Physiological and interactomic analysis reveals versatile functions of Arabidopsis 14-3-3 quadruple mutants in response to Fe deficiency

To date, few phenotypes have been described for Arabidopsis 14-3-3 mutants or the phenotypes showing the role of 14-3-3 in plant responding to abiotic stress. Although one member of the 14-3-3 protein family (14-3-3 omicron) was shown to be involved in the proper operation of Fe acquisition mechanisms at physiological and gene expression levels in Arabidopsis thaliana, it remains to be explored whether other members play a role in regulating iron acquisition. To more directly and effectively observe whether members of 14-3-3 non-epsilon group have a function in Fe-deficiency adaptation, three higher order quadruple KOs, kappa/lambda/phi/chi (klpc), kappa/lambda/upsilon/nu(klun), and upsilon/nu/phi/chi (unpc) were generated and studied for physiological analysis in this study. The analysis of iron-utilization efficiency, root phenotyping, and transcriptional level of Fe-responsive genes suggested that the mutant with kl background showed different phenotypes from Wt when plants suffered Fe starved, while these phenotypes were absent in the unpc mutant. Moreover, the absence of the four 14-3-3 isoforms in the klun mutant has a clear impact on the 14-3-3 interactome upon Fe deficiency. Dynamics of 14-3-3-client interactions analysis showed that 27 and 17 proteins differentially interacted with 14-3-3 in Wt and klun roots caused by Fe deficiency, respectively. Many of these Fe responsive proteins have a role in glycolysis, oxidative phosphorylation and TCA cycle, the FoF1-synthase and in the cysteine/methionine synthesis. A clear explanation for the observed phenotypes awaits a more detailed analysis of the functional aspects of 14-3-3 binding to the target proteins identified in this study.

www.nature.com/scientificreports/ iron-sufficient condition (20 µM Fe), while the shoot weight of Wt was significantly lower compared to all the mutants (Fig. 2). At 0 µM Fe, all genotypes produced less biomass, whereat the fresh weight of both shoot and root of unpc were less than those of Wt. At 2 µM Fe, only klpc showed a significant growth reduction of both shoot and root (Fig. 2). In the same plants as described above, the Fe contents of dry shoot and root material were measured. As expected, Fe shortage treatment reduced Fe content in the root and shoot tissues of all genotypes, with the greatest difference measured in the roots (Fig. 3). The comparison between genotypes showed that under iron-deficient condition (0 µM Fe-EDTA) and low-iron condition (2 µM Fe-EDTA), the Fe content in Wt shoots and roots were lower than those of klun while no difference was observed when plants were grown under iron-sufficient condition (20 μM Fe-EDTA). When plants were grown at ½ strength Hoagland with 0 µM and 2 µM Fe-EDTA, the Fe content in Wt shoots and roots did not differ from the Fe content in unpc (Fig. 3A,B). As a measure of Iron-Utilization Efficiency (IUE), we calculated the ratio between fresh weight and the Fe content of the shoot and root (Fig. 3C,D). Strikingly, the IUE of Wt shoot and root at 0 µM and 2 µM Fe-EDTA were much higher than those of klpc and klun, while unpc only showed significantly lower IUE than Wt when the plants were Fe starved (Fig. 3C,D).

Root growth of Wt and 14-3-3 qKOs as affected by Fe-deficiency. The reduction strategy can lead
to an improved capacity for Fe uptake via inducing a series of root morphological changes, e.g. root hair length and root tip swelling 27 . Figure 4 shows root growth of mutant plants as compared with the Wt plants grown on the same agar plate. We observed that the growth of the main root of klun and unpc plants is less than that of Wt plants under Fe-sufficient condition (Fig. 4 B-D). Under Fe-deficient condition, the main root of klpc was shorter than that of Wt (Fig. 4B), while the main root length of klun was significantly longer as compared to Wt (Fig. 4C). There was no obvious difference in the main root length of unpc and Wt under Fe-deficient condition (Fig. 4D). Phylogenetic tree of 14-3-3 superfamily in Arabidopsis. 14-3-3 s with the three closely related gene pairs of which T-DNA insertion lines are used in this study shown in blue boxes. The phylogenetic tree was constructed using the full-length amino acid sequence and Pi was used as outlier, the support branch values was marked in red. The division between the epsilon and non-epsilon group is indicated. www.nature.com/scientificreports/ We also measured the growth of lateral roots in the same 14-3-3 mutant plants. The total lateral root length of klpc and unpc were indistinguishable from that of WT plants, while total root length of lateral roots in klun plants was significantly longer than that of Wt plants under both Fe-sufficient and Fe-deficient condition (Fig. 4).   www.nature.com/scientificreports/  www.nature.com/scientificreports/ Quantitative affinity-purification mass spectrometry analysis of the 14-3-3 interactome of Wt roots as affected by Fe deficiency. In view of the reported Fe deficiency-induced changes in protein amounts 11 and phosphorylation 13 , we addressed the question whether and how Fe-deficiency affects the root 14-3-3 interactome, which is phosphorylation dependent. To this end, we carried out a semi-quantitative interactome analysis by affinity-purification mass spectrometry (qAP-MS) to assess Fe deficiency induced changes in 14-3-3/target interaction. To exclude proteins that bind non-specifically to the beads, we also performed a mock pull-down with empty beads along with the test pull-down experiments (Fig. S1). In this experiment, we included the "unused" values generated from the software ProteinPilot (version 3.0; Applied Biosystems, Foster City, CA, USA; MDS Sciex, https:// sciex. com/ produ cts/ softw are/ prote inpil ot-softw are) 31 . After removal of contaminant and false positive interactors (for details see M&M), a curated list of 117 proteins was identified in the pull-down with either Wt or klun roots protein extract, with unused value > 2 in at least 2 independent samples remained (Table S1). These proteins are either so-called primary interactors (bind directly to 14-3-3) or secondary interactors (part of a multiprotein complex of which one protein is a primary interactor). A comparison between our results (Table S1) and published 14-3-3 interactome studies shows that 24 proteins identified in this study have been previously reported as 14-3-3 interactors interactors 19,21,32,33 , such as Glucose-6-phosphate 1-dehydrogenase 3 (G6PD3), ATP synthase (ATP1, ATP5). 16 proteins from our list were reported as Fe-responsive proteins, amongst which several enzymes involved in S-adenosylmethionine synthesis, two cytosolic invertases, Ferretin-1, germin-like protein GLP5 (a plasmodesmata-localized protein involved in the regulation of primary root growth) 34 and others 11,13,35 . We assigned the identified binding proteins to different KEGG categories, as shown in Fig. S2. The majority of the identified proteins are related to "carbohydrate and energy metabolism", "amino acid metabolism", "protein folding, sorting and degradation", and "transport and catabolism" (Table S2). As will be discussed later, the 14-3-3 interactome is not a random collection of proteins, but consists of distinct networks of interrelated proteins and protein complexes, involved in glycolysis, methionine metabolism, oxidative phosphorylation. Further, three major protein complexes were identified based on known and predicted interactions: the head structure of the FoF1-synthase and vacuolar V-ATPase, tubulins and the TCP-1/cpn60/HSP chaperones, as well as four functional networks: elongation factors, chaperones, glycolysis and TCA cycle and cysteine/methionine metabolism (Table S3).

Comparison of 14-3-3 interactome under control and Fe deficient conditions.
To study the changes of 14-3-3-target interaction in response to Fe deficiency, we further analysed the abundance of binding proteins identified from the pull-down experiment conducted with roots extract under Fe-sufficient or Fe-deficient condition in both Wt and klun plants, based on normalized intensity-based absolute quantification (iBAQ) values (normalized to the bait proteins in that run). The iBAQ intensities act as a measure of protein abundance and can be used to compare protein abundance between different samples. Among the 117 identified proteins, 27 and 17 proteins were significantly changed upon Fe deficiency in Wt and klun, respectively (Tables S4 & S5). According to the Venn analysis, 21 and 10 proteins were uniquely changed in Wt and klun pull-down experiment, respectively (Fig. 6A). Moreover, we found 7 proteins that were commonly changed upon Fe deficiency in both the Wt and klun pull-down experiment, and most of them showed similar trends in klun as well as those in Wt upon Fe deficiency (Tables S4 & S5). To visualize the entire data set, a heatmap was generated to represent the normalized abundance of proteins in each pull-down experiment (Fig. 6B). A total of 37 14-3-3 binding proteins showed clear separation between pull-down experiment on plants grown under Fe-deficient and Fe-sufficient condition (Table S6). Especially, S-adenosylmethionine synthase 1 (MAT1, AT1G02500), Elongation factor EF-2 (LOS1, AT1G56070), ATP synthase subunit alpha (ATP1, ATMG01190), and mitochondrial ion transporting ATP synthase beta-subunit (AT5G08670) were significant enhanced in their interaction with 14-3-3 proteins in both Wt and klun Fe-deficient plants (Table S6). Five proteins, including mitochondrial-processing peptidase subunit beta (MPPBETA, AT3G02090), ATP synthase subunit delta (ATP5, AT5G13450), TCP-1/cpn60  www.nature.com/scientificreports/   www.nature.com/scientificreports/ chaperonin family protein (AT1G67760), PHOS32 (AT5G54430) and Dirigent protein 6 (AT4G23690) solely increased in the pull-down with extract of Wt plants upon Fe deficiency. Moreover, the abundance of Glucose-6-phosphate 1-dehydrogenase 3 (G6PD3, AT1G24280), Neutral invertase 2 (CINV2, AT4G09510), Calciumdependent protein kinase 3 (CDPK6, AT4G23650) and ZW9 (AT1G58270) were uniquely reduced in the pulldown from extracts of Wt plants grown under Fe-deficient conditions (Table S6). KEGG pathway enrichment analysis was also conducted with the changed binding proteins from the aforementioned comparisons (Fig. 6).
The enriched KEGG pathways with the highest representation of the changed proteins in both Wt and klun were mostly associated with amino acids metabolism and carbohydrate and energy metabolism, such as "Cysteine and methionine metabolism" (ko00270), "Oxidative phosphorylation" (ko00190), and "Glycolysis / Gluconeogenesis" (ko00010) ( Table S7 & S8). Comparison of the klun interactome with that of Wt plants grown under Fe-sufficient condition, showed significant quantitative differences in the pull-down from the two genotypes (Fig. 7). The klun interactome lacks 7 proteins of the Wt interactome and vice versa the Wt interactome lacks 4 proteins of the klun interactome. Moreover, 17 proteins are 3-to fourfold enriched in the klun interactome as compared to Wt, whereas only 3 proteins are enriched in the Wt interactome as compared to klun (Table S6). So, the absence of the four 14-3-3 isoforms in the klun mutant has a clear impact on the 14-3-3 interactome and it is noteworthy that identified proteins are enriched in the klun interactome, what cannot be the result of a loss of in vivo phosphorylation in the absence of the four 14-3-3 proteins.
A. The Venn diagram of changed interactors identified in the comparison between Fe-sufficient and Fedeficient condition in Wt and klun. B. Heatmap of changed interactors identified in the comparison between Fe-sufficient and Fe-deficient condition in Wt and klun pull-down experiment.

Discussion
14-3-3 Proteins regulate the activities of a wide array of targets via direct protein-protein interactions and play a crucial role in many metabolic pathways 36 . Several studies have implied that 14-3-3 proteins have crucial roles in Fe deficiency However, analysis of the function of 14-3-3 proteins through mutant phenotyping is a challenging task due to gene specificity and functional redundancy as previously reported 18,20,24 . To avoid missing the phenotype of mutants due to redundancy we studied three high order mutants (klpc, klun and unpc) from the non-epsilon group which were generated by crossing two double mutants: kl*pc, kl*un and un*pc. In this study, we found the distinct physiological responses to Fe deficiency of Wt and the three 14-3-3 quadruple mutant lines and conducted a qAP-MS to analyze the 14-3-3 interactomes of roots in response in both Wt and klun.
Fe deficiency is a major nutritional disorder that causes decreases in vegetative growth and marked yield and quality losses 37 . In this study, the 14-3-3 qKOs performed distinctly in response to long-term Fe deficiency at physiological levels. Under Fe-deficient condition, unpc showed the strongest growth reduction of shoot and root, and growth of klpc showed a significant reduction of both shoot and root under low Fe condition (Fig. 2). The uptake of Fe by klun is clearly better than that of Wt (Fig. 3A,B). Nevertheless, Fe deficiency induced a much higher IUE (Iron Use Efficiency) of Wt shoot and root than that of the 14-3-3 mutants (Fig. 3C,D), what points to a more efficient use of available Fe in the Wt plants and thus to a role for 14-3-3 proteins in Fe-translocation and/or use after uptake from growth medium.
Studies on the micronutrient requirements of the rhizosphere suggest that root morphology (e.g. root length, root hair density) are critical for plants to acquire available Fe 38 . Iron-deficient plants must allocate their limited energy resources between finding available iron and physiological maintenance. Low Fe bioavailability induce plants to address their efforts to acquire the nutrient by increasing the root surface. Our results indicated that the length of both main and lateral roots in klun was significantly longer than those of Wt plants under Fe-deficient condition (Fig. 4). The increased root length of klun were likely correlated to the higher Fe uptake. In addition, we found that the main root length of klun and unpc was shorter under Fe-sufficient condition as compared to Wt. According to our data, the mutant plants that combine kl with un (klun) showed a better Fe uptake and longer root length as compared with Wt at Fe deficiency, while these phenotypes were absent in mutant plants that combine un with pc (unpc). This indicates that the combination of kappa and lambda present isoform specificity amongst the 14-3-3 genes tested. Whether there is redundancy between kappa and lambda requires further testing by analyzing the single and double mutants.
Along with the diverse physiological responses to Fe deficiency observed in 14-3-3 qKOs, changes in gene expression related to physiological responses were also observed in the present study. H + -ATPases (AHA2), FIT, FRO2 and IRT1 together mediate Fe deficiency induced medium acidification and Fe uptake in Strategy I plants [39][40][41] . In this study, the expression of 14-3-3 OMICRON, FIT, FRO2 and IRT1 were all induced by Fe deficiency, whereat the induction in klun mutant was strongest among all the tested genotypes (Fig. 5). Under Fe deficiency, high-level expression of the genes encoding the abovementioned proteins is controlled by the bHLH transcription factor FIT, where the induction of FIT expression is dependent on the 14-3-3 Omicron protein 9. In a detailed microarray analysis, GRF11 was significantly induced in the root elongation and maturation zone after subjecting plants to Fe deficiency for 24 h 42 . Loss of function of GRF11(14-3-3 Omicron) resulted in failure of acidification and ferric chelate reductase (FCR) induction, and thus decreased Fe uptake. However, another two 14-3-3 isoforms, GRF9 (14-3-3 Mu, Epsilon group member) and GRF1 (14-3-3 Chi, non-Epsilon group member) did not show a noticeable difference in expression in a Northern analyses in Arabidopsis upon iron deficiency 18 . Although we observed that the 14-3-3 qKOs affected the expression of Fe deficiency responding genes, it remains a question whether this regulation is affected directly by the mutations or indirectly by the 14-3-3 OMICRON taking over other's function (redundancy). Moreover, combining un with pc (unpc) and kl with pc (klpc) showed no different effect on expression level of 14-3-3 omicron, FIT, and AHA2 as compared with WT under Fe deficiency, raising the possibility that KL but not UN isoforms are involved in the iron deficient phenotype (Fig. 5). www.nature.com/scientificreports/ Our results from the pull-down experiments using recombinant 14-3-3 s and protein extract from Fe-deficient or Fe-sufficient roots showed a total of 117 proteins identified as putative 14-3-3 clients. It must be noted that proteins found in the 14-3-3 s affinity purification include not only primary or direct 14-3-3 targets but also secondary or indirect targets as members of multi protein complexes containing 14-3-3 s. Of these, 24 proteins have been identified in other 14-3-3 s interactome studies or characterized as 14-3-3 binding targets in vivo or in vitro, such as EIF4A (AT3G13920) 43 , CINV1 (AT1G35580) 44 and CDPK6 45 . In addition, 16 proteins were reported as Fe-responsive proteins including a fructose-bisphosphate aldolase (AT3G52930), two cytosolic invertases (AT1G35580, AT4G34860), a 5-methyltetrahydropteroyltriglutamate-homocysteine methyltransferase (AT5G17920), and two S-adenosylmethionine synthase (AT1G02500, AT2G36880) 11,35,46 (Table S1). According to KEGG classification, the most significantly enriched terms were associated with "amino acid metabolism", "carbohydrate and energy metabolism", "protein folding, sorting and degradation", and "transport and catabolism" (Fig. S2), which are consistent with previous studies on 14-3-3 interactomics 21,32,47 .
Since most proteins were found with high abundance in all pull-down experiments, it is difficult to directly assess the dynamic response of 14-3-3 interactions to Fe deficiency. Thus, quantitative analysis of the identified 14-3-3 clients is essential to reveal responses to Fe deficiency. By comparing the normalized iBAQ, we found that 27 and 17 proteins were significantly changed in the 14-3-3 interactome of Wt and klun upon Fe deficiency treatment, respectively (Tables S4 & S5). Moreover, a comparison of the Wt and klun interactome of Fe sufficient roots showed a quantitative or absolute difference in the interaction of 32 proteins (Table S6). So, the 14-3-3 interactome was clearly affected by Fe deficiency (Wt, ± Fe) as well as by the absence of four 14-3-3 isoforms (Wt/klun, + Fe). The observed changes are mainly in proteins with a function in carbohydrate and energy metabolism and cysteine/methionine synthesis and we will therefore limit our discussion to these pathways (Fig. S2). In this study, changes of protein abundance in carbohydrate and energy metabolism enzymes upon Fe deficiency are consistently identified among different proteomics studies 11,46 . For example, G6PD3 (AT1G24280) (down-regulated, 0.31-fold) and fructose-bisphosphate aldolase (AT3G52930) (up-regulated, 4.03-fold) was significantly changed in the pull-down with Wt root upon Fe deficiency, whose abundance have shown the similar trends (0.60-fold and 1.41-fold, respectively) in iTRAQ protein profile analysis of Fe-deficient Arabidopsis roots 11 (Tables S5 & S6). An increase in NADH + ATP production by activation of the glycolytic pathway and mitochondrial respiration can provide the reducing equivalents to keep the Fe (III) reductase working and fuel the plasma membrane ATPase: processes that are essential for the Fe-uptake mechanism in iron-deficient roots 48 . In this study, we found that three subunits of the mitochondrial ATP synthase head structure (α, β and δ), NADH dehydrogenase and the mitochondrial-processing peptidase MPPBETA. FoF1-synthase has been reported as 14-3-3 target 49,50 and the identification of only subunits of the head-structure is in line with the evidence that the β-subunit is the direct target for 14-3-3. Functionally, the mitochondrial synthase activity is reduced by interaction with 14-3-3 49 and in that respect it is surprising that in Wt, Fe deficiency enhances the interaction with the FoF1-subunits and the mitochondrial processing peptidase 2-to tenfold (Table S5). This increased interaction suggests a down-regulation of the ATP-synthase, what is contrary to the expected increase in respiration. Interaction with MPPBETA and CI51 is increased 3-to tenfold respectively and this warrant further study of the functional consequences of 14-3-3/MPPBETA and CI51 interaction. It should be noted that the related head structure of the vacuolar V-ATPase (A, B and E1 subunits) was isolated as well, but Fe deficiency had no effect on the interaction with the V-ATPase.
Methionine synthesis is important in the Fe deficiency response as it provides the precursors for ethylene and nicotianamine (NA), an important chelator with a crucial function role in Fe homeostasis and transport 51,52 . Fe deficiency induced multiple S-adenosylmethionine synthases (gene expression and protein amount) involved in S-adenosyl-Met 53 biosynthesis, an important precursor for ethylene production 11 . 14-3-3 s are linked to the ethylene biosynthesis by interacting with S-adenosylmethionine (SAM) synthase 53 , ACC synthase and 1-aminocyclopropane-1-carboxylate synthase (ACS) 19,32,54,55 . Here, we identified four S-adenosylmethionine synthases (MAT1, MAT2, MAT3, and MAT4) and the 5-methyltetrahydropteroyltriglutamate-homocysteine methyltransferase (ATMS1) in the 14-3-3 s interactome, underlining the importance of 14-3-3 s in this pathway (Table S1 and S2). In addition, SAM synthases MAT1, MAT2 and MAT3 were found to be significantly increased at the protein level in the presence of iron deficiency in Arabidopsis roots 11,56 , while our data suggested that Fedeficiency stimulated the interaction of MAT1/14-3-3 s, but reduced the interaction with MAT4 in Wt and klun (Tables S5 & S6). It may imply that the interaction between 14-3-3 and adenosylmethionine synthetases involved in multiple molecular mechanisms underlying the Fe deficient stress responses, and it is of prime importance to address the question how 14-3-3 s affect the SAM-synthase activities. Amongst the proteins that were lost in the 14-3-3 pull-down with protein extract form Wt plants, ECIP1 (EIN2 C-terminus interacting protein 1) may be relevant in view of the role of ethylene in Fe deficiency adaptation. ECIP1 is an MA3 domain-containing protein that interacts with EIN2, a central membrane protein that acts downstream of ethylene receptors, and upstream of ethylene regulated transcription factors. ECIP1 directly interacts with EIN2 and loss-of-function of ECIP1 resulted in enhanced ethylene response 57 . If ECIP1/14-3-3 interaction prevents ECIP1 breakdown as is e.g. the case for ABF3 , then loss of interaction between ECIP1 and 14-3-3 will result in enhanced ethylene signaling. So, this warrants further investigation.
Fe deficiency may cause changes in post-translational modifications as well. Several protein kinases (e.g. MPK3/MPK6) accumulate differentially upon Fe-deficient plants, what suggests that alterations in protein phosphorylation induced by Fe-deficiency are involved in Fe homeostasis 58 . In this study, we found that the interaction of 14-3-3/CDPK6 (AT4G23650) only significantly decreased accumulation in Wt upon Fe deficiency, suggesting that phosphorylation level of 14-3-3 binding targets is different under Fe-deficient conditions, as reported for CINV1, AHA1 and MAPKKK 13 . As the changes in 14-3-3 interaction may be due to changes in protein abundance and protein phosphorylation, it is difficult to predict how the absence of four 14-3-3 proteins, as is the case in the klun mutant, will affect the interactome. It may reduce in vitro interaction because in vivo binding www.nature.com/scientificreports/ of 14-3-3 proteins to a phosphorylated motif in target proteins enhances the in vivo level of phosphorylation as the site is then protected from dephosphorylation by phosphatases. On the other hand, the absence of 14-3-3 s may result in more target protein in the pull-down because 14-3-3 can act as repressors of transcription factors or kinases. An example of the latter is the suppression of SOS2 activity by 14-3-3 binding 59 . Thus, quantification of abundance as well as phosphorylation level of the 117 14-3-3 putative binding proteins in the respective input fractions needs to be done in the future. Altogether, these findings provide novel insights into the role of 14-3-3 non-epsilon group in response of Fe deficiency. Our data highlight that the combination of kappa and lambda presents isoform specificity in iron deficient phenotype. In addition, the absence of the four 14-3-3 isoforms in the klun mutant has a clear impact on the 14-3-3 interactome upon Fe deficiency. This work has greatly reduced the scope of research objects for follow-up work towards the role of tested 14-3-3 isoforms in Fe acquisition. Analysis of single, double mutants of k and l will reveal whether redundancy exists for the observed phenotypes. In addition, we found that the interaction of 14-3-3/ SAM synthases plays an essential role in plants responding to Fe deficiency. How 14-3-3 s affect the SAM-synthase activities and thereby regulate the ethylene signaling pathway needs to be addressed.

Materials and methods
Plant growth conditions. All plants used are in the Arabidopsis thaliana Columbia ecotype (Col-0) background. The 14-3-3 quadruple KO mutants were generated by crossing the double mutants as described in the previous study reported by van Kleeff et al. 20 . The mutant seeds collection and plants growth were carried out with the permission from the Vrije University Amsterdam. Experimental research using plants comply with institutional, national, or international guidelines. Plants were grown in ½ strength Hoagland solution (pH 5.8) with 20 μM Fe(III)-EDTA in a growth chamber at 14/10 h day/night regime, 22/18 °C day/night temperature and a photon flux density of 170 μmol•m-2•s-1. After 22 days of germination, plant roots were washed with once with 10 mM EDTA for 10 min followed by two times wash in Milli Q, before transfer to either iron-sufficient (20 μM Fe(III)-EDTA) or iron-deficient (iron omitted) culture medium for 24 h. Then the roots were harvested and immediately cleaned with Milli Q. After that, the roots were dried on tissue paper and snap frozen in liquid nitrogen. Total roots were ground in liquid nitrogen and weighed, and stored at -80 °C. The plant material was used for 14-3-3 pull-down experiments. For growth and Fe content measurement, plants of Wt and 14-3-3 qKO plants were grown in ½ strength Hoagland solution (pH 5.8) with 20 μM Fe(III)-EDTA for 14 days, then young seedlings were transplanted to a new medium supplemented with 0, 2, 5, and 20 μM Fe(III)-EDTA. Leaves and roots were harvested 12 days after Fe deficient treatment. To get rid of surface constituents, the roots were washed in 10 mM EDTA for 10 min and then rinsed twice in Milli-Q before harvest. The collected shoots and roots from two plants were pooled for each biological replicate and harvested separately and dried at 150 °C for 2 days.

Analysis of the Fe content in Arabidopsis shoots and roots. Fe content in Arabidopsis leaves and
roots was determined by the BPDS (bathophenanthrolinedisulfonicacid) method as described previously by Schmidt 60 . In brief, 4 to 8 mg dried sample was well mixed in 2 mL Eppendorf tubes and heated at 95 °C in 75 µL nitric acid (65%) for 6 h. After the samples were completely digested, 50 µL of H 2 O 2 (30%) was added, and the solution was incubated at 56 °C for 2 h. The volume was adjusted to 200 µL with sterile water. 20 µL of this solution was diluted in 980 µL of BPDS buffer (1 mM BPDS, 0.6 M sodium acetate, and 0.48 M hydroxylammonium chloride). The concentration of Fe-BPDS was measured at 535 nm. A standard curve was prepared by dilution of a stock FeSO 4 solution dissolved in 0.1 M HCl. The iron utilization efficiency (IUE) was calculated based on Fe and dry weight accumulation, as described by Fageria and Baligar 61 .
Root phenotyping. Seed sterilization and germination was according to our previous study 20 . For the root phenotyping, seeds were germinated on 120 mm × 120 mm petri dishes containing 0.5 × MS medium (pH 5.8) solidified with 12 g/L of plant agar (Sigma A1296) after three days of stratification. After 4 days, three Wt and three mutant plants were transferred to new plates containing Fe-deficient or Fe-sufficient medium and grown vertically. Plates were scanned using a flatbed scanner after 7 days and the images were analyzed using EZ-Rhizo 62 . The parameters chosen in this study are: 1) main root length, 2) Total length of the lateral roots.
Real-time PCR assay. Total RNA was extracted from root tissues using the NucleoSpin® RNA Plant Kit (MACHEREY-NAGEL), and first-strand cDNA was synthesized from 2 μg of total RNA using the Superscript II Kit (Invitrogen, USA) with oligo d(T)18 primers according to the manufacturer's instructions. Quantitative RT-PCR reaction contained 100 ng cDNA, 1 pmol of each primer, 2 × Sybr Green PCR buffer (Bio-Rad, Hercules). The PCR conditions for the amplification of 14-3-3 omicron, FIT, FRO2, IRT1, AHA2 and ubiquitin 10 were as follows: 1 min at 94 °C, followed by 30 cycles of 45 s at 94 °C, 60 s at 54 °C and 75 s at 72 °C. The PCR products were examined according the 2-∆∆CT method. Each sample was assayed three times. The relative expression was calculated against that of the internal control gene ubiquitin 10 (UBQ10). Primer pairs used for each gene are listed in Table S9.
Affinity purification of 14-3-3 target proteins. Recombinant His-tagged proteins were purified as described in 44 . Protein concentrations were determined by Bradford micro-assay (Bio-Rad) using BSA as a standard. Arabidopsis roots from WT and klun plants were ground in liquid nitrogen and extracted with extraction buffer (50 mM HEPES-NaOH (pH 7), 10 mM MgCl 2 , 1 mM Na 2 EDTA, 2 mM DTT, 10% ethylene glycol, 0.02% Triton, 1 × protease inhibitor cocktail and 1 × phosSTOP). Protein extracts were centrifuged twice at 20,000 g for 15 min. To avoid the isoform specifically binding of the 14-3-3 target proteins, equal amount of www.nature.com/scientificreports/ four 14-3-3 isoforms (His-KAPPA, His-LAMBDA, His-NU, His-UPSILON) were well mixed and then coated to PureProteome Nickel magnetic Beads (Millipore) according to manufactures' protocol. Protein extracts were collected from roots of klun and WT treated with 0 and 20 μM Fe (III)-EDTA for 24 h. After coating of 40 µg of His-14-3-3 to 100 µl nickel beads and extensive washing with binding buffer (50 mM sodium phosphate, 300 mM sodium chloride, 10 mM imidazole, pH 8), 2 mg protein extract was added to the 14-3-3 beads and incubated overnight at 4 ºC. To distinguish background proteins, we performed mock pull-down with empty beads, where the extracts were incubated with beads that were not coated with His-14-3-3. Beads were extensively washed in 1 ml wash buffer containing 10 mM imidazole for 5 min and this was repeated 5 times. The bound protein complexes were eluted off from beads with 100 μl of wash buffer containing 100 mM imidazole for 20 min. The experiment was conducted three times with independent biological replicates, meaning that in the end each genotype had 6 profiles. CBB-stained gel was used as the loading control (Fig. S1). The affinityenriched proteins were separated on SDS-PAGE, and then characterized by LC-MS-MS semi-quantitatively.
In-gel digestion and LC-MS/MS analysis. LC . In short, 20 μL peptides were loaded on a 5 mm Pepmap 100 C18 (Dionex) column (300 μm ID, 5 μm particle size) and separated on a 200 mm Alltima C18 homemade column (100 μm ID, 3 μm particle size) with an Eksigent HPLC system, using a linear gradient of increasing acetonitrile concentration from 5 to 35% in 45 min, and to 90% in 5 min. The flow rate was 400 nL/min. The eluted peptides were electro-sprayed into the LTQ-Orbitrap discovery.

Data analysis.
To increase the data quality, we removed the low confident interacting proteins from the protein list. First, contaminant proteins like keratin and trypsin were removed from this results file. Second, considering the reproducibility of the pull-down experiment, we excluded proteins that were identified only once in all replicates. False positives were removed from the bait-prey matrix by comparing the abundance of proteins identified in each pull-down against their abundance in the matching empty bead controls, in which the true positive bait-prey interactions should be > tenfold enriched in the APs. Also, the four bait proteins were excluded from the list. Intensity-based absolute quantification (iBAQ) values were calculated in the Max-Quant suite as previously described 63,64 . The protein abundance was calculated on the basis of the normalized iBAQ intensity. For short, quantifiable proteins in the analysis defined as those identified in at least two of the three biological replicates in at least one type of sample. Missing values were imputed using row mean imputation. The relative protein intensities were calculated as the ratio of their intensity to the bait proteins in that run. The relative protein intensities for each pull-down experiment were combined in a matrix, and false positives were removed the proteins identified in the background. The minimum two requirements for the differentially expressed proteins are: (1) identification of a protein with unused value > 2; (2) the fold change of protein quantities protein quantities in Fe-deficient treated samples against Fe-sufficient samples with more or less than 1.5 times with significant difference Student's t test P value < 0.05). Functional analysis of identified proteins was obtained by performing Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses using the KEGG Orthology-Based Annotation System 65,66 .