Soybean (Glycine max cv. Primus obtained from Die SAAT Austria) seeds were soaked in diluted industrial Radicin inoculant (B. japonicum) (7.5 ml filled up to 100 ml with HQ water) for 5 h. For germination, seeds were sown into acid-washed (2% HNO₃) plastic boxes containing perlite and a germination solution (0.6 mM CaSO₄ ^* 2H₂O, 0.002 mM H₃BO₃, pH 6.2, conductivity of 136.2 μS cm-1).

After 1 week seedlings were transferred into pots containing a substrate mixture of vermiculite and perlite in a 5/3 (v/v) ratio, and watered with a modified half strength Hoagland nutrient solution (N-free, pH 7.2, conductivity of 1066–1160 μS cm⁻¹ between control and 0.5 mM W) (Hoagland and Arnon, 1950). After 2 weeks, seedlings were again inoculated by directly applying 1 ml diluted (as described above) industrial Radicin inoculant (B. japonicum) to each root.

To test the effects of different N regimes (N fertilization, N₂ fixation) on tungsten uptake by the plants, three different W treatments (control, 0.1 mM, 0.5 mM) were applied for 6 weeks. One half of the plants (N fed) was supplied with 10 mM KNO₃ from week two onwards, the other half (N fix) received 0.25 mM KNO₃ for 2 weeks and then were only watered with N-free nutrient solution. For each treatment 5 biological replicates where prepared. The plants were grown under controlled conditions with a 12/12 light/dark cycle at 29/21°C and an approximate photosynthetic radiation of 400 μmol m⁻¹s⁻¹ at canopy level.

Tungsten concentrations were chosen based on a germination test (Data not shown) as well as literature data. The highest tungsten concentration (0.5 mM corresponding to 91.9 mg W L⁻¹) was chosen because it was the highest concentration where plants still germinated and growth was still was not yet significantly hampered. The second tungsten concentration (0.1 mM W or 18.4 mg W L⁻¹) was chosen as intermediate between the decided on highest concentration and the control.

Plants were harvested after 49 days as they reached the full pod stage (Fehr and Caviness, 1977). A subsample of roots, nodules and leaves was cut off using a stainless steel razorblade, transferred into a precooled 2 ml Eppendorf tube and immediately put on liquid nitrogen (LN₂) for proteomic analysis and stored at −80°C.

The rest of the biomass was harvested for acid digestion and ICP-MS/OES analysis. Roots and nodules were washed and ultra-sonicated with CaCl₂ (0.01 M) for 5 min followed by further 3 min of ultra-sonication with HQ-water. Roots, shoots and nodules were separated and dried at 60°C for 1 week and subsequently ground to a homogenous fine powder using a Retsch mill (Retsch MM20).

To evaluate the effect tungsten and nitrogen regimes had on biomass production, four parameters were measured: fresh and dry weight of root and shoot as well as length of the major root and shoot. Accordingly, also the nodule biomass production was assessed by counting and weighing the nodules.

Aliquots of the dried plant material were used for determination of total starch content according to Nagler et al. (2015). Briefly, soluble sugars were removed by boiling approximately 15 mg of the plant material twice for 30 min with 80% ethanol at 80°C. The pellets were then re-suspended in 0.5 M NaOH and heated to 950°C for 30 min and subsequently acidified with 1 M CH₃COOH and digested for 2 h at 55°C with amyloglucosidase. Starch content was then assessed by determination of glucose content by further incubating 200 μl of the obtained supernatant at 30°C with 400 μl glucose oxidase reagent (50 ml Tris-Glycerin-Buffer pH 7, 15 mg glucose oxidase, 1.5 mg peroxidase, 5 mg o-Dianisidin-HCl). After 1 h, the reaction was stopped by adding 800 μl of ice cold HCl and absorbance was determined at 540 nm against a calibration curve using a glucose standard processed in the same manner as the amyloglucosidase digests.

Approximately 200 mg of dry, ground plant material was used for acid digestion. The digestion matrix contained 5 ml HNO₃, 1 ml H₂O₂ and one drop 1-Octanol. For biomass < 100 mg half of acid and peroxide was used. The digestions were performed using an open digestion unit (Velp Scientifica). Digests were filled up to a volume of 50 ml/25 ml with HQ-water resulting in ~6.5% HNO₃. If dry weight was lower than 50 mg, digestion was performed with a Multiwave 3000 (64 MG5 rotor) using only 0.5 ml HNO₃ (sub-boiled) and 0.1 ml H₂O₂. After the digestion, micro-digests were acidified with hydrofluoric acid to a final concentration of 0.01% in order to prevent polymerization or precipitation of W.

For measurements of nutrient and W content in nodules, biomass was not always sufficient to perform digestion for individual replicates. If present, nodules of treatments with low biomass (plants supplied with KNO₃ and those treated with W) were pooled. Results only represent estimates and are thus not extensively discussed (Table S2). Micronutrients and W concentrations in the digests were measured with inductively coupled plasma mass spectrometry (ICP-MS; Perkin Elmer, Elan DRCe 9000, Waltham, MA, USA); macronutrients were analyzed by ICP-OES Optima 3000 XL (Perkin Elmer).

Nitrate reductase assay was performed using fresh plant material according to Sanderson and Cocking (1964) and Stöhr and Ullrich (1997). Briefly, 6 h into the light period a composed sample was taken from the three uppermost fully expanded trifoliates, taking one leaf each. Samples were ground and homogenized in a prechilled (−20°C) porcelain mortar at with an extraction buffer (1 ml/100 mg FW), and centrifuged at 4°C with 5,000 g for 5 min. The extraction buffer was used to stabilize the enzyme in solution and contained following substances in given concentrations HEPES 100 mM, EDTA 1 mM, sucrose 330 mM, riboflavin 10 μM, Na₂MoO₄ 1 μM, casein 1 g l-1, glycerin 10% (v/v), cysteine 10 mM, DTT 1 mM, Na-ascorbate 100 mM, PVPP 1.5% (w/v), PVP 5% (w/v), PMSF 1 mM. PH of the buffer was adjusted to 7.5 with NaOH and buffer prechilled at −20°C until usage. Protein content was determined as previously described using Bradford assay.

The NR-assay was performed using a reaction buffer containing 200 μl 0.1 M phosphate buffer, pH 7.5, 50 μl of 0.01 M NADH, 50 μl of 0.1 M KNO₃ as well as 200 μl of enzyme extract making up a full volume of 500 μl (Sanderson and Cocking, 1964). A blind control for each sample was processed without adding NADH to the buffer. Samples and blinds were incubated for 30 min at 30°C.

The incubation was immediately stopped by adding 1 ml of a mixture (ratio 1:1) of 2.5% (w/v) sulfanilamide in 3.75 N HCl and 0.5% (w/v) N-(1-Naphtyl) ethylenediamin-dihydrochlorid (Snell and Snell, 1957). After another incubation of 10 min at room temperature the samples were centrifuged for 5 min with 10,000 g and the extinction determined at 540 nm. Absorption values where then converted into nmol of nitrite (NaNO₂) by using a standard curve and the previously determined protein content.

Symbiotic N₂ fixation and nodule fixation activity was estimated based on natural N isotope fractionation between plant organs using the ¹⁵N natural abundance method (Högberg, 1997). Briefly, dried and finely ground plant tissues were weighed into tin capsules and analyzed for N content and ¹⁵N:¹⁴N ratios by an elemental analyzer (EA 1110; CE Instruments, Milan, Italy) coupled via a ConFlo II interface (Finnigan MAT, Bremen, Germany) to a gas isotope ratio mass spectrometer (IRMS, DeltaPLUS; Finnigan MAT). Reference N₂ gas was calibrated to the atmospheric N₂ standard using certified IAEA reference materials. The relative abundance of ¹⁵N in plant samples was expressed in δ units, which denote the deviation in %0 of the sample ¹⁵N:¹⁴N ratio from that in atmospheric N₂. The standard deviation of repeated measurements of a laboratory standard was below 0.15%0. The δ¹⁵N of total plants (roots + shoots + nodules) were calculated as the N mass-weighted mean of the δ¹⁵N of the respective plant parts (Högberg, 1997). The δ¹⁵N signatures were used to estimate the percentage of plant N derived from the atmosphere (%NdfA) as a proxy of N₂ fixation/nitrogenase activity (Equation 1).

Where δ¹⁵N_ref represents the δ¹⁵N value (−4.49) of non-nodulating (Glycine max cv Primus) grown under the same control conditions (including 20 mM NO₃ supply) as the soy plants from the control and W treatments from the experiments as described above; δ¹⁵N_fix represents the δ¹⁵N value of the individual, experimental test plants (Glycine max cv Primus) and B is the δ¹⁵N value (−0.56) of soy grown with N₂ as the sole source of N.

Nodule fixation activity (g N fixed g⁻¹ nodule dwt) was calculated according to Equation (2)

with N_planttot being the total amount of N accumulated during the growth period (mg N) and M_nodule being the dry nodule biomass (mg).

Overall, of 2,479 identified proteins, 2,096 were detected in at least three out of five replicates of at least one treatment. The analysis of variance followed by a Tukey honest significant difference post hoc test revealed that 678 of the 2,096 identified proteins showed a significantly different regulation between W stressed plants and corresponding control, in at least one organ and one N regime (p = 0.05). A set of 156 proteins was significantly changed between the two nitrogen treatments (p = 0.05) irrespective of W concentration applied. A full list of significantly affected proteins can be found in the supplements (Tables S4, S5).

Overall 375 root-proteins significantly changed due to tungsten addition, with an almost equal amount of proteins changing in both modes of nitrogen supply as well as in each group exclusively (N fed 34.1%, N fix 32% and 33.9% in both) (Figure S2, Table S4).

Generally, exposure to high W concentrations (0.5 mM) lead to a depletion of protein abundances in roots of both N treatments (Figure 4A). Of the 247 proteins changed in N fix plants, 209 (i.e., 85%) showed a depletion while only 38 proteins accumulated. In the N fed treatment, 194 of the 255 changed proteins (i.e., 76%) were significantly depleted and 61 showed an increase in abundance. Functional categories that were significantly affected by the presence of tungsten are shown in Figure 4B. The root PCA shows a clear separation between control treatment (zero W) and high tungsten concentrations independent of N regime (0.5 mM W) through PC1, which was responsible for 76.6% of the variance (Figure 5A, Table S6B). PC2 only contributed 4.35% to the variance and neither lead to a separation between the different tungsten concentrations, nor between nitrogen fixing (N fix) and nitrogen fed (N fed) plants. In order to identify the proteins responsible for this separation, a threshold for proteins with the largest negative and positive loadings was set at −0.07 and 0.07 (Figure 5C). The PCA revealed that proteins involved in response to stress (germin-like proteins, thaumatin-like proteins, starvation associated messenger (SAM-22), protein P21, pathogenesis-related protein 10, chitinases) (Figure S4), peroxidases as well as protease inhibitory proteins (trypsin inhibitors, alpha amylase/subtilisin inhibitors) contributed most to PC1 and thus to a differentiation between the different tungsten treatments (Figure 6).

The majority (82.2%) of the 303 proteins affected by tungsten addition (0.5 mM), exclusively changed in nitrogen fixing plants (Figure S2). Only 20 (6.6%) were affected in both nitrogen treatments; 34 proteins (11.2%) exclusively changed in nodules of nitrogen fed plants (Figure 4). Of overall 269 proteins that significantly changed in nodules of the N fix treatment, 109 showed an increase, and 160 were significantly decreased at high levels of tungsten (Figure 4A, Table S5). Functional categories that were significantly affected by the presence of tungsten are shown in Figure 5B. Principal component analysis revealed that PC1 is responsible for 63.8% of the variation and separates the nodule proteins between control and high tungsten independent of N regime (Figure 5B). PC2 (13.61%) additionally separates N fix control from N fed control and the 0.5 mM W treatments. To identify the proteins most responsible for the observed separation, a cutoff of +0.06 and −0.06 was set for PC1 loadings (Figure 5D, Table S6A). The largest negative loadings for PC1 can be attributed to proteins lower in abundance in control treatments compared to plants exposed to W (Figure 6). The majority (21) of the 30 proteins with the largest negative loadings below said cutoff were rhizobia proteins depleted in high tungsten treatments as well as in control treatments of N fed plants. Figure 6 further shows that, after proteins without assigned function, especially “redox” (Leghemoglobin [A,C1,C2,C3], SOD, glutaredoxin), “metal handling” (Ferritin), “signaling” (nitrogen regulatory protein PII) and “mitochondrial electron transport, ATP synthesis” (Electron transfer flavoprotein, cytochrome b-c1 complex, ATP synthase) as well as “N-metabolism” (nitrogen fixation protein) and “transport ABC” (Amino acid ABC transporter ATP-binding) negatively contribute to PC1. The largest positive loadings can be attributed to proteins that were increased in presence of tungsten. Functional categories “stress” (Thaumatin-like protein, germin-like proteins, stress-induced proteins SAM22, probable protein P21, mRNA from stress-induced gene (H4), 18.5 kDa class heat shock protein) (Figure S4), “secondary metabolism” (Chalcone—flavonone isomerase 1A/1B-1, NAD(P)H-dependent 6'-deoxychalcone synthase, isoflavone reductase like), “peroxidases,” “endopeptidase inhibitor” (putative Kunitz-trypsin protease inhibitor), “protein degradation” (Subtilisin-like protease), as well as “hormone metabolism” (linoleate 9s-lipoxygenase) and “glutathione S-transferases” are positively contributing to PC1 (Figures 5D, 6).

Of 621 significantly changed proteins, 57 proteins changed in roots and nodules upon 0.5 mM W exposure compared to the control treatments, irrespective of N regime (Figure S3A). Of protein changed in both roots and nodules, functional categories “stress,” “peroxidases,” “secondary metabolism,” “protein degradation,” and “endopeptidase inhibitor” were most affected (Figure S3A). Additionally, proteins of “TCA, organ transformation,” “cell wall,” “glutathione S-transferases,” “glycolysis” and “metal handling” and others (Figure S3B) significantly changed in both organs. Of these Proteins 5 changed in both organs and both N regimes between control and high tungsten (Table S3).

Of the 57 proteins that changed in roots and nodules, 14 also contributed to the principle component responsible for a separation between the different tungsten treatments (PC1) of their respective organs (Figure 6). Twelve of these proteins showed same accumulation patterns and were significantly changed in at least three of four combinations (Table 2). These proteins can be assigned to the functional categories “cell wall,” “protease inhibitors,” “stress,” “peroxidases,” and “metal handling.”

Of 127 proteins changed in the zero tungsten control treatments upon different N regimes (N fix, N fed) (Tables S8A,B), 54 were constitutively elevated in the nitrogen fed plants. These proteins can be primarily assigned to “peroxidases,” “CHO metabolism” (“glycolysis,” “major” and “minor CHO metabolism,” “gluco-, galacto-, mannosidases,” “OPP”) “stress,” “not assigned,” “glutathione s transferases” and “cell wall,” as well as “redox” and “cell.” Slightly more (73) proteins were less abundant upon nitrate addition, with the majority (54 proteins) of these proteins being rhizobial proteins. Many of these proteins have still unknown function and thus result in “not assigned” being the largest category. Further functional categories that were significantly decreased are “protein” (“protein synthesis,” “protein folding,” “protein posttranslational modification”), “redox” (Leghemoglobin (A, C1, C2, C3), glutaredoxin, superoxide dismutase), “CHO metabolism” (sugar hydrolase, oligosaccharide deacetylase, etc.), “metal handling” and “amino acid metabolism” (see Tables S8A,B).

Only a small set of 32 proteins significantly changed between the two high tungsten treatments upon differing N supply (Tables S8A,B). Proteins increased due to nitrate addition (N fed) and high W were assigned to the functional categories “RNA” and “protein degradation” but also “peroxidases,” “protein,” “amino acid metabolism,” “nucleotide metabolism,” “lipid metabolism,” “N-metabolism” and “myrosinases-lectin-jacalin” as well as “not assigned.” Proteins only accumulating in N₂ fixing plants exposed to high W were assigned to “stress,” “peroxidases,” “acid and other phosphatases,” “hormone metabolism,” “signaling,” “metal handling,” “glutathione s transferases” (see Tables S8A,B).

The high dose of nitrate supplied (10 mM) to N fed plants resulted in an overall higher biomass of N fed compared to N fix plants irrespective of W addition. Depending on cultivar and growth conditions, the majority of soybean N demand can be covered by biological N₂ fixation, a certain amount of N is usually derived from naturally occurring soil N (Salvagiotti et al., 2008). Thus, while the continuously administered dose of NO₃ in N fed plans resulted in enhanced growth of N fed control plants, the initial dose of 0.25 mM NO₃ was apparently not providing additional growth in the N fix treatment.

Increasing W concentrations in the nutrient solution differentially affected root and shoot biomass of N fix and N fed soybean plants, despite similar W tissue concentrations within each W application level (Figures 1A,B). While root biomass of N fix plants was not affected, root biomass of N fed plants significantly decreased with increasing W concentrations (Figures 1A,B). This suggests that under W induced stress, a strong dependency of rhizobial activity (N fix plants) results in a change in resource allocation that allowed maintenance of root biomass to ensure symbiosis functioning and plant growth. The size of below ground biomass is crucial for maintenance of nutrient uptake during adverse condition and can have beneficial effects on plant health during stress (Glick, 2003). This is supported by the fact that despite reduced nodulation (0.5 mM W), N₂ fixation (nodule activity) of N fix soybean was not affected by increasing W concentrations (Figure 2D).

Our results correspond well with findings of Harper and Nicholas (1978), who reported a significantly decreased biomass production (~30% of the control) of 21 day old soybean plants grown on 0.4 mM W. While in non-symbiotically grown Pisum sativum Adamakis et al. (2008) found reduced root length at 200 and 500 mg L⁻¹ sodium tungstate (329.85 g mol⁻¹) after 8 days of growth, Jiang et al. (2004) already found significant reduction of organ length at concentrations higher than 10 μM sodium tungstate in Hordeum vulgare after 9 days of growth.

The necrotic blackening of roots (typically derived from oxidation of phenolic compounds) and a coral like morphology due to reduced side root development of soybean at 0.5 mM W (Figure S1) in both N regimes corresponded well with findings of other studies investigating W phytotoxicity in plant species such as oat, radish and lettuce (Bamford et al., 2011; Adamakis et al., 2012). Similar morphological changes have also been described for lead and arsenic treated soybeans (Békésiová et al., 2008) suggesting similar metal toxicity response mechanisms for W as for other heavy metals.

At high tungsten concentrations (0.5 mM), not only above ground biomass of both nitrogen treatments were negatively affected, NR activity also dropped to only 20% of the activity of healthy N fed plants (Figure 2B). This does not come to surprise, since it has been shown that, due to the competition with molybdenum as enzymatic co-factor, tungsten leads to a decreased NR activity but does not inhibit its synthesis (Heimer et al., 1969; Deng et al., 1989; Adamakis et al., 2012; Xiong et al., 2012). Harper and Nicholas (1978) found that while nitrate reductase activity of N fed soy plants (6 mM KNO₃) continuously decreased with increasing tungsten levels to 8.6% at 0.4 mM compared to control plants (23% at 0.5 mM in our experiment), acetylene reduction showed a 3-fold increase between control and 0.4 mM. The authors suggested that the observed patterns together with an increase in nodule biomass were an effect of the inhibition of the NO3- metabolism by W resulting in an increase in C:N ratio (Harper and Nicholas, 1978). Such a compensation of NR activity by increased nodulation and enhanced N₂ fixation activity of soy was also observed in one of our previous studies investigating bioavailability, speciation and phytotoxicity of tungsten in soil (Oburger et al., 2018). In the present study, however, no increase or compensation of N₂-fixation was observed, possibly since the higher dose of NO₃ of N fed plants (we compared 10 mM KNO₃ fed plants with plants receiving only 0.25 mM KNO₃ for the first 2 weeks of growth) already impeded nodule formation and thus N₂-fixation (Kanayama et al., 1990; Salvagiotti et al., 2008).

Unlike nitrate reductase, the effect of tungsten on bacterial nitrogen fixation (BNF) via nitrogenase seems not to be simply explainable by the inhibition of catalytic function of the enzyme by substitution of Mo with W, since the N₂ fixation activity (mg N fixed g⁻¹ nodule dwt.) of the fewer nodules formed under high W was not fully inhibited by tungsten (~55% activity remained at 0.5 mM W) (Figure 2D). Consistent with that, some studies found indices for an active nitrogenase even in presence of tungsten and that inhibitory effects of tungsten on nitrogenase are more pronounced in Mo deficient conditions (Kletzin and Adams, 1996; Schwarz et al., 2007; Ringelberg et al., 2009; Strigul et al., 2009). In Rhodospirillum rubrum it was shown that, although lower than activity in Mo grown cells, nitrogenase activity (measured via H₂ evolution) was not inhibited by W addition (Paschinger, 1974). Nagatani and Brill (1974) found that in Azotobacter Vinelandii specific activity of nitrogenase component one (NifD and NifK) but not component II was reduced in Mo-free, W containing medium. Furthermore, they found that relative amounts of component I protein were still present in W grown cultures (56% of normal component I protein) and that the activity of the component was reinstated 3 h after the addition of Mo to the medium. More recently, however, studies on Rhodobacter capsulatus suggest that, although possessing proton reduction activity, tungsten substituted nitrogenase is incapable of N₂ fixation and acetylene reduction (Siemann et al., 2003; Schwarz et al., 2007).

The direct effects of W on the rhizobial nitrogenase, which also exhibits a Fe-Mo co-factor could not be tested with our approach. However, proteomics, data revealed that in N fix plants relative abundance levels of N₂-fixation relevant proteins such as the nitrogenase precursors (NifD, NifT/FixU), a molybdenum cofactor biosynthesis protein (UniProt: G7D2E8) involved in the final step of Fe-Moco biosynthesis, as well as leghemoglobin, were strongly reduced in the presence of high W concentrations (~10-fold) (Figure 7 and Table S8C). NifD and leghemoglobin (Lb) are known to have an important role in abiotic stress response of nodules and have already been shown to be negatively affected by heavy metals such as Cd (Balestrasse et al., 2004; Marino et al., 2013). The decrease of the redox sensitive leghemoglobin potentially affects nitrogenase expression and thus activity by causing alterations in nodule oxygen concentrations (Marino et al., 2013). This is consistent with the depletion of proteins involved in Fe-Mo cofactor biosynthesis as well as the decline in the amount of N derived from N₂ fixation in N fix plants and indicates that W might be indirectly affecting nitrogenase abundanceby inducing oxidative stress (Figures 2A, 7). Considering this decrease in abundance of proteins important for N₂ fixation, the still considerable N₂ fixation activity (mg N fixed g⁻¹ nodule dwt.) suggests an increase in substrate turnover in W stressed plants. Furthermore, it is still not fully understood how Mo (and W) is taken up and transported into and between cells of higher plants as well as distributed or stored internally (Hagen, 2011; Bittner, 2014; Tejada-Jiménez et al., 2017; Vigani et al., 2017; Gil-Díez et al., 2018). Besides a proposed transport via sulfate transporters, three molybdate specific transporters MOT1(Chlamydomas reinhartii and Arabidopsis thaliana) and MOT2 (Chlamydomas reinhartii) have been identified (Tejada-Jiménez et al., 2013). More recently, two molybdenum transporters (MtMOT1.2 and a nodule-specific MtMOT1.3) have been identified in Medicago truncatula, living in symbiosis with sinorhizobium meliloti. While MtMOT1.2 is said to transport Mo through endodermis cells into the symplast, MtMOT1.3 was shown to mediate the transport into nodule cells and thus appears to be key in Mo supply for nitrogenase biosynthesis (Tejada-Jiménez et al., 2017; Gil-Díez et al., 2018). Interestingly, these two transporters are highly specific for molybdate anions, and appear not to be able to transport similar anions such as sulfate and possibly also tungstate (Gil-Díez et al., 2018). Furthermore, molybdenum an possibly also tungsten are proposed to be delivered to bacteroides via SST1 transporters and bacterial modABC transporters (Gil-Díez et al., 2018). Interestingly, we found rhizobial ABC transporters (mediating amino acid and phosphate transport), as well as proteins involved in ATP synthesis and N- and amino acid metabolism to be depleted at high tungsten (Table S7D).

Taken together, our data indicates that high W concentrations (0.5 mM) hamper N-assimilation and N₂-fixation differently but to a similar degree with no clear advantage for either one of the N-regimes. However, our data suggests that while NR activity is inhibited by the incorporation of W into the enzyme, N₂ fixation is indirectly affected by a decrease of Lb and NifD and thus reduced nitrogenase synthesis. This seems to be consistent with Nagatani and Brill (1974), who showed that although W reduced nitrogenase synthesis, specific activity of nitrogenase was only inhibited if Mo was omitted from the medium.

Similar to other HMs, the presence of increasing tungsten resulted in changes of nutrient tissue concentrations in soybean. Nutrient level changes induced by W could potentially be driven by (i) competition for uptake through the same transporters as other anions (phosphate, sulfate, molybdate), (ii) inhibition of uptake due to the reduction of root growth, (iii) decrease of nutrient availability due to polymerization with tungsten (particularly P and S) in soil solution and/or in acidic compartments of the cell such as the vacuole (Foy et al., 1978; Bednar et al., 2008; Johnson et al., 2010; Tran et al., 2010; Adamakis et al., 2012).

Several micronutrients (Mo, Cr, Fe, Cu) accumulated in the presence of high tungsten concentrations (0.5 mM W), particularly in soybean shoot tissue; however, these trends were not consistent across the different W and N treatments (Table 1). An increased micronutrient demand could be explained by their importance for molybdenum-specific metabolic processes, since there is a proposed crosstalk between molybdenum, copper and iron metabolism (Hänsch and Mendel, 2009). Copper as well as iron are involved in MoCo biosynthesis since some of the molybdoenzymes require iron prosthetic groups, and copper is essential for the formation of MoCo intermediates (Hänsch and Mendel, 2009; Bittner, 2014). In order for the plant to keep the nitrogen metabolism running, the production of NR apo-enzymes and thus enhanced uptake of Fe, Cu and Mo is required. Consequently, the lack of active NR combined with enhanced Fe, Cu and Mo uptake indicates that NR function rather than NR biosynthesis itself is disturbed (Heimer et al., 1969; Deng et al., 1989; Xiong et al., 2012). A similar tungsten induced increase in Mo shoot, root and nodule concentrations was also observed when soybean was grown on limed soils spiked with increasing concentrations of metallic W (Oburger et al., 2018). This is in line with Strigul et al. (2009) who found that the presence of tungsten promotes hyperaccumulation of molybdenum and vice versa. Fe is among the most important nutrients for N₂-fixation as it is also needed for synthesis of the Fe-S complex of the nitrogenase as well as for leghemoglobin, which is the reason why iron levels are constitutively higher in the roots of N fix control plants (Tang et al., 1992). While iron levels increased in shoots of both treatments at W 0.5 mM, there was no significant change Fe concentrations in roots at the highest tungsten concentration compared with the respective control (Table 1). This in combination with the significant reduction in S levels in roots indicates that, unlike nitrate reduction, N₂ fixation was negatively affected by decreased co-factor biosynthesis, which also corresponds to our proteomic data. The increase in Fe and Cu could have led to disturbances of the oxidative balance, since both are redox active transition metals that take part in fenton and fenton-like reactions, which have highly reactive oxygen species as a product and thus have detrimental effects on the organism (Mithöfer et al., 2004; Tran et al., 2010).

At high W concentrations, we found a significant decrease in levels of Mg, Ca and Mn in shoots, and Mg, K and S in roots of both N-regimes (Table 1). Nevertheless, concentrations of all decreased macro-nutrients (K, Mg, Mn, Ca, S) were within the physiological ranges reported in literature (Marschner, 2012). The decrease in root S concentrations, as indicated by our proteomic data, might be due to the depletion of several ABC transporters and ATPases in response to tungsten in N fix nodules and roots of both treatments (Table S7D). Since the cellular uptake and transport of W is suspected to be mediated by transporters of the same family as oxyanions such as sulfate and phosphate, the observed disturbances in sulfate uptake were most likely a result of tungsten induced depletion or blockage of these transporters (Bevers et al., 2009). Considering that plants have been shown to increase S uptake for increased production of phytochelatins such as glutathione during HM exposure (e.g., Cd) (Nocito et al., 2006). The found depletion, especially in roots, thus indicates reduced chelation and reduced capacity to combat oxidative stress (Ben Ammar et al., 2008). Besides limitation in S, also a decrease in Mg can have consequence for the plants toxicity resistance, since Mg is a prerequisite for the functionality of the enzyme glutathione synthetase, which is the key enzyme of glutathione biosynthesis, and has been shown to alleviate HM stress by decreasing negative electrical potential (Marschner, 2012; Rengel et al., 2015). Furthermore, Mg is most notably required as central atom of chlorophyll, as well as for a variety of different enzymes such as fructose-1.6-biphosphate, involved in sugar-starch partitioning and glutamine synthetase, involved in ammonia assimilation, as well as other physiologically important reactions such as transfer of phosphate and carboxyl groups (Marschner, 2012). Since Mn serves as co-factor in allantoate amidohydrase, a ureide degrading enzyme (allantoin, allantoate as major transport form of symbiotically fixed nitrogen in soy), the difference Mn leaf tissue concentrations between N fix and N fed can be explained by the higher demand for Mn of N₂ fixing legumes (Marschner, 2012). Accordingly, The decrease of Mn in N fix leaf tissue at the highest W concentration might be due to the observed reduction of N₂ fixation with increasing tissue W. Calcium, which was also decreased at high W, has important function as structural and regulatory component of macromolecules (especially cell walls), as well as in osmo-regulation and cation-anion balance and secondary messenger (Marschner, 2012). The cations Mg, Mn, K, and Ca are taken up by other channels and transporters than S and P (and prospectively also W), which indicates that W causes a depletion of these nutrients through other mechanisms. We could identify a potassium channel (UniRef100 = I1NEH6) which was significantly depleted at high W, which could explain the observed decrease in root K concentrations.

Although not reflected in total P concentrations, one reason for the decrease in macronutrient tissue concentrations (Mg, K, Ca, Mn) could be a W induced phosphate deficiency (Kavka and Polle, 2017). It is known that W polymerizes with itself or other ions in acidic environments (Strigul et al., 2010; Oburger et al., 2018) and also in biological systems (Johnson et al., 2010). Polymerization of W with phosphorous in the acidic compartments of the cell, such as the vacuole, which is where HMs but also nutrients are normally sequestered and stored, could lead to P immobilization and thus P deficiency. This would also explain the increase of P in roots in both N treatments as well as the accumulation of purple acid phosphatases in N fix nodules of the high tungsten treatments (Tran et al., 2010) (Table 1, Figure 4B). Acid phosphatases are enzymes that mediate vacuolar P cycling but also catalyze the hydrolysis of organically bound phosphate in extracellular environments and have been shown to increase under phosphate malnutrition as well as oxidative and salt stress (Tabaldi et al., 2007; Tran et al., 2010; Solanki and Dhankhar, 2011; Kavka and Polle, 2017). N₂ fixing legumes possess higher acidic phosphatase activity under P limitation or at high N supply, as nodules are strong P sinks, and P is essential for both, nodulation and N₂ fixation (Zahran, 1999; Png et al., 2017). Additionally, it has been shown that N₂ fixation dependent plants require more phosphorus than those additionally supplied with nitrogen, which becomes especially critical for the symbionts during P limitation (Zahran, 1999). The fact that acid phosphatases only increased in N fix plants mirrors the higher demand of P for N fixation and indicates that the plants tried to meet the bacterial P requirement. This in turn potentially led to the observed maintenance of N₂ fixation and thus the only slightly reduced N₂ fixation activity.

High W concentrations strongly increased starch content in shoots of both N regimes and in roots of N fed plant (Figure 3). It has already been reported that HM can lead to an accumulation of starch and soluble sugars accumulate in aerial plant parts due to HM induced blockage or/and reduction of source to sink transport (Pahlsson, 1989).

One reason for an increase in starch tissue concentration could be the aforementioned polymerization of tungsten and phosphate in the vacuole, since an often reported effect of phosphate starvation is an increase in starch and sucrose levels (Hermans et al., 2006; Rouached et al., 2010). Phosphate deficiency has direct effects on plant physiological processes such as glycolysis, respiration and photosynthesis, ATP-synthesis and calvin cycle, which were found to be significantly affected in our study (Figure 4B and Tables S4, S5, S7A,D). As discussed above, we additionally found an increase in root phosphorous and iron content as well as an increased root:shoot biomass ratio in N fix, all consistent with symptoms of phosphate starvation (Table 1 and Table S1) (Hammond and White, 2008; Rouached et al., 2010). Root biomass, nutrient and starch content data (Figures 1B, 3B, Table 1)indicates that N fix plants were better able to cope with W induced P starvation due to elevated levels of acidic phosphatases, which might have resulted in a maintained carbon supply to roots of N fix plants (Figure 1B).

Besides polymerization with nutrients (e.g., P), tungsten could bind to -SH groups of enzymes and other proteins (Sharma and Dietz, 2009; Aloui et al., 2011; Solanki and Dhankhar, 2011; Hossain et al., 2012), similar to other heavy metals. We found higher levels of alpha-amylase/subtilisin inhibitors (serine protease) and kunitz-trypsin protease inhibitory proteins as well as some respective proteases (subtilisin proteases) in the presence of high W in both N regimes, however, increase was more pronounced in N fed plants (Figure 6, Table S7C). The increase of these bifunctional protease inhibitory proteins could be another reason for the increased starch content. It has already been shown that during germination proteins involved in starch breakdown such as α-amylase, β-amylases are inhibited by heavy metals such as Zn, As, Cr and Cd (Rodríguez-Celma et al., 2010; Solanki and Dhankhar, 2011; DalCorso et al., 2013).

HMs have been reported to act as an impediment to normal metabolic functioning on various levels (Emamverdian et al., 2015). Several HM induced plant stress response mechanisms are not specific for HM stress but can be observed due to a variety of biotic and abiotic environmental stresses (e.g., herbivore and pathogen attack, salt, drought and logging).

We identified a set of twelve proteins that seem to be majorly involved in tungsten stress response, and accumulated in roots and nodules at high W, irrespective of N regime (Table 2). Most of them were general stress responsive proteins, so called pathogenesis related proteins (PR proteins), which have been shown to respond to pathogen attack and biotic stress, and to accumulate during abiotic stress such as nutrient deficiency or heavy metal stress (Didierjean et al., 1996; Maksymiec, 2007; Ahsan et al., 2009; Hossain and Komatsu, 2012; Sharmin et al., 2012). Among the identified proteins were two general PR proteins (C6T3A2, C6SYB8), a germin-like protein (C7S8D4), stress associated message SAM22 (C6T3A2) as well as a thaumatin-like protein (I1K6M2, I1L927). Thaumatin-like proteins are PR-5 group proteins that have been reported to be activated due to biotic as well as abiotic stress such as chemical treatment, high UV light, wounding, salt stress and HM toxicity (Frendo et al., 1992; Didierjean et al., 1996; Führs et al., 2008; Aloui et al., 2009; Tachi et al., 2009). The exact role of PR-5 group proteins is still unknown, it is proposed that they have similar functions to α-amylases/trypsin inhibitors, zeamatin (α-amylase inhibitors), osmotin-like proteins (increased salt and drought tolerance, leads to increases of proline and quenches ROS) as well as anti-fungal proteins (van Loon, 1985; Malehorn et al., 1994; Zhu et al., 1995; Koiwa et al., 1999; Kishor et al., 2015). Stress induced protein SAM22 (starvation-associated messessager) was found to be increased by various elicitors such as salicylic acid, hydrogen peroxide, methyl viologen or chitosan as well as depletion of cytokinin and auxin (Crowell et al., 1992). Germin-like proteins (GLPs), have been found to be accumulating due to pathogen attack, temperature stress, water stress, heavy metals and nutrient starvation (N, P, and K) but also during germination and senescence (Dunwell et al., 2008). Their exact function has so far not been described, however, they are proposed to have Mn-Superoxide dismutase (SOD) activity. SODs dismutate superoxide to H₂O₂ in order detoxify the reactive oxygen, an thus are proposed to be involved in cell wall stiffening by supplying H₂O₂ for lignification (Bernier and Berna, 2001; Dunwell et al., 2008). Accordingly, we found an increase of peroxidases (I1J7M4, Q9XFI8) and polyphenol oxidase (K7KJM5) across all treatments, and laccase (in roots) which mediate the formation of lignin from its precursors. Peroxidases and polyphenol oxidases are furthermore involved in detoxification of oxygen radicals (Pourcel et al., 2007). It is proposed that lignin biosynthesis is induced under HM stress in order to strengthen and stiffen cell walls and to reduce root growth, reducing uptake of the HM into the plant (Gall et al., 2015). Quite fittingly, we also observed that chitinases were increased in roots of both treatments. These, fungal and bacterial cell wall degenerating enzymes, are thought to be induced during HM stress by changes in cellular oxidative status (Javed et al., 2013) as they contribute to water retention, decrease cell wall elasticity and permeability for metals (Kasprzewska, 2003; Mészáros et al., 2014). However, in addition, we found another cell wall associated enzyme, a ß-expansin (I1MV71) to be highly responsive to tungsten. Interestingly, expansins lead to a loosening of the cell wall and expansion and thus cell growth, which seems antagonistic to the observed indices of lignification. Additionally, we observed increased levels of pectin esterases and annexin in nodules of N fix plant, which both have been shown to be involved in Al tolerance in medicago truncatula (Chandran et al., 2008). Pectin plays an essential role in HM binding and chelation. Pectin esterases mediate the de-esterification of pectin, resulting in higher affinity to bind to HM, and thus sequestering them into the cell wall (Gall et al., 2015). However, expansins have been reported to be involved in environmental stress tolerance (heat, drought, and nutrient stress i.e., increase in P uptake efficiency in soy) as well as to interact with cell wall bound peroxidases and thus to improve tolerance to oxidative stress (Marowa et al., 2016). This does not only hint at a tungsten induced oxidative imbalance but again, at an effect of tungsten stress on phosphate uptake and utilization which is also indicated by the increase of expansins (but also GLPs) since they have been shown to be involved in alleviation during phosphate deprivation. As discussed above, polymerization of P with W in the vacuole could lead to cellular phosphate depletion and thus an activation of the plants phosphate deficiency response. An expansin mediated increased P uptake efficiency would also explain the increase in phosphate concentrations in roots.

Since most of the above discussed proteins serve a variety of different functions, it cannot conclusively be answered which specific function they serve during W stress. It appears, however, that one major site of W stress response is the reorganization of the cell walls in roots. The accumulation of expansin concurrent with lignification of secondary cell walls could hint at two different strategies to alleviate W toxicity: I.) by chelation with pectin and sequestration into cell walls in nodules, possibly to reduce effects on N₂ fixation, II.) by enforcing cell walls via lignification, making them less permeable for W (roots and nodules).

However, reactive oxygen species such as hydrogen peroxide (H₂O₂) and O2- are not only essential for lignification processes, they can also be transformed into highly reactive hydroxyl radicals in so called Fenton or Fenton-like reactions although in presence of redox active transition metals such as Cu(I), Fe(II) (Mithöfer et al., 2004), but also Cr III and VI (Sharmin et al., 2012) and, in vitro, also MoO42- (Aubry and Cazin, 1988; Aubry et al., 1989; Boehme and Brauer, 1992; Popivker et al., 2013). The observed increase in abundance of a variety of peroxidases but also other enzymes involved in antioxidative defense (glutathione S-transferases (GST) in both N regimes; monodehydroascorbate reductase (MDHAR) in N fed plants) as well as an altered regulation of the glyoxalase system (lactoglutathione lyase in nodules of N fix only), (Tables S4, S5) indeed indicate that tungsten interferes with oxidative homeostasis by inducing accumulation of ROS. Interestingly, many of the above discussed PR proteins have been shown to also accumulate during oxidative stress (SAM22, Thaumatin-like, chitinases).

Excessive amounts of heavy metal ions have been found to lead to oxidative damage through (i) increased production of ROS, (ii) the aforementioned transformation of ROS (H₂O₂ and O2-) to highly toxic hydroxyl radicals, (iii) protein degradation due to binding to –SH groups, (iv) inhibition of enzymatic activity, and (v) displacement of nutrient cations such as K, Ca, Mg, Fe and Mn from binding sites (Hall, 2002; Maksymiec, 2007; Ahsan et al., 2009; Hänsch and Mendel, 2009; Hossain and Komatsu, 2012). Since tungsten and molybdenum share certain chemical and structural similarities (Lassner et al., 2000; Koutsospyros et al., 2006; Bevers et al., 2009), excessive W might also contribute to ROS formation in a similar manner to Mo and other transition metals and take part in Fenton-like reactions. However, direct evidence for these reactions in presence MoO42- to also happen in cellular environments, as well as between tungsten and H₂O₂ has not been presented to date.

Nonetheless, we also found that many proteins reported to be accumulating during oxidative stress such as superoxide dismutase (SOD), catalase, glutathione reductase, ascorbate peroxidase and thioredoxin remained unchanged or were depleted in both N regimes at the highest concentrations of tungsten (Tables S4, S5). Depending on HM-dose, growth conditions and organism, differential expression of SODs, APX, peroxidases and other enzymes involved in antioxidative defense was reported for several heavy metals, such as Cr (Sharmin et al., 2012), Cu (Lingua et al., 2012), Cd (Sandalio et al., 2001; Schützendübel et al., 2001; Ferreira et al., 2002; Sobkowiak and Deckert, 2006; Kieffer et al., 2009), As (Requejo and Tena, 2005) and Al (Guo et al., 2004; Sharma and Dietz, 2009). The depletion of antioxidative enzyme levels in our experiment (Tables S4, S5) might be due to the severity of stress induced by high (0.5 mM) W concentrations or the consumption of metabolites such as antioxidative glutathione and ascorbate via other paths (Sharmin et al., 2012). Accordingly, our results indicate that the detoxification of ROS via SODs and the glutathione-ascorbate cycle as well as catalase and peroxidases (Gratão et al., 2005; Hossain et al., 2012) was either blocked or dysfunctional at the administered tungsten concentration.

Furthermore, we observed an accumulation of glutathione S-transferases (GSTs) under W stress (Figure 4B), which were found to be increased in the presence of other HM like Al, AS, Cd and Cu (Marrs, 1996; Ahsan et al., 2009; Hossain et al., 2012). GSTs are said to have an antitoxic function by catalyzing the conjugation of GSH to electrophilic xenobiotics as well as endogenous secondary metabolites (e.g., anthocyanins, phytoalexins) for sequestration into the vacuole (Edwards et al., 2000; Pompella et al., 2003). In addition, they act as non-enzymatic carrier proteins for intracellular transport (e.g., IAA) and can possess a glutathione peroxidase activity and directly detoxify hydroxyl radicals (Marrs, 1996).

In order to combat oxidative stress, plants need to reduce ROS production by keeping the mitochondrial electron transport chain, a major source of ROS, sufficiently oxidized and bringing the system back into balance (Møller, 2001). The upregulation of glycolysis and TCA-cycle is a prerequisite for adequate NADH supply to the mitochondrial electron transport and can thus be expected (Kieffer et al., 2009; Hossain and Komatsu, 2012; Sharmin et al., 2012). However, our proteomic analysis revealed quite the opposite. Almost all key enzymes of glycolysis and TCA-cycle were depleted roots and nodules (Figure 4). This general metabolic breakdown apparently resulted in a disruption of the mitochondrial electron transport chain and ATP synthesis probably making it even more difficult for the cell to reinstate normal redox conditions. With minor exceptions, proteomic data from both N regimes showed this breakdown of TCA-cycle, glycolysis, lipid metabolism and mitochondrial electron transport (Figure 4B). Nevertheless, different sets of enzymes were depleted at 0.5 mM W depending on the mode of N supply. While in N fix and N fed roots of the high W treatment almost all key enzymes of primary metabolic pathways were depleted, there was an accumulation of some enzymes involved in glycolysis, lipid metabolism, TCA-cycle and amino acid metabolism in N fix nodules (Figure 4). This indicates that although metabolic processes of N fix and N fed plants were similarly affected, certain energy producing metabolic functions such as sugar metabolism were maintained or even enhanced in N fix nodules at high W concentrations, which might have contributed to uphold low levels of N₂ fixation and plant growth (Figures 1,2).

Our data indicated a symbiotically enhanced activity of hormone biosynthesis and secondary metabolism in nodules, not observed in roots in the presence of high W. In contrast to roots of N fed plants, where oxygen radicals were most likely scavenged by phenol/ascorbate/peroxide system, plants depending on functional symbiosis additionally enhanced flavonoid biosynthesis (chalcone synthase, chalcone isomerase, dihydroflavonol 4-reductase and isoflavone reductase), which also have been shown to function as radical scavengers (Table S5). This suggests different coping strategies between the two nitrogen treatments. Hale et al. (2002) showed that anthocyanins change color because of a complexation with either molybdenum or tungsten in Brassica and that they are most likely involved in the sequestration of transition elements to peripheral cell layers. However, the authors did not find a direct correlation between tungsten tolerance and anthocyanin content. In corn, flavonoids have been found to be involved in aluminum resistance and are thought to be able to chelate HM (Winkel-Shirley, 2002).

In nodules of the N fix treatment, lipoxygenases and 12-oxophytodienoate reductase, both enzymes involved in the biosynthesis of jasmonate (i.e., phytohormon) were accumulated. The jasmonate biosynthetic pathway has been found to be induced by a variety of biotic and abiotic stressors such as wounding, osmotic stress, herbivore attack as well as heavy metals (Maksymiec, 2007). Noteworthy, they were recently observed to be induced in leaves of N fix compared to non-nodulated N-fertilized Medicago truncatula plants and proposed to be involved in drought stress alleviation (Staudinger et al., 2016). An increase in jasmonic acid and methyl jasmonate results in a decrease in growth and photosynthetic activity, as well as an increased production of secondary metabolites (e.g., flavonols) and an accumulation of so called jasmonate inducible proteins (JiPs) (Maksymiec et al., 2005). Such Jips, two Kunitz-trypsin protease inhibitors (I1KW54, I1MI59), were among the 12 identified tungsten responsive proteins that were accumulating across all organs and treatments. Alpha-amylase/subtilisin inhibitor activity was reported to be induced by various heavy metals as well as other abiotic stressors (Frendo et al., 1992; Solanki and Dhankhar, 2011). In drought tolerant soybean Kunitz trypsin protease inhibitors and acid phosphatase showed a significant increase (37-fold and 114-fold increase respectively) during severe drought (Yu et al., 2016). Our data suggests that both proteins are also involved in tungsten stress response.

Even though, severe W stress seems to affect G. max independent of symbiotic functioning, these findings suggest a possible explanation as to how N fix plants maintained their root growth and nodule fixation activity at lower W levels.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.