The small ARF-like 2 GTPase TITAN5 is linked with the dynamic regulation of IRON-REGULATED TRANSPORTER 1

Plants adjust to environmental changes, such as variations of soil nutrient availability. Cellular membranes are very dynamic in plants, and this is crucial for sessile organisms. Plants can scan their environment for nutrients and control the activities of plasma membrane transporter proteins by vesicular trafficking. Many open questions still exist with regard to the environmental responsiveness and the complexity of the plant cell endomembrane system.

One essential element, iron (Fe), is abundant in the soil. However, given that Fe is mostly present in its ferric Fe³⁺ form in soil, it tends to be immobilized into insoluble complexes. Fe³⁺ must therefore be mobilized into a bioavailable form. Plants like Arabidopsis thaliana are capable of this. One mechanism consists of reducing Fe³⁺ via plasma membrane FERRIC REDUCTION OXIDASE 2 (FRO2) (Robinson et al., 1999). The reactive ferrous Fe²⁺ is taken up by IRON-REGULATED TRANSPORTER 1 (IRT1) into root epidermis cells (Eide et al., 1996; Vert et al., 2002).

IRT1 is one of the founding members of an evolutionarily conserved divalent metal ion transporter family named the ZINC-REGULATED TRANSPORTER, ZRT, IRT-LIKE PROTEIN (ZIP) family (Eide et al., 1996; Guerinot, 2000). Because divalent metal ions are potentially cytotoxic, the activity of IRT1, like that of other ZIP proteins, is controlled at the level of protein abundance at the plasma membrane (Connolly et al., 2002; Barberon et al., 2011; Shin et al., 2013; Barberon et al., 2014; Hu, 2021). Importantly, IRT1 protein activity is likely controlled through transport from the trans-Golgi network/early endosomes (TGN/EE) to the plasma membrane, degradation by the lytic pathway and recycling back to the plasma membrane (Ivanov et al., 2014). However, the molecular machinery for this process is barely known.

The large cytosolic variable region (vr) between transmembrane domains three and four is a characteristic regulatory feature of ZIP proteins (Guerinot, 2000; Gaither and Eide, 2001). In IRT1, this regulatory variable region, IRT1vr, is crucial for the control of IRT1 abundance. IRT1vr has metal ion-binding sites and residues that can be phosphorylated and ubiquitylated, which are processes that precede vesicular trafficking (Grossoehme et al., 2006; Kerkeb et al., 2008; Dubeaux et al., 2018; Cointry and Vert, 2019). In a search for new components binding with IRT1vr, we identified peripheral membrane proteins that directly or indirectly control the activity of IRT1 (Khan et al., 2019; Hornbergs et al., 2023). Among them is C2-domain protein ENHANCED BENDING 1 (EHB1, also known as CAR6) (Khan et al., 2019), which belongs to the ten-member Ca²⁺-dependent C2-DOMAIN ABSCISIC ACID-RELATED (CAR) protein family (Rodriguez et al., 2014). CAR proteins act as tethers between the membrane and signaling or transport proteins that are involved in drought, defense or nutrition (Cui et al., 2023). They can also oligomerize and cause membrane deformations, making them candidate proteins to be involved in vesicle budding (Chen et al., 2023). EHB1 interacts with IRT1vr and it has an inhibitory effect on Fe acquisition (Khan et al., 2019). Another IRT1vr interactor is PATELLIN 2 (PATL2). It has a SEC14 domain coupled with a GOLD domain. The SEC14 domain forms a lipid-binding pocket that serves to present or transfer lipophilic substances at membranes (Montag et al., 2023). For example, PATL2 can bind α-tocopherol and it has been proposed that it prevents oxidative stress upon Fe acquisition, possibly by transferring and exchanging phospholipids and α-tocopherol at the membrane (Hornbergs et al., 2023). PATL2 might also initiate vesicle formation, as it binds proteins of the endomembrane system (Hornbergs et al., 2023). Another hint for the involvement of endomembrane trafficking in IRT1 protein control stems from the observation that SORTING NEXIN 1 (SNX1), a crucial component of the retromer complex for transmembrane protein sorting, binds to endomembrane structures containing IRT1, and it is a positive regulator of IRT1 recycling at the TGN (Ivanov et al., 2014). However, IRT1 does not seem to bind SNX1 directly.

Altogether, it looks like IRT1vr has a regulatory role as protein interaction platform to recruit components for vesicular trafficking of IRT1. Fe acquisition is affected by regulation of the vesicular transport mechanism, but how this is achieved and which proteins are involved are open questions that still remain largely elusive. Whether and how the various IRT1vr-interacting peripheral membrane proteins might be connected with each other to control localization of a plasma membrane protein like IRT1, is not clear yet.

Cycling of plasma membrane proteins is an important function of endomembrane trafficking (Valencia et al., 2016; Ivanov and Vert, 2021). One protein family involved in this process in mammalian cells and yeast is the Ras superfamily of small GTPases. Small GTPases are important in vesicle-related processes as they function as molecular switches, due to their GTPase activities, to transmit signals. Small GTPases have usually low intrinsic nucleotide exchange and hydrolysis activity, and therefore, require the support of guanine nucleotide exchange factors (GEFs) and GTPase-activating proteins (GAPs). GEFs recruit the inactive, GDP-bound, GTPase to their site of action and lead to nucleotide exchange by GDP release. GTP binding leads to a conformational change of two regions referred to as switch I and II. The active, GTP-loaded, GTPases then exert their function, for example, the recruiting and subsequent assembly of coat proteins by ARF1 or SAR1, until hydrolysis of the GTP by GAPs. Most of the known interactions occur in the active conformation of the GTPases (Sztul et al., 2019; Nielsen, 2020; Adarska et al., 2021). Although the ARF GTPase family is well-studied in mammals, there are only a few well-established examples of an ARF GTPase functioning in plasma membrane protein trafficking in plants, like that of ARF1-GNOM-mediated recycling of auxin carrier PIN FORMED 1 (PIN1), with GNOM being an ARF GEF (Steinmann et al., 1999; Geldner et al., 2003). The presence of 21 ARF and ARF-like (ARL) GTPases encoded by the Arabidopsis genome raises the question whether other ARF members might act in endomembrane trafficking and what are their functions in the diverse physiological and nutrient assimilation pathways besides auxin signaling and cell division.

TITAN5 (TTN5), also known as HALLIMASCH (HAL), ARL2 and ARLC1, is an ARF-like protein. It was identified in two independent screens for abnormal embryo mutants arrested soon after egg cell division, indicating a central cellular function (Mayer et al., 1999; McElver et al., 2000), in agreement with its ubiquitous expression in plants (Mohr et al., 2024). TTN5 is more closely related in sequence to human ADP-ribosylation factor-like 2 (HsARL2) than any Arabidopsis protein. HsARL2 is associated with very diverse roles in cells, ranging from microtubule development, which have also been identified for yeast and Caenorhabditis homologs (Bhamidipati et al., 2000; Fleming et al., 2000; Radcliffe et al., 2000; Antoshechkin and Han, 2002; Tzafrir et al., 2002; Mori and Toda, 2013), adenine nucleotide transport in mitochondria (Sharer et al., 2002) and control of phosphodiesterase activity in cilia (Ismail et al., 2011; Fansa and Wittinghofer, 2016). HsARL2 is not only involved in a diverse set of functions and signaling cascades, but it also binds very different protein partners for that. The same can be expected for TTN5. TTN5 is expressed in the root epidermis (Mohr et al., 2024), like IRT1 (Vert et al., 2002). TTN5 is present at the plasma membrane and in the endomembrane compartment, and is among the rare plant small GTPases for which GTPase activities are known. TTN5 can be considered atypical as it does not require a GEF to be in its active GTP-bound form and has a slow GTPase activity relying on a GAP (Mohr et al., 2024). Its unusually high intrinsic GDP-GTP exchange activity, which causes TTN5 to be present predominantly in a GTP-binding form, seems a prerequisite for its dynamic cellular localization and association with the dynamic endomembrane system and potential vesicular trafficking (Mohr et al., 2024). The physiological functions of this small GTPase, on the other hand, relevant during the entire plant life cycle, have remained elusive.

A yeast two-hybrid screen had identified EHB1 and PATL2 as IRT1 regulators (Khan et al., 2019; Hornbergs et al., 2023). We report here, that with this same strategy, we retrieved the HsARL2-related, small GTPase TTN5. We validated protein interactions of TTN5 and IRT1vr as well as the IRT1vr interactome and found a physiological Fe reductase phenotype linked with root Fe acquisition. Colocalization analysis indicated that active GTP-bound TTN5 had a greater degree of colocalization with IRT1 in intracellular vesicles than a mutant TTN5 form. Hence, our work indicates that the dynamic small GTPase TTN5 acts in a nutritional environmental context in the vesicular trafficking system that controls the root Fe transporter IRT1.

To identify new cytoplasmic proteins regulating IRT1 we had developed a strategy to identify candidate interaction partners of IRT1vr (residues 145–192) and used IRT1vr as bait against a cDNA expression library prepared from Fe-deficient Arabidopsis roots (Khan et al., 2019; Hornbergs et al., 2023). We report here that 25 of the colonies (representing 25%) obtained in the reported yeast two-hybrid (Y2H) screen carried a fragment of the coding sequence of the gene AT2G18390, encoding the ARL-type small GTPase TTN5. The interaction was confirmed in a targeted Y2H assay (Fig. 1A). Small GTPases can act as molecular switches in cells due to their rapid reactions and interactions with effectors, and TTN5 might not require a GEF for nucleotide exchange as it has high affinity for GTP (illustrated in Fig. 1B) (Mohr et al., 2024). We further verified the protein interaction in plant cells using bimolecular fluorescence complementation (BiFC), which is visible as a reconstitution of YFP in positively transformed cells that expressed the control marker mRFP (Fig. 1C). We used previously characterized mutants of conserved amino acids in the GTP-binding pocket, Thr30-to-Asn (T30N, previously found to be dominant-negative or fast cycling) and Gln70-to-Leu (Q70L, with reduced GTP hydrolysis activity) (Fig. 1B) (Mohr et al., 2024). Both the cYFP–TTN5^T30N and cYFP–TTN5^Q70L forms interacted with nYFP–IRT1vr (Fig. 1D,E). The BiFC interactions were specific as neither protein partner was able to bind negative controls in this assay (Fig. 1F,G). We further elucidated whether IRT1vr might discriminate between the conserved TTN5 variant forms using a quantitative analysis of the interaction by Förster resonance energy transfer-acceptor photobleaching (FRET-APB). This approach is based on close-proximity-dependent energy transfer of an excited GFP-tagged TTN5 donor to a mCherry acceptor, here IRT1vr–mCherry. Energy transfer was quantified as FRET efficiency (Fig. 1H–J). FRET efficiencies between all GFP–TTN5 forms and IRT1vr–mCherry were significantly higher compared to the donor-only sample, confirming protein interactions between all TTN5 forms and IRT1vr. As final evidence, we performed an in vivo pulldown experiment to show interaction of full-length IRT1 and TTN5 (Fig. 1K) using an Arabidopsis plant line expressing IRT1–mCherry together with hemagglutinin-tagged (HA)–TTN5. The mCherry part was integrated in the first small cytosolic loop of IRT1 at position 80 (proIRT1::IRT1Q80–mCherry, hereafter IRT1–mCherry), and we determined that this fusion protein was functional as it fully complemented the severe irt1-1 (SALK_054554) Fe deficiency phenotype (Fukao et al., 2011) (Fig. S1). pro35S::HA₃-TTN5 (hereafter HA₃–TTN5) was also functional as it complemented the ttn5-1 phenotype as previously described (Mohr et al., 2024). We pulled down protein complexes from whole seedlings using anti-HA beads. This resulted in an increased protein abundance of HA₃–TTN5 (28 kDa) in the elution fraction, as expected (Fig. 1K). Importantly, IRT1–mCherry was simultaneously co-precipitated and detected in this elution fraction whereas it was not detected in a control pulldown experiment without HA₃–TTN5 (Fig. 1K), indicating that the two full-length proteins interact closely in Arabidopsis plant cells.

In summary, we conclude that TTN5 interacts with IRT1 through the IRT1vr domain surface. We hypothesize that TTN5 is a potential link in the coordination of IRT1 regulation with vesicle trafficking typical for ARF GTPases.

Following up on the IRT1–TTN5 interaction, we tested whether TTN5 might affect plant growth under Fe-deficient conditions and/or the regulation of the Fe deficiency response. Homozygous ttn5 knockouts are embryonic lethal (Mayer et al., 1999; McElver et al., 2000), and initial growth experiments with heterozygous plants (ttn5-1^+/−) showed no obvious developmental defects in comparison with wild-type sibling plants under regular growth conditions (Fig. 2A, Fig. S2; Mohr et al., 2024). Likewise, when plants were germinated in Fe-sufficient and -deficient conditions, we did not note any difference in root length. Both, wild-type and ttn5-1^+/− siblings had increased root length under Fe-deficient conditions (Fig. 2B), as expected in our growth system (Gratz et al., 2019), and there was no difference between them. SPAD values indicating chlorophyll contents were also similar in the wild-type and ttn5-1^+/− seedlings, indicating that the ttn5 mutation has no effect on leaf color (Fig. 2C). Also at advanced growth stages, ttn5-1^+/− plants did not differ in growth and leaf color from their wild-type siblings, neither in in control soil (pH 6.2) nor in an alkaline, calcareous soil (ACS, pH 8). The latter conditions require Fe mobilization capacities given that Fe acquisition is inhibited at high pH values in the presence of bicarbonates, resulting in smaller plants with leaf chlorosis (Ohwaki and Sugahara, 1997; Schmid et al., 2014;  Fig. S2B). Both wild-type and ttn5-1^+/− plants showed similarly lower aerial biomass increase and lower chlorophyll in ACS conditions over a period of 43 days compared to what as seen in the control conditions (Fig. S2A–D). We examined the heterozygous seedlings at the molecular and physiological level. TTN5 expression was nearly 50% reduced in ttn5-1^+/− compared to wild-type siblings, consistent with single allele presence (Fig. 2D). Very interestingly, we detected a 15% decrease of TTN5 gene expression under Fe deficiency compared with Fe sufficiency in both wild-type and ttn5-1^+/− plants, indicating that TTN5 might be perhaps less needed under low Fe supply (Fig. 2D). FRO2 encodes an Fe uptake component, the ferric reductase oxidase enzyme (Robinson et al., 1999), that closely interacts with IRT1 in the root plasma membrane (Martín-Barranco et al., 2020). FRO2 and IRT1 gene expression increase in response to Fe deficiency, and this serves as a molecular marker for the root Fe deficiency status (e.g. Gratz et al., 2019). No difference of gene expression levels of FRO2 and IRT1 occurred in ttn5-1^+/− roots compared with wild type, and both genes were induced under low Fe versus high Fe as expected (Fig. 2E,F). This indicates that root Fe deficiency responses might not be compromised in ttn5-1^+/−. However, we noted that in two out of three experiments (each including the indicated biological replicates), that root Fe reductase activity was altered; it was up-regulated under Fe-deficient versus -sufficient conditions in wild-type and ttn5-1^+/− roots, as expected (e.g. Robinson et al., 1999; Gratz et al., 2019) but the induction was lower in ttn5-1^+/− compared with that in wild-type siblings (Fig. 2G).

Taken together, heterozygous ttn5-1^+/− seedlings have an Fe reductase phenotype that indicates that TTN5 has a positive effect of at the physiological level on Fe acquisition responses in the plasma membrane. This was interesting, because the two other interactors of IRT1vr, EHB1 and PATL2, have an opposite effect on Fe reductase activity (Khan et al., 2019; Hornbergs et al., 2023).

To investigate whether TTN5 might be located in similar places to IRT1 in root cells, as expected for interacting proteins, we grew and compared side by side localization of fluorescent signals in Arabidopsis plant roots expressing either IRT1–mCitrine (proIRT1::IRT1–mCitrine/irt1-1; described in Dubeaux et al., 2018), or YFP–TTN5 (pro35S::YFP–TTN5; described in Mohr et al., 2024). A similar approach has been used to compare intracellular localization of IRT1 and PATL2 fusion proteins in root cells (Hornbergs et al., 2023). As previously described, the fluorescence signal of IRT1–mCitrine was present at the plasma membrane of root epidermis cells in our growth system (Fig. S3A) (Dubeaux et al., 2018; Hornbergs et al., 2023). A fluorescence signal at the plasma membrane was also detected for YFP–TTN5-expressing seedlings (Fig. S3B), and this corresponded to the previously identified pattern (Mohr et al., 2024). Virtual cross sections of both roots highlighted the described polar localization for IRT1–mCitrine, visible in only a very few epidermis cells, as previously seen (Fig. S3C; Dubeaux et al., 2018; Hornbergs et al., 2023) compared with an equally distributed YFP localization in the plasma membrane for YFP–TTN5-expressing seedlings in the epidermis (Fig. S3D). Next to plasma membrane localization, the fluorescence signals of both of IRT1–mCitrine- and YFP-TTN5-expressing seedlings were also present in vesicle-like structures in root cells (Fig. S3E,F). These findings suggest that, indeed, IRT1 and TTN5 locate to similar structures inside root cells, supporting an interaction at these places.

To confirm the obtained Arabidopsis data and to investigate potential colocalization of TTN5 with IRT1 in the same cells, we used the transient expression system of Nicotiana benthamiana. This system has the advantage that fusion proteins of TTN5 and IRT1 are well detectable, and through well-established pharmacological approaches, the localization patterns of TTN5 and IRT1 proteins can be confirmed (Ivanov et al., 2014; Mohr et al., 2024). We used a GFP-tagged IRT1 with GFP inserted inside IRT1vr (hereafter termed IRT1–GFP). This IRT1 fusion protein was present at the plasma membrane. Fluorescent spots were also found in close proximity to the plasma membrane (Fig. S4A), consistent with previous reports (Barberon et al., 2011; Ivanov et al., 2014). To further evaluate its usefulness for localization studies, we compared IRT1–GFP signals with already described C-terminally tagged IRT1–mCherry signals in N. benthamiana pavement cells before and after wortmannin treatment (see Ivanov et al., 2014). Wortmannin is a fungal metabolite with an inhibitory effect on phosphatidylinositol-3-kinase (PI3K) function, leading to swollen multivesicular bodies (MVBs) and is therefore commonly used to investigate endocytosis (Cui et al., 2016). Signals upon expression of these two fusion proteins showed colocalization (Fig. S4B) and, following wortmannin treatment, were observed, as well as at other locations, in swollen structures, suggesting a partial presence in the MVBs (Fig. S4C) and indicating that the IRT1–GFP fusion protein is suitable for IRT1 localization studies. IRT1–GFP also partially complemented an early rosette growth deficit of irt1-1 (in pro35S::IRT1-GFP/irt1-1 plants), indicating partial functionality of the IRT1–GFP construct for this trait (Fig. S4D).

We detected colocalization of IRT1–GFP and mCherry–TTN5, including fluorescence signals for mutant TTN5 at the plasma membrane or just below (Fig. 3A–C). Previously, we have described YFP–TTN5 localization in vesicle-like structures and the close proximity of YFP–TTN5 to the plasma membrane (Mohr et al., 2024). Moving fluorescent structures upon YFP–TTN5^T30N expression were less mobile than those seen for YFP–TTN5 and YFP–TTN5^Q70L (Mohr et al., 2024). Very interestingly, the different mCherry–TTN5 had different abilities to colocalize with IRT1, as determined by examining Pearson coefficients (Fig. 3D). Whereas the Pearson coefficients of signals of IRT1–GFP with mCherry–TTN5 and mCherry–TTN5^Q70L were 0.80 and 0.77, it was higher, namely 0.90, for IRT1–GFP and mCherry–TTN5^T30N, demonstrating higher general colocalization between IRT1–GFP and the low-mobile TTN5^T30N form (similar results were found with overlap coefficients, Fig. S4E). Signals of all three mCherry–TTN5 forms were also localized in IRT1–GFP-positive structures close to the plasma membrane, suggesting colocalization in plasma membrane-derived vesicles (Fig. 3E–G). Surprisingly, the Pearson coefficients changed when these spotted structures were analyzed (Fig. 3H; Fig. S4F shows similar results for overlap coefficients). Whereas the values for fluorescent signals of mCherry–TTN5 (0.80) and mCherry–TTN5^Q70L (0.78) constructs with IRT1–GFP remained relatively similar to those for plasma membrane colocalization, we now obtained, at 0.78, a similar coefficient with mCherry–TTN5^T30N. Similar results were achieved upon checking the overlaps of mCherry–TTN5 structures with IRT1–GFP and vice versa with a centroid-based analysis. We obtained overlaps of 36–41% for mCherry–TTN5 signals with IRT1–GFP (Fig. S4G), and 54% for IRT1–GFP-positive structures colocalized with mCherry–TTN5^T30N signals, whereas 64% and up to 83% was found for colocalization with mCherry–TTN5 and mCherry–TTN5^Q70L, respectively. From this, we hypothesize that IRT1 is more likely to be colocalized with TTN5^T30N at the plasma membrane than with wild-type TTN5 or TTN5^Q70L as the latter might be active in promoting intracellular cycling. This is consistent with the above observation that mCherry–TTN5^T30N signals were slower moving in certain cells than those of the other two TTN5 fusion proteins. We therefore deduce that TTN5^T30N might slow down the removal of IRT1 from the plasma membrane. In an attempt to identify the intracellular structures in which colocalization of TTN5 and IRT1 occurred, we treated the cells with wortmannin (Fig. 3I). Structures positive for both IRT1–GFP and mCherry–TTN5 fluorescent signals became doughnut shaped, suggesting that IRT1–TTN5 colocalization occurs in MVBs (Fig. 3J,K). The presence of IRT1 in these late endosomal compartments has been observed previously and might be part of the pathway to IRT1 vacuolar degradation (Dubeaux et al., 2018), as is seen for mCherry–TTN5 (Mohr et al., 2024). Thus, our findings emphasize the involvement of TTN5 in the endomembrane trafficking of the Fe transporter IRT1.

IRT1 can associate and colocalize with peripheral membrane proteins SNX1, EHB1 and PATL2, which are associated with events of vesicular trafficking (Ivanov et al., 2014; Khan et al., 2019; Hornbergs et al., 2023). This opens up the possibility that TTN5 might associate not only with IRT1 but also with proteins of IRT1 interactome in close proximity in cells, prompting us to investigate this further.

At first, we tested the IRT1-interacting protein EHB1. CAR proteins like EHB1 have a plant-specific insertion of an extra region within the C2 domain known as a signature (sig) domain that is involved in binding proteins including IRT1 (Rodriguez et al., 2014; Khan et al., 2019). We demonstrated above that nYFP–EHB1Δsig did not interact with cYFP–TTN5 in BiFC (Fig. 1F). However, when using for interaction with cYFP–TTN5 we did find evidence of protein interaction for the intact nYFP–EHB1 protein and the EHB1 sig-nYFP domain alone being in close proximity to the plasma membrane using BiFC, in contrast to EHB1Δsig (Fig. 4A–C). Surprisingly, the cYFP–TTN5 mutant forms both failed to complement with nYFP–EHB1 in the split-YFP assay (Fig. 4D,E). This was surprising because FRET-APB experiments validated the interaction between EHB1–GFP as the donor and TTN5–mCherry, as well as of the two TTN5 mutant variants, as the acceptors (Fig. 4F). Surprisingly, the FRET pairs EHB1–GFP with TTN5^T30N–mCherry and TTN5^Q70L–mCherry resulted in an even higher FRET efficiency compared to that seen with TTN5–mCherry (Fig. 4F). It is possible that the mutant behavior and difference between BiFC and FRET-APB were due to the different tag orientations in these two types of assays.

In another BiFC experiment, cYFP–TTN5 was able to complement split-YFP together with SNX1–nYFP close to the plasma membrane (Fig. 4G). The cYFP–TTN5^T30N mutant failed in BiFC (Fig. 4H), whereas the hydrolysis impaired TTN5 GTPase variant interacted with SNX1–nYFP (Fig. 4I). SNX proteins are classified by their characteristic domains. SNX1 represents the class I SNX proteins, characterized by the presence of PHOX-homology (PX) and Bin-Amphiphysin-Rvs (BAR) domains (Heucken and Ivanov, 2018). Membrane and phosphoinositide binding is mediated by the PX domain, whereas the BAR domain is involved in membrane curvature sensing and the formation of endosomal tubules (Peter et al., 2004; Frost et al., 2009). Interestingly, ARF GAPs, which are needed for the GTP hydrolysis, contain a BAR domain as well (Memon, 2004). We found evidence that cYFP–TTN5 binds to both the SNX1 PX–nYFP and BAR–nYFP domains using BiFC (Fig. 4J,K).

Finally, it is known that the plasma membrane-localized SEC14 protein PATL2 binds IRT1vr (Hornbergs et al., 2023), hence, we tested interaction ability of cYFP–TTN5 with PATL2–nYFP in BiFC experiments. YFP signals indicating complementation and interaction were visible at the plasma membrane, which is in accordance with the localization studies for both PATL2 and TTN5 (Fig. 4L). We were not able to observe YFP signal complementation for the combination of PATL2–nYFP and either of the two GTPase variants cYFP–TTN5^T30N (Fig. 4M) and cYFP–TTN5^Q70L (Fig. 4N). This was very interesting as it is either a limitation of the method or could indicate that the TTN5 switching mechanism might be required for a potential interaction between the proteins. However, this needs to be tested further in the future. To further identify the region of interaction, we tested already published deletion mutants of PATL2 (Montag et al., 2020). PATL2 can be divided into two protein halves, consisting of either an N-terminal part with no particular conformation or the C-terminal part with CRAL-TRIO-N-terminal extension (CTN)-SEC14-Golgi dynamics (GOLD) domains, which are conserved among SEC14-GOLD proteins in plants (Montag et al., 2020). The N-terminal part is needed for the interaction with IRT1vr (Montag et al., 2020). We used the PATL2 deletion mutants PATL2ΔC lacking the CTN-SEC14-GOLD part and PATL2ΔN lacking the IRT1vr-interacting N-terminal part. PATL2ΔC–nYFP together with cYFP–TTN5 complemented YFP fluorescence indicating interaction (Fig. 4O), whereas no signal was obtained by PATL2ΔN–nYFP in combination with cYFP–TTN5 (Fig. 4P), which is similar to the case for IRT1vr (Hornbergs et al., 2023). Interestingly, the fluorescent complementation signal between cYFP–TTN5 and PATL2ΔC–nYFP was no longer present at the plasma membrane, consistent with the observation that the PATL2 plasma membrane localization is dependent on the C-terminal PATL2 domain region (Montag et al., 2020).

We also analyzed EHB1, SNX1 and PATL2 gene expression. All genes were expressed in similar levels in both wild type and ttn5-1^+/− plants (Fig. S5A–C) indicating that there was no feedback at the level of regulation of genes encoding the TTN5 interactors.

In summary, we show that TTN5 is not only able to interact with IRT1vr but also with three peripheral membrane proteins that can be present in the vicinity of IRT1 at membrane sites in root cells.

Here, we report that the small GTPase TTN5 is a potential link in IRT1-related Fe homeostasis and trafficking of the transporter within the endomembrane system (Fig. 5). Such a finding is highly interesting. It was previously known that IRT1 is post-translationally controlled through a mechanism involving the vesicular trafficking system, but components that might bind with IRT1 during this process had remained obscure. Moreover, the connections of small GTPases, like TTN5, with endomembrane trafficking were mostly not ascribed to concrete physiological roles. Hence, our report is very timely, and highlights direct protein connections between cargo and vesicle trafficking components that fills a gap in our understanding.

The protein interaction between TTN5 and IRT1 is supported by multiple findings. First, the interaction was detected in a Y2H screen, which yielded only few interaction partners, all of which could all be validated (Khan et al., 2019; Hornbergs et al., 2023). Second, the interaction was verified by using independent methods [targeted Y2H, BiFC, FRET-APB and co-immunoprecipitation (co-IP) experiments]. Third, fluorescence signals of mCherry–TTN5 and IRT1–GFP reporter proteins colocalized in cells, and their localization overlapped in intracellular membrane-associated locations in cells of the root epidermis, in which both genes are expressed. Furthermore, TTN5 was found to interact with proteins of the IRT1vr environment (Fig. 5A,B). These findings altogether indicate that, in plant roots, TTN5 and IRT1 can indeed interact in root epidermis cells in locations related to Fe acquisition. The previously described TTN5 localization pattern in vesicles and at the plasma membrane (Mohr et al., 2024) was confirmed upon colocalization with IRT1 in this study. Hence, IRT1 localization dependent on endomembrane trafficking might be linked with the presence of TTN5 in this process.

A very interesting finding was that the mCherry–TTN5^T30N variant tended to have higher colocalization levels at the plasma membrane with IRT1–GFP than did wild-type TTN5. This could explain previous colocalization data where an MVB marker was found to overlap more with YFP–TTN5 and the hydrolysis impaired mutant YFP–TTN5^Q70L than YFP–TTN5^T30N (Mohr et al., 2024), and reinforces our suggestion that TTN5 has functions in vesicle trafficking. Although a GEF is not needed for nucleotide exchange of TTN5 (Mohr et al., 2024), TTN5^T30N might have a conformation that facilitates binding of a GEF so that TTN5^T30N, along with IRT1, might be retained at the plasma membrane by the unknown TTN5-interacting GEF. The presence of mCherry–TTN5 signals together with IRT1–GFP in wortmannin-induced swollen structures, probably indicating MVBs, supports the idea of a role in transporter degradation. MVBs are known to be the final step before vacuolar degradation (Kolb et al., 2015; Arora and Van Damme, 2021). This could indicate that the presence of TTN5 is crucial point for the IRT1 sorting fate. We therefore propose a role of TTN5 in targeting intracellular IRT1 protein (Fig. 5B). In such a process, TTN5 could be present at the plasma membrane in a resting state. After GTPase activation, it could then be recruited to the place of action, and active TTN5 would play a role in IRT1 cycling. TTN5 is thought to predominantly bind GTP and be in an active state in the cell (Mohr et al., 2024). Furthermore, there is support for a role of TTN5 in IRT1 recycling through the binding of TTN5 and SNX1. SNX1 is involved in the recycling of IRT1 at these sites (Ivanov et al., 2014). Hence, TTN5 might be in a functional complex with SNX1 to control the recycling of the cargo protein IRT1.

The observed roles of TTN5 fit well with the general understanding of ARF signaling in plasma membrane protein trafficking. The auxin transporters PIN1 and PIN2 and the boron transporter BOR1, like IRT1, are polarly localized in the plasma membrane and undergo a constant degradation-recycling process (Benková et al., 2003; Takano et al., 2010; Barberon et al., 2014). Apart from vesicle formation, as described for ARF1 or SAR1, a role in this process has been shown for other regulatory mechanisms of ARF signaling through the GEFs and GAPs (Nielsen, 2020). GNOM and BIG5, two ARF GEFs and ARF1 regulators, are described to be involved in the correct localization of the plasma membrane transporter PIN1, functioning at different regulatory steps (Geldner et al., 2003; Anders et al., 2008; Kleine-Vehn et al., 2008; Tanaka et al., 2009). In addition, another ARF GTPase, ARF1A1C, might not only partially colocalize with GNOM, but modified activity also results in impaired localization of PIN1 to the plasma membrane and also altered vacuolar targeting of PIN2 (Tanaka et al., 2014). BOR1 plasma membrane localization depends on GNOM as well (Yoshinari et al., 2021), showing that there is a good connection between ARF GEFs and plasma membrane protein localization.

Taken together, TTN5, which is an endomembrane system protein (Mohr et al., 2024) is linked with a transporter protein whose abundance is controlled by endomembrane trafficking. For future studies, stable coexpression of both proteins under the control of their endogenous promoters will be needed to investigate the effect of TTN5 on IRT1 localization. Finally, owing to the very low intrinsic GTP hydrolysis activity of TTN5, identification of TTN5 GAPs will shed light on TTN5 function in Fe homeostasis and potentially other cellular signaling events. It will also be interesting to explore whether a membrane-localized GEF exists that can act upon TTN5. Future identification of potential TTN5 GEF and GAP proteins will be of great interest to clarify the role of TTN5 in endomembrane trafficking with particular reference to cues controlling IRT1.

TTN5 does not only have an influence on IRT1 but also on other aspects in Fe nutrition. A partial lack of TTN5 results in decreased Fe reductase activity. The decrease in Fe reductase activity is likely caused by an effect on FRO2. It has been shown that IRT1 can interact with FRO2 to form a complex for a highly effective Fe uptake machinery (Martín-Barranco et al., 2020). Reduced Fe reductase activity could be due to an indirect influence of TTN5 on the complex formation or it could be directly driven by TTN5. Interestingly, both ehb1 and patl2 seedlings have a higher Fe reductase activity under Fe deficiency compared to what is seen in wild-type plants (Khan et al., 2019; Hornbergs et al., 2023). If EHB1 and PATL2 also act on FRO2, this points to an opposing role of TTN5 on the Fe deficiency response in the protein complex. Gene expression analysis revealed no great impact on Fe-related genes in ttn5-1^+/− roots, implying that the effect of TTN5 is mostly at the posttranslational level.

It is a possibility that TTN5 has very different effects depending on its localization in the cell. For example, the effect on Fe reduction could be different from its effect on colocalization with IRT1 in MVBs. Different types of interaction partners have been reported for the human ARL2 with connection to distinct cellular functions in tubulin folding and microtubule dynamics (Bhamidipati et al., 2000) or in mitochondria (Sharer et al., 2002). Based on these distinct locations and the associated functions of ARL2, it has been proposed that ARL2 might act at a high level in signaling (Francis et al., 2016). This might be also the case in Arabidopsis. For example, TTN5 might bridge TTN5 microtubule function (Mayer et al., 1999) with IRT1 regulation.

The close interaction of TTN5 with EHB1 was suggested by BiFC and FRET-APB, and with PATL2 and SNX1 by BiFC. The interactions are further supported by deletion mutant approaches. Moreover, the interacting sites of PATL2 and EHB1 with TTN5 are similar to those for IRT1vr, again supporting the possibility that these domains and motifs are at the basis of a protein complex forming at the IRT1vr regulatory platform.

The interaction of TTN5 with the IRT1 regulators EHB1, PATL2 and SNX1 suggests that it could not only be linked with Fe nutrition but also the associated Ca²⁺ and reactive oxygen species (ROS) signaling events (Khan et al., 2019; Hornbergs et al., 2023). EHB1 is additionally interesting as a member of the CAR family. Several CAR proteins can initiate Ca²⁺-dependent membrane curvature, which might provide a direct link to vesicle trafficking. Interestingly, Ca²⁺ is bound by EHB1 via its C2 domain, which is also a common feature of class III ARF GAPs (Vernoud et al., 2003; Knauer et al., 2011). Indeed, EHB1 is directly linked to ARF signaling in an antagonistic interaction with the ARF GAP AGD12 in gravitropic bending (Dümmer et al., 2016; Rath et al., 2020). Further research is needed to investigate whether TTN5 can also interact with the C2-domain containing ARF GAPs AGD11 and AGD13. Additionally, crystallization experiments have revealed a potential Ca²⁺-binding site between HsARF6 and its GAP ASAP3. The GTP hydrolysis was stimulated in the presence of Ca²⁺, suggesting that there is a link between Ca²⁺ and ARF signaling (Ismail et al., 2010). We therefore consider the possibility that the interaction between TTN5 and EHB1 occurs in a Ca²⁺-dependent manner. PATL2 might interact with IRT1 to prevent lipid peroxidation stress (Hornbergs et al., 2023). Prolonged Fe deficiency leads to an increase in hydrogen peroxide concentration as well as an activity increase of catalase CAT2 (Le et al., 2015; von der Mark et al., 2020; Gratz et al., 2021). Some catalases are present in the PATL2 interactome in roots (Hornbergs et al., 2023). Future studies should address whether TTN5 is connected with oxidative stress and Ca²⁺ signaling.

SNX1 is a positive regulator for IRT1 recycling to the plasma membrane in the sorting endosomes (Ivanov et al., 2014) and for endosomal recycling of the auxin efflux carrier PIN2 (Jaillais et al., 2006). SNX1 is therefore part of the system that mediates recycling of specific plasma membrane proteins. Interestingly, deficiency of BLOS1 leads to accumulation of PIN1 and PIN2 at the plasma membrane (Cui et al., 2010). BLOS1 seems to be an ortholog of the mammalian biogenesis of lysosome-related organelles complex 1 (BLOC-1), which is responsible for vesicle transport from endosomes to lysosomes (Li et al., 2007; Raposo et al., 2007). SNX1 can interact with both BLOS1 and BLOS2, and could therefore also play a role in vesicle-mediated transport through these interactions (Cui et al., 2010; Heucken and Ivanov, 2018). SNX1 possesses a BAR domain, which is also typical for class I ARF GAPs (Vernoud et al., 2003; Heucken and Ivanov, 2018). A connection between the SNX1 BAR domain and an ARF6 GEF has been identified in mouse by their interaction. This interaction was associated with a positive effect of ARF6 function (Fukaya et al., 2014), and colocalization analysis revealed overlapping expression in endosomes, which fits the known SNX1 localization in Arabidopsis (Fukaya et al., 2014; Ivanov et al., 2014), so this could be a possible interconnection between ARF signaling and SNX1. Here, we identified an interaction between TTN5 and SNX1, which could be the crucial link in the decision between SNX1-dependent recycling and degradation of IRT1. SNX1 could be needed for a potential TTN5 GEF and subsequent recruitment of TTN5 to membranes. Based on the opposing roles of EHB1 and SNX1 in IRT1 regulation, and their connection to TTN5, we suggest a coordinating role for TTN5 in transporting IRT1 between plasma membrane and endosomes by interacting with peripheral membrane proteins such as EHB1, PATL2 and SNX1. The findings that some ARF GAPs have a C2 or BAR domain, and that SNX1 interacts with an ARF GEF, strengthen the interest in the identification of TTN5 GEFs and GAPs.

Further research has to focus on physiological aspects in either combined knockout mutants or tagged-protein lines to arrange and decipher the IRT1 regulatory network. An intriguing open question is whether a large IRT1vr–EHB1–PATL2–TTN5–SNX1 complex is assembled. If so, the next question is how, when and in which order the individual or all components of the IRT1vr complex are assembled. Another question is what the structural requirements for proper protein conformations and membrane interfaces are (Fig. 5A,B).

This study provides hints that the ARF-like GTPase TTN5 contributes to endocytosis and vesicle trafficking of IRT1 by way of a protein complex at IRTvr comprising TTN5 and other IRT1 interactors. Evidence that small GTPases can bind with plasma membrane and peripheral membrane proteins is instrumental to understanding how membrane proteins are controlled by vesicular trafficking components in plants. An intriguing question is whether TTN5 also targets other transmembrane cargo proteins. Future work might uncover additional functions of TTN5 or of other GTPase protein complexes of this kind and determine the requirements and order of events for complex assembly leading to vesicular trafficking.

The Y2H screen for identification of IRT1vr interactors was performed previously and is described in Khan et al. (2019). Targeted Y2H interactions were tested using the GAL4 transcription factor system with histidine as selection marker. The TTN5 coding sequence was amplified using TITAN5 nter B1 and TITAN5 stop B2 primers (for primer sequences, see Table S1). Obtained fragments were cloned into pDONR207 via Gateway BP reaction (Life Technologies) with following LR reactions into pGBKT7-GW (binding domain, BD) or pACT2-GW (activation domain, AD). Both vectors were a kind gift from Dr Yves Jacob (Institut Pasteur, Paris, France). The yeast strain AH109 was co-transformed with the corresponding AD- and BD-tagged expression vectors. The combination pACT2:SNX1 and pGBKT7:SNX1 (Pourcher et al., 2010) was used as a positive control. Negative controls were co-transformations with the respective non-recombined pACT2-GW or pGBKT7-GW. Drop tests were performed with synthetic defined (SD) –Leu, –Trp medium (Takara Bio) cultures at an optical density at 600 nm (OD₆₀₀) of 0.1 and three additional 1:10 dilution steps (10¹-10⁴) on SD –Leu, –Trp and SD –Leu, –Trp, –His with 0.5 mM 3-amino-1,2,4-triazole plates. One of each plate was incubated either at 30°C or at room temperature, wrapped in aluminum foil, for up to 2 weeks. The experiment was performed three times.

N. benthamiana plants were grown on soil for 2–4 weeks in a greenhouse facility under long-day conditions (16 h light, 8 h darkness, 20°C). N. benthamiana leaf infiltration was performed with the Agrobacterium (Rhizobium radiobacter) strain C58 (GV3101; kind gift from Jörg Kudla, University of Münster, Germany) carrying the respective constructs for confocal microscopy. Agrobacteria cultures were grown overnight at 28°C, centrifuged for 5 min at 4°C at 5000 g, resuspended in infiltration solution (5% sucrose, a pinch of glucose, 0.01% Silwet Gold, 150 µM acetosyringone) and incubated for 1 h at room temperature. Bacterial suspension was set to an OD₆₀₀=0.4 and infiltrated into the abaxial side of N. benthamiana leaves.

Expression of pABind and pMDC7 (Curtis and Grossniklaus, 2003; Bleckmann et al., 2010) constructs was induced with a β-estradiol solution (20 µM β-estradiol and 0.1% Tween 20) 16 h before imaging.

BiFC was used to study protein interaction of TTN5 and the respective GTPase TTN5^T30N and TTN5^Q70L variants with other proteins in plant cells. At first, entry clones were generated. TTN5, TTN5^T30N and TTN5^Q70L coding sequences with a stop codon were amplified using the primers TITAN5 n-ter B1 and TITAN5 stop B4 (Table S1). The IRT1vr coding sequence with a stop codon was amplified using the primers I1L1B3 and I1LB2. For amplifying full-length EHB1, the primer pair EHB1 n-ter B3 and EHB1 stop B2 was used. Amplification of EHB1Δsig is described in Khan et al. (2019). The SNX1 coding sequence without stop codon for C-terminal fusion was amplified using the primer SNX1 B3 and SNX1ns B2. Cloning of SNX1 deletion mutants SNX1PX and SNX1BAR was done using the primers SNX1 B3 and PXSNX1 rev and BARSNX1 fwd and SNX1ns B2 respectively. The used PATL2 constructs are described in Hornbergs et al. (2023). The amplified PCR products were cloned via Gateway BP reaction (Life Technologies) into pDONR221-P1P4 (Invitrogen) or pDONR221-P3P2 (Invitrogen).

The obtained constructs were then used to prepare final recombinant destination vectors in an LR reaction (Life Technologies) for cloning into pBiFCt-2in1 vectors (Grefen and Blatt, 2012). Agrobacteria were transformed with correct constructs and used for N. benthamiana leaf infiltration. After 48 h, the mRFP expression control signal and YFP signals were detected by fluorescent microscopy (LSM 780, Zeiss) with a 40× C-Apochromat water immersion objective. YFP constructs were detected at 491–560 nm after exciting at 488 nm, and mCherry fluorescence was excited at 561 nm and emission detected at 570–633 nm.

The BiFC constructs were tested in three independent replicates with three infiltrated leaves each. The vectors pBiFC-2in1-NN and pBiFC-2in1-CN were kindly provided by Dr Christopher Grefen, Tübingen, Germany.

FRET-APB was used to verify protein–protein interactions of TTN5 TTN5^T30N, and TTN5^Q70L. The coding sequences without a stop codon were amplified with the primers TITAN5 B1 and TITAN5 ns B2 (Table S1), and cloned into the pDONR207 (Invitrogen, BP reaction, Life Technologies). Cloning of pDONR207:IRT1vr and pDONR207:EHB1 is described in Khan et al. (2019). The obtained constructs were used for LR reactions (Life Technologies) with the pABind-GFP, pABind-mCherry and pABind-FRET for C-terminal tagging (Bleckmann et al., 2010). N-terminal GFP–TTN5 constructs were cloned by overlap extension PCR simultaneously with mCherry–TTN5 constructs (see section ‘Subcellular localization of fluorescent protein fusions’ below). In the first step, the primer pairs GFP B1 with GFP R ns BIND and GFP to TTN5 with TITAN5 stop B2 were used with pABind-FRET (Bleckmann et al., 2010) as template for the fluorescent protein coding sequence. The final construct was obtained with the primer pair GFP B1 with TITAN5 stop B2. Constructs were cloned into pDONR207 for further cloning into pMDC7 via LR reactions. Agrobacteria were transformed and used for N. benthamiana leaf infiltration. FRET-APB was performed using laser-scanning confocal microscopy (LSM 780, Zeiss) with a 40× C-Apochromat water immersion objective. GFP was excited at 488 nm and detected at 491–560 nm, and mCherry fluorescence was excited at 561 nm and detected at 570–633 nm. GFP and mCherry channels were recorded for five scans. The mCherry signal was then bleached with 70 iterations of maximum laser intensity in a specific region of interest (ROI) which was set to the nucleus. Both channels were detected for additional 20 post-bleaching scans. FRET-APB measurements were performed with a minimum of 10 repetitions (n≥10).

The Arabidopsis ttn5-1 mutant was previously described (McElver et al., 2000). Heterozygous seedlings were selected by PCR on gDNA using the primer TTN5 intron1 fwd and pDAP101 LB1 (Table S1). pro35S::YFP-TTN5 and pro35S::HA₃-TTN5 lines were previously described (Mohr et al., 2024) as were proIRT1::IRT1-mCitrine/irt1-1 (Dubeaux et al., 2018). irt1-1 (Ws) was characterized in Vert et al. (2002) and irt1-1 (Col-0, SALK_054554) in Fukao et al. (2011).

Reporter lines were constructed as follows: proIRT1::IRT1-mCherry, with the mCherry at position 80 in the first cytoplasmic loop, was cloned by overlap-extension PCR. mCherry was amplified from pABind:mCherry (Bleckmann et al., 2010) using primer pair link mCh F and mCh ns link R. IRT1 halves were amplified using IRT1 B1 and IRT1Q80 link R and link IRT1Q80 F and IRT1 stop B2. In a final PCR, IRT1–mCherry was amplified with primer pair IRT1 B1 and IRT1 stop B2. BP and LR reaction (Life Technologies) were performed as described above with pIM5 as the destination vector. The Gateway-suitable plasmid pIM5 was created by exchanging pro35S from pMCD32 with proIRT1 using AQUA cloning (Beyer et al., 2015). The IRT1 promoter was amplified from gDNA using primer pair pIM5-pIRT1 F and pIM5-pIRT1 R with pMDC32-complementary overhangs and the restriction sites HindIII and KpnI, respectively. The IRT1–GFP construct, with the GFP at position 173 in the variable cytoplasmic loop, was cloned by overlap-extension PCR. GFP was amplified from pABind:GFP (Bleckmann et al., 2010) using primer pair IG173 F and GI173 R. IRT1 halves were amplified using IRT1 B1 and IG173 R and GI173 R and IRT1 stop B2. In a final PCR, IRT1–GFP was amplified with primer pair IRT1 B1 and IRT1 stop B2. BP and LR reaction were performed as described above with pMDC32 (Curtis and Grossniklaus, 2003) as destination vector. Agrobacterium cultures containing the plant protein expression vectors, generated as described above, were used for floral dip transformation of Arabidopsis. Plants were multiplied, PCR-genotyped using specific primer pairs (Table S1) and selected according to standard procedures.

Immunoprecipitation (IP) was performed with Arabidopsis seedlings grown for 10 days under Fe-deficient conditions. Plants with proIRT1::IRT1-mCherry in either the HA₃–TTN5 or wild-type background were grown and analyzed in parallel. Approximately 60 seedlings were flash-frozen in liquid nitrogen and ground using the Retsch MM200 with cooled grinding jars in the presence of two metal balls for 2 min. Ground material was solubilized in 1 ml IP buffer (50 mM Tris-HCl pH 8, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100 and cOmpete Mini Protease Inhibitor). The solution was incubated for 10 min at room temperature followed by a 10 min centrifugation at 4°C and 17,000 g (maximum speed). The supernatant was transferred into a fresh 1.5 ml reaction tube. The centrifugation step was repeated until all cellular debris were removed and 50 µl was taken as the crude extract (CE) aliquot. Next, 25 µl anti-HA magnetic beads (Thermo Fisher Scientific, Pierce) were added to the solution and the reaction was incubated overnight at 4°C while rotating. A magnet was used to collect the beads at the side of the reaction tube and solution was transferred to a fresh 1.5 ml reaction tube (Flow through). Then, 1 ml IP buffer was added to wash beads (Wash I). The collection of the supernatant and washing procedure was repeated twice. proIRT1::IRT1-mCherry immunodetection was performed using primary anti-RFP antibody (rabbit, 1:1000, abcam, ab62341) and secondary anti-rabbit-IgG conjugated to HRP antibody (goat, 1:5000, Agrisera, AS09 602). Immunodetection of HA₃–TTN5 was conducted using an anti-HA–HRP antibody (rat, 1:1000, Roche, clone 3F10, 11867423001). IP was performed twice in two replicates each (n=2).

Arabidopsis seeds were sterilized with sodium hypochlorite solution (6% Sodium hypochlorite and 0.1% Triton X-100) and stored 24 h at 4°C for stratification. Seedlings were grown upright on Hoagland medium [1.5 mM Ca(NO₃)₂, 0.5 mM KH₂PO₄, 1.25 mM KNO₃, 0.75 mM MgSO₄, 1.5 µM CuSO₄, 50 µM H₃BO₃, 50 µM KCl, 10 µM MnSO₄, 0.075 µM (NH₄)₆Mo₇O₂₄, 2 µM ZnSO₄ and 1% sucrose, pH 5.8, supplemented with 1.4% Plant agar (Duchefa)] with sufficient (50 μM FeNaEDTA, +Fe) or deficient (0 μM FeNaEDTA, –Fe) Fe supply in growth chambers (CLF Plant Climatics) under long-day conditions (16 h light at 21°C, 8 h darkness at 19°C). Seedlings were grown on Fe-sufficient or -deficient medium for 6 or 10 days or in a 2-week growth system with plants growing 14 days on Fe-sufficient medium and then being transferred to either fresh Fe-sufficient or Fe-deficient medium for an additional 3 days.

For root length comparison, plants were growing for 10 days in Fe-sufficient or -deficient conditions and photographed including a scale ruler in the picture. Root lengths were measured from root tip to hypocotyl using the JMicro Vision program (https://jmicrovision.github.io) (+Fe: WT n=19; ttn5-1 n=28; –Fe: WT n=16; ttn5-1 n=32).

For chlorophyll content comparison, plants were grown in the 2-week system in Fe-sufficient or -deficient conditions. Chlorophyll content was measured by SPAD values using SPAD-502plus (Konica Minolta Optics, 2012) with up to ten biological replicates (+Fe: n=10; –Fe: n=9). Chlorophyll content was determined in three individual experiments.

Sterilized seeds were placed on Fe-sufficient Hoagland plates in three rows with 12–15 seeds, stratified at 4°C in darkness for 3 days and then [on the day of sowing (DAS)] grown upright for 9 days. A total of 16 wild-type and 16 ttn5-1^+/− seedlings were selected for transfer onto two soil conditions, control and alkaline calcareous soil (ACS). The control condition consisted of peat-based soil (Floraton 1, Floragard, Oldenburg, Germany), 20 g vermiculite (Agrivermiculite Floragard, Oldenburg, Germany) and 400 ml distilled water per liter dry soil. The ACS soil additionally contained 8 g CaCO₃ (AppliChem, Darmstadt, Germany) and 4 g NaHCO₃ (Fisher Scientific, Hampton, USA). The pH in water (1:10 soil:water) was 6.2 for the control soil and 8.0 for the ACS soil 30 min after mixture. The soil was filled into 7 cm×7 cm×6.5 cm pots (Pöppelmann GmbH & Co. KG Kunststoffwerk–Werkzeugbau, Lohne Germany) and afterwards 7.5 cm×7.5 cm large squares of ultramarine blue plotter foil (Europe Warehouse GmbH & Co. KG, Wuppertal, Germany) were stuck on the pots. Seedlings were then planted into 1 cm diameter round holes in the foil. Eight plants were grown per soil and condition. After 1 week, the 1 cm holes were closed around the seedlings.

Plants were watered for 1 week with deionized water. Then control plants were watered with deionized water and ACS plants with 20 g/l NaHCO₃ (second week) and 25 g/l NaHCO₃ (following weeks). At 28 days after sowing the plants were scanned with a PlantEye MicroScan (Phenospex, Heerlen, The Netherlands), a multispectral scanner determining positions and color of points using a laser and four additional wavelengths (red, 624–634 nm; green, 530–540 nm; blue, 465–485 nm; and near-infrared, 720–750 nm). By filtering out the blue color (hue 200–360) from the images the background around the plants was removed from the measurements. Each plant was scanned twice, and the average of the technical replicates was calculated. The parameter ‘3D leaf area’ and ‘normalized differential vegetation index’ (NDVI) average were determined with the PlantEye MicroScan according to the instruction manual, corresponding to an approximation of aerial plant biomass and chlorophyll content. Photos of the plants were acquired with a camera (α 600, Sony, Tokyo, Japan). The entire experiment was conducted once with n=8.

For gene expression analysis, plants were grown in the 2-week system. Total RNA was isolated from roots using the RNeasy Mini Kit (Qiagen) and cDNA synthesis was performed with the RevertAid RT Reverse Transcription Kit (Thermo Fisher Scientific). The CFX96 Touch™ Real-Time PCR Detection System (Bio-Rad) was used for RT-qPCR performance. Data processing was undertaken with Bio-Rad CFX Manager™ (version3.1) software. Mass standard curve analysis was used for determination of the absolute gene expression and the elongation factor EF1Bα expression served as a reference for normalization. RT-qPCR was performed in three biological replicates (n=3) and two technical replicates each. All primer pairs used in this study are listed in Table S1.

Fe reductase activity assay was performed as described previously (Gratz et al., 2019). For Fe reductase activity, plants were grown in the 2-week system in Fe-sufficient or -deficient conditions. Two plants per replicate were washed in 100 mM Ca(NO₃)₂, then incubated for 1 h in the dark in 1.5 ml Fe reductase solution (300 mM ferrozine and 100 mM FeNaEDTA). The Fe reductase activity-dependent color change was measured at 562 nm using an Infinite 200^® PRO (Tecan) plate reader. The activity calculation was undertaken using the ferrozine extinction coefficient ε=28.6 mM⁻1 cm⁻¹ and was normalized to root weight. The assay was performed with a minimum of three replicates (n≥3), each consisting of a pool of two plants. Fe reductase activity was determined in three individual experiments.

The creation of TTN5^T30N and TTN5^Q70L constructs by introduction of point mutations and generation of pro35S::YFP-TTN5 is described in Mohr et al. (2024). mCherry-tagged constructs were created by overlap-extension PCR. Fluorescent protein was amplified from pABind:mCherry (Bleckmann et al., 2010) using primer pair mCherry B1 and mCh R ns BIND. TTN5, TTN5^T30N and TTN5^Q70L CDS were amplified with a 20 bp overlap of mCherry using mCh to TTN5 and TITAN5 stop B2. In a third PCR mCherry-tagged TTN5 constructs were generated with the mCherry B1 and TITAN5 stop B2 primer pair. The BP and LR reaction was performed as described above. The XVE-driven β-estradiol inducible vector pMDC7 (Curtis and Grossniklaus, 2003) was chosen as destination vector. Cloning of IRT1–GFP is described above (see section ‘Arabidopsis plant material’). The IRT1–mCherry construct is described in Ivanov et al. (2014) and proIRT1::IRT1-mCitrine/irt1-1 in Dubeaux et al. (2018). Agrobacteria were transformed with the obtained constructs and tested by N. benthamiana leaf infiltration after 2 days of expression.

Localization studies were carried out by laser-scanning confocal microscopy (LSM 780, Zeiss) with a 40× C-Apochromat water immersion objective. YFP constructs were excited at 488 nm and detected at 491–560 nm. mCherry or FM4-64 fluorescence was excited at 561 nm and detected at 570–633 nm. Each combination was tested a minimum of three times (n≥3).

Wortmannin (10 µM, Sigma-Aldrich) and plasma membrane dye FM4-64 (165 µM, Thermo Fisher Scientific) were infiltrated into N. benthamiana leaves. FM4-64 was detected after 5 min incubation and wortmannin was incubated for 25 min before checking the treatment effect.

Colocalization analysis was performed with the ImageJ (Schneider et al., 2012) plugin ‘Just Another Colocalization Plugin’ (JACoP) (Bolte and Cordelières, 2006). A comparison of Pearson's and overlap coefficients was undertaken. Object-based analysis was performed for spotted-structures, as adapted from the process described in Ivanov et al. (2014). The percentage of colocalization for both channels was calculated based on distance between geometrical centers of signals. Analysis was conducted in three replicates each (n=3).

One-way ANOVA was used for statistical analysis and performed in OriginPro 2019. Fisher LSD or Tukey was chosen as the post-hoc test with P<0.05.

Sequence data from this article can be found in the TAIR and GenBank data libraries under accession numbers: FRO2 (TAIR: AT1G01580), EHB1 (TAIR: AT1G70800), IRT1 (TAIR: AT4G19690), PATL2 (TAIR: AT1G22530), TTN5 (TAIR: AT2G18390) and SNX1 (TAIR: AT5G06140).

We thank Gintaute Matthäi and Elke Wieneke for excellent technical assistance. We thank Ksenia Trofimov for microscopic help and advice, and Rubek Merina Basgaran for help with protein interaction studies. We are thankful for the excellent assistance from Stefanie Weidtkamp-Peters and Sebastian Hänsch, members of the Center for Advanced Imaging (CAi) at the Heinrich Heine University. Funding for instrumentation was from Zeiss LSM780: DFG- INST 208/551-1 FUGG.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.263645.reviewer-comments.pdf