Effect of Substrate Properties and Phosphorus-supply on the Rare Earth Element Facilitation in Mixed-culture Cropping Systems of Hordeum Vulgare, Lupinus Albus and Lupinus Angustifolius

This study presents how nutrient availability and intercropping may inuence the migration of REE when cultivated under P-decient conditions. In a replacement model, Hordeum vulgare was intercropped with 11% Lupinus albus cv. Feodora and 11% L. angustifolius cv. Sonate. They were cultivated on two substrates, A (pH = 7.8) and B (pH = 6.6). Two nutrient solutions were supplied, with N, K, Mg and high P-supply (P+), the other with N, K, Mg, and one-third of P-supply (P-, applied to L0 and Lan only). Simultaneously, a greenhouse experiment was conducted to quantify carboxylate release. There, one group of L. albus and L. angustifolius was supplied with 200 µM K 2 HPO 4 (P+) together with the other nutrients while a second group received 20 µM P (P-). L. albus released higher carboxylates at low P-supply than L. angustifolius. Higher P-supply did not inuence the P concentrations and contents of H. vulgare neither on substrate A nor on substrate B. However, addition of P decreased the concentrations of REEs, especially in plants cultivated on alkaline soil. Nutrient accumulation decreased in H. vulgare in intercropping with L. angustifolius when cultivated on the alkaline substrate A with high P-supply. In the same conditions, the accumulation of REE in H. vulgare signicantly increased. Conversely, on the acidic substrate B intercropping with L. albus decreased REE contents and concentrations in H. vulgare. Intercropping with L. angustifolius opens an opportunity for enhanced phytomining and accumulation of REE. Furthermore, intercropping with L. albus on REE polluted soils may be utilized to restrict REE accumulation in crops used for food production.


Introduction
Carboxylates released by plant roots are an important strategy of plants to access sparingly available phosphorus and micronutrients such as Fe and Mn in soil (Cu et al. 2005). Particularly for P, Fe and Mn the availability is limited by low solubility of the corresponding element-bearing minerals and interactions with other inorganic and organic soil phases. genotypes in a certain species (Krasilnikoff et al. 2003). Forbs in general and legumes in particular are considered to be Pe cient due to a strong ability to acidify the rhizosphere and release large quantities of carboxylates under P and Fe de ciency, while grasses such as Avena sativa and Hordeum vulgare are described as P ine cient ; Wang et al. 2013). However, under Fe de ciency grasses release a variety of phytosiderophores to cope with Fe de ciency.
non-essential elements in the soil. In addition to this, they in uence their availability as it has been demonstrated for Cd, Pb and rare earth elements (REEs) (Wiche et al. 2016a). Among these elements REEs are particularly interesting to study because they i) are present in almost all soils at concentrations comparable to essential plant nutrients, ii) share chemical similarities to essential nutrients, particularly Ca, (iii) interact with nutrient bearing soil minerals (phosphates, Feoxyhydroxides), but (iv) are still not essential to plants nor strongly toxic (Tyler 2004). More speci cally, the REEs comprise a group of 17 elements from the lanthanide series including lanthanum, yttrium (Y) and scandium (Sc) that are abundant in the earth 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) (Davranche et al. 2017, Kabata-Pendias 2010, McLennan 2001. As a special feature in this group, all 16 REEs exhibit ionic radii similar to Ca 2+ ; however, under most pedological relevant conditions REEs form +3 ions (Wyttenbach et al. 1998) which strongly interact with phosphate and other negatively charged soil constituents (Cao et al. 2001;Diatloff et al. 1993;Li et al. 2014;Zhimang et al. 2000). As an exception in this group, Eu and Ce may also occur in the divalent or tetravalent state (Davranche et al. 2017). There are slight but indisputable differences in ionic radii from light REEs to heavy REEs leading to differences in their absorption and complexation behaviour in soil (fractionation). Consequently, this might also in uence their movement in soil-plant systems and availability to plants. Previous studies conducted followed the generic laboratory and eld approach, where synthetic REE were introduced to the cultivation area. In other approaches the cultivated plants were left to grow under natural conditions without any anthropogenic modi cations (Cunha et al. 2012). There is general consensus that rhizosphere processes related to plant nutrition not only affect the availability of nutrients but also of non-essential elements such as Pb Cd (Wenzel 2009) and REEs since these elements can be mobilized through lowering of pH and presence of organic acids (Wiche et al. 2017a). Under eld conditions Wiche et al. (2016aWiche et al. ( , 2016b demonstrated that mixed cultures of P ine cient grasses with P e cient legumes increases the uptake of REEs in the grasses which was most likely due to mobilization of REEs in the rhizosphere of lupins and movement of the elements between intermingling root systems which suggested that not the physiological mechanisms of uptake are of relevance for the accumulation levels of REEs in A. sativa and H. vulgare. In fact, it is generally assumed that uptake of REE 3+ ions is mediated mainly, but not solely by Ca 2+ , Na + , K + channels (Brioschi et  In the present study we conducted a mixed culture study under eld conditions where we cultivated H. vulgare (barley) a P ine cient cereal in presence of 11% lupins using either L. albus, a cluster root forming legume (white lupin) and L. angustifolius a non-cluster root forming lupin (blue lupin). Each of these cultivation forms was set up on two different soils with different soil pH and thus differences in plant-available nutrients and REEs. Additionally, on each soil, the plant stands received P-fertilizer in low and high doses to elucidate effects of P-supply and soil properties on REE accumulation in mono and mixed cultured barley plants. Moreover, in a greenhouse experiment, we characterized the root exudate composition of both lupins depending on P-supply which will give a hint on the plant's behaviour at different P-levels in the eld. In this study we used mixed cultures of P-ine cient barley and P-e cient lupins (legumes) Nobile et al. 2019) as an ecologically derived approach to explore the effects of rhizosphere properties, especially rhizosphere acidi cation and carboxylate release on the availability of REEs to the P-ine cient cereal (Faucon et al. 2015). Knowing the dynamics of the interaction of lupins and P in the rhizosphere, we hypothesise rstly, that there is an interaction between P-supply and REE accumulation in the plants and, secondly this pattern should depend on the initial availability of nutrients in the substrates determining the nutritional status of the plants and REEs mobility in the substrate. Lastly, the effects should depend on the lupin species and consequently on the amount and quality of root exudates interacting with soil phases in the intermingling rhizospheres of barley and lupins.

Field experiment
The experiment was conducted at Bauer Umwelt Business, Hirschfeld (Saxony, Germany), in its off-site recycling and remediation centre. A basin of a total volume of 720 m 3 was lled with two homogeneously sieved top soils both characterized as luvisols. One half of the basin was lled with soil from a road construction location near Freital, Germany (hereafter referred to as substrate A). The second half was lled with topsoil from a mining-affected area in the vicinity of Freiberg, Germany (hereafter referred to as substrate B). Substrate A was a silty loam with a pH (H 2 O) of 7.9.
Substrate B was a silty loam with a pH (H 2 O) of 6.8 (Table 1). A summary of plant-available macronutrient concentrations in the two substrates used for the experiment are shown in Table 1. The elements P, Mg and K were extracted by calcium acetate lactate (CAL) and measured with ICP-MS (inductively coupled plasma-mass spectrometry). For analysis of mineral N, NO 3 and NH 4 + were extracted from soil samples with deionized water and 1 mol L -1 KCl respectively and photometrically determined according to Bolleter (1961) and Hartley and Asai (1963 Table 2 Total concentration and sequential Extraction results (µg g −1 ) for the identi cation of the total concentrations of trace elements in the soil substrates. Given are means ± sd (n = 10). Concentrations within the same element fraction between the substrates were compared by t-tests with Bonferroni adjustment. Means with different letters are statistically signi cantly different at α = 5%.

Plant cultivation
White lupin (Lupinus albus L., cv. Foedora), blue lupin (Lupinus angustifolius L., cv. Sonate) and barley (Hordeum vulgare L. cv. Modena) were grown within eld conditions in both a monoculture and a mixed culture system on 80 plots with an area of 4 m 2 each, to assess the migration of REE in soil in uenced by intercropping. To avoid interactions between adjacent plots (e.g., root interactions, water discharge, nutrients, REE and trace metals), a 0.5 m buffer zone was maintained surrounding each plot without vegetation. In a symmetrically modi ed replacement series con guration, all of the plants were planted in rows (leaving 20 cm between rows). In that sense a total density of 400 seeds/m 2 were used when seeding, for the monoculture (referred to as L0 from henceforth), and also for the mixed culture plots. Mixed barley cultures were obtained from the monocultures by replacement of 11% barley plants with the equivalent proportion of individuals of white lupin and blue lupin (hereinafter referred to as Lal and Lan plots) and thus plant densities were equivalent in both barley monocultures and mixed cultures. Additionally, plants in mixed crops were cultivated in various patterns with regard to agricultural practices, on rows determining the degree of plant interactions.
Eight days after seed germination and plant development had taken place, the rst dose of fertilizer was given to all plants. On each substrate plots with barley monoculture and mixed cultures with white and blue lupin (Lal and Lan) were dosed with 10 g of N m −2 as NH 4 NO 3 , 11.6 g K m −2 as KNO 3 , 3 g of P m −2 as KH 2 PO 4 , 1.5 g Mg m −2 as MgSO 4 representing the fully fertilized reference plants (NPK). Accordingly, on each substrate plots of barley monoculture (L0) and mixed cultures with blue lupin (Lan) received a similar fertilizer composition regarding N, K and Mg but with one third of P (1 g of P m −2 as KH 2 PO 4 ) representing the low dosed variant (NK). To ensure consistency in the provision of nutrients throughout the entire experiment and to avert N de ciency, (e.g., by leaching nitrate), the abovementioned fertilizer was applied in two separate doses at the beginning of the experiment and a second time four weeks later.
Overall, this experimental design allowed a comparison between responses to different culture types on P-supply, a comparison of lupins under fully fertilized conditions and the in uence of the growing substrate. Each of the different treatments, including culture forms and fertilizer treatment were vefold replicated on each of the two substrates and within each substrate the treatments were setup in a fully randomized design. After eight weeks of plant growth shoots of barley in monocultures and mixed cultures were cut 3 cm above the soil surface. Only the shoots of the inner square meter of each plot were further processed for chemical analysis.

Quanti cation of carboxylate release
A separate greenhouse experiment was designed for the determination of root exudates in both L. albus and L. angustifolius depending on P-supply. Seeds of L. albus cv Feodora and L. angustifolius cv. Sonate were surface sterilized by washing the seeds with 0.5% sodium hypochlorite (NaOCl) for three minutes followed by carefully rinsing with deionized water and allowed to germinate in petri dishes in a climate chamber at 20°C. After germination the seedlings of each plant species (one seedling per pot) were planted in 10 plastic pots (2 L total volume) lled with acid (HNO 3  while the other plants received 20 µmol L -1 P (P-references). After a cultivation period of four weeks, the mature plants were carefully removed from the sand by washing with tap water and transferred into glass beakers containing 300 mL of a 2.5 µmol L -1 CaCl 2 solution where they were let to stay for two hours under a growth lamp and allowed to release carboxylates into the collection solutions. Immediately after the collection the resulting solutions were stabilized with 1 mL L -1 Micropur to prevent microbial decomposition of carboxylates according to Oburger et al. (2013) and analysed by means of ion chromatography. Thereafter, the shoots and roots were separated, weight and dried for 24 h at 60°C.

Analysis of trace element concentrations and carboxylates
The harvested biomass of eld grown plants was separated in leaves and stems and dried at 60°C in an oven for 24 h. The dried biomass was ground to ne powder and stored in centrifuge tubes. Thereafter microwave digestion (Ethos plus 2, MLS) was carried out with 0.1 g of subsample taken from the ground biomass measured in duplicates. Samples were mixed with 1.6 mL nitric acid (65% supra) and 0.6 mL hydro uoric acid (4.9% supra − ) and heated to 220°C in the microwave according to Concentrations of acetate, malonate, fumarate, glutarate, malate and citrate in the collection solutions were determined by ion chromatography equipped with suppressed conductivity detection (ICS-5000, 4 mm system, Thermo Scienti c). Inorganic and organic acid anions were separated at 30°C on an IonPac® AS11-HC column (Thermo Scienti c) using gradient elution with sodium hydroxide as eluent and a ow rate of 1.0 mL/min. The measuring program started with an eight-minute isocratic phase and a sodium hydroxide concentration of 1 mmol L − 1 , followed by the gradient analysis with a continuously increasing sodium hydroxide concentration up to 40 mmol L − 1 over a period of 35 min. Finally, the column was ushed for three minutes with 50 mmol L − 1 sodium hydroxide and equilibrated for ten minutes with 1 mmol L − 1 sodium hydroxide, so in total the analysis took 56 min.

Data processing and statistical analysis
Concentrations of light rare earth elements (LREEs) and heavy rare earth elements (HREEs) in the plant and soil samples were calculated according to Tyler (2004)  Prior to the analysis the data was checked for homogeneity of variances using Levenes-test. In case that the assumption of homogeneity was violated, the data was log-transformed. If the assumption was still violated signi cant differences of means were identi ed by using single comparisons of groups of means using Welchs's ANOVA at α = 5%.

Results
3.1 Root exudate patterns in L. albus and L. angustifolius affected by P-supply Compared to L. angustifolius, L. albus produced higher shoot biomass (P+ 203%, P-137%), and root biomass (P+ 400%, P-233%), irrespective of P-supply (Table 3). P-supply did not in uence root and shoot dry mass in L. angustifolius as well as root dry mass in L. albus. However, the shoot dry mass of L. albus responded to differences in P-supply showing a reduction by 35% when plants were supplied with low P doses. From the carboxylates measured, only citrate and malate were detectable in all collection solutions (Table 3), while fumarate was only occasionally present. All other carboxylate signals (acetate, lactate, glutarate, malonate) were below their respective detection limits. Under conditions of low Psupply L. albus strongly responded by 271% increased rates of citrate release per unit root dry mass and showed 71% increased release of citrate per plant (Table 3). In this study P-supply did not alter release of malate by L. albus. In contrast, in L. angustifolius, P-de ciency did not increase release of carboxylates. Instead, in L. angustifolius in adequately P-supplied plants exudation rates of citrate and malate per unit root dry mass were 224% and 243%, respectively, higher than in P-de cient plants. Overall, in L. angustifolius this resulted in a 180% higher release of citrate and 650% higher release of malate in P-supplied plants. A comparison of exudation rates and amounts of carboxylate release per unit root dry mass between two lupin species revealed that there was no difference in the exudation rates under low P-supply. However, when the plants received high P-doses with the treatment solutions, exudation rates of citrate and malate in L. angustifolius per unit root dry mass were 1100% (citrate) and 140% (malate) higher than in L. albus (p < 0.05). Considering the amounts of carboxylates released per plant individual (µM h -1 ) under P-de ciency L.
albus released 140% and 900% more citrate and malate, respectively. In contrast, when P-supply was high, L. angustifolius released 100% more citrate while release of malate was similar. Table 3 Growth parameters and root carboxylates collected from L. albus and L. angustifolius that were semi-hydroponically cultivated under P-de cient conditions (20 µM P: P-) or supply of 200 µM P (P+). The values are means ± sd (n = 4). Signi cant differences among parameters within a species and between species and within a speci c P-treatment were identi ed by a t-test with Bonferroni adjustment.  (Table 5). Substrate properties, culture form (mixed culture with different mixing ratios of L. albus or L. angustifolius) and P-fertilization did not in uence biomass yields of stems of H. vulgare (Table 4 and 5) and there were no differences in leaf biomasses between substrates. Also, intercropping and P-addition did not in uence leaf biomass on substrate B, neither in plant stands with L. albus, nor with L. angustifolius. However, on substrate A, mixed culture cultivation with L. angustifolius slightly increased (p = 0.09) leaf biomass of barley when barley was cultivated at low P-application level (NK) ( A comparison of concentrations in leaves and stems, respectively, and considering data from both substrates and all culture forms and fertilizer treatments revealed that concentrations of all investigated elements were consistently higher in leaves compared to the stems, except for P on substrate B. On substrate A, leaf concentrations were 28% (P), 171% (Ca), 196% (Mn) and 316% (Fe) higher than in stems. On substrate B leaf concentrations were 201% (Ca), 213% (Mn) and 405% (Fe) higher than in stems.

Species
The addition of high doses of P-fertilizer did not affect the concentrations of Ca, Fe, Mn and P in leaves and stems, respectively, irrespective of the growth substrate.   Table 6 Concentrations (µg g −1 ) of light rare earth elements (LREEs) and heavy rare earth elements (HREEs) and their ratio (LREEs relative to HREEs) in the plant parts depending on substrate (slightly alkaline substrate A and slightly acidic substrate B), P addition (low dose P-and high dose P+) and culture form (monoculture: L0, mixed culture with 11% L. albus: Lal and mixed culture with 11% L. angustifolius: Lan). Means ± sd (n = 5). Signi cant differences in yields and concentrations within a plant part and substrate were identi ed by MANOVA followed by Duncan's' post-hoc test. Small letters show differences between means of mono and mixed cultured barley within a speci c substrate and P-treatment. Capital letters denote differences of concentrations in barley plants of a speci c treatment between P-treatments within a substrate. Capital letters in italics show differences of concentrations in barley plants between substrates at α = 5%.

Rare earth element concentrations in different plant parts
Considering both substrate types, all culture forms and fertilizer treatments, concentrations of REEs were constantly higher in leaves compared to the stems with LREE/HREEs > 1 (Table 6). On substrate A leaf concentrations were 442% (LREEs) and 140% (HREEs) higher than in stems (p < 0.01). Also, the LREE/HREE ratio was 46% higher in leaves than in stems (p < 0.01). On substrate B leaf concentrations were 540% (LREE) and 280% (HREE) higher in leaves than in stems (p < 0.01) with very similar LREE/HREE-ratio among the two plant compartments. The addition of P-fertilizer did not affect the concentrations of REEs directly (Tables 4 and 6). However, there were signi cant interaction effects between Papplication and culture form in uencing the REE concentrations in leaves as well as P-application × culture interactions in uencing the REE concentrations in the stems. Overall, the growth substrate strongly affected REE concentrations in leaves but not of stems with a more strongly pronounced effect on LREE (p < 0.01) than on HREE (p = 0.05). Considering data from all mixed culture forms and P-fertilizer treatments, leaf concentrations on substrate B were 64% (LREE) and 72% (HREE) higher (p < 0.05) than on substrate A but with similar LREE/HREE ratio. Application of P-fertilizer in monoculture signi cantly decreased LREE concentrations of leaves (by 48%) and LREE and HREE concentrations of stems both by 50% on substrate A, while on substrate B this effect was not observable. Also, in the mixed cultures there was no direct effect of P-application and there were no differences in element concentrations between mixed cultured plants that received different fertilizers. Moreover, plants that received only low doses of P (NK) showed no differences in elemental composition between monocultures and mixed cultures. However, on substrate A, mixed cultures of barley with L. angustifolius that were treated with P-fertilizer responded by a signi cant increase in concentrations of LREEs by 113% and HREE by 88% in leaves and 225% (LREE) and 200% (HREE), respectively in stems compared to the monocultures.
On substrate A, L. albus did not alter the mineral composition of the mixed cultured plants, irrespective of the Papplication. In contrast, on substrate B, NPK treated mixed cultures with both L. albus and L. angustifolius signi cantly decreased the REE concentrations by a factor of four in the case of LREEs or even roughly one order of magnitude in case of HREEs. It has to be noticed that these effects were only prevailing on the slightly alkaline substrate A when plant stands of barley and mixed cultures of barley and L. angustifolius were treated with higher doses of P-fertilizer.

Accumulation of nutrients and REEs
Considering the biomass of leaves and stems and the herein quanti ed element concentrations, amounts of elements in the respective plant tissues and whole shoot contents were calculated ( Fig. 1 and Fig. 2). Plant leaves consistently contained signi cantly (p < 0.01) higher amounts of Ca (30%), Mn (44%), Fe (87%) and especially of LREEs (265%) and HREEs (158%) than stems, except P which predominantly accumulated in plant stems with 78% higher amounts than in leaves. The growth substrate strongly in uenced the element contents in leaves showing signi cantly higher amounts of all investigated elements in leaves of plants cultivated on substrate B compared to substrate A (Table 7). In stems only contents of P and Mn were in uenced by a general substrate effect ( The element contents in shoot biomass were not in uenced by general effects of culture form and P-fertilizer addition but rather depended on complex responses of different levels of plant tissue accumulation based on interactions of culture form and substrate properties as well as additional interaction effects of P-fertilizer amendment (Table 7). Speci cally, compared to L. angustifolius, intercropping with L. albus did not positively affect the accumulation of the investigated elements except that of Mn in leaves and shoots of barley plants on substrate B. On substrate B the presence of L. albus increased Mn content in leaves by 116% and in shoots by 63% compared to monocultures, while on substrate A L. albus increased the leaf Mn contents by 102% compared to monocultures. However, for LREEs and HREEs L. albus signi cantly decreased the element contents in shoots (by 68% and 71% respectively) and leaves (by 36% and 46%, respectively) when the plants grew on substrate B with high P-addition, while on substrate A no effect of L. albus on REE accumulation in mixed cultured barley was observed.
Unfortunately, in this study, L. albus was solely cultivated on the two substrates with high dosing of P-fertilizer and thus further evaluations of responses of the mixed cultures to different P-availability are not possible. However, considering mixed cultures with L. angustifolius the effect of intercropping on element accumulation was strongly dependent on the growth substrate and P-fertilizer addition. More speci cally, on both substrates, there was no response of mixed cultured barley regarding the contents of P, Ca, Mn, Fe when barley and L. angustifolius were cultivated with high supply of P (NPKtreatment). In contrast, when P-supply was reduced (NK-treatment) and barley was cultivated neighbouring to L. angustifolius, shoot contents of P, Mn and Fe increased on substrate A by 64% (P), 56% (Mn), and 62% (Fe). This was mostly caused by a signi cant increase in leaf contents, except for P, whereas on substrate B the shoot contents of P, Mn and Fe decreased by 37% (P), 50% (Mn) and 37% (Fe), respectively, due to decreased accumulation in stems and leaves.
Concomitantly, on substrate B there were clear tendencies of a reduction of shoot LREE (by 44%) and HREE (by 46%) accumulation when plants were cultivated with L. angustifolius and low P-dosing compared to the monocultures. Under these conditions L. angustifolius signi cantly reduced LREE contents in stems of barley by 69%. Also, on substrate B the presence of L. angustifolius signi cantly reduced stem contents of HREEs by 46% in high P-dosed mixed cultures compared to the monocultures but without striking effects on bulk shoot contents which remained unchanged.
In contrast, on substrate A, mixed cultures with L. angustifolius signi cantly increased contents of LREEs (by 79%) and HREEs (by 96%) in shoots of barley compared to the monocultures. This can be attributed to a combination of increasing contents in leaves (60% increase for LREEs and 50% increase for HREEs) and in stems (169% increase for LREEs and 263% increase for HREEs) when high doses of P were given. For HREEs, this effect was also visible in leaves of plants that were treated with low P-doses (62% increase) but the effect in leaves was not strong enough to in uence bulk shoot contents of HREEs that remained unchanged compared to the monocultures and ,due to a decrease in stem HREES contents, there was no effect on LREEs plant stands treated with low P-doses.

Phosphorus concentrations in lupin plants as affected by substrate and P-supply
Mixed cultures of barley and lupins that received only low doses of P (1 g P m −2 ) did not show signi cant differences in leaf P concentrations when plants cultivated on substrate A and B were compared (Fig. 3). Nevertheless, P concentrations in plants on substrate B were slightly higher (2.3 mg g −1 ) compared to lupins cultivated on substrate A (1.9 mg g −1 ). Generally, on both substrates fertilization of the mixed cultures with P-fertilizer signi cantly increased the concentrations of P and his effect was most visible on substrate B where NPK treated plants reached up to 3.1 mg g −1 P in leaves. Here, plants of L. angustifolius displayed substantially higher P concentrations than plants on substrate A. L. albus was only cultivated under NPK addition of substrate A and thus investigations of responses of the species to substrate and Psupply were not possible. Compared to L. angustifolius, L. albus exhibited similar P concentrations when both species received NPK fertilizer (Fig. 3).

Evaluation of carboxylate release in different lupin species
In the greenhouse experiment exudation experiment was carried out as a means to evaluate the carboxylate release and consequently the nutrient acquisition e ciency of the cultivars of L. albus (Feodora) and L. angustifolius (Sonate) that were later used in the eld experiment for intercropping with barley. Lupins are characterized by an extraordinarily high e ciency to mobilize sparingly available P  (Marschner 1995). The results successfully demonstrate that the response of the two species was divergent (Table 3) showing a higher release of carboxylates in L. albus under P-de cient conditions, whereas L. angustifolius responded with decreased release of carboxylates (Table 3)  and Neumann and Römheld (2000) who reported increased diffusion of citrate and malate as a consequence of metabolic shifts in carbohydrate allocation from shoot to roots in concert with increased biosynthesis of malate and citrate and decreased citrate turnover in the tricarboxylic acid cycle. Concomitantly, the decreased release of carboxylates in L. angustifolius suggests that this species (or the selected cultivar) lacks the ability to alter carboxylate metabolism following P-de ciency similar to chickpea (Pearse et al. 2006). Based on the above it seems that L. albus should be preferably selected for intercropping aiming at improved plant nutrition in mixed culture systems, especially when plant growth is limited by P-availability. Indeed, the total amounts of carboxylates released per plant were higher in L. albus whereas the exudation rates (per root dry weight) of both lupin species were similar under low P supply (Table 3) which can be explained by thicker roots and consequently a higher total root dry mass relative to the number of active root tip regions of L. albus (Egle et al. 2003) compared to L. angustifolius. However, when the plants were adequately supplied with P, L. angustifolius showed substantially higher carboxylate exudation rates and amounts of citrate released per plant individual compared to L. albus which highlights the high potential applicability of this L. angustifulius cultivar for improvement of nutrient supply for mixed culture cropping with nutrient ine cient species on moderately fertile soils.

Effect of substrate properties on plant growth and nutrient availability to the plants
Considering the leaf nutrient concentrations which are commonly used as proxies for the nutritional state of plants (Hayes et al. 2014) it was obvious that on both substrates the barley plants suffered from Mn and P de ciency indicated by leaf P concentrations close or even below to the critical value of 2 mg g −1 P and 50 µg g −1 Mn (Marschner 1995). The lowest concentrations of P and Mn (below 1.9 mg g −1 P and 20 µg g −1 Mn) were observed in plants on substrate A treated with low P-doses (Table 5). Surprisingly, comparing leaf, stem and shoot biomass on both substrates we did not observe signi cant changes in plant yields between the substrates (Tables 4, 5). Compared to substrate A, concentrations in barley leaves as well as bulk shoot contents (Fig. 1, Table 5), were signi cantly higher on substrate B, indicating an improved nutrient supply on this substrate with its slightly acidic pH. Furthermore, on substrate B leaf P concentration of lupin plants were signi cantly higher than on substrate A and signi cantly higher compared to H. vulgare (Table 5, Fig. 3) while on substrate A leaf P concentration in unfertilized plants of L. angustifolius were similar to H. vulgare. Higher nutrient concentrations in lupins compared to H. vulgare can be explained by a higher nutrient acquisition e ciency of lupins (Pearse et al. 2006). Based on P concentrations determined by CAL-extracts both substrates were su ciently supplied with P (Marschner 1995) but the phosphorus was most likely not present in plant-available forms. Substrate A was slightly alkaline (pH 7.9) ( Table 1) which fosters the precipitation of sparingly soluble Ca-phosphates (Mengel et al. 2001) and low solubility of Mn and Fe. In contrast, soil B was slightly acidic (pH 6.8) ( Table 1) so that low speci c sorption of P (Mengel et al. 2001) as well as higher solubility of Mn and Fe can be expected (Gupta and Chipman 1976). Generally, higher accumulation and concentrations of the nutrients on substrate B was not surprising (Fig. 1, Table 7). However, the higher availability of the elements on substrate B exhibited by higher tissue concentrations and shoot contents was not a priori predictable based on data of the sequential extraction where substrate A showed lower concentrations of P, Ca, Mn, Fe in mobile, exchangeable fractions ( Table 2). On the contrary, substrate B was characterized by higher concentrations of P, Fe and Mn bound into organic matter and amorphous Fe-oxyhydroxides ( In our experiment, the addition of the P-fertilizer did not in uence the P concentrations and contents of barley plants neither on substrate A nor on substrate B (Tables 4 and 5). Possibly, the doses were not high enough (1 g m −2 or 3 g m −2 P) to improve the plants' nutrient supply due to a fast P xation e.g. as Ca-phosphates on substrate A so that it was not available for barley and/or the lupin plants strongly competed with barley for phosphate. P concentrations in lupins signi cantly increased when P was added (Fig. 3) indicating a strong root competition for essential elements between lupins and barley. Finally, resource facilitation in mixed cultures strongly depends on the nutrient status of the lupin plants, their responses through the release of carboxylates in uencing the solubility of the elements in the rhizosphere and migration of elements between the intermingling root systems (Cu et al. 2005;Wiche et al. 2016a;Wiche et al. 2017a). The availability of P and micronutrients was higher in substrate B than on substrate A (Table 1, Fig. 1). Therefore, the low performance of L. angustifolius and L. albus in mixed cultures with barley on substrate B (Fig. 1, Table 5) could be explained by the synergetic effects of reduced carboxylate release by the lupins, especially of L. albus (Table 3), and higher substrate-induced solubility of the elements fostering element uptake by the barley plants. Nevertheless, increased Mn concentrations and accumulation (Fig. 1, Table 5) in mixed cultured on substrate B indicate that cluster roots of L. albus were still active even when P-fertilizer was added. It has to be noticed that even on substrate B the plants were still undersupplied with Mn (Table 5, Section 4.2) which is an additional factor triggering carboxylate release by lupins Lambers et al. 2015;Marschner and Römheld 1994). Concomitantly, carboxylates of L. albus are known to strongly affect the availability of Mn as this species is considered a hyperaccumulator of Mn . In this regard, lacking effects in mixed cultures with L. angustifolius might indicate a lower ability of L. angustifolius to respond to de ciency of Mn, while decreased accumulation of P and Mn in presence of L. angustifolius could be due to competition of barley and lupins for these nutrients.
On substrate A, intercropping with L. angustifolius slightly increased leaf P concentrations of low P-dosed plants above the critical level of 2 mg g −1 suggesting that the improved nutritional state of the barley plants was responsible for the increase in leaf biomass (Table 5). On this alkaline substrate, leaf and shoot nutrient concentrations and contents of barley were exclusively positively affected (Table 5, Fig. 1) on experimental plots with low P-addition although the leaf P concentrations of lupins suggested a lower P-supply in L. angustifolius (Fig. 3) which should lead to decreased root activity of this lupin species (Table 3). However, in plots with a higher P-supply, we observed a better plant growth of lupins (data not shown here) so that it is reasonable that the mobilized nutrients were initially taken up by the lupins without any positive effects on barley. Concomitantly, increased concentrations and accumulation of Ca, Mn and Fe in mixed cultures with low P supply (  Zhimang et al. 2000). As such, the higher concentrations (Table 6) and accumulation (Fig. 2) of REEs on substrate B in comparison to substrate A can be attributed to a higher solubility of the elements in this soil. Higher accumulation of LREEs relative to HREEs observed in this study (Table 6, Fig. 2) closely follow the natural abundance of the elements in the substrates (Table 3). Furthermore, the literature indicates a preferential uptake of LREEs compared to HREEs (Censi et al. 2017, Martinez et al. 2018) due to higher stability of HREE-organic complexes and stronger adsorption of HREEs at ion exchange sites in the soil. These, in turn, may have contributed to these results. Surprisingly, in this study leaf concentrations of REEs were constantly higher than in the stems and the plants mostly responded by changes in leaf REE concentrations (Table 6)  . This suggests that either species-speci c or cultivarspeci c traits that we did not consider in this study in uenced the REE distribution in our plants. Nonetheless, we emphasize that differences in substrates as well as intercropping with lupins impacted both leaf and bulk shoot contents of barley (Fig. 2) although in barley the predominant portion of the shoot biomass consisted of stems (Table 5). Leaves only accounted for one-third of the total shoot biomass (Table 5) and changes in foliar REE absorption due to treatment measures was impactful enough to compensate the lower biomass of this plant part when total shoot contents are considered (Fig. 2). Similar to the ndings for nutrients (see Section 4.2), REE concentrations on substrate B were predominantly in uenced by substrate without signi cant effects of P-fertilizer addition or presence of lupins in mixed cultures. However, on substrate B presence of L. albus signi cantly decreased both shoot REE concentrations and contents, especially when the plants were fertilized with P which highlights an immobilization or uptake of the elements by the lupins under conditions where mobility of the elements is high. Unfortunately, our experimental design did not allow to explore the processes beyond these effects. On the alkaline substrate A, the addition of P-fertilizer signi cantly reduced both LREE and HREE concentrations in monocultured barley plants (Table 6). This can be attributed to a precipitation of the elements as hardly soluble REE-phosphates at alkaline conditions (Saatz et al. 2016, Han 2020) or a "dilution" effect originating from slightly higher shoot biomass (Table 5) which is frequently reported for non-essential elements (Chien and Menon 1995). Compared to the monocultures, the presence of L. angustifolius signi cantly increased tissue concentrations and shoot contents of both LREEs and HREEs in mixed cultured barley. Increased REE availability in mixed cultures with lupins was already described by Wiche et al. (2016b) but without considering differences in substrates or nutrient availability. In the present study positive effects of mixed cultures were only visible on the alkaline, P-fertilizer amended soil and in presence of L. angustifolius which releases higher carboxylates under su cient P-supply (Table 3)

Conclusion
We could demonstrate that mixed cultures of barley with L. angustifolius cv. Sonate increased the accumulation of REEs in barley plants when the plants were additionally supplied with P-fertilizer and cultivated on an alkaline soil characterized by low initial availability of REEs and nutrients. In contrast, on soil with high REE mobility, the presence of L. albus cv. Feodora led to decreased REE contents in barley. This highlights the importance of considering plant physiology, nutritional status of neighbouring plants and general substrate chemistry when soil-plant transfer of REEs in mixed culture plants stands or plant communities are subject of evaluation. Considering these factors, mixed culture cropping systems could be a powerful tool to enhance accumulation of REEs in a sense of phytoremediation or phytomining on marginal soils, while at the same time the mixed cultures with L. albus cv. Feodora could be deployed to cope with REE accumulation in crop plants for food production, especially in REE-polluted soils. The processes involved in the results are not yet fully understood, and thus, elucidation of chemical element species in the rhizosphere of neighbouring plants remains a eld of further research. Nevertheless, our ndings suggest that interspeci c root interactions between legumes and grasses impact the soil-plant transfer not only of nutrients but also the nutrient-like rare earth elements which will improve our general understanding of the biogeochemical cycling of these elements in (agro)ecosystems and legumegrass communities. Figure 1 Total accumulation of nutrients in leaves, stems and shoots (total height of bars) of barley plants in monoculture (L0) and mixed cultures with L. angustifolius (Lan) or L. albus (Lal) on the slightly alkaline substrate A and the slightly acidic substrate B. On both substrates the plants in different culture forms were treated with 3 g m -2 P (NPK) or 1 g m -2 P (NK). Means ± sd (n = 5). Differences among means were identi ed by MANOVA followed by Duncan's post-hoc test. Small letters denote differences in element contents within a speci c plant part, substrate and P-addition treatment. Capital letters show differences between shoot contents within the substrates and treatments at α = 5% Figure 2 Total accumulation of nutrients in leaves, stems and shoots (total height of bars) of barley plants in monoculture (L0) and mixed cultures with L. angustifolius (Lan) or L. albus (Lal) on the slightly alkaline substrate A and the slightly acidic substrate B. On both substrates, the plants in different culture forms were treated with 3 g m -2 P (NPK) or 1 g m -2 P (NK). Means ± sd (n = 5). Differences among means were identi ed by MANOVA followed by Duncan's post-hoc test. Small letters denote differences in element constants within a speci c plant part, substrate and P-addition treatment. Capital letters show differences between shoot contents within the substrates and treatments at α = 5% Figure 3 Leaf P concentrations in mixed cultured lupin plants (L. angustifolius: Lan, L. albus: Lal) that received fertilizer with low P doses (1 g m -2 : NK) or high doses of P (3 g m -2 : NPK). Means ± sd (n = 4). Signi cant differences among means were identi ed by t-tests with Bonferroni adjustment. Small letters denote differences between the substrates within a certain P-treatment. Capital letters show differences in P-treatments within a speci c substrate. Means with different letters are signi cantly different at α = 5%.