  NAS4

bHLH

Iron (Fe) is a crucial micronutrient needed in many metabolic processes. To balance needs and potential toxicity, plants control the amount of Fe they take up and allocate to leaves and seeds during their development. One important regulator of this process is POPEYE (PYE). PYE is a Fe deficiency-induced key bHLH transcription factor (TF) for allocation of internal Fe in plants. In the absence of PYE, there is altered Fe translocation and plants develop a leaf chlorosis. NICOTIANAMINE SYNTHASE4 (NAS4), FERRICREDUCTION OXIDASE3 (FRO3), and ZINC-INDUCED FACILITATOR1 (ZIF1) genes are expressed at higher level in pye-1 indicating that PYE represses these genes. PYE activity is controlled in a yet unknown manner. Here, we show that a small Fe deficiency-induced protein OLIVIA (OLV) can interact with PYE. OLV has a conserved C-terminal motif, that we named TGIYY. Through deletion mapping, we pinpointed that OLV TGIYY and several regions of PYE can be involved in the protein interaction. An OLV overexpressing (OX) mutant line exhibited an enhanced NAS4 gene expression. This was a mild Fe deficiency response phenotype that was related to PYE function. Leaf rosettes of olv mutants remained smaller than those of wild type, indicating that OLV promotes plant growth. Taken together, our study identified a small protein OLV as a candidate that may connect aspects of Fe homeostasis with regulation of leaf growth.

Iron (Fe) is a cofactor in many redox reactions and serves multiple functions in fundamental metabolic processes such as photosynthesis and cell respiration. Despite of being essential, Fe can become toxic and promote oxidative stress when it is present in excess. Thus, the micronutrient Fe is a decisive element for plant growth and health, and plants control tightly Fe acquisition and allocation.

Despite the fact that Fe is the fourth most abundant element in the continental crust, most of it is present in the form of poorly soluble Fe3+ oxides in the soil (Yi and Guerinot, 1996) . Inside plants, Fe can be bound or stored by cell walls, plastidial ferritin or precipitates in the vacuole. Plants have a number of transport proteins and enzymes that help them mobilize Fe and allocate Fe across membranes and longdistance, from soil to root epidermis cells, across different tissues, from roots to leaves and from there to sink organs. These transport processes may involve Fe reduction and chelation e.g. by ferric reductase oxidases and nicotianamine (Connorton et al., 2017) . Fe homeostasis of plants depends on the available Fe sources in the soil, but also on the plants’ requirements for Fe during growth. When plants sense Fe deficiency, they mobilize Fe in the soil and within the plants. This is controlled by a cascade of bHLH transcription factors (TFs) that are activated by low Fe. Different subgroups of bHLH TFs steer different aspects of Fe utilization in leaves and roots, and their activities are positively and negatively controlled (Schwarz and Bauer, 2020) . Yet, how the sensing of Fe deficiency is coupled with plant growth and the required uptake and delivery of Fe to target sites bares still many open questions. There are open questions how the bHLH TFs are controlled and which role positive and negative regulation plays.

POPEYE (PYE) is a Fe deficiency-induced bHLH TF, that mediates the internal mobilization of Fe in roots and shoots. PYE is most strongly expressed in the root stele and vascular bundles in leaves although it seems prone to cell-to-cell movement. In the absence of PYE, plants fail to fully utilize Fe and develop a Fe deficiency leaf chlorosis (Long et al., 2010) . pye loss of function mutants are not able to mobilize internal Fe. At the same time, they show up-regulated expression of different target genes (Long et al., 2010; Pu and Liang, 2023) . PYE target genes include FRO3, NAS4 and ZIF1 (Long et al., 2010) . FRO3 is a Fe-chelate reductase involved in mitochondrial Fe import (Jain et al., 2014) . Nicotianamine synthase (NAS4) can produce a metal chelator, nicotianamine, needed for metal ion allocation (Klatte et al., 2009; Schuler et al., 2012) . This nicotianamine may serve to translocate internal cellular metal ions towards the vacuole. Indeed, zinc facilitator ZIF1 is a vacuolar nicotianamine-metal importer (Haydon et al., 2012) . Perhaps, pye mutants retain metal ions in the vacuole instead of mobilizing them to various sink organs upon Fe deficiency. This suggests that PYE may prevent the internal cellular Fe and metal ion storage machinery in the mitochondria and vacuole which then promotes allocation of Fe to other leaves. PYE responds to Fe deficiency signals. It can interact with PYE-LIKE homologs of the bHLH subgroup IVc group (e.g. bHLH104/105/115) (Long et al., 2010; Zhang et al., 2015) . It can also form homodimers (Lichtblau et al., 2022; Pu and Liang, 2023) . A recent study showed that PYE may repress PYE itself and its co-expressed BHLH subgroup Ib genes (Pu and Liang, 2023) . Hence, this varying effect of PYE on different genes indicates that a molecular mechanism must be present that regulates PYE activity to maintain system homeostasis. PYE and related bHLH TFs can interact in protein complexes that may fine-tune the Fe deficiency response, and harbor for example Fe deficiency-induced E3 ligases BRUTUS (BTS)/ BTS-LIKE (BTSL) as negative factors (Li et al., 2021; Lichtblau et al., 2022) .

Small proteins are defined as polypeptides with usually less than 100 amino acids (aa) (Hsu and Benfey, 2018) . They are often mobile and can act in nutrient-related signaling (Takahashi et al., 2019) . In addition, they can be involved in the regulation of protein activity (Staudt and Wenkel, 2011) . A prominent class of protein interactors are small plant protein effectors that are injected by pathogens into plant cells to dampen the plant immune responses by interacting with proteins of many different kinds (Song et al., 2021) . Small proteins can interact and prevent transcription factor complexes from functioning, thus inhibiting their transcriptional regulation activities (Yun et al., 2008) . For example, KIDARI (KDR) is a small protein with Helix-Loop-Helix (HLH) domain that may act as a repressor of light signaling in Arabidopsis by binding to and inhibiting the activity of a light-controlled bHLH transcription factor (Hong et al., 2013) . Another example, BBX31 (length of 121 amino acids) and BBX30 (length of 117 amino acids) allosterically deactivate transcription factors (Song et al., 2020; Cao et al., 2023) . Therefore, identifying and characterizing small protein candidates that modulate bHLH transcription factor activity in the Fe signaling pathway can be relevant for better understanding the molecular mechanisms involved in PYE regulation. Presently, the importance of small effector proteins interfering with Fe deficiency responses is far from being fully understood. It is interesting to note that there are several Fe-deficiency-induced small proteins. Best studied among them is the family of IRON MAN (IMA) small proteins, also known as FE UPTAKEINDUCING PEPTIDEs (FEPs) (Grillet et al., 2018; Hirayama et al., 2018) . IMA small proteins mechanistically regulate the iron homeostasis signaling pathway by attenuating interactions between BTS or BTSLs and bHLHIVc transcription factors through direct binding, thereby stabilizing bHLHIVc transcription factors. PYE differs structurally from related bHLH TFs with regard to the protein interaction sites for BTS/Ls (Lichtblau et al., 2022) . PYE lacks the conserved BTS interaction motif. Therefore, evidence is still pending regarding the effect of BTS on modulating PYE at the protein level. Consequently, the existence of a different regulator of PYE function stands out as salient and warrants further investigation.

Here, we describe that PYE can bind to an Fe deficiency-inducible small protein that we named OLIVIA (OLV). We delimited the protein interaction sites. A molecular-physiological analysis suggests that

OLV has only mild effects on PYE-dependent Fe deficiency responses. OLV also affects biomass and plant growth, indicating a link between Fe deficiency responses during plant growth.

Protein-protein interactions play a crucial role to fine-tune the regulation of Fe uptake and homeostasis, and many of the relevant protein interactions for Fe-regulatory processes involve coexpressed proteins. We have previously reported that co-expression networks help identifying novel protein-protein interactions (Lichtblau et al., 2022) . Among the tested co-expressed proteins was At1g73120 (Supplemental Figure S1). At1g73120 encodes a small protein which we named OLIVIA (OLV) according to its partnership with PYE. PYE and OLV were co-transformed into yeast and tested in targeted yeast two hybrid (Y2H) assays. These assays allowed quantitative conclusions about the interaction strength, when yeast serial dilutions were analyzed. The interactions between PYE and OLV were reconfirmed in reciprocal manner (Figure 1A).

Next, we verified this interaction in plant protein interaction assays by using Förster resonance energy transfer-acceptor photobleaching (FRET-APB) and bimolecular fluorescence complementation (BiFC). FRET-APB is a sensitive method to analyze protein interactions, and FRET efficiency can be used to obtain a quantitative measurement for the interaction strength (Gratz et al., 2019) . FRET-APB analysis was performed in plant nuclei. FRET was detected for the pair PYE-GFP and OLV-mCherry with a FRET efficiency of 3.5 %. This FRET efficiency was significantly higher than the negative control (PYE-GFP only) (Figure 1B), confirming an interaction of PYE and OLV. Controlled BiFC experiments (Grefen and Blatt, 2012; Gratz et al., 2019) provided additional hints. A YFP signal was obtained when both nYFP-PYE and cYFP-OLV fusion proteins were expressed in cells. BiFC indicated that PYE and OLV interacted in the nucleus and cytoplasm (Figure 1C). Negative controls were nYFP-PYE and cYFP-OLV together with either cYFP-bHLH039 or nYFP

To further prove the interaction capacities of OLV protein and describe it at a molecular level, we first examined OLV protein structural predictions and then conducted a mapping of interaction sites for the OLV-PYE pair. OLV is a unique gene in the Arabidopsis genome. Orthologs of OLV are present in all investigated angiosperms. Interestingly, irrespective of the size and total percentage of amino acid (aa) identity, all OLV orthologs exhibited a conserved motif in the C-terminus (in AtOLV aa 71-87), which we named TGIYY motif due to the presence of the respective five aa in the middle of the motif (WVPHEGTGIYYPKGQEK; Supplemental Figure S2A, B; Supplemental Figure S3A, B). This motif does not resemble any known protein domain according to e.g Uniprot (www.uniprot.org) or InterPro-EMBL-EBI (www.ebi.ac.uk). Unlike Arabidopsis, some species have more than one ortholog of OLV or TGIYY-likecontaining protein (for example: Vitis vinifera, Populus trichocarpa, Eucalyptus grandis, Sesanum indicum, Daucus carota). TGIYY-like motif-containing proteins are also encoded in the genomes of green algae and lower land plants (Supplemental Figure S3C), as well as in organisms of other kingdoms such as bacteria, fungi, animals (examples are represented in Supplemental Figure S3D). In non-plant species, the TGIYYlike motif was part of proteins with unknown function that were much bigger than 109 aa (the sizes varied from less than hundred to several hundred aa).

In summary, all plants and other organisms have genes encoding proteins with a TGIYY or TGIYY-like motif. The high level of conservation points to a functional relevance of the TGIYY motif.

To investigate whether the conserved TGIYY motif is relevant for protein-protein interaction with PYE, we constructed OLV deletion mutant forms (Figure 2A) and tested them in BiFC assays. None of the OLV deletion variants devoid of TGIYY, namely OLV-N and OLV-T, interacted with PYE in BiFC, while constructs harboring the TGIYY motif did, OLV-C and TGIYY (Figure 2B). OLV-N, OLV-C and OLV-T were similarly localized in plant cell nuclei and the cytoplasm as OLV (Supplemental Figure S4). OLV-T was also tested in Y2H and FRET-APB assays. While OLV-T did not interact with PYE in Y2H, it was similarly capable of interacting with PYE in FRET-APB as the full-length OLV construct did (Figure 2C, D).

Hence, the TGIYY motif was important in two protein interaction assays but not in FRET-APB (Figure 2E). It may be that several regions of OLV can be involved in the interaction with PYE protein, which can be revealed depending on the tags of fusion proteins. We thus conclude that even through the TGIYY may not be essential it may facilitate the interaction between OLV and PYE.

Next, we determined the required protein domains of PYE needed for interaction with OLV. Different deletion constructs of PYE (Figure 3A) were tested in Y2H and in FRET-APB experiments. PYE harbors an ethylene-responsive element-binding factor-associated amphiphilic repression (EAR) motif in its C-terminal part. EAR motifs are commonly known as repression motifs in plants, commonly associated with TFs that function as adaptors to recruit TOPLESS and -related repressor proteins via EAR motif binding (Plant et al., 2021) . However, a function of the EAR motif in this regard may not be the case for the PYE protein (Pu and Liang, 2023) . Both, PYE and PYE-EAR interacted with OLV in Y2H assays (Figure 3B). On the other hand, all other tested PYE deletion constructs, which were represented by the N-terminal part with bHLH domain (PYE-N), the C-terminal part without bHLH domain (PYE-C), PYE devoid of the bHLH domain (PYE-bHLH) or only the bHLH domain (PYE-bHLH) did not show any interaction with OLV in Y2H assays (Figure 3B). Based on FRET-APB experiments, PYE-N, PYE-C and PYE-bHLH can be suggested to interact with OLV, but the interaction strengths determined as FRET efficiencies were lower in the case of all three PYE mutants compared to PYE (Figure 3C). In one and two out of three experiments, FRET efficiency was lowest for PYE-C and PYE-bHLH, respectively (Figure 3C).

Thus, PYE does not have a unique specific domain for interaction with OLV (Figure 3D). Clearly the EAR motif is not likely involved in the interaction. The Alphafold multimer tool was used to predict the structure of PYE-OLV proteins (Evans et al., 2022) . This model suggests how the three-dimensional structure of PYE can provide a binding site for OLV (Figure 3E). This prediction indicates an interaction interface of OLV near the PYE bHLH domain. An additional interface for OLV interaction is also predicted proximal to the C-terminal end of PYE (Figure 3E). Potentially, OLV may modulate the ability of PYE to control its target genes.

A prerequisite for protein interaction in plants is that PYE and OLV proteins are located in the same cellular compartments and tissues.

The FRET-APB experiments described above already revealed that both proteins localize to the nucleus. More detailed protein localization and co-localization studies were performed to examine in which cell compartments PYE and OLV are localized when expressed alone or together, respectively. Single localization experiments indicated that PYE-mCherry localized primarily to the nucleus in Nicotiana benthamiana (tobacco) leaves, while weak signals were detected in the cytoplasm (Figure 4A). OLV-GFP, on the other hand, localized clearly to nucleus, cytoplasm and at strands of the endoplasmic reticulum (Figure 4B). When PYE-mCherry and OLV-GFP were co-expressed in tobacco leaves, they co-localized inside the nucleus (Figure 4C). The pattern did not change for OLVT-GFP (Figure 4B, C). Therefore, colocalization studies suggest that OLV may interact with PYE primarily in the nucleus, whereby neither the presence of OLV nor that of the TGIYY motif changed the localization of PYE or OLV.

An interesting question was in which root tissues and root zones OLV and PYE are expressed. The promoter activities of both genes were assayed by using Arabidopsis lines that stably express the ßglucuronidase (GUS)-encoding reporter gene driven by the OLV or PYE promoters. Detection of GUS activity is a sensitive method to determine promoter activities in different root zones. The lines were grown for six days under sufficient (+Fe) or deficient (–Fe) Fe supply and analyzed for promoter-GUS activity (Figure 4D, E, F, G). In the proOLV::GUS line, promoter driven-GUS activity was detected mainly in the cortex and epidermis of the rot differentiation and root hair zones. GUS activity was not detected in the root tip and root elongation zone. GUS activity appeared patchy along the upper root zones (Figure 4D, E). proOLV lines grown under -Fe showed stronger GUS activity compared to plants grown under +Fe, implying an induction of the OLV promoter under -Fe conditions, as was expected. ProPYE::GUS expression had been reported previously for our growth conditions (Lichtblau et al., 2022) . In accordance with these published data, strongest PYE promoter activity was detected in tissues of the inner stele, such as the pericycle, along the mature root and root hair zones, in all tissues at the root tip as well as of the root elongation and differentiation zones. In the root elongation and differentiation zones, proPYE-driven activity was detected in the cortex and epidermis (Figure 4F, G). Hence, PYE and OLV promoter activities overlapped in the root epidermis and cortex of the root differentiation zone. In other parts of the root, the promoter activities of the two genes indicated non-overlapping expression.

In summary, PYE and OLV proteins can be present together in the root epidermis and cortex of the root differentiation zone, where the proteins may interact in the nucleus.

The next question was whether OLV affected the growth of plants or the activity of PYE. In our seedling growth assay, roots of Fe-deficient (- Fe) wild-type plants are longer than those of Fe-sufficient (+ Fe) ones (Gratz et al., 2019) , and the root length may change in OLV mutant plants. Since PYE represses FRO3, NAS4 and ZIF1 (Long et al., 2010) , we reasoned that a negative effect of OLV on PYE should result in higher expression levels, while a positive effect on PYE should cause lower expression levels of these PYE target genes.

At first, we assessed a possible impact of OLV overexpression. We analyzed two hemagglutinine (HA3)-tagged HA3-OLV overexpressing lines (pro35S; OX7, OX11) (Supplemental Figure S5A, B, C, D). HA3OLV protein (14.85 kDa) was detected in total protein extracts of 10 d-old seedlings independent of Fe supply, confirming the presence of HA-tagged OLV protein (Supplemental Figure S5B). Root length was not changed upon overexpression of OLV (Figure 5A). PYE gene expression levels were elevated in – Fe roots of wild type and OX7 and in – Fe shoots of wild type, OX7 and OX11 in comparison with the respective + Fe situation (Figure 5B, C). However, there were no significant differences in PYE levels between wild type and any OX line at - Fe or + Fe (Figure 5B, C). FRO3 and NAS4 were up-regulated in roots under – Fe versus + Fe in all tested lines and their expression was higher in pye-1 roots versus wild type and versus

Ox roots (Figure 5D, E). Interestingly, the only OX phenotype was a higher expression of NAS4 compared to wild type, which was significant in the case of OX11 (Figure 5E). ZIF1 was only up-regulated under - Fe versus + Fe in roots of pye-1 versus wild type (Figure 5F). In shoots, there was a significant up-regulation of FRO3, NAS4 and ZIF1 at - Fe versus + Fe in the case of pye-1 (FRO3, NAS4), OX7 (NAS4) or OX11 (NAS4, ZIF1), while there were no significant differences between the expression levels of any mutants with wild type (Figure 5G-I). Since OLV Ox had a mild effect on PYE target genes, we tested whether OLV Ox altered the pye-1 phenotype in any way. This was not the case, and pye-1 OX7 as well as pye-1 OX11 had the pye-1 phenotype with respect to FRO3 and NAS4 up-regulation in roots (Supplemental Figure S6). Hence, we could first of all, reproduce the positive effect that pye-1 has on the PYE target genes FRO3 and NAS4 in roots in our growth system, but we did not reproduce it neither in shoots nor for the reported target ZIF1. The only OLV overexpression phenotype observed was a positive effect on NAS4 expression in roots, suggesting that OLV has a mild or partial effect on the repression of PYE in this case.

Next, we investigated olv loss-of-function mutant lines that we had generated by genome editing. Two alleles, olv-3 and olv-7, had both premature stop codons in the OLV coding sequence (Supplemental Figure S5E, F, G). When grown in the seedling assay, there was again no noticeable difference in the root lengths between wild type, olv-3 and olv-7 (Figure 6A). PYE gene expression was elevated at – versus + Fe in olv-3 and olv-7 roots or shoots, but not to a different level than in wild type (Figure 6B, C). In the olv mutants, PYE target genes FRO3, NAS4 and ZIF1 were similarly expressed as in wild type in roots and shoots (Figure 6D, E, F, G, H, I), indicating that loss of OLV function had no effect on PYE action.

Finally, we studied whether OLV had any effect during later growth stages up to flowering. Interestingly, olv-3 and olv-7 had smaller rosettes in - Fe and + Fe-treated plants, while OLV OX7 and OX11 had no such phenotype (Figure 7A, B, C, D). To study the olv phenotype in more detail, we investigated Fe and Zn contents. Remarkably, Fe contents were lower after three days of -Fe in the old leaves, but not young leaves of wild type, olv-3 and olv-7 (Figure 7E, F). Only olv-7 had a statistically lower Fe content in the old leaves compared with wild type (Figure 7E). No differences were found for Zn (Figure 7G, H).

In summary, the only two phenotypes of OLV mutants detected were first, enhanced NAS4 expression in OX roots, and second, reduced plant rosette growth of olv mutants at an advanced growth stage.

Small regulatory proteins are known that act as effectors to influence TF activities. We report here that a small protein, called OLV, can interact with the bHLH transcription factor and Fe deficiency response regulator PYE. OLV alteration has an effect on expression of NAS4, a target gene of PYE in roots, and plant rosette growth. OLV orthologs are small proteins (mostly between 95 and 120 aa), that share the highly conserved TGIYY motif in their C-terminus. Due to their conservation in angiosperms, OLV may have a function associated with land plant evolution. The acquisition and mobilization of Fe are physiologically highly relevant in higher compared with lower land plants. Hence, small proteins may have become recruited to control Fe homeostasis regulators, such as bHLH TFs to fine-tune responses.

The interaction between PYE and OLV has been validated in independent approaches. First of all, protein interaction studies of three different kinds confirmed that OLV is able to bind PYE in different circumstances and cells. Second, these findings were corroborated by using deletion mutant constructs and mapping of the interaction face. Nevertheless, depending on the method, some of the deletion versions were or were not interacting. We therefore propose that the interaction interface is large between the two proteins, covering different domains and parts of the PYE and OLV protein, as was indicated in an AlphaFold structural prediction approach. Third, OLV and PYE interacted in the nucleus in BiFC and FRET-APB experiments. The two proteins also co-localized in the nucleus of tobacco epidermis cells when co-expressed. It is therefore very likely that the two proteins co-localize in nuclei of root differentiation zone epidermis and cortex where both OLV and PYE promoters are active. Fourth, the upregulated expression of one of the two promising PYE targets, NAS4, in roots upon OLV overexpression indicates that OLV may repress PYE function in the wild-type situation. This interpretation is consistent with the up-regulation of NAS4 upon pye-1 loss of function.

The conserved TGIYY motif is a characteristic of OLV. It was found to be relevant in protein interaction with PYE in Y2H and BiFC assays. Presence and absence of TGIYY did not affect the expression and localization of fluorescence fusion protein in tobacco. The presence of the TGIYY motif is evolutionarily more ancient than the evolving Fe deficiency bHLH regulatory cascade in angiosperms. Perhaps the TGIYY motif has a general role to allow protein interactions. Clearly, the TGIYY motif does not target the conserved EAR motif of the bHLH TF since there was no indication that the EAR motif is required for OLV interaction. Interestingly, the family of small IMA proteins which positively regulates the Fe deficiency response also has a characteristic conserved region in the C-terminus, while the N-terminus is variable (Grillet et al., 2018; Hirayama et al., 2018) . The IMA conserved region is needed for protein interaction with co-expressed E3 ligases of the BTS/BTSL-type (Li et al., 2021; Lichtblau et al., 2022) . Hence, the coexpression and protein interaction represent interesting parallels between the IMA-BTS/L and the OLVPYE systems. It can be proposed that Fe deficiency-induced small proteins harbor a variable region combined with a highly conserved C-terminal region for protein interaction with co-expressed target proteins and regulatory factors of the Fe deficiency cascade, important for the fine-tuning of individual responses.

OLV is mainly expressed in Fe-deficient seedling roots, while PYE is induced upon Fe deficiency in roots and shoots. In roots, OLV and PYE expression occurs in different regions, with the exception of an overlap in a small region in the outer tissue layers of the root differentiation zone. Despite of that, it has been reported that PYE is mobile across the root close to the root tip and has different functions in the tissue layers (Long et al., 2010; Muhammad et al., 2022) . Perhaps the protein interaction of PYE with OLV is physiologically relevant in the root differentiation zone close to the root tip. In this region, OLV may affect NAS4 gene expression, possibly through interacting with PYE. Interestingly, NAS4 promoter is targeted by PYE but also by ILR3, and the PYE and ILR3 TFs interact (Tissot et al., 2019) . Loss of function of ILR3 may also lead to an up-regulation of NAS4 (Tissot et al., 2019) . Perhaps OLV can affect the interaction between PYE and ILR3 or fine-tune the possible TF protein interaction complexes.

An interesting observation was that rosette growth was reduced in olv mutants compared with wild type. The reduced rosette growth is a consequence of the decreased leaf cell elongation in old leaves of olv plants, which can be considered the source leaves for Fe (Nguyen et al., 2022) . Perhaps, there is a long-distance signaling between the action of OLV in roots and the amount of Fe that is translocated to shoots to allow for leaf expansion in the old leaves.

We found that OLV is a novel prospective small protein interactor of PYE that may interact via a conserved previously non-characterized TGIYY motif. OLV had a mild effect on plant growth and Fe deficiency responses involving NAS4. Since NAS4 is specifically targeted by other TFs of the Fe deficiency response, OLV may control a particular sub-function of PYE in controlling NAS4. Hence, our work strengthens the roles that small regulatory proteins play in the fine-tuning of the Fe deficiency responses.

Future studies can address remaining open questions. For example, the functional relevance of the TGIYY motif for formation of the protein complex with PYE can be studied in more detail at the biochemical level. This may provide better evidence on the amino acids involved in the protein complex. Moreover, protein interaction can be further studied in plant roots, e.g. by co-expressing fusion proteins of OLV and PYE expressed from their respective promoters. Such cell-biological studies may indicate in which tissues and root zones both proteins are present and functional. Transcriptome studies of olv mutants and overexpression plants may further indicate subtle phenotypes. This may reveal whether there are significant alterations in the root Fe acquisition programs and leaf cell elongation process gene expression patterns. It will be further of interest to investigate long-distance signaling and how a rootexpressed gene affects leaf development.

Taken together, our work provides evidence for a novel PYE protein interactor and small regulatory protein affecting root Fe deficiency and plant growth responses. This work aligns with several work hypotheses that can be addressed in the future to understand the fine-tuning of Fe deficiencyinduced TF functions.

The Arabidopsis ecotype Col-0 (Columbia-0) was used as wild type (WT) and background for all transgenic lines constructed in this study. To generate triple HA3-tagged overexpression lines, full-length coding sequence (CDS) of PYE and OLV were amplified from cDNA of Fe deficient Arabidopsis WT roots and transferred via Gateway cloning into pDONR207 (Invitrogen) according to the manual (BP reaction, Thermo Fisher Scientific). After sequencing, the respective CDS was shuttled into the plant binary destination vector pAlligator2, via Gateway LR reaction (Thermo Fisher, Scientific). pAlligator2 allows ectopic overexpression of genes N-terminally tagged with a triple HA under the control of a double CaMV 35S promoter (Bensmihen et al., 2004) . Final constructs were sequenced and subsequently transferred into agrobacteria (Rhizobium radiobacter strain GV3101 (pMP90) (Koncz and Schell, 1986) . The Agrobacterium mediated floral dip method (Clough and Bent, 1998) was applied to generate stable transgenic Arabidopsis lines. Transformed seeds were selected based on seed-specific GFP expression and confirmed by genotyping PCR (Supplemental Table S1). Homozygous T3 plants were used for further analysis. To generate proOLV::GUS the promoter sequence of OLV (988 bp upstream of start codon) was amplified from Arabidopsis WT leaf gDNA (Supplemental Table S1), transferred into pDONR207 (Invitrogen) via Gateway BP reaction (Thermo Fisher, Scientific) and then sequenced. To generate the final vector, the promoter sequence was shuttled into the Gateway binary destination vector pGWB3 (Nakagawa et al., 2007) via LR reaction (Thermo Fisher, Scientific), followed by sequencing. Stable Arabidopsis lines were constructed as described above. The proPYE::GUS line was previously described by (Lichtblau et al., 2022) . To obtain olv loss of function mutants, the genome sequence was edited at the OLV locus by Crispr/Cas9 system according to published procedures (Hahn et al., 2017) . sgRNAs were designed using the CRISPR-P2.0 tool (Lei et al., 2014) , targeting the region of the beginning of the second exon up to the start of the TGIYY motif-encoding region. Forward and reverse guide RNA constructs were generated (attgTGAATTGATTTCCTAAGCAT and aaacATGCTTAGGAAATCAATTCA including restriction sites for later genotyping) and transferred under the control of the Arabidopsis U6-26S RNA polymerase III promotor into pFH6_new. Subsequently, the whole sgRNA cassette was transferred into pFH1 (pUB-CAS9) via Gibson cloning (Thermo Fisher Scientific). Following Rhizobacter transformation and floral dipping, as described above, hygromycine-resistant plant lines were selected in the T1 and multiplied up to the T3. Genomic PCR using primers OLV genotyping_F and OLV genotyping_R surrounding the genome-edited region combined with DNA sequencing (primers OLV sequencing_F and OLV sequencing_R) was used to select loss of function lines olv-3 and olv-7. Tobacco (Nicotiana benthamiana) plants were used for transient expression experiments.

Arabidopsis seeds were surface-sterilized and stratified for two days. Propagation and seed production was performed on soil in a climate chamber under long day conditions (16 h light, 8 h dark, 21°C). Three different growth systems were used for phenotypic, physiological, and histochemical analyses. For all growth systems Arabidopsis seedlings were grown on upright sterile plates containing modified half-strength Hoagland medium [1.5 mM Ca(NO3)2, 1.25 mM KNO3, 0.75 mM MgSO4, 0.5 mM KH2PO4, 50 µM KCL, 50 µM H3BO3, 10 µM MnSO4, 2 µM ZnSO4, 1.5 µM CuSO4, 0.075 µM (NH4)6MO7O24, 1% (w/v) sucrose, pH 5.8, containing 1.4% plant agar (Duchefa)] and supplemented with either sufficient Fe (50 µM FeNaEDTA, +Fe) or Fe deficient (0 µM FeNaEDTA). After stratification, the plates were transferred into plant growth chambers (CLF Plant Climatics) with long day conditions for 7 days with +Fe or -Fe (7 d system: whole seedlings were analyzed for root lengths). Plants grown in the 9 + 3 d system were grown under +Fe for 9 days and subsequently transferred on +Fe or -Fe for another three days. Roots and shoots were harvested separately for gene expression analysis. Tobacco plants were grown on soil for three to four weeks in a greenhouse facility under long day conditions. After tobacco leaf infiltration the plants were kept at room temperature under long day light conditions in the lab for two to three days until localization and protein interaction studies were performed.

Construction of OLV and PYE protein mutant versions

pDONR207:OLV was used as a template to generate different OLV truncated versions. For OLV-N (aa 1 to 55), the primer pair OLV_B1 fw/OLV +165 bp _B2 rev was used, and for OLV-C (aa 56 to 109) the primer pair OLV + 165 bp_B1 fw/OLV_B2 rev. To amplify the conserved motif (OLV-TGIYY, aa 71 to 87) the primer pair OLV + 210 bp_B2 fw/OLV + 261 bp_B2 rev was used. Deletion of TGIYY (OLV-TGIYY, aa 1-70 and 88-109) was generated by overlap-extension PCR. Two partially overlapping parts of OLV were amplified with the primer pair OLV_B1 fw/OLV rev and OLV fw/OLV_B2 rev (Supplemental Table S1). Both amplicons served as template in a second PCR to amplify OLV with the underlined outer primers. All primers (except OLV rev and OLV fw) carry B1 and B2 Gateway attachment sites. Subsequently all amplicons were transferred into pDONR207 (Invitrogen), sequenced, and shuttled into destination vectors. pAlligator2:PYE wasused as a template to generate different PYE truncated versions. For PYE-N (aa 1 to 120), the primer pair PYE_B1 fw/ PYE+360_attB2 rev was used, and for PYE-C (aa 121-240) the primer pair PYE+361_attB1 fw/PYE_B2 rev was used. To amplify bHLH domain (PYE-bHLH), aa 27 to 77) the primer pair PYE_attB1 fw +81bp/PYE_attB2 rev +231bp was used. Deletion of bHLH (PYE-bHLH, aa 1-26 and 78-240) was generated by overlap-extension PCR. Two partially overlapping parts of PYE were amplified with the primer pairs PYE_B1 fw/PYE_bHLH rev and PYE_bHLH fw/PYE_B2 rev. Both amplicons served as template in a second PCR to amplify PYE-bHLH with the underlined outer primers. Deletion of EAR motif (PYE-EAR, aa 1-131 and 139-240) was generated by overlap-extension PCR. Two partially overlapping parts of PYE were amplified with the primer pairs PYE_B1 fw/PYE_EAR rev and PYE_EAR fw/PYE_B2 rev. Both amplicons served as template in a second PCR to amplify PYE-EAR with the underlined outer primers. All primers (except PYE_bHLH rev, PYE_bHLH fw, PYE_EAR rev, PYE_EAR fw ) carry B1 and B2 Gateway attachment sites. Subsequently all amplicons were transferred into pDONR207 (Invitrogen), sequenced, and shuttled into destination vectors.

To study protein-protein interactions full-length OLV and mutant versions of OLV were tested with PYE in a targeted Y2H assay. All constructs were N-terminally fused to the GAL4-AD (vector: pACT2-GW), which acts as prey within the Y2H system (AD, activation domain). To generate the bait, all constructs were N-terminally fused to the GAL4-BD (vector: pGBKT7-GW) (kindly provided by Dr. Yves Jacob, Institut Pasteur, Paris, France) (BD, binding domain). The CDS of PYE and OLV were amplified from cDNA of Fe deficient Arabidopsis WT roots with primer pairs carrying B1 and B2 attachment sites (Supplemental Table S1) and transferred via Gateway cloning into pDONR207 (Invitrogen) according to the manual (BP reaction, Thermo Fisher Scientific). pDONR207 constructs were sequenced, the respective CDS shuttled into the destination vectors pACT2-GW/pGBKT7-GW via Gateway LR reaction (Thermo Fisher, Scientific) followed by additional sequencing. Bait and prey constructs were co-transformed into the yeast strain AH109 via the lithium acetate (LiAc)/SS carrier DNA/PEG method based on (Gietz and Schiestl, 2007) . Briefly, a 50 ml AH109-YPDA culture was grown up to OD600 = 0.5 and then made competent by the addition of 100 mM LiAc. 50 µl competent yeast cells were mixed with 33.3 % (w/v) PEG 4000, 0.1 M LiAc, 50 µg denatured Calf Thymus DNA (Invitrogen), 0.5-0.7 µg AD-plasmid, 0.5-0.7 µg BD-plasmid and sterile water to a final volume of 360 µl for each transformation event. Heat shock treatment was performed at 42°C for 20 min. Yeast cells were cultivated on minimal synthetic defined (SD) media (Clontech), lacking Leu (selection for pACT2-GW) and Trp (selection for pGBKT7-GW) for 2-3 days at 30°C, to select for positive double transformants. As negative controls, bait or prey were combined with empty BD or AD plasmids and used in Y2H assays. As positive control the combination of pGBT9.BS:CIPK23 and pGAD.GH:cAKT1 was used (Xu et al., 2006). To test for protein interactions, overnight liquid cultures of transformed AH109 were adjusted to OD600=1, and ten-fold serial dilutions down to 10-4 in sterile water were prepared. 10 µl of each suspension were spotted on SD agar plates lacking Leu, Trp and His and supplemented with 0.5 mM 3amino-1,2,4-triazole (SD-LWH + 3-AT, suppression of background growth, detection of interaction). In parallel 10 µl of the same serial dilutions were spotted on SD agar plates lacking Leu and Trp (SD-LW, positive growth and double transformation control). Plates were cultivated at 30°C for 7 d. Growth was documented by photographing the plates every second day. Final pictures were taken on day 7.

Subcellular protein localization studies were performed to analyse the localization of proteins. Therefore, fluorophore tagged proteins were transiently expressed in tobacco leaf epidermis cells. For Nterminal fusions the pDONR entry clones with PYE and OLV sequences were used for Gateway LR reaction (Gateway, Thermofisher, Scientific) to generate the destination vectors based on pH7WGY2 (N-terminal YFP) (Karimi et al., 2005). Final constructs were sequenced. For C-terminal fusions the CDS of PYE and OLV versions were amplified from cDNA of Fe deficient Arabidopsis WT roots with primer pairs PYE_B1 fw/PYEns_B2 rev and OLV_B1 fw/OLVns_B2 rev carrying B1 and B2 attachment sites without stop codon or from existing pDONR with N-terminal fusion (Supplemental Table S1) and transferred via Gateway cloning into pDONR207 (Invitrogen) according to the manual (Gateway, BP reaction, Thermo Fisher Scientific). After sequencing, the respective CDS was shuttled into the destination vectors pMDC83 (Cterminal GFP fusion) (Curtis and Grossniklaus, 2003) , as well as into the ß-estradiol-inducible pABindGFP/pABind-mCherry (C-terminal GFP and mCherry fusions) (Bleckmann et al., 2010) , via Gateway LR reaction (Thermo Fisher, Scientific). Constructs were sequenced and transformed into Agrobacteria strain GV3101 (pMP90). For tobacco leaf infiltration, an overnight culture of Agrobacteria, carrying one of the constructs, was centrifuged and the pellet re-suspended in AS medium (250 µM acetosyringone (in DMSO), 5 % (w/v) sucrose, 0.01 % (v/v) silwet, 0.01 % (w/v) glucose), according to Bleckmann et al., 2010. The suspension was adjusted to OD600=0.4 and infiltrated with a 1 ml syringe into the abaxial side of two tobacco leaves on two different plants. For co-localization experiments using GFP- and mCherry-tagged proteins, both Agrobacteria suspensions were mixed 1:1 to obtain a final OD600=0.4 for each. Transformed tobacco plants were kept in the lab for 48 to 72 h at RT under long day conditions (16 h light, 8 h dark). To induce the expression of pABind constructs, infiltrated tobacco leaves were sprayed abaxially with 20 µM ß-estradiol (in DMSO, supplemented with 0.1 % (v/v) Tween20) 16 h before imaging. To analyze transgene expression and protein localization, 0.5 cm leaf discs were punched out and imaged using alaser-scanning confocal microscope (LSM780, Zeiss). GFP and YFP were imaged at an excitation wavelength of 488 nm and emission wavelength of 491 to 533 nm. mCherry was imaged at an excitation wavelength of 561 nm and emission wavelength of 562 to 626 nm. Localization and co-localization experiments were performed in three independent experiments with two infiltrated leaves of two different plants.

To verify PYE interactions with OLV in planta the BiFC 2in1 vector system was applied (Grefen and Blatt, 2012) . The CDS of PYE and OLV versions was amplified from pDONR207 constructs (see “Plant Material”) or cDNA of Fe deficient Arabidopsis WT roots using primer pairs carrying B3, B2 and primer pairs carrying B1, B4 attachment sites (Supplemental Table S1). Via BP reaction (Gateway, Thermo Fisher) all amplicons carrying B3, B2 attachment sites were transferred into pDONR221-P3P2 (Invitrogen, for nYFP fusion) and amplicons carrying B1, B4 attachment sites into pDONR221-P1P4 (Invitrogen, for cYFP fusion). Constructs were sequenced. OLV-FL or one of the truncated OLV versions were shuttled simultaneously with PYE into the destination vector pBiFC-2in1-NN (Grefen and Blatt, 2012) (N-terminal nYFP and cYFP fusions) via multisite LR reaction (Gateway, Thermo Fisher). Hereby pBiFC-2in1-NN:OLV-FL-PYE and pBiFC2in1-NN:PYE-OLV-FL (additionally all truncated OLV versions were cloned into pBiFC-2in1-NN combined with PYE as described for OLV-FL) were created and sequenced. An internal mRFP, served as transformation control. As negative controls for PYE and OLV proteins that do not interact with either partners were selected. For PYE ILR3 was used as negative control. As negative control of OLV bHLH39 was chosen. Therefore, pBiFC-2in1-NN:PYE-bHLH39, pBiFC-2in1-NN:bHLH39-PYE, pBiFC-2in1-NN:OLV-ILR3 and pBiFC-2in1-NN:ILR3-OLV were cloned as described above. All constructs were transformed into Agrobacteria strain GV3101 (pMP90) and used for tobacco leaf infiltration as described in “Subcellular (co-) localization”. After 48 to 72 h, cells which were mRFP positive were analysed for YFP signals using the Axio Imager M2 (Zeiss) with ApoTome. mRFP was imaged at an excitation wavelength of 545 nm and emission wavelength of 570 to 640 nm, YFP was imaged at an excitation wavelength of 524 nm and emission wavelength of 520 to 550 nm. Three independent BiFC experiments were performed, using two leaves for each construct.

To investigate protein-protein interaction between PYE and OLV via FRET-APB, the full-length CDS of PYE and OLV were cloned into pABind-GFP, pABind-mCherry and pABind-GFP-mCherry (pABindFRET) (Bleckmann et al., 2010) as described above. For FRET-APB experiments, tobacco leaves were infiltrated with Agrobacteria carrying pABind-GFP:PYE and pABind-mCherry:OLV, or vice versa, to determine the strength of the protein-interaction ability. GFP-tagged proteins with donor only (pABind-GFP:PYE or OLV) served as negative control, the corresponding protein fused to a double tag of GFP-mCherry (pABindFRET:PYE or OLV) as positive control. To induce gene expression, infiltrated tobacco leaves were sprayed with 20 µM ß-estradiol 24 h after infiltration. The experiment was performed 20 h after ß-estradiol treatment.

FRET-APB measurements were taken with a laser-scanning confocal microscope (LSM 780, Zeiss) and controlled by the ZEN2 Black Edition software (Zeiss). For both fluorophores the fluorescence intensity was determined in the nucleus, with 20 frames of a 128 x 128 pixel format with a pixel dwell time of 2.55 µs. mCherry was photobleached after the 5th frame, using 100 % laser power at 561 nm and 80 iterations. The FRET efficiency (FRET E) was calculated in percent of the relative GFP intensity increase after mCherry acceptor photobleaching (Gratz et al., 2019) . Two independent experiments analyzing at least 10 nuclei with equal expression of both fluorophores were performed.

A BLAST search of Arabidopsis OLV-FL aa as well as the TGIYY motif sequence was performed in every order of the angiosperms, selected lower plants and other non-plant organisms (Cole et al., 2019) using NCBI blastp. The protein sequence of the hit with the highest maximum score of each order was reblasted in another BLAST analysis against the Arabidopsis TAIR10 protein sequence collection, applying TAIR BLAST 2.2.8 for validation. Multiple sequence alignments of all members with highest maximum score of each order were performed using the Clustal Omega algorithm (Sievers et al., 2011) and visualized with Jalview 2.10.4 (Waterhouse et al., 2009) .

ProPYE:GUS transgenic plants have previously been reported (Lichtblau et al., 2022) . ProOLV:GUS lines have been generated as described above. Two independent proOLV:GUS and proPYE:GUS lines were chosen and propagated to T2 or T3 for further analysis. Plants were grown in the 6 d system on +Fe and Fe, in two biological replicates, and assayed for histochemical GUS activity. Four to six seedlings of each line were incubated in 1 ml GUS staining solution containing [50 mM Na2HPO4, 50 mM NaH2PO4 pH 7.2, 2 mM K4[Fe(CN)6]Fe2+, 2 mM K3[Fe(CN)6]Fe3+, 0.2 % Triton-X, 2 mM 5-bromo-4-chloro-3-indoyl-b-Dglucuronic acid (X-Gluc)] (Jefferson et al., 1987) for 15 min to four hours at 37°C in the dark, until blue staining was observed. Afterwards leaf chlorophyll was removed by incubation in 70% EtOH for 24 h (EtOH was exchanged every few hours). Whole seedlings were imaged with the Axio Imager M2 (Zeiss, 10x objective magnification). Single images were assembled by using the stitching function of the ZEN 2 BLUE Edition (Zeiss).

Plants were grown on Hoagland agar plates (see “Plant Growth Conditions”) for 7 d on + Fe and Fe. Seedlings were imaged on day 6. Primary root length of individual seedlings was measured using the JMicroVision software (Version 1.2.7, http://www.jmicrovision.com). Root length was measured in two independently grown sets of plants with 45 to 60 plants for each line and condition.

Gene expression analysis was performed as described (Ngigi and Bauer, 2023) . In brief, total RNA was either extracted from whole seedlings grown in the 6 d system (n= 60-70 plants per replicate) or from roots/shoots of plants grown in the 9 + 3 d system (n=20-25 roots/shoots per replicate), using the peqGOLD Plant RNA KIT (PeqLab). Reverse transcription using oligo(dt) primer and the RevertAid firststrand cDNA synthesis kit (Thermo Fisher, Scientific) was performed to obtain cDNA. RT-qPCR was carried out on the SFX96 TouchTM Real Time Detection System (Bio-Rad) with the iTagTM Universal SYBR Green Supermix (Bio-Rad) according to the manual. The Bio-Rad SFX ManagerTM software (version 3.1) was applied to process the data. Absolute gene expression values were calculated by gene specific mass standard curve analysis. Data was normalized to the Arabidopsis elongation factor EF1B. All primer pairs for this study are listed in Supplemental Table S1. The experiment was performed with at least three biological and two technical replicates.

For statistical analysis a one-way analysis of variance (ANOVA) and a Tukey´s post-hoc test, which allow the comparison of more than two groups, were performed. Null hypothesis was rejected for p-values smaller than 0.05. Different letters indicate significant differences (p<0.05).

Total proteins were extracted from ground plant material of either whole Arabidopsis seedlings grown in the 10 d system (Arabidopsis seedlings: n=40) or from tobacco leaves. Protein extraction, SDSPAGE and immunodetection was performed as previously described in (Le et al., 2016). In summary, frozen plant material was homogenized using the Precellys 24 (Peqlab Life Science, VWR) and proteins were extracted with 2x SDG buffer (62 mM Tris-HCL pH 8.6, 2.5 % (w/v) SDS, 2 % (w/v) dithiothreitol, 10 % (w/v) glycerol, 0.002 % (w/v) bromphenol blue). Samples containing equal amounts of protein were separated on 12 % (w/v) SDS polyacrylamide gels via electrophoresis, followed by the protein transfer to a Protran nitrocellulose membrane (GE Healthcare). To control for equal loading of whole protein, proteins on the membrane were stained using PonceauS (0.2 % (w/v) PonceauS, 3 % (w/v) trichloroacetic acid, 3 % (w/v) sulfosalicylic acid). To detect HA-tagged PYE or OLV protein the membranes were blocked in 5 % (v/w) milk solution (Roth) in 1x TBST (20 mM Tris-HCL pH 7.4, 180 mM NaCl, 0.1 mM (v/v) Tween20) for 30 min to avoid nonspecific antibody binding. Afterwards, the membranes were incubated with anti-HA-peroxidase high-affinity monoclonal rat antibody (clone 3F10; Roche) 1:1000 diluted in 2.5 % (w/v) milk-TBST solution, followed by three times washing in 1x TBST for 15 min each. To detect chemiluminescence signals of HAtagged proteins the FluorChemQ System for quantitative western blot imaging (ProteinSimple) was applied and images were processed by the AlphaView® software (version 3.4.0.0, ProteinSimple).

Sequence data of PYE, OLV, AKT1 and CIPK23 can be found in the TAIR library with the following accession numbers: AKT1 (AT2G26650), BHLH39 (AT3G56980), CIPK23 (AT1G30270), FRO3 (AT1G23020), ILR3 (AT5G54680), NAS4 (AT1G56430), PYE (AT3G47640), OLV (AT1G73120), OPT3 (AT4G16370), ZIF1 (AT5G13740).

Figure S1. Co-expression of PYE and OLV genes Figure S2: OLV full-length protein sequence alignments

Figure S4: Localization of various YFP-OLV fusions.

Figure S5: OLV overexpression and olv mutant lines.

Figure S3: Multiple sequence alignment of the TGIYY motifs from OLV orthologs of angiosperms and of Figure S6: Overexpression of OLV in the background of pye-1 did not alter the pye-1 phenotype.

We thank Gintaute Matthäi, Monique Eutebach, and Elke Wieneke for technical support. We thank Christopher Endres for technical help with A. thaliana root cross sections and Florian Hahn for guidance of CRISPR/Cas9 cloning. The authors thank Christopher Grefen, Bochum, and Andreas Weber, Düsseldorf, for providing plasmids. D.L., K.T., and B.S. were members of the international graduate school iGRADPlant/NEXTplant, Düsseldorf. D.B. is member of the international graduate school NEXTplant, Düsseldorf. This work was funded by Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under GRK F020512056 (NEXTplant) and Germany´s Excellence Strategy – EXC-2048/1 – project ID 390686111.

Funding for instrumentation: Zeiss LSM780 + 4-channel FLIM extension (Picoquant): DFG- INST 208/551-1