The ASYMMETRIC LEAVES2 gene of Arabidopsis thaliana regulates formation of a symmetric lamina, establishment of venation and repression of meristem-related homeobox genes in leaves

The left-right symmetry of most living things has been of general interest not only to biologists but also to researchers in other fields. In contrast to animals that have developed bodies with left-right symmetry, which is believed to be tightly linked to the capacity for mobility, plants are basically immovable and their overall body shapes exhibit conic symmetry (Gardner, 1990). It is, nonetheless, generally accepted that leaves of many angiosperms exhibit obvious but approximate left-right symmetry with the rachis as the axis (Hickey, 1973; Hickey, 1979; Sinha, 1999), even though exceptions exist, such as Begonia spp. (Lieu and Sattler, 1976), Tropaeolum (Whaley and Whaley, 1942) and many species of Urticaceae (Dengler, 1999). Regardless of the complexity of leaf shape (e.g., a simple leaf or a compound leaf), the two sides of each leaf are nearly mirror images of one another (Ogura, 1962; Hickey, 1973; Hickey, 1979).

Leaves develop from a shoot apical meristem (SAM) along three-dimensional axes (the proximodistal, transverse and adaxial-abaxial axes; Steeves and Sussex, 1989; Waites et al., 1998), as well as with left-right symmetry. Various mutants with altered leaf morphology that is related to abnormal expansion along these axes, altered adaxial and abaxial identity, and the altered overall shape of leaves have been isolated, and some of the genes responsible for the mutant phenotypes have been cloned and characterized (Hake et al., 1989; Conway and Poethig, 1997; Höfer et al., 1997; Waites et al., 1998; Kim et al., 1998; Berná et al., 1999; Serrano-Cartagena et al., 1999; Timmermans et al., 1999; Tsiantis et al., 1999; Sawa et al., 1999; Siegfried et al., 1999). However, our understanding of the way in which the nearly mirror-image architecture arises during leaf development remains at a descriptive level (see below) and the molecular and genetic basis for this phenomenon remains to be analyzed. Previous studies using Arabidopsis thaliana have focused on two aspects of leaf symmetry. It has been demonstrated that the number and the positions of the serrations on the margin of a leaf lamina are bilaterally symmetric (Tsukaya and Uchimiya, 1997). It has also been demonstrated that venation patterns in the leaf lamina of Arabidopsis are bilaterally symmetric (Candela et al., 1999). A single primary vein, the midvein, is the thickest vein and is located at the center of the leaf lamina (Hickey, 1973; Hickey, 1979; Kinsman and Pyke, 1998). During the development of the leaf from the SAM, the primary vein grows acropetally in the center of the leaf lamina. The primary vein bifurcates at the tip to form secondary veins that elongate basipetally toward the primary vein. Additional secondary veins differentiate as the lamina expands and they are connected almost symmetrically to each side of the primary vein (Hickey, 1973; Hickey, 1979; Kinsman and Pyke, 1998; Poethig, 1997; Nelson and Dengler, 1997; Candela et al., 1999).

The cells of the SAM resemble stem cells in that they have the capacity for self-regeneration and remain in an undifferentiated state, but the SAM can also generate leaf primordia from its peripheral zone (Steeves and Sussex, 1989; Howell, 1998). The SHOOT MERISTEMLESS (STM) gene, a member of the family of class 1 knox homeobox genes, is required for the development of the SAM, as well as for the maintenance of stem-cell identity throughout the life of the plant (Barton and Poethig, 1993; Long et al., 1996). The WUS gene is another type of homeobox gene that also plays an important role in the maintenance of stem-cell identity in the SAM (Laux et al., 1996; Mayer et al., 1998) and it affects heteroblastic leaf development in Arabidopsis (Hamada et al., 2000). The expression of the STM gene is down-regulated in the presumptive region of initiation of a leaf primordium in the peripheral zone of the SAM, and such down-regulation appears to be crucial for development of the leaf primordium (Long et al., 1996; Lynn et al., 1999; Long and Barton, 2000). The genome of Arabidopsis includes other members of the family of class 1 knotted-like homeobox (knox) genes, KNAT1 and KNAT2 (also known as ATK1), and the transcripts of these genes are localized primarily in the region around the SAM and the floral meristem, with down-regulation of expression in the presumptive region of initiation of a leaf primordium (Kerstetter and Poethig, 1998). The roles of KNAT1 and KNAT2 remain to be established but studies of the ectopic overexpression of these genes in Arabidopsis have shown that leaf cells can be converted from the meristematic indeterminate state to the determinate state, and back again, depending on the levels of expression of these genes and, moreover, that their levels of expression are closely related to the extent of leaf serration or formation of lobes (Sinha et al., 1993; Lincoln et al., 1994; Chuck et al., 1996; Serikawa and Zambryski, 1997). Similar observations have been made in other plant species in studies of class 1 knox genes (Hareven et al., 1996; Parnis et al., 1997; Nishimura et al., 2000).

As part of our efforts to clarify the mechanisms responsible for the development of symmetrical leaves, we have taken advantage of the asymmetric leaves2 (as2) mutant of A. thaliana, which was originally isolated by Rédei (ABRC, Ohio). Another similar mutant, asymmetric leaves1 (as1), was also identified by Rédei and it produced leaves with distorted bilateral symmetry (Rédei and Hirono, 1964; Tsukaya and Uchimiya, 1997). In the present study, we analyzed the phenotype of the as2 mutant, focusing on patterns of serration, formation of lobes and venation in leaves and leaf-like organs. We found that leaf serration in as2 was asymmetric, with generation of leaflet-like structures from petioles and malformed entire vein systems. An examination of levels of transcripts of class 1 knox genes revealed that KNAT1 and KNAT2 mRNAs, which are normally present only in the SAM, accumulated ectopically in leaves of as2 and as1 plants, and that STM mRNA accumulated in leaves of as1 plants. In addition, leaves of these mutants exhibited an elevated potential for regeneration of shoots from leaf cells in vitro without exogenous phytohormones, supporting the hypothesis that leaf cells in as2, and also in as1, have characteristics typical of an enhanced indeterminate state. Thus, it seems possible that a close correlation might exist between lamina formation, the establishment of the venation, including the midvein, and the determinate state of leaf cells.

Arabidopsis thaliana ecotypes Col-0 (CS1092) and En-2 (CS1138) and mutants as2-1 (CS3117), as2-2 (CS3118) and as1-1 (CS3374) were obtained from the Arabidopsis Biological Resource Center (Columbus, OH, USA; ABRC; Table 1). Ler-0 (NW20) and as2-4 (N463) were obtained from the Nottingham Arabidopsis Stock Centre (Nottingham, UK; NASC; Table 1). The as2-1, as2-2 and as1-1 alleles are X-ray-induced alleles that were isolated from the ER, ER and Col-1 ecotypes, respectively. The mutants are classified as ‘Form Mutants’ in the G. P. Rédei collection. The as2-1 and as1-1 alleles were mapped to chromosomes 1 and 2, respectively (Fabri and Schäffner, 1994; Machida et al., 1997). Although it was reported that the as2-1 mutant was isolated from the ER ecotype, our analysis of restriction fragment length polymorphism (RFLP) showed that the background coincided with the Col-0 background (data not shown). We outcrossed as2-1 with Col-0 three times and as1-1 with Col-0 once and used the progeny for experiments. The as2-4 (N463) allele was isolated in the En-2 ecotype by Dr Zhuchenko (NASC). Tsukaya and Uchimiya (Tsukaya and Uchimiya, 1997) designated the N463 mutation as2-2, while Rédei gave the designation as2-2 to another allele (ABRC). To avoid confusion, we changed the designation of N463 to as2-4. An M2 population of ethylmethanesulfonate-mutagenized Ler-0 seeds was purchased from Lehle Seeds (Round Rock, TX, USA), and we isolated the as2-5 allele from this population as line no. E56 (Table 1). The as2-4 and as2-5 lines were outcrossed once with En-2 and Ler-0, respectively. For analyses of phenotypes, seeds were sown on soil, and after 2 days at 4°C in darkness, plants were transferred to a regimen of white light at 3,000 lux for 16 hours and darkness for 8 hours at 22°C. Ages of plants are given in terms of numbers of days after sowing.

Plant materials were prepared for sectioning by the procedure described by Nakashima et al. (Nakashima et al., 1998). To determine numbers of branching points (NBPs), leaves from 23-day-old plants were treated as described by Hamada et al. (Hamada et al., 2000).

Leaves were pasted on a brass stage and plunged into nitrogen slush at −210°C for 30 seconds for fixation of tissue. The stage was transferred to the chamber of a scanning electron microscope (Philips, Eindhoven, Holland) and each sample was viewed at an operating voltage of 5 kV.

Seedlings of as1-1 and as2-1 mutants and wild-type (Col-0) were harvested 12 days after sowing. Total RNA from seedlings was isolated and northern blotting was performed as described previously (Banno et al., 1993). To prepare the KNAT1 probe, a 566 bp fragment, corresponding to the 5′ portion of KNAT1 cDNA (Lincoln et al., 1994) was generated by SalI and XbaI cleavage of a plasmid that contained KNAT1 cDNA and was labeled with [α-³²P]dCTP using a High Prime DNA Labeling Kit (Boehringer Mannheim Biochemica, Mannheim, Germany) according to the manufacturer’s instructions.

Details of methods used for fixation of plants, embedding in paraffin and in situ hybridization can be found at http://www.wisc.edu/genetics/CATG/barton/index.html and were described previously (Nakashima et al., 1998). Sections (8 μm thick) were cut with a microtome (ERMA Inc., Tokyo, Japan). KNAT1 cDNA was amplified by PCR with primer 1 (5′GGGTCGACATGGAAGAATACCAGC-ATGACAACAG3′) and primer 2 (5′GGGCGGCCGCTTATGGA-

CCGAGACGATAAGG-TCC3′), cloned into the SalI-NotI sites of pBluescriptII SK(−) (Stratagene, La Jolla, CA, USA) and its identity was confirmed by nucleotide sequencing. Both antisense and sense probes were prepared from the cloned KNAT1 cDNA clones.

Leaves and shoot apices were harvested 19 days after sowing, and then frozen immediately in liquid nitrogen and stored at –80°C. Poly(A)⁺ RNA was purified and the first strand of cDNA was synthesized. Sample volumes were normalized for equal amplification of DNA fragments with primers specific for α-tubulin cDNA. Then, PCR was performed under the same conditions except that the primers were specific for cDNAs of homeobox genes. For semi-quantitation of mRNAs, we examined DNA fragments that had been amplified during increasing numbers of cycles during a series of PCRs. Under the conditions that we used, the amount of PCR products increased quantitatively by a factor of two during single reactions for up to 30 cycles. However, the rate of amplification gradually decreased after 30 cycles (data not shown). The products of PCR were analyzed by agarose-gel electrophoresis and sequenced directly using one of the primers used for amplification. To distinguish products amplified from mRNAs from those generated from contaminating genomic DNA, we selected sites for the design of primer sets in two regions that are separated by introns in the cognate genes. The primer pairs were as follows: for α-tubulin, pU51 (5′-GGACAAGCTGGGATCCAGG-3′) and pU52 (5′-CGTCTCCACCTTCAGCACC-3′); for KNAT1, pU13 (5′-ATGGAAGAATACCAGCATGACAAC-3′) and pU28 (5′-GATGATCCCATATTGTCACTCTTCCC-3′); for KNAT2, pU85 (5′-GCGGCGATCACTGATCGTATC-3′) and pU86 (5′-GCGGCG-ATCACTGATCGTATC-3′); for STM, pU87 (5′-CAAAGCATGG-TGGAGGAGATGTG-3′) and pU88 (5′-GGAGAGTGGTT-CCAACAGCAC-3′); for WUS, pU31 (5′-GTGAACAAAAGTCGA-ATCAAACACACATG-3′) and pU34 (5′-GCTAGTTCAGACG-TAGCTCAAGAG-3′); and for KNAT6, pU161 (ATGTACAAT-TTCCATTCGGCCGGTG) and pU162 (TCATTCCTCGGTAA-AGAATGATCCACTAG). Details of the procedure that we used here will be sent on request.

Rosette leaves of 19-to 21-day-old plants were halved and incubated on plates of Murashige and Skoog (MS) basic medium (Onouchi et al., 1995) at 22°C under continuous white light.

Fig. 1 shows typical leaf phenotypes. In terms of overall shape, the as2 leaf often had many deep and irregularly split lobes; the leaf lamina was often plump and humped at its base, the leaf surface was wavy and plants had leaflet-like structures on petioles, which were relatively short. We also examined the morphology and growth rates of roots, hypocotyls and inflorescence stems but we found no significant differences among these organs between as2-1 and wild-type plants (data not shown).

The positions of the deep lobes on as2-1 rosette leaves corresponded to those of the serrations on wild-type rosette leaves, as judged from their positions relative to those of hydathodes on the respective rosette leaves (data not shown). As shown in Table 2, regardless of the age of plants, leaf lobes were usually observed on leaves at positions higher than the second rosette leaf. As the plants matured, the proportion of leaves with lobes at each leaf position increased. However, no lobes were observed on the first and second leaves, namely, the juvenile leaves of as2-1 plants. Even though they had no lobes, these leaves had obvious humps at the base of the leaf lamina (Fig. 1B,C), which also resulted in an asymmetric lamina (Fig. 2B). The juvenile leaves of wild-type plants had no obvious serrations (Poethig, 1997; Hamada et al., 2000). The allelic mutations as2-2 and as2-4 also generated leaf lobes at similar frequencies (80-86%), whereas as2-5 generated them less frequently (20%), indicating that as2-5 was a weak allele (Table 3).

As as2-1 plants matured, they frequently produced leaflet-like structures on the petioles of the third leaf upwards (Table 2; Fig. 1E), while younger plants did so less frequently (Fig. 1C). These structures were produced on either side of the petiole but, even as the as2-1 plants matured, their first and second rosette leaves did not generate such structures (Table 2). As shown in Table 3, the frequency of leaflet-like structures varied among as2 mutants with different alleles (50%, 60% and 14% for the as2-1, as2-2 and as2-4 alleles, respectively). No such abnormal structures were found on as2-5 leaves (Fig. 1M; Table 3). The transheterozygote as2-1/as2-4 produced leaflet-like structures at a low frequency (20%; Table 3; Fig. 1K), resembling the as2-4 homozygote and indicating that as2-4 was a weak allele.

Petioles of as2-1 and as2-2 leaves were always shorter than those of the wild type (Table 3). There were, however, no significant differences between as2-1 and the wild type in terms of the length and width of the leaf lamina and the number of palisade mesophyll cells (data not shown).

We analyzed the venation in rosette leaves of as2-1 and wild-type plants by dark-field microscopy. The conditions used allowed tracheary elements with lignin to be visualized (Telfer and Poethig, 1994). In the wild type, there was a single, distinct and maximally thick midvein in the center of each leaf lamina and a number of thinner secondary veins were connected to the midvein (Fig. 2A). The severity of the effect of the as2 mutation on venation varied among progeny, even when we considered progeny from a particular parental plant with a specific as2 allele (Fig. 2B-D). In extreme cases, no midvein was obvious, and several veins of similar thickness were evident with a proximodistal orientation (Fig. 2B,C). In less extreme cases, a midvein was present in the center of the leaf lamina but was thinner than the wild-type midvein even when fully matured rosette leaves were compared (Fig. 2A,D). In both mild and extreme cases, several secondary veins failed to connect with the midvein in the leaf lamina and sometimes they ran separately through the petiole (Fig. 2B-F). The venation in each as2 leaf lamina was bilaterally asymmetrical. Moreover, veins at the leaf margin were often insufficiently interconnected or were disconnected from one another (data not shown). The leaf lobes and leaflet-like structures all had single central veins, which were usually connected to a secondary vein in the basal region of the leaf lamina or the petiole (Fig. 2E,F, arrows). In no cases were the veins from these protrusions connected directly to the primary vein of the leaf (Fig. 2E,F), suggesting the limited developmental control of these protruding structures by the midvein.

In addition to abnormalities of the midvein and secondary veins, the patterns of venation in as2 rosette leaves were less complex than those in wild-type rosette leaves. To quantitate the complexity of leaf venation, we counted the total number of branching points (NBPs; Hamada et al., 2000) in rosette leaves of 23-day-old as2-1 and wild-type plants. Fig. 3 shows that the NBPs in cotyledons and in the first to fourth rosette leaves of as2 plants were about 40% of those in the respective wild-type organs. In the fifth rosette leaves and those at higher positions, the NBPs were less than 25% of those in the wild type. The number of NBPs in as2-1 leaves reached a plateau value in the third to fifth rosette leaves, while those in wild-type leaves increased up to the twelfth rosette leaf (data not shown). Taken together, the data suggest that the AS2 gene might be involved in the development of the entire vein system in each rosette leaf.

We investigated the earliest stages of vein development in the first rosette leaves of both wild-type and as2-1 plantlets. Fig. 4A shows the delayed development of the primary vein in as2.

In leaf primordia of most wild-type plants 7 days after sowing, primary veins were visible (Fig. 4Aa), but that was not the case in as2-1 plants, even though the as2-1 leaf primordia were normal in shape (Fig. 4Ag). The emergence of stipules at the base and trichomes at the top of the leaf lamina also occurred in as2-1 plants on day 7 (data not shown). On day 8, a primary vein appeared for the first time in the as2-1 leaf primordium (Fig. 4Ah). In the wild type, the primary vein bifurcated at its distal end on day 8, forming two strands (secondary veins) that extended basipetally (Fig. 4Ab) and connected to middle positions on the primary vein (Fig. 4Ac). By contrast, in almost all the as2-1 primordia, the primary veins bifurcated irregularly and asymmetrically (Fig. 4Ai). The secondary veins of as2-1 leaves developed with bilateral asymmetry, approached the primary vein at a more acute angle than in the wild type and, in some cases, did not connect with the primary vein in the leaf lamina (compare Fig. 4Ad,e,f with Aj,k,l). The development of new higher-ordered veins ceased in as2-1 leaves by day 15 (Fig. 4Ak,l).

We performed similar analysis of vein development in the third rosette leaves of wild-type and as2-1 plants, which often generated leaf lobes (Table 2). As shown in Fig. 4B, abnormal patterns of veins, similar to those in the first leaf, were seen throughout the development of the third leaf of as2-1. At the early stage of the development (Fig. 4Be,f,g), most of the as2-1 leaf lamina was normal in shape but the secondary veins were being developed in an asymmetric manner. As the leaf developed into the late stage, deep asymmetric lobes were generated (Fig. 4Bh). We observed similar patterns of development of venation and the lamina in almost all other rosette leaves (data not shown).

While the phenotype associated with as2-1 was apparent in the vein networks of rosette leaves, differentiation of vascular elements, such as xylem and phloem cells, was unaffected (data not shown).

To examine the structural organization of the vascular bundles in as2-1 rosette leaves, we prepared a series of transverse sections starting from the middle of first rosette leaves and continuing as far as the base of petioles of 15-day-old wild-type and mutant plants. Fig. 5 shows representative sections from both wild-type and as2-1 rosette leaves. In the wild type, there was a single vascular bundle from the transition region between the leaf lamina and the petiole toward the base of the petiole (Fig. 5A-C). In as2-1 leaves, there was a wider vascular bundle in the center, with disconnected secondary vascular bundles on both sides of the wider bundle in the transition region (Fig. 5E). Even at the base of the petiole (Fig. 5D,F), the as2-1 mutant still had a single but wider vascular bundle, which was surrounded by bundle sheath cells (Fig. 5D). Such wider bundles appeared to contain several subvascular arrays, each of which consisted of xylem and phloem and was surrounded by bundle sheath-like cells (Fig. 5F).

In the as2-1 leaf lamina, the central vascular bundle was thinner than in the wild type but several parallel bundles that were almost equivalent in thickness to the central bundle were observed with a proximodistal orientation (Fig. 5H). By contrast, the wild-type leaf lamina had a single and distinct, central vascular bundle (Fig. 5G). Our analysis of fifth leaves revealed similar organizations of veins (data not shown).

Although the cotyledons of the as2-1 mutant did not exhibit severe asymmetry, they were curled, had shorter petioles and bent downwards (data not shown). The complexity of venation in as2-1 cotyledons was lower than that in the wild type. Fig. 6 shows typical differences between the venation of wild-type and as2 cotyledons. We classified the complexity of the venation of cotyledons into three types on the basis of the extent of formation of loops due to the joining of primary and secondary veins (Fig. 6F): type I had three or four loops, which corresponded to NBPs of 5-7; type II had one or two loops, which corresponded to NBPs of 2-4; and type III had no loops, corresponding to NBPs of 1 or 2. The results of our analysis are summarized in Fig. 6F. Among wild-type cotyledons, 97% had NBPs of 5-7 and were classified as type I. By contrast, no as2-1 cotyledons were of type I: 86% of as2-1 cotyledons were type II and 14% were type III. Thus, secondary veins were less efficiently connected to the primary vein in as2-1 cotyledons, as was also the case in the foliage leaves.

We also examined the morphology and patterns of venation of cauline leaves and leaf-like reproductive organs. The as2-1 cauline leaves were severely lobed (Figs 1H, 7D). The relatively narrow sepals and the petals were unusually curled downward, but they were not strongly asymmetric (Fig. 7E,F). Fig. 7 shows typical patterns of venation of cauline leaves, sepals and petals. In all these organs, venation in as2-1 was less complex than in the wild type (Fig. 7D-F). The as2-1 sepals and petals always had open loops of veins (Fig. 7E,F) but in most of these organs in the wild type the loops were closed (Fig. 7B,C). These results showed that the development of vein networks was less extensive in the leaf-like organs of as2-1 plants than in the wild type.

The morphology of as2 and as1 leaves was similar to that of leaves of transgenic Arabidopsis that ectopically express the homeobox gene KNAT1 (see Introduction). We examined the expression of KNAT1 by northern blotting of poly(A)⁺ RNA prepared from 12-day-old wild-type Arabidopsis and as1-1 and as2-1 mutant plants. Fig. 8A shows that levels of KNAT1 transcripts in the two mutants were higher than in the wild type.

We then attempted to quantify relative levels of KNAT1 transcripts in separated rosette leaves of wild-type, as1-1 and as2-1 plants using RT-PCR. We also examined transcripts of other meristem-related homeobox genes, namely KNAT2, STM and WUS. Fig. 8B shows representative results. When primers for amplification of KNAT1 cDNA were used, products of PCR were detected in samples from all the leaves and from shoot apices of as1 and as2 mutants but not of wild-type plants. When we used primers for KNAT2 cDNA, increased levels of products were detected in both as1 and as2 (Fig. 8B). Direct sequencing of the products of PCR confirmed that they included sequences of the KNAT1 and KNAT2 genes (data not shown). Thus, transcripts of both genes had accumulated in all the leaves and the shoot apices of as1-1 and as2-1 plants but only in the shoot apices of wild-type plants. Levels of the KNAT1 and KNAT2 transcripts in as1-1 were slightly higher than those in as2-1 plants.

The Arabidopsis genome contains another knotted-like homeobox gene (AC007945). The deduced amino acid sequence of the product is 71% identical to that of the product of the KNAT2 gene and the gene belongs to the class 1 knox family. We designated this gene KNAT6. Fig. 8B shows that, in the wild type, transcripts of this gene accumulated in the shoot apices, while very low levels of transcripts were detected in rosette leaves. In the rosette leaves of as1 and as2 mutants, levels of transcripts of the KNAT6 gene increased similarly to those of the KNAT2 gene (Fig. 8B). The leaves at lower positions tended to accumulate higher levels of the transcripts of these knox genes.

Transcripts of the STM gene accumulated in the first and second rosette leaves of as1-1 plants as well as in their shoot apices, although the relative levels were lower than those of the KNAT1 gene (Fig. 8B). Levels of STM transcripts in as2-1 rosette leaves fluctuated (data not shown) and wild-type leaves did not accumulate any detectable STM transcripts. No significant increase in the accumulation of WUS transcripts was detected in as1-1 and as2-1 rosette leaves (data not shown).

In situ hybridization showed that the pattern of accumulation of KNAT1 transcripts around vegetative meristems of as2-1 was similar to that of the wild type (Fig. 8C). In both cases, transcripts were detected in the peripheral zones and basal regions of leaf primordia but not in the central zones, presumptive leaf initials and small leaf primordia. The pattern of accumulation of KNAT1 transcripts in the as1-1 meristem was also similar to that in the wild type (data not shown). Thus, mutations in the AS1 and AS2 genes did not affect transcription of KNAT1 in meristems. By contrast, we detected relatively strong hybridization signals at the bases of as2-1 leaf primordia. Furthermore, we occasionally detected weak but significant hybridization signals in some as2-1 leaf primordia and such signals were not detected in the wild type.

Overexpression of the KNAT1 gene results in the formation of ectopic shoots on rosette leaves of some transformants of Arabidopsis with an extreme phenotype (Chuck et al., 1996). However, no ectopic shoots appeared on rosette leaves of as1-1, as2-1 and as2-4 mutant plants. We examined whether as1-1, as2-1 and as2-4 mutant rosette leaves could regenerate shoots during culture in vitro. We prepared sections of rosette leaves from mutant and wild-type Col-0 plants and incubated them on MS medium without exogenous phytohormones. Table 4 shows that shoots were regenerated from 2-3% of leaf sections from as1-1, as2-1 and as2-4 plants, but not from those of wild-type plants. The shoots appeared on adaxial surfaces at the base of petioles of rosette leaves (Fig. 9A,B) and at the sinus of leaf lobes (Fig. 9C). Although small calli were produced from the cut edges of mutant leaves, no shoots formed from these calli. However, such shoots appeared to be generated directly from leaf cells. These results suggested that as1-1, as2-1 and as2-4 rosette leaves had a higher potential for regeneration of shoots in vitro than did wild-type leaves. When wild-type leaf sections were incubated under the same conditions in vitro, roots were sometimes regenerated from the small calli that were produced close to the margins of the sections (Table 4; Fig. 9D). However, roots were rarely produced from leaf sections from as1-1, as2-1 and as2-4 plants (Table 4). Thus, these mutations might reduce the rooting potential of leaf cells.

Results in the present study suggest that AS2 is involved in the establishment of the entire vein system including the prominent midvein, which is the structural axis of left-right symmetry in the leaf, as well as in the development of lamina symmetry. AS2 also appears to function in maintaining leaf cells in a developmentally determinate state and in repressing the expression of class 1 knox genes. These morphological, physiological and molecular phenomena might be related to each other but possible relationships remain to be investigated.

Distortions of the lamina, such as curlings and vein abnormalities, were commonly observed in all the leaf-like organs of as2 mutants. The formation of asymmetric lobes was not, however, seen in all these organs. The cotyledons, first and second leaves, sepals and petals of as2-1 plants had no obvious lobes (Table 2; Fig. 7). Cauline leaves (Fig. 7D) and rosette leaves at positions above the second rosette leaf (Table 2) in as2-1 plants had obvious lobes. In the wild type, the corresponding rosette leaves have obvious serrations, while other leaf-like organs are relatively simple and lack serration. For generation of obvious leaf lobes, the as2 mutation might require some additional function(s), which might be related to the potential of leaves to develop serration.

As depicted schematically in Fig. 10, the developmental patterns of three distinct veins in as2 leaves were abnormal. First, the as2 mutation delayed the formation of the primary vein at the early stage. This vein, however, did develop in the center of the leaf primordium in as2 plants, indicating that the position of the rachis was still correctly determined in the mutant leaf. Second, the secondary veins branched asymmetrically in as2 leaf primordia; they elongated basipetally, rarely joining the primary vein but running through the petiole in parallel with the midvein at the mid stage of leaf development. Thus, even mature as2 leaves failed to develop a thickened and prominent midvein. Third, the complexity of venation of higher-order veins was also reduced at the late stage. These results suggest that the AS2 gene might be involved in the establishment of the prominent midvein and the network of lateral veins.

The shape of the lamina in as2-1 leaf primordia in which the primary vein had not been observed yet was normal. In these primordia, asymmetric and abnormal venation was found. Thus, lamina formation seems to be independent of vein formation at the early stage. However, in the third leaf of as2-1, conspicuous lamina lobes and leaflet-like structures asymmetrically appeared at the late stage of leaf development when both abnormal midvein and secondary veins were already almost fully developed. It is, therefore, likely that AS2 might suppress the formation of abnormal lobes in the lamina and might support the establishment of the vein system, including the midvein in the mature leaves. Relationships between formation of the lamina and that of veins remain to be clarified, and characterization of the AS2 gene might provide clues to this relationship.

Mutations at other loci, such as LOPPED 1 (LOP1), VAN2 (also 3, 4, 5, 6 and 7), SCARFACE (SFC), MONOPTEROS (MP) and CVP1 (CVP2) also affect patterns of veins in rosette leaves (Carland and McHale, 1996; Koizumi et al., 2000; Deyholos et al., 2000; Przemerck et al., 1996; Carland et al., 1999). The lop1, van5 and van7 (emb30-7), sfc and mp mutants also have defects, in terms of thickness and shape, in midveins, as well as in lateral veins of leaves, while cvp1 and cvp2 have defects primarily in lateral veins.

Among the mutations mentioned above, lop1 is associated with the severest disruption in development of the midvein, which include bifurcation and failure to produce the normal series of lateral veins (Carland and McHale, 1996). The lop1 leaf lamina is asymmetric but there are no leaf lobes. Thus, LOP1 appears to function differently from AS2 although the lop1 mutation affects both lamina symmetry and midvein formation. The van5 and emb30-7 mutations produce midveins that are thicker than the wild-type midvein (Koizumi et al., 2000). These mutations also cause alterations in overall morphology of rosette leaf lamina, but this phenotype is again different from that of as2 which has the thinner midvein. The lamina of sfc and mp leaves, which have thinner midveins and fragmented lateral veins, appears to be symmetric (Deyholos et al., 2000; Przemerck et al., 1996). Phenotypes of these mutants suggest that vein disruption might not be sufficient to generate asymmetric lobes in laminae. The formation of an asymmetric leaf lamina appears to be uniquely associated with as2 and as1 mutations.

The presence in as2 leaves of multiple secondary veins, unconnected to the primary vein, is reminiscent of the multiple strands in the center of the leaf lamina and petiole in the pin1 mutant (Mattsson et al., 1999), in which the capacity for polar transport of auxin is reduced (Okada et al., 1991). The vascular pattern in as2 leaves was similar only to the proximal portion of pin1 leaves. Thus, the state of cells in the proximal portion of as2 leaves might be similar to that in pin1 leaves. However, the polar transport of auxin in as2 plants was not different from that in wild-type plants (our unpublished data).

The as1-1 and as2-1 mutations both caused the accumulation of transcripts of the KNAT1, KNAT2 and KNAT6 genes in rosette leaves (Fig. 8). Very recently, other groups have also reported that KNAT1 and KNAT2 transcripts accumulate in the rosette leaves of as1 (Byrne et al., 2000) and as2 (Ori et al., 2000) mutants. Abnormalities in as2 leaves might be related to the ectopic expression of these homeobox genes. In an earlier study, overexpression of KNAT1 revealed that the KNAT genes are responsible for the formation of leaf lobes, the enlargement of secondary veins and the regeneration of ectopic shoots on leaves, and, furthermore, that a relationship exists between the extent of development of leaf lobes and the efficiency of regeneration of ectopic shoots on leaves (Chuck et al., 1996). Thus, the similar morphological changes that we observed in as2 in the present study might be explained by the ectopic expression of these homeobox genes in the mutant leaves (Fig. 8). The STM gene was also expressed ectopically in the lower leaves of as1-1 plants (Fig. 8), and the STM gene might contribute to the mutant phenotype of as1-1 leaves. Since these homeobox genes belong to the class 1 knox family and no other member is known in Arabidopsis, expression of all the members of this family is probably regulated by a similar mechanism involving AS1 and AS2.

Hareven et al. (1996) reported that Tkn2, a tomato homologue of kn1, is expressed in leaf primordia as well as the SAM of wild-type tomato plants which produce compound leaves consisting of many leaflets. They propose that such expression in leaf primordia is related to the formation of the compound leaves in tomato. The overall shape of a tomato leaf is, however, roughly symmetrical. Formation of asymmetric lamina lobes and leaflet-like structures in as2 plants might require ectopic expression of the multiple members of the class 1 knox family and/ or by functions of other unknown factors.

The AS1 and AS2 genes apparently repress the expression of the KNAT1, KNAT2 and KNAT6 genes by inducing reductions in the levels of their transcripts, and at least three plausible models can be proposed for the mode of action of wild-type AS1 and AS2 genes. These genes might repress the rates of transcription of these homeobox genes; they might be involved in the degradation of the transcripts; or one of them might be involved in the former process and either might be involved in the latter. It remains unknown whether such repression is directly or indirectly attributable to the encoded AS1 and AS2 proteins. The as1-1 and as2-1 mutations affected levels of transcription of the STM gene differently. Low levels of STM transcripts accumulated in the first and second leaves of as1-1 plants (Fig. 8). The accumulation of STM transcripts in as2-1 leaves was not reproducible (our unpublished data). Thus, AS1 represses the expression of STM in leaves at least to some extent, but the role of AS2 is still unclear.

With regard to repression of class 1 knox genes, the PHANTASTICA (PHAN) gene of Antirrhinum majus and the ROUGH SHEATH2 (RS2) gene of Zea mays, which encode plant homologues of the myb protein (Waites et al., 1998), have been shown to have such a repressive function (Schneeberger et al., 1998; Timmerman et al., 1999; Tsiantis et al., 1999). Recently, the AS1 gene was shown to encode a domain that is similar to the myb repeat (Byrne et al., 2000). This result suggests that a similar mechanism is involved in regulation of the class 1 knox gene family in Arabidopsis.

Culture of sections of Arabidopsis leaves in vitro revealed that the ability to regenerate shoots from as1 and as2 leaves was higher and the ability to regenerate roots was lower than in wild-type leaves (Table 4 and Fig. 9). Therefore, it seems likely that the cells in the growing leaves of the mutant plants might have an enhanced tendency to shift at random to a more indeterminate state, which might resemble that of cells in the SAM. Since ectopic overexpression of the KNAT1 gene in Arabidopsis plants induces shoot meristems on leaves (Chuck et al., 1996), such an indeterminate state of leaf cells in these mutants might be due to the ectopic expression of the KNAT1, KNAT2 and KNAT6 genes in these leaves (Fig. 8).

For the formation of a bilaterally symmetric leaf, the division and expansion of cells must be coordinated on both sides of the midvein of the leaf lamina. The random regeneration of meristematic cells in mutant leaves might interfere with the necessary coordinated cell division and expansion, with eventual formation of asymmetric leaves with irregular curls, lobes and leaflet-like structures. Therefore, it might be critical for leaf cells to maintain a determinate state if leaves with bilateral symmetry are to be generated.

As described above, the establishment of the vein system, including the midvein, appears to require the wild-type AS2 gene. It will be of interest to examine whether a relationship exists between establishment of the vein system and the maintenance of the determinate state of leaf cells. There are at least three plausible mechanisms by which AS2 might regulate these two phenomena. AS2 might regulate these events independently. Alternatively, it might act first to induce the determinate state of leaf cells, which in turn might establish the vein system, or vice versa. These possibilities should be investigated in future by functional analysis of the corresponding genes.

The authors are grateful to Dr Hidemi Kitano (Dept. of Agriculture, Nagoya University) for helpful discussions and suggestions related to cytological analysis. The authors also thank Professor Fumio Osawa (Aichi Institute of Technology, Japan) for his encouragement and support and Mr. Fumiaki Ogasawara for his technical assistance. This work was supported in part by grants from the Research for the Future Program of the Japan Society for the Promotion of Science (JSPS-RFTF97L00601 and JSPS-RFTF00L01603), by Grants-in-Aid for Scientific Research on Priority Areas (no. 10182101) and for General Scientific Research (no. 12640598) from the Ministry of Education, Science and Culture and Sports of Japan.