Dissection of iron signaling and iron accumulation by overexpression of subgroup Ib bHLH039 protein

Iron is an essential growth determinant for plants, and plants acquire this micronutrient in amounts they need in their environment. Plants can increase iron uptake in response to a regulatory transcription factor cascade. Arabidopsis thaliana serves as model plant to identify and characterize iron regulation genes. Here, we show that overexpression of subgroup Ib bHLH transcription factor bHLH039 (39Ox) caused constitutive iron acquisition responses, which resulted in enhanced iron contents in leaves and seeds. Transcriptome analysis demonstrated that 39Ox plants displayed simultaneously gene expression patterns characteristic of iron deficiency and iron stress signaling. Thereby, we could dissect iron deficiency response regulation. The transcription factor FIT, which is required to regulate iron uptake, was essential for the 39Ox phenotype. We provide evidence that subgroup Ib transcription factors are involved in FIT transcriptional regulation. Our findings pose interesting questions to the feedback control of iron homeostasis.

Adequate plant nutrition with iron is critical for plant development and harvest. The presence of this metal cofactor is essential for multiple biochemical electron transfer reactions in all cells. At the same time, this metal may cause deleterious effects and oxidative stress due to radical formation through the Fenton reaction. Since the bioavailability of iron varies with soil types and is mostly rather low, plants are challenged to adapt and develop an efficient iron acquisition system. Two different strategies are distinguished for iron acquisition in plants. These strategies allow plants to mobilize and take up the amounts of iron they need also under conditions of low iron supply 1 . Strategy I, found in plants except in grasses, is based on the uptake of ferrous iron upon soil acidification and iron reduction, aided by the exudation of small organic compounds like phenolics, riboflavin or acids 2 . Strategy II, on the other side, present in grasses, relies on the uptake of ferric iron chelated by mugineic acid-based phytosiderophores 3 . Both strategies are regulated at the transcriptional level by networks of iron-regulated transcription factors 2,3 . Attempts are underway to characterize the major players of iron regulation by using model plant systems. The knowledge about iron uptake regulation will enable us to design novel strategies for targeted breeding of iron-efficient crops.
Three types or subfamilies of bHLH transcription factors control iron uptake in dicotyledonous plants. The bHLH proteins FER and FER-LIKE IRON DEFICIENCY-INDUCED TRANSCRIPTION FACTOR (FIT) are encoded by single genes and are essential for iron acquisition in tomato and Arabidopsis [4][5][6][7] . While FER/FIT are a specific Strategy I component, the other bHLH subgroups are present in Strategy I and II plants. The bHLH subgroup Ib comprises bHLH038, bHLH039, bHLH100 and bHLH101 in Arabidopsis and the rice homolog IRO2 [8][9][10][11] . The bHLH subgroup IVc is represented by POPEYE (PYE), bHLH104, IAA-LEUCINE RESISTANT3 (ILR3/bHLH105) and bHLH034 in Arabidopsis and the rice homolog IRO3 12,13 .
FIT acts in the root and up-regulates the genes encoding IRON-REGULATED TRANSPORTER1 (IRT1) and FERRIC REDUCTASE OXIDASE2 (FRO2), two major players for ferrous iron uptake in the Strategy I 4,5,7,14,15 . Coumarins aid to mobilize iron in the soil and their production and export are also under control of FIT as well as some other root cell metal homeostasis functions [16][17][18][19] . FIT is induced by iron deficiency. The FIT gene is regulated at transcriptional level, which involves FIT itself 11. FIT overexpression in plants is not sufficient to promote iron uptake 5,20 . FIT requires protein activation upon iron deficiency. At least in part this is mediated by FIT protein-protein interaction and stability control, that serve to fine-tune iron acquisition in response to hormone signals and abiotic stress factors 21,22 . FIT was found in an active state, when FIT was overexpressed together with any one of the four BHLH subgroup Ib genes 23,24 .
The bHLH subgroups Ib and IVc are part of a regulatory cascade leading to the onset of iron deficiency responses. BHLH genes from the subgroup Ib and PYE from the subgroup IVc are co-expressed in Arabidopsis across different plant growth conditions. These BHLH genes are up-regulated by iron deficiency along with genes encoding regulators, transporters and enzymes with iron homeostasis functions in shoots and roots, for example, BRUTUS (BTS), FERRIC REDUCTASE OXIDASE3 (FRO3), NATURAL RESISTANCE-ASSOCIATED MACROPHAGE PROTEIN4 (NRAMP4), OLIGOPEPTIDE TRANSPORTER3 (OPT3) and NICOTIANAMINE SYNTHASE4 (NAS4) 16. Subgroup IVc proteins can interact with each other and with the E3 ligase BRUTUS (BTS) 12,[25][26][27] . BTS is discussed to be an iron/metal sensor since its stability is affected by the binding of iron/metal cofactors 26,28 . bHLH104, ILR3 or bHLH034 can bind directly to promoters of BHLH subgroup Ib genes and activate them 25,27 . Loss of function of bHLH IVc factors comes with an increased sensitivity to alkalinity-mediated or general iron deficiency, while overexpression causes iron uptake 25,27 . BHLH101 overexpression was able to partially rescue the iron deficiency phenotype of bhlh34 bhlh104 mutants by promoting IRT1, FRO2 and FIT expression 25 . Thus, these findings suggest that bHLH IVc factors act upstream of bHLH Ib and FIT transcription factors. Single knockout mutants of the four subgroup Ib BHLH genes did not have any impact on plant growth 11. Double and triple mutants showed different degrees of leaf chlorosis phenotypes upon iron deficiency. On the other hand, no phenotypic differences were apparent under sufficient iron supply 8,23,29 . Generally stronger leaf chlorosis and growth phenotypes were found in those double and triple mutants, in which a bhlh039 knock-out was present 23 . Transcriptome analysis has shown that a subset of FIT target genes was not up-regulated in response to iron deficiency in bhlh039 bhlh100 bhlh101, but among them were not FRO2 and IRT1 8. These results support that bHLH038, bHLH039, bHLH100 and bHLH101 are not fully functionally redundant.
From these observations the idea can be put forward that FIT is not only activated at protein level upon iron deficiency, but that FIT gene expression is induced downstream of the iron regulatory bHLH transcription factor cascade. Here, we present evidence for this idea by studying iron accumulation caused by bHLH039 overexpression. FIT was found essential for this phenotype, and transcriptomic data analysis gave hints on the hierarchy of regulatory responses upstream and downstream of this subgroup Ib bHLH protein.

Results
Physiological phenotypes of bHLH39 overexpression (39Ox) plants in response to iron. Previous reports indicated that subgroup Ib bHLH proteins are active in plant and yeast cells when overexpressed together with FIT 23,24 . However, we show here that overexpression of the subgroup Ib protein bHLH039 in Arabidopsis even without overexpression of FIT can lead to significant differences in iron (Fe) acquisition responses. We have generated stable transgenic Arabidopsis lines expressing BHLH039 behind the double cauliflower mosaic virus 35 S promoter and tagged with a sequence encoding the hemagglutinine (HA) antigen for protein immunodetection. Root growth was diminished upon sufficient iron supply (+Fe) in four independent lines, when compared to the wild type (Suppl. Figure 1A). In the wild type, BHLH039 was induced upon iron deficiency in both roots and shoots compared to the Fe sufficiency situation (Suppl. Figure 1B,C), as expected 11 . In the four transgenic lines, BHLH039 was expressed at higher level than in the respective wild type samples, showing that the transgene caused an over-expression of BHLH039 in leaves and in roots upon + and −Fe (Suppl. Figure 1B,C). Line JM78-5 with high BHLH039 expression was selected for detailed physiological and molecular analysis. This line was designated as 39Ox. In this line, HA-bHLH039 protein was present at +Fe and −Fe in roots and leaves of two week-old plants (Suppl. Figure 1D). Hence, we can deduce that bHLH39 protein was both present and functional at + and −Fe.
At first, we investigated whether the short root growth phenotype caused by bHLH039 overexpression was related to iron supply. 39Ox plants were grown in a six day and in a two week growth system along with wild type plants as the control, exposed to + and −Fe and examined for iron response phenotypes. As described above, six day-old 39Ox seedlings exhibited shorter roots than the wild type upon + and −Fe (Fig. 1A,B). 39Ox seedlings had dark cotyledons and red hypocotyls when exposed to +Fe, indicating stress symptoms and anthocyanin production (Fig. 1C). Red pigmentation was not found in in wild type (Fig. 1C). Anthocyanin measurements confirmed that more anthocyanins were present at + than at −Fe in 39Ox and more in 39Ox than in the wild type at +Fe (Fig. 1D).
At −Fe, Arabidopsis plants typically display an enhanced iron reductase activity in the root to mobilize external iron. Root Fe reductase activity was increased in +Fe-grown 39Ox plants compared to the wild type, as shown by qualitative and quantitative measurements (Fig. 1E,F). The two lines had similar levels of Fe reductase activity upon −Fe (Fig. 1E,F). Fe reductase activity was higher at + than at −Fe in the 39Ox plants, perhaps caused by the lack of a negative Fe regulation. Due to the high standard deviations, the differences between + and −Fe in the wild type were not found to be significant. When we analyzed the six-day seedlings at the molecular level, we found that the Fe deficiency marker genes FRO2 and IRT1 were expressed at higher level in 39Ox than in wild type at +Fe, conform with the iron reductase activity data (Fig. 1G). On the other hand, at −Fe, FRO2 and IRT1 were rather similarly expressed in 39Ox and wild type. In the wild type, FRO2 and IRT1 expression were induced at -versus +Fe, which was not the case in 39Ox (Fig. 1G). FIT was also up-regulated in 39Ox compared to wild type at +Fe (Fig. 1G). BHLH039 expression was found about 100 times increased at +Fe and about tenfold at −Fe in 39Ox compared to the wild type, demonstrating again the overexpression effect (Fig. 1G). Interestingly, FIT, FRO2 and IRT1 were hardly expressed in 39Ox leaves in either Fe supply condition. On the other hand, BHLH039, was induced in wild type leaves at −Fe versus +Fe and was overexpressed in 39Ox leaves (Suppl. Figure 2A,B). FIT, FRO2 and IRT1 were highly up-regulated at +Fe in 39Ox roots, where BHLH039 was expressed ten to 100fold higher than in the wild type (Suppl. Figure 2C,D). BHLH038 and BHLH100 were not expressed in the 39Ox situation neither at + nor −Fe in leaves and roots (Suppl. Figure 2B,D). Lack of BHLH038 and BHLH100 expression could indicate a reduced sensing of Fe deficiency in 39Ox compared to the wild type. On the other hand, BHLH101 showed a distinct expression compared to BHLH038 and BHLH100. BHLH101 was expressed in 39Ox similarly as in the wild type at+Fe, but the expression was reduced at −Fe (Suppl. Figure 2B,D).
Because of the generally increased iron mobilization phenotypes, we suspected that 39Ox plants might have accumulated iron. Iron measurements in leaves and seeds of 39Ox plants grown under Fe sufficiency showed a two-and fourfold increased Fe content, respectively, compared to the wild type (Fig. 1H). It was then interesting to test whether 39Ox seedlings might have experienced oxidative stress caused by the elevated Fe contents. Hydrogen peroxide measurements showed that indeed, H 2 O 2 levels were higher in 39Ox exposed to +Fe compared to all other samples (Fig. 1I).
Taken together, overexpression of bHLH039 protein resulted in high Fe acquisition and high root to shoot and seed translocation of iron in seedlings. Due to this iron accumulation, 39Ox plants developed iron toxicity-based oxidative stress symptoms.
Comparative transcriptome profiling of 39Ox plants. As transcription factor, bHLH039 acts in a regulatory iron sensing and iron regulation cascade. We conducted a transcriptomic microarray analysis to investigate the molecular phenotypes of 39Ox in more detail and to determine how 39Ox plants responded internally to the increased bHLH039 and Fe levels. We compared gene expression between wild type and 39Ox six day-old seedlings exposed to + and −Fe. We chose the six-day growth system because we had previously collected transcriptome data sets for this condition using wild type, fit mutant and a triple bhlh knockout mutant 8,18 . The transcriptomic study was based on the CATMA two-color microarray and included three biological replicates of the samples. Similar CATMA microarray data were recently collected with fit mutant and wild type 18 . Genes with at least 1.5fold statistically significant differential regulation were retained for further analysis (Suppl. Table 1).
We found that 3745 genes were differentially regulated at +Fe comparing 39Ox with wild type, comprising 1786 down-regulated and 1959 up-regulated genes. BHLH039, IRT1, FRO2 and FIT were up-regulated, which confirmed the overexpression effect and the results described in the previous paragraph (Suppl . Table 1A). 1348 genes were differentially expressed between 39Ox and wild type at −Fe, comprising 813 up-regulated and 535 down-regulated genes (Suppl . Table 1B). Several Fe-regulated genes were chosen and successfully used to validate the microarray data by quantitative RT-qPCR (Suppl. Figure 3).
To determine which physiological and cellular pathways were affected in 39Ox based on the transcriptome profiles, we conducted a gene ontology (GO) term enrichment analysis using all genes differentially regulated in 39Ox at +Fe as well as −Fe (Fig. 2, Suppl. Table 2). Genes associated with GO categories related to stress, defense and hormone responses, regulation and metal homeostasis, secondary and cell wall metabolism and hormone biosynthesis were predominantly up-regulated in 39Ox versus wild type at + and −Fe (Fig. 2, Suppl. Table 2A,C). GO categories indicative of stress, light and hormone responses, photosynthesis, rhythm and chlorophyll biosynthesis and lipid biology were predominantly down-regulated in 39Ox versus wild type at +Fe, and to a lesser extent at −Fe (Fig. 2, Suppl. Table 2B,D). At −Fe, we also observed a down-regulation of cell wall biosynthesis and glucosinolate metabolism in 39Ox compared to the wild type (Suppl. Table 2D). The overall transcriptome data therefore indicated that 39Ox suffered from stress symptoms related to metal homeostasis.
We suspected that +Fe-grown 39Ox plants should display gene expression patterns reflecting increased iron mobilization and Fe accumulation. To check this, we compared the list of 3745 differentially expressed genes between 39Ox versus wild type at +Fe with that of 1128 Fe-regulated genes 18 (Suppl. Table 3). We could identify an intersection of 410 regulated genes (Fig. 3, Suppl. Table 3A). 184 out of the 410 genes were down-regulated in the wild type in response to −Fe, a condition in which BHLH039 was highly expressed in the wild type. Among them, 81 genes were also down-regulated in 39Ox, suggestive of a repression of these genes in the presence of bHLH039 protein. These 81 genes included ribosomal genes, chloroplast function-related genes and IRT3 (Fig. 3, Suppl. Table 3A,B). However, 103 of the 184 genes were down-regulated at -versus +Fe in the wild type (down-regulated in the presence of bHLH039 at −Fe) and up-regulated in 39Ox versus wild type at +Fe (up-regulated in the presence of bHLH039 at +Fe). These 103 genes could therefore not have been regulated by bHLH039 itself but by the level of Fe. These 103 genes comprised the three ferritin genes FER1, FER3 and FER4, typical high iron and oxidative stress marker genes 30 , and NAS3, YSL1 and YSL3, involved in long-distance phloem-based iron translocation 31,32 (Fig. 3, Suppl. Table 3A). Among the 103 genes regulated in this group, a large coexpression cluster was found, related to the GO categories stress and oxidation-reduction regulation (Suppl . Table 3B). 226 out of the 410 genes were up-regulated in the wild type upon −Fe versus +Fe, when BHLH039 was highly expressed. Among them, 95 genes were also up-regulated in the +Fe 39Ox situation, reflecting indeed an expression pattern conferred by the presence of bHLH039. These genes included −Fe-responsive genes FIT and FIT targets, e.g. FRO2, IRT1, AHA7, MTPA2, At3g07720, MYB10, At3g12900, F6′H1, PDR9, COPT2, CYP82C3, as well as Fe homeostasis genes NAS1, IREG2 and IRT2 18 (Fig. 3, Suppl. Table 3A). 131 among the 226 −Fe-up-regulated genes were down-regulated in 39Ox, indicative of a response to Fe but not to bHLH039. These genes, being also part of several co-expression clusters, comprised several photo-oxidative stress and circadian rhythm regulators like LHY1, CCA1, COL2, ELIP. Very interestingly this list of genes also contained NAS4 and OPT3, two components for phloem-based iron transport and transmission of a long-distance iron signal 32,33 , as well as At1g47400 and At3g56360, that are part of the Fe deficiency-regulated co-expression cluster of BHLH039 16 (Fig. 3, Suppl. Table 3A,B). Taken together, nearly one half of the iron-responsive genes differentially expressed in 39Ox versus wild type at +Fe were iron deficiency response markers, whereas more than a half of these iron-responsive genes were suggestive of a response to high internal Fe levels. Hence, 39Ox plants displayed at the same time characteristic symptoms of iron deficiency and symptoms of iron sufficiency and even iron stress. In particular, the FIT target coexpression cluster was generally induced in 39Ox at +Fe, indicative of the Fe deficiency response. On the other side, the iron homeostasis co-expression cluster containing BHLH Ib and IVc genes, BTS, NAS4 and OPT3, was either not regulated or down-regulated, which reflected a high iron response of  Table 2. 39Ox. These patterns of 39Ox-mediated iron-related transcriptome changes were confirmed when the expression patterns of 162 robust Fe-regulated genes were studied in 39Ox (Suppl. Table 3C,D).
Next, we used the transcriptome data for confirmation of downstream target genes that depend for their expression on bHLH039. Previously, 134 genes were identified, that are differentially expressed upon −Fe in six day-old seedlings of the triple knockout mutant line 3xbhlh ( = bhlh039 bhlh100 bhlh101) versus wild type 8. This list of 134 genes (denoted groups I and III in ref. 8) contained 52 genes, that we found to be differentially regulated in 39Ox versus wild type. 25 out of the 52 genes were up-regulated in the 3xbhlh and down-regulated in 39Ox versus wild type. They were mostly related to chloroplast, rhythm and photo-oxidative stress reponses in leaves (Suppl . Table 3E,F). 18 out of the 52 genes were down-regulated in 3xbhlh and up-regulated in 39Ox. These genes comprised a large coexpression cluster and four FIT and bHLH039 potential downstream target genes, namely At3g07720, At3g12900, CYP82C4 and COBL6 (Suppl. Table 3E,F). Therefore, the 39Ox transcriptome data indicated again that these latter four genes might be targets regulated by bHLH039.
Finally, we investigated the regulation of 34 FIT-dependent and root-expressed genes that we had previously described 18, 33 of them showed differential regulation in 39Ox versus wild type (Suppl . Table 3G). 29 genes were up-regulated in the presence of FIT and in 39Ox and included the FIT-dependent iron uptake genes IRT1 and FRO2. Two genes down-regulated by FIT were also found to be down-regulated in 39OX, namely ZIP2 and SCPL31. Hence, the regulation of FIT downstream genes in 39Ox plants likely depended on the presence of FIT. Since FIT is an essential protein for iron uptake, the 39Ox effect is well explained by the activation of the FIT response pathway through bHLH039.

Requirement of FIT for 39Ox phenotype.
If FIT is indeed an important target activated downstream of bHLH039, we predicted that FIT should be required for regulation of the FIT targets and the 39Ox phenotype at +Fe.
To test this, we investigated the 39Ox phenotype in the background of fit loss of function mutant plants. Such fit mutant plants do not mobilise sufficient iron and suffer from severe leaf chlorosis 5. We found that in contrast to 39Ox, 39Ox/fit plants showed similar morphological and physiological phenotypes as fit mutants. 39Ox/fit plants had leaf chlorosis and did not display the short root phenotype of 39Ox (Fig. 4A,B). Instead, 39Ox/fit roots grew similarly as fit roots. They were as long as wild type roots at +Fe, and reduced in length at −Fe (Fig. 4B). 39Ox/fit plants did also not have the strong anthocyanin pigmentation of 39Ox (Fig. 4C,D). However, plants with fit background showed more anthocyanin pigmentation than the wild type at −Fe, probably due to the iron deficiency stress. Iron reductase activity was low in 39Ox/fit at + and −Fe like in fit plants (Fig. 4C). Furthermore, gene expression analysis confirmed that BHLH039 was overexpressed. FRO2 and IRT1 gene expression were low and not highly induced by −Fe nor by +Fe in 39Ox/fit, as it was observed in fit mutants (Fig. 4D). Gene expression levels of FIT-dependent genes AT3g07720, At3g12900, MTPA2 and CYP82C were also comparably low in 39Ox/fit and fit mutant (Suppl. Figure 4). HA-tagged bHLH039 protein was equally expressed in 39Ox plants with wild type as well as fit mutant background (Fig. 4E,F). These results indicate that a functional FIT was required for the action of bHLH039 in 39Ox. FIT did not affect the protein abundance of bHLH039.  Table 3.
Above, it was shown that Fe reductase activity was found along the roots of 39Ox seedlings. Thus, we suspected that FIT expression might also have been activated all along the root in response to 39Ox. To investigate in which region of the root the FIT gene was activated, we localized FIT promoter activity based on FIT promoter-driven beta-glucuronidase reporter activity in transgenic pFIT::GUS plants with 39Ox background. In wild type background, FIT promoter activity was detected mainly in the epidermis cells of the elongation root zone in Fe-deficient conditions, as expected 5 (Fig. 5). GUS activity was stronger in plants with 39Ox than wild-type background at +Fe (Fig. 5B). GUS activity was detected along the entire 39Ox roots, but not in hypocotyls and cotyledons (Fig. 5A). Thus, overexpression of bHLH039 activated the FIT promoter in an ectopic manner in roots.

Discussion
Here we showed, by using overexpression plants and combinations of transgenic and mutant plants, that bHLH039 induces directly or indirectly FIT in the Fe deficiency regulatory cascade and that bHLH039 is active even in the presence of sufficient iron.
A remarkable conclusion from this study is that bHLH039 protein was not only abundant in 39Ox plants but also functionally active regardless of Fe supply. On the contrary, FIT and HA-FIT were present at + and −Fe in plant cells in an overexpression situation, but only active at −and not +Fe 5,20 . It was known that BHLH039 is manifold induced at −Fe, while at +Fe hardly any transcripts are produced. In contrast, the FIT gene is expressed at +Fe and induced about threefold at −Fe 5. Hence, the action of bHLH039 is regulated in the plant mainly via gene expression regulation, while the action of FIT is regulated at transcriptional and protein level. The transcriptional activation of BHLH039 makes sense in view of the current models about the iron deficiency regulatory cascade. According to these models gene expression of BHLH039 and of other BHLH subgroup Ib genes is induced by subgroup IVc bHLH factors in response to −Fe 25,27 .
The most striking 39Ox phenotypes were apparent at +Fe but not at −Fe, showing that the utilization of Fe was affected in the 39Ox plants. First, the 39Ox plants displayed high expression of iron acquisition genes, high Fe reductase activity, high Fe contents in leaves and seeds and high ferritin gene expression, in contrast to the wild type. These phenotypes clearly show that 39Ox plants took up more Fe from the growth medium than the wild type. Second, 39Ox plants displayed oxidative stress symptoms at +Fe, like up-regulation of oxidative stress gene expression, anthocyanin pigmentation, short root phenotypes and H 2 O 2 production. Both Fe accumulation and oxidative stress phenotypes of 39Ox were Fe-dependent. We interpret the observed 39Ox phenotypes as follows (depicted in a model in Suppl. Figure 5): The transcription factor bHLH039 up-regulated together with low levels of FIT protein the FIT gene. FIT and bHLH039 then induced iron acquisition genes like FRO2 and IRT1. The resulting FRO2 activity and IRT1 stimulated iron uptake into the roots. Iron was translocated to shoots and high amounts of Fe accumulated in leaves and seeds. The high Fe levels in roots and shoots then caused toxicity symptoms such as oxidative stress, short root growth and anthocyanin pigmentation. At least half of the Fe-regulated differentially expressed genes in 39Ox versus wild type were indicative of a high iron response that normally serves to alleviate iron stress. For example, ferritin FER gene expression was increased, together with NAS3, YSL1 and YSL3. These genes were up-regulated in response to the high Fe status in the shoot, indicating that iron was bound and/or circulated in the plant. Ferritin was presumably used to suppress oxidative stress and to store Fe in plastids [30][31][32] . Even though these protective responses were enhanced, we need to assume that they were not sufficiently effective to alleviate the iron stress in 39Ox. Surprisingly, repressive iron signals, that normally would be expected to delimit excessive iron uptake, were switched off or ineffective in +Fe-grown 39Ox plants. It has been hypothesized that BTS could be a sensor of Fe and negative regulator in the transcription factor cascade 12,26,28 . BTS protein function and regulation in plant cells in response to cellular Fe awaits further experimentation 26. BTS was not up-regulated in 39Ox and therefore BTS could not likely have repressed the cascade. Perhaps low BRUTUS (BTS) E3 ligase protein levels or levels of related proteins were active in 39Ox and effectively suppressed the Fe deficiency cascade at the level of subgroup IVc bHLH proteins. BHLH IVc genes are not Fe-regulated, with the exception of PYE that we did not detect in our list of 39Ox up-regulated genes either. Moreover, very interestingly, BHLH038, BHLH100 and BHLH101 were down-regulated in 39Ox at + and −Fe. Since subgroup IVc bHLH proteins induce the promoters of BHLH Ib genes, we suspect that the Fe deficiency cascade upstream of BHLH subgroup Ib genes was switched off 25,27 . Furthermore, four genes linked with that cascade and normally induced by −Fe and co-expressed with BHLH039, namely OPT3, NAS4, At1g47400 and At3g56360, were even down-regulated in 39Ox. Overexpression of bHLH039 must have counteracted any suppressor effects by BTS or related proteins. Our results are therefore in agreement with previous published data, such as overexpression of a bHLH subgroup IVc protein leading to iron over-accumulation phenotypes similar as seen in 39Ox or bts loss of function plants 25,27 . A long distance iron suppressor signal transported via the phloem from shoot to root has been hypothesized to communicate the shoot iron status to the root 11,32,34,35 . A long-distance iron signal was suspected to be loaded into the phloem of leaf cells via OPT3 33,34 . Such a long-distance signal was either inactive in 39Ox or in case it acted at the level of the transcription factor cascade upstream of bHLH039 not effective in 39Ox. The down-regulation of OPT3 in 39Ox could reflect a potential mechanism for the inhibition of transmission of a long distance iron suppressor signal from leaf to root. However, it could also be that OPT3 may not be involved in the transmission of a signal. Finally, local suppressor mechanisms for iron uptake are described that suppress the activity of individual proteins mediating iron uptake in the root epidermis cell. For example, IRT1 localization is altered in the presence of metals 36 . Under sufficient Fe, IRT1 is more prone to degradation 37 . FRO2 is described to be inactive upon sufficient Fe despite of the presence of this protein in a FRO2 overexpression situation 38 . It is surprising that these negative regulation mechanisms acting inside root epidermal cells were also not in place in the 39Ox background. This suggests that protein regulation factors for IRT1 and FRO2 might be related to the action of FIT.
bHLH039 action required functional FIT. 39Ox/fit plants lost the Fe-dependent 39Ox phenotypes. Therefore, FIT must have been functionally active in 39Ox plants, and again, potential inhibitory components previously reported to negatively control FIT protein activity 20 were not present or in place in 39Ox. This gives support to the idea that the main activator of FIT protein in 39Ox was bHLH039. The same occurs in wild type plants exposed to −Fe (see model in Suppl. Figure 5). Once subgroup Ib bHLH factors are produced in root cells in response to the regulatory Fe deficiency cascade, FIT becomes activated at protein level through direct protein interactions, as shown 23,24 . FIT was found to be involved in its own feed-forward gene expression regulation 11 . In 39Ox plants, the FIT gene promoter was activated in an ectopic manner in roots but not in leaves. Hence through the interaction with bHLH039 in 39Ox plants, FIT might have further stimulated its own transcription all along the root. This scenario can also explain why molecular Fe acquisition responses like IRT1 and FRO2 induction were only switched on in roots but not in leaves of 39Ox. FIT is a root-specific transcription factor 5 and hence a basal level of FIT was required at +Fe for bHLH039 action in 39Ox. Different scenarios of direct or indirect positive bHLH039-mediated FIT promoter induction are conceivable. One possibility is that bHLH039-FIT heterodimers bind to the FIT promoter (direct activation of FIT promoter by FIT-bHLH039 heterodimers, evidence for FIT heterodimerization in 24 ). Alternatively, FIT may bind to its promoter as a homodimer being activated by binding in addition to bHLH039 (direct activation of FIT promoter by multiprotein complexes consisting of FIT homodimers and bHLH039, evidence for FIT homodimerization in 24 ). Finally, an additional yet unknown transcription factor could be induced by bHLH039 and FIT and then serve to activate the FIT promoter and FIT target genes (indirect regulation of FIT promoter by FIT and bHLH039). High Fe stress-related anthocyanin production was not found in 39Ox leaves with fit mutant background. Hence, such shoot responses could not have been under direct or indirect control of bHLH039. Therefore, anthocyanin production was more likely the result of secondary effects related to iron accumulation in the shoot rather than the action of bHLH039 in the cells.
In conclusion, bHLH039 is a direct or indirect activator of FIT transcription in the root. Any suppressive Fe signals were inactive in 39Ox and did not inhibit Fe uptake. By manipulation of bHLH039 expression, it was thus possible to enhance iron levels in shoots and seeds. Our findings raise interesting questions with regard to the coupling of iron signals with post-translational negative control mechanisms acting upon iron uptake components. An interesting question to pursue in the future is to tackle the roles of bHLH subgroup Ib factors in the shoot. Since the Fe levels of 39Ox leaves and seeds were enhanced, it is conceivable that bHLH039 was involved in regulating the delivery of Fe to sinks.

Methods
Plant material and growth conditions. The wild type ecotype used was Col-0. fit-3 (termed fit throughout the text) and pFIT::GUS plants were previously described 5 .
bHLH039 overexpression plants were constructed as follows: A genomic DNA fragment containing exons and introns of BHLH039 was inserted into pALLIGATOR yielding a double cauliflower mosaic virus 35S promoter 2xpCaMV::triple hemagglutinine (HA) tag-gBHLH039 coding cassette, according to a previously described procedure 20 . Transgenic lines were generated, genotyped and multiplied until homozygous lines were obtained. Four lines were investigated and one line selected for further analysis (here designated 39Ox), as described in the text (see Suppl. Figure 1). 39Ox/fit-3 (termed 39Ox/fit) and pFIT::GUS/39Ox plants were generated after crossing and selection of homozygous lines.
For experiments, seeds were surface sterilized and plants raised in a six-day agar plate assay in the presence of 50 microM Fe (+Fe) or 0 Fe (−Fe) or in a two-week agar plate assay with a three-day exposure to medium containing 50 microM Fe (+Fe) or 50 microM ferrozine and 0 Fe (−Fe) using a modified Hoagland medium, as previously described 5,22 and as indicated in the text and figure legends.
Morphological and physiological plant analysis. Seedlings grown on agar plates were pictured with the Axio Zoom.V16 Stereo-Zoom-Microscope (Zeiss, Jena, Germany). Root lengths were measured after photography with the help of the ImageJ software tool.
Anthocyanin extracts were prepared from 15 preweighted seedlings by incubation for 5 minutes at 95° Celsius and then overnight at 25° Celsius with a propanol:HCl:H 2 O (18:1:81 v/v/v) mixture, as decribed 39 . The extract was photometrically measured at OD535. After background subtraction (OD 650 ) the values were normalized to the respective sample weights.
Iron reductase activity was determined using a quantitative liquid and qualitative agar plate assay with ferrozine as indicator for Fe 2+ , as previously described 40 . Quantitative iron reductase activity was normalized to root length.
For iron content determination, leaves were harvested, dried, powdered and processed for graphite furnace atomic absorption spectroscopy, as previously described 22 . Seeds were dried, weighed and microwave-digested (Multiwave 3000, Anton Paar GmbH, Graz, Austria) with concentrated HNO 3 . Fe concentrations were determined by inductively-coupled plasma-optical emission spectrometry ICP-OES with a Czerny-Turner type monochromator (Ultima 2, Horiba Jobin-Yvon, New Jersey, USA, and Longjumeau, France).
Quantitative GUS activity analysis using total protein extract was conducted as previously described, following a fluorescence measurement method with MUG substrate (4-methylumbelifetyl-beta-D-glucuronide) 5 .

Gene expression analysis by RT-qPCR.
A detailed protocol of the RT-qPCR procedure is available 40,42 . RNA from about 60 seedlings per biological replicate was extracted. cDNA was synthetized with poly dT primer from DNase-treated total RNA (300-500 ng) and then analyzed by real time qPCR using the DyNAmo ColorFlash SYBR Green qPCR Kit (Thermo Scientific, Waltham, USA) and the ICycler (Bio-Rad, Hercules, USA), as described 42 . Primers and primer sequences are listed in 8,18 . qPCR data accuracy and reliability was checked as described 40,42 . The initial transcript amount of two technical replicates was determined by standard curve analysis using mass standards and averaged. The absolute expression data were normalized with the constitutive expression control gene encoding elongation factor 1Bα1 (EF1Bα1). The mean of three biological replicates per sample was calculated and used for statistical treatment of RT-qPCR data.
Microarray-based transcriptome analysis. Six-day wild type and 39OX seedlings were raised on +Fe an −Fe in three biological replicates. 60 seedlings were pooled per sample and RNA extracted using the Qiagen plant RNA extraction kit (Qiagen, Hilden, Germany). A two-color CATMAv6 microarray analysis was performed including a dye-swap control, following the published procedure 43 . After the statistical analysis gene lists were retained with log 2 normalized intensities per samples, log 2 normalized ratios, and p-values (FDR correction). Probes with a Bonferroni p-value of 0.05 or less and a fold change of 1.5 or more were considered differentially expressed. The same RNA preparations were also used for validation by RT-qPCR of selected genes.
Transcriptome data sets were compared using Excel. Gene ontology (GO) term and coexpression analysis was conducted and visualized using VirtualPlant 44 , Revigo 45 and STRING 46 , as indicated in the text and legends. Immunoblot analysis. Immunoblot analysis was conducted according to 21 . Around 40 six day-old seedlings were harvested, frozen in liquid nitrogen and grinded using the Precellys (PeqLab Biotechnologie GmbH, Erlangen, Germany). Per 1 mg plant powder 1 microl of 2x Laemmli buffer was added. After centrifugation 10 microg heat-denatured total protein was loaded on a 12% SDS-polyacrylamide gel and proteins separated by electrophoresis. Proteins were electro-transferred to a nitrocellulose membrane (GE Health Care, Amersham) with transfer buffer. HA 3 -bHLH039 protein was detected after immunoblotting with anti-HA-horseradish peroxidase (HRP)-coupled high-affinity monoclonal rat antibody (3F10 clone 1:1.000 dilution; Roche Diagnostics GmbH, Basel, Switzerland) by chemiluminescence (ECL chemiluminescence kit, GE Health Care, Little Chalfont, UK) and the FluoChem Q (Biozym, Hamburg, Germany) device. The membrane was stripped and reused for the detection of actin protein with anti-actin antibodies (rabbit anti-ACT, 1:5000 dilution; Agrisera, Vännäs, Sweden) and goat anti-rabbit-HRP-coupled secondary antibody (Pierce, 1:1000 dilution; Thermo Fisher Scientific, Carlsbad, USA).
The intensities of the bands were determined using the software AlphaView (FluorChem Q). The local backgrounds were subtracted. HA signal intensity was normalized to that for actin. Alternatively and as mentioned in the text and figure legend, normalization was conducted to major bands observed and measured after Ponceau S staining of the membrane for total protein.
Statistical analysis. Statistical analysis was conducted using the one-way ANOVA test followed by the post-hoc Tukey test for pair-wise comparisons. p-values of 0.05 or lower indicate statistical significance between two mean sample values, represented in the figures by different letters. p-values above 0.05 indicate no statistically significant difference between the two compared sample values, indicated by the same letter above the two sample values in the figures. The diagrams represent the mean values of biological replicates and the error bars reflect standard deviations.