Relationships between carboxylate-based nutrient-acquisition strategies, phosphorus-nutritional status and rare earth element accumulation in plants

We explored how phosphorus (P) availability influences accumulation of rare earth elements (REE) in plant species with different P-acquisition strategies beyond the commonly explored REE-phosphate precipitation. Two P-efficient carboxylate-releasing lupin species (Lupinus albus, and L. cosentinii) and four species with less carboxylate release under P-deficiency (Triticum aestivum, Brassica napus, Pisum sativum, Cicer arietinum), were cultivated with a split-root system on two sand types. Phosphorus availability was controlled on one root side by watering the plants with 100 μM P or 0 μM P solutions. Carboxylate release and changes in pH were measured on both sides. Concentrations of nutrients, cadmium (Cd), aluminum (Al), light REE (LREE: La–Eu), and heavy REE (HREE: Gd–Lu, including Y) in roots and shoots were analyzed by ICP-MS. P-deficient T. aestivum, B. napus and C. arietinum did not respond with elevated carboxylate release. These species accumulated more REE when the P supply was low and higher REE concentrations were proportional to declining plant growth. However, P. sativum, L. albus and L. cosentinii accumulated less REE when P supply was low. Plants that strongly acidified the rhizosphere and released low quantities of dicarboxylates accumulated more REE (with higher LREE/HREE ratios) than species that released tricarboxylates. Our findings suggest that REE accumulation strongly depended on rhizosphere acidification, in concert with the amount and composition of carboxylates determining the exclusion of REE-carboxylate complexes. Leaf REE signatures may offer a promising ionomics screening tool for carboxylate release into the rhizosphere.


Introduction
Root carboxylate release is an essential plant strategy to access sparingly-available soil nutrients, especially inorganic phosphate (Pi), iron (Fe), and manganese (Mn) (Shane and Lambers 2005;Lambers 2022). Plants have adapted to conditions of heterogeneously distributed and sparsely-available soil resources and evolved strategies to influence the properties of the soil surrounding their roots (rhizosphere) to create an environment more conducive for nutrient acquisition (rhizosheath). In addition to mutualistic interactions with bacteria and fungi, and alteration of root morphology (Honvault et al. 2021a, b), the most profoundly studied traits involved in direct root-soil interactions include the acidification of the rhizosheath and release of chelating carbon compounds such as carboxylates (Lambers et al. 2015;Honvault et al. 2021b;Lambers 2022). Rhizosphere acidification in the presence of carboxylates increases the solubility and availability of many essential or beneficial elements, including P, Fe, Mn, Cu, Zn and Si, by dissolution, complexation and ligand-exchange reactions (de Tombeur et al. 2021;Lambers 2022). The ability to mobilize Pi and micronutrients in the rhizosphere varies considerably among plant species, functional plant groups (Neumann et al. 2000;Lambers et al. 2013;Lambers et al. 2015) or even genotypes of specific species (Krasilnikoff et al. 2003;Pang et al. 2018). Plant species adapted to P-impoverished or P-sorbing soils, of which Proteaceae and some grain legumes such as Lupinus albus have been most profoundly studied, respond to P deficiency by increased release of citrate and malate (Neumann and Römheld 2001;Pearse et al. 2006). In contrast, non-mycorrhizal phosphophilous species in the Brassicaceae, Chenopodiaceae, Urticaceae and some cereals such as Triticum aestivum, do not respond to P deficiency with elevated carboxylate release (Pearse et al. 2006;Lambers 2022).
Although carboxylate-based P-acquisition strategies are regulated by plant P status and predominantly target the acquisition of essential mineral nutrients, the resulting chemical changes in the rhizosphere are nonelement-specific (Lambers et al. 2015). That means, while nutrient deficiency triggers a shift in metabolism towards elevated proton and carboxylate release, the compounds released solubilize not only nutrients, but also mobilize a number of non-essential elements in the rhizosphere, impacting their chemical speciation and availability to plants as demonstrated for Cd, Pb, Ge and rare earth elements (REEs) (Wenzel 2009;Wiche et al. 2016a, b). In this respect, REEs are particularly interesting to study, because they i) are present in almost all soils at concentrations similar to essential plant nutrients (Reimann et al. 2003;Wiche et al. 2017a), ii) share chemical similarities with essential nutrients, mainly Ca (Tyler 2004;Brioschi et al. 2013), and iii) strongly interact with nutrient-bearing soil minerals (phosphates, Fe-oxyhydroxides), but are neither essential to plants nor strongly toxic (Tyler 2004;Davranche et al. 2017). The REEs comprise a group of 16 elements from the lanthanide series, including lanthanum, yttrium (Y) and scandium (Sc) that are widespread in the earth's crust with concentrations that vary from 66 μg g −1 (Ce), 30 μg g −1 (La) and 28 μg g −1 (Nd) to 0.3 μg g −1 (Lu) (McLennan 2001;Davranche et al. 2017). As a unique feature in this group, all 16 REEs exhibit ionic radii similar to Ca 2+ ; however, under most pedologically-relevant conditions, REEs form trivalent cations (Wyttenbach et al. 1998), which strongly interact with phosphate and other negatively charged soil constituents (Diatloff et al. 1999;Cao et al. 2001;Li et al. 2014). In particular, they can form stable complexes with dissolved organic compounds (Pourret et al. 2007;Wiche et al. 2017b), and their stability depends on the nature of the ligand and the REE involved. There are slight differences in ionic radii from light REEs (LREE: La to Eu) to heavy REEs (HREE: Gd to Lu, including Y), leading to differences in their sorption and complexation behavior in soil and their availability in the rhizosphere (Khan et al. 2016;Schwabe et al. 2021;Monei et al. 2022). For REEs in the soil solution, it is generally assumed that uptake of REE 3+ -ions is mediated mainly by Ca 2+ -, Na + -and K + -channels (Han et al. 2005;Brioschi et al. 2013), while REE-carboxylate complexes are excluded, relative to free ionic forms (Han et al. 2005;Wiche et al. 2017b). After root sorption, due to the element's higher reactivity, the biogeochemical behavior of REEs in the soil-plant system is not simply analogous to Ca 2+ , but may resemble that of other trivalent metals, particularly Al 3+ (Ma and Hirdate 2000). Thus root-shoot transport of REEs depends on their mobility within the plant (Kovarikova et al. 2019), most likely governed by cell-wall absorption, phosphate deposition and intracellular complexation with carboxylates (Ma and Hirdate 2000). Based on the above, plant species that deploy a carboxylate-based nutrient-acquisition strategy will likely exhibit differences in REE sorption. The P status of the plants and the quantity and composition of the compounds released should influence the processes during the mobilization of non-essential elements and their uptake.
In the present study, we conducted a split-root experiment with two P-efficient carboxylate-releasing lupin species (Lupinus albus and Lupinus cosentinii) that typically show carboxylate release under low P supply, and four species (Triticum aestivum, Brassica napus, Pisum sativum, Cicer arietinum) that lack the ability to respond to P deficiency with elevated carboxylate release (Pearse et al. 2006). A split-root approach was used to exclude the direct effects of P addition on REE availability, i.e. by precipitation as REE-phosphates (Fehlauer et al. 2022;Liu et al. 2022). Thus, one root half received all essential plant nutrients except phosphate, which was supplied to the other root half only. Root carboxylate release and shoot element concentrations (selected nutrients, aluminum, cadmium, REE) were measured, to explore the relation between P nutrition and accumulation of non-essential elements, including total REE uptake and LREE/HREE ratios. If we would be able to show such a correlation, this would offer the possibility to use shoot REE signatures to proxy the involvement of carboxylates in nutrient acquisition.

Substrates for plant cultivation
In this experiment, 120 pots (7 × 7 × 18 cm) were filled with 1.2 kg of sand. Half of the pots (60 pots) were filled with quartz sand (0.1-0.4 mm grain size, 1500 kg m −3 ), while the other half was filled with a mixture of 75% of quartz sand and 0.25% of river sand (0.4-2 mm grain size, 1320 kg m −3 ). Here, a second sand type was added to increase the amount of potentially plant-available elements to one half of the split-root systems. The quartz sand had a pH of 5.6 (water/solid 1/10) and 1.1 ± 0.5 mg kg −1 calcium lactate-extractable P (van Laak et al. 2018), whereas the mixed sand had a pH of 5.9 and 2.1 ± 0.3 mg kg −1 P. In both sand types, the total element concentrations were similar (Table 1); however, the sand types differed regarding the distribution of elements in potentially plant-available element fractions indicated by a sequential extraction analysis considering the distribution of elements in five operationally-defined soil fractions according to Wiche et al. (2017a) (Table 1). In these fractions, the mixed sand was characterized by higher concentrations of P, Mn and Fe (Table 1). Furthermore, the quartz sand showed higher concentrations of mobile/exchangeable and acid-soluble Al and higher concentrations of mobile/exchangeable Table 1 Total element concentrations and distribution of elements in exchangeable (F1), acid-soluble (F2), oxidizable (F3) and moderately-reducible (F4) fractions (μg g −1 ) according to Wiche et al. (2017a, b) determined by a sequential extraction method (mean ± sd; n = 10) Differences in means between the two sand types are identified by t-tests with Bonferroni correction. Means with different letters are significantly different at α = 5% LREE. Thus, in quartz sand, these elements are more easily accessible by roots than in mixed sand. However, Al and both LREE and HREE were generally more concentrated in the mixed sand, especially in the more stable fractions 4 and 5, which were also the significant element-bearing fractions of Cd (Table 1). The LREE / HREE ratios in both sand types were > 1 (Table 1). In particular, quartz sand exhibited a 12% higher LREE / HREE ratio in Fractions 1 (mobile/ exchangeable) and a 15% and 33% higher LREE/ HREE ratio in Fractions 4 and 5, respectively, where the elements are predominantly bound to amorphous and crystalline structures of oxides and oxidehydroxides (Table 1). In Fractions 2 and 3, however, the LREE/HREE ratios were similar between the two substrates.

Plant growth
Seeds of Triticum aestivum cv Arabella, Brassica napus cv Genie, Pisum sativum cv Karina, Cicer arietinum cv Kabuli, Lupinus albus cv Feodora, and Lupinus cosentinii cv were surface sterilized by washing the seeds with 0.5% sodium hypochlorite (NaOCl) for 3 min, followed by rinsing with deionized water. Seeds were germinated in Petri dishes in a climate chamber at 20 °C. After germination and development of seminal roots, the seedlings were transferred to a hydroponic culture with a 1/20 strength Hoagland solution (Arnon and Stout 1939), 22 °C room temperature, relative humidity 60% and 600 μmol m −2 s −2 photosynthetically active radiation. After one week, the primary roots of B. napus, P. sativum, C. arietinum, L. albus, and L. cosentinii were cut 1 cm below the first lateral roots to obtain a split root system by stimulation of root branching and lateral root development (Saiz-Fernandez et al. 2021). Triticum aestivum developed several seminal roots; thus, the abovementioned procedure was unnecessary, and the roots could easily be diverted into different compartments. After cutting, all plants were transferred back into the hydroponic solution and cultivated for another 10 days to allow the plants to recover (Saiz-Fernandez et al. 2021). Plant individuals with similarly developed root systems were transferred from hydroponic culture into the previously prepared pots filled with sand. Each experimental unit consisted of one plant with a split root system where one part of the root system was placed in a pot with quartz sand and the other part into a pot with mixed sand. The pots were connected with clamps, and the seedlings were stabilized with a stick to support the shoot growing between the two pots. In total, from each plant species, 10 experimental units were prepared. The plants were grown in a growth chamber at 22 °C and 65% humidity, 600 μmol m −2 s −1 photosynthetically active radiation and watered with a 1/20 strength Hoagland solution containing all essential mineral nutrients, except P. After one week of growth and allowing the plants to extend their roots deeper in the sand substrates, the experimental units were watered with two different nutrient solutions containing either all essential plant nutrients according to a 1/10 strength Hoagland solution except P (P0), or all mineral elements contained in the previous solution with the addition of 100 μM P (P+). Half of the experimental units were watered with P0 solutions at both root sides (50 mL in each pot), whereas the other half received P+ solutions at the root side growing in quartz sand (50 mL) and P0 solutions at the root side growing in mixed sand (50 mL). The addition of treatment solutions was continued every second day over a period of five weeks. Each P treatment was replicated fivefold for each plant species, and the different species and treatments were spatially distributed in a fully randomized design.

Rhizosphere properties and exudate collection
After five weeks, the plants were removed from the sand and carefully shaken to remove loose sand particles. Sand adhering to the root surface was collected by washing the roots with 20 mL of deionized water until 1 g of rhizosheath was obtained. The sand was left in the washing solution for 1 h until the pH was measured using a pH electrode. If necessary, the root was washed a second time without collecting the solution or sand material to remove the remaining sand entirely. The plants were transferred with their individual root systems into a 200 mL sterile Erlenmeyer flasks filled with 100 mL of a 2.5 μM CaCl 2 solution. This allowed the collection of root exudates depending on plant species and P-treatment for each root system separately. The plants in the collection solutions were placed back into the growth chamber and allowed to release root exudates over a time period of 3 h. Immediately after the collection, the resulting solutions were analyzed using ion chromatography. After that, the plants were separated into roots and shoots. Shoots were washed for 1 min with deionized water. The split roots were separately washed for 5 min with ice-cold CaCl 2 solution (5 mM) and 1 min with deionized water to remove adsorbed ions from charged root cell structures (Han et al. 2005). Finally, the shoots and roots were dried at 60 °C for 48 h, weighed and stored in centrifuge tubes until being analyzed by inductively coupled plasma mass spectrometry (ICP-MS).

Determination of carboxylates and element concentrations
The dried plant material was ground to a fine powder using a centrifugal mill equipped with a titanium rotor (Retsch ZM 100) and stored in centrifuge tubes. Afterwards, microwave digestion (Ethos plus 2, MLS, Leutkirch, Germany) was carried out with 0.1 g of subsample taken from the ground biomass and measured in duplicate. Samples were mixed with 1.6 mL nitric acid (65% suprapure) and 0.6 mL hydrofluoric acid (4.9% suprapure) and heated to 220 °C in a microwave, according to Krachler et al. (2002). Concentrations of P, Fe, Mn and REEs (Y, La, Ce, Pr, Nd, Sm, Eu, Gd, Tb, Dy, Ho, Er, Tm, Yb and Lu) from the diluted digestion solutions and soil solutions were determined by ICP-MS (XSeries 2, Thermo Scientific, Dreieich, Germany) using 10 μg L −1 rhodium and rhenium as internal standards. Possible interferences were monitored and corrected if necessary .
Concentrations of acetate, fumarate, glutarate, malate and citrate in the collection solutions were determined by ion chromatography equipped with conductivity detection (ICS-5000, 4 mm system, Thermo Scientific, Dreieich, Germany). Inorganic and organic acid anions were separated at 30 °C on an IonPac® AS11-HC column (Thermo Scientific, Dreieich, Germany) using gradient elution with sodium hydroxide as eluent and a flow rate of 1.0 mL min −1 .

Data processing and statistical analysis
Concentrations of LREEs and HREEs in the plant and soil samples were calculated as sums of La, Ce, Pr, Nd, Pm, Sm, Eu (LREEs) and Gd, Tb, Y, Ho, Er, Yb, Tm, Lu (HREEs) according to Tyler (2004). Significant differences among means of element concentrations in soil fractions, carboxylate concentrations of P+ and P0 plants, and element concentrations in plant parts cultivated with different P supply were compared by t-test with Bonferroni adjustment of p values using IBM SPSS Statistics 25. Carboxylate release and element concentrations in different root parts of the same plants were compared by a t-test for non-independent samples at α = 5%. Element concentrations, contents and root carboxylate release among plant species within a certain P-treatment were compared by one-way analysis of variance (ANOVA) followed by Tukey's HSD post-hoc test. Prior to the analysis, the data were checked for homogeneity of variances using Levene's-test. In case the assumption of homogeneity was violated, the data were log-transformed. If the assumption was still violated, significant differences between means were identified using Welch's ANOVA at α = 5%.

Plant growth and biomass responding to P-supply
The dry biomass varied considerably among the tested plant species (Fig. 1). Brassica napus accumulated the most biomass (3.1 g) when P was supplied, and T. aestivum accumulated the least biomass when no P was supplied (0.2 g) (Fig. 1A). Phosphorus supply increased the total dry mass (shoot and root mass) of T. aestivum, B. napus, C. arietinum and L. albus by 65%, 52%, 27%, and 56%, respectively. In contrast, the total biomass of P. sativum and L. cosentinii did not respond to P supply (Fig. 1A).
Considering the shoot biomass of plants, the addition of P did not significantly (α = 5%) affect the shoot mass of P. sativum, C. arietinum, and L. cosentinii, and there were no differences among P. sativum, L. cosentinii and C. arietinum (Fig. 1B). However, L. albus, B. napus and T. aestivum strongly responded (p < 0.01) to elevated P-supply with 39%, 52%, and 88% greater shoot biomass (Fig. 1B). Considering the whole root system, including both root parts growing in quartz sand and mixed sand, the addition of P did not significantly affect root biomass within a plant species, but tended to increase (p = 0.14) total root Fig. 1 Shoot mass and root mass (g) of the two root halves growing in quartz sand (Q) and mixed sand (M) and root mass ratio considering both root halves depending on treatment with 100 μM P (P+) and no P (0P). The bars represent means ± se (n = 5). ANOVA identified differences among species (capital letters) following Tukey's HSD post-hoc test at α = 5%. Significant differences in shoot mass and root mass ratios between different P-treatments are indicated by asterisks (**p < 0.01; *p < 0.05; (*) p < 0.1).Differences in root masses between the root halves within a P treatment are indicated with lowercase letters. Capital letters denote differences among the species within a P-treatment. Means with the same letters are not significantly different at α = 5% mass in L. cosentinii, by 124% compared with P-deficient plants (Fig. 1C). There were no differences between plant species when the plants were externally supplied with P ( Fig. 1C). However, in P0-treated plants, L. albus and C. arietinum showed the greatest root mass, and the lowest root mass was found for L. cosentinii (Fig. 1C).
Considering the development of the split root systems in different sand types, all species tended to have a greater root mass in mixed sand, especially T. aestivum and L. albus, which showed 200% and 60% more root mass when no P was supplied and C. arietinum (125% more biomass in mixed sand) when P was provided. Without the addition of P, the root mass ratio varied significantly among the species showing decreasing ratios from T. aestivum > C. arietinum > P. sativum, L. albus, L. cosentinii > B. napus (Fig. 1D). Addition of P significantly reduced the root/shoot ratio in T. aestivum, B. napus and L. albus by 45%, 25% and 17%, respectively, while in the other plant species, there were no effects (C. arietinum) or slightly increasing trends (P. sativum, L. cosentinii). When the plants received solutions containing 100 μM P, the root mass ratio was similar in T. aestivum, P. sativum, C. arietinum and L. cosentinii, but lowest in B. napus (Fig. 1D).

Shoot nutrient accumulation
Shoot [P] of plants watered with 100 μM P ranged from 1.21 mg g −1 (B. napus) to 2.46 mg g −1 (T. aestivum) ( Table 2). Triticum aestivum and L. cosentinii showed substantially higher [P] in shoots than all other investigated species did. Shoot [P] of T. aestivum, B. napus, P. sativum, L. albus and L. cosentinii responded to a reduction in P supply by a 57%, 19%, 13%, 12% and 20% decrease of shoot [P], respectively, compared with plants treated with high P (100 μM P). Shoot [P]   In the corresponding mixed sand root part of P-supplied plants, root [P] was the highest in B. napus and P. sativum and the lowest in T. aestivum. When P supply was reduced at the root side in quartz sand, root [P] also declined significantly in the mixed sand root part of T. aestivum (19%) and L. albus (18%), but it was unchanged in the other species. Considering both P treatments, roots in mixed sand of T. aestivum and C. arietinum showed consistently lower [P] than roots growing in quartz sand, while in B. napus and Table 3 Nutrient concentrations in roots of six species cultivated under split-root conditions on two sand types, quartz sand and mixed sand (means ± sd; n = 5) The plants received 100 μM P (P+) or no P (0P) in quartz sand. Capital letters denote differences among plant species within a P treatment, and lowercase letters denote differences between the root halves for a specific element P. sativum root [P] only differed when P was added to the root half in quartz sand (higher [P] in quartz sand than in mixed sand). Lupinus albus and L. cosentinii did not show any differences in root [P] between the roots when P was added, nor in situations of P deficiency. Concerning the micronutrients, the root half in mixed sand generally responded more strongly to the P-treatment than the root half growing in quartz sand (Table 3). Specifically, the reduction in P supply increased [Mn] in B. napus (45%), P. sativum (83%), C. arietinum (54%), L. albus (20%) and L. cosentinii (81%).
Carboxylate release in response to P supply Considering the quantity of carboxylates released by both root parts per root half and unit of time, B. napus released by far the greatest amounts, irrespective of P treatment ( Fig. 2A). In B. napus, T. aestivum and P. sativum, the major portion (more than 98%) of the carboxylates released consisted of malate, and citrate was only occasionally detected. In contrast, C. arietinum, L. albus and L. cosentinii released both malate and citrate ( Fig. 2A). Carboxylate release was not affected by P supply in T. aestivum and C. arietinum. Brassica napus and P. sativum responded to a reduction in P supply with a decrease in carboxylate release by 20% (p = 0.04) and 44% (p = 0.08), respectively. In contrast, in L. albus and L. cosentinii, the reduction of P supply significantly increased total carboxylate release by 159% (p < 0.01) and 115% (p = 0.03), respectively, showing an increase of both malate and citrate, but especially of citrate ( Fig. 2A). Roots growing in mixed sand released greater amounts of carboxylates per unit time in all tested species, except B. napus, which tended to release greater amounts of malate in quartz sand, but only when this root part was supplied with P ( Fig. 2B). Also, in the other species, there were significant differences in the response of the different root halves to P supply. Triticum aestivum showed no response in any of the root halves (Fig. 2B). Brassica napus, P. sativum and C. arietinum predominantly responded in the root half in quartz sand, where P was added with the watering solution and showed a significant reduction in carboxylate release (24%, 65% and 75%) at low P supply. In comparison, in the root half in mixed sand, carboxylate release in P. sativum and C. arietinum was unchanged or increased in B. napus by 80% when P supply in quartz sand was low. Also, L. albus and L. cosentinii did not respond in the root half supplied with P but showed an increase of carboxylate release from the root half in mixed sand only. Here, the exudation of malate increased by 121% and 320%, respectively, and citrate release increased by 192% and 870%, respectively, when P-supply was low (Fig. 2B). Carboxylate release per unit root mass showed far less variation depending on growth substrates and P supply (Fig. 2C). Mixed sand roots still tended to release more carboxylates per unit root mass. However, this trend was only observed in P-supplied T. aestivum, C. arietinum (irrespective of P-supply) and L. albus, but in the latter species only when P was lacking (Fig. 2C). Additionally, in both C. arietinum and L. albus, exudation rates were affected by P supply, showing less carboxylate release from the root half in quartz sand (C. arietinum: 90% decrease) or increasing exudation from the root half in mixed sand (L. albus: 105% increase) (Fig. 2C).

Rhizosheath acidification in response to P supply
In the rhizosheath of all plant species and treatments, the pH was significantly higher than that of the unplanted control soil (Fig. 2D). The pH in the mixed sand rhizosphere was consistently higher (on average 0.3 units considering all species) than that in quartz sand as a consequence of the initial pH of the substrates used (Table 1); however, the pH of the substrates was altered depending on plant species, root half and P supply (Fig. 2D). Considering data from both root halves, the rhizosphere pH of P-supplied plants of B. napus, L. albus and L. cosentinii (pH 7.1 ± 0.2) was on average 0.5 units higher than that of C. arietinum, P. sativum and T. aestivum (pH 6.6 ± 0.1). When the P supply was low, the pH in the rhizosheath of B. napus was still highest and lowest in the soil of L. albus and L. cosentinii, which was predominantly driven by a strong acidification at the root half in mixed sand. In contrast, in P. sativum the pH was low in both root halves, irrespective of the P treatment.
Considering the different root halves under P+ conditions, the pH in the quartz sand rhizosheath was highest for B. napus and showed the pattern B. napus > L. albus = L. cosentinii > P. sativum = T. aestivum = C.
arietinum. When P was lacking, the rhizosphere pH of B. napus, C. arietinum and T. aestivum was 0.3 units lower (p < 0.05) but unchanged (around 6.7 ± 0.2) for L. albus, L. cosentinii and P. sativum. At the root half with mixed sand, the pH of B. napus, L. albus and L. cosentinii was much (7.2 ± 0.1) higher than that of T. aestivum, Fig. 2 Total carboxylate release per plant (A), (B) carboxylate release from root halves growing in quartz sand (Q) and mixed sand (M), (C) exudation rates from the different root halves, and (D) rhizosphere pH depending on treatment of plants with 100 μM P (P+) or no P (0P) from the root half growing in quartz sand (means ± se, n = 5). Capital letters indicate significant differences between species within a P treatment. Small letters indicate i) differences among species concerning a specific carboxylate type within a P-treatment (A), ii) differences in carboxylate release between different root halves within a P-treatment and species (B, C) or iii) differences in the pH between the same root half at different P supply rates (D). Means with different letters are significantly different at α = 5% identified by Tukey's HSD post-hoc test C. arietinum and P. sativum (6.8 ± 0.1) (α = 1%). Here, the low P supply reduced the pH in the rhizosheath of L. albus and B. napus by 0.2 units and strongly reduced the pH in the rhizosheath of L. cosentinii, by 0.6 units.

Shoot accumulation of non-essential elements
Shoot [Cd] was highest in B. napus and lowest in C. arietinum (Fig. 3) In other species, no effects of P addition on the LREE/HREE ratios were observed, except in L. cosentinii, which showed the opposite trend with a lower LREE/HREE ratio at a low P supply (Fig. 3).
Considering the shoot element contents (calculated as shoot biomass × concentration) (Table 4), B. napus showed the highest Cd, Al and REE contents, mainly when the plans were supplied with P and shoot content was lowest in T. aestivum. The low P supply did not significantly affect the shoot REE content of T. aestivum, L. cosentinii, C. arietinum and B. napus. However, in L. albus and P. sativum, LREE and HREE contents were 40-46% (P. sativum) and 58-60% (L. albus) lower at low P supply. Moreover, in L. albus Al, Cd contents were 48% and 71% lower. Shoot Al contents in B. napus was 47%, lower and Cd content in T. aestivum was 43% lower when the P supply was low (Table 4).

Root accumulation of non-essential elements
All investigated species showed significantly higher [LREE] and [HREE] in roots growing in quartz sand, irrespective of the P treatment (Fig. 3). Similarly, quartz sand roots of L. albus and L. cosentinii exhibited higher [Al] and [Cd]. Considering the different P supplies, quartz sand roots of T. aestivum, C. arietinum, L. albus and L. cosentinii did not show differences in their Al, Cd, LREE and HREE concentrations. However, in B. napus, the concentrations of all elements were 102% (Cd), 208% (Al), 275% (LREE) and 248% (HREE) higher in P-deficient roots than in roots supplied with P. In P. sativum, P deficiency also increased [Cd]  Considering data from both root parts and P treatments, the calculated LREE/HREE ratios of L. albus and C. arietinum were substantially higher than those in the other species (on average 4.9-5.0 times) and the lowest ratios were found in L. cosentinii (LREE/ HREE = 3.6 ± 0.7). The LREE/HREE ratios were higher in roots grown in quartz sand than in those in mixed sand, except for T. aestivum and P. sativum. In T. aestivum, the ratios were higher in roots grown in mixed sand of P-supplied plants than in corresponding roots grown in quartz sand but without differences between the P+ and P0 treatments (Fig. 3). In contrast, in P. sativum, adding P to the quartz sand decreased the ratio from 5.0 to 4.3. Similarly, in L. albus, P addition decreased the LREE/HREE ratios in both root halves (0.7 units in quartz sand and 0.5 units in mixed sand).
Considering the root element contents (Table 5), P-supplied quartz sand roots of C. arietinum and P. sativum accumulated the greatest amounts of REE and T. aestivum and L. cosentinii the lowest. In the other root half growing in mixed sand, there were no differences in element contents between the species, except for T. aestivum and C. arietinum, which showed a 3-4 times greater Cd content than B. napus, P. sativum and L. cosentinii did. A low P supply at the root half in quartz sand did not change the contents of Al and Cd in any of the investigated species, neither in quartz sand nor in mixed sand roots. However, in roots grown in quartz sand, LREE and HREE contents were 45% lower in P. sativum. Conversely, in mixed sand roots, a low P supply did not change LREE and HREE contents in P. sativum and did not affect LREE in the other species. However, in B. napus, the low P supply tended to increase the content of HREE by 83%, while in L. cosentinii the HREE content decreased by 44%.

Normalized REE pattern in shoots and responses to P supply
The substrate-normalized [REE] calculated for shoots treated with different P levels showed clear differences among the species and partly depended on the treatment with P (Fig. 4). In all plant species, the normalized REE concentrations were < <1, and the pattern was generally similar among the species tested with curves downward from left to right showing LREE-enrichment and HREE-depletion.
In B. napus and P. sativum, the normalized [LREE] relative to [HREE] was much higher than those in T. aestivum, C. arietinum, L. albus and L. cosentinii showing LREE/HREE >1. Moreover, B. napus and P. sativum exhibited steeper curves than T. aestivum, C. arietinum, L. albus and L. cosentinii did. Concerning the effects of P addition, C. arietinum did not show any differences in the REE pattern between P-supplied and P-deficient plants. When

Plant growth, root biomass and nutritional status
In the present experiment, cultivation of plants with split roots growing in different sand types allowed us to control the P supply to one root half without influencing REE availability directly through the precipitation of Al and REE with phosphate in the presence of P. The treatment with low P supply showed less production of shoot and total plant biomass of T. aestivum, B. napus, and L. albus, whereas there was no effect on C. arietinum, P. sativum and L. cosentinii (Fig. 1). The latter species showed virtually unchanged shoot [P] following P addition (Table 2). Shoot [P] did not exceed the concentration that is adequate for crop growth of 2 mg P g −1 dry weight (Marschner 1995), except in T. aestivum. This was unexpected, given that the plants received a high supply of P (100 μM P as KH 2 PO 4 ) in the nutrient solution. It is possible that a considerable amount of P sorbed onto Al and Fe oxides and hydroxides of the acidic quartz sand. Additionally, after five weeks of plant growth, all plants entered the reproduction phase, so P remobilization to the seeds may have contributed to the low shoot [P] (El Mazlouzi et al. 2020). Shoot [Fe] and [Mn] were largely unchanged at the low P supply. However, lower concentrations of Mn and Fe in P-deficient T. aestivum and P. sativum (Table 2) might indicate a reduced uptake and/or translocation capacity (Fan et al. 2021). Root [P] was higher in all species in the high P-half. This was not only observed in the root half in contact with the nutrient solution, but also in the other root half, grown in mixed sand of T. aestivum, C. arietinum and L. albus, and to some extent in roots of L. cosentinii (Fig. 3). Conversely, [P] of roots grown in mixed sand of B. napus and P. sativum were unaffected. Indeed, the [P] was highly influenced by root growth. Fig. 3 Concentrations of trace elements in shoots (left) and roots (right) of split-root plants treated without phosphorus (low P) or with 100 μM P at the root half growing in quartz sand. LREE = sum of La -Eu, HREE = sum of Gd -Lu plus Y (means ± sd, n = 5). Differences between the P treatments were identified by t-tests with Bonferroni correction. In shoots, asterisks indicate significant differences between P treatments and means with the same capital letters were not significantly different (identified by ANOVA and Tukey's HSD post-hoc test) among plant species within a P-treatment at α = 5%. Capital letters indicate differences among plant species within a specific root half and P treatment. Lowercase letters denote differences between P treatments within a species and the root side. Additionally, for roots, asterisks indicate significant differences between root sides within a specific P treatment (α = 5%) ◂ However, all plants developed more root biomass in the mixed sand. Hence, these findings indicate that the plants allocated a large portion of P absorbed in quartz sand to the other root half growing in mixed sand. The increased root mass ratios of T. aestivum, B. napus and L. albus in P0 treatments (Fig. 1) indicate a relatively increased allocation of dry matter to roots and adjustment of root growth to a low P supply (de Bang et al. 2020). This growth adjustment is determined by the overall nutrient status of the plants (Robinson 1996), and, therefore, might explain the high biomass allocation in T. aestivum, which showed the largest differences in shoot [P] resulting from differences in P supply. In contrast, L. cosentinii, C. arietinum and P. sativum did not respond to differences in P supply with altered root mass ratios. These species presumably relied more heavily on chemical changes in the rhizosphere than on more extensive root systems (Pearse et al. 2006). In the present experiment, the mixed sand (roots without P supply) was characterized by a higher pH and higher P availability (Table 1). Therefore, in the present experiment, resource allocation must be considered not only between shoot and roots but also between the different root halves (Fig. 1), allowing us to explore the capacity to respond to nutrient availability by plasticity in root development. Indeed, when P supply was low in quartz sand (P0), all species (except L. cosentinii) developed more extensive roots in the mixed sand where the plants were exposed to conditions that allowed them to acquire more nutrients. The P0 treatment reduced the root growth of B. napus in quartz sand, but did not affect the root mass of other species at this root side.
When the P supply was higher at the root side in quartz sand, the root mass of L. albus was unaffected in mixed sand, but B. napus showed a lower root mass. In contrast, L. cosentinii and C. arietinum had a higher root mass in mixed sand (Fig. 1) when the plants were supplied with P to the roots in quartz sand. This suggests that under P deficiency, the phosphophile B. napus mainly relies on readily-available P sources and effectively adjusts its root growth to the compartment where P can be most easily acquired. In contrast, C. arietinum, L. albus and especially Table 4 Contents of nonessential elements in shoots of six species cultivated under addition of 100 μM P (P+) or no P (0P) (means ± sd; n = 5) Capital letters denote differences among the species within a P treatment. Differences in element contents between species within a P treatment were identified by ANOVA followed by Tukey's HSD post-hoc test. Differences in element contents in a species between P treatments were identified by t-tests with Bonferroni correction L. cosentinii appeared to follow a more conservative strategy and sustained root development in the mixed sand with higher total nutrient concentrations, increasing the chance to maintain P and micronutrient supply through changes in rhizosphere chemistry.
Modifications of rhizosphere chemistry in response to P supply After five weeks of plant growth, all species tested had entered the reproductive phase and started flowering. So the carboxylate release observed in the present study may not necessarily characterize the plant's nutrient-acquisition efficiency, because carboxylate release typically declines when plants enter the reproductive stage (Mimmo et al. 2011). However, the observed exudation rates among P-supplied and P-deficient plants can be used to characterize the species' general response to the P status (Fig. 2). In this study, the amount of carboxylates released from the different root halves per unit time (Fig. 2B) integrates root mass and carboxylate release per unit mass (Fig. 2C). They characterize the species' ability to chemically influence the root environments.
In contrast, exudation rates per unit of time and root mass characterize the physiological response to environmental conditions. T. aestivum, B. napus, P. sativum and C. arietinum did not show differences in rhizosphere pH in response to P supply (Fig. 2D); however, the rhizosheath pH was lowest in P. sativum and C. arietinum (Fig. 2D), highlighting the capacity of these species to acidify the rhizosphere irrespective of P supply (Pearse et al. 2006). In contrast, L. albus, and L. cosentinii strongly acidified the rhizosphere when P was lacking in the nutrient solution, especially in the mixed sand. Table 5 Contents of light rare earth elements (LREE), heavy rare earth elements (HREE) in roots of six species cultivated under split-root conditions on two sand types, quartz sand and mixed sand, respectively (means ± sd; n = 5) The plants received 100 μM P (P+) or no P (0P) in quartz sand. Capital letters denote differences among the plant species within a P-treatment, and lowercase letters denote differences between the root halves It is generally assumed that the response of plants to nutrient deficiency is determined by the overall nutrient status of the plant, as demonstrated for lupins and some Proteaceae species (Shane et al. 2003a(Shane et al. , 2003bWang et al. 2013). However, In C. arietinum, the production and release of carboxylates appears to be independent of plant P status (Wouterlood et al. 2004) and B. napus and T. aestivum typically show slow and declining carboxylate release under P-deficient conditions (Pearse et al. 2006). Consistently, in the present study, P deficiency increased the carboxylate release of the lupins, reduced carboxylate release in B. napus and P. sativum, but did not affect the amount of carboxylates released in C. arietinum and T. aestivum ( Fig. 2A, B).
Besides changes in amounts, the composition of root exudates is an integral factor determining P-mining efficiency (Jones 1998;Lambers 2022  carboxylate released by C. arietinum, L. albus and L. cosentinii ( Fig. 2A) forms more stable complexes with soil cations and consequently is more efficient at releasing P and micronutrients by complexation and ligand exchange reactions (Jones 1998). Moreover, when P supply was low in the quartz sand substrate, all species responded with decreased amounts of carboxylate release at this root side (Fig. 2B) which was primarily due to reduced root mass as a consequence of P starvation (Fig. 1). In contrast, carboxylate-exudation rates were unaffected in P-starved roots of L. albus and L. cosentinii and these species showed an up-regulation of carboxylate release in roots in mixed sand (Fig. 2B, C) attributing to these species' ability to respond to a low P supply with adjustment of root activity and rhizosphere chemistry.
Accumulation of non-essential elements related to P-supply and carboxylate release A low P supply may affect the accumulation of nonessential elements through i) altered plant growth and thus an enrichment per unit biomass, ii) altered uptake and translocation when uptake is mediated by nutrient transporters that are affected by the growth-limiting nutrient, and iii) altered solubility and chemical speciation in the rhizosphere determining the accessibility for transport mechanisms. If altered solubility is involved, when the availability is limited by mobility in soil, any increase in solubility following changes in chemical speciation will ultimately increase diffusion towards the root and the probability of the element entering the root. Conversely, when the mobility of elements is high(er), changes in the chemical speciation from the ionic form to a metal-organic complex may decrease availability through exclusion at the site of uptake (Barber and Lee 1974). In the present experiment, all plants altered the rhizosheath pH and released carboxylates depending on species and P supply (Fig. 2). The sand substrates contained the elements in sparingly soluble forms (Table 1). Less than 0.1% of Cd, Fe, Mn, and Al were present in mobile forms (Fraction 1). In contrast, the solubility of REE was somewhat higher, especially in the quartz sand (Table 1). Nonetheless, all species contained detectable concentrations of all elements with high variability among the species tested (Fig. 1). Aluminum and REE showed a similar behavior in the shoots, consistent with the literature (Liu et al. 2021;Fehlauer et al. 2022). In B. napus, high shoot and low root [Cd] can be primarily explained by the efficient influx and transport of Cd from roots to shoots (Selvam and Wong 2009).
Concerning the effect of P supply, the REE concentrations in shoots and roots responded more sensitively than those of Al and Cd, given that four out of six species showed significant differences in [LREE] and [HREE] following a reduction of P supply (Fig. 3). Of these species, C. arietinum and T. aestivum did not respond to altered element accumulation and showed a relatively flat normalized REE pattern with a slight decrease in HREE accumulation (Gd-Lu) (Fig. 4). These species did not respond to a low P supply with altered carboxylate release (Fig. 2). The higher LREE and Al concentrations in roots in mixed sand of P-deficient C. arietinum (Fig. 3) corresponded with less root biomass (Fig. 1). Enrichment could largely explain this in the roots which led to unchanged element contents in the plants (Table 3). Thus, the higher concentrations of Al and LREE in shoots and roots of P-deficient T. aestivum (Fig. 3) were accompanied by lower biomass production ( Fig. 1) and unchanged element amounts accumulated in the plant compartments (Tables 3; 4).
Similar to T. aestivum, the shoot biomass of B. napus was lower as a consequence of lower P supply ( Fig. 1) but without changes in Cd, Al and HREE concentrations (Fig. 3), whereas LREE concentrations were significantly higher (Fig. 3) and Al, LREE, and HREE contents were less. (Table 3). Additionally, in P-deficient plants, total carboxylate release was less (Fig. 2), suggesting that the element pattern in shoots resulted from less element uptake in concert with a preferential root-shoot transfer of LREE relative to HREE and LREE accumulation in shoots. HREE form more stable complexes with low-molecular-weight organic anions, for instance, citrate, during long-distance transport in the xylem (Ma and Hirdate 2000;Yuan et al. 2017). However, based on the higher charge density, HREE are preferentially sorbed onto cell walls during radial transport and form more stable complexes with metabolites released into the rhizosphere. Given that REEs are predominantly taken up in ionic form through Ca, K, and Na channels (Han et al. 2005), carboxylates and other chelating compounds would alter the chemical speciation, and hence the uptake and accumulation of REE, including the ratio of LREE/HREE (Wiche et al. 2017b). Element exclusion through extracellular complexation has been studied in detail for Al in Alresistant species (Zheng et al. 1998;Ma et al. 2001;Kochian et al. 2004) and Cd in L. albus (Römer et al. 2000). For a specific carboxylate (e.g., citrate), the complex stabilities decrease in the order HREE > LREE > Al > Cd (Byrne and Li 1995;Martell et al. 2004), while for a given element (e.g., La), the complex stabilities decrease in the order citrate > malate > acetate (Fig. 5). Han et al. (2005) demonstrated that organic acids promote the uptake of La by barley, but the effect of the acid decreased in the order acetic acid > malic acid > citric acid, which can be explained mainly by decreased sorption of La onto the apoplast in the presence of the acid anion but a reduced uptake with increasing complex stability (Han et al. 2005). In the present experiment, B. napus released large quantities of malate (Fig. 2), a dicarboxylate with a lower complexation constant (La: log K 4.37) compared with that of citrate (La: log K 7.63). Nonetheless, the large quantities released should favour complex formation and element exclusion, which might also explain the lower total REE concentrations in B. napus than in P. sativum. Pisum sativum released much smaller amounts of dicarboxylates but strongly acidified the rhizosphere (Fig. 2D) and mobilized the elements in plant-available (ionic) forms (Cao et al. 2001;Wiche et al. 2017b). Slight differences in the complexation behavior between LREE and HREE might have influenced the LREE accumulation in this species at a low P supply (Figs. 3 and 4). Indeed, P-deficient roots exposed to quartz sand with higher mobility of REE (Table 1) showed higher concentrations of Al, LREE and HREE but did not affect net root sorption (Table 3) with lower carboxylate release (Fig. 2). Conversely, P-deficient roots in mixed sand released greater amounts of carboxylates (Fig. 2B) and showed higher concentrations (Fig. 3) and element contents (Table 3), most likely through increased element dissolution followed by decreased internal element transport. This contention is supported by the responses in P. sativum, L. albus and L. cosentinii. Pisum sativum strongly acidified the rhizosheath in both root parts, irrespective of P supply, and released only small amounts of carboxylates, mainly malate (Fig. 3). A reduction in P supply did not change shoot and root biomass (Fig. 1). Still, it decreased the concentrations and contents (Fig. 3) of LREE, HREE, Fe and Mn with higher LREE/HREE ratios in P-deficient plants. Conversely, in shoots of L. albus and L. cosentinii, the concentrations and contents of LREE, HREE and Cd declined (Fig. 3) at a low P-supply which was accompanied by greater exudation of citrate (Fig. 2). Although in L. cosentinii, this effect was somewhat less pronounced than in L. albus, in L. cosentinii P-deficient plants displayed significantly lower  Martell et al. (2004) LREE/HREE ratios indicating a higher HREE translocation relative to LREE when P-supply was low. In contrast, P-deficient roots of L. albus showed higher LREE/HREE ratios, irrespective of the root half, while in L. cosentinii, the LREE/HREE ratios in roots were unaffected. This can be primarily explained by the strong acidification of the rhizosphere of L. cosentinii, shifting the carboxylic acid: carboxylate ratio towards the acid form (Pearse et al. 2006), preventing complex formation and favouring uptake of LREE in L. albus but not in L. cosentinii. In the latter species the presence of carboxylates might have increased the release and uptake of HREE from sparingly-available element forms from the HREE-enriched mixed sand (Table 1).

Conclusion
We demonstrated that plant P status influenced the accumulation of the non-essential elements Cd, Al, and REE, beyond the commonly recognized mechanism of REE-phosphate precipitation in roots. Plants that strongly acidified the rhizosphere and released small quantities of dicarboxylates accumulated the highest concentrations of REE. Conversely, modest rhizosphere acidification and large amounts of carboxylates were associated with a significantly lower accumulation of REE. Phosphophile species or plants that do not respond to P deficiency (B. napus, T. aestivum, C. arietinum) with increased carboxylate release accumulated REE to higher concentrations when P supply was low, which was explained largely by reduced growth and thus enrichment of the elements in the plant biomass. Additionally, in these species REE-phosphate precipitation might have contributed to a lower REE accumulation in P-supplied plants. In contrast, plants that released more tricarboxylates under conditions of P deficiency accumulated more REE when the P supply was high and carboxylate release was low. The proposed mechanism involves the mobilization of the elements in the rhizosphere through carboxylate and proton release, pH-dependent formation of REE-carboxylate complexes with complex stabilities depending on the amount and composition of carboxylates with HREEcomplexes > LREE-complexes and exclusion of the complexes during uptake, radial transport and/ or translocation. This suggests a functional overlap of carboxylate-based belowground traits related to P nutrition and exclusion of REE, which otherwise might become toxic in REE-enriched growth environments. The relationship between plant nutrition and REE accumulation could also explain the large variability in REE accumulation among different plant species and plant individuals growing in the same soil. The proposed model provides a mechanistic explanation for the REE-hyperaccumulation in Proteaceae ( Van der Ent et al. 2023) and highlights the potential of leaf REE signatures to characterize plant species regarding their P-acquisition strategy through changes in rhizosphere chemistry following an ionomic approach.
Author contributions All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by Oliver Wiche, Christine Dittrich, Nthati Monei and Juliane Heim. The first draft of the manuscript was written by Oliver Wiche and all authors commented on versions of the manuscript. All authors read and approved the final manuscript.
Funding OW was supported by a habilitation fellowship granted by the German Academic Exchange Service (grant number 91758065).
Data availability All data obtained during the experiment are contained in the manuscript.

Declarations
Competing interests The authors have no relevant financial or non-financial interests to disclose.