Restoration of DWF4 expression to the leaf margin of a dwf4 mutant is sufficient to restore leaf shape but not size: the role of the margin in leaf development


*(fax +44 114 222 0002; e-mail


The role of the margin in leaf development has been debated over a number of years. To investigate the molecular basis of events in the margin, we performed an enhancer trap screen to identify genes specifically expressed in this tissue. Analysis of one of these lines revealed abnormal differentiation in the margin, accompanied by an abnormal leaf size and shape. Further analysis revealed that this phenotype was due to insertion of the trap into DWF4, which encodes a key enzyme in brassinolide biosynthesis. Transcripts for this gene accumulated in a specific and dynamic pattern in the epidermis of young leaf primordia. Targeted expression of DWF4 to a subset of these cells (the leaf margin) in a dwf4 mutant background led to both restoration of differentiation of a specific group of leaf cells (margin cells) and restoration of wild-type leaf shape (but not leaf size). Ablation of these cells led to abrogation of leaf development and the formation of small round leaves. These data support the hypothesis that events in the margin play an essential role in leaf morphogenesis, and implicate brassinolide in the margin as a key mediator in the control of leaf shape, separable from a general function of this growth factor in the control of organ size.


Leaves exist in many shapes. This is most obvious when comparing different species, but even within a species leaves formed at different developmental times or under different environmental conditions can develop different forms (reviewed by Tsukaya, 2006). Indeed, even a single leaf may develop different relative dimensions of length and width as it grows. The study of the proportional growth rate differences that underpin shape change during development is termed allometry, and any particular organ can be described to have a specific allometric relationship that defines the relative change in organ dimensions as the organ grows. The genetic basis underpinning allometry has long been described (Sinnott, 1936), and progress has been made in identification of the genetic factors influencing leaf form (e.g. Langlade et al., 2005; see also review by Tsukaya, 2006). However, the cellular mechanism underpinning these genetically defined allometric relationships remains very unclear. As has been pointed out, gene products cannot directly encode shape (Green, 1999). Rather, they must encode components of a shape-defining mechanism such that altered temporal or spatial activity of these components leads to a reproducible and robust output in terms of leaf form.

Classical work led to the hypothesis that the leaf margin played an important and deterministic role in leaf form in dicotyledonous species (described in Poethig and Sussex, 1985a). According to this hypothesis, a group of meristematic cells exists around the leaf perimeter (constituting the marginal meristem), and the extent of cell proliferation of these cells dictates the extent of lamina growth. Seminal experiments in the 1980s using clonal analysis showed that, generally, the progeny of cells at the leaf margin did not make a more significant contribution to leaf growth than cells at other positions within the leaf (Poethig and Sussex, 1985a,b). As a consequence, the concept of the marginal meristem fell into disrepute, and interest in this region of the leaf waned. However, a series of later investigations by various groups nevertheless implied that the margin did play a special role in dicot leaf morphogenesis. For example, at a very early stage after leaf initiation, cells at the perimeter cease dividing and undergo extensive elongation parallel to the leaf edge (Donnelly et al., 1999). This specific and unusual pattern of cell division (or, more specifically, termination of cell division) was interpreted as indicative of a specific function of the cells formed. Analysis of chimeras in tobacco plants with distinct genetically determined leaf shapes showed that, when sectors of one genetic background were entirely surrounded by cells of another background, the overall leaf shape followed that of the surrounding tissue. However, when (by chance) internal genetic sectors extended to the leaf perimeter, the tissue in that area seemed to take on the shape appropriate to the genetic background of that sector, i.e. the local leaf shape seemed to be determined by the genotype of the sector encompassing the leaf margin (Marcotrigiano, 2001). Further evidence linking the phenotypes of the margin and overall leaf shape came from studies of tomato leaf shape mutants (Kessler et al., 2001). A variety of mutants were studied, and a correlation was found between the organization and number of leaf margin cells and leaf shape. More recently, data implicating a role of the margin in leaf development have come from analysis of the role of auxin in vascular patterning. Thus, characterization of PIN protein distribution in developing leaves has led to the concept that directed influx of auxin at points around the leaf perimeter determines the sites of vascular differentiation (Scarpella et al., 2006). This model implies that there must be a pre-existing pattern of auxin flow around the leaf perimeter (for which there is some evidence, e.g. Mattsson et al., 1999, 2003), suggesting that the leaf margin is an important route of auxin flux within the leaf. If this is the case, then one might expect an influence of events at the margin on overall leaf growth. Recent data providing a link between margin form and transcription factors regulating leaf development are consistent with such a hypothesis (Zgurski et al., 2005).

The above data are indicative, but do not prove, that the leaf margin does indeed play a significant role in leaf development. To date, although there are good descriptive data on leaf margin differentiation and cellular dynamics (e.g. Donnelly et al., 1999; Poethig and Sussex, 1985a,b), very little is known about the molecular processes occurring in this region of the leaf. Therefore, in a first step to characterize the leaf margin, we set out to identify genes specifically expressed in this tissue. To do this, we screened an enhancer trap library to identify lines displaying a leaf margin-specific element of expression, with the aim of cloning the genes showing this expression pattern. Here we report the characterization of one such line, E1439. Our analysis provides a novel insight into the role of the growth factor brassinolide in the control of leaf shape via its action in the leaf margin, and, in a wider context, the relationship between organ size and shape.


An enhancer trap screen identified line E1439, which shows a leaf margin expression pattern and an altered leaf phenotype

To identify genes specifically expressed in the leaf margin, we performed a screen of an enhancer trap library created in the Poethig laboratory (see also One of the lines identified by this screen (E1439) showed restriction of GFP expression within the leaf, primarily at the margin but with some trichomes towards the proximal base also showing signal (Figure 1a,b). GFP signal was not apparent during embryogenesis, and only became visible in young leaves from about the P2–P3 stage of development. Signal was initially observed at the distal tip of the leaf, but this expression domain gradually extended around the perimeter of the leaf so that, by developmental stage P4–P5, it encompassed the entire leaf margin. At later stages of development, signal was restricted to the proximal margin and was sporadically observed in trichomes (Figure 1b). Confocal scanning laser microscopy of the E1439 line revealed that GFP expression was tightly restricted to the epidermal cell layer at the leaf perimeter (Figure 1c–f). As well as this leaf-specific pattern of GFP expression, signal was also observed in the root apex in a central column of cells proximal to the root apical meristem (Figure 1g).

Figure 1.

 GFP expression pattern and leaf phenotype in the enhancer trap line E1439.
(a) GFP expression in the leaf margin of a heterozygous E1439 seedling.
(b) GFP expression in the margin of a homozygous E1439 seedling.
(c) Confocal image at the margin of a heterozygous E1439 primordium. GFP signal (green) is apparent only in the leaf MCs, with no signal in the internal mesophyll (marked by red chlorophyll fluorescence).
(d, e) Confocal images of individual MCs in homozygous E1439 primordia. GFP signal is present only in the MCs, and does not overlap with the chlorophyll fluorescence, which is restricted to the mesophyll cells.
(f) Confocal image of a WT primordium using the same settings as (d). No GFP signal is apparent.
(g) GFP expression (green) is present in the root but excluded from the root apical meristem of homozygous E1439 seedlings.
(h) Homozygous E1439 seedlings showing a dwarf phenotype and small, curled dark-green leaves.
(i) WT seedlings of same age as in (h).
(j) Image of a P6 leaf from a homozygous E1439 plant.
(k) WT leaf of similar developmental age to that shown in (j).
Scale bars = 2 mm (a), 1 mm (b), 20 μm (c, f), 50 μm (d, e), 100 μm (g), 10 mm (h, i) and 500 μm (j, k).

In addition to this specific pattern of GFP expression, line E1439 displayed a clear phenotype. Seedling growth was very slow and the leaves were small, rounded and dark green compared with segregating wild-type plants (Figure 1h,i). A comparison of leaf shape is shown in Figure 1(j,k), where it appears that although the leaf width of leaves of approximately similar developmental age is similar, leaf length in the E1439 leaves is significantly shorter than that in wild-type (WT). This preferential reduction in leaf length rather than width was also apparent when line E1439 was grown on soil (Figure 2a), and was confirmed by quantitative analysis of the length and width of developmentally staged leaves (Table 1).

Figure 2.

 Leaf form and histology in line E1439.
(a) The first ten true leaves at maturity from a WT and an E1439 plant grown on soil.
(b) Longitudinal section through the 8th leaves from a WT and an E1439 plant.
(c) Distal tip of the WT section in (b).
(d) Distal tip of the E1439 section in (b).
(e) Proximal base of the WT section in (b).
(f) Proximal base of the E1439 section in (b).
(g) Whole-mount GUS assay of a WT plant containing a cyclin B::GUS construct. Signal (blue) is visible as speckles towards the base of the leaf.
(h) As in (g), but for a representative leaf from an E1439 plant containing a cyclin B::GUS construct.
Scale bars = 100 μm (b), 50 μm (c–f) and 500 μm (g, h).

Table 1.   Analysis of leaf growth
GenotypeLengthWidthRatio of length to widthSample size (n)
  1. The axial length (including petiole) and maximum width (both in mm, means ± SE) for leaf 8 at 35 days after germination were analysed in a number of soil-grown plants (n) for each of the genotypes described. WT is wild-type Col-0; E1439 is an allele of dwf4; E1439>>DWF4 are homozygous E1439 plants transformed with a UAS::DWF4 construct; controls are homozygous E1439 plants transformed with the pBIB vector. SE, standard error of the mean. Values in parentheses indicate variance. Populations with the same letter are significantly different from each other at the = 0.01 level using a t-test analysis.

WT55.1 ± 11.5 abc18.46 ± 3.87 d2.99 (0.755) e28
E143934.4 ± 5.81 a18.81 ± 2.661.83 (0.047) e31
E1439>>DWF435.6 ± 8.96 b12.06 ± 2.84 d2.98 (0.302)36
Control30.5 ± 10.1 c13.3 ± 3.422.29 (0.186)36

Scanning electron microscopy (SEM) analysis of the E1439 leaves indicated that there was a significant disruption of margin cell (MC) differentiation. During WT leaf development, cells at the presumptive margin are distinguished at an early stage of development by cessation of cell division (Donnelly et al., 1999). As cell growth continues commensurate with leaf growth, the MCs become massively elongated (Figure 3a,b) to form a cord of cells that defines the leaf perimeter. The MCs develop a relatively thickened, slightly undulating outer epidermal cell wall around a cell that is largely vacuolated and contains no chloroplasts and relatively little cytoplasm compared with the adjacent internal cells (Figure 3e,g). In the E1439 line, MC differentiation appeared to either aberrant or absent. Thus, although the E1439 leaves clearly formed a flattened structure, the cells at the margin did not take on the elongated form apparent in WT leaves of approximately equivalent developmental age (Figure 3c,d). Instead, cells in this region appeared more similar to epidermal pavement cells or formed stomatal complexes (which are generally absent from the leaf margin). As leaves grew, some formation of MCs was apparent, but these cells remained relatively short compared with equivalent WT MCs. Transmission electron microscopy (TEM) analysis revealed that cells at the margin position in line E1439 had an altered cell wall structure. Although the internal structure of these cells appeared similar to that of WT MCs, the outer epidermal cell wall was abnormal, being thinner with finger-like projections and a much thicker outer electron-dense layer (Figure 3f,h).

Figure 3.

 SEM and TEM analysis of the leaf margin in line E1439.
(a) MC formation at the edge of a P3-stage primordium of a WT plant. Asterisks (*) indicate representative MCs.
(b) MC formation at the edge of a P5-stage primordium in a WT plant.
(c) Margin region of a P5-stage primordium in the E1439 mutant. MCs are not apparent and stomatal complexes (arrows) are present on the leaf margin.
(d) As in (c), but viewing along the edge of an E1439 leaf.
(e) TEM of MC in a WT leaf.
(f) TEM of a cell in the margin position in an E1439 leaf.
(g) TEM of the outer epidermal cell wall of an MC in WT.
(h) TEM of the outer epidermal cell wall of a cell in the margin position in an E1439 leaf.
Scale bar = 30 μm (a), 60 μm (b–d), 10 μm (e, f) and 1 μm (g, h).

To investigate whether the mutant phenotype of the E1439 leaves encompassed an altered internal structure, we analysed leaf histology. Comparison of longitudinal sections through a WT leaf and an E1439 leaf (Figure 2b) shows the curvature along the proximal–distal axis that is typical of E1439 leaves. Moreover, analysis of the cellular patterning at the tip and base of these leaves indicates a higher cellular density in the E1439 leaf. Although this difference is relatively minor at the distal tip of the leaves (Figure 2c,d), the proximal base of the E1439 leaf is distinguished by the massive accumulation of very small cells (Figure 2e,f). These observations suggest that cell division is maintained at a higher rate for a longer time in the E1439 leaves. To test this hypothesis, we introduced a cyclinB::GUS construct (which can be used as a marker of mitosis) into the E1439 background (Colón-Carmona et al., 1999). Analysis of these plants indicated general maintenance of cyclinB::GUS activity for a longer time period over a larger leaf area towards the proximal base of the E1439 leaves compared with WT (Figure 2g,h). This tendency for a more intense GUS signal towards the base of developing leaves is consistent with the cell patterning data in Figure 2(c–f).

Molecular analysis reveals that E1439 is an allele of DWF4 and that DWF4 shows a specific pattern of expression in the epidermis during very early stages of leaf development

Analysis of the E1439 line showed that the phenotype was due to a recessive mutation at a single locus that segregated with the T-DNA insertion (data not shown). Our initial efforts to identify this locus using PCR-based methods failed; therefore, we resorted to a map-based cloning approach (as described in Experimental procedures). This revealed that, in the E1439 line, a T-DNA had been inserted into intron 7 of At3g50660/DWF4 at base position 2346 (Figure 4a). RT-PCR analysis of line E1439 confirmed that expression of DWF4 was absent in these plants and that GFP expression was present (Figure 4b).

Figure 4.

 Molecular analysis of line E1439, a novel allele of dwf4.
(a) Diagram showing the genomic arrangement at the DWF4 locus in the E1439 mutant. The T-DNA responsible for the E1439 phenotype is inserted in the 7th intron of At3g50660, which encodes DWF4, a cytochrome P450 (CYP90B1). The insertion at base 2346 relative to the transcription start site led to a 10 bp deletion (bold). Filled rectangles represent exons. The arrow indicates the orientation of the GAL4::VP16 coding region.
(b) RT-PCR analysis of homozygous E1439 seedlings reveals the presence of transcripts encoding GFP and the loss (compared with WT) of transcripts encoding DWF4. RBCS transcripts (present in both backgrounds) act as a positive control. GFP, RBCS and DWF4 indicate the specific primer pairs for these genes used in PCR analysis. E (E1439) and WT indicate the genotypes of plants from which cDNA was synthesized. m, DNA molecular size markers.

DWF4 encodes a cytochrome P450 enzyme (CYP90B1) with a key role in brassinolide (BL) biosynthesis (Choe et al., 1998, 2001). Previous analysis of dwf4 mutants [and other brassinosteroid (BR)-related mutants (Bishop and Koncz, 2002)] identified a phenotype broadly in line with that reported here, but did not indicate any linkage of BR biosynthesis and MC differentiation. To further investigate this possibility, we performed an in situ hybridization analysis of DWF4 transcripts in WT plants, focusing on very early stages of leaf development. The transcript level was generally very low, but our analysis revealed an accumulation of DWF4 mRNA in the epidermis of young leaf primordia (Figure 5a). As development proceeded, DWF4 transcripts became more restricted to the lateral and abaxial epidermis encompassing the MCs (Figure 5b), and, at later stages, also became apparent in the vascular tissue (Figure 5c). In situ hybridizations with an RBCS probe indicated a high level of signal in the mesophyll and exclusion from the epidermis and vasculature (Figure 5d–f). RBCS transcripts contribute a major fraction of mRNA in leaf tissue, thus the distinctive pattern observed in the DWF4 in situ hybridizations does not simply reflect a general pattern of bulk RNA distribution within the tissue. Control hybridizations with a DWF4 sense probe did not reveal any signal (Figure 5g).

Figure 5.

 Analysis of DWF4 and E1439 driven gene expression pattern in young leaf primordia.
(a) Cross-section of a WT leaf primordium (P3 stage) hybridized with an antisense probe for DWF4 reveals accumulation of DWF4 transcripts (dark blue signal) in the epidermis, including MCs, and some abaxial cells, but exclusion from the palisade mesophyll.
(b) As in (a), but at the edge of an older leaf (stage P5). Signal encompasses the margin and abaxial epidermis.
(c) Section through the margin of a P7-stage primordium hybridized as in (a). Signal is visible in the vascular tissue (arrows).
(d) As in (a), but hybridized with an antisense probe for RBCS. Signal is visible throughout the leaf but excluded from the epidermis and vascular tissue.
(e) As in (d), but for the edge of an older primordium (P5).
(f) As in (d), but providing an overview of a P6 primordium.
(g) Cross-section of the edge of a P5 primordium hybridized with a sense DWF4 probe.
(h) Cross-section of a margin of a leaf primordium (P5 stage) from a E1439>>DWF4 plant hybridized with an antisense probe for DWF4. Signal is apparent in the region of the MCs (arrow).
(i) As in (h), but the section is from the non-complemented E1439 line.
(j, k) GUS expression in DWF4::GUS leaves at an early stage of development (P3–P4). Signal (blue) is first apparent at the distal tip of the leaf and then becomes apparent along the leaf margin.
(l, m) Cross-sections of leaf primordia from DWF4::GUS plants. Signal (blue) is restricted to the extreme margin of the leaves. Cells in the marginal epidermis as well as some sub-marginal cells show signal. The counterstain is saffranin red.
(n, o) As in (j) and (k), but for later stages of leaf development (P6–P8). Signal is apparent along the length of the margin but is not detectable in the distal margin at later stages. Scale bars = 50 μm (a, d, e, g), 25 μm (b, h, i, m), 250 μm (c, f), 100 μm (l), 150 μm (j, k) and 1 mm (n, o).

Restoration of DWF4 to the margin of E1439 plants is sufficient to restore leaf shape but not size

The pattern of GFP expression observed in the E1439 driver line (Figure 1) represents a subset of the endogenous DWF4 expression pattern indicated by the in situ hybridization data (Figure 5). Notably, a margin pattern is highlighted in the driver line, whereas the general epidermal and vascular pattern is not observed. This distinction of the E1439 driver line pattern from the endogenous transcript pattern for DWF4 allowed us to perform a partial complementation experiment. Previous work had shown that uniform expression of the DWF4 gene in a dwf4 background was sufficient to restore leaf size and shape (Choe et al., 1998, 2001), and recent data showed that restoration of BR signalling throughout the leaf epidermis was sufficient to restore normal leaf growth (Savaldi-Goldstein et al., 2007). By using the E1439 line to drive expression of a UAS::DWF4 construct, we investigated the outcome of restoring DWF4 expression to the leaf margin in an otherwise totally dwf4 mutant background.

RT-PCR analysis of a number of transgenic lines containing the E1439>>DWF4 construct indicated that the DWF4 transcript was present (Figure 6a), and in situ hybridization analysis detected a low level of DWF4 transcripts only in a small group of cells around the margin of the complemented lines (Figure 5h), in contrast to the lack of detectable signal in the parental E1439/dwf4 line (Figure 5i), indicating that DWF4 expression was indeed restricted to the leaf margin in the E1439>>DWF4 plants. SEM analysis of E1439>>DWF4 leaves revealed restoration of MC differentiation to create a margin similar to that observed in WT leaves at a similar stage of development (compare Figures 6b and 3b) and distinct from that in the E1439 background (Figure 6c).

Figure 6.

 Margin differentiation and leaf shape in dwf4 plants with DWF4 expression restored to the leaf margin.
(a) RT-PCR analysis of a series of seedlings derived from homozygous E1439 plants transformed with a UAS::DWF4 construct identifies a number of lines showing the presence of DWF4 transcripts. cDNA synthesized from various putative E1439>>DWF4 lines was analysed by PCR using a DWF4-specific primer pair.
(b) SEM analysis of a primordium at stage P4 from an E1439>>DWF4 plant reveals the presence of MCs at the leaf edge.
(c) SEM analysis of a homozygous E1439 plant showing the original margin phenotype.
(d, f) Leaf phenotype of E1439>>DWF4 plants. Plants are small and slow-growing, but the leaf shape is more spatulate (less rounded) than a homozygous E1439 plant [shown in (e)] and similar to the WT leaf form [shown in (g)].
Scale bars = 20 μm (b, c) and 1 cm (d–g).

Surprisingly, analysis of the E1439>>DWF4 plants indicated that leaf shape (but not leaf size) was restored to that of control WT plants (Figure 6d–g and Table 1). Thus, the E1439>>DWF4 leaves (Figure 6d,f) appeared to be miniature versions of the WT leaves (Figure 6g) rather than the smaller and rounder versions observed in the E1439/dwf4 mutant (Figure 6e). Using the ratio of leaf length to width as a measure of leaf shape, comparison of leaf 8 from WT and E1439>>DWF4 complemented plants revealed no significant difference, although the growth (measured in terms of axial leaf length) of the E1439>>DWF4 leaves was significantly lower than that of WT plants (t-test, < 0.01) and similar to that observed in the E1439 mutant and control transformed plants. The shape of the E1439>>DWF4 leaves was significantly different from that of both control transformed and E1439 plants (t-test, < 0.01).

One possible interpretation of these data is that DWF4 protein expressed in the leaf margin can move to surrounding tissue, i.e. that it does not act in a cell-autonomous fashion and therefore a wild-type phenotype is restored in a wider field of tissue than that targeted by the E1439 driver line. To investigate this possibility, we created a DWF4::GUS reporter gene fusion under the control of the UAS promoter element and introduced this into the E1439 driver line. Reporter gene activity was restricted to the leaf margin in a dynamic fashion (Figure 5j,k,n,o). Expression was first observed at the distal tip of young primordia, then appeared more towards the proximal leaf margin as development progressed. This dynamic pattern was similar to that described for GFP expression in the E1439 line, but the GUS signal appeared to extend more into the leaf margin than the GFP signal. Analysis of sections of the DWF4::GUS lines indicated that GUS activity was indeed present in the MCs and in adjacent epidermal cells, and was detectable in some adjacent sub-epidermal cells (Figure 5l,m). No signal was apparent in either the vascular tissue or trichomes. Although it is difficult to totally exclude the possibility of some diffusion of the GUS product, the DWF4::GUS protein may have the capacity to move over short (cellular) distances, leading to DWF4 expression in a complex of cells surrounding the MCs. There was, however, no evidence of widespread movement of the protein. In addition, it should be noted that DWF4 expression under the control of the E1439 driver line did not restore all aspects of leaf morphogenesis, i.e. leaf size was not restored whereas leaf shape was. This again argues that the phenotype observed in the E1439>>DWF4 transgenic plants was not due simply to movement of DWF4 throughout the mutant plants.

Targeted genetic ablation of the margin leads to abnormal leaf development

To further investigate the function of MCs, we introduced a UAS::DTA (diptheria toxin A) construct into a heterozygous E1439 driver line with the aim of genetically ablating the GFP-marked cells in this line. During the early stages of leaf development, MCs were formed (Figure 7a), consistent with GFP and UAS::GUS expression not being detectable until after leaf initiation at stage P3–P4. At subsequent stages of leaf development, the MCs in the E1439>>DTA plants collapsed but the adjacent epidermal cells remained intact and turgid (Figure 7b). As growth progressed, the margin appeared to be pulled apart, leading to strands of collapsed tissue along the leaf edge (Figure 7c). Leaf growth was greatly impaired compared with WT, and there was altered leaf morphology. In mild cases, this was observed as some torsional twisting of the leaves (as seen in Figure 7c), and in extreme cases the formation of small concave leaves with abnormal epidermal and vascular differentiation (Figure 7d). The final leaves formed were minute, compact and curved (Figure 7e,f). Calculation of the mean leaf length/width ratio of these leaves gave a value of 1.03 (SE = 0.03, = 11). Comparison with values given in Table 1 for WT leaves indicates that margin ablation resulted in the formation of an approximately radially symmetrical leaf lamina compared with the normal ovate leaf form.

Figure 7.

 Outcome of genetic ablation of cells in the leaf margin.
(a) SEM analysis of MC formation at an early stage of leaf development (P2–P3) of heterozygous E1439 plants containing a UAS::DTA construct. MCs (asterisk) are formed.
(b, c) As in (a), but at a later stage of development (stage P4–P6). Epidermal cells are intact, but collapsed cells are visible in the region of the margin (arrows).
(d) SEM overview of a small, concave leaf from an E1439>>DTA plant.
(e) Light micrograph overview of an older E1439>>DTA leaf.
(f) Side view of the leaf shown in (e) to demonstrate leaf curling.
Scale bars = 20 μm (a–c), 100 μm (d) and 500 μm (e, f).


As outlined in the introduction, a number of authors over the years have proposed that the leaf margin plays an important role in morphogenesis. However, both the molecular events underpinning the specific pattern of differentiation displayed by cells in the margin and the function that the margin plays in the wider context of leaf development have been poorly characterized. The data presented in this paper help to clarify both of these issues.

With respect to margin cell differentiation, our data indicate that, in the leaf epidermis at an early stage of development, there is an accumulation of transcripts encoding DWF4, an enzyme previously shown to catalyse a key step in BR biosynthesis (Choe et al., 1998). These data provide direct evidence for a novel dynamic pattern of gene expression in leaves relating to growth factor (BL) biosynthesis. A previous investigation using a DWF4 promoter::reporter gene construct suggested a higher level of gene expression in young leaf primordia that encompassed a cotyledon and leaf margin component, but did not directly evaluate gene expression at the spatial resolution provided here (Kim et al., 2006). Thus, our data provide a spatial analysis of DWF4 expression during the earliest stages of leaf development, and add to a body of evidence indicating that there is a distinct spatial and temporal element to the control of genes whose products are involved in brassinolide metabolism and signalling (Castle et al., 2005; Kim et al., 2005; Montoya et al., 2005). Furthermore, our data provide support for recent results indicating that BR signalling in the epidermis is likely to be a major determinant of leaf growth (Savaldi-Goldstein et al., 2007). The mechanism underlying this control process awaits elucidation, but our results indicate that regulation of BR activity occurs in the epidermis during early leaf development, and thus could contribute to the control of epidermal-localized BR signalling in leaf growth.

The observation by Savaldi-Goldstein and colleagues that BR signalling in the epidermis plays a key role in leaf growth raises a number of questions. One of these is whether all cells in the epidermis play an equal role in mediation of the effect of BRs on leaf growth. Our data address this issue, as by exploiting the fortuitous pattern of expression of the enhancer trap that led to the dwf4 mutation in line E1439, we were able to restore DWF4 expression (and thus BR biosynthesis) to a sub-domain of the leaf. Thus, in the E1439>>DWF4 plants, DWF4 transcript and a DWF4:GUS fusion protein were restricted to the complex of cells at the leaf margin. The identity of the enhancer sequences responsible for this pattern of gene expression is a topic for future investigation; however, it is notable that the last intron of the DWF4 gene contains the sequence CGTGCG at base position 2394, 48 bp downstream of the E1439 T-DNA insertion site. This sequence has been identified as a potential target for the BZR transcriptional regulator, which plays a role in mediating BL-controlled gene expression (He et al., 2005). The presence of this sequence in the last intron of the DWF4 gene is consistent with the hypothesis that sequences in addition to the ‘standard’ 5′ promoter region are important in the regulation of DWF4 transcription.

Although our DWF4::GUS data indicate that the DWF4 protein targeted to the MCs probably diffuses within a small domain of cells encompassing the MCs, we have no data to suggest that the protein moves widely through the leaf. In addition, previous data from the analysis of chimeras suggest that BRs are relatively poorly mobile within the plant, and tend to act close to or at the site of synthesis (Symons and Reid, 2004). Also, the observation that leaf shape but not leaf size was restored in our transgenic plants indicates that our manipulation of DWF4 expression did not lead to a complete restoration of the normal endogenous pattern of BL expression. Absolute proof of the site of action of BL generated in the margin domain requires the development of high-resolution techniques to visualize the spatial distribution of this growth factor within plant tissue. To our knowledge, no such techniques have yet been established. Indeed, even for more fully explored growth factors, such as auxin, the lack of reliable physico-chemical-based methods for the quantification of growth factor distribution in intact tissue remains a problem. Although promoter::reporter gene constructs can be used to provide an indirect assessment of growth factor activity, the interpretation of such data can be problematic (Nemhauser et al., 2004). Our results highlight again the requirement for progress in the visualization of plant growth factor distribution.

Overall, our data indicate that, in the E1439>>DWF4 plants, DWF4 expression (and thus most probably BL biosynthesis and action) was restricted to a region around where the MCs normally form. As a result of this manipulation, MC expansion was restored. BL has long been associated with cell expansion, and a role for BL in the massive elongation that MCs undergo fits with this general function of BRs (Bishop and Koncz, 2002). However, it should be noted that the dwf4 phenotype also encompassed a major downstream effect on cell division. Thus, in the basal region of dwf4 leaves, there was a massive accumulation of very small cells, i.e. it was not simply the case that lack of BL led to the formation of a normal number of cells which did not undergo expansion. Whether this increase in cell number is a direct outcome of lack of BL or an indirect outcome of the limited tissue expansion resulting from lack of BL activity is a moot point, especially bearing in mind the possibility that the main point of action of BR may be in the epidermis (Savaldi-Goldstein et al., 2007) and not in the mesophyll (where the effect on cellular proliferation is most obvious). Further dissection of the downstream elements of BR signalling will help to resolve this issue (Nemhauser et al., 2004).

In addition to the restoration of MC differentiation, the shape of the dwf4/E1439 leaves was restored to approximately that of wild-type plants. Although BR mutants are often described as having smaller leaves, in actual fact the influence of BR on leaf growth is more complex, comprising both a general growth rate component and an allometric component with regard to the relative growth rates in leaf length and width. Thus, previous work identified BR biosynthesis mutant leaves as being rounder than WT, and showed that this shape defect was rescued by restoring BR biosynthesis throughout the plant (Kim et al., 1998, 1999). The mechanism by which BR might exert a polar influence on leaf growth was unclear. Our data take a step towards resolving this issue by showing that BR activity in the margin plays a key role in the co-ordination of increase in leaf length and width, i.e. they identify a specific region of the leaf involved in mediation of the influence of BRs on leaf allometry.

How might BR in the margin function to co-ordinate leaf length and width? At present we can only speculate. One possibility is that BR is required for MCs to undergo their characteristic elongation, and that the formation of a cord of elongated cells (MCs) at the perimeter has an important function in determination of the shape that the perimeter takes. This might, for example, be a biophysical function of the MCs, in which the physical strength of the cells acts to restrain lateral growth of the internal tissue. MCs clearly have a relatively thick cell wall and could act as mechanical struts around the leaf circumference. Weakening of these struts (as seems to occur in the E1439 line) might lead to the leaf perimeter bulging outwards under pressure generated by growth of the subepidermal cells, thus generating a rounder leaf shape. Alternatively, recent data on auxin flux and vascular patterning have suggested that the position of vascular elements within the growing leaf is dependent on the sites of influx of auxin from the leaf perimeter (Scarpella et al., 2006). If this is the case, then one would expect that disruption of the leaf margin would disrupt the route of auxin flux around the leaf perimeter, disrupting the pattern of auxin flux and thus the pattern of vascular differentiation. As there is a close correlation between vascular pattern and leaf form (Dengler and Kang, 2001), such a mechanism would provide a way of linking events in the margin with both leaf form and vascular patterning. The observations reported here that ablation of cells in the margin or inhibition of their differentiation (via loss of BL synthesis) led to altered leaf shape are consistent with this interpretation. Indeed, the massive disruption to leaf morphogenesis by directed ablation of cells in the leaf margin indicates that a functional margin is essential for leaf development. Future experiments aimed at manipulating the form and function of the leaf margin will help to resolve these possibilities.

Experimental procedures

Plant material

Seeds of Arabidopsis thaliana (Col-0) were surface-sterilized and germinated on solid medium using standard methods (Weigel and Glazebrook, 2002). Where appropriate for selection of transformants and segregation analysis, antibiotics were included in the growth medium. Growth conditions were 100 μmol m−2 sec−1 light, a 16/8 h photoperiod, temperature 24°C. For soil growth, conditions were 160 μmol m−2 sec−1, a 16/8 h photoperiod and 65% relative humidity. Transformed lines of Arabidopsis thaliana containing a UAS::DTA (diptheria toxin A) construct were obtained from colleagues (Laplaze et al., 2005; Laurent Laplaze, Institut de Recherche pour le Développement, Montpellier), crossed with heterozygous E1439 plants, and E1439>>DTA progeny were selected and characterized.

Constructs and transformation

A full-length DWF4 coding sequence was obtained from the ABRC stock centre (U13551) and cloned into the pBIB vector (gift from Jim Haseloff, University of Cambridge, UK). This construct (UAS::DWF4) was then transformed into homozygous E1439 inflorescences using standard techniques (Weigel and Glazebrook, 2002). Putative transgenic seed were selected and genotyped by PCR using appropriate primer pairs. A DWF4::GUS gene fusion was created using Gateway vector pKGWFS7 (Invitrogen;, so that the reporter gene was in-frame to replace the stop codon of the DWF4 sequence.

RNA analysis, in situ hybridization and electron microscopy

RNA extraction and PCR were performed as previously described (Pien et al., 2001). For in situ hybridization, young seedlings (8–12 days old) grown on medium were processed as previously described (Pien et al., 2001). A 822 bp fragment of DWF4 (bp 592–1398 of the full-length cDNA) was used to generate DIG-labelled antisense and sense probes. For green fluorescent protein (GFP) analysis, seedlings were viewed using a Leica MZ10 with epifluorescence light and a plant GFP filter (Leica; Images were captured using a charge coupled device and further processed using Adobe Photoshop. For confocal microscopy, dissected leaves were viewed on a Zeiss LSM 520 ( using an argon laser (488 nm). GFP signal was visualized via a 515–560 nm filter combination, chorophyll autofluorescence using a 590 nm LP (long pass) filter. Images were captured into the Zeiss LSM image browser. Individual images were processed further in Adobe Photoshop. Cryo-SEM was performed as previously described (Fleming et al., 1999).

Positional cloning

Homozygous lines of E1439 (Col-0 background) containing a single insertion locus (as determined by Southern hybridization and segregation analysis, data not shown) were crossed to Landsberg erecta ecotype to create a mapping population. Genomic DNA isolated from individual F2 E1439 plants was probed with CAPS markers (Konieczny and Ausubel, 1993), leading to linkage of the phenotype to chromosome 3. CAPs and SSLP markers were used to further define the position of the gene (Bell and Ecker, 1994), leading to positioning of the E1439 gene between markers T16K5-TGF and CIW4. To identify the mutated gene, we designed primer pairs spanning the candidate open reading frames in this region. A primer designed to the 3′ end of At3g50660/DWF4 (5′-TCCCCACGTCGAAAAACTAC-3′) yielded products when used with the right border primer of the T-DNA (5′-GTTTTGGGAGAGTAGCGACACTC-3′), and a primer designed to the 5′ end of DWF4 (5′-CTCAAGACGAGGCCAAAAAG-3′) yielded products when used with the left border primer of the T-DNA (5′-CCACCCCAGTACATTAAAAACGTC-3′). Sequencing of these PCR products revealed that the T-DNA had been inserted into intron 7 of At3g50660/DWF4, deleting 10 nucleotides of that intron (5′-GACGATGGTA-3′).


Funding for this project was provided by the Swiss National Science Foundation, the University of Sheffield and the Gatsby Foundation. During part of this work, A.F. was supported by a Swiss National Science Foundation START fellowship and laboratory space was kindly provided by Nicholas Amrhein (Swiss Federal Institute of Technology). We thank Asuka Kuwabara (University of Sheffield) for advice on sectioning, Jim Haseloff (University of Cambridge) for the gift of the UAS::DTA seeds, and Martin Goldwater (University of Durham, UK) for help with SEM. The initial work on the enhancer trap library was supported by funding from the National Science Foundation, USA, to S.P.