Epidermis: the formation and functions of a fundamental plant tissue

Differentiation of a protodermal cell layer in the embryo According to established views of angiosperm embryogenesis, the differentiation of the protoderm is one of the three major events occurring during embryo development (Kaplan & Cooke, 1997). This crucial step takes place after the acquisition of apico-basal polarity and before the formation of meristems at opposite ends of the embryo (Fig. 1). In A. thaliana and maize, apico-basal polarity is manifest as a highly asymmetric distribution of cytoplasmic contents in the egg/zygote, and is subsequently fixed by the asymmetric division of the zygote into a highly cytoplasmic apical cell and a vacuolated basal cell, giving rise to the embryo proper and the suspensor, respectively (Goldberg et al., 1994). In the A. thaliana embryo proper, an outermost cell layer becomes demarcated after four rounds of division, at the dermatogen stage (Fig. 1). In the maize embryo, where the first divisions are less synchronized and reproducible, the differentiation of the protoderm becomes apparent at the transition stage c. 6 days after pollination (DAP; Fig. 1). In both species, the cytological differentiation of the protodermal cells, characterized by a more regular and rectangular cell shape and thicker cell walls, becomes more and more evident during subsequent protodermal cell divisions, which are almost exclusively anticlinal.

In A. thaliana, cytological differentiation is preceded by the expression of the two functionally redundant homeo domain leucine zipper class IV (HD-ZIP IV) genes A. thaliana MERISTEM LAYER1 (AtML1) and PROTODERMAL FACTOR2 (PDF2), which are expressed in all cells of the pro-embryo from the one-cell stage to the 16-cell stage, when their expression becomes progressively restricted to the outer cell layer (Lu et al., 1996; Abe et al., 2003; Tanaka et al., 2007). In maize, the HD-ZIP IV gene OUTER CELL LAYER1 (OCL1) shows a comparable expression pattern with a uniform expression in the developing embryo at 3 DAP and a gradient-like expression at 4 DAP which is stronger in the outer cells of the embryo proper than in the inner cells (Ingram et al., 1999). The complete restriction of OCL1 expression to the outermost layer occurs c. 5 DAP and slightly precedes the cytological differentiation of the protoderm in the maize embryo (Ingram et al., 1999). In view of their expression pattern, it has been proposed that AtML1 and PDF2 in A. thaliana as well as OCL1 in maize could regulate the molecular pathway required for the differentiation of the protodermal cell layer in the embryo (Ingram et al., 1999; Abe et al., 2003). While no knockout mutant for OCL1 has been isolated so far (Khaled et al., 2005), the double mutant atml1/pdf2 never forms an organized protodermal layer in the apical part of the proembryo (Abe et al., 2003). These functional data reinforce the expression data suggesting that AtML1 and PDF2 play redundant roles in the differentiation of protodermal cell fate in the apical region of the embryo (Abe et al., 2003).

During A. thaliana protoderm differentiation, A. thaliana DEFECTIVE KERNEL1 (AtDEK1) appears to act upstream of AtML1 and PDF2 (Johnson et al., 2005). Embryos carrying a mutation in AtDEK1 establish an apico-basal polarity but the division patterns of both the suspensor and the embryo proper are severely altered and mutant embryos abort early during development (Johnson et al., 2005; Lid et al., 2005). In fact, no distinct outermost cell layer is formed in dek1 embryos and the protoderm-specific expression of AtML1 is abolished (Johnson et al., 2005). AtDEK1 encodes an integral membrane protein with a cytoplasmic calpain cysteine proteinase domain at the C-terminus (Lid et al., 2002), which may perceive a signal produced outside the cell and transduce it to the cytoplasm through the autolytic cleavage of the calpain domain and its release from the membrane (Johnson et al., 2008). AtDEK1 expression is not restricted to the outermost cell layer but rather AtDEK1 transcripts are found evenly expressed in the embryo during early embryogenesis (Johnson et al., 2005; Lid et al., 2005). An attractive hypothesis is that a positioning signal from the outside of the embryo, perceived by AtDEK1, might be required for the differentiation of the protodermal layer (Fig. 2).

Inspired by findings in Fucus where the cell wall is required for the maintenance of pre-existing polarity in isolated embryo cells (Bouget et al., 1998), some authors have proposed that the cell wall of the A. thaliana zygote could provide positional cues for protodermal cell fate because only the outermost cells of the embryo retained cell wall fragments derived from the zygote (Laux et al., 2004). In citrus, Bruck & Walker (1985) observed the existence of an electron-dense cuticularized layer secreted on the outside of the zygote. The same authors used wounding experiments to show that, at later developmental stages, plant cells appear to be incapable of specifying epidermal cell fate de novo (Bruck & Walker, 1985). As cuticle deposition is a strictly epidermal trait, they proposed that epidermal identity is acquired only once, in early embryogenesis, and that thereafter protodermal identity is maintained only in outermost cells by signalling processes which may require the presence of an intact outer epidermal cell wall and/or cuticle.

Role of signalling proteins in the differentiation of the aleurone layer The second example for the specification of an epidermal layer during seed development is the differentiation of a specialized outer cell layer in the endosperm, known as the aleurone layer in cereals (Fig. 1). Because defects in the layer are easily observable at the cytological level and readily identified at the surface of the maize kernel using colour markers, maize has been a model of choice for studying the genetics of aleurone differentiation (Becraft et al., 1996). The outer cell layer of the endosperm is the first to cellularize in the coenocytic endosperm formed after fertilization (Dumas & Rogowsky, 2008). Unlike the underlying cells of the central endosperm, the outer cells form pre-prophase bands during mitotic divisions, which are essentially anticlinal (Brown et al., 1994). Consequently, the cells at the periphery must be able to interpret their position in order to adopt aleurone cell fate.

There has been some debate as to whether aleurone cell fate is determined by cell lineage or by positional cues. The first concept implies that the cell fate of daughter cells is inherited from the mother cell and that aleurone cell fate is restricted to the outer layer by the absence of periclinal cell divisions. The observed rarity of periclinal cell divisions in the aleurone layer has historically caused several authors to favour this hypothesis (Randolph, 1936; Kiesselbach, 1949; Walbot, 1994). A second hypothesis proposes that cell fate is not inherited but established for every cell based on the reception of signals or signal gradients. According to this hypothesis, aleurone cells receive qualitatively or quantitatively different signals to underlying cells, and differentiate accordingly. This model is supported by the observation of occasional periclinal cell divisions in the aleurone layer of wheat (Morrison et al., 1975), and by the analysis of genetically marked sectors in maize which clearly demonstrates that the aleurone layer contributes cells to the starchy endosperm (Becraft & Asuncion-Crabb, 2000). Today the preponderant role of positional cues rather than cell lineage is widely accepted.

What are the source and nature of positional cues required for aleurone cell fate determination? Earlier studies investigated whether the maternal seed coat could provide at least in part the information necessary for the differentiation of the aleurone layer (Olsen et al., 1998). However, more recent experiments based on in vitro culture of maize endosperm expressing a fluorescent aleurone marker clearly argue against this hypothesis (Gruis et al., 2006). The observation that cavities which form stochastically deep inside the starchy endosperm during culture were lined by a single layer of fluorescent cells indicated that the absence of neighbouring cells on one side of a starchy endosperm cell is sufficient to provide a signal triggering the acquisition of aleurone cell fate.

Consistent with the concept of positional signalling, the proteins identified so far as being involved in aleurone cell specification have putative or demonstrated functions in cell-to-cell signalling. Defects in DEK1, the maize orthologue of AtDEK1, a membrane-bound cysteine protease of the calpain superfamily (Lid et al., 2002), or in CRINKLY4 (CR4), a serine/threonine receptor-like kinase (RLK) with an extracellular ligand-binding domain (Becraft et al., 1996), prevent the differentiation of an aleurone layer and lead to the presence of starchy endosperm cells in peripheral positions (Becraft et al., 1996; Becraft & Asuncion-Crabb, 2000). Phenotypic analysis of the dek1/cr4 double mutant suggests that the two gene products function in partially overlapping pathways, as strong dek1 alleles were epistatic to cr4 (Becraft et al., 2002). Induction of dek1 mutant sectors even late in development caused aleurone cells to loose their identity and to trans-differentiate into starchy cells, while the reversion of an unstable dek1 allele allowed starchy endosperm cells in the peripheral layer to trans-differentiate into aleurone cells, showing that the identity of the outermost endosperm cells remains plastic until late in development (Becraft & Asuncion-Crabb, 2000). Thus, peripheral endosperm cells constantly monitor their position, and positional cues are required throughout endosperm development to maintain aleurone cell fate.

In contrast to DEK1 and CR4, SUPERNUMERARY ALEURONE LAYERS1 (SAL1) appears to be a negative regulator of aleurone cell fate in maize, as sal1 knockout lines produce numerous additional aleurone layers (Shen et al., 2003). A similar phenotype has been described in the hypermorphic Extra cell layer1 (Xcl1) mutant, the molecular nature of which has not yet been determined (Kessler et al., 2002). Both mutants demonstrate that aleurone identity (at least at a cytological level) can be maintained by cells located in nonperipheral positions. One hypothesis is that, in these two mutants, positional signalling is impaired, leading to an extension of the region where aleurone identity can be specified. SAL1 encodes a class E vacuolar sorting protein which could be involved in the trafficking of proteins required for the perception or transmission of a positional signal. Interestingly, SAL1 has been reported to co-localize with DEK1 in a small cellular compartment in maize aleurone cells, suggesting that the proper concentration of DEK1 for specification of aleurone identity may be maintained by SAL1-mediated recycling (Tian et al., 2007). Similarly, SAL1 may mediate recycling of CR4. Thus, it would be interesting to investigate the subcellular localization as well as the co-localization of DEK1 and CR4 in aleurone cells of the sal1 mutant.

Phytohormones may also play a role in the differentiation of aleurone cell fate, as ectopic cytokinin production under the regulation of a senescence-inducible promoter leads to production of a discontinuous aleurone cell layer in transgenic maize kernels (Geisler-Lee & Gallie, 2005).

Positional signalling and the maintenance of epidermal cell fate The early embryo arrest phenotype of protodermal mutants makes investigating the role of the corresponding genes in the maintenance of epidermal identity later during embryogenesis, and after germination, very difficult. In the case of AtDEK1 this problem was solved by the production of transgenic A. thaliana plants expressing an AtDEK1 RNAi construct under the control of the 35S promoter, which is not active during early embryo development, thus allowing the embryo to partially complete its development before silencing of the AtDEK1 transcript by the RNAi construct. Transgenic A. thaliana seedlings form cotyledons with mesophyll-like cells at the lamina surface, and in the most severe lines the cotyledons are fused together. Thus, a relatively late but extensive silencing of AtDEK1 activity causes a loss of epidermal cell identity in the cotyledons (Johnson et al., 2005). These results suggest that molecular elements such as AtDEK1 required for the initial differentiation of protodermal cell fate are also required to maintain the perception of positional signalling, continuously reinforcing epidermal identity during late embryogenesis (Fig. 2).

As DEK1 and CR4 seem to have overlapping functions in maize aleurone differentiation, the potential functional overlap of the orthologous genes was addressed in A. thaliana. There are five CR4-like genes encoding RLKs in the A. thaliana genome, ARABIDOPSIS CRINKLY4 (ACR4) being the most similar to maize CR4 (Cao et al., 2005). ACR4 expression is largely restricted to protodermal cells in the embryo, L1 cells in the SAM and epidermal cells in organ primordia (Tanaka et al., 2001; Gifford et al., 2003). Unexpectedly, acr4 mutant embryos are viable and are covered by a relatively normal protoderm, while the adult plant shows some defects in the differentiation of leaf epidermal cells, ovule integuments, and sepal margins (Gifford et al., 2003; Watanabe et al., 2004). The phenotype of AtDEK1-RNAi lines is exacerbated in an acr4 background (Johnson et al., 2005), suggesting that AtDEK1 and ACR4 act in the maintenance of epidermal identity via distinct pathways (Fig. 2). An interesting feature of ACR4 is its subcellular localization at the lateral and basal plasma membranes in epidermal cells of leaf primordia (Watanabe et al., 2004). Consequently, ACR4 may contribute to the maintenance of epidermal cell fate by receiving and transmitting signals from neighbouring epidermal cells and/or from underlying cell layers, rather than from the outside.

Two other genes, ABNORMAL LEAF SHAPE1 (ALE1) and ALE2, encoding respectively a subtilisin-like serine protease and a serine/threonine RLK, have also been proposed to contribute to the maintenance of epidermal cell fate (Tanaka et al., 2001, 2007). The ale1 mutant is seedling-lethal at low humidity. Developing mutant embryos and seedlings exhibit an irregular morphology of epidermal cells, lack a continuous cuticle layer and have crinkled leaves with graft-like fusions (Tanaka et al., 2001). Although defects in ale1 mutants are entirely restricted to embryo-derived tissues, ale2 mutants show epidermal defects both in seedlings and in adult plant tissues and are largely sterile (Tanaka et al., 2007). ALE1 and ALE2 act in different pathways because double mutants show synergistic phenotypes (Fig. 2). ale1/ale2 embryos fail to develop a uniform outermost cell layer, do not form one or both cotyledon primordia, and do not express the protodermal marker genes FIDDLEHEAD (FDH) and AtML1 in their apical region (Tanaka et al., 2007). An analysis of the genetic interactions with ACR4 revealed a much stronger epidermal phenotype for the ale1/acr4 mutant than for either parent (Watanabe et al., 2004) and a phenotype very similar to that of the ale2 single mutant for the ale2/acr4 double mutant (Tanaka et al., 2007). Consequently, ALE1 and ACR4 act in separate, albeit overlapping, pathways whereas ALE2 and ACR4 probably act in the same pathway (Fig. 2).

In addition to ALE2 and ACR4, two other RLKs have also been implicated in the formation of the embryonic epidermis. GASSHO1 (GSO1) and GASSHO2 (GSO2) are members of the Leucine-Rich Repeats (LRR) XI class of LRR RLKs and have been shown to act redundantly during embryogenesis, with double mutants showing cotyledon fusion and abnormal embryo bending. After germination, double-mutant seedlings are extremely sensitive to desiccation as a result of the production of an abnormally permeable cuticle, a phenotype very reminiscent of that of ale1 mutants (Tsuwamoto et al., 2008). So far the genetic relationships of the GSO genes with ALE1, ALE2 and ACR4 have not been clarified (Fig. 2).

Unlike ALE2, GSO1 and GSO2, which are expressed throughout the embryo (Tanaka et al., 2007; Tsuwamoto et al., 2008), ALE1 is not expressed in the embryo, but instead shows strong expression in the region of the endosperm surrounding the embryo called the embryo surrounding region (ESR; Fig. 1; Tanaka et al., 2001; Yang et al., 2008). Its expression is regulated by the basic helix-loop-helix (bHLH) transcription factor ZHOUPI/RETARDED GROWTH OF EMBRYO1 (ZOU/RGE1) (Yang et al., 2008). The mutants ale1 (Tanaka et al., 2001) and zou/rge1 (Kondou et al., 2008; Yang et al., 2008) and the double mutant gso1/gso2 (Tsuwamoto et al., 2008) are all impaired in the separation of endosperm and embryo and present similar epidermal defects in the seedling, namely a markedly increased permeability of the cotyledon epidermis resulting from defects in the formation of the cuticularized layer during embryogenesis. The zou/rge1 mutant is also impaired in endosperm breakdown in the ESR. Interestingly, however, neither gso1/gso2 double mutants nor ale1 mutants have been reported to show defects in endosperm breakdown, suggesting that regulation of ALE1 expression and of endosperm breakdown may constitute separable functions of ZOU/RGE1. Taken together, the above data lead to a model in which at least three parallel pathways control protoderm-specific gene expression and the maintenance of epidermal cell fate (Fig. 2): In the first, ALE1 potentially processes a signal molecule in the ESR which is then perceived by the embryo, possibly by the GSO receptors, although this remains to be tested. In the second, ALE2 and/or ACR4 perceives another (as yet unidentified) signal. Finally, DEK1 promotes epidermal identity in a pathway, which is likely to be separate to that involving ACR4.

Transcriptional control of epidermal cell fate After the perception and the transmission of positional signals by epidermal cells, the information is probably relayed at the transcriptional level to regulate molecular pathways involved in the acquisition of epidermal features. Consistently with this hypothesis, the expression of AtML1 encoding an HD-ZIP IV transcription factor was found to be down-regulated in AtDEK1-RNAi seedlings (Johnson et al., 2005) and in the apical part of ale1/ale2 or ale1/cr4 double mutants (Tanaka et al., 2007). While embryos mutant in both AtML1 and the functionally redundant gene PDF2 are not viable under normal conditions, they can be partially rescued if grown in vitro under high humidity and on high-sucrose concentrations. The phenotype of atml1/pdf2 seedlings was strongly reminiscent of that of AtDEK1-RNAi seedlings, as the rare leaf-like organs lacked an epidermis with the exception of sporadic stomatal clusters, thereby exposing mesophyll-like cells to the outside (Abe et al., 2003). Consequently, both transcription factors are required for the maintenance of epidermal cell identity, raising the question of how the signalling cascades and the epidermis-specific expression of AtML1 and PDF2 are linked, as well as the nature and function of the genes activated by AtML1 and PDF2.

An answer to the second question came from the comparison of the promoter sequences of the epidermis-specific genes PROTODERMAL FACTOR1 (PDF1), FDH and LIPID TRANSFER PROTEIN1 (LTP1) in A. thaliana, which led to the definition of a conserved 8-bp motif 5′-TAAATG(T/C)A-3′ called the L1 box. More recently, unbiased DNA-binding site selection assays for AtML1 and PDF2 defined the longer consensus binding sequence 5′-GCATTAAATGC-3′ (Nakamura et al., 2006), the palindromic nature of which fits the idea that HD-ZIP proteins bind DNA as dimers (Sessa et al., 1993). Site-directed mutagenesis of the original L1 box in a pPDF1::GUS (β-glucuronidase) fusion suggested the critical requirement of a native L1 box for epidermis-specific expression of PDF1 (Abe et al., 2001). Moreover, PDF1 expression is strongly reduced in the atml1/pdf2 double mutant, supporting in vitro binding data (Abe et al., 2003). The presence of an L1 box in the AtML1 and PDF2 promoters suggests a positive feedback loop (Fig. 2) through which AtML1 and PDF2 could reinforce epidermal identity (Abe et al., 2003). Interestingly, residual AtML1 expression in the atml1/pdf2 double mutant is L1-specific. This could indicate that auto-regulation plays a minor role in determining the L1 specificity of the expression pattern. However, given the residual ability of atml1/pdf2 double mutants to generate at least some epidermal cell types, it seems more likely that other members of the HD-ZIP IV gene family act partially redundantly with AtML1 and PDF2 (Tanaka et al., 2007). Another positive feedback loop may exist between AtML1 and ACR4 (Fig. 2) as ACR4 expression is reduced in the atml1/pdf2 double mutant (Abe et al., 2003) and AtML1 expression is affected in the ale1/acr4 double mutant (Tanaka et al., 2007).

To gain further insight into upstream regulatory elements (cis or trans) which restrict AtML1 expression to the protoderm, its promoter was completely dissected. Green fluorescent protein (GFP) fusions to small regulatory fragments allowed the isolation of a minimal 179-bp region which is sufficient to confer the typical AtML1 expression pattern during the earliest events of embryo development (Takada & Jurgens, 2007). This fragment contains, among others, the L1 box and a WUSCHEL (WUS) binding site, which are both necessary for the maintenance of AtML1 expression until the globular stage of embryogenesis but not for later developmental stages. The analysis of numerous deletion variants of this region revealed three distinct phases in the regulation of AtML1 expression: initial activation, subsequent maintenance and later activation. Thus, three distinct mechanisms may be necessary for initiation and early and late maintenance of protodermal cell fate. Superimposed on this temporal regulation is a further layer of spatial specificity. Specific regions of the minimal AtML1 promoter drive AtML1 expression in distinct domains along the apico-basal axis of the embryo. Taken together, these results suggest that different mechanisms are required for protodermal cell specification and maintenance along the embryo axis and during successive stages of embryo development. Moreover, these results clearly indicate that cis elements other than the L1 box or the WUS box are necessary and sufficient for the L1-specific expression of AtML1 at certain times and in certain tissues (Takada & Jurgens, 2007). A closer look at the expression pattern of WUS-related homeobox (WOX) genes in the A. thaliana embryo (Haecker et al., 2004) identified WOX9, with its specific expression in epidermal cells of the central embryo domain, as a good candidate for regulating AtML1 expression via the WUS box.

Cuticle biosynthesis and transport: acquisition of an environmental interface The cuticle is a highly hydrophobic lipid layer secreted asymmetrically by epidermal cells on their outside surface, thereby providing a protective film. The biochemical composition and ultrastructure vary considerably between species and organs (Jeffree, 2006), but two basic components are common to all cuticles: cutin and waxes. Cutin, a polyester of C16 and C18 hydroxy fatty acids and glycerol, represents the structural backbone of the cuticle. It is interspersed with and covered by waxes, a mixture of C24 to C34 alkanes, alcohols, ketones and wax esters (Nawrath, 2002; Kunst & Samuels, 2003). It is estimated that over half of the fatty acids produced by stem epidermal cells in A. thaliana are channelled into cuticular lipids, illustrating the importance of cuticle biosynthesis in epidermal cell metabolism (Suh et al., 2005). The functions of many biosynthetic enzymes have been deduced from the comparison of lipid profiles between wild-type and mutant plants, which were generally identified by shiny stem or leaf surfaces and called eceriferum (cer) in A. thaliana and glossy in maize (Jenks et al., 1995; Neuffer et al., 1997). Here we will only briefly summarize current knowledge of genes involved in cuticle biosynthesis (Table 1), as the subject has been treated in greater depth in other recent reviews (Pollard et al., 2008; Samuels et al., 2008; Kunst & Samuels, 2009).

During cutin monomer production, several substrates are subjected to multiple oxidation events resulting in an extremely complex pathway. These events are mediated by the cytochrome P450 monooxygenases LACERATA (LCR)/CYP86A8 (cytochrome P450 monooxygenase) and ABERRANT INDUCTION OF TYPE THREE GENES1 (ATT1)/CYP86A6 which catalyse the hydroxylation and oxidation of fatty acids, respectively (Willemsen & Scheres, 2004; Xiao et al., 2004). Another player is ADHESION OF CALIX EDGE/HOTHEAD (ACE/HTH), showing sequence similarity to long-chain fatty acid ω-alcohol dehydrogenases from Candida species (Kurdyukov et al., 2006b). Cutin polymer production probably requires the action of different acyltransferases in order to link its aliphatic, aromatic and glycerol monomers to each other. While the activity of GLYCEROL-3-PHOSPHATE ACYLTRANSFERASE4 (GPAT4) and GPAT8 (Li et al., 2007) seems to be necessary for the aliphatic acylation of glycerol-3-phosphate, the acyltransferase DEFECTIVE IN CUTICULAR RIDGES (DCR) may be involved in the linkage of hydroxylated fatty acids (Panikashvili et al., 2009). The α/β hydrolase BODYGUARD (BDG) is believed to participate in the polymerization of carboxylic esters (Kurdyukov et al., 2006a). Finally, the synthesis of long-chain fatty acids (C16 and C18) and possibly their incorporation into cutin polyesters appear to be achieved by LONG-CHAIN ACYL-COA SYNTHETASE1 (LACS1)/ECERIFERUM8 (CER8) and LACS2 (Schnurr et al., 2004; Lu et al., 2009). These enzymes also participate in elongation of wax fatty acids precursors. Detailed comparison of mutants revealed that lacs1 was more obviously affected in wax production while the lacs2 mutant was preferentially affected in cutin monomer accumulation. However, these enzymes seem to act partially redundantly as the double mutant shows a much stronger phenotype than the single mutants (Lu et al., 2009; Weng et al., 2010).

Although knowledge of enzymes involved in cutin biosynthesis remains fragmentary and relatively scarce, we now have an extremely clear picture of enzymes catalysing successive steps of wax biosynthesis in A. thaliana (Fig. 3). Wax biosynthesis is mediated by elongation of saturated C16 and C18 long-chain fatty acid acyl-CoAs into C20 to C34 very-long-chain fatty acid precursors (VLCFAs) in the endoplasmic reticulum. The extension of fatty acids is carried out by fatty acid elongase (FAE) complexes with unique substrate chain length specificities. FAE complexes integrate four distinct enzymes: β-keto acyl reductase (KCR), enoyl-CoA reductase (ECR), β-hydroxyacyl-CoA dehydratase (HCD) and the condensing enzyme β-ketoacyl-CoA synthase (KCS). Twenty-one KCS genes have been identified in the A. thaliana genome, including FDH, CER6 and FATTY ACID ELONGATION1 (FAE1), whereas only one HCD gene (PASTICCINO2), one ECR gene (CER10) and two KCR genes (KCR1 and KCR2) have been reported (Costaglioli et al., 2005; Zheng et al., 2005; Bach et al., 2008). The diversity of KCS genes confers the chain length specificity to the FAE complex, while divergent expression patterns confer tissue specificity (Joubes et al., 2008).

After elongation, the newly synthesized VLCFAs are modified into the major wax products via two distinct pathways: the acyl reduction pathway and the decarbonylation pathway (Kunst & Samuels, 2003). In the acyl reduction pathway, VLCFAs are reduced to primary alcohols by fatty acid reductases (FARs) such as CER4, and subsequently condensed with C16 fatty acids into wax esters by WAX ESTER SYNTHASE/ACYL-COA:DIACYLGLYCEROL ACYLTRANSFERASE (WSD1) (Rowland et al., 2006; Li et al., 2008). The decarbonylation pathway produces reduced wax compounds (alkanes, secondary alcohols and ketones) with odd numbers of carbons ranging from C21 to C33. In this second pathway, the loss of one carbon atom probably takes place between two reduction reactions probably involving the genes CER1 and/or CER22 (Samuels et al., 2008). The only well-characterized enzyme in this pathway is the cytochrome P450 monoxygenase MIDCHAIN ALKANES HYDROXYLASE1 (MAH1)/CYP96A15 which catalyses the subsequent internal hydroxylation of alkanes for the formation of secondary alcohols and ketones (Greer et al., 2007).

In maize, a collection of 18 glossy mutants affected in total wax load or wax composition of leaves has been established (Neuffer et al., 1997) and a few of the underlying GLOSSY genes have been identified. Among them, the KCRs GLOSSY8a and GLOSSY8b have a clearly defined, partially redundant function in the FAE complex (Dietrich et al., 2005), which may also contain GLOSSY2 (Tacke et al., 1995). GLOSSY1 encodes a desaturase/hydroxylase necessary not only for wax but also cutin biosynthesis (Sturaro et al., 2005).

Once synthesized, cutin and wax molecules must be exported to the plant surface. The mechanisms behind the transport and asymmetric deposition of cuticle components remain poorly understood. Based on the subcellular localization of biosynthetic enzymes, which are all associated with the endoplasmic reticulum, it is likely that the final compounds of the plant cuticle are produced in this compartment (Rowland et al., 2006; Greer et al., 2007; Li et al., 2008). Consequently, one or several export mechanisms from the endoplasmic reticulum to the extracellular matrix must exist. The proposed mechanisms include: Golgi-mediated vesicular traffic through the secretion pathway; loading into lipophilic cytoplasmic proteins to solubilize cuticular compounds; or direct loading by specific sites of close proximity between the endoplasmic reticulum (ER) and the plasma membrane (PM). None of these hypotheses has been substantiated by direct evidence (Kunst et al., 2006; Samuels et al., 2008).

Presently, the best candidates for the channelling of cuticular components through the plasma membrane are the two closely related ATP-binding cassette (ABC) transporters ABCG12/CER5 and ABCG11/WHITE BROWN COMPLEX 11 (WBC11)/DESPERADO (DSO)/CUTICULAR DEFECT AND ORGAN FUSION1 (COF1), located in the plasma membrane of A. thaliana (Pighin et al., 2004; Bird et al., 2007; PanikashviLi et al., 2007; Ukitsu et al., 2007). While there is no direct experimental evidence that these proteins transfer cuticular molecules, the respective mutants are significantly affected in wax deposition and lipid inclusions are found within the epidermal cytoplasm. In addition, CER5 and WBC11 belong to a particular subfamily of ABC transporters, the so-called ABCG or WBC subfamily, for which lipid transport capacity has been demonstrated in animals (Velamakanni et al., 2007).

Once exported from epidermal cells, the extremely hydrophobic cuticular molecules have to pass through the highly hydrophilic cell wall. Again, several mechanisms have been proposed, including the existence of small structures specialized in the transfer of wax molecules, and the binding of wax molecules to lipophilic proteins small enough to diffuse through the cell wall (Kunst et al., 2006). Lipid transfer proteins (LTPs) could carry out this function (Kader, 1996), and first evidence for a role of LTPs in the export of cuticular components was recently provided by the reduced wax load of the ltpg mutant (Debono et al., 2009; Lee et al., 2009). However, the localization of the corresponding protein at the plasma membrane of all cell faces rather than only at the outer cell wall may indicate that LTPG1 could facilitate the export of cuticular compounds through the plasma membrane in collaboration with ABC transporters (Debono et al., 2009).

Transcriptional regulation of cuticle biosynthesis Cuticular wax formation is known to be tightly regulated in response to both developmental and environmental cues, and in particular in response to water stress (Cameron et al., 2006). Among the transcription factors regulating the activity of genes involved in cuticle biosynthesis are several AP2/EREBP (Activator Protein2 (AP2)/Ethylene Response Element Binding Protein (EREBP)) family members. In A. thaliana, lines over-expressing WAX INDUCER1 (WIN1)/SHINE1 (SHN1) (Aharoni et al., 2004; Broun et al., 2004) or the closely related proteins SHN2 and SHN3 (Aharoni et al., 2004) exhibit increased wax production and cuticular permeability. WIN1/SHN1 over-expression also increases cutin production by the induction of cutin biosynthesis genes (Kannangara et al., 2007). Increased cuticular wax accumulation was also observed in Medicago sativa and A. thaliana plants over-expressing the Medicago truncatula WAX PRODUCTION1 (WXP1) protein, which belongs to a different clade of the AP2/EREBP family (Zhang et al., 2005, 2007). Over-expression of the paralogous WXP2 had similar effects. Several MYB family transcription factors also regulate cuticle-related genes. The over-expression of AtMYB41, an R2R3 MYB transcription factor, leads to an increased leaf epidermal permeability and changes in the expression of genes involved in lipid and cuticle metabolism (Cominelli et al., 2008). AtMYB30 directly regulates genes encoding enzymes of the FAE complex and its over-expression enhances accumulation of epicuticular wax alkanes (Raffaele et al., 2008). Finally, wax biosynthesis has been demonstrated to be under the control of CER7, encoding RIBOSOMAL RNA PROCESSING45 (RRP45), a ribonuclease acting at the level of mRNA stability rather than on transcriptional initiation in A. thaliana (Hooker et al., 2007). In summary, cuticular biogenesis seems to be tightly regulated at the transcriptional level by members of the MYB and AP2/EREBP families (Fig. 3).

In addition to their response to environmental or developmental cues, many cuticle-related genes are characterized by an expression restricted to epidermal cells. An example is the expression of FDH or CER5 in aerial parts of the plant (Yephremov et al., 1999; Pighin et al., 2004). Thus, these genes might be under the additional control of transcription factors which are able to promote epidermis-specific expression, such as members of the HD-ZIP IV family (Nakamura et al., 2006). Recent results obtained in tomato (Solanum lycopersicum) and in maize reinforce this hypothesis. In tomato, the identification of a point mutation in an HD-ZIP IV gene as the likely cause for cutin defects of the tomato fruit in the cd2 mutant made the first direct link between HD-ZIP IV transcription factors and cuticle biosynthesis (Isaacson et al., 2009). Furthermore, the over-expression of the HD-ZIP IV gene OCL1 in maize caused qualitative and quantitative changes in wax alcohols on the leaf blade, and in wax esters on both the leaf blade and the leaf sheath. These modifications in wax composition were associated with up-regulation of a FAR-encoding gene and three ABC transporter-encoding genes closely related to A. thaliana WBC11 and CER5 (Javelle et al., 2010). Interestingly, phylogenetic analysis of the HD-ZIP IV proteins in plants revealed that tomato CD2 and maize OCL1 belong to the same clade (P. M. Rogowsky, unpublished results), suggesting a possible subfunctionalization of these HD-ZIP IV proteins in regulation of wax biosynthesis.

Role of L1 cells in meristem function In A. thaliana, the layered organization of the SAM seems to be a prerequisite for meristem function and the presence of an intact L1 a requirement for the establishment and maintenance of this layered organization. The double mutant atml1/pdf2 not only fails to establish a differentiated protoderm at the apex of the embryo, it also loses its meristem structure, and ultimately meristem function (Abe et al., 2003). A very similar phenotype was observed in plants where DEK1 expression had been knocked down using RNAi (Johnson et al., 2005). In tomato, microsurgical laser ablation of the L1 layer of the SAM disorganizes its structure, compromises leaf initiation and finally leads to a gradual loss of meristem identity. L1 ablation also induces local changes in the division pattern of underlying L2 cells which switch from anticlinal to proliferative periclinal divisions, leading to the idea that the L1 restricts the division plane in the L2, thereby preventing periclinal divisions. In contrast to ablations in the L2 or L3, there appear to be no pathways permitting regeneration of ablated L1 cells (Reinhardt et al., 2003a).

How the L1 is involved in meristem maintenance mechanistically is unclear. Several models addressing the molecular mechanisms underlying meristem homeostasis have been published recently. While some models invoke the L1-specific expression of factors involved in meristem maintenance (Jonsson et al., 2005) in order to explain the maintenance of a stem cell pool, a more recent model has shown that a stem cell population can be established and maintained without invoking specific gene expression in the L1 cell layer (Hohm et al., 2010). However, it should be noted that this model none the less uses a meristem-like, dome-shaped template, and as the L1 is probably involved in regulating the growth of meristematic cells to give this form, the model still relies on the presence of L1 functions, albeit indirectly.

The ‘tensile skin’ theory: the control of plant shoot growth by the epidermal layer Because plant cells are glued together through their cell walls, different cell layers and tissue types must co-ordinate their growth to produce physiologically efficient organs. This raises the question of the relative contributions of the different cell layers to the regulation of overall organ growth and morphogenesis. The role of the epidermis in plant growth regulation has been investigated by using layer-specific promoters to modulate the expression of genes involved in cell division and cell expansion. For example, the inhibition of cell division in the epidermis by expression of the cell cycle inhibitor Kinase Inhibitor Protein (KIP)-RELATED PROTEIN1/INHIBITOR1 OF CDC2 KINASE (KRP1/ICK1) under the control of the AtML1 promoter (Bemis & Torii, 2007) resulted in small plants which had abnormally large epidermal cells. Surprisingly, cell division frequency in the mesophyll of these plants was not affected, leading to a reduction in mesophyll cell size and disturbed cell organization. These results suggested that the epidermis does not control cell division in underlying tissues, but that a reduced cell number in the epidermis puts mechanical constraints on cell expansion in the mesophyll (Bemis & Torii, 2007). The massively increased cell size in the epidermis of these plants also shows that, within certain limits, the expansion of the epidermis is entrained by the growth of underlying tissues. These results are intriguing as ectopically expressed KRP1/ICK has been shown to move between epidermal cells (from trichomes into surrounding cells) in the developing leaf (Weinl et al., 2005). It is possible that plasmodesmatal size exclusion limits between layers are different from those within layers during leaf development.

Interestingly, genetic mosaic analyses examining sectors lacking the cell cycle gene HOBBIT (HBT; a CDC27 homolog) in the epidermis demonstrated that lack of cell division in this layer is partially rescued by the presence of HBT in underlying cells (Serralbo et al., 2006). This rescue could be attributable to intercellular movement of cell cycle regulators, although this hypothesis has yet to be tested. Similarly, a recent study, in which the cell-layer specific effects of the growth regulator ANGUSTIFOLIA (AN) were investigated, showed that only a subset of the growth phenotypes shown by the an mutant could be rescued by epidermal expression of AN, whereas subepidermal expression of AN could rescue mutant growth back to normal levels (Bai et al., 2010). Moreover, as in the case of HBT, expression of AN (which is known to act cell-autonomously) in subepidermal cells stimulated cell divisions in the epidermal cell layer, presumably via non-cell-autonomous signalling. These experiments tend to support a hypothesis in which, at least in developing leaves, the division of epidermal cells is entrained to that of underlying cell layers, possibly by mechanical tension generated by cell expansion in the L2/L3.

Control of leaf growth by cell expansion has also been investigated using similar approaches. Brassinosteroids play important roles in controlling plant growth and seem to act largely (but not exclusively) at the level of cell expansion. When dwarf brassinosteroid biosynthesis or insensitive mutants were complemented with AtML1-driven BRASSINOSTEROID-INSENSITIVE1 (BRI1) or CONSTITUTIVE PHOTOMORPHOGENESIS AND DWARFISM (CPD) genes, encoding the brassinosteroid receptor and a biosynthetic enzyme, respectively, normal size was restored in these mutants (Savaldi-Goldstein et al., 2007). By contrast, L3-specific brassinosteroid biosynthesis or perception did not restore the wild-type phenotype. These results support a view in which the plant epidermis controls cell expansion in the shoot by the perception of brassinosteroids and the production of a nonautonomous signal of yet unknown nature, which acts in the L2 and L3 (Savaldi-Goldstein et al., 2007).

Taken together, these results highlight the possibilities as well as the complexities inherent in interpreting results even from extremely highly targeted strategies aimed at dissecting cell-layer-specific regulation of growth. However, it is clear that both biochemical signals and mechanical constraints are likely to contribute to inter-layer growth co-ordination. Many of the above experiments also show that, in expansion-competent backgrounds, the epidermis can compensate for defects in cell division, at least to some extent, by increasing cell size. For example, sectors lacking HBT activity in the L1 show only a partial restoration of cell division, but a normal final organ size, the shortfall in epidermal division being made up for by an increase in expansion. Similarly, excessive epidermal cell expansion is observed in plants expressing cell cycle inhibitors in the L1. These observations indicate that the epidermis can tailor its growth to that in underlying cell layers, presumably by sensing mechanical strain.

It has long been hypothesized that the plant epidermis has mechano-sensing and transducing functions, and this is an underlying tenet of the theory that the epidermis acts as a ‘tensile skin’ which is ‘stretched’ by pressure from dividing, expanding or turgid cells in underlying tissues. That this stretching can ‘harness’ the turgor pressure generated in pith cells, for example in stems, to impart mechanical rigidity is borne out by many classical observations (Cosgrove & Green, 1981; Niklas & Paolillo, 1997; Hejnowicz et al., 2000; Ryden et al., 2003; Kutschera & Nikas, 2007; Kutschera, 2008). Real evidence for such properties of the outer cell layer of meristematic tissues was only recently provided by the demonstration that the epidermal cells of the A. thaliana SAM are able to remodel their division pattern in response to mechanical stress (Hamant et al., 2008). This experimental evidence was correlated with modelling, the result of which reinforces the hypothesis that, at least in meristems, the epidermis acts as a tissue restricting growth. Interestingly, however, this may not be the case in all tissues. low cell density (lcd) mutants have dramatically reduced mesophyll cell numbers in leaves, but show no major changes in leaf size or shape, suggesting that, at least in leaves, epidermal expansion at later stages of development may not be entirely dependent upon mechanical cues (Barth & Conklin, 2003).

L1-mediated positional cues for the establishment of organ primordia and organ polarity Ablation of the meristematic L1 cannot only lead to loss of stem cell maintenance, but can greatly affect the initiation of organ primordia. These results underline the major role played by the L1 in leaf initiation and organogenesis in general. One explanation is that the epidermal layer can perceive, transmit or integrate signalling that promotes organogenesis. These signals include the phytohormone auxin, as the polar auxin transporter PIN-FORMED1 (PIN1) is preferentially expressed in the L1 driving an auxin flow in the epidermal layer towards the tip of the SAM (Reinhardt et al., 2003b).

In addition to the establishment of primordia per se, the epidermis also appears to make an important contribution to the establishment of dorsoventral polarity in leaves. While it has long been established that the specification of the adaxial side in the incipient leaf primordium requires a signal from the SAM (Sussex, 1951), the nature of this signal has been evasive. The so-called ‘Sussex signal’ was first proposed from results obtained in potato (Solanum tuberosum). More recently, microsurgical laser ablation experiments in tomato not only confirmed these data but also provided the first evidence that L1 cells contribute to the transmission of this signal (Reinhardt et al., 2005). In these experiments the disruption of symplastic continuity in the L1 between the tip of the meristem and the incipient leaf primordium led to the formation of leaves with radial rather than abaxial–adaxial symmetry (Reinhardt et al., 2005). As only the L1 was hit by these precise laser pulses it is possible to conclude that L1 cells are necessary to perceive and/or transmit the signal that confers abaxial/adaxial identity during leaf initiation.

Components of the signalling pathway potentially involved in establishing organ polarity were recently identified and characterized in maize. Abaxial–adaxial patterning involves a cascade of small RNAs regulating the expression of HD-ZIP III transcription factors that promote adaxial polarity (Nogueira et al., 2007; Douglas et al., 2010). The polarized expression of both microRNAs (miRNAs) and trans-acting siRNAs (ta-siRNAs) in the incipient leaf primordium directs the patterning of the adaxial–abaxial axis (Fig. 4; Nogueira et al., 2007). More precisely, miR166 promotes abaxial fate by repressing two HD-ZIP III genes, ROLLED LEAF1 (RLD1) and RLD2, which are sufficient to promote adaxial fate (Juarez et al., 2004). The asymmetric expression of miR166 opposes that of tas3-derived ta-siRNAs which are expressed in adaxial tissue and whose production is controlled by miR390. Thus, miR390 transcripts, which accumulate on the adaxial side of the incipient leaf primordium, represent the most upstream component of this pathway. Unexpectedly, the precursor transcripts of miR390 were found exclusively in the L1, while the mature miRNA was also present in subepidermal layers (Nogueira et al., 2009). This result suggests that a polarity signal, or component thereof, is perceived in the L1, promoting the expression of mir390.

In summary, taken together, the findings of experiments in potato, tomato and maize converge to suggest that the L1 is a strategic layer for the integration of positional cues which confer adaxial–abaxial leaf polarity.

The plant cuticle layer: an agent preventing organ fusion Plant organs are surrounded by their epidermis and the cuticle. The cuticle provides a highly hydrophobic barrier contrasting with the aqueous environment of the cell wall, and thus physically defines organ boundaries. The majority of mutants showing post-genital organ fusions also exhibit defects in the cuticle (Lolle et al., 1998). One of the most severely affected is the fdh mutant which is characterized by the fusion of leaves, floral organs and ovules, even though histological analyses indicate that the epidermal cell layer of these organs is intact (Lolle et al., 1992). The fact that FDH codes for a bona fide KCS involved in cuticle biosynthesis (Table 1) suggests that the resulting defect in the cuticle is the cause of organ fusion and not vice versa (Pruitt et al., 2000). The precise mechanism underlying organ fusion remains unknown and it is important to note that not all cuticle-defective mutants exhibit organ fusion, an example being cer5 (Pighin et al., 2004).

In maize and rice, organ fusions have been observed in some cuticle biosynthesis mutants such as the rice wax crystal sparse leaf1 (wsl1) mutant which is impaired in a gene encoding a KCS enzyme (Yu et al., 2008). The cr4 mutant (discussed in the section ‘Role of signalling proteins in the differentiation of the aleurone layer’) (Jin et al., 2000) and the adherent1 (ad1) mutant (Sinha & Lynch, 1998) from maize also show organ fusions. In the ad1 mutant, epidermal cells including specialized cell types such as stomata differentiate normally in large fused regions, but the extracellular matrix is perturbed. Epidermal cell walls in adherent leaves are abnormally thick and epicuticular wax particles appear reduced in size and number and altered in shape (Sinha & Lynch, 1998; Yu et al., 2008). As the molecular nature of AD1 has not been determined, it is difficult to firmly conclude that the cuticle has a causal effect.

The respective contributions of wax and cutins to organ separation have not been clearly established to date (Nawrath, 2002). Transgenic A. thaliana plants which ectopically secrete a fungal cutinase not only show alterations in the structure and properties of cutin, but also exhibit post-genital organ fusions (Sieber et al., 2000), suggesting that cutin may be more critical than waxes in the prevention of organ fusion. A closer look at the cer5 and wbc11 mutants strengthens this argument. While lack of both these ABC transporters led to a similar alteration in wax deposition, only wbc11 mutants were affected in cutin formation and exhibited organ fusion (Pighin et al., 2004; Bird et al., 2007).

Role of epidermal VLCFAs during embryonic develop-ment Embryo lethality has been observed in certain mutants defective in cuticle biosynthetic pathways, with mutants impaired in the elongation of VLCFAs particularly affected. In A. thaliana, pasticchino2 (pas2) seeds are completely collapsed (Bach et al., 2008) and kcr1 embryos undergo a premature arrest after the globular stage (Beaudoin et al., 2009). In maize, glossy8a/glossy8b kernels are not viable and contain a degenerated embryo surrounded by a normal endosperm (Dietrich et al., 2005). PAS2 encodes a VLCFA dehydratase (HCD) and KCR1, GLOSSY8A and GLOSSY8B encode VLCFA reductases (KCRs) which are all enzymes belonging to the FAE complex involved in VLCFA elongation (Fig. 3; Table 1). These genes have a clearly established role in cuticle biosynthesis as a weak pas2 allele, transgenic AtKCR1-RNAi lines and glossy8a or glossy8b single mutants all show a strong decrease in cuticular wax accumulation (Bellec et al., 2002; Dietrich et al., 2005; Bach et al., 2008; Beaudoin et al., 2009). Because the expression of these genes is not limited to the protoderm of the embryo and because the corresponding enzymes may produce VLCFAs for pathways other than cuticle biosynthesis (for instance triacylglyceride or sphingolipid biosynthesis), one may argue that the four mutants reflect more the importance of the quality of the VLCFA pool during embryogenesis than a particular role of the protoderm or the cuticle. However, the characterization of an additional mutant strengthens the hypothesis that defects could be attributable to defective protoderm development. Mutations in ACETYL-CoA CARBOXYLASE1 (ACC1; also called GURKE or PAS3) lead to a lack of layered organization in the apical part of the embryo (Baud et al., 2004; Kajiwara et al., 2004) similar to that seen in the atml1/pdf2 double mutant. Strong alleles are embryo-lethal, while weak alleles have effects on VLCFA accumulation in the seed (Baud et al., 2003). ACC1 is specifically expressed in protodermal cells and in acc1 embryos the expression of SHOOT MERISTEMLESS (STM), AINTEGUMENTA (ANT) and CUP-SHAPED COTYLEDON (CUC), which are marker genes of the SAM, cotyledons and cotyledon boundaries, respectively, is lost in the outermost cell layer but is unaffected in the underlying cells (Kajiwara et al., 2004). While it is clear that epidermal identity is severely compromised in this mutant, it remains to be clarified whether this is a result of a structural role of VLCFAs in protodermal cells or defects in fatty acid-derived signalling.

Although the link between epidermal lipid metabolism and epidermal specification remains unclear, one important point deserves to be highlighted here. The HD-ZIP IV class proteins, whose role in specifying epidermal cell fate has been previously discussed in the section ‘Transcriptional control of epidermal cell fate', contain steroidogenic acute regulatory protein lipid transfer (START) domains. In mammals, these domains have been shown to regulate protein function by binding small hydrophobic molecules, including cholesterol and other lipids (Alpy & Tomasetto, 2005). It is tempting to speculate that the activity of HD-ZIP IVs could be regulated by interaction with molecules generated during cuticle biosynthesis or catabolism, providing a positive feedback mechanism which would serve to maintain epidermal cell fate in cuticle-producing cells. One possibility could be that binding of lipidic molecules by the START domain might be required for movement of HD-ZIP IVs into the nucleus, in a mechanism analogous to that described for the glucocorticoid receptor (discussed in Kumar & Thompson, 2005).

New insight into the role of VLCFAs in plant development recently came from the detailed description of the pas1 mutant in A. thaliana (Roudier et al., 2010), providing a potential link between VLCFAs and polar auxin transport. PAS1, an immunophilin-like protein, was shown to interact with proteins of the FAE complex in the endoplasmic reticulum. The pas1 mutant exhibits defects in cotyledon formation associated with unco-ordinated divisions in protodermal cells. In addition, pas1 seedlings showed defects in VLCFA content and lateral root initiation (Faure et al., 1998; Roudier et al., 2010). Altered patterning at the apex of the embryo and defects in lateral root formation were both associated with defective polarity and polar auxin transport in these organs, indicated by abnormal PIN1 distribution. Sur-prisingly, application of exogenous VLCFAs restored lateral root formation in pas1 mutants, leading to the suggestion that long-chain lipid molecules are required for cell polarity upstream of polar auxin transport and organogenesis.

Cuticle and the development of specialized cells in the leaf While the vast majority of the leaf surface is covered by more or less uniform pavement cells, some cells of the epidermal layer enter a ‘secondary’ differentiation pathway which leads to the formation of specialized structures, such as stomata and trichomes. These epidermal structures are found regularly spaced throughout the leaf lamina in response to complex interaction with neighbouring pavement cells (Glover et al., 1998). Interestingly, their development is also influenced by cuticular properties, which could provide a means of integrating endogenous and environmental cues.

Plant stomata are microscopic valves in the plant epidermis surrounded by two guard cells which control gas exchanges across the central pore. In most plants stomatal density on the leaf surface is reduced in response to increasing atmospheric CO₂ concentrations. This response is impaired in high carbon dioxide (hic) mutants. Interestingly, HIC encodes a KCS enzyme that probably participates in very-long-chain fatty acid biosynthesis during guard cell cuticle development (Gray et al., 2000). Other mutants involved in cuticle biosynthesis show reduced stomatal indices under ambient CO₂ concentrations. They include cer1 and cer6 (Gray et al., 2000), and in particular wax2 (Chen et al., 2003). It is unclear how defects in cuticle composition might affect stomatal density. One possibility is that changes in cuticular permeability to gases could alter physiological signals perceived by the plant epidermis which globally regulate stomatal density. Another is that aspects of the complex cell–cell signalling pathways involved in stomatal patterning, for example the mobility of a ligand molecule, could be perturbed in cuticle mutants.

Trichomes, defined as any appendage of the epidermal layer, are widely represented at the surface of diverse plant organs including leaf, root, stem, flower and fruit. They vary considerably in their morphology, ranging from single cells with three branches in A. thaliana to multicellular, glandular structures in aromatic plant species. The genetic network that controls the initiation, the correct spacing and the final differentiation of trichomes in the model species A. thaliana is now quite well understood (Ishida et al., 2008), even though the initial signal remains elusive (Pesch & Hulskamp, 2004). Although no direct link between trichome patterning and cuticle biosynthesis has been made, several key enzymes involved in wax biosynthesis have been shown to be specifically up-regulated during trichome development in A. thaliana (Marks et al., 2009) and several cuticle-deficient mutants are affected in trichome density in A. thaliana (Yephremov et al., 1999; Kurata et al., 2003; Aharoni et al., 2004) and trichome distribution in maize (Sturaro et al., 2005).