The transcription factor AtGRF5 and the transcription coactivator AN3 regulate cell proliferation in leaf primordia of Arabidopsis thaliana


  • Gorou Horiguchi,

    Corresponding author
    1. National Institute for Basic Biology/Okazaki Institute for Integrative Bioscience, Okazaki, Aichi 444-8585, Japan,
    2. Graduate University for Advanced Studies, Hayama, Kanagawa 240-0193, Japan, and
      (fax +81 564 55 7512; e-mail
    Search for more papers by this author
  • Gyung-Tae Kim,

    1. National Institute for Basic Biology/Okazaki Institute for Integrative Bioscience, Okazaki, Aichi 444-8585, Japan,
    Search for more papers by this author
    • Present address: Faculty of Plant Biotechnology, Dong-A University, Hadan 2-dong 840, Saha-gu, Busan 604-714, Korea.

  • Hirokazu Tsukaya

    1. National Institute for Basic Biology/Okazaki Institute for Integrative Bioscience, Okazaki, Aichi 444-8585, Japan,
    2. Graduate University for Advanced Studies, Hayama, Kanagawa 240-0193, Japan, and
    3. Graduate School of Science, Kyoto University, Kyoto 606-8502, Japan
    Search for more papers by this author

(fax +81 564 55 7512; e-mail


The development of the flat morphology of leaf blades is dependent on the control of cell proliferation as well as cell expansion. Each process has a polarity with respect to the longitudinal and transverse axes of the leaf blade. However, only a few regulatory components of these processes have been identified to date. We have characterized two genes from Arabidopsis thaliana: ANGUSTIFOLIA3 (AN3), which encodes a homolog of the human transcription coactivator SYT, and GROWTH-REGULATING FACTOR5 (AtGRF5), which encodes a putative transcription factor. AN3 is identical to GRF-INTERACTING FACTOR1 (AtGIF1). The an3 and atgrf5 mutants exhibit narrow-leaf phenotypes due to decreases in cell number. Conversely, cell proliferation in leaf primordia is enhanced and leaves grow larger than normal when AN3 or AtGRF5 is overexpressed. Both genes are expressed in leaf primordia, and in the yeast two-hybrid assay, the gene products were found to interact with each other through their N-terminal domains. These results suggest that AN3 and AtGRF5 act together and are required for the development of appropriate leaf size and shape through the promotion and/or maintenance of cell proliferation activity in leaf primordia.


The extent and direction of growth have a major impact on the determination of organ size and shape. The flat morphology of plant leaves is established by the elaboration of these two processes (Tsukaya, 2002a). The leaf blade has three axes, the proximo-distal (longitudinal), central-lateral (transverse), and adaxial–abaxial axes (Bowman et al., 2002). Polarized cell proliferation and cell expansion along these axes influence the final size and shape of a leaf, but these processes are not well understood. To elucidate these axis-dependent processes of leaf development at the cellular and molecular levels, we isolated and characterized a series of mutants of Arabidopsis thaliana, each containing a specific defect in leaf expansion along either the longitudinal or transverse axis. A comparison of the leaf ontogenies of the wild type, the narrow-leaf mutant angustifolia (an), and the short-leaf mutant rotundifolia3 (rot3) revealed that polar cell expansion along the longitudinal axis is controlled independently of that along the transverse axis (Tsuge et al., 1996). The narrow-cell morphology and reduced trichome branch number of the an mutant correlate with an altered arrangement of cortical microtubules in leaf mesophyll cells, epidermal cells, and trichomes (Folkers et al., 2002; Kim et al., 2002). AN encodes a homolog of C-terminal binding protein/brefeldin A ADP-ribosylated substrate (CtBP/BARS), which functions as a transcriptional corepressor and as a membrane-fissioning protein in Golgi stacks in animals; its function in plants, however, is not known (Folkers et al., 2002; Kim et al., 2002). ROT3 encodes a member of the cytochrome P450 family, CYP90C1, and we have proposed that ROT3 and its closest member, CYP90D1, are involved in the late steps of brassinosteroid biosynthesis (Kim et al., 1998b, 1999, 2005).

On the other hand, our recent identification of the ROT4 gene demonstrated that cell proliferation in leaf primordia is also controlled in an axis-dependent manner (Narita et al., 2004). ROT4 is a member of a novel gene family, the ROTFour Like/DEVIL (RTFL/DVL) family, found in seed plants (Narita et al., 2004; Wen et al., 2004). ROT4 encodes a novel small peptide of unknown function that localizes to the plasma membrane, and its overexpression causes a specific reduction in cell number along the longitudinal axis (Narita et al., 2004). Furthermore, there are several narrow-leaf mutants such as curly leaf (clf), deformed roots and leaves1 (drl1), grf-interacting factor1 (atgif1), pointed first leaf2 (pfl2), and struwwelpeter (swp) associated with reduced cell numbers (Autran et al., 2002; Ito et al., 2000; Kim and Kende, 2004; Kim et al., 1998a; Nelissen et al., 2003). These earlier investigations have provided a blueprint for the genetic pathways of the two-dimensional growth of leaf blades at the level of cell proliferation as well as cell expansion.

Despite this progress our knowledge of the mechanisms of polar growth remains fragmented. One of the next challenging issues to be addressed is the identification of the genetic and molecular components involved in each pathway. In this paper, we show that a classical narrow-leaf mutant, angustifolia3 (an3), has a mutation at the AtGIF1 locus. AtGIF1 interacts with AtGRF1 and AtGRF2 in yeast and in vitro, but none of the single atgrf mutants characterized to date shows obvious abnormal phenotypes in leaf development, probably owing to the extensive functional redundancy of this family (Kim and Kende, 2004; Kim et al., 2003). Nevertheless, our genetic analysis indicates that AtGRF5 has a greater role for leaf cell proliferation than other characterized members of the AtGRF family. Our data also indicate that AtGRF5 and AN3 interact with each other and participate in the positive control of leaf cell proliferation, in particular in the promotion of lateral expansion of the leaf blade.


The an3 mutant phenotypes and cloning of AN3

an3 is a leaf-shape mutant originally isolated by Relichova (1976); its stock number at the Arabidopsis Biological Resource Center (ABRC) is CS241. CS241 is referred to here as an3-1. In this study, we identified three additional alleles, an3-2, an3-3, and an3-4, all of which exhibit similar narrow-leaf phenotypes (Figure 1a).

Figure 1.

an3 leaves and petals are narrower than those of wild type.
(a) Shoots of 27-day-old wild type and an3 plants grown in 16L/8D condition. Bar, 1 cm.
(b) Leaves of wild-type (top) and an3-4 (bottom) plants. The leaves, including the cotyledons, are arranged in order of increasing age from left to right. Bar, 1 cm.
(c, d) Flowers of wild type (top) and an3-4 (bottom). Bar, 1 mm.

The most striking feature of the an3 mutation is the narrow-leaf phenotype shown in Figure 1 and Table 1. The morphological defects of an3 are specific to leaves and floral organs. The lengths of roots of 11-day-old seedlings of an3-4 (6.5 ± 0.7 cm) were similar to those of the wild type (6.8 ± 0.6 cm; n = 11). The width and length of the leaf blades and the length of petioles of an3-4 are all reduced to different extents, but the most prominent effect is the reduced leaf width (Figure 1b, Table 1). The width and length of the first leaf blades were reduced by approximately 37 and 16%, respectively, compared with the wild-type values. As a result, the leaf index (the ratio of the leaf length to the leaf width) was changed from 1.12 ± 0.05 in the wild type to 1.49 ± 0.08 in an3-4 (Table 1). The width of the an3-4 petals was also affected (Figure 1c). A difference was also observed in leaf numbers, with 24-day-old wild-type and an3-4 plants having 13.5 ± 0.9 and 18.1 ± 1.6 (n = 6) leaves, respectively (Figure 1b). The flowering times of the wild type and an3-4 were similar (25.8 ± 1.3 and 26.8 ± 1.3 days, respectively, under 16L/8D condition; n = 34), suggesting that an3-4 produces leaves at a more rapid rate than does the wild type.

Table 1.  Effects of an3, atgf5, and atgrf9 mutations on leaf development
GenotypeLeaf area (mm2)Petiole length (mm)Leaf length (mm)Leaf width (mm)Leaf indexbCell no. in longitudinal axisCell no. in transverse axisCell area (μm2)Cell number
  1. Fully expanded first leaves are characterized (n = 6). Data are mean ± SD.

  2. aSignificantly different from the wild type (P < 0.05, Student's t-test).

  3. bRatio of the length to the width of leaf blade.

Wild type50.0 ± 3.58.2 ± 0.78.5 ± 0.47.6 ± 0.31.12 ± 0.05163 ± 10162 ± 53248 ± 24612 992 ± 1203
an3-425.8 ± 2.4a7.4 ± 0.37.2 ± 0.4a4.8 ± 0.2a1.49 ± 0.08a99 ± 7a70 ± 7a5441 ± 397a3684 ± 408a
grf5-144.6 ± 3.3a8.6 ± 0.68.5 ± 0.37.0 ± 0.3a1.22 ± 0.01a153 ± 7a137 ± 8a3559 ± 28811 116 ± 1058a
grf9-153.0 ± 3.19.0 ± 0.78.8 ± 0.37.9 ± 0.3a1.11 ± 0.03161 ± 10155 ± 83267 ± 27713 480 ± 1143

To understand the role of AN3 at the molecular level, we isolated AN3 using a map-based approach (Figure 2). Fine-scale genetic mapping indicated that the AN3 locus resides between At5g28635 and At5g28643, near the centromere of chromosome 5. Large deletions, identified by PCR analysis, were found in the candidate region encompassing At5g28640 in the an3-1, an3-3, and an3-4 genomic DNAs (Figure 2a,b), and sequencing of this region in an3-2 identified a six-base deletion in the second exon of At5g28640 (Figure 2b,c). When At5g28640 was targeted by RNA interference (RNAi) in the wild-type background, the resulting transformants displayed a narrow-leaf phenotype similar to an3 (Figure 2d,e). In contrast, the overexpression of At5g28460 under the control of a cauliflower mosaic virus (CaMV) 35S promoter in an3-1 rescued the narrow-leaf phenotype (Figure 2d,e), demonstrating that At5g28640 corresponds to AN3. In addition to these results, AN3 was found to be identical to AtGIF1, whose loss-of-function mutant displays a phenotype similar to that of an3 (Kim and Kende, 2004). In this paper, we used the genetic nomenclature AN3 instead of AtGIF1 because of the older historical origin of an3-1.

Figure 2.

Map-based cloning of AN3.
(a) A physical map around the AN3 locus. Exons are indicated by filled objects. Arrows indicate the primers used to detect deletions. An arrowhead indicates the position of the an3-2 mutation.
(b) an3-1, an3-3 and an3-4 has deletions that span the AN3 locus. PCR analyses were carried out using wild type (WT), an3-1 (3-1), an3-3 (3-3) and an3-4 (3-4) genomic DNAs with primer pairs that amplify the 2 kb-promoter plus transcribed region (a–e), 600 bp-5′ upstream region (b–c) and 600 bp-downstream transcribed region (d–e), respectively. Control PCR was carried out using a genetic marker KLPNHC on chromosome 5.
(c) Multiple alignment of the SNH domains.
(d) RT-PCR analyses of the AN3 and TUBULINβ4 (TUB4) mRNAs. Lane 1, wild type; lane 2, an3-1; lane 3, an3-1 transformed with the 35S::AN3 cDNA; and lane 4, RNAi-AN3 in a wild-type background.
(e) Seedlings of wild type, an3-1, and an3-1 transformed with the 35S::AN3 cDNA or RNAi-AN3 in a wild-type background. Bar, 1 cm. Arrowheads indicate the first and second leaves.

AN3 encodes a homolog of the human transcription coactivator, synovial sarcoma translocation protein (SYT) (Clark et al., 1994; Crew et al., 1995; de Leeuw et al., 1995; Thaete et al., 1999) and belongs to a small gene family in the Arabidopsis genome (Figure 2c). Although the biological role of SYT is unclear, chromosomal translocations of the SYT locus to the SSX locus are frequently found in synovial sarcomas (Clark et al., 1994). The chimeric SYT-SSX protein may possess altered regulatory functions that could trigger tumor development (de Bruijn et al., 2001; Kato et al., 2002; Nagai et al., 2001). SYT contains a conserved N-terminal domain called the SYT N-terminal homology (SNH) domain, which participates in protein–protein interactions (de Bruijn et al., 2001; Eid et al., 2000; Kato et al., 2002; Nagai et al., 2001; Thaete et al., 1999). This domain also appears to be functionally important for AN3, as the an3-2 mutation eliminates a conserved amino acid residue in the SNH domain (Figure 2c). The C-terminal region of SYT has a transcription activation function and is enriched in Gln, Gly, Pro, and Thr residues (Thaete et al., 1999). The clustering of Gln residues is frequently observed in transcription activators. The corresponding region of AN3 is also enriched in the above amino acid residues, except that Ser is more abundant than Tyr (Figure 2c), and the protein consistently shows transactivation activity when expressed in yeast (Kim and Kende, 2004).

Expression of AN3 and the AtGRFs in leaf primordia

AtGIF1 has been identified by yeast two-hybrid screening using a truncated version of AtGRF1 as bait (Kim and Kende, 2004). In addition, AtGRF1 and AtGRF3 participate redundantly in the control of cell number, as mutations in these genes enhance the atgif1 phenotype such that the cell number in the leaf blade is further reduced (Kim and Kende, 2004). Northern analysis has indicated that the expression of all nine AtGRFs as well as AtGIF1 is associated with young, growing tissues (Kim and Kende, 2004; Kim et al., 2003), suggesting that other AtGRFs may also have a role in cell proliferation in leaf primordia. If this is the case, it is possible that AN3 and certain members of the AtGRF family share a similar expression pattern. To test this possibility, we determined the gene expression patterns of these genes using a promoter::β-glucuronidase (GUS) approach (Figure 3). AN3 and all of the AtGRF genes except AtGRF8 were expressed at detectable levels in the shoot tip, although with somewhat different spatial expression patterns (data not shown). AN3 was strongly expressed in the basal region of leaf primordia of 5-day-old plants (Figure 3a), the root tips, and floral buds but only weakly in the shoot apical meristem; AN3 expression was undetectable in mature leaves (data not shown).

Figure 3.

Expression of AN3 and the AtGRFs in the first leaf primordia.
Four- to 5-day-old transgenic plants harboring promoter::GUS constructs for AN3 (a), AtGRF1 (b), AtGRF2 (c), AtGRF3 (d), AtGRF4 (e), AtGRF5 (f), AtGRF6 (g), AtGRF7 (h), or AtGRF9 (i) were subjected to histochemical staining for GUS activity. Bar, 100 μm.

Closer examination of the first leaves of 4- to 5-day-old plants showed that AtGRF5 and AtGRF9 were expressed uniformly in the primordium (Figure 3f,i) but were undetectable in mature leaves (data not shown). AtGRF7 was strongly expressed in cells surrounding the vasculature (Figure 3h). The expression domains of AtGRF4 and AtGRF6 partially overlapped with the AN3 expression domain, mainly near the midvein (Figure 3e,g). However, AtGRF1, AtGRF2, and AtGRF3 were expressed in regions where AN3 expression was observed at a low or undetectable level. These results support the idea that some of the AtGRFs, especially AtGRF5 and AtGRF9, are potent interacting partners of AN3.

Our results strongly indicate that there are substantial differences in the promoter activities of the AtGRF family members, but it should be noted that promoter::GUS analysis does not always reflect the authentic expression pattern of the gene of interest. Therefore, we examined the levels of AN3, AtGRF5, and AtGRF9 transcripts in the upper and lower halves of leaf primordia by reverse transcription (RT)-PCR. As controls for cell proliferation activity, CYCB1;2 and CYCD3;1 were also included in this analysis (Figure 4). We tried this experiment using leaf primordia from day 5 or earlier, but their surgical dissection into two parts was technically difficult. Thus, we used leaf primordial from day 6. First, we checked the expression pattern of the AN3, AtGRF5, and AtGRF9 genes as well as the CYCB1;2 gene by using day 6 promoter::GUS plants (Figure 4a). At this stage, GUS activities in AN3, AtGRF5, and AtGRF9 reporter plants were clearly restricted to the lower half of the leaf primordium. In contrast, in the CYCB1;2 reporter plants, GUS-positive spots, which represent cells at the G2/M phase of the cell cycle, were observed throughout the leaf primordia with a mild longitudinal gradient (Figure 4a). Consistent with this analysis the RT-PCR analysis showed that AN3, AtGRF5, and AtGRF9 are expressed at stronger levels in the lower half than in the upper half of the primordium (Figure 4b). These sharp differences were not observed for the CYCB1;2 and CYCD3;1 transcripts (Figure 4b), suggesting that there is spatial specificity in the expression of AN3, AtGRF5, and AtGRF9 in the populations of proliferating leaf cells.

Figure 4.

RT-PCR analysis of AN3 and AtGRFs in upper and lower halves of the first leaf primordia.
(a) GUS histochemical staining of day 6 first leaf primordia in AN3, AtGRF5, AtGRF9 and CYCB1;2 promoter::GUS plants. Bar, 200 μm.
(b) RT-PCR analysis of CYCD3;1, CYCB1;2, TUB4, AN3, AtGRF5, and AtGRF9 on day 6 leaf primordia. Leaf primordia are surgically dissected into upper and lower halves. Then total RNAs were extracted from each tissue sample.

AN3 interacts with AtGRFs in the yeast two-hybrid system

To test the interaction between AN3 and the AtGRFs, we chose AtGRF5 and AtGRF9. AtGRF4 was also tested, although its expression pattern differed from that of AN3 (Figure 5). As expected, AN3 interacted strongly with AtGRF5 and AtGRF9, but only weakly with AtGRF4 (Figure 5c). The homodimerization of AN3 was examined, as SYT self-interacts at its C-terminal region (Perani et al., 2003), but no evidence was found for homodimerization (Figure 5c).

Figure 5.

Interaction of AN3 with the AtGRF proteins in the yeast two-hybrid system.
(a) Multiple alignment of conserved domains in the AtGRF family, hBRM, and BRG1.
(b) Multiple alignment of zinc-finger-like domains in the AtGRFs. An atypical C3H finger is indicated by asterisks.
(c) Interaction of AN3 with AtGRF4, AtGRF5, and AtGRF9. The interaction of a fusion of AN3 and the DNA-binding domain (DB) of GAL4 with the GAL4 transactivation domain (AD) alone as well as AD fused with AN3, AtGRF4, AtGRF5, or AtGRF9 were examined.
(d) The an3-2 protein does not interact with AtGRF5. The interaction of DB:AtGRF5 and AD fused with either AN3 and an3-2 was examined.
(e) The SNH domain of AN3 and the N-terminal region of AtGRF5 are important for the interaction. The interaction of DB:AtGRF5 and AD fused with the SNH domain or the AN3 lacking the SHN domain was examined (left panels). The interaction of DB:AN3 with AD fused with C- or N-terminally truncated AtGRF5 was also examined (right panels).

SYT interacts with several factors related to transcription control; these include human BRAHMA (hBRM) and BRAHMA HOMOLOG1 (BRG1), which are subunits of the SWI/SNF chromatin remodeling complex, as well as the AF10 transcription factor (de Bruijn et al., 2001; Kato et al., 2002; Nagai et al., 2001; Thaete et al., 1999). The SNH domain of SYT interacts with the N-terminal regions of hBRM and BRG1. Interestingly, the N-terminal sequence of hBRM displays homology with the N-terminal region of the AtGRFs (Figure 5a; Kim et al., 2003). In addition, the AtGRFs contain one or two zinc-finger-like domains (Figure 5b; Kim et al., 2003). Therefore, we tested the importance of these domains in protein–protein interactions. We used the an3-2 protein (Figure 2b), and found that this protein was unable to interact with AtGRF5 (Figure 5d). In line with this result, the SNH domain was sufficient for the interaction with AtGRF5 (Figure 5e). Furthermore, the removal of the hBRM-like domain from AtGRF5 also abolished the interaction with AN3 (Figure 5e), indicating that AN3 and AtGRF5 interact with each other through their N-terminal domains.

AN3 and AtGRF5 are required for active leaf cell proliferation

To obtain genetic evidence for a role of AtGRF5 and AtGRF9 in the control of cell numbers, we examined loss-of-function phenotypes of these genes (Figure 6). Two mutants, Salk_086597 and Salk_140746, have a T-DNA insertion at the first intron of AtGRF5 and the fourth intron of AtGRF9, respectively. The amplification of the entire coding regions revealed that the accumulated transcripts of both mutants from the disrupted loci were below the detection level (Figure 6a), suggesting that the functions of these genes are at least partially reduced. Therefore, these two lines were named atgrf5-1 and atgrf9-1. We characterized the leaves of these lines histologically. The measurements of leaf length and width as well as the determination of the leaf index showed that the leaves of atgrf5-1 were slightly narrower than wild-type leaves (Figure 6b, Table 1). The total cell number of palisade cells in the subepidermal layer was significantly reduced in an3-4 and modestly reduced in atgrf5-1 as compared with the wild type (Table 1). These results are in contrast to the atgrf1 atgrf2 atgrf3 triple mutation, which showed only a marginal effect on cell number (Kim and Kende, 2004), suggesting that cell proliferation is substantially more dependent on AtGRF5 than on the other family members examined. The cell numbers along the longitudinal and transverse axes were both reduced (significantly in an3-4 and mildly in atgrf5-1) as compared with the wild type. Importantly, the reduction in cell number in the leaf-width direction was more severe than that in the leaf-length direction in both mutants, demonstrating that the uneven reduction in cell numbers along the leaf axes is a major factor influencing leaf shape. However, the palisade cells were larger in an3-4 than in the wild type or atgrf5-1 (Figure 6c, Table 1). In contrast, atgrf9-1 was almost indistinguishable from the wild type except that the leaf width was statistically different between them (P = 0.036) (Figure 6, Table 1), suggesting that AtGRF9 plays a minor or no role in cell proliferation in the wild-type background or that atgrf9-1 is a weak allele. We also found that an3-4 is a stronger allele than atgif1, judged by the degree of reduction in cell number relative to the corresponding wild type (comparing the palisade cell numbers in the transverse axis of an3-4 and wild type in Table 1 to those of Figure 7 in Kim and Kende, 2004). These similarities in leaf phenotypes support our hypothesis that AN3 and AtGRF5 act in the same pathway that controls leaf cell proliferation.

Figure 6.

an3 and atgrf5 mutants have fewer leaf cells than wild type.
(a) RT-PCR analysis of AtGRF5 and AtGRF9 mRNAs in atgrf5-1 and atgrf9-1 mutants.
(b) First leaves of wild type, an3-4, grf5-1 and grf9-1 (left to right). Bar, 5 mm. Note that the atgrf5-1 leaf is slightly narrower than the wild-type leaf.
(c) Paradermal view of palisade cells in the subepidermal layer in wild type (top left), an3-4 (top right), atgrf5-1 (bottom left) and atgrf9-1 (bottom right). Bar, 100 μm.

Figure 7.

Overexpression of AN3 increases the cell number and the leaf size.
(a) RT-PCR analyses of the AN3 and TUB4 mRNAs in AN3 overexpressers. The numbers indicate the independent transgenic line strain number.
(b) First leaves of wild type and the AN3 overexpressers #23 and #28 (left to right). The leaf blades were cut at several sites to flatten the leaves. Bar, 5 mm.
(c) Paradermal view of palisade cells in first leaves of wild type and the AN3 overexpressers #23 and #28 (left to right). Bar, 100 μm.
(d–g) Leaf area, cell size, palisade cell number, and leaf index of first leaves of wild type and the AN3 overexpressers #23 and #28 (n = 9). The data shown are the mean ± SD. For all experiments, 25-day-old plants were used.

Overexpression of AN3 and AtGRF5 results in increased leaf size

To further explore the functions of AN3 and AtGRF5, we examined the phenotypes caused by the overexpression of these genes. Several AN3 and AtGRF5 overexpressers were generated (Figures 7a and 8a), and the phenotypes of these transgenic lines were very similar (Figures 7 and 8). Each transgenic line developed leaves that were 20–30% larger than those of the wild type (Figures 7b,d and 8b,d). The increase in leaf area in the AN3 and AtGRF5 overexpressers was mediated directly by the increase in cell number because cell sizes were normal (Figures 7e,f and 8e,f). As an3-4 and atgrf5-1 produced narrow leaves, one might expect that the overexpression of AN3 or AtGRF5 would lead to the development of leaves wider than those of the wild type. Contrary to this assumption, the cell numbers increased in a polarity-independent manner, as the leaf indices of these transgenic plants were similar to those of the wild type (Figures 7g and 8g). The similarities in the effects of AN3 and AtGRF5 overexpression suggest that AN3 and AtGRF5 share a common function of promoting leaf growth via cell proliferation.

Figure 8.

Overexpression of AtGRF5 increases the cell number and the leaf size.
(a) RT-PCR analyses of the AtGRF5 and TUB4 mRNAs in AtGRF5 overexpressers. The numbers indicate the independent transgenic lines.
(b) First leaves of wild type and the AtGRF5 overexpressers #10 and #29 (left to right). Bar, 1 cm.
(c) Paradermal views of palisade cells in first leaves of wild type and the AtGRF5 overexpressers #10 and #29 (left to right). Bar, 100 μm.
(d–g) Leaf area, cell size, palisade cell number, and leaf index of first leaves of wild type and the AtGRF5 overexpressers #10 and #29 (n = 9). The data shown are the mean ± SD. For all experiments, 25-day-old plants were used.


In this study, we identified two regulatory components controlling leaf width and size: the AN3 and AtGRF5 genes that encode a putative transcription coactivator and a putative transcription factor, respectively. These two gene products are likely to act in the same developmental pathway, given that the leaf phenotypes caused by loss of function of these genes are similar to each other. The large-leaf phenotypes generated upon the overexpression of AN3 and AtGRF5 also support this hypothesis. In addition, these genes are expressed in developing leaf primordia, and their gene products interact with each other in yeast, raising the further possibility that AN3 and AtGRF5 act as a transcription factor complex. AN3 appears to be a homolog of the human transcriptional coactivator, SYT. Importantly, the AN3-interacting domain of AtGRF5 is homologous to the SYT-interacting domain of hBRM and BRG1, but the residual part of AtGRF5 is not homologous to these proteins. Therefore, it could be possible that AN3 obtained a novel interacting partner during evolution and that this new partnership has been incorporated into a mechanism of planar tissue construction characteristic to seed plants. Below, we discuss the roles of the AN3-AtGRF5 system focusing on the control of leaf width and leaf size.

AN3 and AtGRF5 play a key role in the control of leaf shape through cell proliferation

How does the AN3-AtGRF5 system control leaf width at the level of organ growth? Previously we showed that the ROT4 gene has a specific function in negatively controlling the cell number along the longitudinal leaf axis (Narita et al., 2004). Notably, the underlying mechanisms by which the AN3-AtGRF5 system widens leaf lamina could be substantially different from the ROT4 system, as both the an3-4 and the atgrf5-1 mutations reduce cell numbers along both the transverse and longitudinal axes. Moreover, the overexpression of AN3 and AtGRF5 results in the promotion of cell proliferation without affecting overall leaf shape, suggesting that the AN3-AtGRF5 system controls cell proliferation in a polarity-independent manner. How, then, do the loss-of-function mutations of the AN3 and AtGRF5 genes cause narrow-shaped leaves? That occurs, in part, because the reduction of leaf length in an3-4 mutants is partially compensated for by the increase in leaf-cell size (we have called this phenomenon ‘compensation’: Tsukaya, 2002b, 2003, see also below). Moreover, it occurs because the defects in cell number in an3 and atgrf5 mutants are more severe in the transverse direction than in the longitudinal direction. A leaf primordium is initiated as a small bump and undergoes a successive transformation of its shape from triangular through elliptical and into rounded (Donnelly et al., 1999); thus there is a dynamic change in the direction of growth. The lateral expansion of a leaf blade during this morphological transition coincides with the marginal meristem activity and is completed by plate meristem activity (Donnelly et al., 1999). Perhaps the AN3-AtGRF system has a role in the promotion or maintenance of the activity of cell proliferation after the initiation of marginal meristem activity, given that just after initiation, an3 leaf primordia are indistinguishable from those of the wild type (our unpublished observation). It is possible that failure to maintain the plate meristem activity prevents the leaf primordium from active expansion, leading to the narrow leaf shape which is a remnant of the immature shape. We will examine this possibility by more detailed characterization of an3 and wild-type leaf development.

Control of leaf size through cell proliferation

In plants, the initiation and subsequent growth and development of organ primordia are generally mediated by gradients of the plant hormone auxin, with a maximum accumulation at the organ apex (Benkováet al., 2003). This strategy is similar to those in animals, in which gradients of morphogens, such as Wingless in Drosophila, are utilized (Cardigan, 2002). The diverse morphologies of each organ are attained by recruiting a different set of regulatory genes, including those for the control of organ size. As the determinate nature of cell proliferation is a quantitative parameter of morphology specific to lateral organs of shoots, genes that control cell number in a leaf without influencing other aspects of leaf development are pivotal in relation to the determination of leaf size. The AINTEGUMENTA (ANT) gene, which encodes an AP2-like transcription factor, is a representative of such genes (Mizukami and Fischer, 2000). Unlike several cell cycle genes, such as CYCD3;1, E2 PROMOTER BINDING FACTOR a (E2Fa), and E2F DIMERIZATION PARTNER a (DPa) whose overexpression causes the hyperproliferation of leaf cells and significantly affects subsequent differentiation (De Veylder et al., 2002; Dewitte et al., 2003), overexpression of ANT simply extends the duration of cell proliferation in leaves (Mizukami and Fischer, 2000). As in the case of ANT, the overexpression of AN3 and AtGRF5 enhanced cell proliferation, resulted in the production of larger leaves than the wild type, and did not interfere with differentiation, suggesting that these two genes also fulfill the conditions required for size-control genes. Although it is not clear how the functions of ANT, AN3, and AtGRF5 are linked to cell cycle regulators, the downregulation of these genes at an appropriate time during leaf development could be one mechanism involved in the determination of leaf cell number because these three genes are no longer expressed in mature leaves (this study, Mizukami and Fischer, 2000). Therefore, besides the promotion of cell proliferation by these genes, there may be an additional mechanism, as cell proliferation in leaves is eventually terminated even if AN3, AtGRF5, or ANT is overexpressed.

Interestingly, an3-4 has leaf cells much larger than the wild type, but atgrf5-1 does not. This phenomenon, in which a decrease in cell number leads to an increase in cell size, is referred to as compensation (Beemster et al., 2003; Tsukaya, 2002b, 2003). Compensation has been observed in other leaf mutants such as swp and ant (Autran et al., 2002; Mizukami and Fischer, 2000), and in transgenic plants in which the cell cycle progression is affected, such as by the overexpression of the ICK1 and Kip-related protein 2 (KRP2) genes encoding cyclin-dependent kinase inhibitors (De Veylder et al., 2001; Wang et al., 2000). Although the mechanisms of compensation are mostly unknown, the process of compensation suggests a potential link between cell proliferation and cell expansion and therefore might be important in better understanding the mechanisms of leaf size determination.

AtGRF5 has a greater role in cell proliferation in leaf primordia than other characterized members of the AtGRF family

The AtGRF gene family consists of nine members. To date, AtGRF1, AtGRF2, and AtGRF3 have been shown to have redundant roles in the promotion of leaf growth, acting at the processes of cell proliferation and cell expansion (Kim and Kende, 2004; Kim et al., 2003). In this study we showed that the overexpression of AtGRF5 has no detectable effects on the size of mature leaf cells, while the loss and gain of its function is sufficient to reduce or increase leaf cell number, respectively. Thus, the AtGRF5 gene seems to have a more important role in cell proliferation than in cell expansion. With regard to cell proliferation, AtGRF5 seems to have a greater role than the other characterized AtGRF members because the defect of cell proliferation is not evident in the atgrf1atgrf2 atgrf3 triple mutants (Kim et al., 2003) or the atgrf9 single mutant. Such a defect is detectable only when both of the atgrf1 and atgrf3 mutations are combined with the atgif1 mutation (Kim and Kende, 2004). Although it is unclear in other AtGRF members, this illustrates that the characterized members of the AtGRF gene family have, to differing extents, overlapping roles with respect to cell proliferation. Finally, our promoter analysis seems to conflict with the previously reported physical interactions of AtGIF1 with AtGRF1 and AtGRF2 (Kim and Kende, 2004), but these results should be carefully interpreted because transcription factors often move from cell to cell (Gallagher et al., 2004). This possibility should be tested in the future.

Experimental procedures

Plant materials

The wild-type accession used in this study was Col-0. an3-1 was isolated from an X-ray-mutagenized population of S96 (J. Relichova, personal communication). an3-2 was identified using activation tagging in Landsberg erecta (J. Bowman, personal communication). Both an3-1 and an3-2 were introgressed three times into the Col line. Both an3-3 and an3-4 were identified by screening for leaf-shape mutants in M2 populations generated by fast neutron bombardment and X-ray irradiation. For histological analyses, wild-type plants, mutants, or transgenic plants were grown side by side in the same container to minimize variables that might arise from differences in the microenvironments of the growth room. Although, some results, such as the leaf area cell number and cell size, varied to some extent between trials, the qualitative differences among wild-type plants, mutants, and transgenics were reproducible.

Microscopic observations

The leaves were fixed with FAA, were cleared with chloral solution (200 g chloral hydrate, 20 g glycerol, and 50 ml dH2O) as described (Tsuge et al., 1996). Whole leaves and leaf cells were observed by a stereoscopic microscopy (MZ16a) and a Nomarski differential interference contrast microscopy (DMRX E) (Leica Microsystems, Tokyo, Japan). Palisade cells in the subepidermal layer in the center of the leaf blade, between the midvein and the leaf margin, were analyzed. The density of palisade cells per unit area of this region was determined, and the area of the leaf blade was divided by this value to calculate the total number of palisade cells in the subepidermal layer. Palisade cells some distance away from the midvein were counted to determine the cell number along the leaf-length direction. The cell number along the leaf-width direction was counted at the widest point of the leaf primordium. To determine the cell area, 20 palisade cells were measured in each leaf. We carried out two trials for the whole set of histological experiments and obtained similar results. To visualize promoter activity, promoter::GUS transgenic plants were subjected to GUS staining, according to Donnelly et al. (1999), and cleared in chloral solution.

Vector construction and plant transformation

Total RNAs were extracted from shoots of 10-day-old seedlings using the RNeasy plant kit (Qiagen, Tokyo, Japan). First-strand cDNAs were prepared using the SuperScript first-strand synthesis system for RT-PCR (Invitrogen, Carlsbad, CA, USA). The AN3, AtGRF4, AtGRF5, and AtGRF9 cDNAs were amplified by PCR using the following first-primer pairs, with lower-case letters indicating the extension sequence for Gateway cloning: for AN3, 5′-aaaaagcaggcttgATGGCTGGTTACTACCC-3′ (143F) and 5′-agaaagctgggtAATTCCCATCATCTGATGATTTC-3′ (143R); for AtGRF4, 5′-aaaaagcaggctcaATGGACTTGCAACTGAAACAATG-3′ and 5′-agaaagctgggtaATGAAAAACTTGAGTAGAGATTCC-3′; for AtGRF5, 5′-aaaaagcaggctcaATGATGAGTCTAAGTGGAAGTAG-3′ (179F) and 5′-agaaagctgggtaGCTACCAGTGTCGAGTCTTG-3′ (179R); and for AtGRF9; 5′-aaaaagcaggctcaATGAAGATGCAGAGCCCTAAA-3′ (177F) and 5′-agaaagctgggtaAACACCTGGTGAAAACAAAGAC-3′ (177R). Next, a second round of PCR was carried out using the primer pair 5′-GGGGACAAGTTTGTACAAAAAAGCAGGCT-3′ (attB1) and 5′-GGGGACCACTTTGTACAAGAAAGCTGGGT-3′ (attB2). The amplified cDNA fragments were inserted into pDONR201 with the BP reaction, according to the manufacturer's protocol (Invitrogen). The resulting vector series was used to construct overexpression vectors with pH35G, which was derived from the binary vector pSMAH621, using an LR reaction. For promoter::GUS constructs of AN3 and AtGRF1 through AtGRF9, about 2 kb 5′ upstream region of each gene was amplified using the following primers: for AN3, 5′-aaaaagcaggctcTCGGATCCATTTTTGGTACC-3′ and 5′-agaaagctgggtCAGCCATCATGGGCTGCAT-3′; for AtGRF1, 5′-aaaaag=caggctGGTTATTGTGTTCAGGGTTC-3′ and 5′-agaaagctgggtGATCCATAAAAAATGGATTCAGAAG-3′; for AtGRF2, 5′-aaaaagcaggctTGCTATAACTTAATCTCTCGTC-3′ and 5′-agaaagctgggtTATCCATAAGAAGTAAGTAATATAA-3′; for AtGRF3, 5′-aaaaagcaggctGAGACTTTTCCCTTGGATATG-3′ and 5′-agaaagctgggtAATCCATTGAAGAAAGAGAGAGAG-3′; for AtGRF4, 5′-aaaaagcaggctACCATGTATAACCTAATAACC-3′ and 5′-agaaagctgggtAGTCCATGGAACAAAGAGAG-3′; for AtGRF5, 5′-aaaaagcaggctCGTGGTCATCGCAAGGTAAC-3′ and 5′-agaaagctgggtGACTCATCATCTTATTCTCTAG-3′; for AtGRF6, 5′-aaaaagcaggctCCTAGAGGTTCCGACTCC-3′ and 5′-agaaagctgggtTAGCCATGAAGAGAGAAAGAGAA-3′; for AtGRF7, 5′-aaaaagcaggctTGGTGGGTTGGAAGTTCA-3′ and 5′-agaaagctgggtAGTCCATGAAGGTTAGAGATTT-3′; for AtGRF8, 5′-aaaaagcaggctGGCTAAATAACATGTGGCTG-3′ and 5′-agaaagctgggtTCCTCATTGTTGGAATGACTAG-3′; and for AtGRF9, 5′-aaaaagcaggctGATGAATCGAGTTTGGAACC-3′ and 5′-agaaagctgggtTCTTCATAACAAAAATTGGATTTTTGG-3′. After the second round of PCR, each promoter fragment was cloned into pDONR201 and then into the reporter plasmid pBGGUS, which is a derivative of the binary vector pSMAB704. To construct truncated versions of the AN3 and AtGRF5 cDNAs, the following primers were used: for AN3ΔC, 5′-agaaagctgggtAACTTGGTGGCTGAGGCTGAG-3′ and primer 143F; for AN3ΔN, 5′CACCGCAGATTCTCAGCCTCAGC-3′ and 5′-ATTCCCATCATCTGATGATTTC-3′; for AtGRF5ΔC, 5′-CACCATGATGAGTCTAAGTGGAAG-3′ and 5′-ATCAAGAGATTTTCTGGCAC-3′; and for AtGRF5ΔN, 5′CACCCAATCCCTTGGATGGGGGTGT-3′ and 5′-GCTACCAGTGTCGAGTCTTG-3′. These cDNA fragments were inserted into pENTR D-Topo and used for the construction of the DB and AD fusion genes with pDEST22 or pDEST32 through the LR reaction (Invitrogen). For an RNAi construct, the first intron of FATTY ACID DESATURASE3 (FAD3) was amplified such that several restriction sites were added on both ends using the primer pair 5′-CCCGGGATATAAGCTTACTGGTATAATTTCTTAATCTTACG-3′ and 5′-agaaagctgggtGAGCTCAATTGCGGCCGCTTCCACTGTATCATGATTTACACA-3′. This genomic DNA fragment was cloned into pDONR201 by BP reaction and the resulting vector was named pENTI. Next, AN3 cDNA fragments were amplified using the two primer pairs, 5′-GCCGAGAATCAAGCAAGGCT-3′ and 5′-ATTCCAAGCTTGCTATGGTGC-3′, and 5′-GGTACCGAGAATCAAGCAAGGCTT-3′ and 5′-ATTCCAAGCTGGCTATGGTGC-3′. The fragments were inserted into pENTI, after which the inverted-repeat region of AN3 interrupted by the intron was inserted into pH35G using the LR reaction. The overexpression, RNAi, and promoter constructs were introduced into Arabidopsis plants using the floral-dip method. At least 10 promoter::GUS plants were tested for the staining pattern of each construct, and the staining patterns, if strongly detectable, were reproducible in more than three independent lines.

Genetic mapping

The AN3 locus was genetically mapped using various genetic markers (simple sequence length polymorphism, cleaved amplified polymorphisms, and small insertion/deletions) according to the sequence information available at The Arabidopsis Information Resource (TAIR) database ( Deletions in an3 alleles were detected with the following primers: a, 5′-AAAAAGCAGGCTCTCGGATCCATTTTTGGTACC-3′; b, 5′-AGTATGACTCGTCACGTGAC-3′; c, 5′-AAAAGTTATAAGGTACTTGCAAGAG-3′; d, 5′-CTTGTGTTCTGTTGAGTAACAAGA-3′; e, 5′-AGAAAGCTGGGTAATTCCCATCATCTGATGATTTC-3′.

RT-PCR analysis

RT-PCR was carried out using the 143F/R, 179F/R, and 177F/R primer pairs for AN3, AtGRF5, and AtGRF9, respectively. For controls, the TUB4, CYCD3;1 and CYCB1;2 transcript levels were examined using the following primer pairs: for TUB4, 5′-AATACGTCGGCGATTCTCCG-3′ and 5′-CTTAGGAGAAGGAAACACTG-3′; for CYCD3;1, 5′-GACTTAAGAACAATGCTCACTG-3′ and 5′-TTCGATTATGGAGTGGCTAC-3′; for CYCB1;2, 5′-ATGGCGACGAGAGCAAACG-3′ and 5′-TTTCTTGCAGCAAGAGGTTGT-3′.

Yeast two-hybrid assay

Yeast two-hybrid experiments were carried out using the ProQuest two-hybrid system with a Gateway Technology kit (Invitrogen).


We thank Drs J. Bowman at the University of California, Davis and J. Celenza at Boston University for an3-2 and CYCB1;2::GUS seeds, respectively; Dr H. Ichikawa at NIAS, Japan for pSMAB704 and pSMAH621; and Mss T. Kadowaki, M. Kondo, Y. Ogawa, K. Sakai, E. Takabe, and C. Yamaguchi at NIBB, Japan for technical assistance. This study was supported by a Grant-in-Aid for Scientific Research on Priority Areas from the Ministry of Education, Science, and Culture of Japan, and by a grant from the Bio-Design Program of the Ministry of Agriculture, Forestry, and Fishes of Japan.

Numbers for the seed stock: an3-1 (CS241), atgrf5-1 (Salk_086597) and atgrf9-1 (Salk_140746)