To define the spatial and temporal expression of ARR22 in Arabidopsis, we first performed quantitative RT-PCRs using RNA from different tissues and developmental stages. The data demonstrated that the ARR22 transcript was only detectable in flowers and more predominantly in siliques (Fig. 1A). Subsequently, we translationally fused the ARR22 promoter to the GFP gene. The used 2003 bp long ARR22 promoter fragment covered the entire sequence 1608 bp upstream of the transcription start and 395 bp of the 5' UTR and was expected to contain all significant regulatory elements 3435. Confocal laser scanning microscopy (CLSM) of transgenic Arabidopsis plants carrying this reporter construct indicated that GFP fluorescence was restricted to the chalaza cells of developing seeds (Fig. 1B, C). The activity of the ARR22 promoter was first detectable in the seeds shortly after fertilization (emerging silique) and persisted through all the following developmental stages until the seed was shed. We also fused the ARR22 promoter fragment to the uidA reporter gene and generated P_ARR22::uidA-containing transgenic plants. The GUS staining slightly extended to non-chalaza cells in the developing seeds. However, we could not detect specific GUS activity in organs other than the developing seeds (data not shown; 33). To confirm the specific chalazal expression of ARR22, we performed fluorescence immunolabelling on 300–500 nm cryosections of seeds from elongating siliques (Fig. 1C). The GFP protein was only detected in the chalaza indicating the spatially and temporally restricted expression of ARR22 in Arabidopsis (Fig. 1D). The observed P_ARR22 activity pattern is supported by recent microarray data which demonstrate the accumulation of ARR22 transcript solely in developing seeds 36.

Using the P_ARR22::GFP reporter lines, we also tested the activity of the ARR22 promoter in the chalaza in response to cytokinin and ethylene. For this purpose, we soaked excised inflorescences in either water or solutions containing different concentrations of cytokinin (benzyladenine) or the ethylene precursor 1-aminocyclopropane-1-carboxylic acid. At different time points after hormone application we excised the siliques and analysed the chalaza for its fluorescence intensity. However, no induction of the ARR22 promoter by cytokinin or ethylene was observed (data not shown).

We next examined the subcellular localization of wild type ARR22 and point-mutated versions in which the conserved phosphorylatable aspartate (D74) was changed to either non-phosphorylatable glutamate (E) or asparagine (N). For this purpose, we expressed wild type ARR22 fused to RFP and the modified ARR22^D74E and ARR22^D74N proteins fused to GFP, under the control of the constitutive Cauliflower Mosaic Virus (CaMV) 35S promoter, transiently in tobacco (Nicotiana benthamiana) leaf cells. For the verification and identification of the nuclear and cytoplasmic compartments we used B-type ARR2:GFP (nucleus; 37), AHK5:GFP (cytoplasm and plasmalemma; 38) and GFP (cytoplasm and nucleus) as markers. The leaves were co-infiltrated with Agrobacterium strains carrying either the RFP or GFP constructs and a strain conferring the expression of the silencing inhibitor gene p19 39. After transformation, the abaxial epidermis of the tobacco leaves was subjected to CLSM. We observed a strong RFP fluorescence in the cytoplasm and a weaker signal inside the nucleus indicating that ARR22 is predominantly a cytosolic protein as compared to the nuclear marker ARR2:GFP protein (Fig. 2, RFP). This subcellular localization of ARR22 was not affected by the exchange of Asp74 to Glu or Asn (Fig. 2).

To identify TCS partners of ARR22 potentially involved in a chalaza-based phosphorelay, we screened for protein interactions in the yeast two-hybrid system. We cloned the ARR22 cDNA in the "bait" vector pGBKT7-DEST and the cDNAs of all AHPs (AHP1 to AHP6) as well as the cDNA fragments encoding the histidine kinase (HK) domain of the cytokinin receptors (AHK2 to AHK4) and the ethylene receptor ERS1 in the "prey" vector pGADT7-DEST. The HK domain of the AHKs and ERS1 was used to address the question whether ARR22, which is more similar to the receiver domains of hybrid His kinases than to the receiver region of response regulators 30, could interfere with either the hybrid His kinases or with ERS1, the only plant non-hybrid His kinase. The BD-ARR22 construct was first transformed into the yeast strain PJ69-4A and then a BD-ARR22 expressing clone was re-transfected with the prey AD constructs shown in Figure 3. The interaction of the proteins was assessed by growth assay on interaction-selective media and measuring LacZ reporter gene activity (β-galactosidase) (Fig. 3A, B). Although all AD fusion proteins were expressed (Fig. 3C), only the yeast clones containing the AHP2, AHP3 or AHP5 construct together with BD-ARR22 grew on interaction selective medium and displayed β-galactosidase activity above the background level (Fig. 3A, B). No interaction of ARR22 was found with AHP1, AHP4 and AHP6 and the HK domains of the histidine kinases (Fig. 3). Furthermore, ARR22 did not form homodimers in yeast (Fig. 3).

To substantiate the ARR22 interactions observed in yeast, we studied protein-protein interactions in plant cells using bimolecular fluorescence complementation (BiFC; 40). For this purpose, we generated 35S promoter-driven BiFC constructs expressing ARR22 in fusion with the C-terminal YFP fragment (YFP-C) and the AHP fusions with the N-terminal YFP fragment (YFP-N; Fig. 4). Western-blot analysis confirmed that the fusion proteins were co-expressed in N. benthamiana leaf cells in the presence of the p19 silencing inhibitor (Fig. 4B). As shown in Figure 4A, a significant BiFC signal and, thus, interaction was only observed when ARR22 was co-expressed with AHP2, AHP3 and AHP5. No ARR22 interaction was observed with AHP1, AHP4 and AHP6 (Fig. 4A). Thus, in support of our yeast data, ARR22 showed specific interaction with a subset of AHPs in planta and not with any of the other two-component signalling elements tested.

For the further analysis of the ARR22 function, we identified two T-DNA insertion mutant alleles of the ARR22 gene by screening the Cologne collection, representing 90,000 independent T-DNA insertion lines in the Col-0 ecotype 41. Both mutant lines (arr22-2, arr22-3) contained the T-DNA insertion in the second intron of the ARR22 gene (Fig. 5A). In the arr22-2 mutant the T-DNA was inserted 288 bp downstream of the start codon, whereas in the arr22-3 line it was located 372 bp downstream of the start ATG, corresponding to a position 3 bp upstream of the 3' splicing site of intron 2 (Fig. 5A). After the isolation of homozygous lines, RNA from siliques was tested for the presence of the ARR22 transcript by end-point RT-PCR (Fig. 5B). In contrast to wild type, no ARR22 transcript was detected in either of the insertion mutant lines. The weaker band with slower electrophoretic mobility in the wild type preparation represents an additional ARR22 transcript with an unspliced second intron 33. A contamination with genomic DNA was ruled out by PCR amplification without reverse transcription which did not produce ARR22 amplicons. Compared to wild type, however, both mutant lines showed no aberrant phenotype with respect to their vegetative development.

The function of the chalaza is to support the developing seed with nutrients 42. Therefore, we focused our further analysis on the seeds of the arr22-2 and arr22-3 mutants. However, we found no microscopically detectable difference in the size and morphology of the chalaza cells between the mutants and wild type. Moreover, embryonic development was not affected in arr22-2 and arr22-3, and the seeds also developed normally with respect to their size and mass (data not shown). We next investigated whether the loss of ARR22 function influences the metabolic state of the developing seeds. Siliques from different developmental stages were collected, homogenized and lyophilized. The mutant and wild type siliques exhibited the identical dry mass at each developmental stage tested (Fig. 6A and data not shown). The quantification of inorganic ions using HPLC chromatography, revealed no significant difference between wild type and the arr22 mutants at any developmental stage (Fig. 6B, C and data not shown). Similarly, there was no significant difference with respect to the carbohydrate and and amino acid content (Fig. 7 and data not shown). In conclusion, the loss of ARR22 function in the chalaza has no major detectable impact on silique and seed development and the seeds' nutrient content and composition.

To genetically complement the insertion mutations, we introduced a genomic ARR22 DNA fragment into the arr22-2 and arr22-3 mutant lines. The ARR22 genomic fragment extended from 2003 bp upstream of the ATG start codon to 643 bp downstream of the stop codon. The T1 population of more than 200 arr22-2 and arr22-3 transformants carrying the ARR22 genomic construct (gARR22) showed a broad phenotypic variation in contrast to the plants transformed with the empty vector (Fig. 8). 22% of these transformants were dwarfed plants (category I), 35% had a reduced rosette size (category II) and 43% showed a normal rosette size (category III). All category I and category II plants were sterile. 78% of category III plants were sterile and 5% showed wild type fertility. 17% of category III plants had reduced fertility ranging from a few seeds to a few siliques per plant. The growth of the siliques in these plants was arrested at an early stage of development, and the abscission of the floral organs was delayed (Fig. 9A). A similar phenotype distribution was observed when the gARR22 construct was transformed into wild type Arabidopsis plants (data not shown).

Semi-quantitative RT-PCR using total RNA fom the dwarf and sterile arr22/gARR22 plants revealed that there was an over-accumulation of ARR22 transcript in the apical inflorescence compared to wild type (Fig. 9B). We, subsequently, compared the seed development in arr22/gARR22 transgenic and wild type plants using differential interference contrast microscopy. As shown in Figure 9C, the female gametophyte developed phenotypic identically in the wild type and arr22/gARR22 transgenic plants. In contrast to the wild type, the fully developed female gametophyte persisted longer in the arr22/gARR22 transgenic plants, the central nucleus did not divide and both synergids degenerated at the time when the floral organs were abscissed (Fig. 9C). However, the few seeds, which were occasionally produced in the siliques of arr22/gARR22 plants, were normal. In conclusion, our data suggest that a block of the fertilization, rather than a defect in early embryonic development, is the reason for arr22/gARR22 plants' sterility.

The observed fertilization block was not due to a difference in pollen tube attraction because wild type and arr22/gARR22 transgenic plants attracted GUS-stained wild type pollen tubes to an identical extent (data not shown). Reciprocal crossing experiments, however, revealed that the fertilisation of emasculated wild type and arr22/gARR22 plants using mature anthers from arr22/gARR22 plants was not successful (Table 1). The reason for this infertility was the anthers' physical resistance and reduced ability to open and to release the pollen rather than the sterility of the pollen (data not shown). As ARR22 is normally not expressed in anthers 36, the observed phenotype does not provide information with respect to its native function.

To address the question, whether ARR22 functions as a canonical response regulator or a histidine phosphatase in planta, we generated transgenic arr22 lines expressing ARR22 variants, in which the phosphorylatable Asp74 was mutated either to Glu or Asn. A Glu mutation mimicks a phosphorylated Asp and generates a dominant-active variant of a response regulator, whereas an Asp-to-Asn mutation mimicks a non-phopshorylated version, which acts dominant-negatively. The more than 200 independent arr22 plants carrying either the gARR22^D74E or gARR22^D74N construct did not show any developmental abnormalities but displayed the arr22 phenotype (Table 1; Fig. 8). These findings indicate that ARR22 does not act as a canonical response regulator in planta.

By using a P_ARR22::GFP construct in combination with real-time RT-PCR and GFP immunolabelling, we located ARR22 gene activity to the chalaza cells of developing Arabidopsis seeds. This pattern is in general agreement with observation of Gattolin and colleagues 33, who also reported ARR22 expression in the funiculus of developing seed using a P_ARR22::GUS reporter line. This difference in localisation could be due to a diffusion of the indigo stain out of the chalaza into the funiculus which we avoided in our transgenic line by using the GFP marker. Our intracellular localisation studies revealed that 35S promoter-expressed ARR22:GFP is predominantly but not exclusively located in the cytoplasm of plant cells. The mutation of the phosphorylatable Asp74 to Glu or Asn had no effect on this pattern suggesting that the phosphorylation of ARR22 does not influence its intracellular distribution.

Recently, Kiba and colleagues 30 reported that ARR22 has strong phospho-histidine phosphatase activity on phosphorylated AHP5 in vitro. Here we addressed the question whether ARR22 may target AHP5 only or whether it also may interact with other Arabidopsis TCS elements. Yeast two-hybrid data showed that ARR22 specifically interacts with AHP2, AHP3 and AHP5 but not with AHP1, AHP4 and AHP6 and the His kinase domains of AHK2, 3, 4 and ERS1. As indicated by the quantitative data, the interaction appears to be of low affinity and/or high kinetical dynamics. The in planta BiFC data substantiated the specific interaction of ARR22 with AHP2, AHP3 and AHP5 and suggest that ARR22 exerts its phospho-histidine phosphatase activity not only on AHP5 30 but probably also on AHP2 and AHP3. AHP2, AHP3 and AHP5 are expressed in siliques and developing seeds 3643 and are reported to be particularly important for Arabidopsis seed development 7.

We identified two novel arr22 mutant alleles and proved the absence of the ARR22 transcript in their siliques. In comparison to the wild type, we could not detect any difference in the number, size and morphology of the chalaza cells during embryo development and we found no difference in seed size and silique mass in the mutants. A detailed metabolic analysis of ions, amino acids and carbohydrate content in the seeds also revealed no significant difference between arr22 mutant and wild type at any stage of seed development. These data show that ARR22 is not involved in these processes. In the view of the extremely restricted ARR22 expression in the chalaza and its phospho-histidine phosphatase activity, the lack of any detectable morphological and metabolic phenotype in the seeds of the arr22 mutants is surprising. A functional replacement of ARR22 by its paralog, ARR24, in the chalaza is unlikely because ARR24 gene activity is only observed in pollen and an arr22/arr24 double mutant also showed no phenotypic defects 33.

The complementation of the arr22-2 and arr22-3 mutants with a genomic ARR22 fragment (gARR22) resulted in T1 transformants with severe developmental abnormalities, such as dwarfism, delayed abscission of flower organs and a pollen release defect resulting in sterility. The inability of the arr22/gARR22 anthers to release their pollen suggests a defect in the differentiation of the functional endothecium which is responsible for the rupture of the pollen sacs. The appearance of this pleiotropic phenotype depended on the presence of the phosphorylatable Asp74 30 in gARR22 and was not observed in arr22-2/gARR22^D74N and arr22-3/gARR22^D74E transgenic plants. This finding indicates that both the gain-of-function (ARR22^D74E) and the loss-of-function mutants (ARR22^D74N) of the response regulator have lost their biological activity. If ARR22 functions as a canonical phospho-regulated response regulator like the A-type and B-type ARRs, we would have expected to identify plants within the arr22/gARR22^D74E T1 transformant population showing a pleiotropic phenotype comparable to that of the arr22/gARR22 T1 transformants. The absence of such a phenotype in the arr22/gARR22^D74E plants suggests that ARR22 acts as a phospho-histidine phosphatase exclusively. This genetic finding is in accordance with the biochemical evidence that although ARR22 is able to remove the phosphoryl residue from AHP5, it does not use it for autophosphorylation 30.

The pleiotropic arr22/gARR22 phenotype is reminiscent to that of wol allele in AHK4/CRE1 gene 4445, ahk2/ahk3/ahk4 triple cytokinin receptor mutants 1046, some higher order ahp 7 and B-type arr mutants 26. From these data we suggest that even small traces of ARR22 in tissues, which normally do not contain the response regulator, may interfere with the hormone homeostasis and especially the cytokinin response pathway (Fig. 10). Mis-expression of ARR22 appears to have a particular negative effect on vegetative growth and possibly on the differentiation of connecting tissues such as the endothecium of the anthers. The observed dramatic developmental defects can be explained by the action of ARR22 as a very efficient AHP phosphatase and, therefore, as a strong phosphate "sink", which disturbs the two-component phosphorelay network in the extrinsic tissues (Fig. 10). Similarly, the enhanced phosphatase activity of the AHK4 receiver domain in the wol mutant is responsible for the wol phenotype 47 (Fig. 10).

An ectopic mis-expression of ARR22 in the frame of a genomic fragment can occur when the transgene is inserted in transcriptionally active, more-or-less open chromatin areas, especially in 5'-promoter regions – a process which is favoured by the selection of the transgenic plants 4849. This suggests that the activity of the ARR22 locus might be under the tight control of chromatin condensation mechanisms, which restrict the expression of the response regulator to the chalaza. A similar regulation of gene activity is known for many tightly controlled developmental transformations during the plant's life cycle 505152.

Tobacco (Nicotiana benthamiana) and Arabidopsis thaliana (Col-0) plants were cultivated in the greenhouse (tobacco, temperature: day 25°C/night 19°C, humidity 60%, photoperiod: 14 h; Arabidopsis, temperature: 21°C, night 18°C, humidity 45%, photoperiod: 14 h).

T-DNA mutants were isolated according to Ríos and colleagues 41 from the collection at Max-Planck Institute in Cologne. The insertion alleles arr22-2 and arr22-3 were identified with a T-DNA left border primer (5'-CTACACTGAATTGGTAGCTCAAACTGTC-3') and ARR22 specific primers (forward: 5'-CCTCGTTCTATACTAGCAGATGGGTTCGAT-3', reverse: 5'-GACTTGCATGATTTTACCTCGGACGATAAG-3') PCR products containing the T-DNA/gene junctions were cloned into pCR4-TOPO vector (Invitrogen) and sequenced.

RNA for RT-PCR was isolated using RNaqueous^®Kit (Ambion) and traces of genomic DNA were removed by TURBO DNA-free™ (Ambion). Subsequently, 1.5 μg of total RNA were reverse transcribed using oligo-dT primer with RevertAid™ H Minus M-MuLV Reverse Transcriptase (Fermentas) and resulting cDNA was used as a template for PCR with HotStart Taq polymerase (Genaxxon). PCR products were separated via agarose gel electrophoresis after different number of cycles for comparison with ACTIN2 at non-saturating conditions 55. The sequences of the RT-PCR primers were as follows: ACT2 detF 5'-CTGCTCAATCTCATCTTCTTCC; ACT2 detR 5'-GACCTGCCTCATCATACTCG; ARR22 detF 5'-AACGATCGGAGGAATTTCTCAGACT-3'; ARR22 detR 5'-CGCTCTTCTTCTTGGTCAGCTACTGA.

For quantification of ARR22 transcript, RNA and cDNA was prepared as described above with addition of random primers for reverse transcription and used as a template for qRT-PCR with Platinum Quantitative PCR SuperMix-UDG (Invitrogen) in three replicas. TaqMan^®Probes in combination with ABI PRISM 7700 (Applied Biosystems) detection system according to manufacturer recommended protocol were used to monitor the amplification. To normalize the expression data, 18S rRNA was quantified using Eukaryotic 18S rRNA Endogenous Control (VIC/TAMRA Probe) kit (Applied Biosystems) and the amplifacation efficiency of ARR22 and 18S rRNA was proved to correspond. Sequences of ARR22 oligonucleotides could be sent upon request.

All clones used in our experiments were constructed using Gateway™ technology (Invitrogen). The Entry clones were either obtained via BP-reaction in pDONR201 or pDONR207 or through TOPO-reaction using the pENTR/D-TOPO vector (CTR1, AHK5; all vectors Invitrogen). The templates used to clone all genes were cDNA preparations derived from Arabidopsis roots, leaves or siliques. To avoid spontaneous mutations, the Entry clones of AHK2, AHK3, AHK4 and AHK5 were cloned and propagated in the E. coli strain CopyCutter™ (Epicentre). The reverse primers contained no stop codon to enable C-terminal fusions or a stop codon was included for N-terminal fusions. Sequences of forward and reverse primers could be sent upon request. Construction of clones containing only histidine kinase domains was performed using the full-length Entry clones of the AHKs as templates and following primers: AHK2^HK (aa562-873) – 5'-attB1-TAATGAACCGAATTGCGACAGTTGAAGAG and 5'-attB2-TTACGACGTATTTGTTTCTGCTTTCC; AHK3^HK (aa425-728) – 5'-attB1-TAATGAGTCGAATACACAAAGTTGAAGAA and 5'-attB2-TTAAGCTGGTTGCATCCCATTGGAAAATAC; AHK4^HK (aa424-743) – 5'-attB1-TAATGCACATAGTAAAAGTCGAAGATGAT and 5'-attB2-TTACGCACTGCATTTATCGCATTTCTCTAA; ERS1^HK (aa317-613) – 5'-attB1-TAATGCACGCTCGTGACCAGCTTATG and 5'-attB2-TCACCAGTTCCACGGTCTGGTTTGTGA. Site-directed mutagenesis of ARR22 was carried out on the ARR22 Entry clones using QuikChange^® Site-Directed Mutagenesis Kit (Stratagene) and a pair of fully complementary primers for the designed region GGCGAAGCTAGCTTCGACCTTATTCTAATGGAGAAGG containing D74E mutation and NheI restriction site or CGACCTTATTCTCATGAATAAGGAAATGCCTGAG containing D74N mutation and PagI restriction site.

The binary vectors for expression of the GFP/RFP fusion proteins under the control of 35S promoter were constructed via LR-reaction using the corresponding Entry clones and destination vectors pH7FWG2.0, pH7WGF2.0 and pB7RWG2.0 56. The p19 protein from tomato bushy stunt virus cloned in pBIN61 57 was used to suppress gene silencing. All vectors were transformed in Agrobacterium tumefaciens strain GV3101 pMP90 and prior infiltration resuspended in AS-medium (10 mM MgCl2, 150 μM acetosyringone and 10 mM MES pH 5.7) to OD₆₀₀ 0.8 according to Grefen and colleagues 58 Abaxial epidermis of infiltrated tobacco leaves was assayed for fluorescence by CLSM 2–3 days post infiltration. For BiFC the AHP cDNAs were recombined via LR-reaction into pSPYNE-35S and ARR22 into pSPYCE-35S 40 and infiltrated to the tobacco leaves as described elsewhere 58. The expression of the BiFC fusions was determined by SDS-PAGE and western blot using the tissue of transfected leaves according to Walter and colleagues 40.

Yeast two-hybrid experiments were performed using the Matchmaker™System (Clontech). Plasmids were constructed by LR-reaction of corresponding Entry clones and destination vectors pGBKT7-DEST or pGADT7-DEST. Yeast strain PJ69-4A was transformed using lithium acetate/SS-DNA/PEG method 59. After 3 days of growth on vector selective media (CSM-L^-, W^-), 5 independent clones were pooled and propagated to an OD₆₀₀ 1.0 in liquid, vector-selective media. Subsequently, 7.5 μl of culture were dropped on vector- and interaction-selective media (CSM-L^-, W^-, Ade^-) and incubated at 28 C. At day 2 the growth of the clones was monitored. In addition, liquid yeast cultures were harvested and analyzed by western-blot to determine the correct expression of the fusion proteins 59. Activity of β-galactosidase activity was measured using 3 independent yeast clones per transformation. 50 ml yeast culture was grown in vector-selective media to OD₆₀₀ 2.0, harvested and lysed using freeze-thaw method and vortexing with glass beads in 700 μl of Z-buffer (60 mM Na₂HPO₄, 40 mM NaH₂PO₄, 10 mM KCl, 1 mM MgSO₄, pH 7.0). The total protein amount of the lysate was quantified by Bradford protein assay and β-galactosidase activity measured according to Grefen and colleagues 59.

ARR22 promoter DNA fragment was amplified from A. thaliana genomic DNA using Phusion™ polymerase (Finnzymes) and the primers ARR22pF 5'-attB1-GCAGCAGATGACTTAACTCTCCA, ARR22pR 5'-attB2-GGGTACCTCCGGTGGATTTTGT. The promoter fragment was cloned into pDONR207 via BP-reaction and verified by sequencing. The fragment comprises the region 2003 bp upstream of the start codon including the first 27 bp of the ARR22 coding sequence. The fragment was translationally fused to the GFP reporter gene via LR-reaction in the destination vector pMDC107 60, and the construct transformed in the Agrobacterium tumefaciens strain GV3101 pMP90. Arabidopsis plants were transformed using the flower-dipping method. For selection of transgenic plants, seeds were vapor sterilized as described in Grefen et al 58 and T1 seedlings selected for hygromycin resistance in a growth chamber at 21°C under long-day conditions for 14 days on 0.5× Murashige-Skoog medium supplemented with 1% (w/v) sucrose, 0.8% (w/v) phytoagar and 25 μg/ml hygromycin B. Resistant plants were propagated on soil and siliques from several independent transgenic lines (T2) were tested for fluorescence by CLSM.

For GFP imunolabelling, seeds were fixed for 90 min in 4% (v/v) formaldehyde in MTSB, pH 7.0, followed by a treatment with 8% (v/v) formaldehyde in MTSB for additional 90 min, then infiltrated with a mixture of sucrose and polyvinyl pyrrolidone 61 and frozen in liquid nitrogen. 300–500 nm cryo-sections were cut at -80°C, mounted on coverslips, labelled with a rabbit anti-GFP antibody (diluted 1:250 in blocking buffer; Abcam) and a goat anti-rabbit antibody coupled to Cy3 (diluted 1:400 in blocking buffer; Dianova) for 1 h, each. Nuclei were stained with DAPI. Cryo-sections were embedded in Moviol for microscopy. In the control, the rabbit anti-GFP antibody was omitted. Labeled cryo-sections were viewed with a Zeiss Axiophot light microscope (63×/1.4 oil immersion objective).

For sample preparation as well as for analytics Milli-Q water and chemicals in highest purity were used exclusively. The siliques were collected in five replicas for each line at the developmental stage, where the first senescing siliques appeared on the inflorescences, frozen in liquid nitrogen, homogenized and lyophilized. To 50 mg lyophilized sample 750 μl of a chloroform/methanol/water [5/15/3 (v/v/v)] mixture were added, mixed and afterwards centrifuged for 5 minutes at 14.000 rpm and 4°C. 400 μl of the supernatant were transferred into a new vial and pooled with another 400 μl obtained by similar treatment of a second replicate of the same sample. To the 800 μl pooled supernatant 8 μl of a 2.5 mM norleucin solution were added as internal standard for the amino acid analysis. After thorough mixing 376 μl water and 250 μl chloroform were added, mixed and afterwards centrifuged as described before. 900 μl supernant were collected without touching the interface, evaporated to dryness and redissolved in a mixture of 400 μl water and 600 μl acetonitrile. Undissolved residues were removed by centrifugation (30 seconds,14.000 rpm, 4°C). The supernatant was transferred into a new vial and again evaporated to dryness. The pellet was redissolved in 400 μl water. Further workup of the sample was done by solid phase extraction (SPE) with Sep-Pac C₁₈ columns (Waters). For this SPE columns were washed with 1 ml acetonitrile followed by 1 ml water for column equilibration. Then the sample was loaded onto the SPE column and eluted with 500 μl water. Vacuum was used to collect the entire liquid of the column. Resulting 900 μl eluate were evaporated to dryness afterwards. For amino acid analysis the resulting pellet was dissolved in 100 μl lithium diluent. For sugar and ion analytics pellets were dissolved in 100 μl water. If necessary, storage of samples was done at – 20°C. Before analysis samples were filtered through a 0.5 μm PVDF filter. For amino acid analysis ninhydrin post column derivatization technology was used. The modular HPLC system consisted of two high pressure pumps, autosampler, and UV/Vis – detector from Kontron/BioTek. For post column derivatization Pickering Laboratories ninhydrin pump, column oven and reactor were used. UV/Vis detection of the amino acid derivatives was done at 440 and 570 nm. For amino acid separation the lithium cation exchange column from Pickering Laboratories (4 × 100 mm, particle size: 5 μm) was used with the following ternary gradient at a flow rate of 340 μl/min. 0 min: 100% A (96.0% H₂O, 0.7% lithium citrate, 0.6% lithium chloride, pH 2.8) 12 min: 100% A 48 min: 35% B (97% H₂O, 0.9% Lithium citrate, 2.0% lithium chloride, pH 7.5) 90 min: 100% B 95 min: 100% B 120 min: 94% B and 6% C (99% H₂O, 0.6% lithium chloride, 0.4% lithium hydroxide, pH 11.8) 150 min: 94% B and 6% C. For Ion analysis a Dionex DX120 system equipped with suppressed conductivity detection was used in isocratic mode. For cation separation an Ion Pac-CS12A column (4 × 250 mm) was eluted with 20 mM methansulfonic acid at a flow rate of 1 ml/min. Anions were separated on an Ion Pac AS 9-HC column (4 × 250 mm) with 9 mM sodium carbonate as mobile phase at a flow rate of 1 ml/min. For carbohydrate analysis a Kontron/BioTek HPLC system equipped with an electrochemical detector (Bischoff) was used for high pH anion exchange chromatography (HPAEC). Isocratic separation was done on a CarboPac PA1 column (4 × 250 mm) from Dionex. For elution a 30 mM sodium hydroxide solution was used at a flow rate of 0.80 ml/min.

A DNA fragment containing the ARR22 gene was amplified from Arabidopsis genomic DNA using Phusion™ polymerase (Finnzymes). The forward primer used previously for the cloning of the ARR22 promoter was combined with the reverse primer ARR22gR (5'-attB2-CCCTATCTTCTTAATGACATATAGTATTGG-3'), and the obtained DNA fragment comprising the region 2003 bp upstream of the ATG start codon and 643 bp downstream of the stop codon was cloned into pDONR207 via BP-reaction and verified by sequencing. Site-directed mutagenesis of Asp74 was carried out on the Entry clone as described above. All DNA fragments were recombined via LR-reaction into destination vector pMDC123 60. The arr22-2 and arr22-3 lines were transformed as described above and transgenic seedlings selected on soil by spraying with 0.05% (v/v) BASTA. Transgenic T1 plants were sorted to different phenotypic categories, photographed and siliques or upper parts of inflorescences were collected and tested for the ARR22 transcript amount by RT-PCR (see above). For microscopy of developing seeds, the inflorescences were submerged in ethanol:acetic acid (3:1) fixation solution for several hours. Subsequently, the tissue was rehydrated for 30 min in a series of 70%, 50% and 20% (v/v) ethanol followed by 20 min incubation in a clearing solution consisting of chloral hydrate : water : glycerol mixture (8:3:1 w/w). Seeds were excised from the siliques and mounted on slides in clearing solution for differential interference contrast microscopy.

CLSM was performed using a Leica TCS SP2 confocal microscope (Leica Microsystems GmbH). All CLSM images were obtained using Leica Confocal Software and the HCX APO LW 20×/0.5 or the HCX PL APO 63×/1.2 W water-immersion objective. GFP and RFP channels were acquired by simultaneous scanning using 488-/568-nm laser lines for excitation. The signals were detected between 500–530 nm for GFP and 590–630 nm for RFP. YFP signal was detected between 550 and 580 nm after excitation by 488 nm laser. Microscopy of developing seeds was carried out using a Nikon Eclipse 90i microscope equipped with a 40×/0.75 objective and a CCD camera applying differential interference contrast. The images were acquired using MetaMorph software (Molecular Devices) and processed using Adobe Photoshop 9.0.