Novel as1 and as2 defects in leaf adaxial-abaxial polarity reveal the requirement for ASYMMETRIC LEAVES1 and 2and ERECTA functions in specifying leaf adaxial identity

The establishment of polarity is a fundamental theme in leaf development. Generally, leaf primordia initiate from the peripheral zone of shoot apical meristem (SAM), and establish their adaxial-abaxial, proximodistal and mediolateral axes. Genetic studies of leaf axis formation have uncovered a number of mutants that exhibit abnormalities in leaf polarity(Talbert et al., 1995; Waites and Hudson, 1995; McConnell and Barton, 1998; Schneeberger et al., 1998; Berna et al., 1999; Clarke et al., 1999; Serrano-Cartagena et al.,1999; Kerstetter et al.,2001; Semiarti et al.,2001; Ha et al.,2003). Phenotypic and molecular genetic analyses of some of these mutants have led to the identification of genes that play important roles in the leaf polarity establishment (for a review, see Bowman et al., 2002).

In Arabidopsis, the PHABULOSA (PHB) and PHAVOLUTA (PHV) genes, together with a closely related gene, REVOLUTA (REV), encode members of a homeodomain/leucine-zipper (HD-ZIP) family of proteins. Semi-dominant gain-of-function mutations in either PHB or PHV result in the transformation of abaxial leaf tissues into adaxial ones(McConnell and Barton, 1998; McConnell et al., 2001). Phenotypes in loss-of-function rev mutants could be interpreted as a partial loss of adaxial identity (Talbert et al., 1995; Otsuga et al.,2001). It was suggested that these genes are required for promoting the adaxial cell fate in lateral organs(McConnell and Barton, 1998; McConnell et al., 2001). In addition, YABBY and KANADI (KAN) genes are expressed in the abaxial face of lateral organs and specify the abaxial cell identity in Arabidopsis (Chen et al., 1999; Sawa et al.,1999a; Sawa et al.,1999b; Eshed et al.,1999; Eshed et al.,2001; Kerstetter et al.,2001). Members of the YABBY and KAN gene families are candidate abaxial-promoting factors because mutations in these genes cause abnormality in the specification of the abaxial fate(Siegfried et al., 1999; Eshed et al., 2001; Kerstetter et al., 2001; Bowman et al., 2002).

Other key regulators of leaf polarity include a group of functional homologs: PHANTASTICA (PHAN) in Antirrhinum, ROUGH SHEATH2 (RS2) in maize and ASYMMETRIC LEAVES1(AS1) in Arabidopsis(Waites and Hudson, 1995; Schneeberger et al., 1998; Serrano-Cartagena et al.,1999). PHAN, RS2 and AS1 all encode MYB-domain containing putative transcription factors, with a high degree of sequence similarity among them (Waites et al.,1998; Timmermans et al.,1999; Tsiantis et al.,1999; Byrne et al.,2000; Sun et al.,2002). In situ hybridization and immunolocalization experiments demonstrated that transcripts or proteins of members in the class 1 KNOX (knotted-like homeobox) gene family are ectopically accumulated in leaves of phan, rs2 and as1 mutants(Waites et al., 1998; Timmermans et al., 1999; Tsiantis et al., 1999; Byrne et al., 2000). These results suggest that PHAN, RS2 and AS1 act to down-regulate KNOXgenes in leaf initials, or these genes might initiate a process by which KNOX gene expression is epigenetically repressed.

Furthermore, mutations in the Arabidopsis AS2 gene, another important gene in leaf development, cause very similar phenotypes to those of as1 mutants (Serrano-Cartagena et al., 1999; Ori et al.,2000; Sun et al.,2000; Semiarti et al.,2001). In addition, as2 mutants show increased accumulation of KNOX transcripts in leaves(Semiarti et al., 2001),similar to that in as1 mutants(Byrne et al., 2000). It was proposed that the AS1 and AS2 genes function in the same regulatory pathway (Serrano-Cartagena et al., 1999; Byrne et al.,2002; Xu et al.,2002). AS2 has been cloned recently and the gene encodes a protein with a leucine-zipper motif(Iwakawa et al., 2002; Xu et al., 2002). AS2is expressed in almost all of the above ground portion of the wild-type plant except the stem (Iwakawa et al.,2002; Xu et al.,2002).

Although the isolation and characterization of the AS1 and AS2 genes have provided important insights into the mechanisms that control the establishment of polarity during leaf development, they also raised further questions. First, what is the molecular basis for AS1 and AS2 action? Do they form a complex if they function in the same regulatory pathway? Second, do AS1 and AS2 also regulate leaf polarity in the adaxial-abaxial axis, in addition to their roles in proximodistality and mediolaterality in leaves (Byrne et al., 2000; Tsiantis,2001)? Finally, are there any other genes required for leaf polarity formation in the AS1 and AS2 regulatory pathways?

To address these questions, we previously isolated and characterized new as1 and as2 alleles in the Landsberg erecta(Ler) genetic background (Sun et al., 2000; Sun et al.,2002; Xu et al.,2002). Unlike other as1 and as2 alleles in the Columbia, ER and En-D backgrounds, the alleles in the Ler background showed a novel leaf phenotype: in some rosette leaves the petiole is attached to the under surface of the leaf lamina. We referred to this structure as a lotus-leaf. Here, we further characterize the lotus-leaf defects and demonstrate that the primary AS1 and AS2 functions in the establishment of leaf polarity are the regulation of adaxial-abaxial axis. We also provide evidence that ER function acts in the AS1-AS2pathway to regulate polarity formation during leaf development. We report a physical interaction between AS1 and AS2 proteins in vitro and in yeast. Based on these results as well as the phenotypes of 35S::AS1 and 35S::AS2 transgenic plants, we propose a model of AS1, AS2and ER actions in leaf polarity formation.

Seeds of mutant brevipedicellus (bp), as1-1,as2-1 and wild-type Landsberg (Lan, with the wild-type allele for the ERECTA gene) were obtained from the Arabidopsis Biological Resource Center (ABRC). The as1-101 and as2-101 mutants are in the Ler genetic background and have been previously described(Sun et al., 2000; Sun et al., 2002; Xu et al., 2002). Plants were grown on soil according to our previous conditions(Chen et al., 2000).

The cDNA fragments encoding the entire AS1 and AS2 predicted proteins were amplified using polymerase chain reaction (PCR) and cloned into the NdeI and BamHI restriction sites of the MATCHMAKER two-hybrid vectors pGADT7 and pGBKT7 (Clontech, USA), to generate pGADT7-AS1,pGBKT7-AS1, pGADT7-AS2 and pGBKT7-AS2, respectively. The PCR primers were as follows: 5′-gccatATGAAAGAGCGTCAACGTTGG-3′ and 5′-gtggatccTTAT CAGGGGCGGTCTAATCTG-3′ (for AS1), and 5′-gccatATGGCATCTTCTTCAACAAAC-3′ and 5′-gtggatccTTATCAAGACGGATCAACAGTAC-3′ (for AS2). In each of the above primer sequences, the lowercase letters represent additional nucleotides to introduce restriction sites. All PCR fragments were verified by sequencing.

Construct combinations pGADT7-AS1/pGBKT7-AS2 and pGADT7-AS2/pGBKT7-AS1 were co-transformed into the yeast strain AH109, and transformants were selected for growth on media lacking tryptophan and leucine. The interaction between the AS1 and AS2 proteins was tested by growth of the transformants on media lacking histidine and adenine, indicating expression of the reporter genes HIS3 and ADE2. Analysis of the relative β-galactosidase activity was as described in the Yeast Handbook (Clontech, USA).

For synthesis and purification of recombinant AS1 and AS2 proteins, cDNAs containing the entire coding regions of these two proteins were amplified by PCR. The amplified AS1 cDNA was cloned into the vector pET-14b(Novagen, USA) by using the NdeI and BamHI sites to yield His-AS1, and the AS2 cDNA was cloned into the vector pGEX-4T1 (Pharmacia, USA)by using the BamHI and SalI sites to result in GST-AS2. The PCR primers for the AS1 amplification were 5′-gccatATGAAAGAGCGTCAACGTTGG-3′ and 5′-gtggatccTTATCAGGGGCGGTCTAATCTG-3′, and those for the AS2 amplification were 5′-caggatccATGGCATCTTCTTCAACAAAC-3′ and 5′-cagtcgacTTATCAAGACGGATCAACAGTAC-3′. In each of above sequences the lowercase letters represent additional nucleotides to introduce restriction sites. All constructs were verified by sequencing. Production and purification of His-AS1 and GST-AS2 fusion proteins were according to the manufactures' recommended protocols (Novagen and Pharmacia, USA). The resultant proteins were analyzed by SDS-PAGE before enzyme-linked immunosorbent assay (ELISA) experiments. ELISA was performed by coating wells of microtiter plates (Nunc., USA) with the GST-AS2, followed by addition of the His-AS1 at different concentrations to the coated wells. The retained His-AS1 was determined by incubation with a primary antibody against the His tag (Sigma, USA), at 4°C overnight. Then the second antibody, a POD-conjugated anti-mouse antibody (Sigma, USA), and the substrate 3,3′,5,5′-tetramethylbenzidine (TMB) were added. The reaction was examined by recording the absorbance at 655 nm, using a 450 Microplate Reader(Bio-Rad, USA).

For reverse transcription-polymerase chain reaction (RT-PCR), total RNA was extracted as described previously (Huang et al., 1995). After treatment with DNase (Promega, USA),complementary DNA was synthesized using a reverse transcription kit (Promega,USA). PCR reactions were performed with KNAT1 gene-specific primers(5′-TGTCAGAGTCCCATTCAC-3′ and 5′-GCAACGAGAGGTTGTTATT-3′), which span the exon3/exon5 region. PCR products were examined by separating on a 1.0% agarose gel.

The overexpression construct 35S::AS1 was constructed previously(Sun et al., 2002). For overexpression of the AS2 gene, a 0.6 kb genomic DNA containing the entire AS2 coding region was PCR-amplified from the Lerplants and sequenced. This DNA fragment was then inserted into a binary T-DNA vector pMON530 (Monsanto, USA), downstream of a 35S promoter. The constructs 35S::AS1 and 35S::AS2 were introduced into the Ler,as1-101 and as2-101 plants by Agrobacterium-mediated transformation. Ten 35S::AS1/Ler, thirty-two 35S::AS2/Ler, five 35S::AS1/as2-101 and fifteen 35S::AS2/as1-101 transgenic lines were obtained. Gene overexpression was verified by RT-PCR from the 35S::AS1/Ler and the 35S::AS2/Lertransgenic lines that were used for phenotypic analysis in this work (data not shown), and primers used in the PCR experiment were described previously(Xu et al., 2002). 35S::AS1/as2-101 and 35S::AS2/as1-101transgenic lines were verified by PCR using a forward 35S primer(5′-GCTCCTACAAATGCCATCA-3′) and reverse primers(5′-ttgaattcCATTACAAGTTACAAC-3′ for the AS1 and 5′-GTTTTCTCATCACCAAGCG-3′ for the AS2). Phenotypes of the transgenic lines were consistent among progeny from each transformation.

Fresh leaves and whole seedlings of wild-type and mutant plants were examined using a SZH10 dissecting microscope (Olympus, Japan), and photos were taken using a Nikon E995 digital camera (Nikon, Japan). Preparation of thin section specimens and scanning electron microscopy (SEM) were as described previously (Chen et al., 2000),using the first pair of rosette leaves.

Previous characterizations of the as1 mutants showed that the mutant pattern reflects a change in proximodistal and mediolateral patterning of the leaf, but not in the adaxial-abaxial axis(Byrne et al., 2000). Our more recent results demonstrated that as1 and as2 mutants in the Ler genetic background displayed a novel leaf structure: in some rosette leaves the petiole is attached to the abaxial surface of the leaf lamina, showing a lotus-leaf structure(Sun et al., 2002; Xu et al., 2002). This structure might suggest a defect in the adaxial-abaxial axis in leaves. To understand better the AS1 and AS2 functions, we characterized this lotus-leaf structure extensively. Since as1 and as2 mutants have very similar overall phenotypes, we focused our phenotypic analyses mainly on the as2-101 mutant, except where otherwise noted. In comparison with wild-type plants(Fig. 1A), all as1 and as2 alleles that we have obtained in the Ler ecotype produced the lotus-leaf structure (Fig. 1B,C, arrows), and this type of organs usually appears among the first two rosette leaves. In our growth conditions, 15-30% (depending on individual alleles) of all first two rosette leaves in as1 and as2 seedlings, were lotus leaves.

Petioles of the Ler plants have an asymmetric adaxial-abaxial axis with a flat and a slightly wider adaxial side(Fig. 1D). In the as2-101 mutant, however, each petiole showed a radially symmetric proximal portion, the length of which varied in a continuous series, depending upon leaf positions and leaf ages. Some leaves showed a radially symmetric portion at the very proximal end (Fig. 1E), while in others it was more distal(Fig. 1F). If the radially symmetric tissue reached high enough to affect the region where leaf lamina grew, the lotus-leaf was formed (Fig. 1G,H). The higher the radial portion ended, the smaller the whole leaf structure became. If the radially symmetric portion extended extremely distally, lamina development would be severely affected, resulting in either leaves with a very small lamina (Fig. 1I) or even needle-like organs without any lamina growth(Fig. 1J). We also analyzed petioles of as1-1 and as2-1 plants, these are the previously identified alleles that are in the mixed Col/Ler and ER genetic backgrounds, respectively. Although lotus-leaves appeared in as2-1 at a very low frequency (Xu et al.,2002), all first pairs of rosette leaves that we have analyzed contained a radially symmetric portion in the petiole(Fig. 1K). as1-1exhibited a less severe petiole phenotype, having neither lotus-leaves nor radially symmetric petioles. However, margins of the as1-1 petioles curled upwards (Fig. 1L),resembling the portion distal to the radially symmetric one in petioles in as1-101 (Sun et al.,2002), as2-101 (Fig. 1E,F) and as2-1 (Fig. 1K). These results indicate that the as1 and as2single mutants, regardless of genetic backgrounds, exhibited defects in leaf adaxial-abaxial axis.

The as1 and as2 mutants were identified and first characterized several decades ago (Redei,1965). However, the lotus-leaf phenotype was not reported from previous analyses of the as1 and as2 alleles(Redei, 1965; Serrano-Cartagena et al.,1999; Byrne et al.,2000; Ori et al.,2000; Semiarti et al.,2001). Previously, we reported the observation of lotus-leaves in newly isolated as1 alleles in the Ler background(Sun et al., 2002). We also compared as2 alleles in different genetic backgrounds, and found that only those in the Ler background produced lotus-leaves at relatively high frequencies (Xu et al.,2002). These results indicate that the lotus-leaf phenotype is likely to be sensitive to the genetic background. Ler carries a mutated ER gene. To determine whether the lotus-leaf morphology is associated with the er mutation, we crossed as2-101(Ler) with a wild-type Landsberg ER (Lan) plant. Since the er mutation causes distinctive morphologies from those of the ER (Lan), it is easy to score the F₂as2-101 erand as2-101 ER plants for the lotus-leaf phenotype. Our data showed that the as2-101 er plants had a much higher frequency of lotus-leaves than that in the as2-101 ER plants(Table 1), indicating that ER function is indeed involved in the leaf polarity establishment.

To identify abnormalities of adaxial-abaxial polarity in as2leaves at the cellular level, we analyzed cell patterns by transverse sectioning of as2 petioles. In a Ler rosette leaf, adaxial and abaxial surfaces of a petiole can be recognized by the pattern of their epidermal cells (Fig. 2A). The adaxial epidermal cells of a mature petiole are usually large and irregularly shaped, whereas the abaxial epidermal cells are relatively small. Between adaxial and abaxial epidermis, there are small and dense epidermal cells of the petiole margins. The adaxial and abaxial asymmetry of petioles was lost completely in lotus-leaf petioles (Fig. 2B) and the radially symmetric portion of non-lotus-leaf petioles(data not shown) in the as2-101 mutant. The overall epidermal characters of the radial petiole resembled those of abaxial epidermis in the wild type (for comparison, see Fig. 2A). Moreover, sub-epidermal cells in the lotus-leaf petioles seemed abnormal in shapes compared with those in the Ler plant: most cells in Ler petioles were irregularly shaped(Fig. 2A) while cells in the lotus-leaf petiole were arranged in an orderly fashion(Fig. 2B). In the asymmetric portion of the non-lotus-leaf petioles of the as2-101 mutant,epidermal cells on the adaxial side were also aberrant as shown in Fig. 2C. In addition to the abaxial epidermal cells, cells similar to the margin epidermis seemed to occupy the adaxial positions. The phenotypic analysis of the lotus-leaf petiole indicates that the AS2 gene plays an important role in the formation of leaf adaxial-abaxial polarity.

To further investigate AS1 and AS2 functions in the leaf polarity formation, we fused AS1 and AS2 cDNAs to the`constitutive' CAMV 35S promoter and introduced the constructs into Ler and the corresponding as1 or as2 mutant plants,respectively. We analyzed the first pair of rosette leaves, and found that 35S::AS1/Ler and 35S::AS2/Ler transgenic plants displayed dramatically different phenotypes, although the overall phenotypes of as1 and as2 mutants are very similar. Ler and as2-101 plants carrying 35S::AS2 had narrow leaves with laminae curled upwards (Fig. 3A,B). In comparison, Ler plants containing 35S::AS1 displayed a reduced plant stature with normally shaped leaves (Fig. 3C). We also analyzed lamina epidermis that was located midway up the length of the lamina and midway between the margin and the midvein of Ler and 35S::AS1/Ler plants by SEM. For 35S::AS2/Ler plants, we analyzed the central portion of the laminae, as the adaxial epidermis in this region could be viewed in a curled leaf. The Ler adaxial epidermis of leaves was characterized by an undulating surface composed of uniformly sized cells with a low density of stomata (Fig. 3D). In contrast,the Ler abaxial epidermis was characterized by a flat surface and a high density of stomata with jigsaw-puzzle-shaped cells(Fig. 3G).

On the adaxial side of laminae, epidermal patterns in Ler and the 35S::AS1/Ler transgenic plants were similar(Fig. 3D,F), although the 35S::AS1 lamina contained more stomata(Fig. 3F). Abaxial epidermal cells in Ler (Fig. 3G)and the 35S::AS1/Ler transgenic plants(Fig. 3I) were also similar in shape. In comparison, the identity of adaxial and abaxial epidermal cells on laminae of the 35S::AS2/Ler transgenic plants was altered dramatically. Abaxial-like epidermal cells appeared on part of the adaxial side of laminae of the first pair of rosette leaves(Fig. 3E), whereas the abaxial side was almost entirely covered in cells with adaxial features(Fig. 3H). The other rosette leaves also displayed adaxial-abaxial transformation, albeit weaker: only ectopic patches of adaxial and abaxial epidermal cells appeared on the abaxial and adaxial sides, respectively (data not shown). These results further support the hypothesis that the AS2 function is required for the adaxial-abaxial polarity in leaves.

To examine AS1 and AS2 functions in leaf polarity along the proximodistal axis, we further analyzed adaxial epidermal identity in the as1, as2 and 35S::AS2/Ler leaves. Fig. 4A,B shows the adaxial epidermis in the proximal part of a Ler lamina. There were two distinctive cell types: elongated cells of the midvein and the relatively uniform epidermal cells that covered most of the lamina. The equivalent region of the as2 leaf epidermis (Fig. 4D,E) contained only one type of cell that was long and narrow in shape. These cells resembled the epidermal cells on the margin of Lerpetiole (Fig. 4C, arrowhead),and were very similar to the epidermal cells on the adaxial side of the as2 petiole (Fig. 4F),consistent with the results from transverse sections(Fig. 2). Epidermal patterns in the as1 mutant were very similar to those in the as2 mutant(data not shown). In comparison, 35S::AS2/Ler petioles contained the uniformly shaped epidermal cells (white arrowhead) and elongated midvein-like cells (black arrowhead, Fig. 4G,H). This type of cell is usually positioned in the more distal region in the Ler lamina. In the more proximal portion of the petiole, epidermal cells were mosaic with a mixture of adaxial- and abaxial-type cells (Fig. 4I). This abnormal proximodistal differentiation, however, was not seen in the 35S::AS1/Ler plants (data not shown). These results indicate that the AS1 and AS2 functions are also required for promotion of cell fate along the proximodistal axis.

Interestingly, 35S::AS2/Ler transgenic plants also produced needle-like leaves amongst the first appearing rosette leaves,similar to those in as2-101 mutant in terms of whole organ structure and size (Fig. 1J; Fig. 5A,E). However, the epidermal cells of these structures are markedly different in the as2-101 mutant and 35S::AS2/Ler transgenic plants(Fig. 5A,E). Epidermal cells on most of the as2-101 needle-like structure were long and narrow(Fig. 5B), similar to those on the petiole margin in the Ler plant(Fig. 4C). However, the epidermal cells on 35S::AS2/Ler needle-like leaves looked more similar to the lamina adaxial epidermal cells(Fig. 5F). In more distal regions of the as2-101 needle-like leaves, the long and narrow epidermal cells were partially developed into lamina abaxial cells(Fig. 5C, arrowheads), or even completely abaxialized (Fig. 5D). However, in the equivalent region in 35S::AS2/Ler needle-like leaves, the surface was undulating with dense stomata, reflecting a trend to being expanded into lamina(Fig. 5G), or with three-branch trichomes (Fig. 5H). This type of trichome is usually associated with the adaxial surface of the wild-type laminae in the early-appearing rosette leaves. Therefore, the needle-like structure in 35S::AS2/Ler seedlings was an adaxialized organ. This needle-like structure was not observed in any of the 35S::AS1/Ler transgenic plants. Phenotypic analysis of the needle-like organs of the as2 mutant and 35S::AS2 transgenic plants supports the hypothesis that AS2 function is required for adaxial cell differentiation.

AS1 and AS2 down-regulate KNAT1, an Arabidopsis class I KNOTTED1-like gene, in leaves(Byrne et al., 2000; Semiarti et al., 2001). To determine whether the AS1 or AS2 gene is sufficient for this down-regulation, we analyzed inflorescence phenotypes and KNAT1expressions in inflorescence of 35S::AS1/Ler and 35S::AS2/Ler transgenic plants. The Arabidopsis brevipedicellus (bp) mutant carries a knat1 mutation,which causes altered inflorescence architecture, with reduced internode and pedicel lengths, bending at the nodes, and downward-oriented flowers and siliques (Douglas et al., 2002; Venglat et al., 2002). Therefore, if KNAT1 expression is suppressed in the transgenic plants, we expected to see bp-like phenotypes. In comparison to the Ler (Fig. 6A), as1 (Fig. 6B) and as2 (Fig. 6D) plants,the 35S::AS2/Ler transgenic plants exhibited downward-pointing siliques (Fig. 6E) with very short pedicels, very similar to those in the bp mutant (Fig. 6F). Plants overexpressing AS1 did not show such an inflorescence phenotype (Fig. 6C). RT-PCR results showed that KNAT1 expression was dramatically reduced in the inflorescence in the 35S::AS2/Ler transgenic plants, but was not affected in plants that carried 35S::AS1/Ler(Fig. 6G). This result suggests that although either AS1 or AS2 is required for the down-regulation of KNAT1 in leaves(Byrne et al., 2000; Semiarti et al., 2001), the amount or location of AS2 may be more crucial for the down-regulation in inflorescence.

The Arabidopsis as1 and as2 mutants display very similar phenotypes, which led to the hypotheses that the AS1 and AS2 function in the same regulatory pathway (Serrano-Cartagena et al., 1999) or that they may interact with each other(Byrne et al., 2002; Xu et al., 2002). To gain more direct evidence that AS1 and AS2 function together in the leaf polarity formation, we constructed 35S::AS1/as2-101 and 35S::AS2/as1-101 transgenic plants. Phenotypes including leaf shapes (Fig. 7A,B) and epidermal patterns (data not shown) of 35S::AS1/as2-101 and 35S::AS2/as1-101 plants resembled those of the as2 and as1 single mutants, respectively. The curled rosette leaves(Fig. 3A) and the downward-pointing flowers (Fig. 7C) in the 35S::AS2/Ler plants were not seen in 35S::AS2/as1-101 plants (Fig. 7A,D). Again, the increased stomata on the adaxial side in the 35S::AS1/Ler leaves (Fig. 3F) were not observed in the 35S::AS1/as2-101transgenic plants (data not shown). These results indicate that the AS1 function in the regulation of leaf polarity formation needs the presence of the AS2 function, and vice versa. Flower phenotypes of 35S::AS1/as2-101 and 35S::AS2/as1-101 plants differed from those of the corresponding as2-101 and as1-101 mutants. Briefly, as2-101 (Xu et al.,2002) and 35S::AS1/as2-101 (data not shown) had similar flower shapes, but fertility in 35S::AS1/as2-101 flowers was reduced. Phenotypes of as1-101 (Sun et al., 2002) and 35S::AS2/as1-101 flowers differed markedly. Flowers in the 35S::AS2/as1-101 plants were all sterile with shortened sepals, petals and stamens. The aberrant flower phenotypes in 35S::AS1/as2-101 and 35S::AS2/as1-101 transgenic plants indicate that AS1 and AS2 have separate functions in flower development in addition to their regulations in the leaf polarity formation.

To test the possible physical interaction between AS1 and AS2 proteins, we first carried out a yeast two-hybrid assay. Yeast cells that coexpressed the AS1 bait and the AS2 prey fusion proteins, and the AS2 bait and the AS1 prey fusion proteins both showed clear β-galactosidase activity(Fig. 8A). However, cells coexpressing AS1 or AS2 together with a vector-only control wereβ-galactosidase negative. A parallel yeast two-hybrid experiment showed that coexpression of AS1 and AS2 also promoted the expression of HIS3and ADE2 reporter genes, allowing cells to grow on media lacking tryptophan, leucine, histidine and adenine(Fig. 8B). These results demonstrate that AS1 and AS2 bind each other in yeast cells. To further confirm the physical interaction between AS1 and AS2 proteins, we performed ELISA experiments using purified His-AS1 and GST-AS2(Fig. 8C). Our results showed that the increased absorbance could be recorded only in the presence of both AS1 and AS2 proteins (Fig. 8D),indicating these two proteins can indeed associate physically.

The as1 and as2 mutants in the Ler genetic background show a higher frequency of lotus-leaf structure. In the most severe case, a needle-like organ forms in place of some rosette leaves. This structure is very similar to that in the phan mutant in Antirrhinum (Waites and Hudson,1995). A phan allele, phan-607, grown at 17°C produced almost exclusively needle-like leaves. Epidermal cells on these radialized leaves in the phan mutant are long and narrow,resembling those on the wild-type abaxial epidermis of leaves(Waites and Hudson, 1995). Epidermal cells on needle-like leaves of the as1 and as2mutants are similar to those of the phan mutants, suggesting that the needle-like leaves in as1 and as2 are also abaxialized organs. Although petioles in the less severe as1 and as2leaves can grow asymmetrically, cell specialization at the adaxial surface is aberrant. All these results plus the fact that AS2 is preferentially expressed adaxially in cotyledons in the embryonic stages(Iwakawa et al., 2002)strongly suggest that AS2 is an adaxial-promoting factor in leaves.

Of the three axes of leaves, establishment of adaxial-abaxial polarity is the primary and most essential process for leaf development. It was previously proposed that the establishment of adaxial-abaxial polarity is a requirement for proper lamina growth (Sussex,1954; Sussex,1955). as1 and as2 mutant plants and 35S::AS2/Ler transgenic plants all have needle-like leaves,however, the features of these organs are totally different. The needle-like structure in as1 and as2 is due to a reduction in adaxial differentiation, whereas that in the 35S::AS2/Ler transgenic plants shows only the adaxial epidermis. Moreover, the Arabidopsissemidominant mutant phb-1d also has needle-like structures, which were thought to be adaxialized organs(McConnell and Barton, 1998; McConnell et al., 2001). The epidermal pattern of needle-like leaves in 35S::AS2/Lertransgenic plants is very similar to that of needle-like organs in phb-1d, indicating the needle-like leaves in 35S::AS2/Ler transgenic plants may also be adaxialized organs. Needle-like leaves cannot develop further to form laminae, regardless of their adaxialized or abaxialized nature. This is consistent with the proposal of Sussex that proper establishment of adaxial-abaxial polarity is required for lamina development (Sussex,1954; Sussex,1955).

Interestingly, the lotus-leaf in as1 and as2 mutants is also very similar to the trumpet-shaped leaves in the phb-1d mutant(McConnell and Barton, 1998). However, the inside and outside cell identities in lotus-leaves and trumpet-shaped leaves is reversed (data not shown)(McConnell and Barton, 1998). Cells with adaxial identity are on the inside surface of the as1/as2lotus-leaf, while such cells are on the outside surface of the phb-1dtrumpet-shaped leaf. The analysis of leaf phenotypes in 35S::AS2/Ler transgenic plants, especially needle-like structures in the as2 mutants and the 35S::AS2/Lertransgenic plants further supports the hypothesis that the primary function of AS2 is related to the promotion of the adaxial cell differentiation. Since AS1 associates with AS2, and the as1 mutant also showed lotus-leaves and needle-like leaves (Sun et al., 2002) (data not shown), it is possible that AS1also functions as an adaxial promoting factor in leaf polarity formation. Our recent results using RT-PCR showed that expression of the PHB gene was enhanced in the 35S::AS1/Ler and 35S::AS2/Ler transgenic plants, and expression of the FILAMENTOUS FLOWER (FIL), a member in the YABBYfamily, was promoted in the as1-101 and as2-101 mutant plants (L.X., H. Li and H.H., unpublished). These results suggest that the AS1 and AS2 are genetically upstream to the PHB and FIL genes in the regulation of leaf polarity.

We previously demonstrated that lotus-leaves appeared at a much higher frequency in as1 and as2 mutants in the Lerbackground than that in any other genetic backgrounds analyzed(Sun et al., 2002; Xu et al., 2002). Although genomes from different Arabidopsis ecotypes contain polymorphisms, a major difference between Ler and other Arabidopsis strains is that Ler carries a mutated form of the ER gene. This mutation confers plants with a compact inflorescence, blunt fruits, and short petioles (Torii et al., 1996). In this work, we provide direct evidence that the higher frequency of lotus-leaves in as2 mutant was caused by the er mutation. Therefore, both AS1 and AS2 (possibly the AS1-AS2 complex), as well as ER contribute to the leaf polarity formation. The ER gene encodes a receptor protein kinase with extracellular leucine-rich repeats(Torii et al., 1996). It is widely expected that ER regulates signaling in plant development.

The Arabidopsis bp mutant carries mutated KNAT1 and ER genes. It was proposed that ER functions redundantly with KNAT1 to regulate plant architecture and stem differentiation(Douglas et al., 2002; Venglat et al., 2002). Although AS1-AS2 and ER also seem to be redundant in the promotion of the adaxial cell fate, similar to the KNAT1 and ER pair in the bp mutant, we hypothesize that AS1-AS2 and ER may play different roles in the establishment of leaf polarity. First, we have observed that as2 mutations in genetic backgrounds other than Ler also showed lotus-leaves, although at much lower frequencies (Xu et al.,2002). In addition, petioles of the first pair of rosette leaves in all as2 alleles, regardless of genetic backgrounds, contain a radially symmetric portion. Although petioles of as1-1 plants in the mixed Col/Ler background did not show even the radially symmetric portion, petioles in as1-1 and plants with the other as1 and as2 alleles all reflect a same defect. These observations indicate that the AS1 and AS2 functions, but not the ERfunction, are primarily necessary for the normal adaxial-abaxial polarity in leaves. Second, the length of the radially symmetric portion in as1 er (data not shown) and as2 er was highly variable: from fully expanded leaves to needle-like leaves. Nevertheless, as2 ER showed very few lotus-leaves and less variable portions of radially symmetric petioles, and as1 ER did not contained any radially symmetric position. These observations suggest that the ER function may reduce the sensitivity of plant cells to yet unknown internal or environmental signals for leaf development.

Arabidopsis as1 and as2 mutants have very similar leaf morphology (Redei, 1965; Serrano-Cartagena et al.,1999; Sun et al.,2000). Both mutants also show misexpression of the class 1 KNOX genes (Byrne et al.,2000; Ori et al.,2000; Semiarti et al.,2001; Byrne et al.,2002) and suppression of the LATERAL ORGAN BOUNDARIES(LOB) gene (Shuai et al.,2002). All these suggest that these two genes function in the same regulatory pathway. In this work, we provided direct genetic evidence showing a requirement of the AS1 and AS2 functions together in the leaf development: 35S::AS1/as2 and 35S::AS2/as1 transgenic plants demonstrated only the as1- or as2-like leaf phenotypes, which are markedly different from those in the corresponding 35S::AS1/Ler and 35S::AS2/Ler plants. To explore the underlying molecular mechanisms of AS1 and AS2actions in leaf development, we previously examined AS1 expression in the as2 mutant and AS2 expression in the as1 mutant to determine if these two genes are regulated by each other. There were no obvious changes in either AS1 or AS2 transcripts when one gene was expressed in the other mutant background(Xu et al., 2002), suggesting that the direct transcriptional regulation of one by the other is not likely.

In this work, we tested the possibility of protein-protein interactions between AS1 and AS2. We showed that AS1 and AS2 can indeed associate together both in vitro and in yeast cells. These results suggest that AS1 and AS2 may form a complex to regulate their downstream genes during leaf development.

This regulatory model is similar to that with products of floral organ identity genes APETALA3 (AP3) and PISTILATA(PI) in Arabidopsis. AP3 and PI can associate to form a complex, and mutation in either AP3 or PI results in very similar floral phenotypes (Jack et al.,1992; Goto and Meyerowitz,1994).

Although as1 and as2 have comparable defects in leaf development, transgenic plants carrying 35S::AS1/Ler and 35S::AS2/Ler exhibited dramatically different phenotypes,not only morphologically but also at the molecular level, such as the suppression of KNAT1 expression. One possibility is that the AS1 and AS2 proteins are not present at similar levels in wild-type plants. The AS1 protein may be more abundant than AS2, such that the increase of AS2 dosage results in the formation of more functional AS1-AS2 complexes. Another possibility is that the different phenotypes are caused by the endogenous gene expression pattern. AS1 is expressed throughout the leaf, a pattern similar to that of the 35S-driven gene expression in leaves. The same expression pattern of the AS1 gene may not generate altered phenotypes. AS2 is expressed only adaxially as reported in the embryonic cotyledons (Iwakawa et al.,2002), and therefore ectopic AS2 expression under the control of the 35S promoter may cause dramatic phenotypic changes.

It is known that both AS1 and AS2 are needed to down-regulate class 1 KNOX genes, because loss-of-function mutations in AS1 and AS2 result in the ectopic expression of KNOX genes in leaves (Byrne et al., 2000; Semiarti et al.,2001). Based on the analyses of 35S::AS1/Ler and 35S::AS2/Ler transgenic plants, only the AS2ectopic overexpression suppressed KNAT1 expression in the inflorescence and generated bp-like phenotypes. This phenomenon can also be accounted for by the less abundant AS2 dosage and (or) the strict AS2 distribution in wild-type inflorescence failing to form enough complexes to suppress KNAT1, although the AS1 appears in the same-stage inflorescence (Byrne et al.,2000).

Based on the AS1 and AS2 expression patterns(Byrne et al., 2000; Iwakawa et al., 2002) and the results in this work, we propose a model of AS1, AS2 and ERactions in the establishment of leaf polarity(Fig. 9). AS1 and AS2 can bind each other (evidence from the yeast two-hybrid assays and the in vitro protein binding experiment). The AS1-AS2 complex may efficiently suppress KNAT1 expression in leaves (KNAT1 was expressed ectopically in the as1 and as2 leaves, but was repressed in wild-type leaves). The AS1-AS2 complex can efficiently promote adaxial leaf identity(the as1 and as2 single mutants both showed defective epidermal cells on the adaxial surface, indicating that the independent AS1 and AS2 functions cannot normally promote the adaxial identity; and evidence also from 35S::AS2/as1-101 plants as they failed to reproduce the phenotypes of 35S::AS2/Lerplants, which had adaxialized leaves). The ER function is required for promoting adaxial-abaxial polarity formation in the AS1-AS2 regulatory pathway (as1 ER and as2 ER plants show much weaker adaxial-abaxial defects of leaves than as1 er and as2 erplants, respectively), however the exact involvement of ER action in this pathway remain to be elucidated.

We noted that the strongest phenotypes in the as1 and as2mutant plants and the 35S::AS2/Ler transgenic plants all appear in the earliest rosette leaves (the first pair of rosette leaves). It is possible that there might be some other regulators in Arabidopsisto promote leaf adaxial identity in addition to AS1/AS2 and PHB/PHV. These participants may have partially overlapping action domains with AS1/AS2 and PHB/PHV but play major roles in promoting leaf adaxial identity only in the late appearing leaves. Identification and characterization of new genes and elucidation of the regulatory network for all these genes will refine our views of axis formation in plants, and therefore provide new insights into the leaf development.

The authors would like to thank the Ohio State University Arabidopsis Stock Center for providing as1-1, as2-1, bp and Lan seeds, J. Mao and Y. Dou for SEM, X. Gao for thin sectioning, and W. Shen,X. Chen and H. Ma for helpful discussions and comments on the manuscript. This work was supported by grants from the Chinese Administration of Science and Technology (863 and 973), the Chinese National Scientific Foundation and the Shanghai Scientific Committee, to H.H.