The analysis of Arabidopsis nicotianamine synthase mutants reveals functions for nicotianamine in seed iron loading and iron deficiency responses

Nicotianamine chelates and transports micronutrient metal ions in plants. It has been speculated that nicotianamine is involved in seed loading with micronutrients. A tomato ( chloronerva ) and a tobacco mutant have been utilized to analyze the effects of nicotianamine loss. These mutants showed early leaf chlorosis and had sterile flowers. Arabidopsis thaliana has four NAS genes. We constructed two quadruple nas mutants: One had full loss of NAS function, was sterile and showed a chloronerva -like phenotype ( nas4x-2 ). Another mutant with intermediate phenotype ( nas4x-1 ) developed chlorotic leaves, which became severe upon transition from the vegetative to reproductive phase and upon Fe deficiency. Residual nicotianamine levels were sufficient to sustain the life cycle. Therefore, the nas4x-1 mutant enabled us to study late nicotianamine functions. This mutant had no detectable nicotianamine in rosette leaves of the reproductive stage but low nicotianamine levels in vegetative rosette leaves and seeds. Fe accumulated in the rosette leaves while less Fe was present in flowers and seeds. Leaves, roots and flowers showed symptoms of Fe deficiency, whereas leaves also showed signs of sufficient Fe supply as revealed by molecular-physiological analysis. The mutant was not able to fully mobilize Fe to sustain Fe supply of flowers and seeds in the normal way. Thus, nicotianamine is needed for correct supply of seeds with Fe. These results are fundamental for plant manipulation approaches to modify iron homeostasis regulation through alterations of NAS genes. nas mutant lines grown for 17 days on 60 µM Ni and control Hoagland agar medium (n= 10-35). Three leaf phenotypic classes were defined, namely light green leaf appearance of mildly affected plants, intermediate yielded approx. 15 pmol nicotianamine per injection of both pure plant sample (14.8 pmol NA) and non-extraction treated nicotianamine standard (15.4 pmol NA). Addition of nicotianamine standard prior to extraction yielded 29.7 pmol nicotianamine per injection. From this result a recovery rate of 98% ± 7% was calculated. Thus, it can be concluded that nicotianamine was not lost during the extraction procedure. n= 3 technical repetitions.


INTRODUCTION
One of the breeding goals for high quality nutrition food crops is the production of micronutrient (Fe, Zn)-enriched plants. The relevance of Fe for human nutrition is evident from the most severe human micronutrient deficiency disease which is Fe deficiency anemia (see 2004 report of the World Health Organization). In plants, Fe deficiency is also the most widespread micronutrient deficiency that is seen frequently on calcareous and alkaline soils where Fe is almost unsoluble. On the other hand, beneficial biochemical properties can render this same metal potentially toxic. Free Fe is a catalyst in the formation of hydroxyl radicals which can unspecifically damage biological molecules. To balance these effects, Fe uptake and homeostasis are tightly controlled.
Genetic and transgenic approaches targeting specific Fe homeostasis genes are underway to yield micronutrient-enriched crops (Goto et al., 1999;Takahashi et al., 2001;Sautter et al., 2006;Uauy et al., 2006). It is obvious that a precise breeding approach would be more efficient if more genetic components would be known controlling Fe acquisition, transport and storage.
Among the key elements for Fe homeostasis are small chelators of metals, both to render the metals soluble and to detoxify them (Colangelo and Guerinot, 2006;Briat et al., 2007). For example, organic acids such as citric acid can bind free metal ions. This mechanism is primarily used for transport of Fe in the xylem, where it is considered that the majority of Fe is bound to citrate. On the other hand, phloem Fe is bound by nicotianamine or other amino acids such as part of proteins (von Wiren et al., 1999;Kruger et al., 2002).
The plant metal chelator nicotianamine is a free non-proteinogenic amino acid (Scholz et al., 1992;Stephan, 2002;Hell and Stephan, 2003). Nicotianamine is mobile in the plant and has been detected in root and leaf cells as well as in phloem sap. It can bind metal ions like Fe, Zn, Cu and Ni (Scholz et al., 1992;Schmiedeberg et al., 2003). Nicotianamine is synthesized by a one-step condensation reaction of three molecules S-adenosyl-methionine by the enzyme nicotianamine synthase (NAS) (Shojima et al., 1989;Herbik et al., 1999;Higuchi et al., 1999). In graminaceous plants like barley, rice and maize nicotianamine has a prominent role in Fe acquisition from the soil. It is the direct precursor of phytosiderophores of the mugineic acid family (Mori, 1999). Phytosiderophores are strong Fe 3+ chelators that are extruded to the rhizosphere where they serve to chelate and solubilize Fe 3+ . NAS genes form gene families in grasses that cover a range of expression patterns (Higuchi et al., 2001;Inoue et al., 2003;Mizuno et al., 2003). NAS genes involved in phytosiderophore production are induced by Fe deficiency in the root.
Phytosiderophore-Fe 3+ complexes are imported into the rhizodermis by the transporter of the type YELLOW STRIPE1 (YS1) from maize that is up-regulated by Fe deficiency in roots (Curie et al., 2001). YELLOW STRIPE-LIKE genes (YSL) are encoded by multigene families in grasses as well as in dicot plants (Curie et al., 2001;DiDonato et al., 2004). This indicates that NAS and YSL genes have diversified roles throughout plant development in addition to Fe uptake into the root.
In Solanaceous plants nicotianamine was shown to act in Fe homeostasis which has been mostly found through analysis of the nicotianamine synthase tomato mutant chloronerva and of a transgenic tobacco line with ectopic nicotianamine aminotransferase (NAAT) activity that consumed the available nicotianamine (Scholz et al., 1992;Ling et al., 1999;Takahashi et al., 2003). Tomato and tobacco showed leaf chlorosis as the result of reduced nicotianamine and sterility. Solanaceous plants and other dicots rely on a Fe reduction-based mechanism to solubilize and acquire Fe 2+ . Fe is reduced by a membrane-bound ferric chelate reductase (FRO2 in Arabidopsis) (Robinson et al., 1999). Fe 2+ is taken up into the root epidermis by a membrane-bound divalent metal transporter (IRT1 in Arabidopsis) (Eide et al., 1996;Vert et al., 2002). FRO2 and IRT1-type genes are induced by an Fe deficiency-dependent basic helix-loop-helix transcription factor protein (FER in tomato, FIT in Arabidopsis) (Ling et al., 2002;Colangelo and Guerinot, 2004;Jakoby et al., 2004;Yuan et al., 2005). In tomato, it was found that chloronerva plants up-regulate FER, IRT1 and FRO1 even upon sufficient Fe supply (Bereczky et al., 2003;Li et al., 2004;Brumbarova and Bauer, 2005). Since chloronerva plants can accumulate Fe in leaves, but not Cu (Pich and Scholz, 1996), nicotianamine is not directly needed for Fe uptake and translocation into leaves in Solanaceous plants (Scholz et al., 1992). Most likely, nicotianamine plays a role in intracellular and intercellular distribution of Fe, whereas it may be involved in Cu translocation to leaves. However, contrasting effects were found in tobacco NAAT plants which had reduced Fe levels in leaves compared to wild type, as well as reduced levels of Cu and Zn (Takahashi et al., 2003). Hence, there were clear distinctions between tomato and tobacco mutant nicotianamine lines. Another disadvantage of the two Solanaceous plant studies was that only mutants with severe defects were available. Thus, it remains unknown which phenotypes were pleiotropic in nature and which phenotypes were directly caused by nicotianamine loss. For example, chloronerva and NAAT tobacco both had abnormal flowers.
A function of nicotianamine in seed metal loading has been inferred from studies of the presumptive nicotianamine-Fe transporter family YSL (Curie et al., 2001;Schaaf et al., 2004). Loss of YSL1 resulted in decreased nicotianamine and Fe levels in ysl1 mutant seeds (Le Jean et al., 2005). YSL genes have clearly redundant functions. ysl1ysl3 double mutants are chlorotic and have lower Fe levels in leaves than wild type (Waters et al., 2006). Their seeds also contain less Fe, and fertility is reduced (Waters et al., 2006). ysl3 and ysl2 single mutants have no apparent phenotype (DiDonato et al., 2004;Waters et al., 2006). Expression patterns of the three YSL genes suggest that the transporters are involved in exiting nicotianamine-Fe from the vascular system to leaves upon sufficient Fe supply, whereas reduced exit upon Fe deficiency may serve to translocate Fe towards the growing apex (DiDonato et al., 2004;Le Jean et al., 2005;Schaaf et al., 2005;Waters et al., 2006).
Indeed, various families of metal transporters and regulators were found and are to date best described in Arabidopsis. The Arabidopsis genome has four NAS genes (Suzuki et al., 1999;Bauer et al., 2004). Thanks to the reverse genetic resources it was possible to identify mutant T-DNA alleles of all four genes. To be able to study nicotianamine function in a broad genetic network of known players in metal homeostasis we have decided to develop Arabidopsis as a model system for investigating nicotianamine function. Here, we present the construction and analysis of quadruple nas mutants, one which had an intermediate phenotype, named nas4x-1. In contrast to chloronerva, the tobacco NAAT line and nas4x-2, which is the second nas mutant we identified that had full loss of NAS function, nas4x-1 had strongly reduced but not fully eliminated nicotianamine content. Due to its survival and maintained seed production we could analyze nicotianamine involvement in Fe allocation to reproductive organs and seeds.

Generation and initial characterization of multiple nas mutants
The four members of the NICOTIANAMINE SYNTHASE (NAS) gene family are distributed on both arms of chromosomes I and V. All NAS genes have a single exon of 963 bp (NAS1,NAS2,NAS3) and 975 bp (NAS4). With the aim of ultimately studying nas mutants, we identified and confirmed nas T-DNA insertion mutants for the four NAS genes (Fig. 1A). For the genes NAS1, NAS3 and NAS4 we could confirm that the positions of T-DNA insertions were in the exon coding sequences (alleles nas1-1, nas3-1 and nas4-1, respectively). For the NAS2 gene we could confirm the mutant nas2-1 allele harboring a T-DNA insertion 70 bp upstream of the start codon in the 5'-UTR of NAS2 and the mutant nas2-2 allele with an insertion 144 bp after the start codon in the exon.
Since single T-DNA insertion lines did not show any visible phenotype on soil (Bauer et al., 2004) we crossed together different nas mutants. First, we generated homozygous double mutants by crossing and selfing T-DNA insertion mutants of genes located on the same chromosome, namely nas1-1nas2-1 and nas3-1nas4-1, respectively. Homozygous double mutants were confirmed by genotyping. We found that the crossing over events took place with a recombination rate of approximately 5 % on the two chromosomes. Homozygous nas1-1nas2-1 double mutants were then crossed with nas3-1nas4-1 to generate heterozygous quadruple mutants. Quadruple heterozygous plants were selfed. We used genomic PCR to identify and select the quadruple homozygous mutant as well as double and triple mutants. A quadruple homozygous mutant that was fertile was identified as nas1-1nas2-1nas3-1nas4-1 and termed nas4x-1 mutant. We observed that the nas4x-1 mutant had a minor leaf chlorosis during vegetative growth which intensified during the transition to reproductive growth on soil (Fig. 1B). Non-chlorotic leaves from the vegetative stage of the nas4x-1 mutant had darker green veins than wild type when observed at transmitted light conditions (Fig. 1C). The nas4x-1 mutant was crossed with a homozygous nas2-2 mutant. The resulting multiple heterozygous plant was selfed and its progeny genotyped. We identified a homozygous quadruple nas1-1nas2-2nas3-1nas4-1 mutant that we termed nas4x-2. nas4x-2 plants had a strong leaf chlorosis already upon vegetative growth and they were sterile (Fig. 1D). We therefore suspected that nas4x-2 was a strong loss of function mutant similar to chloronerva, while nas4x-1 had an intermediate nas phenotypic expression.
As a first experiment we studied whether any full-length NAS gene products were made in the quadruple homozygous mutants (data not shown). We found that in nas4x-1 and nas4x-2 mutant plants full-length coding transcripts were not produced for nas1-1, nas2-2, nas3-1 and nas4-1.
However, in the case of the nas2-1 allele present in the nas4x-1 mutant it was possible to amplify a DNA fragment which contained the full-length coding NAS2 region after 35 PCR cycles indicating that despite of the insertion in the 5'UTR a PCR product was made. Very interestingly, nas2-1derived transcripts were found expressed not only in roots as was the case in wild type but also in leaves (data not shown, see later quantitative reverse transcription-PCR analysis in Fig. 3A). This result indicated that the NAS2 gene was aberrantly expressed from the nas2-1 allele.
Next, we analyzed nicotianamine contents. We found that nas4x-1 mutant rosette leaves harvested in the vegetative phase had 10 % nicotianamine levels compared to wild type (11 hour light period; Since vegetative growth resulted in the highest ratio of nicotianamine levels in nas4x-1 mutant versus wild type plants we analyzed nas4x-2 nicotianamine contents at the vegetative stage. We found that in the strong nas4x-2 mutant nicotianamine was not detectable in rosette leaves demonstrating full loss of NAS function ( Fig. 2A). This finding explained the strong nas4x-2 phenotype.
Arabidopsis has four functional NAS genes which are differentially expressed (Bauer et al., 2004).
One might therefore assume that NAS genes may have acquired specific functions during their evolution. The genetic data however indicate that NAS genes are functionally largely redundant. Therefore, we assessed the contribution of individual nas alleles for nicotianamine production by analyzing nicotianamine contents in the four single mutants with a single nas gene knockout and in triple mutants containing only a single viable NAS gene, respectively. We found that all nas single mutants had wild type nicotianamine levels ( Fig. 2A; Suppl. Fig. 1A). All triple mutants analyzed had about 30 -50 % of the nicotianamine content compared to wild type plants ( Fig. 2A; Suppl. Fig. 1B). No single NAS genes may therefore play any predominant role for overall nicotianamine synthesis. We also utilised these studies to refine the function of the nas2-1 allele in the nas4x-1 mutant. We found that nas1-1nas3-1nas4-1 had 35 % nicotianamine compared to wild type. On the other hand, the nas2-1 and nas2-2 mutant had the same amount of nicotianamine as wild type ( Fig.   2A). Thus, nas2-1 does not lead to an increased nicotianamine production. Also NAS2 loss of function could be compensated by other functional NAS genes so that overall nicotianamine content remained constant.
To test whether the redundancy could be due to compensatory effects of NAS gene transcription levels we tested NAS gene expression levels in the single or triple mutant backgrounds (Suppl. Fig.   2). We observed that the presence of other mutant nas alleles did not affect gene expression of NAS2 and NAS3. NAS1 expression levels did not seem to be influenced by other nas mutations, however due to high variability of NAS1 expression in these same plant samples no clear conclusions were possible (see comment later on variable NAS1 expression in response to Fe deficiency). The only case where NAS gene expression may have been influenced was that of NAS4 wild type allelic expression in the background of a nas2-1 allele. These control experiments indicate that redundancy of the four NAS genes was not generally due to compensation of defective NAS genes by increased expression of intact NAS genes.
The reduced nicotianamine content of nas4x-1 during the reproductive phase could be due to removal of nicotianamine through remobilisation and transport and to reduced production. To test this, we grew plants in a hydroponic system for 6 weeks either under short day conditions (8 hour light period, harvest at the vegetative stage) or long day conditions (16 hour light period, harvest at the reproductive stage) and analyzed NAS gene expression (Suppl. Fig. 3). We found that NAS1 and NAS2 were expressed at 8-10fold higher level in roots at the vegetative stage than in the reproductive stage. NAS3 on the other hand was expressed at about fourfold higher level in leaves at the reproductive stage than at the vegetative stage. NAS4 was expressed at a similar level in the two stages. These results are in agreement with the microarray data available from public databases (data not shown). In the nas4x-1 plants NAS2 was expressed twofold higher upon the vegetative than the reproductive stage whereas all other NAS genes were down-regulated due to the mutations (Suppl. Fig. 3). Thus, NAS genes were differentially expressed with respect to the growth stage. At every stage at least two NAS genes were expressed in the wild type. nas4x-1 had NAS expression in roots at a higher level than in leaves, whereby in roots it was lower upon the reproductive phase than the vegetative stage. Therefore, reduced nicotianamine production along with increased transport away from leaves may explain the leaf chlorosis phenotype upon reproduction.
This prompted us to speculate that nicotianamine plays an important role in seed metal homeostasis

Effect of Fe, Zn and Cu deficiency on nas4x-1 mutants
Nicotianamine chelates Fe, Cu and Zn. Therefore, we tested first whether deficiency of these metals may play a role in the expression of the mutant phenotype.
We grew wild type and nas4x-1 plants in a hydroponic system and then exposed five week-old plants for 5 days to control medium (10 µM Fe) or Fe deficiency (0 Fe). We observed that nas4x-1 mutants had developed a leaf chlorosis after the first 5 weeks. The leaf chlorosis became severe when plants were transferred to 0 Fe (Suppl. Fig. 4).
When we tested 0 Cu and 0 Zn supply we did not observe any strong leaf chlorosis in the mutant after 7 days of exposure to deficiency. Cu and Zn contents in leaves of 0 Cu or 0 Zn-grown mutant and wild type plants were however reduced demonstrating that the plants had been indeed Cu-and Zn-deficient (Suppl. Fig. 5).
The results show that nas4x-1 mutants were sensitive to Fe deficiency but not to deficiency of Zn and Cu. Hence, in nas4x-1 plants nicotianamine is present in insufficient amounts to properly distribute the low Fe, whereas no problem exists for distribution of low Cu and Zn.

Gene expression of nas4x-1 plants in response to Fe deficiency
To analyze the susceptibility of nas4x-1 plants to Fe deficiency further we harvested roots and leaves from hydroponically grown nas4x-1 and wild type plants and studied gene expression in response to Fe deficiency by quantitative reverse transcription-PCR (Fig. 3).
As a control NAS gene expression is first shown. NAS1, NAS2 and NAS4 were expressed in roots whereas NAS3 was not expressed in roots (Fig. 3A). NAS2 and NAS4 were induced by Fe deficiency in roots to about three times. NAS1 was not found to be generally induced by Fe deficiency in the hydroponic growth system, whereas on agar it was found induced (compare to Fig. 4A; note that in the various nas mutant backgrounds NAS1 also showed most variable gene expression levels). In leaves, we found expression of NAS1, NAS3 and NAS4. NAS4 was induced by Fe deficiency in leaves, whereas NAS3 was repressed under Fe deficiency. NAS1 was not Fe-regulated in leaves. In the nas4x-1 plants NAS1, NAS3 and NAS4 were down-regulated due to the respective nas mutations, whereas NAS2 was found to be expressed in roots and leaves due to the nas2-1 allele (Fig. 3A).
Hence, in roots NAS genes are regulated by Fe-deficiency suggesting a functional involvement of these genes in Fe deficiency responses. It was confirmed that NAS genes are deregulated in nas4x-1 due to T-DNA insertions.
Next, we tested whether the susceptibility of the mutant to Fe deficiency was due to altered expression of Fe uptake genes (Fig. 3B). We investigated expression of the transcription factor genes FIT, encoding an essential regulator for Fe acquisition (Colangelo and Guerinot, 2004;Jakoby et al., 2004;Yuan et al., 2005), and BHLH100, which is induced by Fe deficiency in roots and leaves (Wang et al., 2007). Furthermore, we studied gene expression of the Fe acquisition genes FRO2 (Fe reductase) and IRT1 (Fe transporter) which are up-regulated by Fe deficiency dependent on FIT (Eide et al., 1996;Robinson et al., 1999;Vert et al., 2002). We found that FIT, IRT1, FRO2 and BHLH100 were induced by Fe deficiency in roots of wild type plants (Fig. 3B). In wild type leaves BHLH100 was also induced by Fe deficiency. Therefore, the marker genes were up-regulated by Fe deficiency in the hydroponic system as previously shown for the agar plate system (Jakoby et al., 2004;Wang et al., 2007). The levels of induction of IRT1 and FRO2 were lower in hydroponically grown plants than in agar plate grown plants, due to the different growth condition and age of the plants. Interestingly, we found that in the nas4x-1 mutant plants Fe deficiency response genes were expressed at higher level than in the wild type plants upon sufficient Fe supply in roots and in leaves (Fig. 3B). Upon 0 Fe the Fe deficiency response genes were expressed at lower or equal level in nas4x-1 plants than in wild type. Hence, nas4x-1 plants experienced Fe deficiency upon sufficient Fe supply, but the mutant phenotype was not explained by inability to induce Fe deficiency response genes.
We then studied whether altered distribution of Fe might cause the phenotype. We therefore studied gene expression of YELLOW STRIPE-LIKE genes YSL1, YSL2 and YSL3 encoding potential transporters for nicotianamine-metal complexes (DiDonato et al., 2004;Le Jean et al., 2005;Schaaf et al., 2005;Waters et al., 2006). We confirmed that in wild type all three YSL genes were downregulated by Fe deficiency in roots and leaves compared to + Fe (Fig. 3C). In nas4x-1 mutant roots grown at + Fe the three genes were expressed at comparable level as in wild type -Fe. The same was found upon -Fe. In leaves of nas4x-1 mutants the three YSL genes were up-regulated or equally expressed upon sufficient Fe supply, whereas they were rather down-regulated upon Fe deficiency.
Hence, YSL gene expression suggested that upon Fe sufficiency nas4x-1 mutant roots behaved like Fe deficient roots, whereas leaves of nas4x-1 plants had characteristics of Fe-sufficient leaves despite of the weak leaf chlorosis.
The previous results were contradictory as nas4x-1 leaf cells were on one side Fe-sufficient (as suggested by YSL up-regulation) and on the other side Fe-deficient (as revealed by BHLH100 upregulation). To further investigate this we analyzed expression of the two leaf marker genes for either Fe sufficiency or deficiency. One marker gene was ferritin gene FER1 which encodes a Fe storage protein in leaf chloroplasts expressed upon sufficient Fe (Petit et al., 2001). The other was Fe deficiency-inducible ferric reductase gene FRO3 (Mukherjee et al., 2006). As expected, FER1 was up-regulated in Fe-sufficient wild type plants (Fig. 3E). FRO3 on the other hand was upregulated in wild type leaves upon -Fe (Fig. 3E). However, in nas4x-1 mutant plants FER1 and FRO3 were surprisingly regulated as in the wild type upon both + and -Fe demonstrating that the leaf Fe regulatory system perceived Fe sufficiency in nas4x-1 leaves upon + Fe (Fig. 3E). Upregulation of BHLH100 and up-regulation of Fe deficiency response genes in the root must therefore be linked with another signalling system which is defective in nas4x-1.
In order to assess a physiological marker for Fe deficiency we determined Fe reductase activity of nas4x-1 plants. We found that as expected root Fe reductase activity was induced in the wild type upon transfer to Fe deficiency in the hydroponic system. In nas4x-1 roots Fe reductase activity was increased compared to the wild type upon sufficient Fe supply (Fig. 3D).
Taken together we conclude from these molecular and physiological results that nas4x-1 mutants suffer from Fe deficiency when sufficient Fe is supplied. Roots showed symptoms of Fe deficiency (up-regulation of FIT, IRT1, FRO2, BHLH100) upon sufficient Fe. Leaves show at the same time symptoms for Fe deficiency (BHLH100 induction) and Fe sufficiency (YSL and FER1 upregulation, FRO3 downregulation) which can be explained by cellular or subcellular mislocalization of Fe.
Obviously, low nicotianamine resulted in an altered regulation of Fe deficiency responses.

Genetic interaction of NAS and FIT genes
The gene expression analysis supports a role of nicotianamine in the regulation of Fe deficiency responses. We analyzed this further by studying the phenotypes of nas4x-1 mutations in the background of a fit-3 mutant. nas4x-1fit-3 multiple mutants were not able to reduce Fe in the ferric reductase assay as was also the case of fit-3 mutants (Fig. 4B) nor could they induce FRO2 gene expression (Fig. 4C). nas4x-1fit-3 multiple mutants showed a similar yellow leaf chlorosis phenotype as fit-3 mutants (data no shown).
Since NAS genes were regulated by Fe supply we investigated whether NAS gene expression in roots might require the Fe deficiency response regulator FIT. fit loss of function mutants are chlorotic and show down-regulation of IRT1 and FRO2. We grew fit-3 mutants and p2x35S::FIT overexpression plants for two weeks on sufficient Fe supply and transferred them for three days to Fe deficiency as previously described (Jakoby et al., 2004) and analyzed NAS gene expression. We found that among the four NAS genes tested only NAS1 was down-regulated in roots of fit-3 mutant plants compared to wild type. However, it was up-regulated in leaves of FIT overexpression plants exposed to Fe deficiency (Fig. 4A). From previous experiments we know that in FIT overexpression plants FRO2 and IRT1 are ectopically expressed (Jakoby et al., 2004), Hence, NAS1 is partially under the control of the FIT transcription factor as well (see also Colangelo and Guerinot, 2004). The results also show that NAS1 can be induced by -Fe in roots upon agar growth ( Fig. 4A), whereas in the hydroponic system it was not the case (see above, Fig. 3A).
Taken together, these results show that fit-3 is epistatic to nas4x-1, and that FIT controls expression of NAS1. Hence, NAS and FIT genes act in the same pathway of regulating Fe mobilization.

Gene expression during reproduction
Since nas4x-1 leaf chlorosis phenotypes became severe upon transition to flowering, we reasoned that most likely increased nicotianamine levels were needed at this stage in the growing inflorescence.
We investigated whether in flowers the nas4x-1 mutation had an influence on the expression of Feregulated genes (Fig. 5A). IRT1 was found previously to be expressed in flowers, whereby loss of irt1 function had no effect on reproduction (Vert et al., 2002). Interestingly, we found that IRT1 and BHLH100 were strongly up-regulated in nas4x-1 flowers and siliques in contrast to the wild type.
As expected, FIT and FRO2 were not expressed significantly in these organs in neither genetic background. Hence, this could be a hint that IRT1 plays a role in Fe transport in flowers.
We studied expression of YSL transporter genes (Fig. 5B). Whereas YSL3 was expressed at similar level in between mutant and wild type, YSL1 was up-regulated to about 30 % in nas4x-1 flowers and siliques. YSL2 expression was drastically suppressed in nas4x-1 mutant flowers and siliques to about 10-15 fold. Hence, YSL1 and YSL3 expression patterns behaved in flowers and siliques similar as in leaves in wild type compared to nas4x-1. If YSL1 and YSL3 expression were indeed correlated with Fe sufficiency to promote lateral removal of Fe, this observation may indicate that despite of Fe deficiency flowers and siliques were also Fe-sufficient. YSL2 down-regulation in nas4x-1 flowers and siliques was very surprising, and in contrast to the expression in leaves.
Perhaps down-regulation of YSL2 in nas4x-1 reproductive organs was advantageous for Fe supply for example of seeds while nicotianamine was reduced.
Finally, we tested whether FRO3 and FER1 might also be regulated in reproductive organs. FRO3 was found up-regulated in nas4x-1 flowers and siliques compared to wild type (Fig. 5C). This was an indication that in contrast to leaves the reproductive organs were rather Fe-deficient. FER1 expression was highly variable in both wild type and nas4x-1 so that no conclusions were drawn (data not shown).
In summary, gene expression in nas4x-1 mutant flowers and siliques indicated that these organs suffered from Fe deficiency and suggested that an alternative rescue pathway for Fe uptake of cells might have been switched on.

Metal contents of nas4x-1 plants
It was then interesting to analyze whether low nicotianamine levels influenced the transport of metals (Fe, Cu, Zn) into various organs. In roots where metals are found in the vascular system, inside root cells and in the apoplast we did not detect changes between nas4x-1 and wild type (not shown). Differences in metal contents were observed in above-ground organs. Vegetative leaves of nas4x-1 plants contained about the same Fe than wild type, whereas Zn was significantly reduced about 38 % (Fig. 6A). Cu contents were unchanged. In rosette leaves of the reproductive stage Fe contents were significantly increased by 216 % in nas4x-1, whereas Zn remained significantly decreased half (Fig. 6B). Cu was if at all slightly reduced (ca. 11 %) in these nas4x-1 leaves. On the other hand, in flowers, siliques and seeds Fe contents were decreased by 32 % (significantly), 13 % and 46 % (significantly), respectively (Fig. 6C). Zn contents were also decreased in nas4x-1 flowers (significantly) and siliques, whereas it was similar in seeds when compared to wild type. Cu contents were slightly reduced in nas4x-1 siliques, but not in the other plant parts tested. Thus, significance tests of metal determination showed that nicotianamine was not needed to take up and transport Fe to leaves but it was required to efficiently mobilise Fe from leaves during the reproductive phase to flowers and seeds. Nicotianamine was also beneficial for Zn uptake and transport to leaves and flowers.

Phenotypes of nas4x-1 plants in response to toxic metal supply
Our results showed that reduced nicotianamine caused reduced Fe transport to reproductive plant parts and reduced Zn in the plant. Hence, we wondered whether the loss of the ability to translocate metals inside the plant could lead to an altered response to metal toxicity. To test this we studied the responses of nas4x-1 plants to supply of excess concentrations of Fe (up to 200 µM), Zn (up to 200 µM) and Cu (up to 28 µM). Fe, Zn and Cu had caused shorter roots and smaller shoots than control conditions, Zn and Cu also caused yellow leaves (not shown). In none of the three cases we observed a significant phenotypic difference between nas4x-1 and wild type plants (not shown). We conclude therefore that wild type nicotianamine levels did not alleviate Fe, Zn and Cu toxicity.
As a control we tested the responses to Ni which is known to be chelated by nicotianamine.
Interestingly, nas4x-1 seedlings were highly sensitive to 60 µM Ni in contrast to wild type plants. nas4x-1 mutants developed strongly chlorotic leaves and short roots at two weeks of age and older, whereas wild type plants still remained green under these conditions (Fig. 7A). Multiple mutant analysis showed epistatic interactions of nas mutant alleles. The single and multiple mutant combinations with nas4-1 were most severely affected, the single and multiple mutants with nas3-1 (but not nas4-1) were second most affected by Ni, whereas nas1-1, nas2-1 and nas1-1nas2-1 double mutants were not affected more than wild type (Fig. 7B).
We also tested whether NAS genes were affected by Ni treatment. NAS1, NAS2 and NAS4 genes but not NAS3 were induced by Ni treatment in roots although only to a 2-3fold level, whereby general transcript levels were low for NAS4 compared to NAS1 and NAS2 (Suppl. Fig. 6). Interestingly, FRO2 as a marker for Fe deficiency was also up-regulated in Ni-treated roots suggesting that Nitreated plants might experience Fe deficiency. In leaves, NAS3 and NAS4 were up-regulated 2-3fold by Ni treatment (Suppl. Fig. 6). Hence, NAS gene expression was affected by Ni but induction was quite low. NAS3 and NAS4 responses to Ni were in agreement with Kim et al. (2005).
Hence, wild type levels of nicotianamine were used for detoxification of Ni but not of Zn, Cu and Fe. Moreover, NAS4 and NAS3 conferred higher Ni-resistance than NAS1 and NAS2 suggesting that at the level of Ni toxicity there was not full redundancy of NAS genes.

DISCUSSION
Our analysis of the first Arabidopsis nas mutant may form the ground for targeted improvement of the Fe nutritional value of crops. Our results support that by way of altering expression of NAS genes using transgenic and classical breeding approaches the seed and leaf Fe contents can be manipulated.
We have shown here that complete loss of nicotianamine resulted in a chloronerva phenotype in Arabidopsis (nas4x-2 mutant), namely in severe leaf chlorosis and sterility. Since Arabidopsis and Solanaceous plants belong to two different subclasses of the eudicots, namely Rosidae and Asteridae, we conclude that nicotianamine function is conserved in most if not all dicots.
Our study concentrated on the analysis of a knockdown mutant of nicotianamine synthase function, nas4x-1, that had residual nicotianamine. This mutant had chloronerva-like features such as leaf chlorosis, susceptibility to Fe deficiency, increased Fe uptake and Fe levels in leaves and fit/nas epistatic interactions. The mutant also was susceptible to Ni supply. Remaining nicotianamine contents in nas4x-1 still allowed the plants to complete their life cycle and to produce seeds.
Mutants with intermediate phenotypes yielded interesting results in plant biology and were often discovered in forward genetic screens. Interpretation of null mutants is frequently limited by the onset of pleiotropic responses and/or a precocious block in development. Examples for intermediate phenotypes are found in many of the developmental mutants, and they have also been valuable in the Fe homeostasis field, where for most gene families only single mutant analysis was performed.
The four NAS genes act in a functionally redundant manner since the enzyme product nicotianamine is a mobile compound in the plant. Despite of redundancy, individual NAS enzymes have gained specific properties, for example due to differential gene expression patterns such as in different tissues and by regulation through metal supply. Functional specification of NAS1 can also explain the variable NAS1 gene expression we have observed while analyzing multiple nas mutants and Fe deficiency responses. NAS1 may have responded specifically to certain developmental or environmental cues that occurred in variable manner in these latter growth conditions.
We have used this mutant to ask several questions: Why have nas4x-1 leaves been more chlorotic during the reproductive than the vegetative stage?
The late leaf chlorosis was the first remarkable phenotype of nas4x-1 plants. We found that rosette leaves of the reproductive stage of nas4x-1 plants did not have any nicotianamine whereas before during the vegetative stage such leaves contained nicotianamine although only 10 % of wild type level. We deduce from these observation that nicotianamine had disappeared from the rosette leaves during the reproductive phase and was no longer replenished through synthesis or transport. The removal of nicotianamine caused a Fe distribution problem in leaf cells which was the cause for the leaf chlorosis. Clearly, some cells or compartments of nas4x-1 plants were Fe-starved so that Fe deficiency signalling took place in the leaves. In support of this is the fact that leaves turned chlorotic and that expression of BHLH100 was induced. The transcription factor gene BHLH100 responds to local Fe deficiency as shown by previous grafting studies (Wang et al., 2007). It is very interesting to note that in nas4x-1 plants BHLH100 regulation was decoupled from the presence of Fe in the leaves. It could indeed have been that Fe deficiency-mediated stress responses that had occurred in nas4x-1 leaves up-regulated BHLH100. In support of Fe deficiency of nas4x-1 plants was also that root Fe acquisition responses had been switched on, such as IRT1 and FRO2 induction. On the other hand, nas4x-1 plants contained Fe, during the reproductive phase even more than twice the amount compared to wild type. Obviously, this Fe did not enter all target places.
However, the Fe in the leaves was sensed so that FRO3, a ferric reductase gene normally induced by -Fe in leaves (Mukherjee et al., 2006), was not induced. Therefore, FRO3 was probably not required for uptake and transport of Fe into leaves of nas4x-1 plants. Similarly, nas4x-1 leaves had similar ferritin gene FER1 expression levels as Fe-sufficient wild type leaves. Therefore, the signalling machinery that regulated FRO3 and FER1 expression in response to Fe operated and sensed Fe in nas4x-1 leaves.
The genetics of the nas4x-1 mutant also are in support that in leaves nicotianamine had disappeared during the reproductive phase. Gene expression studies and mutant T-DNA analysis had shown that only the nas2-1 allele was responsible for residual nicotianamine production in the mutant. This allele was essentially expressed in roots, whereas it was expressed to a low level in leaves.
Moreover, its gene expression levels were higher in the vegetative than the reproductive phase. For example, in the reproductive phase the total number of NAS transcripts of nas4x-1 plants was 25 % higher in roots compared to wild type, but on the other hand it was 88 % lower in leaves. This suggests that roots were able to continuously supply nicotianamine in nas4x-1. We can exclude that the higher and partially ectopic NAS2 gene expression caused higher nicotianamine production from nas2-1 than from the wild type NAS2. In fact we found that a nas1-1nas3-1nas4-1 mutant with wild type NAS2 had 3-4 times higher nicotianamine levels than the mutant with nas2-1 allele (which was nas4x-1). Hence, nas2-1 resulted in a reduced function of NAS2. The NAS leaf isoforms are normally encoded by NAS1, NAS3 and NAS4. Since NAS3 and NAS4 were mutated in nas4x-1 they could not replenish nicotianamine in leaves, especially not during reproduction when NAS3 was induced in leaves.

It remained the question, what happened to the nicotianamine that had been depleted from leaves?
This answer led to a second striking phenotype of nas4x-1 plants, namely reduced Fe contents in seeds. Our results show that wild type nicotianamine levels were essential for normal Fe supply of seeds. Nicotianamine was present in wild type seeds. nas4x-1 seeds contained nicotianamine to a lower level than wild type seeds, and they also contained Fe to a lower level. However, neither in flowers nor in developing siliques of wild type or mutant we could detect NAS gene transcripts above background level in our experiments (data not shown). In both wild type and mutant it therefore seemed impossible that reproductive parts had own nicotianamine synthase activity.
Hence, any nicotianamine that was detected in seeds must have been produced elsewhere in the plant and transported to the flowers before its deposition into developing seeds. The disappearance of nicotianamine from leaves accompanied by the onset of leaf chlorosis during reproduction suggests an important function of nicotianamine at this stage. Therefore, we explain our results such that nicotianamine was transported away from leaves towards developing seeds. Presumably nicotianamine was transported as Fe chelate. Since the nas4x-1 mutant had less nicotianamine a consequence was that less Fe reached the developing seeds, and more Fe remained in leaves. In addition due to diminished NAS activity in nas4x-1 roots we assume that less nicotianamine-Fe reached the reproductive organs directly from the roots but may have ended up in leaves in a nicotianamine-independent manner. ysl1 and ysl1ysl3 double mutants shared phenotypes of nas4x-1 such as low Fe content in seeds, Fe deficiency responses and leaf chlorosis which had suggested that YSL genes were involved in seed Fe partionning (Le Jean et al., 2005;Waters et al., 2006;Waters and Grusak, 2008). We found that YSL gene expression was affected in nas4x-1 plants. These results strongly suggest that indeed YSL and NAS genes acted in a similar biological process. The best explanation is in favour of the idea that YSL proteins are indeed the transporters for nicotianamine-Fe.
Although seeds were clearly reduced in Fe, they were only slightly reduced in Zn content but not reduced in Cu. This shows that high amounts of nicotianamine were needed for Fe deposition but low amounts were sufficient for Zn and Cu transport into seeds. An interesting phenotype was observed in naat mutant rice plants. These mutants had increased nicotianamine levels due to dysfunctional phytosiderophore synthesis which then led to reduced Fe mobilisation via strategy II.
But in return, these plants could take up Fe through an IRT-like transporter which may have been aided by the increased nicotianamine (Cheng et al., 2007). Since these plants had higher seed Fe contents it may have been possible that in these rice plants increased nicotianamine was implicated in the loading of seeds with Fe, the opposite case of nas4x-1.
Why did the nas4x-1 plants increase Fe acquisition during the reproductive phase while Fe accumulated in leaves?
Previous systemic signalling studies of a number of mutants had suggested that systemic Fe deficiency signals must exist that stem from leaves and that regulate Fe acquisition in the root (Grusak and Pezeshgi, 1996;Vert et al., 2002;Enomoto et al., 2007;Wang et al., 2007). As discussed above, nas4x-1 leaves showed signs of Fe sufficiency but also of Fe deficiency. It is possible that the FER1/FRO3 regulatory system was simply not connected with the systemic signalling mechanism for root Fe acquisition responses, whereas BHLH100 might have been.
Another explanation for induction of Fe acquisition in roots could be the high demand for Fe in the plants. Very interestingly, we have observed that IRT1, BHLH100 and FRO3 were all induced in flowers and siliques showing that these organs not only had low Fe levels but also responded to that. It is possible that due to reduced nicotianamine-Fe transport in flowers and siliques alternative routes for Fe uptake were switched on based on FRO3/IRT1. Perhaps these reproductive organs with their Fe deficiency were capable of producing a systemic signal which overruled the potential Fe sufficiency of leaves so that they acted as sink organs for Fe.
Could the reduced Fe content of nas4x-1 seeds result in a germination defect?
The answer is that a 60 % reduction of Fe in nas4x-1 seeds did not result in any germination defect on regular medium. However, upon germination on Fe-deficient medium, nas4x-1 seedlings had stronger chlorotic leaves than wild type (unpublished results). Hence, nicotianamine helped in Fe mobilization during germination when external Fe supply was low. Interestingly, this susceptibility to Fe deficiency was also apparent later in development when plants were transferred from regular to Fe-deficient medium for several days. These plants had been able to take up Fe during their postgermination growth but yet Fe deficiency was more severe than in wild type. Since NAS genes and multiple mutants) nicotianamine was needed during the Fe acquisition process. One possible explanation could be that nicotianamine is used to translocate Fe such as from the epidermis to the vascular system in roots upon Fe deficiency.
The final question for discussion is why have nas4x-1 plants been sensitive to nickel but not to elevated Fe, Zn or Cu? Obviously the normal nicotianamine levels were not used in prevention of the Fenton reaction and other toxic effects of Fe, Cu or Zn. If it was used for preventing toxic effects of these metals we would have expected the mutant to react in an oversensitive manner to these metals. On one side this finding is surprising. In fact, hyperaccumulators have mostly strongly up-regulated nicotianamine synthesis (Becher et al., 2004;Weber et al., 2004;Mari et al., 2006), and NAS overexpression can lead to increased tolerance to metals including Zn (Kim et al., 2005).
The function of high nicotianamine for Zn detoxification remains therefore dubious. Interestingly, overexpression of Thlaspi TcNAS1 in Arabidopsis led to increased sensitivity to Fe deficiency (Cassin et al., 2009). This was also surprising, as Fe deficiency is normally conferred by loss of nas function. As discussed by Cassin et al. (2009), the apparent contradictory results between nas loss of function and NAS overexpression could be explained by altered location of nicotianamine. Our metal determination analyses have shown that nas4x-1 leaves have lower Zn content than wild type.
Thus, nas4x-1 mutants translocate less Zn to leaves than wild type, and hence this may partially protect the mutants from Zn toxicity. On the other hand, less nicotianamine was available in nas4x-1 leaves and roots to chelate Zn and prevent toxic effects in the cells, and in addition stronger Fe deficiency was present. In turn, this may explain why nas4x-1 mutants and wild type reacted in the end similarly to high Zn, while only NAS overexpression was protective as shown by Kim et al. (2005).
Indeed, we have confirmed that nicotianamine can be used for detoxification of Ni through induced nickel sensitivity in nas4x-1, whereas other groups showed nickel tolerance upon increased nicotianamine in Arabidopsis or tobacco (Douchkov et al., 2005;Kim et al., 2005;Pianelli et al., 2005). With respect to this heavy metal we also found non-redundant functions of NAS genes. NAS4 and to a lower degree NAS3 contributed essentially to nickel tolerance whereas NAS1 and NAS2 could be neglected. Ni may directly affect NAS function so that perhaps NAS3 and NAS4 could be the Ni-tolerant NAS isoforms. Upon Ni treatment increased nicotianamine might have been needed for Fe mobilisation. Indeed, our results based on FRO2 expression indicate that Ni treatment had caused Fe deficiency, perhaps through competition of binding sites to nicotianamine and other compounds in the cells. Interestingly, previous studies have shown that increased nicotianamine in Arabidopsis not only resulted in tolerance to Ni but also in tolerance to Fe deficiency (Douchkov et al., 2005;Kim et al., 2005). Hence, the two processes could be physiologically linked.
Multiple mutants were obtained through crossing as indicated in the text. Genotypes of multiple mutants were confirmed by genomic PCR and subsequent genotyping in the progeny of the plants.
fit mutants and the p2x35S::FIT overexpression line were described by Jakoby et al. (2004).

Plant growth
Seeds were surface-sterilized and stratified for 2-3 days at 4°C. The agar plate system using 1x Hoagland agar square plates was conducted as described in Jakoby et al. (2004). Fe-deficient agar medium was devoid of Fe or supplemented with 50 µM ferrozine. For Ni stress conditions Hoagland medium was modified to contain 60 µM Ni (NiCl 2 ). Zn stress medium contained 200 µM ZnSO 4 (instead of the regular 2 µM Zn), Cu stress medium 28 µM Cu SO 4 (instead of the regular 1.5 µM Cu). For the hydroponic system seedlings were germinated on ¼x Hoagland agar medium in 500 µl support tubes containing a hole at the bottom for root growth. After two weeks plants were placed into aerated ¼x Hoagland liquid medium for another 2 weeks. Medium was exchanged weekly. ¼x Hoagland medium contained a quarter of the Hoagland salts and 10 µM FeEDTA (control). Solid germination medium of the hydroponic system contained no sucrose and 0.6 % plant agar. Metal deficiency in the hydroponic solution was achieved by using ¼x Hoagland lacking Fe, Cu or Zn salts, respectively. Cultivation took place at 21°C/19°C and 16 hour light/8 hour dark cycles or as indicated in the text.

Physiological plant analysis
Leaf chlorosis phenotypes were determined by manual inspection as indicated in the text.
Ferric reductase activity was measured as described by Jakoby et al. (2004).
Metal contents were determined on dried plant samples using graphite furnace atomic absorption spectroscopy with direct solid sampling (AAS vario ® 6 -graphite furnace and flame technique, Analytik Jena, Jena, Germany).

Gene expression analysis
Gene expression analysis was performend by reverse transcription-quantitative real-time PCR as previously described (Wang et al., 2007). Briefly, DNase-treated RNA was used for cDNA synthesis. SYBR green I-based real-time PCR analysis was performed by using Ex Taq R-PCR (TaKaRa, Japan) in a Mx3000P (Stratagene, USA) real-time PCR cycler. For each gene the absolute quantity of initial transcript was determined by standard curve analysis. Absolute expression data was normalized against the averaged expression values of the internal control genes EF1BALPHA2 and UBP6. Primer sequences are provided in Suppl. Table 1. Specific PCR oligonucleotides were designed using the Primer3 Software. NAS primer specificity was thoroughly verified in test amplifications.

Determination of nicotianamine contents
Nicotianamine was extracted by using a modified protocol (Neumann et al., 1999). The identity of nicotianamine was evidenced by spiking of chemically synthesized nicotianamine to plant samples.
The nicotianamine-free tomato mutant chloronerva was used as a negative control. In leaves of Arabidopsis, a recovery rate of 98% ± 7% was achieved after addition of external standard to the plant sample prior to nicotianamine extraction (Suppl. Fig. 7). The detailed protocol of the modified method is available online as Suppl. Material 1.    showed that significant differences with p< 0.05 were found for the comparisons wild type + Fe vs. -Fe, wild type + Fe vs. nas4x-1 + Fe and wild type + Fe vs. nas4x-1 -Fe (E) FER1 and FRO3 gene expression.

Determination of nicotianamine contents
Nicotianamine was extracted by using a modified protocol based on Neumann et al. (1999). Plant tissue (0.1 g) was grinded to a fine powder in liquid nitrogen and extracted in 1 ml H 2 O for 20 min at 80 °C, followed by sonification for 15 min in a waterbath (Bandelin, Berlin, Germany). Cell debris was removed by two centrifugation steps for 20 min at 15000 x g. 25 µL of the resulting supernatant was diluted with 75 µL of 0.5 M sodium-borate buffer (pH 7.7) containing 50 mM EDTA and derivatized for 45 s by the addition of 50 µL 12 mM 9-fluorenylmethyl chloroformate (FMOC, Sigma) according to Gustavsson and Betner (1990). Excess of FMOC was precipitated by the addition of 50 µL 40 mM adamantan-1-amine hydrochoride (ADAM, Sigma) in acetone:water (3:1, v/v) and removed by two centrifugation steps for 30 min at 15,000 x g and 4 °C. The derivatisation-assay (10 µl) was injected on a Nucleosil 100-5 C18, 250/4 column (Macherey-Nagel, Düren, Germany) that was connected to a Waters 600E HPLC system. Fluorescence of FMOC-derivatives was detected by using a fluorescence detector (Jasco FP 920, excitation: 263 nm, emission: 313 nm, Gain: 10, Response: Fast) and quantified by external NA standards (T. Hasegawa Co., Tokyo, Japan). Data were collected and processed with the Millenium32 software (Waters). The identity of nicotianamine was evidenced by spiking of chemically synthesized nicotianamine to plant samples. The nicotianamine-free tomato mutant chloronerva was used as a control. In leaves of Arabidopsis, a recovery rate of 98% ± 7% was achieved after addition of external standard to the plant sample prior to nicotianamine extraction (Suppl. Fig. 7 Suppl. Table 1: Gene-specific primers used for real-time qPCR analysis.

Sequence of the oligonucleotides
Gene (AGI code)