White lupin seeds (Lupinus albus L. cv. Amiga; Südwestdeutsche Saatzucht, Rastatt, Germany) were soaked for 24 h in aerated water and germinated on a plastic net placed at the surface of an aerated 0.5 mM CaSO₄ solution in a growth chamber at 25°C in the dark. Thereafter, 7-day-old seedlings were transferred to a hydroponic system, containing a P-free nutrient solution (μM): 5000 Ca(NO₃)₂, 1,250 MgSO₄, 1750 K₂SO₄, 250 KCl, 20 Fe(III)EDTA, 25 H₃BO₄, 1.25 MnSO₄, 1.5 ZnSO₄, 0.5 CuSO₄, 0.025 (NH₄)₆Mo₇O₄. Phosphorus-deficient plants were grown on P-free nutrient solution, while 0.25 mM KH₂PO₄ were added to nutrient solution for P-sufficient condition. Plants were grown under controlled conditions for 4 weeks (day/night photoperiod, 16/8 h; radiation, 220 μE m⁻² s⁻¹; day/night temperature, 25/20°C; relative humidity, 70–80%).

In P-deficiency, lupin plants modify the root architecture developing particularly root structures, called cluster roots or proteoid roots. In order to differentiate the developmental stages of root clusters (Supplementary Figure S1), the root system was immerged in a pH-indicator solution (0.04% w/v bromocresol purple). Depending on their morphology and capability to acidify the solution, at the end of the growing period different regions of cluster root were sampled from P-deficient plants (juvenile, immature, mature, senescent), as described by Massonneau et al. (2001). Root apices and cluster root parts were sampled from a pool of 12–16 P-sufficient and P-deficient plants. The samples were immediately frozen in liquid nitrogen for the RNA extraction or were rinsed twice in 0.5 mM CaSO₄ solution and the root exudates were collected for 1 hour in 0.5 mM CaSO₄ 10 mM 2-[N-Morpholino] ethanesulfonic acid (MES)-KOH pH 6.0 at a ratio 1:10 W/V. The samples were conserved at −80°C until processing. Six independent experiments were performed.

To investigate the plant response to N-deficiency, white lupin seeds were grown for 2 weeks on watered Whatman paper in Petri dishes and then transferred to the magenta boxes containing PFR N-free solution described by Strozycki et al. (2003). For N-sufficient conditions KNO₃ (0.1 mM), NH₄H₂PO₄ (5 mM), Ca(NO₃)₂ (2.46 mM) were added to the PFR medium. Plants were grown for further 2 weeks under controlled conditions (day/night photoperiod, 16/8 h; radiation, 220 μE m⁻²s⁻¹; day/night temperature, 23/20°C; relative humidity, 70–80%). The samples were collected at 7 and 14 days after the transfer in N-deficient condition. Three independent experiments were performed; data shown are from the third experiment.

White lupin seedlings were transformed using Agrobacterium rhizogenes ARqua1 strain (Quandt et al., 1993) carrying binary vector pRedRoot::LaMATE2 RNAi or empty vector (EV) pRedRoot. After germination 5 mm root tips were removed from radicles. The sectioned surface was coated with A. rhizogenes and seedlings were placed on solid Fahraeus medium supplemented with kanamycin (15 mgL⁻¹).

Nitrogen-deficient plants were grown and analyzed as described above. For P-deficiency, after 3 weeks post germination, plants were transferred to P-free nutrient solution (as described above). After 2 weeks, white lupin plants were moved in hydroponic solutions in P-free nutrient solution and grown for 3 additional weeks. LaMATE2 expression analyses were performed, and root release was collected as described previously on the fully developed cluster root (immature, mature). Six biological replicates were performed for each sample.

Three-week-old lupin composite plants were transferred to the pots filled with sterile perlite (0.75l). Plants were nourished with PFR N-free solution. After 2 weeks of N-deficiency, plants were inoculated with B. japonicum (strain UPP 133 (Stępkowski et al., 2011)) and grown for further 2 weeks under controlled conditions (day/night photoperiod, 16/8h; radiation, 220 μE m⁻²s⁻¹; day/night temperature, 23/20°C; relative humidity, 70–80%). For complementation experiments 1h before inoculation with B. japonicum plants were additionally supplemented with 1 μM genistein solution. Fourteen-day post-inoculation, plants were removed from pots, nodules were counted for each single root, and roots were collected for each plant. Afterwards all roots of single plant were grounded and divided into aliquots dedicated for gene expression and metabolomic analysis. Only plants revealing expression of marker gene encoding fluorescent protein DsRed were considered during analysis. Three independent transformations were used with sample sizes 6, 8 and 21.

The vivianite suspension (Fe₃(PO₄)₂^.8H₂O) was prepared by mixing H₃PO₄ with FeSO₄·as described by Eynard et al. (1992). A solution of 0.035 M H₃³²PO₄ and 0.05 M FeSO₄·7H₂O was brought to pH 6.0 with 0.05M KOH under stirring. A bluish suspension of vivianite was obtained, after precipitation via centrifugation (8,000 g, 10 min), the precipitate was washed (by 3 successive centrifugations and decantations) with deionized water. Mobilization of PO₄²⁻ from Vivianite (Fe₃(PO₄)₂^.8H₂O) by exudates after purification with the SPE methods, in order to remove carboxylates, from LaMATE2 RNAi or empty vector (EV) pRedRoot transformed P-deficient roots was performed as follows: 1ml of a suspension containing ³²P-vivianite (70 μmol PO₄, specific activity 60 KBq μmol⁻¹ P) was transferred into a dialysis tube (ZelluTrans/Roth 6.0, ∅ 16 mm, exclusion limit of 4÷6kDa, ROTH) and mixed with 5 ml of mobilization solution [0.5 mM CaSO₄, 20 mM Mes-KOH (pH 6.0), 50 μl methanol]. Thereafter, the dialysis tube was transferred into continuously mixed 34ml of mobilization solution. At the beginning of the experiment, exudates corresponding to the amount release for 1 h by 1 gram of juvenile cluster root tissues (Figure 1) were added to the external solution. After 15, 30, 45 and 60 min, samples from the solution outside the dialysis tube were collected and the amount of ³²P was measured by liquid scintillation counting. The ³²P mobilization was estimated from the difference between the ³²P concentration measured in the presence of the mock and in the absence of exudates and was expressed as μmol P h⁻¹.

Frozen root tissue (500 mg) was grounded and extracted with 80% methanol. Root exudates were extracted from the medium by SPE (Solid Phase Extraction) method using octadecylosilane matrix and methanol according to Staszków et al. (2011). Luteolin was used as an internal standard. Samples were analyzed by liquid chromatography-electrospray ionization tandem mass spectrometry (LC/ESI/MS) using a Waters UPLC coupled with Bruker micrOTOF-Q mass spectrometer. The analysis was performed in a gradient mobile phase consisting of 0.5% formic acid (v/v) in water (A) and 0.5% formic acid (v/v) in acetonitrile (B). The m/z range of the recorded spectra was 50–1,000. Analyses were performed in the ion-positive mode.

RNA extractions were performed using the InviTrap Spin Plant RNA Mini Kit (Stratec Molecular, Berlin, Germany) following the manufacturer’s instructions and contaminant genomic DNA was removed using 10 U of DNase I (GE Healthcare, Munich, Germany). The quantity and the quality of RNA were checked using a spectrophotometer, followed by a migration in a 1% agarose gel. One microgram of total RNA for each sample was retro-transcribed using 1 pmol Oligo d(T)23 (Sigma Aldrich, Saint Louis, United States) and 10 UM-MulV RNase H (Finnzymes, Helsinki, Finland) following manufacturers’ instruction.

The partial sequence was isolated via a cDNA-AFLP approach starting from RNA extracted from juvenile cluster roots which were compared to mature and senescent cluster roots; for details, see Massonneau et al. (2001). The full Open Reading Frame (ORF) of LaMATE2 (LaMATE2_ORF) was isolated from the cDNA of juvenile cluster root tissues from P-deficient plants using the 5'/3' RACE Kit (2nd generation, Roche Diagnostics S.p.a., Monza, Italy) following the manufacturer’s instructions. LaMATE2_ORF sequence was cloned in pGEM-T easy vector (for primers, see Supplementary Table S1, Promega Italia Srl, Milan, Italy) and deposited in the NCBI database (KY464927).

The reaction was performed by adding 0.1 μl of cDNA to RT complete reaction mix, Fluocycle^™ sybr green (20-μl final volume, Euroclone, Pero, Italy). Specific primers were designed for the target, the two closest homologues (www.whitelupin.fr; Hufnagel et al., 2020) and the housekeeping gene using Primer3 software (Koressaar and Remm, 2007; Untergasser et al., 2012) and were synthesized by Sigma Aldrich (Supplementary Table S1). Gene expression analyses were performed using CFX96^™ Real-Time System (C1000TM Thermal Cycler, BioRad) and CFX Manager^™ Software (v 2.0, BioRad). The two closest homologues (Lalb_Chr06g0165721, Lalb_Chr02g0145611) have a low expression in the root tissues and their expression levels are shown in Supplementary Figure S5. Data were normalized in respect to the transcript level of the housekeeping gene (Ubiquitin gene, LaUBI) using the 2^-ΔΔCT method (Livak and Schmittgen, 2001). The efficiency of amplification was calculated using R program (version 2.9.0)¹ with the qPCR package (version 1.1–8), following the authors’ indications (Ritz and Spiess, 2008). Six biological replicates were performed for each sample with 3 technical replicates.

The construction of the yeast expression vector pNEV (Sauer and Stolz, 1994) containing LaMATE2_ORF was performed by amplifying LaMATE2_ORF sequence with primers having both NotI-restriction sites at the 5’end. The cloning of the PCR product (LaMATE2_ORF) was performed into the NotI site of pNEV (pNEV::LaMATE2_ORF), the orientation was verified by sequencing. The transformation of competent yeast cells (S. cerevisiae YPH499 strain) was performed following a standard procedure (Gietz and Woods, 2002) and transformants were selected on synthetic dextrose minimal medium lacking uracil (SD-Ura medium; Burke et al., 2000).

The yeast cells, strain YPH499, were transformed with the pNEV empty vector or pNEV::LaMATE2_ORF and selected on liquid-SD Ura-medium. Thereafter, cells were incubated in YPD medium for 30 min, collected by centrifugation, and digested with lyticase (1,000 Ug⁻¹ fresh weight cells; Sigma Aldrich), and subsequently microsomal vesicles were isolated as described by Klein et al. (2002). Transport assays were performed to study the genistein transport using the rapid filtration technique with nitrocellulose filters (0.45 μM pore size; Millipore, Millipore Co., Bedford, United States). The transport experiment was carried out in the presence of isolated vesicles, transport buffer (0.4 M glycerol, 0.1 M KCl, 1 mM DTT, 1 mM EDTA, 5 mM ATP, 10 mM MgCl₂, 10 mM creatine phosphate, 0.1 mgml⁻¹ creatine kinase, 20 mM Tris-MES pH 7.4) and 5 μM of labelled ³H-genistein (American Radiolabeled Chemicals, Saint Louis, United States; 1850 Bq filter⁻¹). Only for kinetic experiments, the ³H-genistein concentration ranged from 5 μM up to 100 μM (5, 7, 10, 15, 33, 50 and 100 μM ³H-genistein) and the incubation time was 30 s. The concentration-dependency of ³H-genistein uptake were calculated by subtracting uptake rates recorded in the empty-vector vesicles. The concentration-dependency of ³H-genistein uptake were calculated between 7 and 100 μM by subtracting uptake rates recorded in the empty-vector vesicles using the Hanes–Woolf plot.

To test the dependence of genistein transport on a proton gradient, 25 mM NH₄Cl was included in the transport buffer assay and the incubation time ranging between 15 and 120 s (15, 30, 60 and 120 s). Moreover, the capability of LaMATE2 to mediate the uptake of different flavonoids was tested by UPLC under the same experimental conditions reported above (5 μM of flavonoid: genistein, genistin, hydroxygenistein, biochanin A, daidzein, or kaempferol; the incubation time was 30 s). The mixture was loaded on a pre-wetted filter and removed by suction at the end of incubation time. The membranes were rapidly washed twice with 2ml of ice-cold transport buffer.

The radioactive measurements were determined with a beta-counter (Tri-Carb 1900CA, Packard, Downers Grove, United States). As standards, solutions with known amounts of ³H-genistein were used. Vesicle protein content was quantified with BioRad Protein Assay Dye Reagent (BioRad, Hercules, CA, United States), and the data are shown as pmol ³H-genistein μg⁻¹ protein. Data are shown as net pmol ³H-genistein μg⁻¹ protein after removing background, i.e., unspecific adsorption of genistein onto empty-vector yeast membrane. Three independent transformations were performed for each sample.

For transient expression of LaMATE2_ORF in Arabidopsis protoplasts, the plasmid harboring the sequence for the Green Fluorescent Protein (GFP) was fused at the C-terminus of LaMATE2_ORF inside the pUC18-Sp-GFP6 vector (Komarova et al., 2008) using NheI and SphI restriction sites via PCR amplification. A plasmid harboring the sequence for mCherry-fluorescent protein was fused with AtPIP2a, a gene coding for an aquaporin used as a plasma membrane marker (Nelson et al., 2007). Arabidopsis protoplasts were co-transformed with both constructs, LaMATE2_ORF-GFP and AtPIP2a-mCherry, using the polyethylene glycol method (Jin et al., 2001). Protoplasts were examined with a TCS SP5 confocal microscope (Leica Microsystems, Wetzlar, Germany), excited with an argon laser at 458 nm for GFP and 540–552 nm for mCherry (for GFP: excitation BP458, beamsplitter FT500, emission BP 492–511 nm; for mCherry: excitation BP 540–552, beamsplitter FT560, emission BP 575–640).

LaMATE2 silencing was adapted from Uhde-Stone et al. (2005) using binary transformation vectors, pRNAi and pRedRoot (Limpens et al., 2004). The target region (350 bp long, covering the 5' end of LaMATE2_ORF) was first amplified with PCR, and subsequently cloned into pRNAi vector between the restriction sites NcoI–SwaI and BamHI–SpeI. The cloned sequence (LaMATE2 RNAi) was regulated by a double CaMV35s promoter and OCS-3' terminator. Using the KpnI–PacI restriction sites, the LaMATE2 RNAi cassette from the previously produced in pRNAi vector was transferred into the pRedRoot binary vector (pRedRoot::LaMATE2 RNAi).

Phylogenetic analyses were conducted using MEGA software, version 6 (Tamura et al., 2013). The tree was constructed by aligning the protein sequences by Clustal-W and the evolutionary history was inferred using the Neighbor-Joining method. The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test (1,000 replicates) is shown in Supplementary Figure S1 next to the branches. The tree is drawn to scale, with branch lengths in the same units as those of the evolutionary distances used to infer the phylogenetic tree. The evolutionary distances were computed using the Poisson correction method and are in the units of the number of amino acid substitutions per site.

The topology prediction of LaMATE2 was performed using the PROTTER program (Omasits et al., 2014), while the alignment of LaMATE2 protein sequence with orthologous MATE-protein sequences was generated by Clustal-WS using Jalview software version 2 (Waterhouse et al., 2009).

Statistical significance was determined by one-way analysis of variances (ANOVAs) using Holm–Sidak test, p<0.05 or 0.01. Statistical analyses were calculated using SigmaPlot (Systat Software Inc., San Jose, CA, United States).

Former work from our laboratories showed that flavonoids are released mainly from the so-called juvenile, and immature (young states) cluster roots and that genistein and its derivative exudation was induced by phosphate (P) deficiency (Weisskopf et al., 2006). Furthermore, in 2001, we published research on genes differentially expressed amongst different stages of cluster roots in P-deficient white lupin (Massonneau et al., 2001). Within these genes, we identified two LaMATE-type transporters, LaMATE1 was predominantly expressed in mature (Uhde-Stone et al., 2005), citrate excreting cluster roots, while the second (LaMATE2) was predominantly present in juvenile cluster roots.

To identify a flavonoid exporter possibly playing an important role in the establishment of nodules in Fabaceae plants, we looked first whether a link between the expression of LaMATE2 and release of flavonoids exists. Under P-deficiency, the LaMATE2 was predominantly expressed in the root apex and in the juvenile stage of cluster roots (Figure 1A). Nitrogen (N) deficiency, on the other hand, induced LaMATE2 expression mainly in apex and nodule (Figure 2A). Analyses of the root exudates of white lupin revealed that genistein release was strongly induced by P deficiency, particularly in young cluster root tissues (Figure 1B), confirming previous results (Weisskopf et al., 2006). In order to see whether a similar behavior could be observed under N deficiency, plants were grown under N-deficient condition. In this case, cluster root formation was not induced. Hence, we compared the whole roots of plants grown in the presence or absence of N. Nevertheless, we could observe that under N deficiency, genistein release was only slightly higher after 1 and became significantly higher after 2 weeks (+123%; Figure 2B).

As already shown for other MATE transporters, LaMATE2 contains 12-transmembrane helical domains (Supplementary Figure S2) (He et al., 2010). Indeed, LaMATE2 exhibits a good homology to other plant MATE transporters (Figure 3; Supplementary Figure S2), including the functionally characterized vacuolar flavonoid transporters MtMATE1, MtMATE2 and AtTT12 (Marinova et al., 2007; Zhao and Dixon, 2010; Zhao et al., 2011). Up to date, only few MATE transporters have been characterized in roots and most of them mediate either the efflux of citrate, such as AtFRD3, HvAACT1, SbMATE1 (Durrett et al., 2007; Furukawa et al., 2007; Magalhaes et al., 2007) or act as vacuolar flavonoid transporters in Arabidopsis, Medicago or grapevine (for a review see Zhao and Dixon (2010) and Zhao (2015)). Interestingly all citrate transporters contain a large cytosolic loop. In white lupin roots and in agreement with our previous results (Massonneau et al., 2001), the expression of another MATE transporter, LaMATE1, a homolog of the citrate transporter AtFRD3, is highest at the mature stage where a burst of citrate exudation occurs (Uhde-Stone et al., 2005; Wang et al., 2014). Also, this transporter contains a large cytosolic loop.

At the amino acid level, LaMATE2 exhibits a high similarity to a cluster of MATE proteins, which includes a heterogeneous group of transporters (different substrates, subcellular localization and physiological role). Among the characterized members of this cluster, NtJAT1 and AtDTX1 are the closest homologs of LaMATE2, 55 and 50% of identities, respectively (Figure 3). Both are localized at the plasma membrane (PM) of root cells and function as efflux carriers for plant-derived alkaloids and other toxic compounds (Li et al., 2002; Morita et al., 2009).

Several ABC transporters and MATE proteins mediating the export of compounds into the soil have been described (Weston et al., 2012; Baetz and Martinoia, 2014). However, none of them was shown so far to be responsible for the root release of flavonoids and isoflavonoids. Our data showing a connection between flavonoid exudation and LaMATE2 expression together with the fact that LaMATE2 fused with GFP co-localizes with the PM marker AtPIP2a-mCherry (Figure 4) prompted us to investigate whether LaMATE2 acts as a flavonoid transporter.

Genistein is one of the major released isoflavonoids of Fabaceae inducing NOD genes in Rhizobium bacteria, attracting them and inducing nodule formation (Zhang and Smith, 1995; Lang et al., 2008; Liu and Murray, 2016). Therefore, to assess whether the LaMATE2 might affect genistein release and nodule formation, we have used RNA-dependent gene silencing (LaMATE2-RNAi) in roots of white lupin plants. The effectiveness of the silencing was confirmed by a significant reduction (approximatively −80 and 65% of LaMATE2 expression in LaMATE2-RNAi transformants grown in N and P deficiency, respectively (Figures 5A,G). Analysis of these roots revealed that the release of genistein was strongly reduced in these transformants (Figures 5B,H) and concomitantly isoflavonoids, the uppermost as glycosides, were accumulating in the cell content of silenced roots, which can be regarded as a direct consequence of the reduced genistein exudation (Figures 5D-F). To investigate whether the impaired genistein exudation influences nodulation, we used the LaMATE2-RNAi plants to compare the nodule number of silenced and control plants. Our results highlight the importance of LaMATE2-dependent isoflavonoid release in the early step of nodulation, since LaMATE2 silencing leads to a highly significant reduction (a. -80%) of nodules in LaMATE2-RNAi compared to empty-vector transformed roots (Figures 5C; 6B, Supplementary Figure S4). Interestingly exogenous application of genistein onto the LaMATE2-RNAi roots partially restored the nodulation efficiency (Figures 6A,B). Similarly, LaMATE2 expression, genistein release patterns and capability of root exudates to mobilize P were measured under P-deficient conditions in silenced plants (Figures 5G-I). Cluster roots of LaMATE2-RNAi transformants grown under P deficiency released 85% less genistein along with the lower level of LaMATE2 expression (a. −60%). The SPE-purified exudates from those roots also exhibited a limited (a. −66%) mobilization capability from a poorly soluble P source (vivianite). These results suggest that LaMATE2 is the major if not the sole genistein exporter from roots of white lupin plants and in P-deficient condition it contributes to the P solubilization process.

The results showing that genistein export is dependent on the presence of LaMATE2 indicates that within the root this transporter is at least partially localized in the cortex and/or in the epidermis. Attempts in our laboratory to get a more detailed picture on its localization, either using in-situ hybridization or a GUS-promoter construct unfortunately failed. However, it should be mentioned that interestingly, excretion of strigolactones into the soil was independent, whether the transporter was localized specifically in hypodermal passage cells as in Petunia or in the whole cortex as in Medicago (Kretzschmar et al., 2012; Banasiak et al., 2020).

To demonstrate that the reduced genistein release of LaMATE2 silenced plants depends directly on the activity of LaMATE2, we expressed it heterologously in Saccharomyces cerevisiae and performed transport assays using isolated membrane vesicles (Figure 7). Indeed, using this system, we could observe that LaMATE2 mediates an efficient transport of genistein which was characterized by a saturable kinetic exhibiting an apparent Km of 16.2 μM (Figure 7B). This Km is in the same order of magnitude as determined for the vacuolar flavonoid transporters characterized so far (e.g., AtTT12 and MtMATE1: Km of 50 and 36 μM for epicatechin 3'-O-glucoside, respectively; MtMATE2: Km of 88 μM for cyanidin 3-O-glucoside) (Zhao and Dixon, 2010; Zhao et al., 2011). In plasmalemma vesicles isolated from soybean roots, genistein transport exhibited a Km of 160 μM and was strongly dependent on the presence of ATP (Sugiyama et al., 2007).

Up to date, certain MATEs have been shown to exhibit a broad range of substrate specificity (Takanashi et al., 2014), while LaMATE2 displays a strong specificity for the substrate genistein (Figure 7C). Besides genistein, LaMATE2 was able to transport only daidzein at a low rate, but not other tested flavonoids such as genistin (a glycosylated form of genistein), hydroxygenistein, biochanin A (a methylated form of genistein) or kaempferol. Previous observations in soybean showed that the presence of daidzein in the external media strongly limited genistein transport in plasma membrane vesicles of root cells, indicating a possible competition between those molecules (Sugiyama et al., 2007).

A feature of MATE proteins is to couple the substrate transport to an electrochemical gradient, working as H⁺ or Na⁺/substrate antiporters (He et al., 2010; Lu et al., 2013; Tanaka et al., 2013). Our results showed that ATP-dependent ³H-genistein accumulation in vesicles deriving from LaMATE2_ORF-expressing yeast was strongly reduced in the presence of ammonium chloride, which dissipates the transmembrane proton gradient (Marinova et al., 2007; Zhao and Dixon, 2010) (Figure 7A), indicating that LaMATE2 acts as a substrate-proton co-transporter, similarly to AtTT12 (Marinova et al., 2007), AtDTX1 (Li et al., 2002) and NtJAT1 (Morita et al., 2009).