Both M. xiaojinensis and M. baccata are valued in China as native apple rootstocks. M. xiaojinensis performs Fe uptake with high efficiency (Han et al., 1994, 1998, 2005). However, compared with that in M. xiaojinensis, the Fe uptake efficiency in M. baccata is much lower. As shown in Figure 1, typical Fe deficiency symptoms resulting from low Fe treatment for 9 days were quite obvious in M. baccata but not in M. xiaojinensis (Figure 1A). M. xiaojinensis had higher active Fe content in roots than did M. baccata, indicating a variation in Fe uptake ability between the two Malus genotypes (Figure 1B). Further, our microtomography analysis of Fe distribution in roots of the two species confirmed this difference. The X-ray fluorescence (XRF) maps of the Fe distribution pattern in the roots showed the Fe content in M. xiaojinensis roots was higher than that in M. baccata roots (Figure 1C).

The ROS production in roots determined by DCFH-DA fluorescence was intensified at an early stage of Fe deficiency and then weakened after prolonged Fe deficiency (Figure 2D). H₂O₂ localization in the root was monitored by reaction of CeCl₃. A clear signal was observed in the apoplast, particularly in the root of M. baccata at the prolonged Fe deficiency stage (Figure 2E).

The hypothesis that Fe deficiency can trigger ROS production was then tested. Total ROS and H₂O₂ were quantified in roots of M. xiaojinensis and M. baccata exposed to Fe deficiency. As shown in Figures 2A,B and Supplementary Figure S1, Fe deficiency was capable of triggering ROS and H₂O₂ production at the early Fe-deficient stage (9 h) in M. xiaojinensis, but not in M. baccata. Interestingly, after prolonged exposure to Fe deficiency, the ROS and H₂O₂ content was reduced in M. xiaojinensis after 1–3 days and was not significantly different from that in the Fe-sufficient treatment. This result was not observed in M. baccata. Catalase (CAT), surperoxide dismutase (SOD), and peroxidase (POD) can act as ROS scavengers, keeping cellular redox homeostasis under control. Total CAT, POD, and SOD levels quantified in roots exposed to Fe deficiency were increased at 9 h in M. xiaojinensis but were not affected in roots of M. baccata (Figure 3). These results suggest Fe deficiency can trigger ROS scavengers in order to maintain the cellular redox homeostasis in the early stage of Fe deficiency.

These results demonstrate that Fe deficiency is indeed capable of causing a significant accumulation of ROS in roots; however, the Fe-efficient species M. xiaojinensis could activate scavenging mechanisms to preserve the redox homeostasis during prolonged Fe deficiency treatment.

As shown in Figure 2, Fe deficiency induced a significant increase in root ROS contents. Consistent with ROS production, the results showed that Fe deprivation caused a significant increase in root Fe (III) reductase activity of M. xiaojinensis at 9 h (Figure 2C). An attempt was therefore made to assess whether the Fe deficiency-induced alterations in root Fe (III) reductase activity and proton extrusion were mediated by ROS signaling. It was found that the Fe deficiency-induced increase in root ROS levels was correlated with the increased expression of MxFIT and MxBTSa (Figures 4A,D). The Arabidopsis ortholog of the FER gene, FIT, was identified from the 161 predicted bHLH proteins by microarray analysis (Bauer et al., 2007). FIT also plays an important role in positively regulating various iron deficiency inducible genes, including IRT1 and FRO2 (Yuan et al., 2005; Bauer et al., 2007). BRUTUS (BTS), a putative E3 ligase protein, with metal ion binding and DNA binding domains, which negatively regulates the response to iron deficiency (Long et al., 2010).

Next, it was determined whether ROS inhibitors blocked the Fe deficiency responses. The responses to Fe deficiency were markedly repressed by exposure to diphenyleneiodonium (DPI), which can block ROS production (Figures 5A–C). DPI application strongly inhibited the Fe deficiency-induced increase in reductase activity (Figures 5B,E) and proton extrusion (Figure 5C). These results imply that a certain endogenous ROS level is necessary for the Fe deficiency-based induction of expression of transcription factors, such as FIT (Figure 5F). Taken together, these results demonstrate that ROS play a critical role in Fe deficiency-induced signaling at an early stage.

Malus xiaojinensis seedlings were propagated on Murashige and Skoog medium (MS) (Huayueyang Biotechnology, Co, Ltd, Beijing, China) containing 0.5 mg/L 6-benzylaminopurine and 0.3 mg/L indole-3-butyric acid (IBA) for 1 month, and then transferred to ½ MS medium containing 0.5 mg/L IBA for 1 month for rooting. The rooted seedlings were transferred to Hoagland’s nutrient solution (Gao et al., 2011). The pH was adjusted to pH 6.0 with 1 M NaOH. The solution was refreshed every 6 days. Plants were grown in a growth room at 25 ± 2°C day/17 ± 2°C night with a 16 h photoperiod at a light intensity of 250 mmol quanta m²/s, 85% relative humidity. After 1 month, the plants were transferred into Fe-sufficient medium (40 μM FeNa-EDTA) or Fe-deficient medium (0 μM FeNa-EDTA) with 50 μM DPI or Fe-deficient medium (0 μM FeNa-EDTA) with 50 mM H₂O₂. The roots were collected at 0 days, 9 h, and 9 days during these treatments. A completely randomized experimental design with three biological replicates and three plants for each replicate was used.

Total RNA was isolated from the roots using an improved cetyltrimethylammonium bromide (CTAB) method (Gasic et al., 2004). The RNA was digested by DNase I (Takara Biotechnology, Co, Ltd, Dalian, China) and reverse transcribed by using an oligo-dT primer and reverse transcriptase (Takara Biotechnology, Co, Ltd, Dalian, China).

Reactive oxygen species were imaged using DCFH-DA and a confocal laser scanning microscope (TE2000-E, Nikon, Co., Tokyo, Japan). Roots (3–5 mm from root tip) were loaded with 20 μM DCFH-DA in 20 mM HEPES/NaOH pH 7.5 buffer for 90 min, washed three times with fresh buffer, and analyzed microscopically (excitation 488 nm, emission 525 nm). The images were analyzed using EZ-C1 3.00 Free View software. For fluorescence quantification per gram of root tissue, roots were loaded with DCFH-DA and washed; the complete root system (approximately 50 mg) was subsequently ground in 0.5 mL HEPES/NaOH pH 7.5 buffer. The supernatants were then centrifuged (13, 000 × g, 10 min) at 4°C, and DCFH-DA concentration was determined by using a fluorescent spectrophotometer (F-7000, Hitachi, Tokyo, Japan) using excitation at 488 nm and emission at 525 nm. DCFH-DA was similarly processed without roots for comparison as a blank. Data are expressed as the mean ± SD of two independent experiments performed in triplicate.

Hydrogen peroxide was detected using the method described by Mukherjee and Choudhuri (1985). A root sample of 200 mg was ground with 1.5 mL pre-cooled acetone, then centrifuged (800 × g) at 4°C for 10 min. Then, 5% titanium sulfate and ammonia water were added to the supernatant, and it was centrifuged again (800 × g) at 4°C for 10 min. The precipitate was rinsed three times with acetone and vortexed. It was then redissolved in 10 M H₂SO₄. The absorbance of the supernatant at 415 nm was measured using an UV spectrophotometer (UV 1800, Shimadzu, Tokyo, Japan); 1.5 mL acetone was similarly processed without roots for comparison as a blank. The H₂O₂ content was determined based on a standard curve plotted using known H₂O₂ concentrations. The H₂O₂ content was measured as micromoles per gram fresh weight (μmol/g FW).

Freshly harvested roots were cut in 5 mm sections close to the root tip and subjected to protocols for cytochemical localization of H₂O₂. H₂O₂ production was assessed cytochemically via determination of cerium perhydroxide formation after reaction of CeCl₃ with endogenous H₂O₂ (Bestwick et al., 1997). Root sections of M. xiaojinensis and M. baccata were incubated for 1 h in 5 mM CeCl₃ in 50 mM 3-(N-morpholino) propanesulphonic acid (MOPS) pH 7.2, fixed in 1.25% glutaraldehyde and 1.25% paraformaldehyde in 50 mM sodium cacodylate buffer (CAB) pH 7.2 by vacuum evacuation of samples for 30 min, and after drying washed three times in CAB buffer for 10 min (Bestwick et al., 1997). Samples were then embedded in resin and observed by transmission electron microscopy (JEM-1230, JEOL, Tokyo, Japan).

Root tissue (0.1 g) was frozen with liquid N₂ and homogenized in 2.0 mL chilled 50 mM phosphate buffer (pH 7.0). The homogenate was centrifuged at 1, 3000 × g, for 20 min at 4°C. The supernatant was stored at 4°C and used for enzyme assays within 4 h.

Catalase activity was measured in a reaction mixture (3.0 mL) containing 200 μL of the extract, 0.3 mL 0.1 mM H₂O₂, and 1 mL distilled water in 0.2 M phosphate (pH 7.8) containing 1% (w/v) insoluble polyvinylpyrrolidone (PVP). The decrease in absorbance at 240 nm was monitored for 3 min, and the amount of enzyme digesting 1 μM substrate in 1 min was indicated as enzyme activity (μ) (Tewari et al., 2012). CAT activity was presented as μmol H₂O₂ decomposed/min (unit)/mg protein.

Peroxidase activity was estimated in a reaction mixture (3.0 mL) containing 50 μL 50 mM phosphate buffer (pH 7.0), 2.9 mL 0.3% guaiacol (v/v), 60 μL 0.004% H₂O₂ (v/v), and 50 μL root extract, and the change in absorbance was measured at 470 nm (Ranieri et al., 1997). Enzyme activity was expressed as ΔAbs/0.01 min (unit)/mg protein.

Superoxide dismutase activity was assayed by measuring its ability to inhibit the photochemical reduction of nitro blue tetrazolium (NBT) at 560 nm. The reaction mixture (3.0 mL) contained 1.5 mL 50 mM phosphate buffer (pH 7.8), 300 μL 750 μM NBT, 300 μL 20 μM riboflavin, 50 μL root enzyme extract, 300 μL 130 mM methionine (Met), and 1 mL distilled water (Bradford, 1976). The reaction mixture was exposed to light at 4000 Lx for 10 min. The amount of SOD corresponding to 50% inhibition of the reaction was defined as one unit of enzyme. SOD activity was expressed as U (unit)/mg protein.

The root FCR activity was quantified using a ferrozine assay (Schikora and Schmidt, 2001). All reduction assays were performed in 15 mL lightproof centrifuge tubes. The roots were initially washed in 0.5 mM CaSO₄ for 5 min, and then the color reaction was carried out in a solution of 0.5 mM ferrozine, 0.5 mM FeNa-EDTA, and 0.5 mM CaSO₄ at 25°C for 40 min. The reaction was adjusted to pH 6.0 with 1 M NaOH. Reduction activity was measured at 562 nm using a spectrophotometer (UV 1800, Shimadzu, Japan). The reduction rate was calculated after subtraction of the appropriate blanks (assay solution without roots). The rate of root FCR activity was calculated as moles of Fe²⁺-ferrozine per gram of fresh weight per hour. Data, expressed as the mean ± SD, represent three biological replicates, with each replicate containing four plant root samples. To visualize FCR activity spatially, the plant roots were initially pre-processed and then embedded in agarose (0.7%, w/v) medium containing the assay solution. Photographs were taken after 40 min (Li et al., 2000).

Root pH of apple M. xiaojinensis was monitored with the cell-permeant and pH-sensitive fluorescent dye BCECF-AM (Tang et al., 2012). Root pH was quantified by a ratio analysis of the fluorescence at pH-dependent (488 nm) and pH-independent (458 nm) excitation wavelengths from a calibration curve (Supplementary Figure S2), and ratio images were produced using the ion concentration tool of Zeiss LSM confocal software.

Approximately, 1 g of roots was maintained at 70°C for 7 days. One hundred milligrams of each dried sample was then treated with 10 ml 1.0 M HCl in an orbital shaker for 6 h, and the filtered extract was analyzed by polarized Zeeman atomic absorption spectrophotometry (Z-5000; Hitachi, Tokyo, Japan) (Takkar and Kaur, 1984; Han et al., 1994).

DCFH-DA (Dichlorofluorescein diacetate) and BCECF-AM (3′-O-Acetyl-2′,7′-bis(carboxyethyl)-4) were purchased from Beyotime Institute of Biotechnology, Product codes S0033 and S1006¹; DPI, HEPES, and ferrozine were from Sigma².

The quantitative measurement of gene expression was performed using the cDNA samples with an AB7500 Real-time PCR System and the SYBR Green fluorescence dye (Takara) (Li et al., 2012).

The genes analyzed were the following: Fe deficiency-induced transcription factor MxFIT1 (NCBI: XP_008360009.1), Fe²⁺ transporter MxIRT1 (NCBI: AAO17059.1), Malus × domestica zinc finger protein MxZAT12c (NCBI: XM_008371622), putative E3 ligase BRUTUS MxBTSa (NCBI: XP_008393323.1). The sequences of all of the genes in the apple genome were obtained from the website http://genomics.research.iasma.it/ via its BLAST (basic local alignment search tool) service. Based on the BLASTP results in NCBI⁴ for those proteins encoded by the genes examined, primers were designed using Primer Premier 5 (Primer, Co., Canada). The β-Actin was used as the reference gene. The primer sets used are listed in Supplementary Table S1.

Statistical analysis was performed using the Statistical Product and Service Solutions (SPSS) software (IBM, Co., Armonk, NY, USA). All experimental data were tested by analysis of variance (ANOVA) and Duncan’s multiple-range test.