’ that were thought to give rise to the leaf blade in simple-leafed species has been ruled out by clonal analysis experiments (Poethig & Szymkowiak, 1995). It will therefore be interesting to see whether further work supports the existence of marginal blastozones as discernible groups of cells.

IV. Conclusions and Future directions

Parallel genetic studies of leaf development in multiple species have greatly enriched our knowledge of how leaf form is controlled and have given us glimpses on how leaf developmental pathways may have evolved during land plant evolution.

In the future, it will be important to extend these studies further, and to gain a full picture of the evolution of genetic hierarchies that control leaf development in multiple species with divergent morphologies. This will require research on species that not only have specific developmental attributes but also are amenable to genetic experimentation. Work on such species should also allow detailed functional studies such as cross-complementation studies, or promoter swaps, to understand the evolution of cis-regulatory elements that may be responsible for taxon-specific expression of developmentally important genes.

Additionally, it will be important to identify loci that are directly responsible for species-specific morphological variation. This can be done using quantitative trait loci (QTL) analysis and hence is only feasible in closely related interfertile species. The power of this approach has been demonstrated by the isolation of genes controlling evolution of apical dominance in maize and fruit size in tomato (Doebley et al., 1997; Frary et al., 2000). This approach has already been initiated to study variation between Lycopersicon esculentum and Lycopersicon pennelli and led to the identification of QTLs that contribute to leaf shape and size (Holtan & Hake, 2003; Frary et al., 2004). Refining these QTLs and eventually cloning the relevant genes will be an important step in understanding how natural variation in leaf form is controlled. It will also be interesting to link this approach with ecological studies to determine whether evolutionary changes in leaf shape have adaptive value.

Finally, it will be critical to interpret comparative development data in a firm phylogenetic context if meaningful conclusions about evolution of form are to be reached. For this, it is essential that active research in taxonomy and evolution continues and that substantial effort is put in the challenging area of reconstructing species level phylogenies. For example, leaf dissection appears to have been gained and lost multiple times during Angiosperm evolution (Bharathan et al., 2002). Therefore the precise genealogy of regulatory events that condition morphological shifts in leaf dissection can only be done with confidence in groups where these shifts are clearly understood. Similar considerations apply for studying any other character associated with changes in leaf form during land plant evolution.


We would like to thank Angela Hay for providing the maize SEM, John Baker for photography and reviewers for helpful comments. We apologise to authors whose papers we could not cite because of space constraints. P.P. is a trainee on the European Community's Human Potential Programme HPRN-CT-2002–00267, [DAGOLIGN]. M.T. receives funding from the BBSRC, the Gatsby Foundation and the EU.