Metal-sensing properties of the disordered loop from the Arabidopsis metal transceptor IRT1 Open Access

The first-row transition metal iron is the most abundant and used metal in cells [1]. Iron is widely used as a cofactor, participating in oxidation–reduction reactions. In plants, iron is of utmost importance due to its involvement in photosynthesis electron transfer reactions [2]. In nature, iron is often unavailable to organisms including plants because it is found as insoluble ferric oxide/hydroxide. Lack of iron deeply affects plant growth and leads to severe leaf chlorosis [2]. On the other hand, an excess of iron is detrimental to cells due to its redox properties and participation in the Fenton reaction that leads to reactive oxygen species formation [3]. For these reasons, living organisms and plants, in particular, need to tightly control metal uptake.

Dicotyledonous plants, such as the model plant Arabidopsis thaliana, use a three-step iron uptake process based on the acidification-based solubilization of soil ferric iron (Fe³⁺) insoluble oxides and the reduction in ferrous iron (Fe²⁺) prior to transport. This is mediated by the AHA2 proton pump, which secretes proton, and the FRO2 ferric chelate reductase, which reduces Fe³⁺ to Fe²⁺ [4-7]. In Arabidopsis, the main transporter involved in the uptake of Fe²⁺ is the iron-regulated transporter 1 (IRT1) [8], one of the founding members of the family of ZIP transporters that are found across all kingdoms of life [9]. Arabidopsis IRT1 has a broad substrate range, as it transports other essential divalent heavy metals such as zinc (Zn²⁺), manganese (Mn²⁺), cobalt (Co²⁺), and toxic cadmium (Cd²⁺) ions [8,10-16]. In planta, IRT1 is the major root iron transporter responsible for iron uptake from the soil under iron-limited conditions. IRT1 protein indeed localizes to the plasma membrane, and IRT1 gene is strongly expressed in iron-deficient root epidermal cells [8,15]. Consistently, irt1-1 knockout mutant plants are highly chlorotic and show growth impairment, a phenotype that can be reversed by the addition of high amounts of iron in watering solutions [8]. The chlorosis of irt1 mutants is, however, not complemented by the other divalent metal substrates of IRT1, indicating that these constitute secondary substrates that are nonspecifically transported.

In recent years, the molecular mechanisms underlying the regulation of IRT1 expression by its metal substrates have been uncovered. The transcription of IRT1 is up-regulated upon iron starvation through a cascade of bHLH transcription factors [17-23]. Most importantly, post-translational regulation of IRT1 by its non-iron metal substrates Zn²⁺, Mn²⁺, Co²⁺, and Cd²⁺ has recently been reported [16,24]. Non-iron metal availability indeed regulates the subcellular localization and stability of IRT1. In the absence of non-iron metals, IRT1 sits at the cell surface to take up low available iron from the soil. Increasing non-iron metal levels lead to partial IRT1 internalization in early endosomes, as a result of its multi-monoubiquitination on lysine residues K154 and K179 by a yet-to-be-characterized E3 ubiquitin ligase [15]. Higher non-iron metal levels trigger IRT1 phosphorylation by the CIPK23 kinase and the recruitment of the E3 ubiquitin ligase IDF1 [16,24]. IDF1 elongates monoubiquitin moieties into K63 polyubiquitin chains, thus leading to IRT1 targeting the vacuole for degradation [16,24]. Fe²⁺ availability has, however, no influence on IRT1 protein localization since IRT1 is not specifically degraded upon iron resupply under physiological metal conditions [16,24].

The cytosolic regulatory loop of IRT1 located between transmembrane domains TM4 and TM5 (amino acid 144–185) has been shown to bind metals in vitro [24]. A remarkable feature of IRT1 and related transporters is the presence of a histidine-rich motif at the center of such loop. Such repetition of histidine has been shown to be required for the non-iron metal-dependent endocytosis of IRT1, as an IRT1_4HA mutant with the four histidines mutated to alanine fails to be degraded upon non-iron metal excess [24]. Metal binding to histidine residues was shown to drive the recruitment of CIPK23 [24]. IRT1 is, therefore, considered a bifunctional transporter–receptor capable of sensing elevated non-iron metal levels and initiating a signaling cascade culminating in its self-degradation to limit the entry of highly reactive non-iron metals. The intricate mechanism allowing metal binding to IRT1 histidine-rich motif to trigger CIPK23 recruitment remains, however, still unclear but was proposed to involve the folding of IRT1 [25].

To obtain better mechanistic insight into how IRT1 senses non-iron metals and how this mediates the recruitment of CIPK23, we biochemically characterized the metal-binding properties and the structure of the IRT1 loop by a combination of circular dichroism (CD), NMR spectroscopy, and microscale thermophoresis (MST). We report that the regulatory loop of IRT1 binds Zn²⁺ and Mn²⁺in vitro using the four histidine residues with nanomolar affinities, and we also uncover the role of residue D173 in Zn²⁺ and Mn²⁺ co-ordination and sensing during IRT1 degradation in response to non-iron metal excess. Furthermore, we show that the regulatory loop of IRT1 is disordered in both the absence and presence of Zn²⁺, despite experiencing small structural changes in the presence of Zn²⁺. Overall, our work provides additional biochemical and structural insight into the ZIP family of metal transporters.

Considering that, to date, eukaryotic ZIP transporters cannot be expressed and purified from heterologous systems, we focused on the regulatory loop of IRT1 that carries important residues for metal sensing and response [24], as previously done for hZIP4 [26,27]. To get a first glimpse of the structural features of the regulatory loop of IRT1 (a.a. 144–185), we performed far-UV CD on the corresponding peptide chemically synthesized. The ¹⁴³DSMATSLYTSKNAVGIMPHGHGHGHGPANDVTLPIKEDDSSN¹⁸⁶ peptide sequence encompasses all important residues required for phosphorylation and ubiquitination, and contains the four histidine residues that have been previously associated with metal sensing [16,24]. Structure prediction using the PONDR (http://www.pondr.com) software suggests a high degree of disorder in this region (Supplementary Figure S1), with 45% of the residues being disorder promoting (A, R, G, Q, S, P, E, and K), and only 21% being rather order promoting (W, C, F, I, Y, V, L, and N). In accordance with software predictions, CD spectra recorded from 190 to 240 nm at pH 6.7 at room temperature showed a spectrum typical of a peptide in a random coil conformation, with a single negative peak at 198 nm [28] (Figure 1A, gray line, 0 eq).

To obtain deeper insight into the structure of the IRT1 loop, we turned to NMR spectroscopy. One-dimensional (1D) ¹H NMR spectra were recorded at pH 6.7 in the absence of metals. In line with the results obtained by CD, 1D ¹H NMR spectra were poorly dispersed, a characteristic of peptides in random coil conformations (Figure 1B, black line). Furthermore, two-dimensional (2D) H¹-H¹ NOESY spectra of the IRT1 loop showed that despite our previous observations of the peptide adopting a rather unstructured conformation, medium-range nuclear Overhauser effect (NOE) existed, notably pointing to the presence of a turn involving the ¹⁵⁵AVGI¹⁶⁰ hydrophobic residues (α156/HN159, α157/HN159) (Supplementary Figure S2A). Besides, we observed that the cross-peak intensities of the histidine-rich domain from IRT1 decreased relative to the other signals of the peptide, resulting probably from chemical exchange (Figure 2A). Despite the lack of a clear secondary structure, we generated structures using the NOE-derived distance restraints with the CYANA software (Supplementary Figure S2). As expected, except for the ¹⁵⁵AVGI¹⁶⁰ turn (Supplementary Figure S2), no secondary structure was observed.

In order to determine whether the structural properties of the IRT1 loop are affected by interaction with the secondary metal substrates of IRT1, we performed a titration by sequential equimolar additions of Zn²⁺ on the peptide followed by recording of CD spectra (Figure 1A). With each addition of Zn²⁺, the final concentration of the peptide was corrected to avoid dilution effects. We observed a saturable increase in ellipticity with each addition of the substrate, but the overall shape of the spectra remained unchanged. Overall, the binding of Zn²⁺ did not seem to affect the structure of the IRT1 loop, as visualized by CD. Similarly, 1D H¹ NMR spectra in the presence of Zn²⁺ showed a similar dispersion of the signals compared with the spectra recorded in the absence of metal, again suggesting that even in the presence of Zn²⁺, the IRT1 loop remains disordered (Figure 1B, blue line). Interestingly, the signals are broadened upon sequential addition of Zn²⁺, indicating that the IRT1 loop undergoes small structural changes probably induced by the binding of Zn²⁺ ions (Figure 1A, blue-colored lines).

To obtain a deeper understanding of the molecular interaction of the IRT1 loop with its substrates, 2D H¹-H¹ NOESY spectra of the IRT1 peptide were compared at pH 6.7 in the absence of metal substrates and in the presence of two equivalents of Zn²⁺. Superimposition of these spectra indicated that the histidine-rich motif of IRT1 is implicated in the binding of Zn²⁺ since all the proton resonances of these histidines (H162, H164, H166, and H168) disappeared upon the addition of Zn²⁺ (Figure 2A). This suggests that these residues are involved in the exchange between several conformations in an intermediate regime. Further analysis of our data also indicated that signals corresponding to D173, which is located C-terminal of the histidine-rich stretch, behaved similarly to the histidine residues (Figure 2A). NMR signals corresponding to D173 indeed disappeared after the addition of Zn²⁺, suggesting a possible implication of D173 in metal co-ordination.

To better characterize the role of individual histidine residues and aspartic acid D173 present in the wildtype IRT1 loop, we also decided to record H¹-H¹ NOESY spectra of different mutant peptides. We designed three mutants for the PHGHGHGHGPAND motif where histidine residues were replaced with phenylalanine (F), threonine (T), or asparagine (N) to remove the charges at these positions and maintain closest residue similarity; and a mutant where aspartic acid D173 was replaced by the uncharged asparagine amino acid. We observed that in the presence of Zn²⁺, the line width for most of the peptides increased drastically (Figure 2A,C,D,F and Supplementary Figure S3) except for the mutant peptide with four mutated histidine residues (Figure 2B and E). Broadening of the signals indicates that the metal-bound IRT1 loop undergoes dynamic interchange among different metal-bound states. Interestingly, the mutant peptides with only the first two (H162 and H164) or last two histidine residues (H166 and H168) from the stretch mutated to uncharged residues were still able to bind Zn²⁺ (Supplementary Figure S3). Furthermore, the signal for the D173N mutant peptide no longer responded to the addition of Zn²⁺ (Figure 2C and F), suggesting that D173 is indeed involved in metal co-ordination.

Finally, we analyzed the chemical shift variations in the absence and presence of two equivalents Zn²⁺ of each residue. Here, we found that the residues located at the N-terminal side of the histidine-rich fragment were very little affected by Zn²⁺ binding, whereas those located on the C-terminus underwent significant chemical shift variations (Supplementary Figure S4A). The same behavior was observed for peptides with mutations of two histidine residues and for the aspartic acid mutant (Supplementary Figure S4B-D), though to a lesser extent than for the wildtype peptide. As expected, the peptide with the four histidine residues mutated presented no chemical shift variation (Supplementary Figure S4E).

These results are in line with the amino acid composition of the IRT1 loop, which is rich in negatively charged amino acids (E and D) on its C-terminal side with an isoelectric point of 3.84 (a.a. 168–185), instead of 5.83 for its N-terminal side (a.a. 144–167).

In order to quantify more precisely the interaction between the regulatory loop of IRT1 and non-iron metals, we performed MST experiments. We first used Zn²⁺ as substrate, as we previously showed that it binds to the IRT1 peptide by ICP-MS [24], MST [16], and CD (this study), and Zn²⁺ excess was reported to drive IRT1 endocytosis [24]. A constant concentration of the peptide was titrated with a one in half-serial dilution of the ligand, ZnCl₂, and thermophoresis was measured. A binding affinity of 33 ± 2.1 nM, when fitted to the Kd model, was obtained for the wildtype IRT1 loop assuming a 1:1 stoichiometry of the reaction (Figure 3A). Data points from the peptide mutated for the four histidine residues (4 HA) were recorded and failed to fit a binding curve, confirming the absolute requirement of the histidine stretch for metal binding (Figure 3A) [16]. To evaluate deeper the contribution of the residues from the histidine stretch, we decided to test double mutant peptides harboring H162A/H164A or H166A/H168A mutations. Interestingly, both double histidine substitutions resulted in reduced affinities for Zn²⁺ compared with wildtype, with a Kd of 240 ± 73 nM for the H162A/H164A mutant and a Kd of 234 ± 86 nM for the H166A/H168A mutant (Figure 3A).

To inquire about the role of the D173 residue in metal-binding affinity, the IRT1 loop carrying the D173N mutation was subjected to MST analyses. Binding curves for Zn²⁺ were determined in the same concentration range as the previous experiments and fitted to the Kd model. The binding curve obtained for the D173N mutant was slightly shifted to the right compared with the one obtained with the wildtype peptide (Figure 3B), with a Kd of 64 ± 2.3 nM (Figure 3B). Because the contribution of D173 is likely minor compared with histidine residues, we sought to investigate the contribution of D173 in a peptide carrying a double histidine mutation where Zn²⁺ co-ordination is already destabilized, as shown above. The peptide carrying the three mutations showed significantly lower affinity to Zn²⁺ compared with the peptide with only two histidine residues mutated (Figure 3B). Such decrease in the affinity due to the removal of the negative charge from D173, in the context of the H162A/H164A mutation, argues for a contribution of D173 to Zn²⁺ co-ordination during sensing by the IRT1 transceptor.

Considering that IRT1 also transports Mn²⁺ and that this non-iron metal was also shown to regulate IRT1 degradation [8,10,11,14,15,24], we monitored Mn²⁺ binding to the IRT1 loop by MST. We observed that the IRT1 loop is also able to bind Mn²⁺ at nanomolar range, with lower affinity than for Zn²⁺, showing a Kd of 114 ± 62 nM (Figure 4A). Similar to what has been observed with Zn²⁺, the four histidine IRT1 mutant loop was no longer able to bind Mn²⁺ and the double histidine mutants presented lower affinity compared with wildtype IRT1. Interestingly, the absence of two histidine residues showed less impact on the affinity for Mn²⁺ (Figure 4A) compared with Zn²⁺ (Figure 3A), pointing to the possibility of others residues being actively involved in Mn²⁺ co-ordination. To test whether D173 could also participate to Mn²⁺ binding, we studied the affinity of the mutants harboring D173N mutation. Single D173N mutation barely affected affinity for Mn²⁺ compared with wildtype (Figure 4B). Nevertheless, D173N combined with H162A/H164A mutations resulted in a strong decrease in Mn²⁺ affinity (Figure 4B), also revealing a strong contribution of these three residues to Mn²⁺ co-ordination.

To characterize the role of residue D173 in metal transport and sensing, we first decided to express full-length IRT1_D173N in the iron uptake-defective fet3fet4 yeast mutant to evaluate whether the corresponding protein is still active for iron transport. Yeast expressing wildtype IRT1 were able to complement the growth defect of fet3fet4 in low-iron conditions, as previously reported [10]. Expression of IRT1_D173N protein yielded comparable growth compared with wildtype IRT1, indicating that IRT1_D173N is fully functional for iron transport (online supplementary figure 7A).

An excess of non-iron metal substrates of IRT1 (Zn²⁺, Mn²⁺) leads to IRT1 depletion from plasma membrane and increases IRT1 accumulation in endosomes for later degradation in the lytic vacuole [24]. This safety mechanism to limit heavy metal toxicity in plants due to metal overaccumulation through IRT1 involves metal sensing by the histidine-rich motif located in the regulatory loop of IRT1 [16,24]. Considering that the combination of D173N and H162A/H164A mutations impairs the IRT1 loop affinity for Zn²⁺ and Mn²⁺ to a greater extent than the H162A/H164A mutation (Figures 3B and 4B), we wondered if this aspartic acid also participates to metal sensing and metal excess-dependent degradation of IRT1 protein. To test this hypothesis, we investigated the functional impact of the sole D173 mutation on IRT1 degradation in response to non-iron metal excess in planta. For this purpose, we took advantage of the previously reported fluorescently tagged version of IRT1, IRT1-mCitrine, which has been demonstrated to complement the irt1-1 knockout Arabidopsis plants and fet3fet4 [24]. IRT1-mCitrine variants were transiently expressed in Nicotiana benthamiana leaves under the control of the 35S constitutive promoter to study IRT1 localization in response to high levels of non-iron metals. We, however, observed that the D173N substitution yielded retention of IRT1 into intracellular structures resembling the endoplasmic reticulum (ER) (online supplementary figure 8A). This is likely explained by the unexpected generation of an N-glycosylation site at the corresponding position (online supplementary figure 8B). We, therefore, substituted the D173 residue with glutamine (Q) to prevent spurious N-glycosylation and ER retention IRT1 mutated at D173 residue (online supplementary figure 8B). As expected, IRT1_D173Q localized at the plasma membrane similarly to wildtype IRT1 (online supplementary figure 8A). IRT1-mCitrine was observed mostly at the plasma membrane when expressed in N. benthamiana in the absence of non-iron metals (− metals; Figure 5A), similar to what has been reported in stable Arabidopsis transgenic plants [15,16,24,29]. A 3-h treatment with non-iron metal excess (+++ metals) led to IRT1-mCitrine depletion from plasma membrane and internalization in endosomes (Figure 5A and B), as already observed in roots [24]. IRT1 variants harboring the D173Q single mutation and the H162A/H164A double mutation exposed to a non-iron metal excess also displayed reduced plasma membrane localization (Figure 5A and B) and an increased number of endosomes (Figure 5C). Consistently, the interaction of the corresponding IRT1 variants with CIPK23 measured by trimolecular functional complementation (TriFC) is not (IRT1_D173Q) or marginally (IRT1_H162A/H164A) affected (Figure 5D and E). Interestingly, the triple mutant for H162A/H164A/D173Q showed significantly less protein removal from the plasma membrane (Figure 5A and B) and a significantly lower number of endosomes compared with double mutant H162A/H164A after treatment with high non-iron metals (Figure 5C). Such observation is not explained by a defect of IRT1_{H162A/H164A/D173Q} to transport metals as its expression in the fet3fet4 yeast mutant and in an Arabidopsis irt1_crispr mutant complemented growth to the same extent as IRT1 and IRT1_H162A/H164A (online supplementary figure 7B and C). The reduced metal-dependent endocytosis of IRT1_{H162A/H164A/D173Q} is accompanied by a lower ability to interact with CIPK23 compared with IRT1 carrying only the H162A/H164A mutations (Figure 5D and E). These observations clearly argue for an implication of D173 residue in Zn²⁺ and Mn²⁺ sensing and proper IRT1 degradation when plants are experiencing non-iron metal stress. Notably, IRT1_K154R/K179R double mutant, which is defective in endocytosis due to its inability to be ubiquitinated, failed to be endocytosed upon metal excess (online supplementary figure 9).

Altogether, these observations indicate that N. benthamiana leaves possess the cellular machinery responsible for the internalization and degradation of IRT1 under metal excess and highlight the combined role of residues H162, H164, H166, H168, and D173 in co-ordinating and sensing non-iron metals to fine-tune iron and metal uptake.

To cope with changing nutrient availability and nutrient demand, plants developed strategies that allow them to control the expression of nutrient transporters. This is the case for the Arabidopsis IRT1 iron transporter that is controlled by various metal substrates at different levels, that is transcriptional regulation by low Fe²⁺ and post-translational regulation by Zn²⁺ and Mn²⁺ excess [24,29,30]. Previously, we have shown that the post-translational regulation of IRT1 at the plasma membrane in response to non-iron metal excess involves its phosphorylation by the CIPK23 kinase followed by decoration with K63 polyubiquitin chains by the IDF1 E3 ligase, mechanism that requires the histidine-rich motif in the IRT1 regulatory loop [24].

To gain further insight into the molecular mechanisms of metal sensing by IRT1, we here investigate the structural basis of the IRT1 regulatory loop using a combination of NMR spectroscopy, CD spectroscopy, MST analyses, and protein prediction algorithms. We show that the IRT1 loop is an intrinsically disordered region (IDR), which can adopt various conformations. This characteristic appears to be conserved among ZIPs, as it is also observed for the human ZIP4 zinc transporter [26,27]. This is, however, not a specific feature of ZIPs since other families of transporters, like the Arabidopsis MTP1 vacuolar zinc transporter, also possess disordered intracellular loops with metal-binding motifs [31]. Our observations also indicate that the IRT1 loop undergoes small structural changes in the presence of Zn²⁺, although it remains largely disordered. This observation points to the formation of a ‘fuzzy’ complex between an IDR and a small ligand, as previously reported for hZIP4 [27]. Intrinsically disordered proteins (IDPs) or IDR interactions can exhibit varying mechanisms, ranging from complete folding upon binding to the formation of such 'fuzzy' complexes. The term 'fuzzy complex' refers to a class of interactions in which the involved IDPs or IDRs retain conformational flexibility upon binding. Unlike traditional stable protein complexes, these interactions are characterized by dynamic behavior, with the complexes not adopting a single, fixed conformation. The enhanced flexibility and conformational plasticity of IDRs in proteins provide a platform for post-translational events allowing enhanced interactions and chemical reactions between partners [32,33]. For IRT1, such capacity may facilitate the recruitment of downstream factors such as the CIPK23 kinase for the IRT1 loop phosphorylation upon non-iron metal excess [24]. Moreover, the importance of the histidine stretch of the IRT1 regulatory loop in its post-translational regulation has been previously established, where plants harboring a protein mutated for the four histidine residues fail to show IRT1 phosphorylation and degradation upon non-iron metal excess [24]. Interestingly, we observed chemical shift variations in the residues located at the C-terminus of the regulatory loop in the presence of Zn²⁺ compared with its absence, a region where the predicted phosphorylated T175 by CIPK23 is found [24]. Nevertheless, because the histidine resonances disappear upon the addition of Zn²⁺, we are not able to determine whether this particular stretch adopts a specific conformation in such conditions.

The nanomolar range dissociation constants recorded for Zn²⁺ and Mn²⁺ binding to IRT1 are consistent with experimentally determined nanomolar concentrations of such metals in plants grown in excess conditions [34]. Importantly, we unambiguously reveal that the histidine stretch is of absolute importance for direct Zn²⁺ and Mn²⁺ co-ordination, as MST experiments demonstrated intense perturbations upon mutation of all four histidines, completely abolishing metal binding. Surprisingly, we uncovered that the IRT1 loop is still able to bind Zn²⁺ when two histidine residues are mutated, probably due to the other two histidines and aspartic acid 173 being responsible for co-ordination. This is supported by the NMR data obtained with the double histidine mutants where proton resonance of the remaining two histidine residues and aspartic acid 173 still disappeared in the presence of Zn²⁺. Although a co-ordination number of three for Zn²⁺ is not common, Zn²⁺ is described to adopt a variety of distorted co-ordination geometries without significant energy penalty [35]. Besides, we speculate that the IRT1 loop flexibility, as an IDR, could explain the remaining ability of triple mutant H162A/H164A/D173N to bind Zn²⁺, by allowing surrounding residues contribute to Zn²⁺ co-ordination, as predicted by Aphafold server 3 [36] for aspartic acid residue 144 (online supplementary figure 10).

Based on our NMR and MST observations, five residues from the regulatory loop of IRT1 could co-ordinate Zn²⁺. Since two Zn²⁺ ions have already been described to be co-ordinated by five ligands in proteins with one of the ligands would be acting as a bridge [37], two Zn²⁺-binding sites may also exist in the IRT1 loop. This hypothesis is also supported by prediction made using Alphafold server 3 for the IRT1 loop in the presence of two Zn²⁺ ions, where histidine 168 is presented as the bridge co-ordinating both ions (Figure 6). Unfortunately, the possible existence of two Zn²⁺-binding sites in the IRT1 loop cannot be distinguished with our data. As postulated for the histidine-rich cytosolic loop of hZIP4 [27], we claim that a single Zn²⁺-bound state probably does not exist, but rather the Zn²⁺-bound state is likely a conformational ensemble with the Zn²⁺ co-ordinated by multiple combinations of the histidine residues and aspartic acid 173, allowing transient Zn²⁺-binding modes within IRT1 depending on cytosolic Zn²⁺ concentrations. We also uncovered that the same residues involved in Zn²⁺ binding also participate to Mn²⁺ co-ordination, probably along with neighboring residues. We, however, demonstrate here that the IRT1 loop shows higher affinity to Zn²⁺ compared with Mn²⁺, consistent with the stronger degradation of IRT1 observed in plants facing Zn excess [24]. Metal ion availability and relative local cytoplasmic concentrations will likely dictate whether the IRT1 loop will be co-ordinating Zn²⁺, Mn²⁺, or both.

Altogether, our work offers a framework for the analysis of metal-sensing properties of ZIP transporters and deepens our understanding of how IRT1 protein senses metals through its regulatory loop. This understanding is crucial to grasp how plants optimize iron uptake and limit the absorption of highly reactive non-iron metals in plant tissues and to consider biotechnological approaches to modulate heavy metal accumulation in plants.

CD spectra were recorded on a Jasco J-815 spectropolarimeter equipped with a temperature controller operating at room temperature. CD spectra ranging from 190 to 240  nm were recorded in a 1-mm path length cuvette. Sample concentrations were at 25 µM, in 10 mM Tris, pH 7.0, and the data presented are an average of 128 scans.

The wildtype IRT1 loop peptide and the four mutant peptides were synthesized by Proteogenix with a purity grade of 95% minimum. Peptides were first dissolved at pH 3.5 to a final concentration of 500 µM in the presence of 150 mM NaCl. The pH was then adjusted to 6.7. Then, two equivalents of ZnCl₂ for the peptide concentrations were added, and the pH was again adjusted to 6.7. A 2D phase-sensitive 1H Clean-TOCSY [38] with 60 ms spin lock and NOESY experiments [38] with 200 ms mixing time were recorded at 5°C and 20°C on an AVANCE Bruker 800.13 MHz spectrometer, with a spectral width of 8013 Hz, without sample spinning, with 2 k real points in t2 and 512 t1-increments. Pulsed-field gradient-based WATERGATE was used for water suppression [39]. Three independent samples were prepared to test different conditions, such as salt concentration, pH, and temperature. The data were processed using TopSpin 4.0.6 software (Bruker). π/3 and π /6 phase-shifted sine bell window function was applied before Fourier transformation in both dimensions (t1 and t2). Data processing and analysis were performed using the Topspin ® 4.0 and CCPN NMR software [40].

Interproton distance restraints were derived from the 2D ¹H NOESY (with a 200 ms mixing time and at 5°C) using CcpNmr 2.4 [40] and used to generate the IRT1 loop structures with the program CYANA version 3.98.5 [41]. We used the standard CYANA protocol of seven iterative cycles of NOE assignment and structure calculation, followed by a final structure calculation. In each cycle, the structure calculation started from 200 randomized conformers, and the standard CYANA simulated annealing schedule was used with 10,000 torsion angle dynamics steps per conformer. Graphic representations were prepared with PyMOL [42].

The wildtype IRT1 loop (a.a. 144–185) was cloned in the pMalc2x vector, in frame with the sequence encoding the maltose-binding protein (MBP) tag. H162A, H164A, H166A, H168A, and D173N mutations were introduced on the pMal-IRT1 vector by primer reactions in parallel SPRINP method [43], using primers listed in Supplementary Table S1. The different constructs generated are found in Supplementary Table S2. Vectors were transformed into BL21 DE3 E. coli, and recombinant proteins were purified using the manufacturer’s recommendation. Briefly, 1l of LB supplemented with glucose at 0.2% (w/v) final concentration, and ampicillin was inoculated with 10 ml of an overnight high-density culture of cells containing the fusion plasmids pMal containing wildtype and mutants IRT1. Cultures were grown at 37°C upon agitation until OD₆₀₀ reached 0.5. The expression of the wildtype and the IRT1 loop mutants was induced by the addition of IPTG to a final concentration of 0.3 mM, and cell cultures were incubated at 20°C overnight. Pellets were recovered by centrifugation at 10,000×g for 40 min at 4°C, then suspended in HEPES 10 mM pH 7, NaCl 150 mM in 30 ml/l culture. Suspended cells were stored until further processing at −80°C.

Affinity purification of the MBP-IRT1 loop variants was performed in batches. Frozen cell suspensions were thawed. Lysozyme was added for cell wall disruption, and then the suspensions were sonicated with eight pulses for 45 seconds while keeping on ice. Clarification of the lysate was performed at 20,000×g for 30 min. The supernatant (crude extract) was recovered and stored in ice. Separately, 500 µl of amylose resin (NEB) was washed according to the provider’s specifications. Washed amylose resin was incubated with the crude extract in rotation at 4°C for 1 h 30 min. The resin was then pelleted at 500 g, and flow through was discarded. Two washing steps were performed with HEPES 10 mM pH 7, NaCl 150 mM, 1 mM EDTA, followed by two more washes in the same buffer without EDTA. Three consecutive elutions were carried out with HEPES 10 mM pH 7, NaCl 150 mM, and Maltose 10 mM.

Binding experiments were performed by MST with a Monolith NT.115 (NanoTemper® Technologies, Munich, Germany). The MBP-IRT1 loop variants were labeled with the Monolith NT™ Protein Labeling Kit RED according to the instructions provided by the manufacturer, using a 1:3 protein:dye molar ratio. For binding experiments, the labeled proteins (20 nM) were incubated with a range of titrant concentrations made by serial dilutions (1:2), in 50 mM Tris buffer pH 7.4, 10 mM MgCl₂, 150 mM NaCl, 0.05% Tween 20, in PCR tubes, at room temperature for 10 min. Premium treated capillaries (NanoTemper® Technologies) were loaded and the measurements were performed at 25°C, 40% LED power and 40% MST power, 20 s laser‐on time and 1 s laser‐off time. All the experiments were repeated three times with two independent protein-labeling reactions. Binding data were analyzed using MO. Affinity Analysis software (NanoTemper® Technologies).

IRT1 mutations H162A, H164A, and D173Q were introduced into the pDONR-IRT1mCitrine [24] or the pDONR-IRT1-ALFA vectors [44] using the SPRINP method [43], using primers listed inSupplementary Table S1. For plant expression, final destination vectors were obtained by MultiSite Gateway® recombination using the entry vector described above, the pDONR-P4P1R-p35S entry vector containing the 2 × 35S promoter sequence, the pDONRP2RP3 entry vector containing a mock sequence, and the pGm43GW destination plasmid for expression in plants [24,45,46]. For yeast expression, final destination vectors were obtained by MultiSite Gateway® recombination using the entry pDONR-IRT1mCitrine vector harboring the mutations of interest and pDR195 destination vector.

The complementation of the wildtype yeast strain Y00000 (Mata; leu2Δ0; ura3Δ0; met15Δ0) and fet3fet4 (Mata; leu2Δ0; ura3Δ0; lys2Δ0; YMR058w::kanMX4;YMR319c::kanMX4) yeast mutant was performed by expressing IRT1 or IRT1 mutated versions (IRT1_D173N, IRT1_D173Q, IRT1_H162A/H164A, or IRT1_{H162A/H164A/D173Q}) in the pDR195 yeast expression vector. Transformants were selected on a selective medium lacking uracil. For complementation assays, strains were grown at 28°C for three or four days on a selective medium without iron or containing 100 µM of Fe-EDTA.

To generate a CRISPR/Cas9 construct for genome editing of A. thaliana (Col-0), two guide RNAs (gRNA) targeting IRT1 were cloned into the pHEE401 plasmid, as described in [47]. The sgRNA sequences were designed using CRISPR-P v2.0 [48] (Supplementary Table S1). T1 mutants were screened based on their chlorotic phenotype. Genomic DNA was extracted, and the targeted IRT1 region was amplified by PCR using gene-specific primers flanking the gRNA target site. PCR amplicons were subjected to Sanger sequencing to confirm the mutation. Mutant plants were then backcrossed with wildtype A. thaliana to eliminate the transgene with the Cas9. irt1_crispr Col-0 mutant was transformed with pIRT1:IRT1_H162A/H164A-mCitrine and pIRT1:IRT1_{H162A/H164A/D173Q}-mCitrine constructions by floral dip method [49]. Phenotype of transgenic lines were observed after growth in soil for 25 days.

N. benthamiana plants were grown for four weeks on soil under 16 h light at 22°C and 55% humidity before infiltration with Agrobacterium tumefaciens GV3101 strain.

IRT1-mCitrine variants were infiltrated into A. tumefaciens GV3110, and IRT1 protein localization assays were performed 48 h after agroinfiltration. Leaves transiently expressing IRT1-mCitrine variants were infiltrated with the treatments of study, consisting of control solution: half-strength Murashige and Skoog (MS/2) medium [50] lacking iron and non-iron metals; or metal excess solution: MS/2 lacking iron and containing Zn²⁺ (150 µM) and Mn²⁺ (300 µM). After 2 h, leaves were re-infiltrated with correspondent treatment supplemented with 300 µM of cycloheximide for 1 h (in order to distinguish between newly IRT1 protein synthesized and endocytosed IRT1), and confocal images were taken.

N. benthamiana leaves were infiltrated with IRT1-ALFA, ALFA NB-mCitN, and CIPK23-mCit_C vectors, as previously described [44]. Interaction between ALFA-tagged IRT1, recognized by the ALFA Nanobody-mCit_N fusion, and CIPK23-mCit_C allows reconstitution of mCit. CIPK24-mCit_C was used as negative control.

For IRT1-mCit fluorescence and TriFC imaging in N. benthamiana, infiltrated leaves were mounted in discs in each correspondent treatment solution, and confocal imaging was performed using a Leica TCS SP8 confocal laser scanning microscope (www.leica-microsystems.com/home/). A 25×-water-immersion objective was used to collect images at 1024 × 1024 pixel resolution. mCitrine excitation used the 514- nm laser line, and emission was collected from 525 to 580 nm. Laser intensity and detection settings were kept constant in individual sets of experiments.

Quantification of the ratio of the plasma membrane to cytosolic fluorescence signal and quantification of particles were calculated from the maximum projection of 10–15 optical sections taken using 1-µm z-distance. For the ratio, mean fluorescence was quantified using ImageJ software, and results are presented as the mean value ± standard deviation of n = 25–50 cell leaves from three independent experiments. Same cytoplasmic ROIs designed for measuring the ratio were used for quantifying particles using ComDet v.5.5.5 plugin from Image J software. One-way ANOVA was performed and the Tukey post-test method was applied to establish significant difference between means (P<0.0001). Statistical analyses were performed using GraphPad Prism version 7.00 software.

To quantify TriFC experiments, the mean mCit fluorescence intensity at the plasma membrane was quantified in a total of ten cells belonging to two different biological replicates. To do so, the ‘polygon selection’ tool from the ImageJ software was used, and ten values across the plasma membrane were measured per cell. One-way ANOVA was performed, and the Tukey post-test method was applied to establish significant difference between means (P<0.0001). Statistical analyses were performed using GraphPad Prism version 7.00 software.

Alphafold tool [51] (https://alphafold.ebi.ac.uk/) and Alphafold3 server (https://golgi.sandbox.google.com) were used for IRT1 protein structure prediction; UCSF ChimeraX (https://www.rbvi.ucsf.edu/chimerax) for modeling IRT1 structure predictions and performing metal ion contact analyses; PONDR [52] (http://www.pondr.com) for predicting IRT1 degree of disorder; and NetNGlyc [53] (https://services.healthtech.dtu.dk/service.php?NetNGlyc-1.0) was used for prediction of N-glycosylation sites in IRT1 protein.

Authors agree to make any materials, data, and associated protocols available upon request.

The authors declare that they have no conflicts of interest with the contents of this article.

This work was supported by research grants from the French National Research Agency [ANR-21-CE20-0046 to G.V.] and the French Laboratory of Excellence [project 'TULIP' grant nos. ANR-10-LABX-41 and ANR-11-IDEX-0002-02 to G.V.]. Financial support from the IR INFRANALYTICS FR2054 for conducting the research is also gratefully acknowledged. This work was also supported by a PhD fellowship from the Fondation pour le Recherche Médicale [FRM, ECO20170637545, to V.C.] and a postdoctoral fellowship from the Alfonso Martín Escudero Foundation (to R.R).

V.C.: Investigation, Formal analysis, Writing—original draft. R.R.: Investigation, Formal analysis, Writing— original draft, Writing—reviewing and editing. N.M.: Investigation, Formal analysis, Writing—original draft. S.F.: Investigation, Methodology, Formal analysis. V.C.: Validation, Writing—reviewing and editing. J.N.: Conceptualization, Validation, Supervision, Writing—reviewing and editing. G.V.: Conceptualization, Validation, Supervision, Writing—reviewing and editing, Funding acquisition, Project administration.

We would like to acknowledge the Fédération de Recherche Agrobiosciences Interactions et Biodiversité of Toulouse (FRAIB) for the access to confocal microscopes and microscale thermophoresis systems.

CD

circular dichroism

ER

endoplasmic reticulum

IDP

intrinsically disordered protein

IDR

intrinsically disordered region

IRT1

iron-regulated transporter 1

MST

microscale thermophoresis