ROXY1, a member of the plant glutaredoxin family, is required for petal development in Arabidopsis thaliana

Flower formation in higher plants serves as a model system with which to understand the regulatory processes controlling initiation and morphogenesis of organs that form a complex reproductive structure. In Arabidopsis,flowers develop four different organ types that are arranged in concentric whorls. In the first whorl, four sepals are formed; in the second whorl, four petals are initiated alternately. As the reproductive organs, six stamens develop in the third whorl and two carpels fuse in the fourth whorl,generating the gynoecium. The control of floral organ identity has been intensively studied in the past 15 years. Analysis of floral homeotic mutants led to the proposal of a simple genetic model, explaining how three groups of regulatory genes (class A, B and C genes) control alone or by interaction the organ identity of the four floral whorls(Coen and Meyerowitz, 1991). Class A genes control sepal organogenesis, class A and B genes together petal formation. Class B and C genes govern stamen and class C alone, carpel development. Recently, newly identified genes were incorporated encoding proteins that interact with the class A, B and C proteins in larger complexes(Honma and Goto, 2001). Interactions between the class A genes APETALA2 (AP2) and APETALA1 (AP1), the latter also functioning in the specification of floral meristem identity, and the class C gene AGAMOUS (AG), refine their expression pattern during later stages. AG represses AP1 in whorl 3 and 4 and AP2functions in repressing AG in the first and second whorl, as indicated by ectopic AG expression in ap2 mutants(Gustafson-Brown et al., 1994; Drews et al., 1991). Additionally, further genes participate in the repression of AG,among which LEUNIG (LUG) and SEUSS (SEU)encode for proteins that act also as transcriptional repressors in yeast and animals (Conner and Liu, 2000; Franks et al., 2002).

Whereas the ABC genes control the identity of the organs, initiation of organ primordia is determined before the onset of the class A, B and C activity. For example, in mutants from the class B gene PISTILLATA(PI), feminization of the third whorl organs becomes visible only after initiation of the respective primordia(Hill and Lord, 1989). Primordia formation requires the proper allocation of progenitor cells and regulated control of cell divisions, processes that are presumably under genetic control. Petal primordia are initiated when cells in the L2 layer of the floral primordium divide periclinally, rather than anticlinally(Hill and Lord, 1989). Several mutants have recently been characterized in Arabidopsis that disturb primordia initiation and thus develop an either increased or decreased floral organ number. In clavata mutants, floral organ number is increased because of an enlargement of the floral meristem(Clark et al., 1993). In perianthia mutants, five organs are formed in the first three whorls(Running and Meyerowitz,1996), indicating that PAN normally establishes a tetramerous whorl architecture. UFO encodes a F-box protein required for the activation of the class B gene expression, and strong ufomutants resemble class B mutant flowers(Levin and Meyerowitz, 1995; Wilkinson and Haughn, 1995). Additionally, UFO is also involved in the early control of petal outgrowth, probably by counteracting inhibitory effects non-autonomously exerted by AG (Durfee et al.,2003; Laufs et al.,2003). Two other genes have been shown to affect predominantly initiation of petal development: PETAL LOSS (PTL), a trihelix transcription factor (Brewer et al., 2004); and RABBIT EARS (RBE), a SUPERMAN-like zinc finger protein (Takeda et al., 2004). In ptl mutants, the number of petals is reduced and if they develop, the orientation is altered and the size reduced. PTL is not expressed in petal primordia, but between sepal primordia,indicating a rather indirect effect on petal development(Brewer et al., 2004). Similarly, petal initiation in rbe mutants is disturbed and in a strong mutant only a reduced number of filamentous organs is formed in the second whorl. RBE is expressed in petal precursor cells and petal primordia and controls second whorl organ development independently of organ identity (Takeda et al.,2004). As ptl and rbe mutants still form aberrant petals, primordia initiation seems to be a complex process that is regulated by additional factors.

Here, we report on roxy1 mutants that displays defects in petal initiation and also abnormalities during later petal development. ROXY1 encodes a glutaredoxin (GRX), of which many isoforms exist in different organisms, such as E. coli, yeast, animals, humans and plants. GRXs are oxidoreductases that are involved in many different cellular processes, mainly in the response to oxidative stress(Fernandes and Holmgren,2004). Our data show a novel function during the regulation of flower development for a GRX that belongs to a subgroup specific for higher plants.

The three recessive alleles of roxy1-1, 1-2 and 1-3 were obtained from the GABI-KAT collection(Rosso et al., 2003). Mutants and wild-type plants are all Columbia (Col) ecotypes. Seeds for ag-1/+ in Landsberg erecta (Ler); ap3-3/+in Ler; ap1-10 in Col; ap2-5 in Col; lug-1in Ler and ptl-1 in Col(Griffith et al., 1999) were obtained from the Nottingham Arabidopsis Stock Centre. The rbe-2allele is from the sequence-indexed T-DNA insertion library from the Salk Institute Genomic Analysis Laboratory (accession number SALK_037010). Seeds of AG-I::GUS transgenic plants contain the pAG-I::GUS plasmid, driving GUS reporter expression identical to wild-type AG expression(Sieburth and Meyerowitz,1997). Seeds were sown on soil and treated at 4°C for 2-4 days. Plants were grown in the greenhouse at 21-23°C under long day conditions (16 h light).

Inflorescences or flowers from GFP transgenic plants were observed under the binocular microscope LEICA MZ-FLIII. Images were made using a digital camera (KY-F70). For scanning electron microscopy (SEM), we followed the protocol described by Zachgo et al.(Zachgo et al., 1995).

To identify the flanking sequence adjacent to the T-DNA in roxy1-1, genomic DNA was prepared from the mutant and digested with DraI. An adaptor (a mix of two oligonucleotides:5′-GTAATACGACTCACTATAGGGCACGCGTGGTCGACGGCCCGGGCTGGT-3′ and 5′-ACCAGCCC-3′) was ligated to the ends of digested genomic DNA using ligase from Promega (Mannheim, Germany). First, PCR was carried out with adaptor primer 1 (5′-GTAATACGACTCACTATAGGG-3′) and T-DNA left border primer (5′-ACGACGGATCGTAATTTGTCG-3′); then, a second PCR was performed with the nested adaptor primer 2(5′-ACTATAGGGCACGCGTGGT-3′) and a nested left border primer(5′-ATATTGACCATCATACTCATTGC-3′). Expand High fidelity polymerase(Roche, Mannheim, Germany) was used in the reaction. PCR products were gel purified and sequenced. The cDNA cloning was performed using the 5′/3′ cDNA RACE kit (Roche, Mannheim, Germany) as indicated in the manufacturer's instructions.

To obtain double mutants, roxy1-3 was crossed either with homozygous (ptl-1; rbe-2; ap1-10; ap2-5; lug-1; ap3-3 and ufo-2) or heterozygous(ag-1) lines. All F1 plants produced wild-type flowers. Novel phenotypes were identified in segregating F2 populations and genotypes of double mutants were confirmed by PCR.

Total RNA was extracted with the RNeasy kit (Qiagen). Different organs of 6-week-old plants were used, except for the root, which originates from a three-week liquid culture. First strand cDNA was synthesized from 2 μg of total RNA using 200 U SuperScript™ II Reverse Transcriptase (Invitrogen)according to the supplier's instructions. PCR reactions (58°C annealing temperature with 40 seconds extension for 31 cycles) were carried out using ROXY1-specific primers from the 5′ UTR (100 bp upstream of the start codon; 5′-TCGCGAATTCCCAACAAACTTTAGCCAATCCCTC-3′) and from the 3′ UTR (220 bp downstream of the stop codon;5′-ACGCGAATTCTGTTTACTATTATGATTTAATGAGAGC-3′). PCR products were loaded on an ethidium bromide-stained 1.2% (w/v) agarose gel. Images were made with the PhosphorImager Typhoon 8600 (Amersham Biosciences). 18S rRNA(Katz et al., 2004) was used as an internal control.

A 4468 bp ROXY1 genomic fragment was amplified using the Expand High fidelity PCR system (Roche) with the gene-specific primers 5′-GCGTAGATCTCAATAGTCGAGGATCATTCGGAGTGC-3′ and 5′-GTACGCTAGCCTTCAAGCTTCACCTATCTCACTCATAGTC-3′, digested using BglII and NheI and subcloned into pGSA1252(www.chromdb.org/plasmids/pGSA1252.html). A recombinant plasmid containing the correct coding sequence was transformed into the Agrobacterium tumefaciens strain GV3101, and introduced into roxy1-3 mutants using the floral infiltration method(Clough and Bent, 1998). T1 transformants were obtained by BASTA selection.

RNA in situ hybridization was performed as previously described by transcription from PCR templates containing a T7 polymerase binding site at the 3′ end (Zachgo,2002). To avoid cross-hybridization with other GRXs, the ROXY1 antisense probe was prepared using a unique 245 bp fragment from the ROXY1 5′ end(5′-GCGGAATTAACCCTCACTAAAGGGCCAACAAACTTTAGCCAATC-3′ and 5′-GCTCGTAATACGACTCACTATAGGGCACGTGCTCACGCTGAAGATC-3′) and a unique 225 bp fragment from the ROXY1 3′ end(5′-GCGGAATTAACCCTCACTAAAGGGTCTGATCCCTTCCTCTGCTTTC-3′ and 5′-GCGCGTAATACGACTCACTATAGGGCTGTTTACTATTATGATTTAATG-3′) as templates. Control experiments on roxy1-3 inflorescences confirmed probe specificity. The AG antisense probe was generated by PCR with the primer pair:5′-GTGGAATTAACCCTCACTAAAGGGAGCTTACGAGCTCTCTGTTCTTTG3′ and 5′-GTGCGTAATACGACTCACTATAGGGCAATTCACTGATACAACATTCATGG-3′. T7 RNA polymerase (Roche) was used for in vitro transcription. GUS staining was done as described by Müller et al.(Müller et al.,2001).

The GFP ORF with stop codon was amplified using primer 5′-TAGGCGCGCCCATGGGTAAAGGAGAAGAACTTTTC-3′ and 5′-GAACGAGCTCGTCGACCGGTGAGGATCCTTATTTGTATAG-3′, and subcloned into the binary vector pGPTV-HPT. The regions 3912 bp upstream from the ROXY1 stop codon TGA and 445 bp downstream of the stop codon were further amplified using primer pairs:5′-TGCGTCTAGAACTACATAAAAGCCTTTCAG-3′ and 5′-TAGGCGCGCCGAGCCAGAGAGCGCCGGCGTCTTTGAGAA-3′; and 5′-GCCGACCGGTTCCCTTCCTCTGCTTTCTTTTTTCTTTTC-3′ and 5′GTACGAGCTCCTTCAAGCTTCACCTATCTCACTCATAGTC-3′. The two fragments were fused up- and downstream to the GFP ORF, respectively. The clone,selected for transformation into plants, was confirmed by sequencing.

Point mutations in specific amino acids of ROXY1 were introduced by a PCR-mediated mutagenesis strategy as described by Rouhier et al.(Rouhier et al., 2002). To introduce a serine/cysteine exchange mutation, pairs of mutagenic oligomers(mutagenic bases are written in lowercase) were used:5′-TGATCTTCAGCGTGAGCACGTcCTGC-3′ and 5′-GCAGgACGTGCTCACGCTGAAGATCA-3′ (C49S);5′-TGATCTTCAGCGTGAGCACGTGCTGCATGTcCCA-3′ and 5′TGGgACATGCAGCACGTGCTCACGCTGAAGATCA-3′ (C52S); and 5′-CCTCATTCGTCTCCTCGGCTcCT-3′ and 5′-AGgAGCCGAGGAGACGAATGAGG-3′ (C90S). After sequencing, the three mutated ROXY1 genes were cloned into the pBAR-35S vector(Müller et al., 2001) and transformed into the roxy1-3 mutant. Complementation was analyzed in transgenic T1 plants, obtained by BASTA selection.

In Arabidopsis, four petals develop in the second whorl of the flower. Two petal primordia are initiated next to the two lateral stamen primordia and develop initially slightly slower than the stamens. Petal growth progresses rapidly when organ differentiation starts with the formation of a stalked base and flattened blade (Fig. 1E). We identified a mutant, named roxy1, displaying severe defects during petal development in the T-DNA GABI-KAT collection(Fig. 1B-D). Three different roxy1 alleles were investigated that show similar phenotypes(Table 1). The petal organ number is decreased and varies between 0 and 4 petals. On an average, mutant flowers form about 2.5 petals instead of 4.0 petals(Fig. 1B-D; Table 1). SEM analysis shows that absence of petals is due to lack of petal primordia initiation. Either one or two of the petal primordia initiated next to the lateral stamens can be missing in roxy1 mutants (Fig. 1G,H). Analysis of 100 mutant flowers did not reveal a positional bias: ab- and adaxial petal organogenesis were affected equally. Furthermore,strength of the defect was not dependent on the position of the flowers in the inflorescence. These observations indicate that the ROXY1 function is required for the initiation of petal primordia. Further petal development is often delayed and abnormal. In wild-type flowers, petal primordia start to differentiate into a stalked structure at stage 9(Smyth et al., 1990) and commence rapid growth that leads to the formation of a tongue-shaped petal. The size of roxy1 petals is frequently reduced and blades often fold inwards such that petals are not protruding out of the flower(Fig. 1B,I). The identity of the differentiated epidermal cell layers of the folded petals is not affected by roxy1 mutations. Conical cells, typical for the adaxial (inner)epidermal cell layer are formed at the tip of these petals(Fig. 1J). Occasionally,filamentous structures emerge in the second whorl that develop on their tips stigmatic papillae, a carpeloid feature(Fig. 1F,L). Rarely, petals form spoon- or tube-shaped organs (Table 1, Fig. 1K). These data demonstrate that ROXY1 has an additional function during petal development by maintaining normal petal morphogenesis throughout later stages of petal development. As similar flower phenotypes are displayed by the ptl and rbe mutants(Griffith et al., 1999; Takeda et al., 2004), we excluded the possibility that roxy1 is allelic to ptl and rbe (data not shown). To a lesser degree, roxy1 mutations also affect stamen development. Rarely, fusions of lateral and medial stamens were observed predominantly occurring between their filaments(Fig. 1H). Overall, the number of stamens was reduced by about 10%, mainly caused by lack of lateral stamen formation (Table 1). Sepal development was normal, although at higher temperature (26°C), stigmatic structures were occasionally produced at the tips of these organs (data not shown). Fourth whorl organogenesis was not affected.

The roxy1-1 mutant was identified in a T3 population of the GABI-KAT T-DNA tagging collection (Rosso et al., 2003). For isolation of the mutation causing the roxy1-1 phenotype, genomic sequences flanking the T-DNA were obtained by adapter ligation. Blast searches using the whole Arabidopsisgenome revealed that the T-DNA was inserted into the promoter region of a gene with the Accession Number At3g02000. At least two T-DNA borders connected and inserted in the same position, 175 bp upstream of the putative At3g02000 start codon (Fig. 2A). Thereupon, two other T-DNA lines, named roxy1-2 and roxy1-3, were identified in the GABI-KAT collection. Reciprocal crosses confirmed that the three mutants are allelic (data not shown) and further work was focused on the roxy1-2 and roxy1-3 alleles, as they carry insertions in the coding region of At3g02000 (Fig. 2A). In the roxy1-2 mutant, the T-DNA insertion was localized in the 5′ end of the coding region, which comprises a single exon. The right T-DNA border was detected at the 5′ end, but neither a left nor a right border could be detected at the 3′ end of the insert,indicating rearrangements. In the roxy1-3 mutant, the T-DNA inserted in the 3′ end of the coding region and a left border could be identified at the 3′ insertion side (Fig. 2A).

Complementation studies were carried out employing a 4468 bp genomic sequence of At3g02000 spanning over 3600 bp of the promoter region(Fig. 2A). Binary vectors were transformed into roxy1-3 mutants, and 20 independent transgenic T1 lines were analyzed. In total, 18 lines produced wild-type like flowers and two plants formed mutant-like flowers. These results show that the 4468 bp genomic fragment of At3g02000 is sufficient to complement the roxy1-3mutant phenotype.

For isolation of the full-length ROXY1 cDNA, 5′ and 3′Rapid Amplification of cDNA ends (RACE) reactions were conducted. The 5′RACE allowed us to locate the transcription initiation site 100 bp upstream of the ATG start codon of translation. By 3′ RACE, three cDNAs with 3′ UTRs of different lengths were isolated. The longest 3′ UTR comprises 220 bp downstream of the TGA stop codon, the other two are 159 and 147 bp long. By RT-PCR analysis, no ROXY1 gene expression could be detected in the roxy1-2 and roxy1-3 mutants(Fig. 2C), indicating that they are null mutants.

The ROXY1 open reading frame consists of one single exon, encoding 136 amino acids (Fig. 2B). Homology searches revealed that the ROXY1 protein belongs to the family of glutaredoxins (GRXs), which are represented by a number of isoforms in different species, including procaryotes as well as lower and higher eukaryotes (see below). GRXs are small redox proteins that can reduce disulfides via a dithiol or monothiol mechanism by way of conserved cysteines. The dithiol mechanism depends on an active site that contains a conserved N-and C-terminal cysteine (CXXC), and can reduce both, protein disulfides and glutathione (GSH) mixed disulfides. Monothiol GRXs (CXXS) can only reduce GSH mixed disulfides (Vlamis-Gardikas and Holmgren, 2002; Lemaire,2004). These oxidoreductases have thus far mainly been studied in E. coli, yeast and mammal cells, and play a major role in the response to oxidative stress (Fernandes and Holmgren, 2004). For plants, information on the function of GRXs is scarce. roxy1 represents the first mutant revealing a function for a plant glutaredoxin during flower development.

In Arabidopsis, GRXs of the CPYC type exist (e.g. At5g20500),sharing the active site with other GRXs from E. coli, S. cerevisiaeand human GRXs. However, ROXY1 and its closest Arabidopsishomolog At5g14070 belong to the CC type of GRXs, which are specific for higher plants (Lemaire, 2004). In the ROXY1 protein, the putative active site is composed of two cysteines (C49 and C52; Fig. 3A), separated by another cysteine and methionine (CCMC). Furthermore, these two proteins have another cysteine in common (C90; Fig. 3A), which could represent (but only for ROXY1), a conserved N-terminal cysteine of an additional monothiol site (CSGS).

To investigate how the different cysteines identified in the ROXY1 protein contribute to its function during petal development, cysteines at the positions 49, 52 and 90 were converted into serines. Thereby, we aimed to remove the capacity to form dithiol bridges and thus to abolish the putative reducing power of the ROXY1 protein. Three mutated (C49S, C52S, C90S) and a wild-type cDNA were brought under the control of the cauliflower mosaic virus promoter, transformed into roxy1-3 mutants and the complementation capacity of the mutagenized proteins was analyzed. Fig. 3B-E shows representative inflorescences from T1 transformants. The wild-type control construct was able to complement the roxy1-3 mutant(Fig. 3B). Sixty-four percent of the transgenic T1 lines (41/64) formed wild-type flowers. Similarly, 51%from the C52S (35/68) and 68% from the C90S (37/54) transgenic roxy1-3 T1 plants resembled wild-type flowers, respectively(Fig. 3D,E). However, only 4%C49S T1 plants (4/92) were complemented(Fig. 3C). These data suggest that the N-terminal C49 in the CXXC motif is crucial and less indispensable than the other investigated cysteines for the ROXY1 protein to exert its function as an oxidoreductase during petal development.

Organ-specific expression of ROXY1 was investigated by RT-PCR,showing that ROXY1 is strongly expressed in inflorescences, roots and siliques. Weaker expression was detected in mature flowers, and no expression was observed in stems and leaves (Fig. 4A). Inflorescence-specific expression was further analyzed by in situ hybridization. The earliest detectable ROXY1 signal was localized in the inflorescence apex, delineating the area where future floral primordia will emerge (pre-stage 1). After formation of young floral primordia(stage 1; Fig. 4B,C), ROXY1 expression is restricted to young floral organ primordia (stage 2 and early stage 3). Signal was detected in the areas where sepal primordia will be initiated (Fig. 4B,C)and vanishes once sepal primordia start to overgrow the flower primordium(Fig. 4D). Then, at stage 4 and 5, ROXY1 mRNA appears in the petal and stamen primordia. Again, onset of expression starts slightly before the respective organ primordia are formed(Fig. 4D). In early stages, ROXY1 is expressed throughout the whole petal and stamen primordia. After onset of stamen differentiation, ROXY1 expression is restricted to the vascular tissue of these two organs(Fig. 4E,F) and was also detected in young ovule primordia (data not shown).

For analysis of the in vivo protein expression pattern, a fusion-protein of ROXY1 and GFP was produced, driven by a 3560 bp long promoter fragment and comprising 445 bp located downstream of the ROXY1 stop codon. Comparison of Fig. 4C,D with Fig. 4G,H shows that the ROXY1 protein and mRNA expression patterns overlap. We therefore conclude that the GFP construct contains all regulatory elements required for the proper expression of ROXY1.

To determine whether the reduction of petal organs in roxy1mutants is whorl or organ specific, double mutants with class B and C mutants were constructed. Owing to loss of the class B gene function in the APETALA3 (AP3) mutant ap3-3, petals are transformed into sepals and stamens into carpeloid organs or filamentous structures(Fig. 5A). In the roxy1-3 ap3-3 double mutant, an additive flower phenotype was observed. In the second whorl, the organ number and size of sepals is reduced and some are folded inwards, as observed in roxy1-3 single mutants (compare Fig. 1B with Fig. 5B). In the class C mutant ag-1, third whorl stamens are transformed into petals and another mutant flower is produced in the place of fourth whorl carpels(Fig. 5C). In the roxy1-3 ag-1 double mutant, the number of second whorl petals is reduced and their shape altered (Fig. 5D,E). However, petals formed in the third whorl are not affected in their development, resembling ag-1 third whorl organs. Additionally, some double mutant flowers produced a secondary flower in the second whorl, a feature not observed in the single mutants under our growth conditions (Fig. 5F). Overall,these data show that ROXY1 exerts its function in a position- and not organ-dependent mode.

In the roxy1 single mutants, occasional appearance of stigmatic papillae on filamentous structures formed in the second whorl(Fig. 1L) indicated that loss of the ROXY1 function might cause ectopic expression of the class C gene AG. Analysis of AG-I::GUS expression in roxy1-3 mutants led to an occasional weak detection of GUSexpression in second whorl petals (data not shown). However, when we analyzed double mutants with the class A mutant ap1-10, the phenotype differed dramatically from both single mutants. In the ap1-10 single mutant,first whorl sepals are transformed into leaf-like organs or form rarely sepal-carpel intermediate organs (Schultz and Haughn, 1993). Second whorl organs are absent or reduced and often, secondary flowers are produced at the base of the first whorl organs(Fig. 6A). In the double mutant, first whorl organs are carpeloid, an effect that was strongest in younger flowers no longer producing secondary flowers(Fig. 6B). Correlation of the double mutant phenotype with ectopic AG expression was investigated by in situ hybridization and histochemical staining of roxy1-3 ap1-10 AG-I::GUS flowers. In ap1-10, as in wild-type flowers, AG is expressed in the center of the floral meristem from stage 3 on and is later restricted to the center of the flower(Fig. 6C,E,G). Loss of AP1 function does not cause ectopic AG expression(Gustafson-Brown et al.,1994). By contrast, in the double mutant onset of AGexpression occurs prematurely in the inflorescence meristem and in young flower primordia (Fig. 6D). During further development, ectopic AG expression extends into first and second whorl organs (Fig. 6D), where it is detectable until late stages, as indicated by GUS staining in mature feminized first whorl organs of roxy1-3 ap1-10 AG-I::GUS flowers (Fig. 6F,H).

In summary, ectopic AG expression is not strongly detectable in roxy1-3 or ap1-10 single mutants. However, the combination of roxy1-3 with ap1-10 results in a synergistic defect revealed by premature onset and ectopic expression of AG, which is likely to be responsible for carpeloid transformation of first whorl organs.

As our data indicate that ROXY1 is involved in the negative control of AG expression we tested genetic interactions of roxy1-3 with two other negative regulators of AG, AP2 and LUG. Single ap2-5 and lug-1 mutants(Fig. 7A,C) show premature and ectopic AG expression (Jofuku et al., 1994; Liu and Meyerowitz,1995), and their phenotypes resemble the roxy1-3 ap1-10double mutant forming carpeloid first whorl organs. Double mutants of roxy1-3 with the intermediate ap2-5 and the weak lug-1 alleles reveal that a lack of the ROXY1 function strongly enhances single mutant phenotypes. In the roxy1-3 ap2-5double mutant, first whorl organs are fused at their base and stigmatic papillae are formed at the tips (Fig. 7B). No petals are formed and occasionally abnormal stamens develop in the third whorl (data not shown). Similarly, first whorl carpeloidy is enhanced in the roxy1-3 lug-1 double mutant and rarely second whorl organs develop (Fig. 7D).

We also investigated genetic interactions with mutants affecting petal development. UFO has a function in controlling the identity of whorl 2 and whorl 3 organs, probably by degrading a repressor of the class B genes AP3 and PI (Levin and Meyerowitz, 1995). Recently, it was shown that UFOcontrols also petal initiation in a dose-dependent manner(Durfee et al., 2003; Laufs et al., 2003). We tested whether the defects during petal initiation and morphogenesis represent a disruption of the same process as in roxy1-3 by scoring roxy1-3 ufo-2 double mutant flowers. In ufo-2, second whorl organs are often absent or transformed into sepaloid organs and the third whorl produces carpeloid or filamentous structures (Fig. 7E). In the roxy1-3 ufo-2 double mutant, a reduced number of carpeloid second whorl organs forms, again indicating ectopic class C activity (Fig. 7F). Additionally, inflorescence growth in the double mutant terminates prematurely in a pistil or in a cluster of carpeloid structures (data not shown).

The rbe-2 mutant is very similar to the roxy1 mutant, as its defects are restricted to petal initiation and morphogenesis(Fig. 7G) (Takada et al.,2004). Double mutants between roxy1-3 and rbe-2 were generated to determine whether ROXY1 and RBE function in the same pathway during petal development. RBE is expressed in petal precursor cells and petal primordia, overlapping with ROXY1expression (Takeda et al.,2004). In roxy1-3 rbe-2 double mutant flowers, instead of filamentous petals, stamens develop in the second whorl(Fig. 7H), producing fertile pollen (data not shown). First whorl organs are not affected. The double mutant phenotype might correlate with ectopic AG expression,affecting only the second whorl. As ectopic stamen formation depends on class B function (Jack et al.,1997), it seems that sufficient class B activity either remains in the roxy1-3 mutant or is enhanced in the double mutant to control,together with ectopic C function, stamen organogenesis in the second whorl.

Taken together, double mutant analyses indicate a function for ROXY1 in the temporal and spatial expression regulation of the key floral regulator AG, as all observed organ transformations are probably caused by ectopic class C function. However, double mutants differ with respect to the degree of the ectopic AG function, such that double mutants with ap1-10, ap2-5 and lug-1 indicate AG activity in the first whorl, whereas this activity is restricted to the second whorl in double mutants with ufo-2 and rbe-2.

Although much research has been conducted on the function of homeotic genes that control flower organogenesis, still little is known about the genes required for the initiation of the flower organ primordia, an event that is not controlled by these key regulatory transcription factors. Here, we demonstrate that a novel gene, named ROXY1, is required for proper petal initiation and organogenesis. Furthermore, double mutant analysis shows that it functions in restricting AG activity to the third and fourth whorl. Surprisingly, the gene encodes for a GRX and belongs to a family of genes that have thus far been associated with the response to oxidative stress.

Sequence analysis revealed that ROXY1 codes for a GRX. GRXs are small (∼12 kDa) disulfide oxidoreductases that possess a typical glutathione-reducible dithiol CXXC or monothiol CXXS active site, required for the reduction of target protein disulfides. Together with thioredoxins (TRXs),GRXs form the large thioredoxin superfamily, whose members share one or several common thioredoxin folds, defined successions of β-sheets andα-helices (Martin,1995). GRXs, like TRXs, may operate as dithiol reductants, and perform fast and reversible thiol-disulfide exchange reactions between their active site cysteines and cysteines of their disulfide substrates. Additionally, GRXs uniquely reduce mixed disulfides in a monothiol mechanism(Fernandes and Holmgren,2004). In most organisms, TRXs and GRXs are the major reducing molecules, and are involved in many cellular processes. Very recently,comparative studies including phylogenetic tree analysis were conducted with oxygenic photosynthetic organisms, comprising Arabidopsis,Chlamydomonas and Synechocystis. Plant GRXs can be grouped into three classes that differ in their active site composition(Lemaire, 2004; Rouhier et al., 2004). In Arabidopsis, 30 GRX genes have been identified. Six belong to the classical CPYC type, intensively studied in E. coli and yeast; and four to the CGFS type that have thus far been analyzed in yeast and humans. ROXY1, together with 19 other Arabidopsis GRXs, forms a third and novel CC type group, defined by the presence of CCMC or CCMS motifs. These GRXs have thus far been identified only in higher plants, suggesting a specifically evolved function (Lemaire,2004). In addition to the 49CCMC52 site, ROXY1 contains another motif, 90CSGS93, that could represent an active site of the CXXS monothiol type. To investigate the contribution of the different cysteines to the ROXY1 function during flower development, site-directed mutagenesis experiments were carried out. Mutation of C49 to S49 caused a strongly reduced capacity of the protein to complement the roxy1-3mutant. Similar results were obtained with E. coli, human and recently also a poplar GRX, showing that the catalytic cysteine was found to be located at the N-terminus of the active site(Foloppe et al., 2001; Padilla et al., 1996; Rouhier et al., 2002). Given that C52 and C90 are less important, our data indicate that C49 might function in the monothiol pathway, even though it is part of a CCMC motif.

In the monothiol mechanism, GRXs catalyze the reduction of glutathione(GSH)-mixed disulfides by using only the N-terminal cysteine thiol, a process known as deglutathionylation (Bushweller et al., 1992). This reversible mechanism is considered to be the more general function of GRXs (Fernandes and Holmgren, 2004; Cotgreave and Gerdes, 1998). However, little is known about glutathionylation of proteins in photosynthetic organisms. In bacteria, yeast and animal cells,several proteins have been identified as being glutathionylated, especially in response to oxidative stress, including chaperones, cytoskeletal proteins,metabolic enzymes and kinases (Lind et al., 2002). Glutathionylation of transcription factors, as shown for NF-κB and Jun reveals a mechanism for a redox-induced inhibition of DNA-binding (Pineda-Molina et al.,2001; Klatt et al.,1999). Our data indicate a novel function of GRXs in flower development, probably the modification of factors involved in the regulation of floral organogenesis post-translationally.

In roxy1 mutants, petal development is affected at very early and also during later stages, as revealed by initiating a reduced number of petal primordia and abnormally bended mature petals. SEM analysis suggests that lack of primorida initiation does not seem to be caused by a reduced size of the flower meristem in the area where petal primordia are normally initiated. Rather, it might be due to a lack of initiation of proper cell divisions required for primordia outgrowth. Similar phenotypes have been observed in mutants of the RBE and PTL genes(Takeda et al., 2004; Brewer et al., 2004). Only the RBE and ROXY1 expression domains overlap in young petal primordia and second whorl organ number is reduced similarly in the single mutants and in the double mutant. This suggests that ROXY1 and RBE might have a common function during second whorl organ initiation. A combination of both roxy1-3 or rbe-1 with the class B ap3 mutants results in additive effects, demonstrating that the function of both genes is organ identity independent. Interestingly, the RBE protein contains two cysteines in the zinc-finger motif and one further cysteine close to the C terminus that could represent putative target sites for modification by ROXY1.

During later petal differentiation, abnormal petal bending was observed in over 40% of the remaining roxy1 petals, a phenotype not observed in rbe mutants. The organ identity of folded petals is not altered, as they still form conical cells typical for adaxial wild-type epidermal cell layer. Curvature of otherwise flat organs can be caused by deregulation of cell division processes, as demonstrated by leaf mutant analysis(Nath et al., 2003). In the roxy1 mutant, disturbance of cell division regulation could be responsible for both, lack of primordia initiation at early stages and altered curvature of petals at later stages. As ROXY1 expression is confined to the vasculature in older petals, this effect would probably be exerted non cell-autonomously. Two observations indicate redundancy for the ROXY1function. First, roxy1 mutants still produce a reduced number of normal petals. Second, we observed only a defect in the second whorl, although early ROXY1 expression is not restricted to this whorl. As ROXY1 belongs to the CC type group of GRXs, which comprises 20 genes,redundancy could be due to the activity of related proteins. Ectopic ROXY1 expression does not disturb wild-type flower development(Fig. 3B), which could be due to high substrate specificity of ROXY1 and/or lack of ubiquitous target gene expression.

Several genes are known to be involved in repression of AG,including AP2, LUG, SEU, AINTEGUMENTA, STERILE APETALA, BELLRINGERand CURLY LEAF (Drews et al.,1991; Conner and Liu,2000; Franks et al.,2002; Elliott et al.,1996; Byzova et al.,1999; Bao et al.,2004; Goodrich et al.,1997). All these genes restrict the expression of AG to the reproductive organs of the flowers. Occasional formation of filamentous structures topped with stigmatic papillae in the second whorl of roxy1 flowers indicated that ROXY1 is an additional component in this repressive mechanism. This effect is enhanced in double mutants of roxy1-3 with ap2-5, lug-1 and also ap1-10, where flowers develop strongly feminized organs in the first whorl and largely lack second whorl formation. Surprisingly, the influence of ROXY1 on AG expression is also revealed in combination with ufo-2 and rbe-2 mutants that do not obviously control AG activity in single mutants. In roxy1 ufo-2 double mutants lack of class B activity in the second whorl (due to mutation in UFO)(Levin and Meyerowitz, 1995),along with an ectopic AG activity (due to the lack of ROXY1function) explains second whorl transformation into carpeloid structures. In double mutants with rbe-2, enough residual class B activity probably still resides in the second whorl that causes together with an ectopic AG activity transformation of second whorl organs into stamens. ROXY1 could participate in AG regulation by modifying its repressors post-translationally. In fact, GRXs have been shown in animals to glutathionylate transcription factors and thereby alter their DNA-binding activity (Pineda-Molina et al.,2001; Klatt et al.,1999). Recently, also a redox-sensitive plant transcription factor, TGA1, has been reported, the reduced form of which displays enhanced DNA-binding during systemic acquired resistance(Després et al.,2003).

Our data indicate for the first time that a GRX plays a crucial role during flower development, probably by means of a conserved cysteine that mediates posttranslational modifications. The spectrum of ROXY1 target genes seems to be rather broad, as indicated by the roxy1 mutants displaying distinctive early and late defects throughout petal development,and by diverse genetic interactions involved in repressing ectopic AGexpression. Characterization of the roxy1 mutants can now serve as a starting point to investigate the function of the plant GRX family in more detail, aiming for an understanding of their biochemical properties, the identification of target proteins and unraveling the intriguing connection between redox regulation and flower development.

We thank Zsuzsanna Schwarz-Sommer, Heinz Saedler and Peter Huijser for comments on the manuscript, and Anja Hörold for assistance in the molecular work. We thank the Salk Institute Genomics Analysis Laboratory for rbe-2 T-DNA insertion mutant seeds (SALK_037010), the Nottingham Arabidopsis Stock Centre for providing mutant seeds, and Rüdiger Simons,University of Düsseldorf, for AG-I::GUS seeds. S.Z. is grateful to Heinz Saedler for ongoing support, advice and stimulating discussions. This work has been initiated in the Arabidopsis Functional Genomics Network which is supported by the Deutsche Forschungsgemeinschaft.