ICP-MS based metallomics and GC-MS based metabolomics reveals the physiological and metabolic responses of Dendrobium huoshanense plants exposed to Fe3O4 nanoparticles

It is found that the growth of Dendrobium huoshanense was dependent on Fe3O4, while the bioavailability of plants to ordinary Fe3O4 was low on the earth. In order to improve the growth, quality and yield of D. huoshanense, we used Fe3O4 NPs (100 or 200 mg/L) that was easily absorbed by plants as nano-fertilizer to hydroponically treat seedlings of D. huoshanense for 3 weeks. Fe3O4 NPs induced not only earlier flowering and increased sugar content and photosynthesis, but also stressed to plants, increased MDA content and related antioxidant enzymes activities. Inductively Coupled Plasma Mass Spectrometry (ICP-MS) revealed that Fe3O4 NPs caused a significant accumulation of Fe and some other nutrient elements (Mn, Co, B, Mo) in stems of D. huoshanense. Metabolomics revealed that the metabolites were reprogrammed in D. huoshanense when under Fe3O4 NPs exposure. Fe3O4 NPs inhibited antioxidant defense-related pathways, demonstrating that Fe3O4 NPs have antioxidant capacity to protect D. huoshanense from damage. As the first study associating Fe3O4 NPs with the quality of D. huoshanense, it provided vital insights into the molecular mechanisms of how D. huoshanense responds to Fe3O4 NPs, ensuring the reasonable use of Fe3O4 NPs as nano-fertilizer.

It is found that the growth of Dendrobium huoshanense was dependent on Fe O , while the bioavailability of plants to ordinary Fe O was low on the earth. In order to improve the growth, quality and yield of D. huoshanense, we used Fe O NPs ( or mg/L) that was easily absorbed by plants as nano-fertilizer to hydroponically treat seedlings of D. huoshanense for weeks. Fe O NPs induced not only earlier flowering and increased sugar content and photosynthesis, but also stressed to plants, increased

Introduction
Dendrobium huoshanense, a perennial epiphytic herb, is commonly used as a precious high-end medicinal herb and is only found in the mountainous areas of the northern Yangtze River in China. Harboring diverse chemical constituents including polysaccharides, alkaloids, etc. (1), it has several medicinal properties, mainly including hypoglycemic, hypolipidemic, hepatoprotective, immunoregulatory, cataract prevention and tumor proliferation inhibition (2). Wild-type D. huoshanense grows on rocks, but better on magnetite rocks than ordinary ones (3). The effects of metallic elements upon D. huoshanense have been investigated. It has found Fe 2+ strongly effected the proliferation (4) and polysaccharide GRAPHICAL ABSTRACT | biosynthesis (5) of its protocorm-like bodies. Therefore, there is an urgent need to study the interaction of Fe on plants (6). Although Fe element is enough in the Earth's crust, the Fe 3+ is not easily absorbed by plants in nature since its insolubility (7).
Nanoparticles (NPs) are extremely tiny, with multiple binding sites and a large surface area, rendering them excellent characteristics as nanocarriers for bioactive molecules (8). The unique photoelectric, physiochemical, and catalytic properties of NPs make them an excellent tool to improve the growth and photosynthesis of plants (9). Accordingly, nanotechnology could be employed to better protect plants and make their production sustainable (10). Even more materials are converted into nanoparticles to promote growth and stress tolerance of plants through foliar spraying, hydroponics or soil route.  (14), with the released iron is a potential key resource for plants (15). Many previous studies have illustrated that iron-based NPs have effects on plant growth. Liu et al. (16) cultured seedlings of Lactuca sativa with FeO X NPs, which significantly promoted the length of lettuce shoot. The unique peroxidase (POD)-like activity of Fe 3 O 4 NPs can increase the chlorophyll content and biomass of Quercus macdougallii (17) and increase POD enzyme activity of Cucumis sativus seedlings (18), respectively. However, iron-based NPs can also have phytotoxic effects on plants. Lee et al. (19) observed that Arabidopsis root elongation was inhibited with Fe 3 O 4 NPs. Ding et al. (20) found that Fe 3 O 4 NPs significantly hindered the growth of Eichhornia crassipes, markedly reducing its chlorophyll content and catalase activity while increasing its malondialdehyde (MDA) content. Therefore, it is paramount that we investigate how to apply Fe 3 O 4 NPs as nano-fertilizer in the growth of D. huoshanense. Considering uncertainty about Fe 3 O 4 NPs' application, it is primarily necessary to study the metabolite reprogramming of D. huoshanense under Fe 3 O 4 exposure.
As a powerful high-throughput tool, metabolomics can capture and analyze the metabolic status of plants in targeted (21) or untargeted (22), aiming at detecting small molecular metabolites in cells or tissues under specific conditions (23). As the products of gene expression, metabolites can reflect mild variances in gene and protein expression level. Yan et al. (24) treated Zea mays with Fe 3 O 4 NPs and used GC-MS to detect their root metabolites, which revealed the exposure to Fe 3 O 4 had significant effects on maize root growth and development, as well as cell membrane integrity.
Here, D. huoshanense seedlings were grown hydroponically with Fe 3 O 4 NPs for 3 weeks, and their biochemical parameters, biomass, chlorophyll content, lipid peroxidation, and polysaccharide contents were detected to evaluate the influences of Fe 3 O 4 NPs. ICP-MS was chosen to determine the content of metal elements. An untargeted metabolomics method using GC-MS was applied to detect the content and composition of metabolites responsing to Fe 3 O 4 NPs. Our study provides valuable information to improve the quality of D. huoshanense.

Plant materials and chemicals
The biennial D. huoshanense seedlings, from Huoshan County (Anhui Province, China), were grown at 25 ± 2 • C and 23 ± 2 • C during the day and night, respectively, under relative humidity at 60-70% and a light/dark cycle at 14 h/10 h in the greenhouse of Anhui University of Chinese Medicine (Hefei, China). Fe 3 O 4 NPs were bought from the Shanghai Macklin Biochemical Co., Ltd. (Shanghai, China) with 99.5% purity; their original size was ca. 10-30 nm. Standards of (+)glucose, mannose, glucosamine hydrochloride (>98% purity) were the products of Chengdu Push Bio-technology Co., Ltd. HPLC-grade methanol was bought from Oceanpak. All of other chemicals were analytical grade. Water (HPLC grade) was prepared by a purification system from Pall Filter Co., Ltd. (Beijing, China).

Experimental design
Uniformly selected D. huoshanense seedlings were transferred to a hydroponic system. Every six clusters of seedlings were fixed on a polystyrol-plate in a pot which contained the 1/2 Hoagland solution (25). To prepare Fe 3 O 4 NPs stock solutions at 100 and 200 mg/L in nanopure water, Fe 3 O 4 NPs were dissolved in Hoagland solution and sonicated at 45 kHz for 60 min in an ice bath (AS20500BDT, Tianjin Autoscience Instrument Co., Ltd, China), and these applied as treatment groups to D. huoshanense seedlings, while the same solution lacking NPs was the control. All seedlings cultivated in a greenhouse for 21 days with their nutrient solution changed every 3 days until harvested. The growth of D. huoshanense was observed macroscopically at 0, 7, 14, and 21 days of treatment.

Biomass and photosynthetic parameter measurements
The seedlings of D. huoshanense were treated with their solution for 21 days, and then washed thoroughly with running water for more than 30 min. At last, ultrapure water was used to remove traces of nutrients and Fe ions on their roots' surface by 3 times. After wiping the plants with paper towels, their fresh biomass was measured before oven-drying them (at 65 • C for 72 h). Photosynthetic pigment content of seedlings was quantified in situ, at the 0, 7, 14, and 21 days (26).
The absorbance of total carotenoids and chlorophyll a and b were, respectively, detected at 470, 663, and 645 nm, with a spectrophotometer.

Malondialdehyde and enzyme crude extract
The crude solutions for Malondialdehyde and the resistance-related enzymes superoxide dismutase (SOD), catalase (CAT), and peroxidase (POD) were extracted as described (27) with several modifications. Leaf samples (each 0.50 g) were ground with an extraction buffer (0.05 M phosphate buffer, pH 7.8) (2 mL) at 4 • C, then transferred into centrifuge tube with another 6 mL of extraction buffer. Each mixture was cold centrifuged at 10,000 × g for 20 min. Their supernatants were used to enzyme activities and the MDA content measurement in each treatment group (or control); all measurements were performed in triplicate.

Lipid peroxidation
MDA content was determined using the 2-thiobarbituric acid (TBA) reaction (28). Two milliliters of the extracted supernatant were mixed with 2 mL of 0.6% TBA, followed by 4 mL of 5% trichloroacetic acid (TCA), mixed and reacted on a boiling water bath for 10 min. The tube was removed to be cooled and centrifuged, and the supernatant was taken to determine its absorbance values at wavelengths of 532, 600, and 450 nm, respectively. The MDA content was expressed as the amount of substance per gram of fresh leaves (µmol/g Fw).
Total SOD activity was quantified by the inhibition of the of nitro blue tetrazolium (NBT)'s photochemical reduction (29). The sample tube reaction system consisted of 1.5 mL 0.05 mol/L phosphate buffer (pH 7.8), 0.3 mL 130 mmol/L methionine (Met), 0.3 mL 0.75 mmol/L NBT, 0.3 mL 0.1 mmol/L EDTA-Na 2 , 0.3 mL 20 µmol/L riboflavin, 0.05 mL enzymatic extract, and 0.25 mL distilled water in a total volume of 3 mL for the reaction mixture. Enzymatic extract in reaction mixture replaced by phosphate buffer was as control tube. Place the sample tube under fluorescent lamp for irradiation (50 µmol/m 2 /s) for 10-20 min (adjust the reaction time appropriately according to the enzyme activity), shade the control tube from light, and record the absorbance at the wavelength of 560 nm. SOD activity was 50% inhibition of NBT reduction by superoxide produced from photo-reduced riboflavin with oxygen. The total SOD activity was expressed in units per gram of fresh leaves (u/g Fw).
The guaiacol method was used for the determination of POD activity (29). A reaction mixture was prepared using 50 mL 0.05 mol/L phosphate buffer (pH 7.8), 28 µL guaiacol, and 19 µL 30% H 2 O 2 (v/v); 3 mL of the reaction mixture solution was placed into a cuvette with a 1 cm path length. The increase in absorbance at the wavelength of 470 nm was recorded over 4 min at 30 s intervals after the addition of 1 mL enzyme extract. One unit of POD was defined as the amount of enzyme produced a 0.01 absorbance change per minute per gram of fresh leaves at 470 nm [u/(g·min) Fw].
The CAT activity was determined by Gao method (30). A reaction mixture was prepared using 1.5 mL 0.05 mol/L phosphate buffer (pH 7.8), 0.2 mL enzyme extract, and 1 mL ultrapure water; 2.7 mL of the reaction mixture solution was placed into a cuvette with a 1 cm path length. The decrease in absorbance at the wavelength of 240 nm was recorded over 4 min at 30 s intervals after the addition of 0.3 mL 0.1 mol/L H 2 O 2 . One unit of CAT was defined as the amount of enzyme that lowers 0.1 absorbance per minute per gram of fresh leaves at 240 nm [u/(g·min) Fw].

Bioactive components in D. huoshanense
According to the current study, the main bioactive components of D. huoshanense are polysaccharides, alkaloids and flavonoids, among which the polysaccharides are mainly composed of glucose and mannose. Using glucosamine as internal standard, the water-soluble polysaccharides was obtained from D. huoshanense stems via the hotwater extraction and ethanol precipitation (31). And the monosaccharide solution was obtained by hydrolysis of polysaccharide at high temperature (for 1 h at 110 • C). The pre-column derivatization was used by 1-phenyl-3-methyl-5-pyrazolone (PMP) between a given sample solution and standards (mannose and glucose), these then, respectively, washed with chloroform for 2 or 3 times. The supernatant of the water layer was injected for high performance-liquid chromatograph (HPLC) (Agilent 1260 System, Agilent Technologies Inc., CA, USA), to determine each monosaccharide's content.
Following a method described by Wang et al. (32), the contents of total alkaloids and flavonoids in D. huoshanense were determined. The former was quantified according to .
the absorbance value at 620 nm of Dendrobium reference substance, the latter by the absorbance value at 510 nm of Rutin reference substance.

Metabolomics analysis
The metabolites of D. huoshanense seedlings were analyzed by GC-MS (Agilent Technologies Inc., CA, USA) in the control and treatment groups at their harvest time. Full details of the metabolite isolation and purification, GC-MS based metabolomic analysis, and statistical analysis are given in the following sections.

Metabolite extraction
Once harvested, D. huoshanense seedlings were thoroughly rinsed to remove any residual particles from their surfaces, and then dried with Kimberly wipes. Next, each fresh whole D. huoshanense seedling was ground into powder. From each, accurately weighed 60-mg of subsample and transferred it to a 1.5 mL centrifuge tube containing two small steel balls. The 2-chloro-L-phenylalanine (CAS: 1036) dissolved in methanol (0.3 mg/mL) was set as the internal standard; while the samples were added 40 µL 2-chloro-L-phenylalanine and 360 µL cold methanol and laid aside at −80 • C for 2 min, 60 Hz ultrasound for 2 min. All sample tubes were ultrasonicated at room temperature for 30 min, and then chloroform (200 µL) and water (400 µL) were added by another vortex and ultrasonic at room temperature for more than 30 min. Centrifuge each of the above samples at 13,000 rpm for 10 min at 4 • C. Samples of quality control were made by mixing all experimental samples in aliquots. Take 100-µL of each supernatant into a glass vial for vacuum-dry at ambient temperature, then add methoxylamine hydrochloride (CAS: 5919) dissolved in pyridine (15 mg/mL, 80 µL). The obtained mixture was vortexized for 2 min and then incubated for 90 min at 37 • C, solution of N,O-Bis (trimethylsilyl) trifluoroacetamide (BSTFA) (80 µL) (with 1% trimethylchlorosilane) and n-hexane (20 µL) was then added to the mixture, vigorously vortexed for 2 min and derivatized at 70 • C for 60 min. Before GC-MS analysis, all prepared samples were placed at room temperature for 30 min.

Metabolomics analysis based on GC-MS
Derived samples after treatment were analyzed using an Agilent 7890B gas chromatography (GC) system (Agilent Technologies Inc., CA, USA), which coupled with a mass selective detector system (Agilent 5977A) and equipped with a fused-silica capillary column in 30 m × 0.25 mm × 0.25 µm size (Model: DB-5MS; Agilent J & W Scientific, Folsom, CA, USA). Helium (>99.999%) was used as the carrier gas through the capillary column at a constant flow rate of 1.2 mL/min. The initial oven temperature was set at 60 • C, ramping to 125 • C at 8 • C/min, 210 • C at 5 • C/min, 270 • C at 10 • C/min, 305 • C at 20 • C/min, and finally maintained at 305 • C for 5 min. The 1-µL sample was injected, and the temperature of the injector was set to 300 • C in the splitless mode. The temperature of ion source (electron impact) and MS quadrupole was set to 330 and 280 • C, respectively. The collision energy was 70 eV, mass data was obtained in a full-scan mode (m/z 50-500), and solvent delay time was set to 5 min. The quality control samples were injected periodically throughout the GC-MS analysis (every 10 samples) to provide an evaluable repeatability data set.

Multivariate statistical analysis
In order to visualize the metabolic differences among experimental and control groups, PCA (Principal Component Analysis) and OPLS-DA (Orthogonal Partial Least-Squares-Discriminant Analysis) were performed based on data from GC-MS, via an online analytic method (http://www. metaboanalyst.ca/) (33), following Cao et al. (34). Before analyzed, data were normalized (summation normalization) for differences among samples in general-purpose adjustment, and transformed logarithmically performing to make individual characteristics much more comparable.
Significance of difference among samples was defined as follow: VIP (Variable Importance in Projection) values > 1 in OPLS-DA model analysis, and statistically significant P-value (>0.05) from a student's t-test (two-tailed) of peak areas in different groups.

Data processing and statistical analysis
For flowering time and number, chlorophyll, MDA, resistance-related enzyme activity, mineral nutrient contents, and sugar contents were statistically analyzed using one-way ANOVAs (p < 0.05) to determine whether mean contents significantly differed among the Fe 3 O 4 NPs treatments and control. A two-tailed distribution was used to calculate the p-values; data are described as mean ± SD (standard error), expressed in three digits. (Zetasizer Nano ZEN3690, Malvern), and pH of 6.86 ± 0.06 and 7.02 ± 0.17, respectively.

Plant growth and chlorophyll content
Throughout the 21-day growing period, the treated D. huoshanense seedlings grew normally and none had any toxicity symptoms. Furthermore, application of Fe 3 O 4 NPs had no significant impact on their biomass (Supplementary Figure 2A). Interestingly, the  Figures 2B,C). This early flowering of D. huoshanense may be due to oxidative stress caused by exposure to the Fe 3 O 4 NPs. Stress-induced early flowering is considered a stress-escape response, to ensure species persist, by shortening their life cycle to produce seeds before stress-induced death (35). Similar results were found on peanut seeds treated with ZnO particles in nanoscale: seed germination rate and seedling vigor were significantly greater under the 1,000 ppm nanoscale ZnO treatment, which hastened blossoming and increased the leaf chlorophyll content (36).
It was found that chlorophyll a and b in the 200 mg/L Fe 3 O 4 NPs group gradually increased in content comparing with the control group, peaking at the 14th day, after which they tended to remain stable. In comparison, the content of chlorophyll in group of 100 mg/L Fe 3 O 4 NPs stabilize sooner, by 7 days. On the 21st day of harvest, the seedlings' chlorophyll content was significantly higher in the 200 mg/L Fe 3 O 4 NPs treated group than the control or treating with 100 mg/L Fe 3 O 4 NPs group (Supplementary Figures 2D,E). Although the content of carotenoids fluctuated during the growth of treated group, it was similar to the control group at the 21st day (Supplementary Figure 2F). These results suggested that Fe 3 O 4 NPs increased photosynthetic activities and promoted chlorophyll biosynthesis, which was in coordination with many previous studies. Previous study showed that an application of Fe 3 O 4 NPs and 50 mg/L Fe NPs by spraying on foliar to maize significantly promoted its chlorophyll content about 26.1 and 19.3% over the controls, respectively (37). They speculated that Fe-based NPs might accelerate the electron transport rate within the thylakoid membrane, leading to enhanced photosynthesis. In another report, the superparamagnetic NPs could be transferred into the soybean from the nutrient solution, thereby increasing the chlorophyll content of soybean's sub-apical leaves (38). While in some studies, Fe 3 O 4 NPs did not affect chlorophyll content in plants. Adding Fe 3 O 4 NPs to maize grown soil had no impact on the photosynthesis parameters of their leaves as described (24 NPs induced ROS overproduction and oxidative press, MDA, as a known physiological biomarker of lipid peroxidation in plants, its content in each group was determined. Evidently, MDA content of the control group increased only at the 21st day and did not change before that, whereas in response to 200 mg/L Fe 3 O 4 NPs it increased significantly at the 7th day and then fluctuated. After being exposed to 100 mg/L Fe 3 O 4 NPs, the MDA concentration initially fell and then rose continually during the entire cultivation period. The MDA contents of the two treatment groups significantly exceed that of the control group at 21 days ( Figure 1A). This indicated that 200 mg/L Fe 3 O 4 NPs caused oxidative stress to D. huoshanense sooner, while oxidative stress was induced by 100 mg/L Fe 3 O 4 NPs obvious latter. Belonging to metalloenzyme family, SODs can be categorized on the basis of their metal cofactors into 3 groups such as Mn-SOD, Fe-SOD and Cu/Zn-SOD (41). The activity of SOD depends on the presence of the metal co-factor in its vicinity (42). Here, the change in SOD activity roughly mirrored the changed MDA content ( Figure 1B). As explanation is that stress augmented SOD activity to eliminate more ROS, to which more Fe and Mn in plants also contributes. Curiously, from the 7th to 14th day, there was less peroxidase activity in treatment group than the control one ( Figure 1C). This could be attributed to the Fe 3 O 4 NPs possessing an intrinsic peroxidase-like activity (43). The reason why peroxidase activity of D. huoshanense was lower under exposure to Fe 3 O 4 NPs was that they can substitute for the plant's natural enzymes in breaking down excess H 2 O 2 . Further, the catalase levels in Fe 3 O 4 NPs-treated group were similar to the control one ( Figure 1D). Compared with the control group, the changes of MDA and antioxidant enzyme activities in D. huoshanense during culture in the experimental group showed that Fe 3 O 4 NPs induced D. huoshanense stress, but excessive ROS were decomposed to avoid plant damage after a series of enzyme system reactions. Moreover, 200 mg/L Fe 3 O 4 .

Main bioactive components in D. huoshanense
In D. huoshanense, photosynthesis metabolites, flavonoids, and alkaloids are dominant bioactive metabolites supporting its development and medical efficacy (44). Given that Fe 3 O 4 NPs increased the chlorophyll content of D. huoshanense leaves and enhanced their photosynthesis, the latter would have influenced carbon fixation. To understand how Fe 3 O 4 NPs exposure impacted carbon fixation in D. huoshanense, total contents of polysaccharides and monosaccharides in its stems were determined (Figure 2). The polysaccharide content of D. huoshanense peaked at the 7th day, and then declined and stabilized under 200 mg/L Fe 3 O 4 NPs, that under 100 mg/L Fe 3 O 4 NPs showed slowly increased after the 7th day and peaked at the 14th day, while that in the control group was slightly increased. The polysaccharides of D. huoshanense mainly consisted of mannose and glucose, and the former's trend in variation was roughly consistent with that of polysaccharides. Interestingly, both polysaccharide and mannose content reduced from 7th day to 21st day by 31.98% in the 200 mg/L Fe 3 O 4 NPs group (Table 1). We speculated the change was caused by the early flowering of D. huoshanense, resulting in the excessive consumption of mannose, which also affected the total polysaccharide content. Glucose contents of D. huoshanense exposed to Fe 3 O 4 NPs in concentration of 100 and 200 mg/L were already significantly increased by 7 days, after which they were stable. In the control group, glucose content increased slowly during plant growth and development. It was similar with two treatment groups at 21 days of cultivation. We also found that gene expression in the sugar synthesis pathway in D. huoshanense might be influenced by Fe 3 O 4 NPs, followed by regulating the production of mannose and thus changing the total polysaccharide content in D. huoshanense. The changed polysaccharides contents of D. huoshanense were consistent with their changed MDA contents. Therefore, we suggest that oxidative stress in D. huoshanense  Alkaloids are a kind of secondary metabolites responsing to habitat stress (46). The total alkaloid content of D. huoshanense reached the lowest level at the 7th and 14th day in 100 and 200 mg/L Fe 3 O 4 NPs treatment groups, respectively. Yet only at the 21st day did the control's alkaloid content did not fall to same level as the Fe 3 O 4 NPs treatment groups. This may be attributed to the special properties of Fe 3 O 4 NPs, which protect the plant from oxidative stress, enabling D. huoshanense to quickly adjust to its nutrient environment.

Iron and other minerals
The results suggested that the Fe bioaccumulation in the stem of D. huoshanense after 21 days of exposure to NPs was 100.325 ± 7.537, 149.281 ± 4.948, 161.227 ± 10.540 µg/g DW (dried weight) for the 0 (control), 100, and 200 mg/L Fe 3 O 4 NPs treatment groups, respectively by ICP-MS analysis ( Table 2). The cellular changes in roots, stems, and leaves of D. huoshanense were observed by electron microscopy at 21 days (Figure 3). The ICP-MS results uncovered significantly increased contents of Mn, Co, B, Mo, and I for stems exposed to 100 mg/L and 200 mg/L Fe 3 O 4 NPs, vis-à-vis the control ( Table 2) perhaps because of complementary effects of nutrient elements in D. huoshanense. By contrast, the Cu and Zn content were unchanged. In maintaining plant normal growth and development, some micronutrients should be used as cofactors during photosynthesis. Within the plant chloroplasts, 60% to 80% of Mn, Cu, and Fe were found in thylakoids. According to photosynthetic electron transport chain, a total number of 4 Mn atoms, 1 Cu atom, and 22 Fe atoms were required for per chain (41). Therefore, the exposure of plants to the 200 mg/L Fe 3 O 4 NPs was with higher chlorophyll content than that of control group, which may be caused by their increases of Fe and Mn contents. Cobalt (Co) is the metal component of the cobalamin coenzyme that can promote stem and bud elongation and stabilize protein complex on chloroplast membrane. Boron (B) has structural functions in primary cell walls (47). The above results showed that hydroponic treatment with Fe 3 O 4 NPs could promote D. huoshanense to absorb some metal ions related to plant chloroplasts, and then improve photosynthesis, thereby promoting plant growth and development. However, the mechanism by which Fe 3 O 4 NPs promote root uptake   The data are means of three replicates ± standard deviation. Same letters within column indicate no significant difference and different letters stand for significant differences at P < 0.05, highlight the significant differences within the same column by bold values.
Frontiers in Nutrition frontiersin.org . /fnut. .   Figure 5A). A supervised clustering method of PLS-DA is generally providing a greater discriminative power than a PCA. To gain deeper insight into the metabolite changes, the biased PLS-DA model was applied to the data: the ensuing score plot revealed their clearly separation (Supplementary Figure 5B). The VIP value was used to gauge the explanatory power and influence intensity of each metabolite expression profile on the classification and discrimination of each group's samples. The VIP value represents the overall contribution a given variable to the PLS-DA model, and those metabolites screened at VIP ≥ 1 were the discriminating compounds underpinning differences between groups. Of the 447 metabolites in D. huoshanense, 47 were significantly changed by exposure to Fe 3 O 4 NPs based on VIP score from the PLS-DA model and the P-value for student's t-test.

Perturbed biological pathways in D. huoshanense
Results for the biological pathway analysis showed that 200 mg/L Fe 3 O 4 NPs induced the perturbation of 8 of 42 biological pathways (p < 0.05), including those for pyrimidine metabolism, tyrosine metabolism, beta-alanine metabolism, lysine degradation, alanine, aspartate and glutamate metabolism, pantothenate and CoA biosynthesis, butanoate metabolism, propanoate metabolism (Supplementary Figure 5D). Among . /fnut. . them, the first five perturbed pathways were related to nitrogen metabolism. Pantothenate (vitamin B5) was the universal precursor for coenzyme A (CoA); it was also a cofactor in energy yielding reactions, including fatty acid synthesis and carbohydrate metabolism (48). Hence, changes to pantothenate and CoA biosynthesis will inevitably impact downstream carbohydrate and fatty acid metabolism in D. huoshanense. Butanoate metabolism and propanoate metabolism were also formed of carbohydrate metabolism. Nitrogen metabolism and carbon metabolism may be the most vulnerable pathways of exposing to Fe 3 O 4 NPs. Both photosynthetic carbon metabolism and NO − 2 assimilation occur in chloroplasts, and carbon and nitrogen metabolism require the consumption of organic carbon and energy from CO 2 assimilation and photosynthesis as well as other electron transport chains.
However, in this experiment, Fe 3 O 4 NPs had a greater effect on chlorophyll in leaves of D. huoshanense, which well-explained why carbon and nitrogen metabolism was the most sensitive pathway for NPs exposure.

Metabolic changes in D. huoshanense
A heatmap using 47 differential metabolites revealed significant changes between the treatment group and the control (Supplementary Figure 6). From this, it can be concluded that a marked change in metabolites' content occurred when D. huoshanense was hydroponically grown in Fe 3 O 4 NPs-containing nutrient solution. According to different metabolic functions and pathways, 47 metabolites were divided into six main categories, including carbon metabolism, amino acids and derivatives, antioxidants, fatty acids, signaling molecule, and plant hormone. These compounds will be discussed in the following paragraph on the basis of their own category.  (Figure 4). The core of cellular respiratory machinery was the TCA cycle. Down-regulation of the TCA cycle's intermediates may indicate the limited respiration in D.
huoshanense. Enrichment analysis of metabolic pathways also confirmed that the TCA cycle was significantly affected by the exposure to 200 mg/L Fe 3 O 4 NPs (Supplementary Figure 5D). Photosynthesis and respiration are highly entwined in plant cells, sharing carbon dioxide and oxygen as the product and substrate, respectively (49 1.44). Both G6PD and 6pGD were considered rate-limiting enzymes of PPP because their reactions were irreversible in organisms (50). In our study, the Fe 3 O 4 NPs treatment significantly increased content of 6-phosphogluconate. Combined with the results that the accumulation of glucose in the growth of Fe 3 O 4 NPs hydroponically cultured D. huoshanense was higher than that in the control group, we concluded that Fe 3 O 4 NPs inhibited the downstream enzyme activity of 6PGD in the PPP pathway of D. huoshanense, resulting in the accumulation of the intermediate product (6-phosphogluconate). Several reports demonstrated G6PD activity in plants could be modulated by some metals (51), and it may also act in cell division and salt press (50). In summary, the application of Fe 3 O 4 NPs may modulate the activity of key enzymes in PPP, thereby changing its 6-phosphogluconic acid content and improving the stress resistance of D. huoshanense to better adapt to the environment. Galactoic acid, the main component of pectin, was a dicarboxylic acid produced by the oxidation of D-galacturonic acid (52). Exposing D. huoshanense to 200 mg/L Fe 3 O 4 NPs significantly reduced their galactoic acid content. Sugar acids containing carboxyl and hydroxyl groups can chelate metal ions through multiple binding modes. At physiological pH, galactaric acid coordinates Fe 3+ through carboxylic oxygen and deprotonated α-hydroxylic group (53). Therefore, the decreased galactoic acid content may be closely related to the increased Fe 3+ content of D. huoshanense. Both the TCA and carbohydrates were important components of carbon metabolism. In conclusion, Fe 3 O 4 NPs hydroponics exerted a certain influence on carbon metabolism of D. huoshanense plants.

Amino acids (nitrogen metabolism) and derivatives
Amino acids are essential components of primary metabolism in plants. They not only constitute proteins, but also figure prominently in many physiological processes, by acting as osmolytes, regulating ion transport, contributing to redox-homeostasis, participating in the detoxification of heavy metals, being precursors of plant signaling molecules and defense metabolites (54). Our results showed that most amino acids were influenced by Fe 3 O 4 NPs negligibly, except α-aminoadipic acid (α-AAA), L-alanine, norleucine, and the amino acid derivative N-acetylglycine ( Figure 4). In plant tissues, α-aminoadipic acid is obtained from the catabolism of lysine via the saccharopine pathway (55). The level of α-AAA was promoted significantly in D. huoshanense based on its exposing to Fe 3 O 4 NPs, perhaps because of its response to oxidative stress. Similar to our results, Moulin et al. found that the saccharopine pathway was associated with the stress response in rapeseed (56). Norleucine is generated by the leucine biosynthetic pathway with pyruvate or α-ketobutyric acid replacing α-ketoisovaleric acid as the initial substrate (57). The norleucine content was also significantly increased upon exposure to Fe 3 O 4 NPs, whereas the contents of L-alanine and N-acetylglycine both decreased. The TCA cycle can produce abundant energy and small molecule precursors, providing the basis for downstream metabolism, for example, intermediates during carbon assimilation and oxidation provide carbon shelves for nitrogen assimilation. hence, downregulation of the TCA cycle could lead to a substantial reduction in downstream response metabolites, such as purine metabolite (adenine, deoxyadenosine, 1-methyladenosine) and pyrimidine metabolite (3-hydroxypropionic acid, cytidine, thymidine, thymine, uracil).

Antioxidants
Non-enzymatic antioxidants in plants include ascorbic acid (ASA), glutathione, tocopherol, and phenolic compounds. Compared with the control, the contents of ASA and tocopherol in the Fe 3 O 4 NPs treatment group were not significantly changed. Based on the available genetic evidence, chlorophyll degradation and sequential phytol phosphorylation yield phytyl diphosphate (PDP), which is the pentenyl precursor for tocopherol biosynthesis (58). The fact that tocopherol content did not increase in D. huoshanense with 200 mg/L Fe 3 O 4 NPs suggested their degradation was not accelerated, which allowed . /fnut. . chlorophyll to accumulate. Chlorogenic acid is one of the most widespread soluble phenolic chemicals in plants, as a byproduct of phenylpropanoid pathway acid pathway and an important precursor of flavonoid biosynthesis (59). The Chlorogenic acid content was significantly reduced in the treatment group; moreover, a number of metabolites with the capacities of ROS scavenging, namely 4-hydroxybenzoate, 3-hydroxybenzoic acid, 3, 4-dihydroxyhydrocinnamic acid, salicylic acid, hydroquinone, noradrenaline, 1, 2, 3-trihydroxybenzene, 1, 2, 4-benzenetriol were decreased by 16-49% in response to the Fe 3 O 4 NPs compared with the control ( Figure 5). On the one hand, these substances are consumed as reducing agents, acting as hydrogen donors to inhibit or quench a free radical (60). Given the large amounts of ferric and ferrous iron available on the surface of Fe 3 O 4 NPs, they exhibit peroxidase-like activity because of their ability to catalyze the oxidation of peroxidase substrates (61). We suggest that the POD mimetic activity exerted by Fe 3 O 4 NPs scavenges some ROS in treated plants thereby protecting the endogenous nonenzymatic antioxidant system of D. huoshanense. Additionally, there was abundant evidence that phenolic substances and polyphenol oxidase (PPO) activities temporally declineed during flower development and fruit ripening (62, 63). We found that the flowering of D. huoshanense could be hastened by the iron treatment, which might also be a reason why the polyphenolic content was reduced in the treatment group. In sum, the changes in antioxidant-related compounds indicated that Fe 3 O 4 NPs may modulate antioxidant defense pathways in D. huoshanense.

Fatty acids
Not only served as the major source of stored energy, fatty acids also constitute an essential component of cellular membranes and are involved in stress responsive activities (64). Fatty acids induce stress resistance in plants through remodeling cell membrane fluidity (65) and work as modulators of defense gene expression (66). Elaidic acid, an unsaturated fatty acid, was trans-isomeric of oleic acid and reported as a membrane component (Supplementary Figure 7). Pentadecanoic acid, a saturated phospholipid fatty acid, was an essential component of the phospholipid bilayer. In this study, both pentadecanoic acid and elaidic acid were significantly increased in D. huoshanense exposed to Fe 3 O 4 NPs (Figure 6), perhaps as an adaptation of this plant to environmental changes. Another significantly upregulated one was 3-hydroxymethylglutaric acid; in contrast, L-2-hydroxyglutaric acid and several unsaturated fatty acids (itaconic acid, trans-trans-muconic acid and linolenic acid) were down-regulated significantly. In addition, unsaturated fatty alcohol dodecanol was also significantly reduced whereas the content of fatty acyl glycosides (maltitol) was significantly increased. All these metabolite changes may indicate Feinduced reprogramming of lipid membrane composition of D. huoshanense. Lipid peroxidation caused by unsaturated fatty acids is a chain reaction, which can influence free radical in cell membranes. Clearly, a potential reason for the observed changes of fatty acids is lipid peroxidation. It might also because that D. huoshanense regulates the fluidity of cell membranes in roots and restricts the infiltration of excess iron ions Frontiers in Nutrition frontiersin.org . /fnut. . Signaling molecule/plant hormone Salicylic acid (SA) not only acts as a phenolic antioxidant but also as a pivotal plant hormone that takes part in multiple plant physiological processes, including immune responses, modulation of opening and closing of stomatal aperture, seedling germination, and photosynthesis (67). SA can act as a signaling molecule that coordinates effective defense responses by activating defense gene expression (68  Research has shown that Fe 3 O 4 NPs can selectively adsorb phytochemicals containing carboxyl groups or o-phenolic hydroxyl groups, and Fe(III) ions may bind more strongly to phytochemicals containing a SA moiety (70). Similar study of different iron-based nanomaterials on the physiological effects of rice, found that a low dose (50 mg/L) Fe 3 O 4 NPs could reduce oxidative press and decrease the contents of stress-related phytohormones, such as indole-3-acetic acid and gibberellin (71). Taken together, the down regulation of SA may represent an adaptation of D. huoshanense to Fe 3 O 4 NPs exposure.

Other di erential metabolites
Cholesterol is also present in plant cells although is generally below the 1% of total sterols in plants (72). In plants, cholesterol is the precursor of many metabolites with biological activities (73). In addition, it is a component of cell membranes and leaf surface lipids (74). The Fe 3 O 4 NPs treatment reduced the cholesterol content in D. huoshanense plants, perhaps because cholesterol was used to regulate the fluidity of cell membrane phospholipid molecules. Finally, some substances not annotated in the KEGG database also underwent significant changes in response to Fe 3 O 4 NPs treatment, including glucosaminic acid, 3,6-anhydro-d-galactose, xylofuranose, citrazinic, betagentiobiose, 1-butylamine, isobutene glycol, L-alanine-alanine, 3-dehydroshikimic acid.

Conclusion
In this study, we investigated the response of Fe 3 O 4 NPs to D. huoshanense by hydroponics. Our data showed that Fe 3 O 4 NPs at 100 and 200 mg/L had no significant toxic effect on D. huoshanense. In contrast, exposure to Fe 3 O 4 NPs increased leaf chlorophyll content in D. huoshanense. In addition, the content of major bioactive substances (polysaccharides) was also increased to different extents, and according to ICP-MS metallomics results, antioxidants in D. huoshanense, the decrease of most phenolic metabolites and fatty acids implied the Fe 3 O 4 NPs altered the lipid membrane of D. huoshanense cells, but did not require the accumulation of phenols to enhance their antioxidant defense system to better cope with stress. In addition, Fe 3 O 4 NPs also interfered with amino acid metabolic pathways, which suggested they can alter the dynamics of nitrogen metabolism and reallocation of energy. These results indicate that Fe 3 O 4 NPs have potential applications as nano-fertilizers on D. huoshanense, which is a reference for making some nutrients that are not easily absorbed by plants into nanomaterials to improve their utilization. In this study, the contribution of ferric iron, divalent iron, and NPs in Fe 3 O 4 NPs to metabolite changes is still unclear, and future work should address the effects of nanoparticles with different valence states iron ions on polysaccharides (molecular weight, glycosidic bonds, and spatial structure) and metabolomics in D. huoshanense at the same level of bioavailability, their benefits or toxicity to D. huoshanense should be interpreted with corresponding nanomaterials with larger concentration gradients.

Data availability statement
The data presented in the study are deposited in the MetaboLights repository, https://www.ebi.ac.uk/metabolights/, accession number MTBLS5735.
replicates. FW represents the fresh weight of the samples. Error bars represent standard deviation. Di erent letters stand for statistical di erences at p < . .

SUPPLEMENTARY FIGURE
Principal Component Analysis (PCA) (A) and Partial least squares discriminate analysis (PLS-DA) (B) The number of di erential metabolites between control group and mg/L Fe O NPs, (C) Score plots of metabolic profiles in D. huoshanense treated with mg/L Fe O NPs, (D) Top enriched pathways, red line dotted line shows P-value is . and blue dotted line shows P-value is . .

SUPPLEMENTARY FIGURE
Heatmap of di erential metabolites.

SUPPLEMENTARY FIGURE
Elaidic acid content of D. huoshanense in mg/L Fe O NPs and control group (n = ).