NUBBIN and JAGGED define stamen and carpel shape in Arabidopsis

As a result of evolution, lateral organs have assumed a vast array of shapes and sizes that adapt the sessile plant to diverse environments. In order for shape differences to emerge during organogenesis, morphogenesis must be carefully regulated by pathways that promote, maintain or suppress growth. Recently, several genes have been identified that play crucial roles in determining organ shape by regulating growth(Crawford et al., 2004; Dinneny et al., 2004; Ha et al., 2003; Hepworth et al., 2005; Nath et al., 2003; Norberg et al., 2005; Ohno et al., 2004; Palatnik et al., 2003). The C₂H₂ zinc-finger transcription factor encoded by JAGGED (JAG) is expressed in all organs and appears to act as a region-specific promoter of growth. In a loss-of-function jagbackground, lateral organ growth is partially suppressed. In petals, a correlation between the expression domain of JAG and the abundance of cells actively engaged in the cell-cycle suggests that JAG may regulate growth by maintaining or activating cell-cycle activity. Furthermore,the phenotypic similarity between jag loss-of-function mutants and transgenic lines in which cyclin-dependent kinase inhibitors are overexpressed, suppressing cell-cycle activity, further hints at such a connection (De Veylder et al.,2001; Wang et al.,2000). Unlike other genes that promote growth, such as AINTEGUMENTA and BIG BROTHER(Disch et al., 2006; Krizek, 1999; Mizukami and Fischer, 2000), JAG is also sufficient to promote morphogenesis in multiple types of lateral organs. In rosette leaves, for example, where JAG activity is normally limited to the blade region, ectopic expression can promote blade tissue expansion into the petiole region.

The ability of ectopic JAG activity to promote blade tissue led to the hypothesis that the BLADE-ON-PETIOLE (BOP) genes might negatively regulate JAG expression(Ohno et al., 2004). Like JAG gain-of-function lines, loss of BOP activity results in the ectopic development of blade tissue in the petiole region of leaves(Ha et al., 2003). Recent work characterizing the bop1,2 double mutant has further elucidated the role of BOP genes in lateral organ development. The BOPgenes negatively regulate JAG expression in lateral organs,particularly in the cryptic bract, a vestigle organ in Arabidopsisthat subtends floral primordia, but which develops as a macroscopic leaf-like organ in other species (Hepworth et al.,2005; Norberg et al.,2005). Ectopic expression of JAG in the cryptic bract rescues organ outgrowth, whereas loss of JAG activity suppresses bract outgrowth in mutants that ectopically develop bracts(Dinneny et al., 2004; Ohno et al., 2004). However,when JAG activity is removed from bop1,2 double mutants,which also develop ectopic bracts, no suppression of the bop-mutant phenotype was seen (Norberg et al.,2005). The observation that ectopic expression of a gene closely related to JAG, At1g13400 [previously referred to as JAGGED-LIKE and renamed NUBBIN (NUB) in this work],is also detected in bop1,2 mutants suggests that other factors in addition to JAG may be responsible for the ectopic tissue growth in bop mutants (Norberg et al.,2005). Independently, we have hypothesized that NUB may function redundantly with JAG(Dinneny et al., 2004), because JAG does not equally affect all lateral organs in which it is expressed. Here, we have used gain-of-function and loss-of-function approaches to reveal the function of NUB in development. We focus our analysis on the role of NUB and JAG in stamen and carpel development,where NUB is expressed in a polar manner on the organ surface oriented toward the shoot axis, termed the adaxial side. The polar expression of NUB on the adaxial side correlates with the growth and patterning defects observed in jag nub mutants.

Wild type was Col-0. jag-1(Dinneny et al., 2004) and SPL::GUS (Ito et al.,2004) have been described before. Seeds for the T-DNA insertion line 244A08 (also known as nub-1) were obtained from GABI-Kat(Rosso et al., 2003).

Genotyping of jag-1 has been described before(Dinneny et al., 2004). To genotype nub-1 we used primers oJD153 (AAG ACA GCG GAG GAT AAA GAT ATG) and oJD154 (GCT TCT CTT CAT CTT CTT CTT CTG G) to detect the wild-type allele and oJD155 (GTT CAT GTG GCC ACC GAG AGC TTG) and GABRB (GTG GAT TGA TGT GAT ATC TCC) to detect the nub-1 allele.

The NUB-RNAi contruct was made by amplifying a 326 bp NUBcDNA fragment using primers N-1392 (ATA GCT TTC CTC CTC ATC AAG GAC) and N-1393 (ATG TGG AGC AAC TCT AGA ACC ATT A), and cloning this into the pHANNIBAL vector (Wesley et al.,2001) in forward and reverse orientations (pJD93). The NUB-RNAi cassette was then inserted 3′ of an APETALA1promoter fragment derived from pAM571(Mandel et al., 1992) using BamHI restriction sites (pJD108). The FIL::NUB construct was made by amplifying a 4 kb FILAMENTOUS FLOWER promoter fragment, using primers oJD182 (gag ctc CAA CCA TTG AAC CAT CAC CGA TAT TC) and oJD183 (ggt acC TTT TTT GTA AGA AGG GGA AAA ATA TTG GAA GCT G) that introduced SacI and KpnI restriction sites (indicated by the lowercase letters) to the 5′ and 3′ ends, respectively. This fragment was then cloned into the pMX202 (Wu et al.,2003) binary vector (pJD136). A full-length NUB cDNA fragment was then amplified from pPY1(Dinneny et al., 2004), using primers oJD211 (ggt acc ATG AGA GCT GAT GAA AAT AAC ACT TTA G) and oJD212 (ggt acc TTA TAG CCC ATG ATG TGG AGG TAG ATG) that introduced KpnI restriction sites to the 5′ and 3′ ends, and cloned into pJD136(pJD148). Transgenic Col-0 plants were generated by the floral-dip method(Clough and Bent, 1998).

RNA amplification and reverse transcription reactions were performed as previously described (Kardailsky et al.,1999). PCR amplification was performed on 2 μl of reverse-transcription reaction. A NUB fragment encompassing the entire coding region was amplified using oJD119 (AAG CTT ATG AGA GCT GAT GAA AAT AAC AC) and oJD120 (GGA TCC TTA TAG CCC ATG ATG TGG AGG). A TUBULIN fragment was amplified as previously described(Dinneny et al., 2004).

In situ hybridization was performed as previously described(Dinneny et al., 2004). An anti-sense NUB probe was transcribed using T7 RNA polymerase(Promega) from a partial NUB cDNA clone (pJD109) linearized with HindIII. The JAG (Dinneny et al., 2004), FIL(Siegfried et al., 1999) and PHB (McConnell et al.,2001) probes have previously been described. Scanning electron microscopy (SEM) was performed as previously described(Dinneny et al., 2004). Embedding of plant material in JB-4 media, sectioning and staining with Toluidine Blue was performed as previously described(Roeder et al., 2003). Saffranin O and Alcian Blue staining of paraffin sections performed as previously described (Roeder et al.,2003). GUS staining performed as previously described(Blázquez et al.,1997).

Cell layer number and cell width measurements were determined from the same transverse cross-sections of stage-13 gynoecia. For each genotype, 16 sections were imaged from four gynoecia. Sections were taken from the approximate middle of the gynoecium and were separated by at least 30 μm from each other. Cell layer number was measured in four regions of the carpel. Two counts were obtained from valve tissues flanking the main vascular bundle of the valve (lateral region) and two counts were obtained from valve tissues flanking the presumptive valve margins (medial region). The average width of cells in the ena, enb and outer epidermal cell layers were determined as follows. First, the total width of all cells in a particular cell layer was measured using Image J software(http://rsb.info.nih.gov/ij/). Next, the number of cells in this layer was determined. The average cell width was then calculated by dividing the width of all cells in a layer by the number of cells in that layer. Both valves of the gynoecium in each section were counted separately.

Abaxial trichome number during leaf development was determined for the area of the leaf blade outside of the midvein, as this region normally contains trichomes in most leaves. Segregating T2 transgenic FIL::NUB plants were selected on MS agar plates containing 20 μM kanamycin then transferred to soil. Wild-type control plants were also germinated on MS agar plates before being transferred to soil. Trichome counts were performed on plants after bolting. Kanamycin is unlikely to affect abaxial trichome development,as a FIL::GFP transgenic line germinated on kanamycin-supplemented agar plates did not develop ectopic abaxial trichomes (data not shown).

In order to determine what roles NUB may play in development, we examined its expression in wild-type plants(Fig. 1). During vegetative development, NUB is expressed in initiating leaf primordia(Fig. 1A-C). Expression is restricted to the inner surface of primordia, termed the adaxial side. Unlike other genes in Arabidopsis expressed on the adaxial side of leaves,such as PHABULOSA (PHB), PHAVOLUTA and REVOLUTA, NUB is not expressed in the shoot meristem(Emery et al., 2003; McConnell et al., 2001; Waites et al., 1998).

Like JAG, NUB expression is excluded from the cryptic bract(Fig. 1D). Unlike JAG,however, NUB expression is also excluded from sepal primordia(Fig. 1D). The expression of NUB in petals is transient and can only be detected during the initial outgrowth of the primordia (Fig. 1K). During stamen development, NUB is strongly expressed throughout initiating primordia during stage 5 and then becomes restricted to the adaxial side by stage 6 (Fig. 1E-G). At stage 8, NUB expression becomes restricted to a small cleft of tissue in between the two sets of microsporangia(Fig. 1H). This expression pattern differs from JAG, which is expressed throughout stamen primordia (see Fig. S1 in the supplementary material). During carpel development, NUB is initially expressed in cell layers of both abaxial and adaxial halves (Fig. 1G), it then becomes restricted to the adaxial cell layers at stage 7 (Fig. 1H,I). NUB expression can also be seen on the adaxial side of initiating ovule primordia (Fig. 1J). In summary, NUB expression is more restricted than that of JAGand is found predominantly in leaf, stamen and carpel primordia. Furthermore,although JAG is expressed throughout all tissue layers of organ primordia, the adaxial expression of NUB suggests that it may play a role in promoting adaxial-specific development.

We identified a T-DNA insertion allele of NUB, 244A08, in the GABI-Kat collection (Rosso et al.,2003). Sequencing of the insertion site by GABI-Kat indicates that line 244A08 has a T-DNA disrupting the second intron of NUB(Fig. 2A). Examination of NUB transcript levels shows that the nub-1 mutation strongly diminishes the accumulation of the NUB full-length transcript(Fig. 2B). The homozygous mutant does not display an abnormal phenotype (data not shown). This was not entirely unexpected, as the NUB expression domain is fully encompassed by the JAG expression domain. The jag nub double mutant, however, results in a strong enhancement of defects in leaf, stamen and carpel development compared with jag single mutants (see below). To ensure that these enhanced defects are specifically due to the nub-1 mutation, we generated a NUB-RNAi construct and introduced this into a jag mutant background and observed similar defects in stamens and carpels as in jag nub mutants(Fig. 2C,D). Sepal development is also affected indicating that our NUB-RNAi construct may result in some non-specific degradation of other transcripts (data not shown).

Examination of flower development in jag nub mutants reveals that the effect of the nub mutation is restricted to the center two whorls. Defects in sepal development observed in jag single mutants are not enhanced by nub (Fig. 3A-C). Petal growth also tends to be similar between jagsingle mutants and jag nub double mutants; however, the phenotype of jag nub petals is more consistent (data not shown), whereas defects in petal growth are more variable in jag flowers(Dinneny et al., 2004).

Stamens are divided into two regions, an apical anther and a basal filament(Bowman, 1993). In jag nub mutants, anther development is nearly abolished, whereas the filament does not appear to be affected (Fig. 3, see also Fig. S2 in the supplementary material). We examined the anther region more closely to determine whether any anther-specific cell types remained in jag nub mutants. On the adaxial side of wild-type anthers, four microsporangia develop, which enclose the developing pollen in chambers termed locules (Fig. 3D,K). During anthesis, the pollen is released from the anthers at a dehiscence zone that develops between the two locules in each pair. On the abaxial side, a small lump of tissue, termed the connective, joins the two sets of microsporangia to the filament(Fig. 3G,K). When we examined the cells present in the stub of tissue that remains in the anther region of jag nub mutants, we saw that it is composed of two different domains(Fig. 3F). In the apical half of the stub, cells with a slight puzzle shape form with cuticular ridges(Fig. 3F,I). In the basal region, cells with a rounded shape develop without cuticular ridges(Fig. 3F). The apical cells resemble those normally found in the connective region of the anther(Fig. 3H), whereas the basal cells do not resemble any cell type found in the anther.

We next examined anthers in plastic thin sections to determine the cell types that develop internally. Wild-type anthers develop four locules within the microsporangia (one locule per microsporangium) where the pollen grains form (Fig. 3K). In jag nub mutants, no such tissues can be found(Fig. 3M). These data show that locule development does not occur in jag nub mutants and that the stub of tissue that remains may mostly be composed of abaxial connective. We have also found that the effect of the nub mutation on anther development is semi-dominant when in a jag mutant background. In jag nub^+/- mutants, some microsporangia growth is still apparent (Fig. 3J). In addition, these plants are partially fertile, indicating that they are capable of producing pollen (data not shown).

We also took a closer look at the anthers of jag single mutants in plastic sections and found that there seemed to be fewer locules(Fig. 3L). This was at first surprising, as the anthers appeared to have four microsporangia when imaged by SEM (Fig. 3E). When the sections of jag anthers were examined more closely it could be seen that a narrow crease still develops and runs down the center of the microsporangia at the point which would normally divide the two locules and where the anthers normally dehisce (Fig. 3L). Thus, although the external structure suggests that four microsporangia have formed, the internal structure suggests that only two locules are patterned in jag mutants, or that the tissue that normally separates the locules does not develop properly.

To further characterize the defects in stamen development seen in jag nub mutants, we examined the expression of genes that mark different regions of the anther. We first examined FIL, which is initially expressed on the abaxial side of anthers(Fig. 4A). Later, as tissues on the adaxial side proliferate to form the microsporangia, FILexpression remains on the abaxial side in tissues that do not proliferate as much and become the connective (Fig. 4B). In jag nub mutants, FIL expression initiates in the anthers as it does in wild type(Fig. 4C). During stage 7,however, FIL expression extends toward the adaxial side(Fig. 4D). The apparent expansion of FIL onto the adaxial side of the anther may indicate that abaxial fate has expanded to the adaxial side in jag nubanthers. Alternatively, the expansion of FIL may be an indirect consequence of unequal growth of the abaxial side in relation to the adaxial side in jag nub mutants. In either case, these data suggest that the anthers of jag nub mutants are largely composed of connective tissue,marked by FIL expression, and that the adaxial microsporangia tissues fail to proliferate.

We next examined genes that are normally expressed in tissues on the adaxial side of the anther. PHB is initially expressed in the vasculature and on the adaxial side, but it soon becomes restricted to cells that denote the two dehiscence zones of the anther(Fig. 4E). In jag nubmutants, where no microsporangia development is apparent, PHBexpression is limited to a few cells in the center of the jag nubanther, presumably marking the vasculature(Fig. 4F). Note that PHB expression on the adaxial side of the carpels is not affected in jag nub mutants (Fig. 4E,F).

To further explore why the adaxial microsporangia region of the anther does not properly develop in jag nub mutants, we examined the expression of SPOROCYTELESS (also known as NOZZLE) using an SPL::GUS reporter line that marks cells of the microsporangia(Ito et al., 2004; Schiefthaler et al., 1999; Yang et al., 1999). During wild-type anther development, the SPL reporter is expressed in two domains on the peripheral flanks of the anther(Fig. 4G)(Ito et al., 2004). Surprisingly, in jag nub mutants, where microsporangia formation is not apparent, SPL reporter activity is still initiated in the proper domains (Fig. 4I). Later during anther development, SPL reporter activity is localized to the basal region of jag nub anthers, where we observed rounded cells with little differentiation (Fig. 4J). Finally, reporter activity is lost in the anther region of jag nub mutants (Fig. 4K), whereas in wild type, expression is maintained until anthesis(Fig. 4H). These data show that, although microsporangia development is blocked in jag nubmutants, the patterning of the microsporangia domains appears to occur normally. Thus, JAG and NUB may not be required to specify microsporangia per se, but instead act in parallel to such specification pathways to promote the growth of these tissues. This function of JAGand NUB may be similar to the role that JAG plays in petal development where it promotes distal petal growth.

To further test whether NUB regulates microsporangia identity or plays a more restricted role in promoting the growth of this region, we used the FIL promoter to drive NUB expression on the abaxial side of anthers in the connective region. We reasoned that if NUB is sufficient to promote microsporangia identity, then FIL::NUBtransgenic lines might be expected to promote ectopic microsporangia formation in place of the connective. Alternatively, if NUB regulates growth,but does not control the identity of microsporangia tissues, then we might expect the connective to become enlarged in FIL::NUB anthers but still retain the same cell types as in a wild-type connective.

As shown in Fig. 5,introduction of a FIL::NUB transgene into wild-type plants results in anthers with expanded abaxial tissue. In moderately affected lines, such as JD148-B, most of the ectopic growth is localized to the connective region,resulting in an enlarged and elongated structure(Fig. 5A,B). In stronger lines,such as JD148-A, the entire abaxial side appears to become overgrown and can be seen exceeding the limits of the adaxial microsporangia(Fig. 5C,D). To determine whether this ectopic tissue is composed of microsporangia-like or connective-like cell types, we analyzed plastic thin sections of FIL::NUB anthers and found it to be composed of cells most similar to the connective (Fig. 3K, Fig. 5E).

The observation that FIL::NUB transgenic lines have an abnormal phenotype suggests that the FIL promoter is active in these lines. We confirmed our prediction by examining FIL expression in FIL::NUB flowers and found FIL to be expressed on the abaxial side in two cell layers (Fig. 5F). Interestingly, the expression of FIL in anthers appeared to have expanded laterally, probably because of the positive effect of NUB expression on tissue proliferation. Thus, despite being normally expressed in mutually exclusive domains, FIL expression is compatible with NUB expression in FIL::NUB transgenic lines. These findings differentiate NUB from other adaxially expressed genes such as PHB, which are sufficient to suppress FIL expression if mis-expressed on the abaxial side(Siegfried et al., 1999).

Although the gynoecium is mildly affected in jag single mutants(Dinneny et al., 2004; Ohno et al., 2004), jag nub double mutants show a strong decrease in the growth of the valves near the apical region of the gynoecium, resulting in exposure of the ovules(Fig. 6G,H). Tissues not associated with the valves, such as the replum and style seem to be less affected and grow apically beyond the limits of the valves. The splitting of the style, which was occasionally observed in jag mutants, is further enhanced in jag nub gynoecia (Fig. 6) (Ohno et al.,2004).

We next followed these defects throughout gynoecium development. In wild-type flowers, the gynoecium develops as a cylinder of tissue(Fig. 6A)(Bowman, 1993; Smyth et al., 1990). During this early stage, jag nub gynoecia develop notches near the apical end (Fig. 6B). As the gynoecium grows, these notches enlarge and appear to be restricted to the valve regions(Fig. 6C,D). At stage 11, the gynoecial cylinder closes with the formation of stigmatic papillae in the apical region (Fig. 6E). In jag nub mutants, however, the gynoecium does not close, leaving a gaping hole in the ovary (Fig. 6F). The valve tissue in this apical region is often shredded in appearance (Fig. 6H). These data show that JAG and NUB promote the growth of valve tissues during gynoecium development, and that failure of jag nubvalves to enclose the ovules is caused by this defect. In agreement with this interpretation, we also have found that strongly affected FIL::NUBtransgenic lines develop valve tissues that grow beyond their normal limits and expand into the style region, showing that NUB is also sufficient to ectopically activate valve growth (Fig. 8A,B). Interestingly, the epidermal cells of the ectopic valve tissue develop cuticular wax similar to epidermal cells found in the style region.

The adaxial-specific expression of NUB during carpel development suggested that NUB may play a role in the development of tissues in this region. To explore this possibility, thin sections were made of wild-type, jag, nub and jag nub gynoecia(Fig. 7; data not shown). In wild type, the valves or carpel walls of the gynoecium are composed of six layers of cells: an outer epidermis, three layers of mesophyll, a modified layer of mesophyll termed the enb and finally the inner epidermal layer termed the ena (Fig. 7A). Most of the cell layers are established early during gynoecium development as the dome of tissue, which forms the gynoecium, begins to develop a cleft down the middle to create a cylinder of tissue at stage 7(see Fig. S3 in the supplementary material)(Alvarez and Smyth, 1998). Later, during stage 9-11, cells near the valve margins of the valve divide,resulting in a final count of six cell layers. At flower opening (stage 13),each cell layer is distinguishable as a result of differences in cell size. The enb cells, for example, are very small owing to cell divisions that begin at stage 9. By contrast, cells of the ena are very large and also begin to differentiate in size from the other cell layers around stage 9.

In jag nub mutant gynoecia, we found that the number of cell layers was reduced (Fig. 7A,C). We measured cell layer number in two regions of the carpel walls and found a one to two cell layer reduction (Fig. 7D). This reduction always appeared to be at the expense of the mesophyll. These defects suggest that JAG and NUB are necessary to promote the cell divisions that established the correct number of cell layers. We also found that jag single mutants display a slight decrease in the number of cell layers, but not to the extent of jag nub mutants (Fig. 7A-D). We examined nub single mutants but did not observe any apparent defects in cell layer number or differentiation (data not shown)

The differentiation of specific cell types in the gynoecium is also affected in jag nub mutants. As mentioned before, cells in the enb layer undergo cell divisions during stage 9-12, resulting in cells with a very small width. Late in fruit development, these cells become lignified and contribute to fruit opening (Spence, 1996). In jag nubmutants, however, these cells are enlarged, presumably as a result of the reduction in the number of cell divisions that occur(Fig. 7A,C,E). We examined jag nub fruit and could observe cells in the enb layer that were enlarged and did not stain for lignin (see Fig. S4 in the supplementary material). The overall size of the cells and the lack of lignification suggest that these cells may have acquired partial mesophyll identity. Only the largest cells of the enb layer tended to be unlignified.

Development of the ena layer was also affected in jag nubmutants. Whereas enb layer cells were larger in jag nubmutants, ena cells were smaller than in wild type(Fig. 7A,C). Both the ena and enb defects are seen in jag single mutants,but to a lesser degree (Fig. 7A-C,E). The outer epidermal cells were slightly wider in jag and jag nub mutants compared to wild type (two-tailed t-test, P=0.0025 and 0.0210, respectively), but there was no significant difference between jag and jag nub mutants(P=0.8257), consistent with the lack of NUB expression in this cell layer (see Fig. S5 in the supplementary material). These data indicate that JAG and NUB play a specific role in promoting the differentiation of adaxial cell types in the carpel walls.

JAG and NUB promote the cell divisions that establish the proper number of cell layers in the gynoecium. We explored the role of JAG and NUB in this process further by examining the effect of ectopic NUB expression on cell layer number in FIL::NUBtransgenic lines. Consistent with our loss-of-function studies, we find that ectopic NUB expression is sufficient to promote the amplification of cell layer number in valves, sepals and petals. In carpels, most of the increase in cell layer number is restricted to tissues overlying the main vascular bundle (Fig. 8C,D). In sepals, the effect is more uneven (Fig. 8E,F). Petals, which normally have only two to three layers of mesophyll, develop up to six layers in FIL::NUB(Fig. 8G,H). Although cell layer number is affected, no effect on the identity of the tissues that develop on the abaxial side can be seen in the flower. In the gynoecium, where JAG and NUB play a role in the differentiation of adaxial cell types, no ectopic enb or ena layer development is seen in FIL::NUB transgenic lines (Fig. 8C,D; data not shown). These data indicate that JAG and NUB may regulate the differentiation of adaxial-specific cell types in the gynoecium in parallel with genes that control tissue distribution,similar to their role in development of the microsporangia.

The development of ectopic blade tissue along the petioles of jag-5D rosette leaves demonstrated that JAG is sufficient to promote leaf blade growth. The blade area of wild-type leaves is, however,similar to that of jag-1 mutants(Dinneny et al., 2004). More recently, we have shown that combining jag-1 with the filand yabby3 (yab3) mutations results in a dramatic loss of blade development, demonstrating that JAG does indeed promote leaf blade expansion (Dinneny et al.,2005). When we examine the leaves of jag nub mutants, we also see a reduction in leaf blade area compared with wild type(Fig. 9A-D). The shape of jag nub leaves is somewhat variable and notches can form in the leaves (Fig. 9C). These notches may represent enhanced versions of the serrations found in jag single mutants (Fig. 9B). Leaves of jag nub mutants also develop outgrowths on the abaxial side(Fig. 9E-G). These outgrowths are often radialized and develop a high concentration of trichomes. In kanadi mutants, such outgrowths, termed enations, are also observed(Eshed et al., 2004). The enations of jag nub mutants are most common overlying vascular tissues, and tend to occur more frequently near the tips of the leaves and serrations. Scanning electron microscopy of the epidermal cells that constitute jag nub enations shows that they are shaped similarly to cells overlying vascular bundles (Fig. 9G).

Although abaxial expression of NUB does not have an effect on the differentiation of abaxial cell types in the flower, we found that leaf epidermal development is affected. During early vegetative development in wild type, leaves form trichomes (leaf hairs) only on the adaxial surface(Fig. 10)(Kerstetter et al., 2001). After the plant has produced about six leaves, trichomes begin to develop on the abaxial side and increase in number until the formation of the cauline leaves. In both strong (Fig. 10) and moderate (see Fig. S6 in the supplementary material) FIL::NUB transgenic lines, trichome development is precocious in comparison with wild type. For example, the transgenic line JD148-A begins to develop abaxial trichomes as early as the third leaf. The early development of trichomes on the abaxial side of leaves suggests that the FIL::NUBtransgene is able to impart partial adaxial identity to the abaxial side of leaves. We have also noticed that FIL::NUB transgenic lines occasionally develop trichomes on the adaxial side of floral organs, including petals, stamens and carpels (Fig. 8A; data not shown).

Early during lateral organ development in plants, a coordinate system is established that differentiates the adaxial side from the abaxial side. Soon afterwards and possibly coincident with organ polarization, organ outgrowth is initiated, marking the first morphogenetic event in the development of the lateral organ. In Arabidopsis, we have shown that the subsequent morphogenetic events that determine organ shape rely on the closely related JAG and NUB genes to specify organ domains competent for growth. In the absence of JAG and NUB activities, both morphogenesis and the subsequent late events of histogenesis fail to occur properly in lateral organs. Although JAG and NUB are able to activate morphogenesis and are necessary to promote the differentiation of cell types, they are not sufficient to promote the differentiation of these cell types in most cases. We therefore propose that JAG and NUB regulate cellular differentiation in cooperation with other factors that provide positional information.

In this work, we have extended our understanding of the role that JAG plays in shaping lateral organs by unraveling a layer of genetic redundancy caused by NUB, which masks the effects of jagmutations on stamen and carpel development. In stamens, JAG and NUB function together to promote microsporangia formation. Our analysis, using the expression of genes that mark different regions of the anther, reveals that the jag nub anther is composed largely of connective tissue. Interestingly, however, the initial expression pattern of the SPL microsporangia marker appears normal in jag nubanthers. These data suggest that JAG and NUB are not required to specify the identity of microsporangia tissues, but instead promote growth on the adaxial side, which is a prerequisite for microsporangia formation. In the gynoecium, JAG and NUB act to drive sufficient growth of the valves necessary to enclose the ovules. Growth defects in the gynoecium of jag nub mutants appear to be limited to the valves, while replum and style growth is unaffected. This function is the converse of that performed by genes such as CRABS CLAW and SPATULA, which affect the growth of stigma and style tissues without causing defects in valve growth (Alvarez and Smyth, 1999; Alvarez and Smyth, 2002).

Characterizing the redundant roles of JAG and NUB has allowed us to determine the relationship between their spatial patterns of gene expression and their activity. This is particularly important for determining directness, and to establish whether these genes control distinct morphogenetic events or act more globally to control growth throughout an organ. Although the expression of JAG suggests that it may play a role in stamen and carpel development, it was necessary to examine the function of NUB to elucidate the nature of this activity. Taking into account the mutant phenotypes of jag and jag nub mutants reveals that the expression domains of JAG and NUB directly correlate with defects in growth that occur when their functions are disrupted. Furthermore misexpression studies demonstrate that both JAG and NUB are able to activate growth in ectopic locations. Together, these data show that JAG and NUBcontrol specific morphogenetic events during lateral organ development that are crucial for determining final shape.

Unlike JAG, which is expressed in all cell layers of lateral organs, NUB is limited to the adaxial side of lateral organs. Many studies have focused on elucidating the pathways that are involved in determining adaxial/abaxial polarity. Expressed on the abaxial side, the GARP-type transcription factors encoded by KANADI 1, 2 and 3promote abaxial development by activating the expression of FIL and by negatively regulating the expression of adaxial promoting genes such as the miRNA-regulated class of HD-ZIP transcription factor genes, PHB,PHAVOLUTA and REVOLUTA(Emery et al., 2003; Eshed et al., 2001; Eshed et al., 2004; Kerstetter et al., 2001; Mallory et al., 2004; McConnell et al., 2001; Pekker et al., 2005). Ectopic expression of PHB has been shown to negatively regulate FILexpression, whereas ectopic KAN expression on the adaxial side can suppress PHV and REV expression. One interesting peculiarity with the expression domain of adaxial identity factors is that they are often expressed in both the meristem and the adaxial side of lateral organs. NUB expression, however, is limited to the adaxial side of lateral organs and is excluded from the meristem. This suggests that NUB may function downstream of, or in parallel to, the polarity pathways. Examination of FIL and PHB expression in jag nub mutants shows no clear mis-regulation of either gene in the gynoecium. In stamens, FIL expression appears to extend circumferentially into the adaxial domain; however, this may simply be a consequence of the altered growth dynamics of a jag nub anther, where the normally proliferative adaxial microsporangia region is stunted. Furthermore, ectopic expression of NUB on the abaxial side does not suppress FIL expression,nor does it result in ectopic PHB expression (see Fig. S7 in the supplementary material). Consistent with these observations, phenotypic analysis of NUB gain-of-function and loss-of-function lines does not reveal any homeotic conversion of abaxial or adaxial tissues. These data suggest that JAG and NUB regulate histogenesis in parallel with pathways that establish polarity. The polarity pathways, which are known to regulate the spatial distribution of tissues, may require input from pathways that regulate growth. Although the nature of this relationship is unclear, recent work demonstrating that FIL and YABBY3 act in cooperation with JAG to pattern the distribution of tissues in the fruit supports the existence of such a partnership(Dinneny et al., 2005). It may be that, in addition to promoting growth, JAG and NUB also promote the competence of tissues to respond to factors that determine tissue type and distribution, thus linking morphogenesis to histogenesis.

Two observations complicate any strict delineation of JAG and NUB function in adaxial development, however. First, jag nubmutants develop enations on the abaxial side of their leaves. Guided by the model proposed by Waites and Hudson, that the juxtaposition of adaxial and abaxial domains drives tissue outgrowth(Waites and Hudson, 1995),enation formation in jag nub leaves suggests a role for JAGand NUB in establishing/maintaining the boundaries between adaxial and abaxial halves. The second observation, that NUB expression on the abaxial side is sufficient to drive precocious trichome development further suggests that NUB may also play a more active role in regulating adaxial identity.

We would like to thank Bernd Weisshaar and the GABI-Kat Consortium for the nub-1 allele, Toshiro Ito and Elliot Meyerowitz for the SPL::GUS reporter line, and Jeff Long for the PHB in situ probe. We thank Evelyn York for technical help with SEM work, which was done at the Scripps Institution of Oceanography. We thank B. Crawford, K. Gremski and A. Roeder for careful reviewing of this manuscript. This work is funded by a National Science Foundation grant to M.F.Y. and a Howard Hughes Medical Institute predoctoral fellowship to J.R.D. D.W. is funded by the National Institutes of Health (GM062932) and the Max Planck Society.