An Iron-induced Nitric Oxide Burst Precedes Ubiquitin-dependent Protein Degradation for Arabidopsis AtFer1 Ferritin Gene Expression*

Ferritins play an essential role in iron homeostasis by sequestering iron in a bioavailable and non-toxic form. In plants, ferritin mRNAs are highly and quickly accumulated in response to iron overload. Such accumulation leads to a subsequent ferritin protein synthesis and iron storage, thus avoiding oxidative stress to take place. By combining pharmacological and imaging approaches in an Arabidopsis cell culture system, we have identified several elements in the signal transduction pathway leading to the increase of AtFer1 transcript level after iron treatment. Nitric oxide quickly accumulates in the plastids after iron treatment. This compound acts downstream of iron and upstream of a PP2A-type phosphatase to promote an increase of AtFer1 mRNA level. The AtFer1 gene transcription has been previously shown to be repressed under low iron conditions with the involvement of the cis-acting element iron-dependent regulatory sequence identified within the AtFer1 promoter sequence. We show here that the repressor is unlikely a transcription factor directly bound to the iron-dependent regulatory sequence; such a repressor is ubiquitinated upon iron treatment and subsequently degraded through a 26 S proteasome-dependent pathway.

As the major cofactor of proteins involved in essential processes like photosynthesis, respiration, DNA replication, or nitrogen fixation, iron is an essential element for life. Nonetheless, in the free ionic form, iron is toxic as it can catalyze the formation of reactive oxygen species through the Fenton reaction. These reactive oxygen species damage the cell membranes, DNA, and proteins (1,2). Thus, iron homeostasis has to be tightly regulated, to avoid starvation that impairs the metabolism, and to avoid excess that may lead to cell death. Iron homeostasis is strongly dependent on ferritins, which are iron-storage proteins, found in bacteria, animals, and plants. Plant and animal ferritin structures are very similar, and are formed by 24 subunits arranged to form a hollow sphere able to sequester iron in a non-toxic and bioavailable form (3).
In animals, ferritin synthesis is mainly regulated at the posttranscriptional level (3,4). Ferritin mRNAs contain iron-responsive elements in their 5Ј-untranslated regions that function as binding sites for two related trans-acting factors, namely iron regulatory proteins IRP1 and IRP2. When bound to the iron-responsive element in the ferritin mRNA, the IRP inhibit translation of the transcript (4). IRP1 is a bifunctional protein that when iron is abundant possesses a 4Fe-4S cluster and acts as cytoplasmic aconitase. When iron levels are low, the 4Fe-4S cluster disassembles and the apoprotein acquires IRP 3 activity, thus repressing ferritin translation. High levels of iron lead to the 4Fe-4S cluster reconstitution and therefore the protein aconitase activity. In contrast to IRP1, IRP2 cannot assemble a iron-sulfur cluster and lacks aconitase activity. IRP2 shares about 60% amino acid sequence identity with IRP1, but differs only in having a 73-amino acid insertion in its N-terminal region. This region contains a cysteine-rich sequence responsible for targeting the protein for degradation via the ubiquitinproteasome pathway when cellular iron level is high (5,6). NO has been shown to play an important role in iron metabolism by modulating both IRP1 and IRP2 activities (7,8). Exposure to NO ⅐ was shown to disassemble the iron-sulfur cluster of IRP1, promoting binding to ferritin mRNA (4,9). By contrast, IRP2 binding to iron-responsive elements is negatively regulated by NO (10 -12). An oxidized form of NO, the nitrosonium ion NO ϩ (11,13) may cause the S-nitrosylation of a cysteine found in the Fe 2ϩ -dependent degradation domain of IRP2, leading to a subsequent and specific down-regulation of IRP2 by the ubiquitin/26 S proteasome pathway (14,15).
plasm, and their main location is in the plastids (17). In addition, their synthesis is regulated at the transcriptional level in response to iron excess (17,18), and not at the translational level as described above for animal cells. In plants, ferritin mRNA abundance has been shown to be regulated by several environmental factors including iron (17, 19 -21), H 2 O 2 (22), photoinhibition (23), pathogen attacks (24), by the stress hormone ABA (25), and by NO donors or scavengers (26,27). Experiments based on serial deletions and site-directed mutagenesis of maize ZmFer1 and Arabidopsis AtFer1 ferritin promoter sequences allowed to identify a 15-bp cis-acting element necessary for the iron-dependent regulation of the transcription of these genes (18). This sequence, named IDRS, for iron-dependent regulatory sequence, has been shown to be involved in the repression of ZmFer1 and AtFer1 gene expression under iron-deficient conditions (18,28). Thus, iron addition leads to the de-repression of ZmFer1 and AtFer1 gene expression rather than to their induction. Despite the growing number of physiological conditions reported to date leading to plant ferritin synthesis, little is known about the regulatory molecules acting downstream of iron. By using an Arabidopsis cell culture system, we show in this work that iron excess and oxidative stress, mimicked by exogenous H 2 O 2 application, promote AtFer1 gene expression through two independent and additive pathways. We show also that iron application leads to a rapid NO burst in the plastids of the cell. This NO accumulation, which does not involve NOS1 nor nitrate reductase activities, is leading to AtFer1 de-repression. The factor that represses AtFer1 transcription under iron-deficient conditions is ubiquitinated and degraded by a 26 S proteasome-dependent pathway after iron application. This repressor is not a transcription factor directly bound to the IDRS present in the AtFer1 promoter region.  (29), and gЈ4-3 (30) mutants were surface-sterilized by immersion in a 4% (w/v) Bayrochlor, 50% ethanol solution for 20 min. Seeds were washed three times with ethanol and left to dry in sterile conditions. Seedlings were grown in 100 ml of half-strength Murashige and Skoog medium (Sigma), pH 5.7, supplemented with 1% sucrose, 0.5 g liter Ϫ1 MES, and 50 M Fe(III)-EDTA. After 1 week of culture at 24°C under continuous light (100 E m Ϫ2 s Ϫ1 ) and shaking (60 rpm), medium was discarded and replaced by 100 ml of fresh medium. Plants were grown 4 additional days in these conditions before treatments.

EXPERIMENTAL PROCEDURES
Chemicals-One volume of a 100 mM FeSO 4 stock solution in 0.06 M HCl was mixed with 1 volume of 200 mM Na 3 -citrate for a concentration of 50 mM FeSO 4 , 100 mM Na 3 -citrate. This mixture was used at final concentration of 300 M FeSO 4 , 600 M Na 3 -citrate in the culture medium. Except where indicated, all chemicals were purchased from Sigma. Okadaic acid and cycloheximide were dissolved in ethanol and used at final concentrations of 250 nM and 100 M, respectively. MG132 was dissolved in Me 2 SO and used at a final concentration of 50 M. Pefabloc (Roche Applied Science), cPTIO, L-NMMA, and SNP were dissolved in sterile water and used at final concentrations of 100 M, and 1, 5, and 2.5 mM, respectively. After treatments, cells were filtered or plantlets were collected and immediately frozen in liquid nitrogen and stored at Ϫ80°C.
Microscopy-NO imaging was performed by using 4-amino-5-methylamino-2Ј,7Ј-difluorofluorescein diacetate (DAF-FM DA, Molecular Probes) dissolved in Me 2 SO at a stock concentration of 5 mM. For confocal laser-scanning microscopy, cells were loaded with 5 M DAF-FM DA for 20 min. Then a solution containing 300 M FeSO 4 , 600 M Na 3 -citrate or 300 M K 2 SO 4 , 600 M Na 3 -citrate was added. The cell suspension (30 l) was transferred on the slide. After overlaying by the glass cover, the slides were placed under the microscope, and the images were taken within 5 min on the same cells for each treatment. Settings and laser of the Zeiss Axiovert 100M inverted microscope were as described previously (31). Microscope, laser, and photomultiplier settings were held constant during the course of an experiment to obtain comparable data. Images were processed and analyzed using the Zeiss LSM 510 software.
RNA Preparation and Analysis-Total RNA were extracted from cells and plantlets as indicated in Ref. 25. For Northern blot analysis, 10 g of total RNA were loaded in each lane, separated by electrophoresis through a 1.2% (w/v) agarose/ formaldehyde gel, and blotted onto a nylon membrane (Hybond N; Amersham Biosciences). Hybridizations with 32 Plabeled probes were performed overnight at 42°C in the presence of 50% formamide (32). After washes, filters were exposed for a few hours at Ϫ80°C to Fuji Medical X-Ray film Super RX (Fujifilm) with an intensifying screen. AtFer1 mRNA relative abundance was determined by measuring hybridization signal intensities of AtFer1 and EF1␣ on the same blot. Quantifications were performed with the Imager Reader Bas-5000 software (Fuji). The AtFer1 mRNA relative abundance was defined as the ratio of AtFer1 and EF1␣ signal intensities.
Protein Preparation and Analysis-Total protein extracts were prepared from 1 g of each sample as described (33). Protein concentration was determined according to Schaffner and Weissmann (34) using bovine serum albumin as standard. Proteins were subjected to electrophoresis on a 13% polyacrylamide, 0.1% SDS gel according to Laemmli (35). After electroblotting onto Hybond-P membrane (Amersham Biosciences), immunodetection of ferritin was performed using a rabbit polyclonal antiserum raised against purified AtFer1p (24) and the Aurora Western blotting kit (ICN) following the manufacturer's recommendations.
Preparation of Nuclear Extracts-All procedures were carried out at 4°C. Frozen cells (50 g) were ground in a Waring Blender in 300 ml of homogenization buffer (250 mM sucrose, 10 mM NaCl, 25 mM Pipes, 5 mM EDTA, 0.15 mM spermine, 0.5 mM spermidine, 20 mM ␤-mercaptoethanol, 0.1% Nonidet P-40, and 0.2 mM phenylmethylsulfonyl fluoride, pH 7.0). The homogenates were filtered through two layers of Miracloth (Calbiochem). Nuclei were recovered by centrifugation at 4,200 ϫ g for 20 min at 4°C, then were gently resuspended, and washed four times with homogenization buffer with subsequent centrifugations at 2,000 ϫ g for 10 min, then at 1,500 ϫ g for 10, 8, and 6 min. Nuclei were resuspended in a minimum volume of freezing buffer (100 mM NaCl, 50 mM Hepes, 10 mM KCl, 5 mM MgCl 2 , 1 mM DTT, 0.5 g ml Ϫ1 leupeptin, and 50 g ml Ϫ1 antipain, 50% glycerol, pH 7.6), frozen in liquid nitrogen, and stored at Ϫ80°C until use. Nuclear extracts were prepared by thawing nuclei on ice and lysing by adjusting the NaCl concentration to 0.47 M with lysing buffer (2.5 M NaCl, 50 mM Hepes, 10 mM KCl, 5 mM MgCl 2 , 1 mM DTT, 0.5 g ml Ϫ1 leupeptin, 50 g ml Ϫ1 antipain, 20% glycerol, pH 7.6), and then shaking at 4°C for 30 min. Chromatin was pelleted by centrifugation at 13,000 ϫ g for 15 min, and the supernatant containing nuclear proteins was dialyzed for 4 h against dialysis buffer (20 mM Hepes, 40 mM NaCl, 0.2 mM EDTA, 1 mM DTT, 20% glycerol, pH 7.6). Nuclear proteins were concentrated with centrifugal filter devices (Amicon, Ultracel 10k). Protein concentration was determined according to Schaffner and Weissmann (34) using bovine serum albumin as standard. Nuclear extracts were frozen in liquid nitrogen, and stored at Ϫ80°C.
DNA Probes and Labeling Reactions-The specific probe used for AtFer1 detection consists in a chimeric fragment containing the 5Ј-and 3Ј-untranslated regions of the AtFer1 cDNA. The 5Ј-untranslated region was amplified with thermostable Pfu DNA polymerase (Promega) using primers 5Ј-GGTACCTATATA-AACCCTTCCTCCTTCC-3Ј and 5Ј-GAATTCCATCGGCAT-GTTGTTTGTGTCC-3Ј introducing KpnI and EcoRI sites at the 5Ј and 3Ј ends of the amplified fragment, respectively. The 3Ј-untranslated region was amplified using primers 5Ј-GAACTA-GAATTCGACCTCTATAAG-3Ј and 5Ј-TAGAAACTAGTA-AAACAAAAACTTCATTG-3Ј introducing EcoRI and SpeI sites at the 5Ј and 3Ј ends of the amplified fragment, respectively. The fragments were cloned at the corresponding sites in pBluescript (Stratagene) and sequenced. The specific probe (469 bp) was obtained by digesting the resulting construct by KpnI and SpeI. The EF1␣ probe (550 bp) was obtained by amplification with the Pfu DNA polymerase on the EF1␣ cDNA using the forward 5Ј-CCACCACTGGTGGTTTTGAGGCTGGTATC-3Ј and reverse 5Ј-CATTGAACCCAACGTTGTCACCTGGAAG-3Ј primers. The resulting fragment was cloned at the EcoRV site of pBluescript and sequenced. The probe was obtained after digestion of the plasmid with BamHI and HindIII. The fragments were purified on agarose gel prior to labeling. Probes were labeled with [␣-32 P]dCTP with the use of Prime-a-Gene Labeling kit (Promega).
For the gel shift experiments, DNA probes were amplified from genomic DNA with Pfu DNA polymerase and different primers introducing a BamHI site at the 5Ј end of the amplified fragment and a XhoI site at the 3Ј end. Amplified fragments were cloned in pBluescript at the corresponding sites and sequenced. For probes A, B, C, and D (see the location of the amplified fragments on Fig. 6), the primer located at the 5Ј end of the amplified fragment is 5Ј-GGATCCGAGCGAGTAGGA-AATA-3Ј. At the 3Ј ends, the primers were 5Ј-CTCGAGAAA-GGCGTGTGGTCACCGTTGG-3Ј, 5Ј-CTCGAGCCGTTGG-ATTGAGATCC-3Ј, 5Ј-CTCGAGTGGATATGAAAGCCAG-ATGT-3Ј, and 5Ј-CTCGAGGATAGTGTGAACTGTGAG-3Ј for probes A, B, C, and D, respectively. The probe E was obtained with primers 5Ј-GGATCCCAGATTTACACGTCT-AACTT-3Ј and 5Ј-CTCGAGCATCTCTCCAAATAAAGTT-TGTCC-3Ј. For labeling and competitions, the fragments were obtained by digestion of the corresponding plasmids by BamHI and XhoI and subsequent purification on agarose gel. For labeling, 100 ng of DNA was introduced in a medium containing 50 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , 1 mM DTT, 50 g ml Ϫ1 bovine serum albumin, dATP, dGTP, dTTP (600 M each), 50 Ci of [ 32 P]dCTP and 10 units of Klenow fragment. After 30 min at room temperature, DNA was purified by phenol/chloroform extraction and ethanol precipitation. After a 15-min centrifugation at 10,000 ϫ g at 4°C, the pellet was washed twice with 70% ethanol, dried, and re-suspended in 50 l of water. Specific activity of the probe was determined by scintillation counting, and the probe was diluted to 20,000 cpm/l.
Mobility Shift Assay-The mobility shift reaction was done in a volume of 30 l using 1 l of 32 P-labeled DNA fragment, 2 g of poly(dI-dC), and 5 g of nuclear protein in the fixation buffer (25 mM Hepes-KOH, 70 mM KCl, 1 mM EDTA, 1 mM DTT, 50% glycerol, pH 7.6). The binding reaction was performed for 30 min at room temperature prior to loading reactions onto 6% polyacrylamide non-denaturing gel in 45 mM Tris, 45 mM boric acid, 0.5 mM EDTA, 5% glycerol, pH 8.0. The gel was run at 120 V in 45 mM Tris, 45 mM boric acid, 0.5 mM EDTA, pH 8.0, buffer for ϳ6 h. After migration, the gel was dried during 2 h under vacuum at 80°C and exposed for a few hours at Ϫ80°C to Fuji Medical X-Ray film Super RX (Fujifilm). For the binding competition assays, a 50-fold molar excess of unlabeled fragments was included in the reaction.

Iron and H 2 O 2 Increase AtFer1 mRNA Abundance by Two
Independent and Additive Pathways-Both maize ZmFer1 and Arabidopsis AtFer1 ferritin mRNA accumulates in response to iron treatment. This response, dependent of the cis-acting element IDRS (18), is antagonized by antioxidants like N-acetylcysteine and GSH (20,22), indicating that an oxidative step is involved in the pathway leading to the iron-dependent ferritin mRNA increase in abundance. Furthermore, it has been previously reported that H 2 O 2 treatment increases maize ZmFer1 and Arabidopsis AtFer1 mRNA abundance (21,22). It can therefore be hypothesized that H 2 O 2 could act in the iron/IDRSdependent pathway, and this is the first point we analyzed. We checked whether H 2 O 2 treatment could mimic the potential oxidative effect of iron overload. Arabidopsis cell cultures were treated either with 300 M iron-citrate, 5 mM H 2 O 2 , or with both inducers. Above 300 M iron-citrate or 5 mM H 2 O 2 , no significant difference in AtFer1 mRNA abundance was observed (data not shown). Cells were harvested at different time points from 1 to 48 h. Total RNA was purified and subjected to Northern analysis for determining AtFer1 mRNA abundance (Fig. 1A). In response to iron treatment, the AtFer1 steady state mRNA level increased 1 h after iron application, reached a maximum after 3 to 6 h of treatment, and decreased. This result is in accordance with those obtained with Arabidopsis plantlets (20,21). The AtFer1 mRNA abundance was also increased after H 2 O 2 treatment, but with a different time course; the maximum level of transcript was observed 12 h after H 2 O 2 treatment, and AtFer1 mRNA was barely detectable after 24 h (Fig. 1A). AtFer1 mRNA abundance after cotreatment with both effectors was compared with the effect of iron or H 2 O 2 applied alone. AtFer1 mRNA abundance was normalized relatively to the EF1␣ mRNA abundance. As shown on Fig. 1B, the addition of both effectors led to a signal intensity quite close to the sum of the values obtained with iron or H 2 O 2 applied alone. This result indicates that iron and H 2 O 2 are acting in two independent and additive pathways leading to AtFer1 mRNA accumulation. However, addition of iron could cause the production of a certain amount of H 2 O 2 , potentially leading to a further AtFer1 mRNA accumulation. Such an hypothesis is unlikely because when cells were treated with catalase prior to iron addition, no change in AtFer1 mRNA accumulation was observed compared with catalase untreated cells (Fig.  1C). At the protein level, ferritin was accumulated after 24 h of iron treatment. It was also accumulated at the same time point after H 2 O 2 treatment, but to a lower extend. Co-treatment with iron and H 2 O 2 led to a ferritin protein accumulation to a level close to the one that was observed in response to iron treatment (Fig. 1B). Moreover, we have tested the effect of okadaic acid (OA) on AtFer1 mRNA abundance in response to iron excess or H 2 O 2 treatments. It has to be reminded that OA has been shown to antagonize maize ZmFer1 gene expression both in response to iron excess and H 2 O 2 treatment. When Arabidopsis cells were treated with OA, iron-induced AtFer1 mRNA abundance was decreased, whereas H 2 O 2 -induced mRNA abundance was increased (Fig. 1D). This indicates that a PP2A-type phosphatase is a positive regulator of the iron pathway and a negative regulator of the H 2 O 2 pathway. This result enforces the hypothesis of independent pathways. Although we cannot rule out that reactive oxygen species other than H 2 O 2 could be involved in the increase of AtFer1 mRNA abundance, it is more likely that the response is specific to H 2 O 2 because catalase addition prior to H 2 O 2 treatment abolished AtFer1 mRNA accumulation (Fig. 1C). Our main goal being to decipher the iron-and IDRS-dependent pathway, we further investigate only the response of AtFer1 gene to iron treatment.
An Iron-induced Plastidial NO Burst Precedes AtFer1 mRNA Accumulation-Several reports have involved nitric oxide in the control of iron homeostasis and in AtFer1 regulation (26, 27, 36 -38). NO was shown to be involved in iron-and IDRSdependent AtFer1 regulation (26,27). SNP (a NO donor) induced an AtFer1 mRNA abundance increase and cPTIO (a NO scavenger) antagonized iron-induced AtFer1 mRNA accumulation. We have first examined whether iron treatment may lead to a NO production in the cells by using the NO-sensitive fluorescent probe DAF-FM DA. Cells were incubated for 20 min with the probe before the application of treatments. NO production was visualized using a confocal microscope (Fig. 2). The iron-citrate treatment led to a very rapid increase of the fluorescence. A clear signal was detected 5 min after iron-citrate addition. Its intensity increased and reached a maximum after 30 min (Fig. 2A). A control treatment made with potassium replacing iron as cation did not lead to such a strong fluorescence increase, indicating that the NO production observed was specific of iron. Furthermore, the DAF-FM DA fluorescence pattern observed at the early times of the kinetic merged with the red chlorophyll autofluorescence (Fig. 2B). This indicates that iron treatment leads to a rapid NO burst in the plastids. After 30 min of iron treatment, the fluorescence was very intense and was also detected in both the nucleus and the cytosol of the cells.
The next step was to examine further the origin of the NO production regulating AtFer1 expression. Cells were treated with animal NO synthase inhibitors prior to iron addition. A treatment with L-NMMA decreased AtFer1 mRNA abundance in response to iron treatment (Fig. 3A). The same decrease was obtained with N -nitro-L-arginine methyl ester (data not shown). We used both SNP and cPTIO as controls, and, as previously reported (26,27), these compounds increased and  AUGUST 18, 2006 • VOLUME 281 • NUMBER 33 decreased, respectively, AtFer1 mRNA abundance (Fig. 3A). The decrease observed with L-NMMA treatment suggests that a NO synthase activity could be required for the iron-dependent AtFer1 up-regulation. So far, two enzymes have been implicated in NO production in plants (39,40): the nitrate reductase and the NO synthase NOS1. To determine whether these enzymes are involved in NO production leading to AtFer1 mRNA accumulation in response to iron excess, the loss of function mutants gЈ4-3 (30) and atnos1 (29) were used. The gЈ4-3 mutant is deficient in both nia1 and nia2 gene activities and displays only 0.5% of the wild-type shoot nitrate reductase (NR) activity. Arabidopsis cell lines of these two mutants are not available. Therefore, experiments were performed on plantlets grown under sterile conditions. In this plantlet system, AtFer1 response to iron treatment is not altered (20), and NO treatment led to an increase of AtFer1 mRNA abundance, in the same manner as in the cell culture system (data not shown). In the two mutants, the abundance of AtFer1 mRNA after iron application was similar to the one observed in wild-type plants (Fig. 3B). These results indicate that neither nitrate reductase nor NOS1 are involved in iron-dependent NO production leading to AtFer1 regulation. AtFer1 Repressor Is Degraded by a 26 S Proteasome-dependent Pathway-The cis-acting element IDRS has been shown to be involved in iron-dependent AtFer1 de-repression (18). This suggests that a repressor acts in the pathway, and that iron addition leads to a de-repression of AtFer1 transcription, rather than to activation (18,27). This also suggests that iron treatment may lead to the inactivation or the degradation of a repressor. To test this latter hypothesis, and more specifically the involvement of a 26 S proteasome-dependent protein degradation, we used the specific 26 S proteasome inhibitor MG132. Cells were treated with MG132 prior to iron treat-ment. A general serine protease inhibitor, Pefabloc, was used in the same conditions as a control (Fig. 4A). Pefabloc did not affect the iron-dependent AtFer1 regulation, whereas MG132 treatment completely abolished the response. This result indicates that a protein, involved in the repression of AtFer1 transcription in low iron conditions, is ubiquitinated and degraded by a 26 S proteasome-dependent pathway after iron treatment. This factor was named repressor.

Regulation of Ferritin by NO and Protein Degradation
As protein degradation appears to be involved in AtFer1 derepression, we have also checked whether protein synthesis A, effect of NO scavenger, NO donor, and NOS inhibitors. Cells were pretreated for 1 h with 1 mM cPTIO or 5 mM L-NMMA prior to 300 M iron-citrate addition. SNP (2.5 mM) was added without iron. Cells were collected at different time points. Northern blot analysis was performed as described in the legend to Fig. 1. B, implication of NR and NOS1 in AtFer1 iron-mediated expression. Ten-day-old wild-type Arabidopsis (Columbia ecotype), nos1 and gЈ4-3 mutants plantlets, cultivated in vitro, were treated, where indicated, with 300 M iron-citrate for 3 and 6 h. RNA extracted from plantlets were analyzed by Northern blotting and hybridized successively with AtFer1 and EF1␣ probes. 600 M Na 3 -citrate was added in culture medium. Cells were collected before, and 3 and 6 h after iron treatment. Northern blot was performed as described in the legend to Fig. 1. B, effect of cycloheximide. Cells were pre-treated, where indicated, with 100 M cycloheximide for 1 h prior the addition of 300 M iron-citrate. Samples were collected at different time points, and AtFer1 mRNA abundance was examined as before. AtFer1 mRNA relative abundance was determined as described in the legend to Fig. 1. Bars correspond to the standard deviation from three independent experiments. may arise in the pathway. Cells were treated with cycloheximide prior to iron addition, and harvested at different time points from 3 to 12 h. A control kinetic was performed with untreated cells. As shown on Fig. 4B, 3 h after the iron addition, AtFer1 mRNA abundance was not affected by the cycloheximide. By contrast, after 6 h and later, mRNA abundance was increased by cycloheximide treatment. The decrease of mRNA abundance observed 9 and 12 h after iron treatment was greatly reduced by addition of cycloheximide. This suggests that de novo synthesis of the repressor, and/or of protein(s) involved in the AtFer1 mRNA turnover, is necessary to decrease ferritin mRNA abundance in the latter times of the kinetic after iron addition.
So far, three elements have been identified in the signal transduction pathway leading to the AtFer1 response to iron. Two positive regulators of the mRNA accumulation: NO and a PP2A-type phosphatase, and a negative regulator: the repressor. To position them with respect to the others in the pathway, cells were co-treated with two effectors at the same time. We used SNP, which promotes an AtFer1 mRNA increase in abundance, and OA and MG132, which leads to a decrease of the transcript level. The positive effect of SNP treatment on the AtFer1 mRNA level was abolished by both OA and MG132 treatments (Fig. 5A). Therefore, NO production by SNP cannot bypass the effect of the two inhibitors, MG132 and OA. This indicates that NO production, acting downstream of iron, could act upstream of the PP2A-type phosphatase and the repressor. However, because all protein degradation (i.e. not only the repressor) by the 26 S proteasome is blocked by MG132, it is possible that protein degradation acts further upstream in the pathway. It can be hypothesized that degradation of some proteins could be required for NO production. Thus, altering the NO production by MG132 could also explain why AtFer1 mRNA does not accumulate in response to treatment with this proteasome inhibitor. To show that NO generation in response to iron treatment occurs independently of the proteasome, Arabidopsis cells were treated with iron and MG132, and NO production was monitored by DAF imaging. The fluorescence observed was similar to the one of control cells treated with iron in the absence of MG132, showing that iron-dependent NO production was not altered by MG132 (Fig. 5B).
AtFer1 Repressor Is Not Bound to the IDRS-As the IDRS cis-acting element has been shown to be involved in AtFer1 de-repression, it was tempting to postulate that the repressor could be a transcription factor bound to the IDRS in low iron conditions, and degraded upon iron addition. To test this hypothesis, we first checked whether the AtFer1 IDRS could bind nuclear factors. Nuclear extracts were prepared from untreated cells and from cells treated with iron (de-repression condition), or with MG132 (repression condition). A 32 P radiolabeled DNA probe corresponding to a 200-bp region of the AtFer1 promoter sequence, and containing the IDRS (probe A; Fig. 6A), was incubated with nuclear proteins. Only one complex was observed by gel shift (Fig. 6B). To check the specificity of this complex, the binding reaction was performed with a 50-fold molar excess of different unlabeled DNA fragments. With DNA fragments containing the IDRS (probes A, B, and C, Fig. 6A), the signal corresponding to the complex was completely abolished, whereas it was not modified with probes that did not contain the IDRS sequence (probes D and E, Fig. 6A). Results indicate that the complex observed consists of nuclear proteins bound to the IDRS. The intensity of the DNA-protein(s) complex signal was almost the same with extracts prepared from untreated, iron-treated, and MG132-treated cells (Fig. 6B), suggesting that complex formation is the same under repressive and de-repressive conditions. Taken together, results suggest that a repressor is acting upstream of the nuclear protein(s), which are stably bound to the IDRS, both in repressive or in de-repressive conditions. Such repressor is degraded by the proteasome in de-repressive conditions.

DISCUSSION
Our current knowledge on ferritin gene expression in plants is largely based on work on maize ZmFer genes (18,19,22,25) and on the Arabidopsis AtFer1 gene (20,21,23,27,28). We used an Arabidopsis cell culture system to further characterize the pathway leading to AtFer1 de-repression after iron addition.
The AtFer1 mRNA level is enhanced by both iron and H 2 O 2 treatments (Fig. 1A). This result is consistent with the studies on the maize ZmFer1 gene, which is orthologous to AtFer1 (22). As a highly reactive transition metal, iron may lead to oxidative stress. By sequestering free iron, the accumulation of ferritin may prevent oxidative damage. In maize, ZmFer1 mRNA abundance in response to iron treatment is antagonized by antioxidants like N-acetylcysteine and GSH, indicating that the iron effect on ferritin mRNA abundance is dependent of an oxidative step (22). Both iron-and H 2 O 2 -dependent ZmFer1 mRNA regulation are sensitive to okadaic acid. This is consistent with the hypothesis that these inducers could act through the same oxidative pathway leading to an increased ZmFer1 mRNA abundance (22). In contrast to ZmFer1, OA strongly decreased AtFer1 mRNA abundance in response to iron (Fig. 1D). Furthermore, the amount of the AtFer1 transcript in response to H 2 O 2 is increased after okadaic acid treatment. Thus, a PP2Atype phosphatase may be an activator of the iron-dependent pathway, and a repressor of the H 2 O 2 -dependent pathway. The two pathways appear therefore totally independent. This is in full agreement with the observation that the abundance of the AtFer1 mRNA in response to the addition of the two effectors at the same time is closed to the sum of the transcript abundance observed in response to each effector when applied alone (Fig.   1B). However, the iron-dependent AtFer1 mRNA increase in abundance has been previously shown to be antagonized by N-acetylcysteine treatment, revealing the involvement of an oxidative step in this response (20). This suggests that two different oxidative signals are involved in both iron-and H 2 O 2 -dependent pathways. At the protein level, we observed a higher amount of ferritin accumulated in response to iron treatment than in response to H 2 O 2 treatment (Fig. 1B). This is consistent with previous data indicating that iron is required for ferritin protein stabilization (25).
In the present work, we have shown that iron addition leads to NO accumulation in the plastids (Fig. 2). NO is an essential growth regulator in plants, and serves as a signal in biotic and abiotic stress responses, programmed cell death, hormone responses, root and xylem development, flowering, and iron homeostasis (reviewed in Refs. 37 and 39 -45). The inhibitory effect of L-NMMA on AtFer1 expression in response to iron excess indicates that NO synthase activity is involved in the pathway (Fig. 3). In plant cells, the two enzymes so far reported to be implicated in NO synthesis are the NR and a nitric-oxide synthase (NOS1). NR is a cytosolic located enzyme, and NOS1 has recently been shown to be targeted to the mitochondria (46). Thus, none of these enzymes could produce NO in the plastids. Such a conclusion is consistent with the observation that AtFer1 expression in response to iron is not altered in the mutants gЈ4-3 and atnos1 (Fig. 3). Although there is so far no conclusive evidence for enzymatic NO production in the plastids, there are two indirect observations in favor of this hypothesis. First, an immunoreactive NOS protein has been detected in plastids (47). Second, the addition of the plant defense elicitor cryptogein on epidermal tobacco cells leads to a N -nitro-L-arginine methyl ester-sensitive NO production in the plastid (48,49). These results indicate that NO synthase activity is probably present in the plastid, but the nature of the enzyme involved remains to be determined.
A NO scavenger, cPTIO, completely abolishes iron-dependent AtFer1 expression, clearly establishing that NO is a major element in this signal transduction pathway. In contrast to cPTIO, L-NMMA application decreases AtFer1 response by only 50% when compared with untreated cells (based on relative AtFer1 mRNA abundance compared with EF1␣; data not shown). Such a partial inhibition could be attributed to an incomplete action of this inhibitory compound. However, it cannot be excluded that both an enzymatic and a non-enzymatic (insensitive to the inhibitor used) pathway may lead to NO production in response to iron. Indeed, it is known that in plants, non-enzymatic NO production can arise from reactions between nitrite and various plant metabolites (50 -52). Such a non-enzymatic NO production from nitrite has in particular been reported to occur at acidic pH in the apoplasm for example (52) and could explain the NO effects on germinating seeds (52)(53)(54). However, such a nitrite-dependent NO production is unlikely to occur in the stroma of plastids where the pH value is 7.0 or higher (55,56).
Regardless the origin of the NO produced in the plastid in response to iron, this NO burst is an early event in the signal transduction pathway. This suggests that a retrograde signal, of unknown nature, could be produced in the plastid and lead to FIGURE 6. Specific binding of Arabidopsis nuclear protein(s) to the IDRS. A, localization of the DNA probes used for the gel shift experiments on AtFer1 promoter region. The putative regulatory elements indicated by boxes are those defined by Petit et al. (18). TATA, TATA box. B, mobility shift assays. Five g of nuclear protein extracts from Arabidopsis cells untreated (ϪFe), or treated for 3 h either with 300 M iron-citrate (ϩFe) or 1 h with 50 M MG132 (ϩMG132) were incubated for 30 min at room temperature with radiolabeled probe A. For the competition assays, a 50-fold molar excess of unlabeled fragments was added in the reaction. the transcription of the nuclear-encoded AtFer1 gene. Two major future prospects arise from this work and concern the identification of the site of action of NO in the pathway, and the nature of the retrograde signal. In animals, it is documented that the redox-related species of NO can have simultaneous effects on cellular iron metabolism and homeostasis via mechanisms that might involve S-nitrosylation (57), ligation of NO to iron-sulfur clusters (58,59) or to heme-containing proteins (60,61). NO and free iron may also form complexes with low molecular weight thiols like glutathione and cysteine (62)(63)(64). These dinitrosyl-iron complexes are relatively stable in contrast to the highly reactive free NO molecule (65,66) and were shown to be potential NO carrier molecules in mammals (62). In plants, interactions of NO with hemes (67-69) or iron-sulfur clusters (70), and S-nitrosylation reactions (71) have been shown, and dinitrosyl-iron complexes have been detected (72). A post-translational modification of a plastidial protein by NO could be involved in the pathway leading to AtFer1 de-repression. Alternatively, a dinitrosyl-iron complex with glutathione, which has been shown to permeate quickly through membranes (73), could be a good candidate for a retrograde signal.
Plant ferritin mRNA accumulation can be promoted by ABA (25), H 2 O 2 (Refs. 21 and 22, this work), and NO (Ref. 27, this work). Whether or not these inducers act in the same pathway has not been documented. Interestingly, by combining pharmacological, biochemical, and genetic approaches, it has recently been demonstrated that ABA-induced NO production via NR is required for ABA-induced H 2 O 2 -mediated stomatal closure (74). It is, however, unlikely that such a pathway occurs in the regulation of AtFer1 gene expression for the following reasons. First, we show that NO production in response to iron treatment is not mediated by NR (Fig. 3B). Second, among the ferritin gene families, only AtFer2 in Arabidopsis and ZmFer2 in maize have been reported to be regulated by ABA (19,21). The AtFer1 gene is not regulated by ABA, and the ABA-regulated AtFer2 gene is not modulated by H 2 O 2 (21).
A growing number of reports involve ubiquitination of positive or negative regulatory proteins, and their subsequent 26 S proteasome-dependent degradation as key steps of control within signaling pathways (75)(76)(77). Protein ubiquitination and subsequent degradation have been involved in light, auxin, ethylene, pathogen resistance, and more recently in phosphate starvation responses (78). The presence of a repressor in ironmediated AtFer1 regulation of expression prompted us to examine whether inhibitors of the proteasome-dependent protein degradation could alter the iron response or not. Indeed, the use of MG132 dramatically decreases AtFer1 mRNA abundance. This result indicates that a protein acting in low iron conditions, called the repressor, is ubiquitinated after iron addition and subsequently degraded by the 26 S proteasome. This iron-triggered degradation leads to a de-repression of AtFer1 transcription and to the increase in abundance of the corresponding mRNA. Protein synthesis inhibition by cycloheximide perturbs such an increase, because at later time points of the kinetic (from 6 to 12 h) protein synthesis inhibition leads to a higher increased abundance of the AtFer1 mRNA compared with control cells untreated with cycloheximide (Fig.  4B). Therefore, it can be proposed that de novo synthesis of the repressor is necessary to decrease AtFer1 mRNA abundance about 6 -9 h after iron addition. In fact, such a cycloheximide-promoted superinduction has been largely documented and can be attributed not only to a transcriptional de-repression (79,80) but also to a decrease of mRNA degradation (79 -82). Our results cannot discriminate between these two possibilities.
Mutagenesis of the IDRS cis-element within the maize ZmFer1 or Arabidopsis AtFer1 promoter sequences leads to de-repression of the expression of reporter genes, in the absence of iron treatment (18). It can be therefore hypothesized that the repressor could bind to the IDRS in low iron conditions and that iron treatment would promote its ubiquitination and degradation, resulting in AtFer1 gene expression. MG132 treatment of the cells would avoid the repressor degradation, maintain its binding to the IDRS and repress AtFer1 gene expression, even after iron treatment. However, nuclear protein(s) from cells either treated with iron or MG132, or untreated bind equally to the IDRS, without significant difference in size or intensity of the complex formed (Fig. 6). Therefore, the protein ubiquitinated and degraded in response to iron treatment, and defined as the repressor, is unlikely to be a trans-acting factor able to directly bind to the IDRS in low iron conditions. This result is consistent with the observation that no significant difference in the intensity of the ZmFer1 IDRS-protein complex was observed by gel shift experiments using nuclear extracts of iron-treated or untreated maize plants (18). Thus, it is proposed that the transcription factor(s) bound to the IDRS is(are) essential but not sufficient for AtFer1 de-repression, and that the repressor acts upstream of this transcription factor.
It is worth noticing that in animal cells, one of the ferritin trans regulators involved in the translational repression of ferritin mRNAs in low iron conditions, namely IRP2, is regulated by NO and ubiquitin. After iron addition, IRP2 has been shown to be nitrosylated and subsequently ubiquitinated, and degraded by the proteasome, thus leading to ferritin mRNA translation (14,15). Our work shows that NO and protein degradation via proteasome are also involved in the regulation of ferritin gene expression in plants. It is remarkable that molecular effectors of the response to iron excess, such as NO and ubiquitination, are conserved between the translational regulation of animal ferritin and the transcriptional regulation of plant ferritin.