| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
|
First published online March 4, 2005; 10.1105/tpc.104.029447 © 2005 American Society of Plant Biologists
The Mutant crispa Reveals Multiple Roles for PHANTASTICA in Pea Compound Leaf Development
|
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONRESULTSDISCUSSIONMETHODSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMETHODSREFERENCES |
|---|
Leaf formation is initiated at the shoot apical meristem (SAM). This process involves the establishment of a separate proximo-distal axis of growth, orientated away from the shoot axis. Once initiated, leaves usually elaborate along this axis to generate a bilaterally symmetrical structure. The different developmental patterns of leaf primordium cells adjacent to (adaxial) and away from (abaxial) the SAM is proposed to determine the medio-lateral plane of lamina growth (Waites and Hudson, 1995
). Studies in the simple-leaved species Antirrhinum majus and Arabidopsis thaliana have shown that the PHANTASTICA (PHAN) gene (Waites et al., 1998; Byrne et al., 2000; Xu et al., 2003; Qi et al., 2004) and class III HOMEODOMAIN-LEUCINE ZIPPER gene family members (McConnell et al., 2001; Emery et al., 2003) specify adaxial lamina identity, whereas members of the YABBY (Siegfried et al., 1999; Golz et al., 2004) and KANADI (Kerstetter et al., 2001; Emery et al., 2003) gene families specify abaxial identity, though it is not yet clear how the activities of these genes are coordinated. In addition to, or as part of, their roles in specifying polarity in lateral organs, PHAN and YABBY genes downregulate class 1 KNOTTED1-like homeobox (KNOX) gene expression (Tsiantis et al., 1999; Kumaran et al., 2002; Golz et al., 2004), ensuring that apical meristematic activity is excluded.
The effects of the phan mutation on lamina dorsoventral identity and blade expansion are conditional. In Antirrhinum, the phan phenotype is low temperature sensitive, manifesting as planar leaves with occasional loss of adaxial identity at 25°C and as completely abaxialized, needle-shaped leaves at 17°C (Waites and Hudson, 1995; Waites et al., 1998). The Arabidopsis phan ortholog asymmetric leaves1 (as1) (Byrne et al., 2000) has a higher frequency of radialized leaves in a Landsberg erecta genetic background (Xu et al., 2003), and this phenotype is high temperature sensitive (Qi et al., 2004). The maize (Zea mays) ortholog of phan (Timmermans et al., 1999; Tsiantis et al., 1999), rough sheath2 (rs2), develops bladeless leaves in some genetic backgrounds, but these were shown to maintain abaxial/adaxial asymmetry (Schneeberger et al., 1998).
A recent anatomical study of antisense NSPHAN Nicotiana sylvestris plants led to an alternative interpretation of radialized leaves (McHale and Koning, 2004). Adult transgenic leaves exhibited a bladeless, radialized morphology analogous to Antirrhinum phan leaves; however, the radialized leaves expressed NSPHAVOLUTA, a molecular marker for adaxial identity, in a wild-type pattern, and the radialized leaves subtended axillary meristems, which are associated with adaxial leaf identity (McConnell et al., 2001). Furthermore, juvenile leaves formed laminae with proliferating adaxial mesophyll layers and delayed palisade cell differentiation. Together, these findings led the authors to interpret the radial leaf phenotype as incomplete differentiation of cell layers rather than loss of adaxial identity (McHale and Koning, 2004), a role that was also proposed for AS1 (Sun et al., 2002). The delay in cell differentiation may be achieved through ectopic expression of BREVIPEDICELLUS (BP), which binds the promoters of cell wall biosynthetic genes and downregulates their transcription (Mele et al., 2003).
Antisense PHAN transgenic plants of another member of the Solanaceae family, Solanum lycopersicum (tomato), revealed an additional role for PHAN in species with compound leaves. Antisense LEPHAN plants exhibited a variable phenotype, ranging from fully radialized, bladeless leaves to peltate and palmate leaves, but in all cases, the wild-type, pinnate morphology of tomato compound leaves was lost (Kim et al., 2003a, 2003b). After examination of PHAN gene expression in diverse compound-leaved species, including a legume, the final leaf morphology of pinnate compound leaves was found to be correlated with an extended domain of PHAN gene expression on the adaxial side of leaf primordia (Kim et al., 2003a). Thus, there are alternative interpretations of the role of PHAN in the adaxial domain of leaves, either conferring adaxial identity (Waites and Hudson, 1995; Waites et al., 1998) or maintaining determinacy (Sun et al., 2002; McHale and Koning, 2004), and this has been found to be associated with leaflet formation in compound leaves (Kim et al., 2003a).
Defects in proximodistal elaboration are also observed in phan, rs2, and as1 mutants (Waites and Hudson, 1995; Schneeberger et al., 1998; Waites et al., 1998; Byrne et al., 2000; Ori et al., 2000; Sun et al., 2002). The earliest stage of proximodistal elaboration is at leaf initiation and at low temperatures; phan mutants fail to initiate leaves (Waites et al., 1998). Defective elaboration later in development is seen in phan leaves (Waites and Hudson, 1995) and in as1 rosette leaves (Tsukaya and Uchimiya, 1997; Byrne et al., 2000; Sun et al., 2002), which are shorter, puckered, and heart-shaped or lobed compared with the lanceolate and spatulate forms of the respective wild-type leaves. A characteristic proximodistal aberration seen in rs2 maize leaves is that proximal sheath and ligule tissues are displaced distally into the blade region of the leaf and the sheath can be shorter than that of the wild type (Schneeberger et al., 1998).
Localization of PHAN gene expression in leaf founder cells is consistent with a role in leaf initiation (Waites et al., 1998). This expression pattern is complementary to that of class 1 KNOX homologs, whose expression is restricted to meristems (Waites et al., 1998; Timmermans et al., 1999; Tsiantis et al., 1999) and whose role in maintaining meristematic activity has been shown to be at least partly mediated by repressing biosynthesis of the growth regulator gibberellin (Tanaka-Ueguchi et al., 1998; Sakamoto et al., 2001; Hay et al., 2002). Arabidopsis has four class1 KNOX genes; SHOOT MERISTEMLESS (STM), BP, KNOTTED1-LIKE in ARABIDOPSIS THALIANA2 (KNAT2), and KNAT6. It has been shown that AS1 represses BP, KNAT2, and KNAT6 because they are ectopically expressed in as1 mutant leaf primordia (Byrne et al., 2000, 2002; Ori et al., 2000; Semiarti et al., 2001), whereas STM is not ectopically expressed in as1 mutants (Byrne et al., 2000), or at least not to the same degree (Semiarti et al., 2001). STM is proposed to have a different regulatory relationship, whereby it downregulates AS1 expression in leaf initials (Byrne et al., 2000, 2002). Analysis of compound leaf formation in tomato suggests that it may have a different mechanism for leaf initiation and development. The PHAN ortholog, LEPHAN, and class 1 KNOX gene, LET6, are coexpressed in the SAM and leaves (Koltai and Bird, 2000; Kim et al., 2003b), though there is conflicting evidence (Pien et al., 2001; Reinhardt et al., 2003).
Proximodistal defects of PHAN mutants can be attributed to a failure to repress class 1 KNOX gene expression. The lobed leaves of as1 mutants (Tsukaya and Uchimiya, 1997) and ectopic stipules produced on as1 pickle (pkl) and as1 serrate (se) double mutants (Ori et al., 2000) resemble those of plants overexpressing KNOX genes (Chuck et al., 1996). By contrast, in tomato, where PHAN and KNOX genes appear to have different regulatory relationships (Koltai and Bird, 2000), the simplified leaf phenotypes of antisense LEPHAN plants (Kim et al., 2003a, 2003b) are the opposite of the supercompound leaf phenotypes of transgenic plants overexpressing class 1 KNOX genes (Hareven et al., 1996; Chen et al., 1997; Parnis et al., 1997; Janssen et al., 1998).
Expression of class 1 KNOX genes in tomato leaf primordia is thought to be important for maintaining the meristematic activity required to produce a compound leaf (Hareven et al., 1996; Chen et al., 1997; Parnis et al., 1997; Janssen et al., 1998) and is also partially mediated through gibberellin (Hay et al., 2002). Wider surveys of angiosperm species suggest that class 1 KNOX gene expression may be found commonly in leaves with complex primordia (Bharathan et al., 2002). Pisum sativum (garden pea) is the only other species in which a genetic analysis of compound leaf development has been performed. The expression pattern of pea class 1 KNOX genes follows that of simple-leaved species: transcripts are excluded from leaf founder cells and primordia, even in supercompound leaf mutant backgrounds (Hofer et al., 2001). In pea, the wild-type, pinnate leaf morphology is dependent on UNIFOLIATA (UNI), the pea ortholog of LEAFY (Hofer et al., 1997), and supercompound leaf phenotypes are correlated with prolonged UNI expression (Gourlay et al., 2000).
To determine the relationship between PHAN and class 1 KNOX genes and ascertain their influence on pea compound leaf morphology, we identified and characterized crispa (cri) (Lamm, 1949), the phan mutant in pea. Expression analysis of PHAN and PISUM SATIVUM KNOTTED1 (PSKN1), the pea ortholog of STM, shows that they are expressed in complementary domains, like simple-leaved species and unlike tomato. Phenotypic characterization of cri shows that the leaves are pinnate, unlike antisense LEPHAN transgenic tomato leaves. Features of cri in common with previously described phan mutants are that the leaves are shorter and develop ectopic laminae. We use double mutant analysis to determine that these are ectopic stipules, and we show that their formation is associated with ectopic expression of PSKN2, the pea ortholog of BP. The leaflets of cri mutants develop ectopic patches of tissue and are sometimes completely radialized. We use wax deposition mutants to demonstrate that this phenotype is due to a loss of adaxial cell identity. These observations indicate many parallels between pea compound leaf development and that of simple-leaved species and highlight fundamental differences with tomato compound leaf development.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMETHODSREFERENCES |
|---|
). Closer inspection of the mutant (see below) revealed an apparent loss of adaxial identity on leaflets and stipules, as observed on the leaves of the Antirrhinum phan mutant (Waites and Hudson, 1995) and a foreshortening of the leaf proximodistal axis, as observed on the maize phan mutant, rs2 (Timmermans et al., 1999; Tsiantis et al., 1999). Because of this phenotypic similarity, we undertook to investigate the pea ortholog of PHAN as a candidate for CRI.
A shoot cDNA library was screened at low stringency using a heterologous PHAN probe from Antirrhinum. A cDNA clone 1506 bp in size was isolated that encoded an open reading frame of 359 amino acids, with two MYB domain repeats at the 5' end, indicating it was a member of the R2R3 MYB gene subfamily (Stracke et al., 2001). Sequence similarity with other PHAN-like genes, including a characteristic Leu residue in the R3 MYB repeat, further suggested that the clone was a PHAN ortholog (Koltai et al., 2001). The maximum likelihood tree shown in Figure 1 is based on aligned MYB domain nucleotide sequences (see supplemental data online), and it supports a close relationship between the pea cDNA and other PHAN orthologs.
|
). The map location, in combination with phenotypic similarities, suggested that CRI was a PHAN ortholog.
To provide further confirmation of this, corresponding genomic DNA sequences from three independent, spontaneous, recessive cri mutants were examined in comparison to isogenic wild-type sequences. The cri-1 mutant-type line (Lamm, 1949) was found to carry a single base deletion. This would result in a frame shift encoding three extraneous residues before premature truncation at position 97, within the second MYB domain of the protein. The cri-2 mutant was found to carry an Arg212Stop transition that would truncate 41% of the protein beyond the MYB domain. The cri-3 mutation was found to be a 9-bp deletion that would result in the loss of residues Glu, Gln, and Lys and the substitution Leu329Val in a conserved region near the C terminus. The cri-1 mutation resulted in the loss of a DdeI site, relative to its wild-type allele, and the cri-2 mutation resulted in the loss of an RsaI site. Cleaved-amplified polymorphic sequence (CAPS) markers were used to show that the cri-1 and cri-2 phenotypes cosegregated with the allele lacking a restriction enzyme site in segregating populations of 32 individuals and 120 individuals, respectively (data not shown). Taken together, these results lead us to conclude that the pea PHAN ortholog corresponded to CRI.
cri Leaves Are Pinnate
Antirrhinum (Waites and Hudson, 1995; Waites et al., 1998), Arabidopsis (Byrne et al., 2000), and maize (Schneeberger et al., 1998) phan mutants and antisense-PHAN transgenic tomato (Kim et al., 2003a) and Nicotiana (McHale and Koning, 2004) plants have been described that show a range of effects of loss of PHAN function on leaf development, including leaf radialization and reduction of the leaf proximodistal axis. We wanted to compare the cri phenotype with those of other phan mutants to determine whether PHAN functions were conserved in pea compound leaves and whether there were any novel functions.
The wild-type, adult garden pea leaf is pinnate and carries pairs of lateral organs with distinct identities: stipules, leaflets, and tendrils arranged along the leaf petiole-rachis axis as shown in Figure 2A. The positions of these organs conveniently mark the leaf proximodistal axis at the base of the petiole, the proximal rachis region and the distal rachis region, respectively. The leaves of all three cri alleles remain pinnate and bear stipules, leaflets, and tendrils (Figure 2A). To confirm this quantitatively, the numbers of leaflets and tendrils were counted on cri-3 mutants and compared with the numbers on isogenic wild-type plants. Figure 3A shows there was no reduction in lateral organ number on cri-3 plants compared with the wild type. At nodes 9 to 17, the number of mutant leaflets and tendrils was greater than the number on wild-type leaves, indicating that the mutant leaf was actually more complex than the wild type during this phase of development.
|
|
To determine when cri leaf development first differed from the wild type, we used scanning electron microscopy to examine vegetative shoots. Early events in wild-type pea leaf development have been described previously (Meicenheimer et al., 1983; Gould et al., 1986; Gourlay et al., 2000). Characteristic stages are the emergence of dorsoventrally flattened stipules during plastochron 1 (p1) and emergence of the first pair of leaflets during p2. These and subsequent leaflets are initialized as radial primordia but become folded, flattened laminae during later plastochrons. Figure 4 shows that the cri-3 mutant is not distinguishable from the wild type until p5. Up to this stage, the cri-3 SAM is approximately the same size as the wild type, and the developing rachis and emerging leaflets have the same shape and rate of initiation (Figures 4A and 4B). By p5, which corresponds to node 9 of these samples, the proximal pair of wild-type leaflets are dorsoventrally flattened and folded along their length (Figures 4C and 4D). The p5 cri-3 proximal leaflet is distinctive, being deeply incised. It is seen developing into two lobes, a smaller proximal lobe and a larger distal lobe. Division of the cri-3 proximal leaflet into unequal lobes is also seen at p6 (Figure 4D).
|
; Marx, 1983). Figure 5B shows that ectopic laminae are still present on cri-1 af double mutant leaves, despite the absence of leaflets in this genotype. The ectopic laminae arise on the petiole, the proximal region of the main leaf rachis, and the proximal region of lateral rachides. The stipules reduced (st) mutation (Figure 5C) has an opposite effect to af, in that leaflets remain intact, but stipules are replaced by small, strap-like organs (Pellew and Sverdrup, 1923; Marx, 1983). Ectopic laminae are not present on the petiole-rachis axis of the cri-1 st double mutant (Figure 5D). In summary, cri-induced ectopic laminae that emerge from the petiole-rachis axis remain present in a cri af genetic background, where leaflets are absent, but are absent in a cri st background, where stipules are absent. Together, these results determined that these ectopic laminae are distally displaced stipules, as opposed to leaflets.
|
; Marx, 1983). Distal leaflets on the cri-1 tl double mutant were spaced apart, rather than bunched together, and no ectopic stipules were observed between leaflet pairs (Figure 5F). We concluded that the distal part of the cri rachis is unaffected, even in the absence of tendrils. The af tl double mutant leaf develops highly ramified rachides that terminate in miniature leaflets (Figure 5G). In combination with cri-1, the petiole, main leaf rachis, and lateral rachides develop ectopic stipules, as they do in the cri-1 af double mutant. Developing leaves of the cri-1 af tl triple mutant have such an extremely foreshortened proximodistal axis that they are almost entirely hidden between the basal stipules (Figure 5H). Leaflet expansion is delayed compared with the af tl double mutant, and eventually only the distal sections of the main and lateral rachides extend and present fully expanded leaflets. One cri-1 af tl leaf that escaped extreme foreshortening was found to have developed a petiole free from contact with the stipules (Figure 5H). This showed that petiole elongation in cri mutants is impeded by ectopic stipules developing in connection with the petiole.
cri Petioles Acquire Adaxial Identity
The proximodistal gradation in the effect of cri was examined further in transverse sections through different parts of the leaf axis: the petiole, distal rachis, and terminal tendril, as shown in Figure 6. Sections through the tendril and distal region of the rachis show the central pith and a radial arrangement of vascular bundles within the peripheral cortex, which is the same in the mutant and the wild type (Figures 6A to 6D). By contrast, sections through cri-2 petioles show that vascular bundles are larger and fewer in number than in the wild type, and their orientation is disrupted (Figures 6E and 6F).
|
cri Stipules and Leaflets Lack Adaxial Identity
Leaflets and stipules of all three cri alleles show dorsoventrality defects characteristic of simple-leaved phan mutants at low temperatures (Waites and Hudson, 1995). Ectopic patches of tissue the color of the abaxial epidermis are present on the adaxial surface of leaflets and stipules as shown in Figure 7A. The boundary between abaxial and adaxial identity around these patches is often associated with bifacial lamina outgrowths (Figure 7A). Ectopic patches form independently of the cri leaflet midrib, in contrast with the midrib-adjacent outgrowths seen in antisense NSPHAN Nicotiana leaves (McHale and Koning, 2004).
|
Further confirmation of loss of adaxial identity in cri leaflets and stipules was obtained by crosses to wax deposition mutants. Apart from their coloration (Figure 7A), the upper and lower surfaces of wild-type pea leaflets are not easily distinguishable, but as Figure 8A shows, in a wachslos (wlo1) background (Macey and Barber, 1970; Marx, 1983), leaflet adaxial epidermal identity is clearly marked by reduced wax deposition on the adaxial surface of the leaflet compared with the waxy abaxial surface. Conversely, in a vix-cerata (wb) background (Macey and Barber, 1970; Marx, 1983), leaflet abaxial epidermal identity is marked by reduced wax deposition compared with the waxy adaxial surface (Figure 8C). Both surfaces of cri-1 wlo1 double mutant leaflets had wax distributions typical of a single wlo1 mutant, except for epidermis within the boundaries of ectopic patches on the adaxial surface. This had a waxy deposition typical of an abaxial surface (Figure 8B), indicating that it had abaxial identity. Likewise, both surfaces of cri-1 wb double mutant leaflets had wax distributions typical of a single wb mutant, except for ectopic patches, which were waxless (Figure 8D), confirming their abaxial epidermis identity. These studies showed abaxial patches on the adaxial surface in two different ways, and they concur with the interpretation of a loss of adaxial identity in pea cri mutant leaflets (Waites and Hudson, 1995) rather than a failure to differentiate (Sun et al., 2002; McHale and Koning, 2004).
|
; Byrne et al., 2000; Ori et al., 2000; Semiarti et al., 2001). To determine if aspects of the cri phenotype were associated with aberrant class 1 KNOX gene expression, we determined orthology relationships between pea and Arabidopsis class 1 KNOX genes and performed RNA in situ hybridizations on pea shoots. Figure 9 shows a phylogenetic analysis (see supplemental data online), which supports the pea class 1 KNOX genes, PSKN1 and PSKN2 (Hofer et al., 2001), as the orthologs of STM and BP, respectively.
|
|
; Ori et al., 2000) is expressed in the wild-type SAM and is excluded from leaf initials (Figure 10E) in a pattern that confirms previous studies (Hofer et al., 2001). The boundary of PSKN1 exclusion from the p1 primordium coincides with the boundary delimiting CRI expression within the p1 primordium (Figures 10B and 10E); thus, these genes are expressed in nonoverlapping domains. PSKN2, like its ortholog BP (Lincoln et al., 1994; Byrne et al., 2000; Ori et al., 2000; Semiarti et al., 2001), is expressed in the peripheral zone of the SAM and at the base of wild-type leaf primordia. At leaf bases, but not in the SAM, it overlaps with CRI gene expression (Figure 10F). In summary, in wild-type shoots, CRI, PSKN1, and PSKN2 are expressed in patterns similar to those of their respective Arabidopsis orthologs (Byrne et al., 2000; Ori et al., 2000; Semiarti et al., 2001) and unlike observed expression patterns of tomato orthologs (Koltai and Bird, 2000; Kim et al., 2003b).
In the cri mutant, ectopic PSKN1 expression is not observed in p0 to p5 leaf primordia (Figure 10G), although occasional patches of expression were observed in older primordia (data not shown), suggesting that in pea, as in Arabidopsis (Byrne et al., 2000; Ori et al., 2000), the STM ortholog is not repressed by the AS1 ortholog. Examination of PSKN2 gene expression in the cri mutant showed that a negative regulatory relationship, as observed in Arabidopsis (Byrne et al., 2000; Ori et al., 2000; Semiarti et al., 2001), is conserved in pea. Ectopic PSKN2 expression is detected on the adaxial side of developing leaves, in the presumptive petiole and rachis, during p3 to p6 (Figure 10H). This pattern was confirmed in transverse sections, showing that ectopic PSKN2 expression is not detected in proximal leaflet lobes of the cri mutant, but is detected on the adaxial side of the petiole, where ectopic stipules eventually emerge (Figure 10I). Serial transverse sections through a single developing p5 leaf (Figures 10J to 10L) verified that this ectopic expression is confined to the adaxial side of the rachis, adjacent to ectopic lamina as it forms (Figure 10J) and where it is predicted to form (Figure 10K), but not in the terminal tendril, where ectopic stipules are not found (Figure 10L). At p6, ectopic PSKN2 expression in the petiole and proximal rachis is diminished, and it is absent from the expanding lamina of the ectopic stipules (Figure 10H).
The cri Mutation Is Pleiotropic
The cri mutation is pleiotropic, affecting other aspects of the shoot apart from leaf development. Adult cri-2 shoots have slightly longer internodes than the wild type between node 3 and the node of flower initiation, but mutant internodes are slightly shorter than the wild type at later nodes (data not shown). As with leaf development, the cri mutation affects flower development in several ways. The inflorescence proximodistal axis, measured as peduncle length, is reduced in the cri-2 mutant compared with the wild type (data not shown). There is no effect on organ identity or organ number, but the laminar floral organs exhibit loss of adaxial identity. These organs, sepals, petals, and carpels are generally reduced in width and distorted by ectopic laminar outgrowths. In wild-type flowers, the keel petals enclose the carpel and stamens and self-fertilization normally occurs, whereas the short, distorted keel petals of the cri mutant fail to enclose the stamens fully. Pollen viability is unaffected, but seed production per pod is lower in cri-2 mutants (1.2 ± 0.1, n = 19) compared with the wild type (2.8 ± 0.4, n = 15). Analysis of ovule numbers showed that cri-2 carpels contain fewer ovules (5.6 ± 0.6, n = 30) than the wild type (7.1 ± 0.3, n = 30). Ovule:seed ratios were 4.6 for the cri-2 mutant, compared with 2.5 for the wild type, showing that a larger proportion of mutant ovules remain unfertilized. Seed-containing cri-2 pods that reach maturity are frequently open along their sutures. This combination of factors contributes to the lower seed yields of cri mutants compared with the wild type.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMETHODSREFERENCES |
|---|
, 2002) also occur in pea. Expression analysis of wild-type and cri mutant shoot apices showed that CRI has different regulatory relationships with the STM ortholog, PSKN1, and the BP ortholog, PSKN2, as has been observed with AS1-KNOX gene relationships in Arabidopsis (Byrne et al., 2000, 2002; Ori et al., 2000; Semiarti et al., 2001). PSKN2 is expressed in a similar domain in the wild-type pea shoot as BP (Lincoln et al., 1994), and it is ectopically expressed in the petiole and rachis of cri mutants. This suggests that CRI negatively regulates PSKN2 as BP is negatively regulated by AS1 in Arabidopsis (Byrne et al., 2000, 2002; Ori et al., 2000; Semiarti et al., 2001).
PSKN1 and CRI are expressed in complementary domains in the pea SAM, as are STM and AS1 in Arabidopsis (Long et al., 1996; Byrne et al., 2000; Ori et al., 2000) and HIRZINA and PHAN in Antirrhinum (Waites et al., 1998). It has been proposed that STM downregulates AS1 in Arabidopsis because apical arrest in stm is rescued in an as1 background (Byrne et al., 2000, 2002). The absence of ectopic STM expression in the as1 mutant supports this relationship (Byrne et al., 2000, 2002); however, there is conflicting data (Semiarti et al., 2001), and patchy ectopic STM expression in as1 mutants cannot be ruled out (Ori et al., 2000). Ectopic STM expression would be predicted because it occurs in ectopic meristems that result from ectopic 35S:BP expression (Chuck et al., 1996). In Antirrhinum, HIRZINA is ectopically expressed in a phan mutant background, and it was proposed that PHAN downregulates HIRZINA (Tsiantis et al., 1999). In pea, we did not observe consistent ectopic PSKN1 expression in cri mutants by RNA in situ hybridization (Figure 10G), though occasional ectopic patches of expression were observed in older leaves (data not shown). Thus, the relationship between PSKN1 and CRI appears to be similar to that of STM and AS1 in Arabidopsis (Byrne et al., 2000, 2002; Ori et al., 2000), but because a pea stm mutant has not yet been identified, we have not been able to assess whether PSKN1 downregulates CRI.
Leaflet and Stipule Development in Compound Leaves
We detected CRI gene expression in the petiole-rachis primordium of the compound leaf as well as in stipule, leaflet, and tendril primordia. This suggests that CRI is involved in the formation of both the primary axis leaf and secondary axis laminae in pea. The relationship between CRI and class 1 KNOX genes suggests that leaf initiation in pea is similar to simple leaf initiation in Arabidopsis; however, the relationships between these genes do not hold for the initiation of secondary axis leaflets because pea class 1 KNOX genes, PSKN1 and PSKN2, are not expressed in the wild-type petiole-rachis primordium. The complementary pattern of expression between PSKN1 and CRI that distinguishes a leaf primordium from the apex does not occur within the leaf primordium; therefore, a different mechanism must be involved in the establishment of the leaflet laminae. This different mechanism must allow a role for CRI because leaflet adaxial identity is lost in cri mutant leaflets, as our wax deposition and sectioning studies showed. It is possible that other pea class 1 KNOX genes, such as a KNAT2/KNAT6 ortholog, play a more prominent role in leaflet formation, or alternatively, these functions may be performed by unrelated genes.
Stipules are flattened laminae that are conventionally described as lateral organs of the pea compound leaf; however, they are attached to the main stem and their primary vascular traces are independent of the rest of the leaf (Sachs, 1972). CRI expression is seen in stipule primordia by p3 (Figure 10D), and PSKN1 expression is downregulated in stipule anlagen (Hofer et al., 2001). Therefore, unlike leaflets, stipule initiation may be regulated by the same mechanism as petiole-rachis initiation.
Ectopic Stipule Formation
Two distinct types of ectopic lamina are present on cri mutant leaves. The first type is ectopic laminae on leaflets and stipules, associated with new surface boundaries between abaxial and adaxial identity (Figure 7A), as seen in the Antirrhinum phan mutant. The second type is ectopic stipule laminae on the petiole-rachis axis, associated with new surface boundaries between central and peripheral identity along this axis (Figures 2B and 6F). The latter boundaries arise as a result of loss of peripheral identity on the adaxial side of the petiole-rachis axis. In the developing, radial, wild-type petiole, central identity is concentric within peripheral identity, whereas in the cri petiole, we propose that central identity would be exposed at the adaxial surface. As a consequence, we propose that two new boundaries between central identity and peripheral identity would be exposed along the adaxial surface of the cri petiole and that these are coincident with the positions of ectopic stipules.
Both the position and timing of ectopic PSKN2 expression in the cri mutant are consistent with a role in ectopic stipule initiation. Ectopic PSKN2 expression appears on the adaxial side of the petiole at p3, before our first observations of ectopic stipules at p5 (Figures 4D and 10J). PSKN2 expression was not detected within the ectopic stipules, suggesting that it is not required at later stages during lamina expansion. In wild-type pea shoots, the limited domain of PSKN2 expression at the base of the leaf primordium corresponds to the position of stipules at their point of attachment to the main stem. This demarcation, and the coincidence of ectopic PSKN2 expression with ectopic stipules in cri mutants, suggests that one function of CRI in pea compound leaves is to define a boundary for stipule position at the base of the leaf by downregulation of PSKN2.
The cri phenotype of ectopic stipule formation in association with ectopic PSKN2 gene expression can be interpreted as an extension of leaf patterning normally confined to nodal regions of the stem (McHale and Koning, 2004), as seen in antiPHAN Nicotiana and maize mutants overexpressing KNOX genes. The fact that the ectopic stipules have serrate margins along their entire length suggests that they may be a distal extension of the anterior part of the stipule, which is serrate and asymmetrical, rather than the posterior part of the stipule, which is bisymmetrical and has an entire margin (Mitra, 1949; Figure 2A). Furthermore, the strongest expression of PSKN2, on the abaxial side of the leaf–stem junction, is detected at p3 to p5 (Figure 10F). This is later than the initiation of the bisymmetrical part of the stipules as flattened outgrowths, which occurs at p1 (Meicenheimer et al., 1983; Gourlay et al., 2000), but it is coincident with the development of the anterior, asymmetric region of the stipule, which occurs after p3 (Figure 4). Also, it is the anterior region of the stipule that is most affected in cri mutants. This can become reduced to a single serration, whereas the posterior region remains a narrow, flat lamina. Finally, the diminutive st stipule has a narrow posterior region but lacks an anterior region (Pellew and Sverdrup, 1923; Marx, 1983), and ectopic stipules fail to develop on cri st double mutants. These observations suggest that the boundary defined by CRI-mediated downregulation of PSKN2 may be very specific and apply to the anterior part of the stipule.
The cri st double mutant expresses PSKN2 ectopically, on the abaxial side of the petiole (data not shown), as observed in the cri single mutant, but it has a marked adaxial groove on the petiole in place of ectopic stipules. This absence of laminate ectopic stipules suggests that although the boundary for stipule position is marked by PSKN2 expression, ST is required for the production of foliaceous stipules.
The cri mutant phenotype of ectopic lamina formation on the petiole resembles the blade-on-petiole mutant of Arabidopsis, which expresses class 1 KNOX genes, including BP, ectopically in leaves; however, analysis of the cell types present in the ectopic laminae of the blade-on-petiole mutant suggested they were ectopic leaf blades rather than ectopic stipules (Ha et al., 2003), the latter being very reduced organs in Arabidopsis. Ectopic stipules are a feature of Arabidopsis as1 mutants, but these are located in the sinuses of leaf lobes, rather than on the petiole, and their formation requires additional mutations at the pkl or se loci (Ori et al., 2000) because they were not observed in as1 single mutants containing a stipule-specific ß-glucuronidase marker (Tsukaya and Uchimiya, 1997). Mutant analysis showed that BP was not required for the rounded, lobed leaf phenotypes of Arabidopsis as1 mutants because the as1 single mutant and as1 bp double mutant phenotypes were the same (Byrne et al., 2002), but it remains to be shown in as1 bp pkl or as1 bp se triple mutants whether or not ectopic stipule formation is BP dependent. The production of lobes and ectopic stipules on 35S:BP transgenic Arabidopsis leaves suggests their formation is BP directed (Chuck et al., 1996). In pea, the BP ortholog, PSKN2, is not expressed within proximal leaflet lobes of cri mutants (Figure 10I), but the distal boundary of ectopic PSKN2 expression appears to be close to the site of lobe formation on proximal leaflets, which suggests that this aspect of the mutant phenotype may be dependent on ectopic PSKN2 expression. If so, this would not be consistent with the as1 bp analysis in Arabidopsis, where the lobed as1 phenotype occurs in the absence of BP (Byrne et al., 2002)
It is possible that CRI represses PSKN2 directly because both genes are expressed in overlapping domains at the base of wild-type leaf primordia. In Arabidopsis, it is known that AS1 can homodimerize (Theodoris et al., 2003) and can bind to the leucine zipper protein AS2 (Xu et al., 2003), but direct targets have not yet been defined. In the cri mutant, where direct or indirect repression of PSKN2 fails, ectopic expression occurs within the petiole-rachis primordium, but not ubiquitously. Ectopic PSKN2 is confined to the adaxial side of the petiole and proximal region of rachis. This may explain why the distal part of the cri rachis is unaffected, and it suggests that other factors repress PSKN2 expression in abaxial and distal regions of the leaf primordium. Candidates for such factors are pea orthologs of members of the YABBY (Siegfried et al., 1999; Golz et al., 2004) and KANADI (Kerstetter et al., 2001; Emery et al., 2003) gene families that are involved in the specification of abaxial fate and are expressed abaxially in Arabidopsis. For example, in the yabby mutant filamentous flower, ectopic BP gene expression occurs on the abaxial side of developing leaves, suggesting that YABBY gene activity normally downregulates abaxial BP gene expression (Kumaran et al., 2002). Although we observed a close temporal and spatial association between ectopic stipule formation and ectopic PSKN2 expression in cri mutants, an alternative possibility is that ectopic stipules result from ectopic YABBY expression. It has been shown that a YABBY gene, GRAMINIFOLIA, is ectopically expressed in adaxial regions of phan leaves (Golz et al., 2004), and polarized YABBY expression has been shown to be associated with lamina expansion (Eshed et al., 2004).
Tendril Formation in Pea
The pea leaf produces both laminar (leaflets) and radial (tendrils) structures along the rachis. In both cases, axillary meristems at the organ-rachis junction are absent, so axillary meristems cannot be used as markers for adaxial identity within the context of the pea compound leaf. Axillary meristems at the stem–leaf junction have been cited as indicators of gained or retained adaxial identity in Arabidopsis (McConnell et al., 2001) and Nicotiana (McHale and Koning, 2004). The pea leaf petiole, which has a radialized anatomy but can be orientated by its larger abaxial vascular bundle, supports axillary meristems.
Analysis of the cri mutant has provided insight on the possible origins of pea tendrils. The resemblance of epidermal cell types present on tendrils and leaflets (Gerrath et al., 1999) and the ability to convert tendrils into leaflets by application of an inhibitor of polar auxin transport (Gould et al., 1991) suggest that tendrils represent leaflets modified by homeotic (Marx, 1983) or heterochronic (Lu et al., 1996) mutation. Thus, tendrils would be postulated to represent leaflets that have undergone an evolutionary loss of adaxial identity, and the tl mutant, which has leaflets in place of tendrils, would represent regained adaxial identity, allowing lamina formation. On the other hand, the anatomical similarity between rachis and tendrils (Gould et al., 1986) (Figures 5A and 5B) suggests that tendrils may rather represent determinate rachis structures, as opposed to modified leaflets.
The needle-shaped leaflets found in cri mutants are radial structures, as are tendrils; however, transverse sections reveal that cri needle leaflets and tendrils have different internal vascular arrangements: needle leaflets are fully radially symmetrical (Figure 6C), whereas tendrils are bisymmetrical (Figure 5A). This difference is consistent with the view that a tendril represents a determinate rachis, rather than a radialized leaflet. It can also be rationalized as consistent with homeotic and heterochronic theories, however, if tendrils represent only partially abaxialized leaflets. If this were the case, then it is unclear why the cri mutation appears to have no effect on tendrils. If tendrils partially lack adaxial identity, then the cri mutation might be expected to result in a complete loss. Perhaps abolition of a second PHAN-like gene is required for complete abaxialization of tendrils.
The lack of effect of cri on tendrils, incomplete radialization of all leaflets, and the fact that we have never observed radialized stipules may be due to redundant gene activity. Preliminary experiments have shown that the cri phenotype is more severe at higher temperatures, suggesting that the redundant activity is high temperature sensitive (data not shown). The Arabidopsis ortholog, as1, is likewise high temperature sensitive, an effect that has been shown to be mediated via the receptor kinase gene ERECTA (Qi et al., 2004). An alternative possibility is that the alleles we have studied are not completely null; however, this is unlikely because the major phenotypic features described here are in common between all three cri alleles, and a fast neutron deletion allele of cri recently identified has a very similar phenotype: pinnate leaves, ectopic stipules, some but not all radialized leaflets, and tendrils unaffected (data not shown).
Evolution of Independent Mechanisms for Compound Leaf Development
Acropetal leaflet initiation in pea means that cells comprising the stem-adjacent, leaflet-free petiole are circumscribed early, as soon as the first pair of leaflets initiate because these define the petiole-rachis junction. In tomato, basipetally initiating leaflets continually encroach on the petiole-rachis axis. Only after the last pair of leaflets initiate is the petiole defined as the remaining leaflet-free region of the axis adjacent to the stem. Mutant cri leaflets are frequently lobed, and those in proximal-most positions are often directly adjacent to the next pair, without an intervening region of rachis. This proximity suggests that additional cri leaflets may arise as a consequence of deep lobing. In support of this, an increase in leaflet number was observed at certain nodes, for example, in node 9 of cri-3 mutants, the same node at which lobed proximal leaflets were observed by scanning electron microscopy. The lobes of these leaflets were seen to be associated with a single common base (Figure 4D), and these are predicted to expand to form unequally lobed, mature leaflets.
In theory, compound leaves could be derived from deeply dissected simple leaves or from modified shoots. Evidence for their independent derivation, more than once, comes from their scattered taxonomic distribution, for example, tomato and pea lie in the euasterid 1 and eurosid 1 clades, respectively, of eudicot angiosperms (Bharathan et al., 2002). Simple leaves are characteristic of the family Solanales, with obvious exceptions being the cultivated species, tomato and potato (Solanum tuberosum). The tomato compound leaf therefore appears to be relatively recently derived from a simple-leaved progenitor (Bharathan et al., 2002). By contrast, the pea family, Fabales, including the Papilionoid subfamily to which garden pea belongs, is generally characterized by compound leaves (Bharathan et al., 2002). This is likely to be a more ancient compound leaf form, which may have been derived from a simple-leaved, or a compound-leaved, ancestor.
Further evidence of the independent derivation of tomato and pea compound leaves has come from studies showing the significant effect of class 1 KNOX genes on tomato leaf dissection (Hareven et al., 1996; Chen et al., 1997; Parnis et al., 1997; Janssen et al., 1998) and the minor effect of FALSIFLORA, the LFY ortholog (Molinero-Rosales et al., 1999), compared with the major effect of UNI (LFY) on pea leaf dissection (Hofer et al., 1997; Gourlay et al., 2000) and the presumed minor role of class 1 KNOX genes, which are not expressed in pea leaf primordia (Hofer et al., 2001). Our analysis of the cri mutant and CRI gene expression in relation to class 1 KNOX orthologs indicates that pea leaf initiation and development has closer parallels with leaf formation in Arabidopsis than with compound leaf formation in tomato. The leaves of the cri mutant remain pinnate. This demonstrates that pinnation in pea is PHAN independent, in contrast with PHAN-dependent pinnation in tomato (Kim et al., 2003a). Furthermore, cri petioles ectopically express PSKN2 and acquire adaxial lamina. Both these aspects of the mutant phenotype more closely resemble wild-type tomato than the peltate, or needle-shaped, leaves of antiLEPHAN transgenic tomato plants (Kim et al., 2003a, 2003b
