Combinations of Mutations Sufficient to Alter Arabidopsis Leaf Dissection

Leaves show a wide range of shapes that results from the combinatory variations of two main parameters: the relative duration of the morphogenetic phase and the pattern of dissection of the leaf margin. To further understand the mechanisms controlling leaf shape, we have studied the interactions between several loci leading to increased dissection of the Arabidopsis leaf margins. Thus, we have used (i) mutants in which miR164 regulation of the CUC2 gene is impaired, (ii) plants overexpressing miR319/miRJAW that down-regulates multiple TCP genes and (iii) plants overexpressing the STIMPY/WOX9 gene. Through the analysis of their effects on leaf shape and KNOX I gene expression, we show that these loci act in different pathways. We show, in particular, that they have synergetic effects and that plants combining two or three of these loci show dramatic modifications of their leaf shapes. Finally, we present a working model for the role of these loci during leaf development.


Introduction
Leaves are the main plant photosynthetic organs. They show a large variety in their sizes and shapes. Regardless of their final shape, leaves are initiated as small, finger-like primordia from groups of undifferentiated cells, the meristem [1,2]. These primordia will either grow out and rapidly differentiate to generate simple leaves, in which the blade forms a unique unit, or go through additional OPEN ACCESS morphogenetic events that allow the formation of lateral structures, thus generating the leaflets of compound leaves. Different leaf shapes can also be seen as the result of different patterns or levels of dissection: large dissection of the blade generate leaflets, less prominent dissections of the leaf or leaflet margins cause lobes or teeth [3].
Recent molecular genetics have revealed a surprising conservation of the regulatory mechanisms acting at different steps of leaf development, from their initiation at the meristem to the formation of leaflets in the case of compound leaves or to the dissection of the leaf or leaflet margins. Class I KNOX (KNOX I, KNOTTED 1-LIKE HOMEOBOX) genes maintain meristematic cells in an undifferentiated state [4] but also prevent precocious differentiation of primordia in most compound leaves to allow the formation of leaflets [5][6][7]. KNOX I proteins act, in part, through the modulation of hormonal pathways. KNOX I proteins promote cytokinin biosynthesis [8,9] to prolong the morphogenetic phase of compound tomato leaves, enabling leaflet formation [10]. Conversely, KNOX I proteins repress gibberellin biosynthesis [8,11] to prevent differentiation. An increased gibberellin signalling reduces leaflet formation in tomato and smoothens the serrated margin of tomato leaflets or the lobed margins of Arabidopsis leaves overexpressing KNOX I genes [12,13]. Whereas KNOX I genes maintain cells in an undifferentiated state, class II TCP genes (TEOSINTE BRANCHED1/CYCLOIDEA/PROLIFERATING CELL FACTOR) promote cell differentiation [14,15]. Five Arabidopsis TCP genes and the tomato TCP LANCEOLATE gene are targeted by a miRNA, miR319 or miRJAW [16,17]. Expression of miR319-resitant TCP genes leads to smaller leaves in Arabidopsis and a simplification of the compound tomato leaf. Conversely, inactivation of TCP genes, via for instance the overexpression of miR319, leads to prolonged growth leading to larger and serrated leaves in Arabidopsis and to super-compound tomato leaves [14][15][16][17][18][19].
Altogether, the opposite effects of the KNOX I and TCP genes, and the balance between the cytokinin and gibberellin pathways contribute to determine the window during which the leaf primordium remains undifferentiated and hence is able to respond to morphogenetic clues determining leaflet or teeth formation. Similar to local peaks of auxin response triggering primordium formation at the meristem [20,21], an increase in auxin response is also required for the patterning of leaf primordium to form serrations or leaflets [22][23][24][25][26]. Such an increase in auxin response contributes to the repression of KNOX I genes in the meristem during leaf primordium formation and in primordia of compound leaves during leaflet formation [22,25]. Proper individualisation of these auxin-mediated new growth axes (leaf primordia, leaflet primordia and developing teeth) require the activity of the CUC (CUP-SHAPED COTYLEDON) boundary genes [27][28][29][30][31][32]. Some CUC genes, like CUC2 in Arabidopsis or SlNAM/Goblet in tomato are targeted by a microRNA, miR164. Expression of a miR164-resistant CUC2 gene leads to deeply serrated leaf margins [32,33]. All these different regulators are not acting independently each of the other and numerous connections have been reported, generating an intricate regulatory network. For instance, CUC genes and auxin interact [23,31,34] and there is a cross talk between KNOX I and CUC genes [29,30,35].
Here, to investigate the genetic basis of leaf development we analyse the morphogenetic and genetic effects of combining different mutations that lead to an increased leaf dissection. We used CUC2g-m4 plants, which expresses a miR164-resistant CUC2 gene, the mir164a-4 mutant in which the MIR164A gene controlling CUC2 expression during leaf development is inactivated [32], the jaw-D mutant, in which the overexpression of miRJAW/miR319 down-regulates five TCP genes [17] and stip-D, which overexpresses STIMPY/WOX9 that maintains cell division and prevents premature differentiation [36]. We showed previously that leaf dissection in these mutants relies for a large part on a functional CUC2 gene and that the regulation of CUC2 by miR164 is still effective [30]. Here, we further investigate the relationship between these pathways and how they control leaf development. In the jaw-D single mutant, the dissection of the leaf margin can be divided in several orders of incision ( Figure 1A,B). The first order consists of the main teeth of the leaf (we will call hereafter "teeth" any clear outgrowth at the leaf margin, whatever its size). A second order appears with incisions of the main teeth. Additional orders of incision can occur and reach an order of five, the highest order being observed in the proximal teeth of old leaves. The stip-D mutant develops narrow and wavy leaves presenting a single order of dissection ( Figure 1C). The jaw-D stip-D double mutant has the same order of incision as the jaw-D single mutant but the dissections are deeper and can almost reach the midrib in the proximal part of the leaf. In addition, the teeth of this double mutant have a particular form, being narrower in the part close to the midrib and larger towards the outside of the leaf ( Figure 1J). double mutants. These ectopic leaves also show the fractal phenotype observed in the main leaves. (P) Reiterative growth of the leaf margin: example of the jaw-D CUC2g-m4 double mutant. The first lobe formed initiates two secondary lobes (label "2", only one labelled here). This process is reiterated to form the lobes of order 3 to 5 visible in this part of leaf (labels "3" to "5"). Bars = 1 cm in all panels except (M) and (N) (bars = 100 µm ) and (Q) (bar = 0.5 cm). The mir164a-4 and CUC2g-m4 mutants have serrated margins with an order of dissection of 1 or 2 ( Figure 1D,G). jaw-D mir164a-4 [30] and jaw-D CUC2g-m4 double mutants have curly leaves, with more pronounced dissections ( Figure 1E,H). Compared to the single mutants, the dissections of the double mutants are wider and deeper, although they never reach the midrib. The order of dissection is also strongly increased: at bolting it can reach 9 for jaw-D CUC2g-m4, and in plants grown for six months in short-day conditions, the order of incision reaches 12 and 15 for the jaw-D mir164a-4 and jaw-D CUC2g-m4 double mutants, respectively. The pattern of the different order of teeth is however simple, and can be compared to fractals: on a tooth of a given order, two new teeth appear at the base of the tooth to create a new order ( Figure 1M,N). This process is reiterated to give the final leaf shape ( Figure 1P). In some cases, ectopic leaves are observed in the sinus ( Figure 1O), hinting to the formation of ectopic meristems.

Stip-D Mir164a-4 and Stip-D CUC2g-m4 Double Mutant Leaf Phenotype
The stip-D mir164a-4 [30] and stip-D CUC2g-m4 double mutants share a similar type of leaf phenotype characterised by an extreme dissection of the teeth ( Figure 1F,I). In stip-D mir164a-4 the sinus nearly reach the midrib, whereas the dissection is even stronger in the stip-D CUC2g-m4 double mutant in which the proximal teeth are completely separated one from the other. The constriction at the base of the teeth is also accentuated compared to stip-D and jaw-D stip-D mutants and in the strongest case the base of the teeth is limited to the central vein of the teeth resembling a petiolule (the petiole-like structure supporting the leaflet). Despite these strong phenotypes, the leaf of these double mutants is still rather flat and not as curled as other double mutants described above.

jaw-D stip-D mir164a-4 and jaw-D stip-D CUC2g-m4 Triple Mutant Leaf Phenotype
As the above-described genetic analysis indicates that stip-D, jaw-D and mir164a-4/CUC2g-m4 control leaf development through different pathways, we constructed the jaw-D stip-D mir164a-4 and jaw-D stip-D CUC2g-m4 triple mutants ( Figure 1K,L). The jaw-D stip-D mir164a-4 triple mutant has a leaf phenotype resembling the jaw-D stip-D double mutant with a higher order of dissection ( Figure 1K). A clear constriction is observed at the base of the teeth of the jaw-D stip-D mir164a-4 triple mutant. The jaw-D stip-D CUC2g-m4 triple mutant displays an even stronger leaf phenotype ( Figure 1L,Q). The incision of the first proximal teeth of the leaf are so deep that they reach the midrib and that the structure is looking more like leaflets than teeth. In some cases, secondary leaflets with a clear petiolule structure can be observed on the primary petiolule-like structures ( Figure 1Q). The lamina is also widely affected, with a higher order of incisions, which gives rise to strong flatness defects. As a result, the leaves look like parsley and the rosette, instead of being flat like in the wild type, adopts a hemispheric shape ( Figure 1L).

Stip-D Promotes the Formation of Petiole
The progressive formation of petiole-like structures in stip-D and multiple mutants containing stip-D suggest that stip-D may promote the formation of a petiole. To further test this hypothesis, we examined the shape of the epidermal cells by scanning electronic microscopy (SEM). In the wild type, the petiole is covered by elongated and rectangular cells. Similar elongated cells are present in the lamina above the midrib. Other parts of the lamina are covered by jigsaw-shaped pavement cells and a progressive increase in the complexity of the epidermal shapes is observed from the midrib to lateral domains (Figure 2A-D). In the stip-D single mutant, the base of the teeth contains some elongated cells ( Figure 2E,H). The region that contained such cells is enlarged in the stip-D mir164a-4 double mutant and in the jaw-D stip-D mir164a-4 triple mutant ( Figure 2F,G,I,J) suggesting that the base of the teeth have acquired petiole characteristics.

Floral Phenotype of jaw-D, stip-D, mir164a-4/CUC2g-m4 Double and Triple Mutants
Floral organs are modified leaves. To investigate if the incision defects observed in leaves are conserved in floral organs, the sepals and petals shapes of the different mutants combinations involving stip-D, jaw-D and mir164a-4 were analysed. The numbers of floral organs in the different combinations of mutants varied slightly but these variations are not statistically significant except an increased number of sepals in the stip-D mir164a-4 double mutant ( Figure 3K).

Leaf Phenotype of TCP2 and TCP4 Insertion Lines
The miR319 miRNA targets five TCP genes: TCP2, TCP3, TCP4, TCP10 and TCP24 [17]. To get some insight into the contribution of individual TCP genes to the phenotype of the jaw-D mutant, we analysed the leaf phenotype of two insertion lines in TCP2 and TCP10 ( Figure 4A). At bolting, tcp2-101 and tcp10-1 single mutants exhibit a weak increase of leaf serration, compared to wild type ( Figure 4B-D), in agreement with previous observations [19]. This phenotype is more pronounced two weeks after bolting, while the wild-type leaf is not modified during these two weeks ( Figure 4E-G). These observations suggest that the down-regulation of any of these two TCP genes leads to a prolonged growth period during which the serration of the mutants becomes more pronounced. The tcp2-101 tcp10-1 double mutant shows a stronger leaf phenotype than the two parental single mutants, with an increased leaf serration already clearly visible at bolting ( Figure 4H). Like for the single mutants, this phenotype becomes more pronounced two weeks later ( Figure 4L). Furthermore, the distal margins of the teeth are curled downwards.
We next combined either tcp2-101 or tcp10-1 with CUC2g-m4. The two resulting double mutants show a similar increased leaf serration compared to CUC2g-m4 single mutant ( Figure 4I-K). The serrations are deeper and more teeth present secondary incisions than in the CUC2g-m4 mutant. Like for the tcp2-101 and tcp10-1 single mutants, the phenotype of the double mutants is more pronounced two weeks after bolting whereas the leaf phenotype of the CUC2g-m4 mutant does not become more pronounced after bolting ( Figure 4M-O). In some leaves of the double mutants, the proximal teeth are almost separated from the rest of the lamina.  The CUC2g-m4 mutant harbours leaves that can not be properly flattened out: in some cases, teeth of the leaf overlap, while the wild-type leaves do not present such phenotype and can easily be flattened out. The tcp2-101 CUC2g-m4 and tcp10-1 CUC2g-m4 double mutants also have leaves that could not be properly flattened out, and more overlapping teeth are present compared to the CUC2g-m4 mutant. A similar phenotype could also be observed in the jaw-D mir164a-4 and jaw-D CUC2g-m4 double mutants, though the phenotype was stronger in the combinations with jaw-D compared to tcp2 or tcp4 mutants ( Figure 1E,H). This suggests that the leaf phenotype of these mutants was not only due to an increased incision of the leaf margin but also to an extra growth of the margin compared to wild type.

Expression of KNOX I Reporters
Proper regulation of the KNOX I genes is central for leaf development [3,37,38]. Because ectopic expression of KNOX I genes in leaves is associated with the formation of increased leaf dissection, we analysed the expression pattern of KNOX I genes in different combinations of mutants using GUS reporter constructs for the four Arabidopsis The pKNAT1:GUS reporter has been shown to be often ectopically expressed in lobed leaves of mutants [6,39,40]. No ectopic KNAT1 expression is observed in the jaw-D, stip-D or CUC2g-m4 single mutants (Table 1 and Figure 5A-D). In contrast, KNAT1 expression could be observed in the tips of the serrations of jaw-D stip-D, stip-D CUC2g-m4 and jaw-D CUC2g-m4 double mutants ( Figure 5E-G,L,N). In addition, in jaw-D CUC2g-m4 and stip-D CUC2g-m4 double mutants, KNAT1 expression is also observed in the sinus of some teeth ( Figure 5M,N).    (Table 1 and Figure 5J,K).
In the wild type, activity of the pKNAT6:GUS reporter is observed in the tips of the serrations and at the base of the trichomes ( Figure 5O). pKNAT6:GUS is similarly expressed in the tips of the serrations and trichomes of the three single mutants. An additional activity of the reporter is detected in the sinus of all mutants ( Figure 5P-T). In the double jaw-D stip-D and stip-D CUC2g-m4 the expression in the sinus is very strong (Figure 5S-T).
No ectopic expression of the pKNAT2:GUS reporter is observed in any of the mutants analysed (Table 1 and Figure 5H-I).
Altogether, this shows that CUC2g-m4 is sufficient to induce ectopic expression of KNAT6 in the sinus. To obtain a similar KNAT1 ectopic expression, the presence of jaw-D combined with either stip-D or CUC2g-m4 is required (Table 1). This revealed a specific effect of CUC2g-m4 on some members of the Arabidopsis KNOX I family and a cooperative effect between the three different mutations analysed here.

Discussion
Here, we investigated the relationship between different genes regulating leaf dissection and analysed their combined effects on leaf shape and expression of the KNOX I genes. First, we show that the abnormal leaf phenotype of jaw-D, stip-D and CUC2g-m4 mutants can be further increased when any of these mutations are combined. Second, jaw-D, stip-D and CUC2g-m4 have different, and sometimes synergistic effects on KNOX I expression on leaves. Altogether, this confirms and extends the previous conclusion [30] that jaw-D, stip-D and CUC2g-m4 control leaf development via independent pathways.
The most striking phenotype of the mutants analysed here is the reiterative dissection process leading to several orders of dissection that dramatically modify leaf shape. On the basis of the leaf shape of the different mutants and combination of mutants, we propose a working model for the effect of the CUC2g-m4, jaw-D and stip-D transgenes on leaf development ( Figure 6) that underlines the role of several factors controlling the formation of leaves with complex shapes. Working model for the effects of CUC2g-m4, stip-D and jaw-D on CUC2 activity and margin development. The CUC2 activity (black) and leaf shape at three stages of development are shown for the four genotypes. The expression of CUC2 in the future distal sinus of the leaf is represented only at a young stage (A). Once the primary lobes have formed, the competence to initiate additional lobes ceases. In the CUC2g-m4 mutant, the expression of CUC2 is extended in width leading to an enlargement of the sinus. At the same time, CUC2 expression is reinitiated more rapidly in the primary lobe, leading to a second order of incision and to a secondary lobe (B). The competence to form lobes ceases rapidly however limiting the lobe order to two. In the stip-D mutant, the CUC2 expression pattern is the same as in wild type except that CUC2 expression is prolonged leading to an increased serration depth (C). In the jaw-D mutant, the duration of the window of competence to form lobes is increased, allowing the reiterative expression of CUC2 and the formation of several orders of lobes (D).
The first factor is the competence to form teeth that seems to be controlled by the TCP genes. We have down-regulated the TCP genes through two different strategies: first by knocking-out individually or in combination the TCP2 and TCP10 genes, and second, via the over-expression of miR319 in jaw-D. In both cases an increased leaf serration was observed, though at different levels. Conversely, expression of TCP genes resistant to miR319 leads to completely smooth leaves [17,18]. Therefore, a window of competence for teeth growth is setup by the action of the TCP genes and may be linked to their role in promoting differentiation [14]. This definition of a competence window by the TCP proteins is similar to the one proposed for tomato (Solanum lycopersicum). In tomato, the down-regulation of TCP genes through the over-expression of miR319 leads to leaflets of higher order. Conversely, a miR319-resistant form of the LANCEOLATE (LA) TCP gene leads to simple tomato leaves [16]. Therefore, the TCP genes may control in both Arabidopsis and tomato a window of competence to form leaf sub-units (either teeth or leaflets, Figure 6A,D).
The second factor is the patterning action of the CUC2 gene to instruct the formation of serrations. Formation of the serrations may involve local growth inhibition where CUC2 is expressed, in agreement with the role of CUC genes during organ primordium formation [41], but CUC2 may also have a promoting effect on teeth growth similar to the role of the NAM/CUC genes during leaflet growth [29]. Indeed, based on the comparison of wild-type and cuc2 mutant leaf outlines, Kawamura et al., suggested that CUC2 promotes teeth outgrowth [31]. In CUC2g-m4, expression of CUC2 is enlarged leading to wider and deeper sinus. Teeth of CUC2g-m4 leaves overlap, whereas wild-type leaves could be m o s t l y flattened out without any overlapping teeth. This also supports the role of CUC2 in the promotion of teeth outgrowth.
Large teeth of CUC2g-m4 also tend to harbour a second order of small teeth at the base of the first order teeth, which is not observed in the wild type. One possible explanation for this would be that formation of higher order of dissections would be controlled by a spacing mechanism such as lateral inhibition. Tips of the first order of serration would locally inhibit the formation of secondary serrations. As it has been shown that auxin distribution underlines first order teeth patterning [23], it is possible that tips of developing serrations deplete auxin from their surroundings, thus inhibiting the initiation of teeth of higher order. The base of the larger teeth of CUC2g-m4 would be outside the zone from which auxin is depleted and able to reinitiate the process of teeth formation. In this view, the increase in the order of dissection in the CUC2g-m4 mutant would be an indirect consequence of the deeper serration. However, in the stip-D mutant an increase of the serration is also observed but there is no increase in the order of incision. This is also the case for the jaw-D stip-D double mutant, which has the same order of incision as the jaw-D mutant despite a much stronger leaf dissection. Therefore, the teeth of higher order in CUC2g-m4 may reflect a shortening of the range of lateral inhibition, due for instance to modified production, distribution or sensitivity to auxin ( Figure 6B). A first step to test these hypotheses will be to perform a precise morphometric analysis of leaf development of these mutants to determine the geometry of the leaf margin at the time when the different orders of serrations are initiated.
When the CUC2g-m4 construct is combined with jaw-D, or with the tcp2-101 or tcp10-1 mutants not only is the growth of the serration increased, but also their dissection, as the teeth are more individualised than in any of the parental mutants. This is difficult to reconcile with jaw-D only controlling teeth growth, and suggests a synergistic effect between both constructs. Koyama et al., showed that TCP proteins activate the expression of MIR164A, a negative regulator of CUC2 [42]. This can, however, not account for the synergistic effect between jaw-D and CUC2g-m4, as CUC2g-m4 is insensitive to miR164. The same group showed previously that the expression of a chimeric repressor form of TCP3 (TCP3SRDX), leads to the ectopic expression of CUC3 that is not a target of miR164. TCP proteins may therefore have an additional effect on CUC gene expression independent of miR164 [18]. Finally, because a snapdragon CUC orthologue CUPULIFORMIS (CUP) interacts with a TCP protein in a two-hybrid test [43], interactions between CUC2 and TCP proteins may also contribute to the regulation of their activities.
In the stip-D mutant, and all mutant combinations including stip-D, a strong increase of the depth of the dissection between the teeth is observed. This phenotype is increased when the regulation of CUC2 by miR164 is disabled, suggesting another synergistic effect between CUC2 and stip-D. As stip-D increases the depth of the dissection between teeth, it could indicate that stip-D prolongs the expression of CUC2 ( Figure 6C). This hypothesis is consistent with the recent finding that two genes related to STIMPY are positive regulators of CUC2 during embryo development [44].

GUS Assay
The GUS staining of shoots harvested from plants grown in long day conditions in greenhouses was carried out as described [49], in the presence of 0.5 mM potassium ferri/ferrocyanide. After an over-night staining, the reaction is stopped by a 70% ethanol solution.

Scanning Electron Microscopy
Scanning electron microscopy was carried out as previously described [50]. The different images obtained where then assembled using Hugin software [51].

Conclusions
Our work reveals the strong flexibility of leaf development, as it is sufficient to modify the expression of only a few genes to generate extremely diverse leaf shapes. We have proposed a working model of the effects of the three transgenes analysed here that may help to understand how leaf shape can be modulated. Validation of this model awaits however tracking-down at their origin the changes in leaf shape that we describe here essentially for mature leaves and linking the dynamic changes in leaf geometry with changes in gene activity.