The higher-plant shoot apical meristem is a dynamic structure continuously producing cells that become incorporated into new leaves, stems and flowers. The maintenance of a constant flow of cells through the meristem depends on coordination of two antagonistic processes: self-renewal of the stem cell population and initiation of the lateral organs. This coordination is stringently controlled by gene networks that contain both positive and negative components. We have previously defined the ULTRAPETALA1(ULT1) gene as a key negative regulator of cell accumulation in Arabidopsis shoot and floral meristems, because mutations in ULT1 cause the enlargement of inflorescence and floral meristems, the production of supernumerary flowers and floral organs, and a delay in floral meristem termination. Here, we show that ULT1 negatively regulates the size of the WUSCHEL (WUS)-expressing organizing center in inflorescence meristems. We have cloned the ULT1 gene and find that it encodes a small protein containing a B-box-like motif and a SAND domain, a DNA-binding motif previously reported only in animal transcription factors. ULT1 and its Arabidopsis paralog ULT2 define a novel small gene family in plants. ULT1 and ULT2 are expressed coordinately in embryonic shoot apical meristems, in inflorescence and floral meristems, and in developing stamens, carpels and ovules. Additionally, ULT1 is expressed in vegetative meristems and leaf primordia. ULT2 protein can compensate for mutant ULT1 protein when overexpressed in an ult1 background, indicating that the two genes may regulate a common set of targets during plant development. Downregulation of both ULT genes can lead to shoot apical meristem arrest shortly after germination, revealing a requirement for ULT activity in early development.
Meristems, as centers of cell proliferation and organ initiation, are the foundation of all plant structures. After germination, the shoot apical meristem (SAM) and the root apical meristem (RAM) grow in opposite directions to generate the aerial and underground parts of the plant. The cells produced by the SAM develop into stem tissue, leaves and flowers, and also form axillary meristems that reiterate the development of the primary SAM. The constant formation of new organs and tissues throughout plant life relies on precise mechanisms that maintain overall meristem integrity for upwards of hundreds of years in some species.
The SAM persists as a cell dome with both a longitudinal and a radial structure (Steeves and Sussex,1989). In most dicots, the SAM is divided into three clonally distinct layers. Cells in the outermost layer (L1) produce epidermal tissues,whereas cells of the sub-epidermal layer (L2) and the internal layers (L3)differentiate into vascular and internal tissues. Superimposed across these cell layers are distinct zones of differential meristem activity. A central zone (CZ) at the very apex harbors the unspecialized stem cells, which maintain themselves and also replenish cells in the adjacent peripheral zone(PZ) as they are lost during the formation of lateral organ primordia on the meristem flanks. Maintenance of SAM integrity requires a precise coordination between the flow of cells leaving the PZ and their replacement by cells from the CZ, which implies that the different regions of the meristem are in communication with one another. In Arabidopsis, one key component of the meristem communication system is the CLAVATA (CLV) extracellular signaling pathway.
In clavata mutants (clv1, clv2, clv3), all aerial meristems produce a greatly increased cell population, resulting in the formation of fasciated stems, supernumerary flowers and flowers with extra organs (Clark et al., 1993; Clark et al., 1995; Kayes and Clark, 1998). At the other extreme, wuschel (wus) mutants undergo premature termination of their shoot and floral meristems(Laux et al., 1996). WUS, which encodes a homeodomain transcription factor(Mayer et al., 1998), is expressed in a small region in the meristem interior referred to as the organizing center (OC), from where it specifies stem cell identity on the overlying L1 and L2 cells (Schoof et al.,2000). The stem cell-promoting activity of WUS is counterbalanced by the CLV proteins, which are members of a signal transduction pathway that limits the size of the WUS-expressing cell population(Clark et al., 1997; Fletcher et al., 1999; Jeong et al., 1999; Trotochaud et al., 1999). In clv mutant meristems, the WUS expression domain expands laterally and upwards, leading to the accumulation of excess stem cells(Brand et al., 2000; Schoof et al., 2000). Thus,the activity of the CLV pathway establishes a negative feedback loop between the stem cells and the underlying organizing center that maintains meristem homeostasis throughout development (Brand et al., 2000; Gallois et al.,2002; Lenhard and Laux,2003; Schoof et al.,2000).
Maintenance of a functional SAM also requires additional factors that act in pathways independent of the CLV pathway. For example, a number of Arabidopsis mutants that are impaired in chromatin assembly or genome maintenance display pleiotropic phenotypes, including severely disorganized cell arrangements at both the shoot and root apices. Among these are fasciata1 (fas1) and fas2(Leyser and Furner, 1992), mre11 (Bundock and Hooykaas,2002), the AtCAP-E1 and AtCAP-E2 condensin mutants (Siddiqui et al.,2003), and tonsoku/mgoun3/bru1(Guyomarc'h et al., 2004; Suzuki et al., 2004; Takeda et al., 2004). Likewise, mutations in the HALTED ROOT (HLR) gene, which encodes a subunit of the 26S proteasome, result in the disorganization of the SAM and the RAM that correlates with a disturbed shoot organizing center and root quiescent center (Ueda et al.,2004). For most of these mutants, the observed stem fasciation phenotypes are linked to the distortion of the WUS expression pattern in the SAM in a broader and more random manner than occurs in the clvmutants. Mutations in the farnesyltransferase gene ENHANCED RESPONSE TO ABSCISIC ACID1 (Running et al.,1998; Ziegelhoffer et al.,2000) and the prenyltransferase gene PLURIPETALA(Running et al., 2004) also increase SAM size in a CLV-independent manner.
We have previously identified ULTRAPETALA1 (ULT1) as an additional factor that negatively regulates Arabidopsis shoot and floral meristem activity, as ult1 mutations cause the enlargement of inflorescence and floral meristems, leading to the production of supernumerary flowers and floral organs (Fletcher,2001). ult1 mutants also have reduced floral meristem determinacy, and ULT1 has been shown to negatively regulate WUSexpression in order for floral meristem termination to occur at the correct stage of flower development (Carles et al.,2004). Here, we report the cloning and characterization of the ULT1 gene and its paralog ULT2. We show that the organization of the SAM is not altered in ult1 mutants, but that ULT1 restricts the size of the WUS-expressing cell population. ULT1 and ULT2 encode small proteins containing a SAND DNA-binding motif and a B box-like domain, and their expression patterns fully overlap in inflorescence meristems, floral meristems and reproductive organs. Both genes are also expressed in embryos, but only ULT1 mRNA accumulates in vegetative meristems and leaf primordia. We discuss the functions of the ULT factors throughout plant development, and in light of what is known about SAND domain-containing factors in animals, we propose that the ULT proteins may act as direct regulators of developmental gene expression.
Materials and methods
Plant material and growth conditions
Landsberg erecta (Ler), the ecotype in which the ult1 ethyl methanesulfonate (EMS) alleles were isolated, was used as the wild-type strain for the ult1 alleles. The STM promoter-GUS line(Ler ecotype) was obtained from Anita Fernandez and Kathy Barton. The ult1-3 allele (SALK 074642) and ult2-1 allele (SAIL 748C4)were initially isolated from T-DNA mutagenesis populations in the Columbia-0(Col-0) background. The ult1-3 allele was introgressed into Ler through three backcrosses. Plants were grown in a 1:1:1 mixture of perlite:vermiculite:topsoil under continuous cool-white fluorescent lights(120 μmol m-2 s-1) at 22°C, and were watered daily with a 1:1500 dilution of Miracle-Gro 20-20-20 fertilizer. Transgenic lines were generated by the floral dip method(Clough and Bent, 1998).
Mapping and molecular identification of ULT1
Using cleaved amplified polymorphic sequence (CAPS) markers(Konieczny and Ausubel, 1993)distributed across the lower arm of chromosome 4, we established that ult1-2 was flanked by markers PG11 (map position 75.16 cM) and g8300(81.22 cM). For the fine mapping of ULT1, we designed 12 new CAPS markers that spanned the region between markers PG11 and g8300, using the TIGR Landsberg erecta random sequence database(www.tigr.org)as a source for single nucleotide polymorphisms (SNPs). The primer sequences,restriction enzyme and number of restriction sites in the Col/Lerecotypes for the CAPS marker sequences generated across this interval are available upon request.
Sequencing was performed on an ABI PRISM® 3100 Genetic Analyzer sequencer (Perkin Elmer), according to the manufacturer's instructions. Computer-based sequence analysis was performed using VectorNTI® Suite(Informax) and Sequencher (Gene Codes Corporation, Ann Harbor) software. Multiple protein alignments were obtained using ClustalX and edited with SeqVu(The Garvan Institute of Medical Research).
Construction of transgenic lines
ULT1-214 complementation construct
A 2745 bp fragment spanning the ULT1-coding region and flanking sequence was digested from BAC F26K10 with NdeI, Klenow-filled and cloned into blunt pBSK+ vector (pBSK+-ULT construct). Then a BamHI/KpnI fragment was cut from pBSK+-ULT vector and cloned into the binary vector pCD214 (kindly provided by Chris Day). Transgenic plants were selected on MS plates containing gentamycin (100μg/ml).
d35S::ULT1/d35S::ULT2 sense and d35S::ULT1 antisense constructs
The full-length ULT ORFs were cloned into the binary vector pCD223 (kindly provided by Chris Day) at the EcoRI site, flanked 5′ by a double CaMV 35S promoter and 3′ by a nopaline synthase transcription termination signal. The clones were then screened by PCR to obtain the ULT cDNA insert in the sense (S) or antisense (AS) orientation. Transgenic plants were selected on MS plates containing gentamycin (100 μg/ml).
35S::ULT-(Ala)10-GFP and 35S::ULT-(Ala)10-GUS-GFP constructs
The pEZS vectors carrying the CaMV 35S-MCS-(Ala)10-EGFP cassette or the CaMV 35S-EGFP-(Ala)10-MCS cassette were kindly provided by David Ehrhardt. The ULT cDNA fragments were cloned into the EcoRI and BamHI sites of pEZS-NL/CL vectors to give pEZS-NL/CL-ULT constructs. To create the CaMV 35S-MCS-(Ala)10-GUS-EGFP and CaMV 35S-GUS-EGFP-(Ala)10-MCS cassettes, we introduced a short synthetic linker at the NcoI site of pEZS. The β-Glucuronidase uidA (GUS) gene was cloned at the newly created NcoI and PmlI sites. Then the ULT cDNA fragments were cloned into the EcoRI and BamHI sites of the pEZS MCS. For stable transformation of Arabidopsis plants, the 35S::ULT-GFP and 35S::ULT-GUS-GFP cassettes were transferred into pART27 at the NotI site (Gleave, 1992). Transgenic plants were then selected on MS plates containing kanamycin (50μg/ml).
For transient assays, the pEZS-ULT fusions were transformed into onion epidermis cells by particle bombardment using a Biolistic PDS-1000/He unit(BioRad, Richmond, CA), as described(Sanford et al., 1993). For GFP visualization, epidermal peels were examined 24-36 hours after bombardment using a Zeiss Axiophot microscope. GFP fluorescence was visualized with the FITC channel and images were acquired with a 12-bit MicroMax cooled CCD camera operated by IPLab software (Scanalytics, Fairfax, VA). GFP and DAPI fluorescence was visualized in plants using a Zeiss LSM510 confocal laser-scanning microscope (CLSM), with the FITC channel and the UV channel,respectively.
For immunodetection of GFP in the transgenic lines, 0.5 g of inflorescence tissues were ground in liquid nitrogen and then extracted with 500 μl of cold buffer [100 mM MOPS pH 7.6, 100 mM NaCl, 5% (v/v) Glycerol, 1 mM EDTA, 14 mM β-mercaptoethanol, 1 mM PMSF, 2 μg/ml pepstatin A, 0.2 μg/ml leupeptin, 1 μg/ml aprotinin] containing protease inhibitors. Each protein extract (15 μg) was separated on a 12.5% SDS-PAGE gel and blotted on a nitrocellulose membrane. Immunoblots were incubated with a 1:500 dilution of an anti-GFP polyclonal antibody (Santa Cruz Biotech).
The GUS staining reaction and subsequent tissue embedding and sectioning were performed as described (Sieburth and Meyerowitz, 1997), with the exception that GUS localization was visualized after 6 hours of staining with 2 mM of 5-bromo-4-chloro-3-indolyl-β-D-glucoronide (X-GLUC, Bioworld).
Total RNA was isolated from various tissues using the RNeasy plant kit(Qiagen), treated with RQ1 RNase-free DNase (Promega) for 30 minutes at 37°C, and then purified with phenol/chloroform. The first-strand cDNA synthesis was performed on 5 μg of total RNA using Superscript II RNase H- reverse transcriptase (Gibco BRL, Life Technologies) and an oligo dT primer (18 mer), according to the manufacturer's instructions. From 20 μl of the reverse-transcription (RT) product, 1 μl was used for each PCR reaction. The annealing temperature was 54°C for all primer pairs and 34 cycles of PCR were performed for all genes, except when mentioned otherwise.
In situ hybridization
Probes for in situ hybridization were transcribed using the digoxigenin-labeling mix (Roche). The WUS antisense probe was generated as described previously (Mayer et al., 1998). The STM antisense probe was generated from the 1.1 kb transcript described by Long and Barton(Long and Barton, 1998). Separate ULT probes were generated from the full-length coding sequences of ULT1 and ULT2, and from the 3′UTR of each gene. Tissue fixation and in situ hybridization were performed as described previously(Ambrose et al., 2000), with the additional steps that siliques were smashed and seedlings tips were chopped before infiltration, to facilitate fixative penetration into the tissues.
ULTRAPETALA1 negatively regulates the size of the WUS-expressing organizing center during reproductive development
Previous analysis of the EMS-induced ult1-1 and ult1-2alleles revealed an increase in inflorescence and floral meristem size,leading to the production of supernumerary flowers and floral organs(Fletcher, 2001)(Fig. 1A-C). To determine the molecular basis of the ult1 meristem enlargement, we performed in situ experiments to analyze the expression pattern of STM as a marker for inflorescence meristem fate. STM is expressed throughout wild-type inflorescence and floral meristems(Long and Barton, 2000; Long et al., 1996), and is absent from the flanking region that corresponds to the incipient flower and floral organ primordia (Fig. 1E). We did not detect any major changes in the overall pattern of STM expression in ult1-1 and ult1-2 inflorescence meristems. STM expression is visible in the central part of the inflorescence and floral meristems and absent from the peripheral region. However, the domain expressing STM is more extensive in ult1mutant meristems than in wild-type meristems(Fig. 1E-G). This result shows that the supernumerary cells present in ult1 mutant meristems(Fletcher, 2001) correspond to meristem cells rather than to lateral organ primordia cells.
To determine whether the ult1 meristem enlargement occurs uniformly across the shoot apex or is confined to one area, we examined the expression patterns of molecular markers for specific meristematic regions in ult1 inflorescences. We have previously reported that the CLV1 expression domain, which corresponds to the most central L3 cells of the SAM, is significantly broader in ult1-1 inflorescence meristems than in wild type (Fletcher,2001). This result suggested a function for ULT1 in restricting cell accumulation in the interior, central region of the meristem. By contrast, the CLV3 expression pattern in the L1 and L2 layers of the central zone appeared to be unchanged in ult1-1 meristems(Fletcher, 2001), either because the CLV3 expression domain is not affected by the mutation,or because its enlargement across such a small group of cells is too slight to be noticed by in situ hybridization.
The WUS gene is expressed in the interior, deeper layers of shoot and floral meristems, overlapping the CLV1 expression domain(Fig. 1I)(Mayer et al., 1998). Mutations in ULT1 result in the lateral expansion of the WUSexpression domain, without altering its layer specificity(Fig. 1J-K). Counting of WUS-expressing cells confirmed that the organizing center is significantly larger in ult1-1 and ult1-2 meristems than in wild-type meristems (Fig. 1P). In wild-type inflorescence meristems, the mean size of the WUS-expressing domain corresponds to 6.12±0.33 cells in width,3±0 cells in height and 14.75±0.97 cells in total. In ult1-1 inflorescence central sections the WUS domain expands to 8.50±1 cells in width, 3.12±0.33 cells in height, and 21.12±2.80 cells in total, while in ult1-2 inflorescence central sections the WUS domain is 7.75±0.66 cells in width,3±0 cells in height and 18±1.66 cells in total. This result shows that the size of the WUS-expressing organizing center is negatively regulated by ULT1 activity. As ult1-1 plants have larger inflorescence and floral meristems than do ult1-2 plants and produce more floral meristems and floral organs(Fletcher, 2001), our results suggest that the size of the WUS-expressing organizing center may directly affect these traits.
Finally, we used a pSTM::uidA(McConnell and Barton, 1998)reporter line as a marker to examine the size of the peripheral zone of the meristem in wild-type and ult1 plants. This reporter construct does not recapitulate the STM expression pattern in the meristem; instead,it is expressed at the boundary between the proper inflorescence meristem and the incipient floral primordia (Fig. 1M). The pSTM::uidA expression pattern is unaltered in ult1 inflorescences, indicating that the peripheral region of the mutant meristems is not significantly enlarged(Fig. 1N-O). Altogether, our expression analyses indicate that ULT1 restricts the lateral expansion of CLV1- and WUS- expressing cells in the interior of inflorescence and floral meristems.
Positional cloning of ULT1
To isolate the ULT1 gene, we used CAPS-based mapping(Konieczny and Ausubel, 1993)of recombination breakpoints in 1366 meiotic events among the F2 progeny of ult1-2 (Ler) × wild type (Col-O). We had previously shown that the ult1 mutations mapped between the visible markers ag and ap2 on chromosome 4(Fletcher, 2001). Using the CAPS markers throughout this interval, we established that ult1-2 was flanked by markers PG11 and g8300(www.arabidopsis.org). Thirty-one plants with recombination events between PG11 and g8300 were identified from the mapping population, and used to refine the position of the ult1-2 recombination breakpoints to the ends of BAC F26K10. We sequenced candidate genes annotated on the BAC and identified a single gene(At4g28190) that was mutated in both ult1 alleles.
To confirm the identity of At4g28190 as the ULT1 gene, a genomic clone (ULT1-214) containing the At4g28190 coding region along with 1 kb of upstream and 0.5 kb of downstream sequence was introduced into ult1-1plants, and this clone partially or fully complemented the mutant phenotypes(Fig. 1H). T1 and T2 ult1-1 plants transformed with the ULT1-214 genomic construct produced meristems and flowers similar to those of wild-type plants. In addition, ult1-1 plants carrying the ULT1-214 transgene flowered at the same time as did wild-type plants, while untransformed ult1-1plants flowered 1 week later on average(Fletcher, 2001). These data confirm that At4g28190 encodes the ULT1 gene.
The complete ULT1-coding region was determined by EST and cDNA analysis, RT-PCR and 5′RACE. This region is 714 bp in length, and consists of three exons and two introns(Fig. 2A), encoding a predicted protein of 237 amino acids with a mass of 26.7 kDa. Genomic sequence analysis indicated the presence of a TATA box, a CCAAT box and a GC box, as well as an in frame stop codon upstream of the transcription start site. Ceres cDNA 96705(www.arabidopsis.org)and the sequencing of RT-PCR products support the annotation of this gene. We have identified a missense mutation in the second exon of this gene in the ult1-1 and the ult1-2 alleles(Fig. 2A). The ult1-1mutation is caused by a G to A transition that changes a cysteine residue to a threonine residue at position 173 relative to the translational initiation site (Fig. 2B). The ult1-2 mutation is due to a C to T transition that replaces a serine residue with a phenylalanine residue at position 83.
Database searches revealed the presence of a sequence on Arabidopsis chromosome 2 that is highly similar to ULT1 at the nucleotide level. This paralogous gene, At2g20825, consists of two exons and a single intron. Because overexpression of this gene can rescue the ult1-1 mutant phenotype (see below), we refer to this locus as ULT2. Conceptual translation of ULT2 gives a putative protein of 226 amino acids (26.1 kDa) with 81% identity and 86% similarity to ULT1 over the full-length of the proteins(Fig. 2B). Twenty-one of the 23 cysteine residues present in ULT1, including C173 which is mutated in the ult1-1 allele, are conserved in the ULT2 protein. The serine residue(S83) that is mutated in the ult1-2 allele is also conserved in ULT2. Notably, ULT1 contains five amino acids at the N terminus (residues 2-5) and six amino acids in the middle of the protein (residues 121-126) that are not present in the putative ULT2 protein.
We have identified sequences corresponding to ULT1- and ULT2-like genes in a number of other plant species, including tomato,maize, cotton, rice, soybean and wheat. So far, only a single ULT-like gene has been identified in these species, compared with two in Arabidopsis. An amino acid alignment of the putative ULT-like proteins for which full-length or nearly full-length genomic and/or EST sequences are available, is shown in Fig. 2B. The overall identity between the proteins ranges from 59% to 72% across the length of the protein. No functions have yet been assigned to any of these ULT-like proteins. The ult1-1 and ult1-2mutations both occur in amino acids that are invariant among all nine of the plant species examined, suggesting that these residues are crucial for protein function.
Sequence analysis of the ULT1 and ULT2 proteins
Two domains can be recognized in the ULT1 and ULT2 protein sequences that have been found in transcription factors. The Prosite program(pit.georgetown.edu)revealed a significant structural homology between the N-terminal region of the ULT proteins (Fig. 2B) and a conserved SAND domain found in animal proteins. The SAND domain is an evolutionarily conserved ∼80-100 amino acid DNA-binding motif that takes its name from the Sp100, AIRE-1, NucP41/75 and DEAF-1/suppressin proteins found in humans and Drosophila melanogaster(Gibson et al., 1998). The ULT1 and ULT2 proteins, as well as the other ULT-like plant sequences, share∼75% identity within the SAND domain(Fig. 2C).
The three-dimensional structures of several SAND domains have been determined by NMR and x-ray crystallography(Bottomley et al., 2001; Surdo et al., 2003). The SAND domain is a compact, strongly twisted α/β fold consisting of five antiparallel β-sheets alternating with four α-helices(Fig. 2C). However, the primary sequence of the SAND domain is poorly conserved between family members. The highest degree of amino acid conservation is found between two otherwise unrelated proteins from C. elegans, CeC25G4.4 and CeC44F1.2, which share 57% identity within the SAND domain. Most of the animal proteins share less than 30% identity within the SAND domain, and the pair-wise comparison score can be as low as 7% identity, as shown for the human AIRE-1 and GMEB1/2 proteins. Thus, the similarity between animal SAND domains instead resides at the secondary and consequent tertiary structure level. Similarly, the major conservation of the ULT SAND domains is at the level of the secondary structure: The PsiPred program (McGuffin et al., 2000) predicts the β1, β2, β3 and β5 strands, as well as the α2 and α4 helices in the ULT proteins(Fig. 2C). The program did not detect the α1 and α3 helices or the β4 sheet, probably because of their extremely small size. Only two conserved cores are highlighted by multiple alignment of the SAND domains, the TPxxFE and the KDWK motifs (Fig. 2C). The TPxxFE motif is perfectly conserved among all the putative ULT-like proteins in plants (Fig. 2B,C). The KDWK core is not conserved in ULT1 and ULT2 nor in the mouse and human AIRE-1 proteins at the primary sequence level, but the secondary structure is conserved. The ult1-2 mutation, which causes a null mutant phenotype(see below), lies within the α2 helix of the SAND domain(Fig. 2B,C).
The ULT1 and ULT2 proteins are highly cysteine rich, with cysteine residues accounting for 9.7% of the total amino acid content of each protein(Fig. 2B). One particular arrangement of cysteine residues near the C terminus of the ULT1 and ULT2 proteins is highly similar to that of a B-box motif found in many eukaryotes(Fig. 2D). In these organisms,the B-box domain has been proposed to function in protein-protein and in protein-RNA interactions (Borden,1998; Torok and Etkin,2000). B-box domains are associated with cysteine-rich zinc-binding motifs in otherwise unrelated proteins, many of them transcription factors, that participate in a wide range of cellular processes(Borden, 1998; Torok and Etkin, 2000). The putative B-box region is more highly conserved between ULT1 and the homologous sequences than the rest of the protein(Fig. 2B).
Subcellular localization of the ULT proteins
In animals, SAND domain-containing proteins are found in the nucleus, in the cytoplasm, or in both compartments(Gross and McGinnis, 1996; Jimenez-Lara et al., 2000; Peterson et al., 2004). Similarly, eukaryotic proteins containing B-box domains have been localized to either the nucleus or the cytosol (Borden,1998; Torok and Etkin,2000). The computer programs Prosite(Hulo et al., 2004; Sigrist et al., 2002), PSORT(Nakai and Kanehisa, 1992),SignalP (Nielsen et al.,1997), and NLSdb (Nair et al.,2003) each predict the ULT1 and ULT2 proteins to be localized to the cytosol, based on the absence of any sorting or signal peptide. However,both ULT proteins are small enough to diffuse passively into the nucleus through the nuclear pores (Raikhel,1992). Subcellular localization experiments using enhanced green fluorescent protein (EGFP) as a marker showed that ULT1-EGFP and ULT2-EGFP fusion constructs transiently transformed into onion epidermal cells are localized in both the nucleus and the cytosolic compartments(Fig. 3A).
To determine the relevance of this localization pattern in vivo, we generated transgenic ult1-1 plants stably expressing either the ULT1-EGFP or ULT2-EGFP fusion protein under the control of the 35S promoter. Transgenic plants that expressed the ULT1 or ULT2 protein with the EGFP moiety attached to either the N terminus or the C terminus had a wild-type appearance, indicating that the fusion proteins are functional in either orientation and can rescue the ult1-1 mutant phenotypes. Visualizing the ULT1-EGFP or ULT2-EGFP fusion proteins in the roots or petals of the transgenic plants, we observed signal in both the nucleus and the cytosol(Fig. 3B-D). Immunoblotting of extracts from the transgenic plant using an anti-GFP antibody showed that the observed localization pattern is not an artifact due to the cleavage of the fusion protein (Fig. 3E). The same ULT-EGFP fusion proteins in combination with a nuclear localization signal (NLS) or a nuclear export signal (NES) also complement the mutant phenotype when expressed in an ult1-1 background (data not shown).
However, as the ULT-EGFP fusion proteins are still smaller than the nuclear pore exclusion size they may enter or exit the nucleus passively, especially when expressed at high levels under the 35S promoter. To prevent passive entry into or exit from the nucleus, we fused each ULT protein to a combined GUS(β-glucuronidase)-EGFP protein(Grebenok et al., 1997). When bombarded into onion epidermal cells, the constructs gave a GFP and a GUS signal primarily in the cytosol for some cells and equivalently in the cytosol and the nucleus for others (Fig. 3F,G). Thus, the ULT1 and ULT2 proteins have a dual localization in the nucleus and in the cytosol, and may function in both compartments.
ULT1 and ULT2 expression analysis
We used RT-PCR to determine the distribution of ULT1 and ULT2 mRNA transcripts in wild-type tissues. As shown in Fig. 4, ULT1transcripts could be amplified from all tissues tested: roots, 8-day-old seedlings, mature leaves, stems, inflorescences, pollen and siliques. ULT2 expression was specific to the reproductive developmental stage,being detected only in inflorescences, pollen and siliques. For both genes,the highest level of expression was observed in inflorescence tissues.
We then performed in situ hybridization experiments to localize the ULT1 and ULT2 mRNAs more precisely in the tissues where they could be detected by RT-PCR. ULT1 and ULT2 transcripts can be detected throughout the inflorescence meristem, and a weak signal can also be detected in the inflorescence vascular tissues(Fig. 5A,B). Neither ULT1 nor ULT2 transcripts are detectable in stage 1 flower meristems budding from the flanks of the inflorescence meristem, but they reappear much stronger in late stage 2 primordia. As soon as the sepal primordia initiate (stage 3), ULT1 and ULT2 expression is excluded from these organ primordia and becomes restricted to the floral meristem (Fig. 5B). As flower development continues, ULT1 and ULT2 transcripts become further restricted to stamen and carpel primordia(Fig. 5D-G). Expression in carpels was detected only in the adaxial domain, corresponding to the region of ovule formation. Hybridization to cross-sections of mature flowers reveals specific signal throughout the ovules and in the tapetum tissue of the anthers(Fig. 5H-I). Thus, the ULT1 and ULT2 mRNA expression patterns are coincident in inflorescence meristems, floral meristems and developing flowers.
Next, we determined the expression patterns of both genes in seedlings and embryos. ULT1 is expressed throughout the vegetative SAM and in young leaf primordia (Fig. 6A). A stronger ULT1 signal is detected on the adaxial side of the leaf primordia, as observed for the carpel primordia. The antisense ULT2mRNA probe did not hybridize to seedling tissues(Fig. 6B), confirming that ULT1 but not ULT2 is expressed during the vegetative stage. In mature embryos, both ULT1 and ULT2 transcripts are detected in the SAM, and ULT2 expression is also observed in the RAM(Fig. 6D-G). The expression is detected in very restricted domains corresponding to meristematic cells localized at the apices. Interestingly, in all earlier analyzed stages - from the eight-cell stage to the bending cotyledon stage - ULT1 and ULT2 transcripts are localized throughout the embryo, occasionally displaying a stronger signal between the developing cotyledons and in the vasculature (Fig. 6J-P). This suggests that ULT expression becomes tissue-restricted only at the time when the embryo enters the maturation phase of development. No signal was detected in the suspensor or in the endosperm at any stage(Fig. 6J-P), showing that ULT1 and ULT2 gene expression is embryo specific.
ULT2 overexpression can rescue the ult1-1 mutant phenotypes
Because the ULT1 and ULT2 gene expression patterns overlap in inflorescence and floral meristems, we asked if ULT2 could mimic ULT1 function in these tissues. We transformed a d35S::ULT2 sense construct into ult1-1 plants, in order to increase the level of ULT2expression in this mutant background. We analyzed the capacity of the d35S::ULT2 transgene to rescue the ult1-1 phenotypes, and compared its effects with those of a d35S::ULT1 transgene, by scoring floral organ number and flowering time.
Transgenic d35S::ULT plants display a gradient of phenotypes that correlates with the level of ULT gene overexpression. Those plants expressing the highest levels of ULT1 or ULT2 show dramatic vegetative phenotypes as soon as a few days after germination (C.C.C. and J.C.F.,unpublished). Consequently, we performed the complementation analysis on d35S::ULT2 ult1-1 lines that had a wild-type appearance at the vegetative stage. As expected, RT-PCR experiments showed that these lines display a more moderate increase in ULT2 gene expression than the dramatically affected overexpression lines (data not shown). By analyzing these moderate overexpression lines, we found that the d35S::ULT2 transgene complements the ult1-1 mutant phenotypes to the same extent as the d35S::ULT1 transgene (Fig. 7). Indeed, ult1-1 plants containing either of these constructs display floral organ number and bolting time phenotypes close to those of the wild type. Thus, although the endogenous level of ULT2 is not sufficient to overcome the effect of the ult1-1 mutation, an increase in the amount of wild-type ULT2 protein in the ult1-1 background allows complete rescue of the ult1-1 mutant phenotypes. These data indicate that, when expressed at higher levels, wild-type ULT2 protein can functionally compensate for mutant ULT1-1 protein.
Identification of ult1 and ult2 T-DNA alleles
In order to examine the full spectrum of biological functions for ULT1 and ULT2, we have obtained an insertion allele of each gene. The ult1-3 allele contains a T-DNA insertion in the first exon of ULT1, 155 bp after the start codon(Fig. 2B). RT-PCR experiments show that ult1-3 plants do not accumulate ULT1 transcripts(Fig. 8A) and place the ult1-3 allele as a null allele. The inflorescence and flower phenotypes of plants carrying the ult1-3 mutation are indistinguishable from those of ult1-2 plants(Fig. 1C,D, Fig. 8B), indicating that the ult1-2 EMS line is a phenotypic null allele probably because of the lack of functional ULT1 protein. WUS molecular marker analysis confirms that the ult1-3 allele phenocopies the ult1-2allele (Fig. 1K,L). In ult1-3 inflorescences, the mean organizing center size is 7.62±0.48 cells in width, 3.12±0.33 cells in height, and 18±1.22 cells in total. These values are not significantly different from ult1-2 (Fig. 1P). Surprisingly, the ult1-1 EMS mutation has a more severe effect on inflorescence meristem size, WUS domain expansion and floral organ number than either the ult1-2 or the ult1-3 null mutations(Fig. 1B-D,J-L,P; Fig. 8B). In addition, ult1-2 and ult1-3 mutant plants flower only two days later than the wild type (Fig. 8C),whereas ult1-1 plants are more dramatically affected, flowering up to 2 weeks later than wild-type plants.
The ult2-1 allele contains a T-DNA insertion in the intron of ULT2, 408 bp after the start codon. Because a weak band corresponding to correctly-spliced ULT2 cDNA could be amplified from inflorescence tissues from some homozygous mutant individuals after 45 cycles of RT-PCR, we cannot conclude that ult2-1 is a null allele(Fig. 8A). ult2-1mutant plants do not display any inflorescence or flower phenotypes, and are indistinguishable from wild-type plants (data not shown). Determining whether the presence of ULT2 protein is required for proper reproductive meristem activity will rely on the identification and analysis of a true null allele for the ULT2 locus.
The ULT1-1 mutant protein has semi-dominant effects
Because the ult1-1 mutant phenotype is more dramatic than that of the ult1-3 null mutant, we analyzed the effect of the ult1-1mutation in the heterozygote state. ult1-1 behaves as a slight semi-dominant allele when heterozygous: of 18 ult1-1/+ plants scored,five had five sepals and/or petals in the first one or two flowers and one had six petals in the first flower. All ult1-1/+ plants are wild type with respect to stamen and carpel number and floral determinacy, indicating that the ult1-1 mutation is recessive with respect to these traits.
To test whether the semi-dominant effect of the ULT1-1 mutant protein is altered in the absence of wild-type ULT1 protein, we compared the floral organ number and flowering time phenotypes of ult1-1/+ plants with those of ult1-1/ult1-3 plants. We found that ult1-1/ult1-3 plants are more severely affected than either ult1-3 homozygous plants or ult1-1/+ heterozygous plants with respect to sepal/petal number and also flowering time (Fig. 8D,E). Thus, eliminating wild-type ULT1 protein enhances the effects of the ult1-1 mutation on flowering time and on floral organ number in the outer two whorls.
Down-regulation of both ULT genes leads to early arrest of the vegetative SAM
Antisense plants carrying a d35S::ULT1 AS construct generated in the Ler wild-type background show a dramatic reduction in the level of both ULT1 and ULT2 transcripts(Fig. 9A). Some plants from the antisense lines fail to germinate (data not shown), while the rest display a range of shoot and floral meristem defects(Fig. 9B-I).
The most strongly affected plants have severely disorganized SAMs that resemble those of fas1 or fas2 plants(Leyser and Furner, 1992; Kaya et al., 2001), with highly aberrant lateral organ initiation(Fig. 9B). Although Ler wild-type seedlings develop four true leaves after 7 days of development, ULT AS seedlings have formed only two cotyledons (class 1), two barely developed filamentous leaves (class 2) or two to three stunted leaves(class 3) after 14 days of development(Fig. 9B, first row). The wild-type SAM is a dome-shaped structure that produces lateral organs in a regular phyllotaxy, but no meristematic structure can be detected between the two cotyledons of class 1 ULT AS plants (arrowhead). Class 2 plants initiate leaf primordia at a greatly reduced rate and their SAMs are very small and flat (asterisk), while class 3 plants produce small leaf primordia(arrowheads) around a reduced SAM composed of few enlarged cells(Fig. 9B, second row). Comparison of sections through 7-day-old Ler and 14-day-old ULT AS seedlings (Fig. 9B, third row)shows that class 1 and class 2 AS seedlings lack more than a few meristematic cells (arrowheads). Sections through class 3 ULT AS shoot apices reveal a small group of enlarged cells that are not organized into layers as in the wild type. After the termination of the primary SAM some ULT AS plants initiate axillary meristems, which generate one or more inflorescences(Fig. 9D) much later than wild-type plants (Fig. 9C). These axillary inflorescence meristems can also arrest precociously, after the production of a couple of flowers (Fig. 9E).
The least severely affected ULT AS plants produce flowers that resemble those of ult1-2 mutants (Fig. 9F-I). These plants form flowers with supernumerous floral organs(Fig. 9G) when compared with wild-type plants (Fig. 9F). Five sepals and five petals are observed in some flowers(Fig. 9G), and others form up to four carpels (Fig. 9H). Flowers from the ULT AS lines also display a partial loss of determinacy, in that supernumerous carpels can develop as fifth whorl structures within the fourth whorl gynoecium (Fig. 9I, arrow).
ULT1 regulates the size of the inflorescence meristem organizing center
It has been well established that the self-perpetuation of shoot and floral meristems requires interactions between the CLV and WUSfactors, which set up a feedback loop between the stem cells and the underlying organizing center (Brand et al.,2000; Schoof et al.,2000). When the negative regulation of meristem cell accumulation is disrupted in clv mutants the WUS expression domain expands laterally and upwards, correlating with an increase in stem cell accumulation and meristem fasciation (Brand et al., 2000). Our analysis in ult1 mutant plants of molecular markers for different meristem regions indicates that like the CLV proteins, ULT1 also plays a role in preventing the lateral enlargement of the WUS-expressing cell population in reproductive meristems. In the absence of ULT1 function, the increased number of cells in the central region of the meristem may enable the production of extra floral primordia from the inflorescence meristem and extra organ primordia from the floral meristem.
Several observations suggest that ULT1 and the CLV loci regulate the size of the WUS-expressing cell population via separate pathways. First,the meristems of clv mutant plants, but not of ult1 mutant plants, are measurably taller than those of wild-type plants(Clark et al., 1993; Clark et al., 1995; Fletcher, 2001). Second, in clv but not in ult1 inflorescence meristems the WUSexpression domain extends one cell layer up compared with wild type. Third, ult1 and clv alleles show synergistic effects on inflorescence and floral meristem size, suggesting that they use separate pathways to regulate a common process(Fletcher, 2001). Finally,although wus mutations have been shown to be epistatic to clv mutations in both shoot and floral meristems, ult1 wusdouble mutants have additive phenotypes, except in the center of the flower(Carles et al., 2004). Thus,ULT1 has both WUS-dependent and WUS-independent functions in maintaining meristem activity, and converges with the CLV pathway primarily at the point of limiting the lateral expansion of the WUS-expressing cell population.
ULT1 and ULT2 proteins resemble transcriptional regulators
ULT1 and ULT2 define a small family of closely related plant proteins that contain conserved SAND and B box-like domains. A high degree of sequence conservation between the Arabidopsis ULT1 and ULT2 proteins and predicted proteins in eight other monocot and dicot species was observed across the length of the proteins (Fig. 2B). In addition, the ULT-like ESTs from tomato, soybean and alfalfa were identified from shoot and/or floral meristem tissues - tissues in which Arabidopsis ULT1 is known to act(Carles et al., 2004; Fletcher, 2001). This observation suggests that the roles ULT1 plays in meristem maintenance and floral determinacy may be widely conserved among angiosperms.
Until now, reports of proteins containing a SAND domain have been restricted to the animal phyla (Bottomley et al., 2001; Gibson et al.,1998; Surdo et al.,2003). Overall, their primary sequences are quite divergent except for two core elements, but the secondary structure of the SAND domain is highly conserved. The same holds true for the SAND domains present in ULT1 and ULT2. The ult1-2 missense mutation, which lies in the α2 helix of the ULT1 SAND domain, changes a serine residue to a phenylalanine. In other SAND domain proteins glycine, alanine or cysteine residues are encountered at the same position. These amino acids, like serine, have small side chains. Thus, the introduction of a highly hydrophobic aromatic phenylalanine residue is likely to disrupt the structure of the SAND domain in the mutant ULT1-2 protein, perturbing potential DNA-binding and/or protein-protein interactions. Moreover, the fact that such a missense mutation behaves as a knockout mutation illustrates the importance of the SAND domain in ULT1 protein function.
Many SAND domain-containing proteins, such as DmDEAF-1, HsNUDR and HsGMEB1/2, have been shown to bind specifically to DNA(Bottomley et al., 2001; Burnett et al., 2001; Gross and McGinnis, 1996; Surdo et al., 2003). The SAND domain itself has been proven to mediate this interaction via the KDWK motif-containing region (Bottomley et al.,2001; Jimenez-Lara et al.,2000; Surdo et al.,2003). Moreover, the SAND domain has been shown to be necessary for the transactivation and homo-multimerization activities of AIRE(Halonen et al., 2004), as well as for its nuclear localization(Ramsey et al., 2002). Altogether, phenotypic and biochemical studies, along with three-dimensional structure modeling, suggest that the SAND domain defines a novel DNA-binding module involved in the regulation of gene transcription. The cloning of the ULT1 gene has led us to the identification of a novel group of plant SAND domain proteins with a conserved secondary structure. By analogy with animal SAND domain factors, we propose that ULT1 and ULT2 may function as transcription regulators, possibly binding directly to target DNA. Alternatively, association of the ULT proteins with DNA might require the presence of a liaison factor between the SAND domain and the target DNA sequence, as observed for the AIRE protein(Pitkänen et al.,2001).
Our subcellular localization studies show that the ULT1 and ULT2 proteins accumulate in, and may be functional in, both the nucleus and the cytosol. One possibility for this dual localization pattern is that the small ULT proteins diffuse freely between the nucleus and the cytoplasm. However, when we generated ULT-GUS-EGFP fusion proteins that were too large to passively enter the nucleus, we still detected signal in the nuclear compartment. This implies that the ULT proteins either contain a functional NLS for nuclear import, or that they enter the nucleus by forming complexes with protein partners that possess an NLS core (Boulikas,1994). A canonical NLS is not detected in the ULT proteins, but both the ULT1 and ULT2 sequences contain a hexapeptide and an octapeptide that each has four arginine or lysine residues. These sequences may correspond to nuclear targeting signals or be part of a bipartite NLS core(Hicks and Raikhel, 1995), as reported for the AIRE protein(Pitkänen et al., 2001). The dual nuclear and cytoplasmic localization of ULT1 and ULT2 proteins is not unprecedented among the SAND domain-containing proteins(Halonen et al., 2004; Ramsey et al., 2002). One possibility is that the dual localization in the nucleus and the cytosol may serve as a modulation mechanism for transcriptional regulation, as shown for some families of transcription factors in plants and animals(Fabbro and Henderson, 2003; Merkle, 2001; Ziegelbauer et al., 2001).
ult1-1 is a dominant negative allele
Side-by-side comparison of ult1-2 and ult1-3 homozygous plants showed that their phenotypes are indistinguishable from one another,revealing that ult1-2 is a phenotypic null allele for the ULT1 locus. However, the ult1-1 EMS allele confers a more severe phenotype than the other two alleles with respect to sepal/petal number and flowering time. Analysis of heterozygous ult1-1/+ plants shows that the ult1-1 mutation is semi-dominant with respect to these traits. When we scored for ult1-1 semi-dominancy in an ult1-1/ult1-3 hemizygous background, we found that the effect of the ULT1-1 mutant protein was more dramatic when wild-type ULT1 protein was absent. In fact, ult1-1/ult1-3 plants closely resembled ult1-1/ult1-1 plants. These results suggest that in ult1-1/+heterozygous plants, wild-type ULT1 protein can compete with the mutant ULT1-1 protein and maintain some normal function, whereas in ult1-1/ult1-3plants no wild-type ULT1 protein is present to compete with the dysfunctional ULT1-1 mutant protein.
It is possible that the ult1-1 missense mutation abolishes protein function but does not prevent binding to other factors. In such a scenario,ULT1-1 mutant protein would compete with wild-type ULT1 protein, sequestering one or more physical interaction partners and preventing or altering their activity. As the ULT1 and ULT2 expression patterns fully overlap in the inflorescence, it is possible that the two proteins themselves physically interact. The example of the SAND domain proteins GMEB-1 and GMEB-2, which share a high amino acid similarity and interact with one another in vitro (Jimenez-Lara et al.,2000), is consistent with this idea. In the ult1-1allele, ULT1-1 mutant protein could sequester wild type ULT2 protein,preventing ULT2 from functioning in shoot and flower tissues. However, because ULT1 and ULT2 expression patterns do not fully overlap throughout development, it seems likely that the ULT proteins also interact with additional factors.
Roles of the ULT1 and ULT2 genes in development
The phenotypes displayed by ult1 null mutant plants reveal that ULT1 plays an important role in negatively regulating inflorescence and floral meristem size, and in maintaining floral meristem determinacy(Carles et al., 2004; Fletcher, 2001). Correspondingly, we find that the ULT1 gene is expressed in inflorescence meristems, floral meristems and developing carpels. Yet despite its role in negatively regulating the size of the WUS-expressing organizing center in a central and interior domain, ULT1 is expressed throughout the shoot and floral meristems, similar to STM, rather than in a region-specific fashion like CLV3, CLV1 and WUS. These data suggest that ULT1 may interact with other, as yet unidentified region-specific factors in the meristem to restrict the accumulation of the WUS-expressing cell population. Moreover, the fact that ULT1is expressed in other domains, such as cotyledon and leaf primordia, shows that ULT activity is not restricted to the meristems. The expression of ULT1 in the developing tapetum and ovules, in particular, implies a specific function(s) in reproduction. However, the absence of detectable phenotypes outside the shoot and floral meristems in ult1 mutant plants again suggests redundancy with other factors, such as ULT2.
The pattern of ULT2 expression in inflorescence meristems, floral meristems and developing flowers appears to coincide perfectly with that of ULT1, yet ult2-1 T-DNA mutant plants do not display any shoot or floral phenotypes. Currently, we cannot exclude the possibility that the presence of very low levels of ULT2 protein translated from rare correctly spliced transcripts is sufficient for proper reproductive meristem activity in the ult2-1 mutant. Nonetheless, the presence of wild-type levels of ULT2 cannot compensate for the loss of ULT1 activity in reproductive meristems, whereas increasing ULT2 expression under the control of a dual 35S promoter can complement the ult1-1 mutation. This observation positions ULT2 as a functional duplicate of ULT1, and suggests that shoot and floral meristem activity may be sensitive to the dose of the ULT proteins. The necessity to fine tune the regulation of genes involved in meristem maintenance could explain the retention of both ULT factors in Arabidopsis. This implies that ULT1 and ULT2 are likely to have multiple common targets, the regulation of which is dependent on ULT dose. That the two genes have other independent targets, as well, is clear from the specific expression of ULT1 in leaf primordia and ULT2 in the embryonic root apical meristem.
ULT1 and ULT2 transcripts are detected throughout the embryo as early as the octant stage, and continue to accumulate in all cells of the embryo proper until maturity, when ULT1 transcripts become restricted to the SAM and ULT2 transcripts to the SAM and RAM. To our knowledge, ULT1 and ULT2 encode the only mRNAs characterized thus far that become restricted to the meristems at such a late stage of embryo maturation. The high level of ULT1 and ULT2expression throughout the developing embryo might be a sign that these genes have important functions early in development that are not revealed in the single mutants. Our previous results have shown that ult1 mutations restore SAM activity to stm and wus null mutant seedlings(Carles et al., 2004),indicating that ULT1 functions to restrict SAM cell accumulation prior to the appearance of a visible phenotype in ult1 mutant plants. Our analysis of ULT1/ULT2 antisense lines provides additional evidence for an early and important function for the ULT genes, as downregulation of both ULT1 and ULT2 can result in aberrant lateral organ production and SAM arrest very early during seedling development.
We thank Chris Day for providing the pCD214 and pCD223 plasmids, David Ehrhardt for the pEZS vectors, and Anita Fernandez and Kathy Barton for the pSTM::GUS line. We also thank Harley Smith and George Chuck for advice concerning the in situ procedure, Doris Kim and Kristen Kwan for assistance with the subcellular localization experiments, Steve Ruzin and Denise Schichnes for help with the confocal microscopy, Guillaume Blanc for thoughtful discussions, and Sheila McCormick, Karen Osmont and Leor Williams for helpful comments on the manuscript. We acknowledge the Torrey Mesa Research Institute (San Diego) and the Salk Institute for Biological Studies(La Jolla) for providing insertion lines, and the Arabidopsis Biological Resource Center for providing clones. This work was supported by the National Science Foundation (IBN-0110667).