Nanoscale-specic bioassimilation of sulfur: Time and coating specic modulation of transcriptomic and metabolomic pathways in diseased tomato

Nanoscale sulfur was investigated as a multi-functional agricultural amendment to simultaneously enhance crop nutrition and suppress disease damage. Pristine (nS) and stearic acid coated (cS) sulfur nanoparticles were added to soil (0, 100, or 200 mg/L) that was planted with tomato (Solanum lycopersicum) and infested with the Fusarium wilt pathogen. Bulk sulfur (bS), ionic sulfate (iS), and healthy controls were included. In two greenhouse experiments, measured endpoints included time-dependent agronomic and photosynthetic parameters, disease severity/suppression, and a range of mechanistic biochemical and molecular endpoints, including the expression of 13 genes related to two S bioassimilation pathways and pathogenesis-response, and tissue-specic metabolomic proles. The impact of treatment on the rhizosphere bacterial microbiome was also evaluated. Disease reduced tomato biomass by up to 87%, but amendment with nS and cS signicantly reduced disease progress by 54 and 56%, respectively, compared to the infested controls. Increased S accumulation was evident in plant roots and leaves, independent of S type. Molecular analysis revealed particle size and coating-specic impacts on the plants. For nS and cS, two-photon microscopy and time-dependent gene expression data revealed a nanoscale specic elemental S bioassimilation pathway within the plant tissues. These ndings correlated well with detailed metabolomic proling of plant tissues at 4, 8, and 16 d, which exhibited increased disease resistance and plant immunity related metabolites with nanoscale treatment. The data also demonstrate a time-sensitive physiological window whereby nanoscale stimulation of plant immunity will be effective. An analysis of the rhizosphere soil bacterial community revealed minimal impacts from S soil treatments. These ndings provide signicant mechanistic insight into non-metal nanomaterial-based suppression of plant disease, and signicantly advance efforts to develop sustainable nano-enabled agricultural strategies to increase food production.


Introduction
Although the Green Revolution led to signi cant increases in agricultural output through agrochemical use and irrigation, many current practices are not sustainable [1][2][3] . For example, the e ciency of agrochemical delivery is quite low, often no more than 30% 4 . This results in growers overapplying pesticides and fertilizers to achieve the acceptable yield, but leads to environmental damage 5,6 . In addition, agricultural production will have to increase by up to 70% by 2050 to feed an expected global population of more than 9 billion people [7][8][9][10] . Such increases will only be possible if new technologies and approaches can be applied across the farm to fork continuum.
In spite of the widespread use of agrochemicals, plant pathogens reduce the yield of most crops by 20% and compromise product quality, often leading to 40% food loss before harvest 1 . There has been increasing interest in developing nanoscale materials for agricultural use, including novel crop management strategies and targeted delivery of nutrients and pesticides [10][11][12][13] . Recent work has demonstrated the unique potential of nanoscale micronutrients to enhance tolerance to both biotic and abiotic stressors. Dimkpa et al. demonstrated that nanoscale zinc amendments to soil not only alleviated (HT7800, Hitachi Ltd., Tokyo, Japan). For TEM analysis, suspensions in acetone of both nanoscale sulfur types were sonicated for 15 min in cold water. Aliquots were then added dropwise to a Cu grid with lacey carbon lm and were allowed to air dry. A Malvern Zetasizer Nano ZS-90 following a laser doppler electrophoresis procedure was used to measure hydrodynamic diameter and zeta potential of the S based compounds at 200 mg/L.

S dissolution and soil bioavailable nutritional elements
The dissolution of sulfate from different sulfur types (200 mg/L) was measured in batch reactors. There were ve replicates for each treatment; each replicate was sonicated on ice for 2 minutes at 120 W (Fisherbrand™ Model 120 Sonic Dismembrator) and kept in the dark to avoid photodegradation. Five ml samples were taken at 4 h, 1 day, 4 d, 8 d, and 16 d and were centrifuged (Eppendorf™ Centrifuge 5810R) at 12,000 rpm for 30 minutes at 4°C. The supernatant was then transferred into clean vessels, acidi ed with HNO 3, and analyzed using inductively coupled plasma-optical emission spectroscopy (ICP-OES) (see below) for S content. To determine the bioavailability of soil nutrients, 2.0 g aliquots of air-dried soil were extracted by two procedures modi ed according to a previous study 25 . First was a DTPA method using with 30 mL mixture of 0.01M CaCl 2 , 0.005M DTPA and 0.1M triethanolamine (TEA) (pH=7.6). The second procedure used 30 mL of 0.01M CaCl 2 (pH= 5). DI water (30 mL) was also used as a third extracting solution. All extractions were conducted on a reciprocal shaker for 2 hours at 180 rpm to facilitate mixing. The samples were then centrifuged for 10 min at 12,000 rpm, and the supernatants were passed through a 0.2 um PTFE lter. The pH and electrical conductivity (EC) of DI extracts were measured to estimate the soil pore water pH and soil salinity ( Fig. S18 and Table S3). All supernatants were analyzed by ICP-OES as described below.

Greenhouse experiments
Seeds of Solanum lycopersicum L. cv Bonnie Best were obtained from Harris Seed Co., Rochester NY; this is a common heirloom variety which is susceptible to Fusarium wilt. Seeds were germinated in Pro-Mix BX potting soil (Premier Hort Tech, Quakertown, PA) (with a background S level of 218.4 ± 5.5 mg/L) in 36 cell (5.66 × 4.93 × 5.66 cm) plastic liners and grown for four weeks. Twenty-ve ml of a low sulfur (0.035%) soluble fertilizer (Premium Water Soluble Plant Food, 24-13-17, NPK, Proven Winners, USA) was applied at the end of the third week, and was given once per week after seedling transplanting. Seedlings of uniform size with three to four leaves were selected for experiments. F. oxysporum f. sp. lycopersici inoculum was prepared according to our previous studies 15,17,[26][27][28] . Brie y, autoclaved millet seeds were seeded with agar plugs colonized by F. oxysporum at 22-25 °C for two weeks. The millet was dried, ground, and passed through a 1 mm sieve. The millet was added at 0.75 g/ liter potting soil and thoroughly mixed to yield approximately 7.5 × 10 4 Fusarium colonies per g potting mix as determined by serial dilution onto 25% potato-dextrose-agar (Difco, Corpus Christi, TX) 15 .
The e cacy of nS and cS at suppressing Fusarium infection on tomatoes was evaluated and compared with corresponding bulk form. To assess the effect of surface coating/charge, the responses to nS and cS were directly compared. All S based materials were applied by a soil amendment. Suspensions/solutions of each S-based compound were prepared in 25 mL deionized water (DI), followed by sonication for 2 min using a probe sonicator to provide a stable dispersion. The prepared solutions were then mixed into the potting soil prepared above to achieve nal concentrations of 200 mg S/L soil, and manually mixed for 20 min. Twenty-ve mL DI water was added to potting mix to serve as an untreated control. Uniform 4-week-old tomato seedlings were transplanted into these soils (one per pot) and were placed in a greenhouse before harvest under the following conditions: 25-28 °C day/20-22°C night, relative humidity of 60-70%, under high intensity sodium vapor lights set for 12-hour photoperiods. The time-dependent effects of treatment were evaluated; here, three plants were harvested at 4, 8, and 16 days after transplanting. In a second experiment of similar design, an additional concentration of 100 mg S/L soil and an additional treatment of ionic sodium sulfate (Na 2 SO 4 ) was added to address the effect of dose and a non-nano ionic S form. There were 10 replicates, and plants were harvested at 35 days after transplanting. Photosynthetic parameters including relative chlorophyll content, leaf thickness, and linear electron ow (LEF) were measured using a portable PhotosynQ (PHOTOSYNQ INC., USA) at 30 d after the transplanting. For both greenhouse experiments, root stem and leaf tissue were separately weighed, washed in tap water, rinsed with 0.01 % nitric acid, rinsed again with DI, frozen in liquid nitrogen, and then stored at -80 o C until further analysis.

Disease progress
Disease severity of F. oxysporum infected tomato across all treatments was determined every other day on a scale of 1 to 5: 1 = asymptomatic, 2 = slightly stunted and/chlorotic, 3 = partially stunted or wilted, 4 = serious stunted or wilted, and 5 = dead or near death. Disease progress was plotted over time, and the area-under-the-disease-progress curve (AUDPC) was calculated using the trapezoid rule where: AUDPC = ∑ (Y i + Y i )/2 × (t i+1 − t i ), where Y i = disease rating at time t i . The pathogen was re-isolated from damaged stem tissue to con rm its association with the disease.

Elemental Analysis
Harvested tissues were analyzed by ICP-OES (iCAP 6500, Thermo Fisher Scienti c, Waltham, MA) to measure S accumulation, as well as secondary and macronutrients (Ca, Mg, P, and K) and micronutrients (B, Cu, Fe, Mn, Mo, Zn). Brie y, 150 mg of dry ground tissue was weighed, and digested in 5 mL plasma pure HNO 3 on a hot block at 115 °C for 45 min. The digesta was amended with 1 mL of H 2 O 2 and further digested for 20 min at 115 °C. The digests were then diluted to 50 mL with DI. To validate the measurements, samples consisting of a blank (no plant tissues), pure nanoscale and bulk S powder, and standard reference materials (NIST-SRF 1570a and 1547, Metuchen, NJ) were prepared, digested, and analyzed by the same procedure. Yttrium (Y) was used as an internal standard, and a continuing calibration veri cation sample was evaluated every 15 samples.

RNA Isolation and RT-qPCR
For gene expression analysis, tissue samples that had been frozen in liquid N 2 were pulverized with a mortar and pestle. Total RNA was extracted using the PureLink™ Plant RNA Reagent (Invitrogen™) according to the manufacturer's instructions and was quanti ed using a NanoDrop spectrophotometer (Thermo Scienti c). RNA quality and integrity were con rmed by the 260/280 Abs ratio. RT-qPCR analysis was conducted as described previously 29 ; RNA was exposed to RNase-free DNAse I (Roche) at 37 °C, and then cDNA was synthesized by the High-Capacity cDNA Reverse Transcription Kit with RNase Inhibitor (ThermoFisher Scienti c). The nal concentration of cDNA was 5 ng/mL. Quantitative PCR (qPCR) was performed in a StepOne™ Real-Time PCR system. PCR reaction tubes were amended with iQ™ SYBR® Green Supermix (Biorad) and primers at a 20 µL nal volume. The PCR conditions were: 1 cycle at 95 °C for 5 min; 40 cycles at 95 °C for 15 s, 55 °C for 25 s, and 72°C 20 s, including a nal melting curve program at 60 °C + 0.3 for 20 s. The selected genes and the corresponding primers are described in Table  S22. The PCR ampli cation e ciency was calculated with different primers with a range between 90 and 110%. Gene expression analysis was done following the 2 ΔCt method.

Two-photon microscopy analysis
Tomato root and leaf samples were cut with a CryoMicrotome (Triangle Biomedical Sciences, Durham, NC) at −20 °C. The resulting sections were mounted on glass substrates and viewed under a waterimmersion objective lens (Olympus LUM Plan FLN) 30,31 using a mode-locked Ti: Sapphire laser (Spectra-Physics, Mai-Tai HP) as the light source. Two-photon excitation was achieved at a wavelength of 720 nm and 150 mW. The pulse duration and the repetition rate were ~100fs width and 80 MHz, respectively. The blue and green/red uorescence signals were de ected with a long-pass dichroic mirror at 665 nm, and split through a long-pass dichroic beam splitter, transmitted through band-pass lters of 417-477 nm, 500-550 nm, and 570-616 nm, respectively, and detected by a photomultiplier tube at 2.10V, 2.20 V, and 0.9 V, respectively. The outputs were fed into red/green/blue channels by a frame grabber. The speed of imaging was 30 frames/s, and a static image was obtained as an average of 100 frames 32 .
Metabolomic analysis and data processing Fresh leaf tissues collected at harvest were immediately frozen in liquid nitrogen for 20 s for complete metabolic inactivation and stored at -80 °C 33 . They were then freeze-dried for 6 h using a lyophilizer. Dry samples (~8 mg) were extracted in 2 ml of cold 80% LC-MS-grade methanol containing 100 ng/mL of metalaxyl as the internal standard. Untreated healthy tomato leaves were used as a control. Mixtures were consecutively vortexed and sonicated in ice cold water bath for 20 min. The extracts were centrifuged for 15 min at 14,000 rpm, and the supernatants were used for detection using a Waters Acquity ultra-high performance liquid chromatography (UPLC) system directly coupled to a Waters Synapt G2Si HDMS high resolution mass spectrometer. For all experiments, 10 µL of metabolite extract was directly loaded onto a Waters Acquity UPLC HSS T3 analytical column with 100 Å pore size, 1.8 µm particle diameter, 2.1 mm x 150 mm, that was held at 40˚C. A reversed phase binary gradient using Fisher Optima LC/MS grade solvents (Solvent A: 0.1% formic acid in water, Solvent B: 0.1% formic acid in acetonitrile) was used to elute metabolites directly into the mass spectrometer. The gradient included a 0.300 mL/min ow rate and the following conditions for elution: 0.2 min initial hold at 1% Solvent B, a linear ramp to 60% Solvent B over 4.8 min, a linear ramp from 60% to 90% Solvent B over 0.5 min, a 0.7 min hold at 90% Solvent B, a linear ramp from 90% to 1% Solvent B over 0.2 min and a re-equilibration wash at 1% Solvent B for 2 min.
The Synapt G2Si mass spectrometer was kept in positive mode electrospray ionization and the following tune parameters: +3 kV capillary voltage, 40 V sampling cone voltage, 80 source offset, source temperature of 100˚C, cone gas ow of 50 L/h, desolvation gas ow of 832 L/h, and nebulizer pressure of 6.5 bar. An 8.4 min alternating full MS and MS E data-independent acquisition method was implemented using positive polarity and the "Resolution" acquisition mode. The following parameters were used for Full MS acquisition: mass range of 100 to 1,000 Da, 0.2 sec/scan time, Trap xed collision energy of 4 V, with no additional Transfer collision energy. MS E scans were collected with the following parameters: mass range of 50 to 1,000 Da, 0.2 sec/scan time, Trap collision energy ramp from 5-40 V, and no additional Transfer collision energy. A lock spray mass correction was performed in real-time using infused Leucine Enkephalin (556.2771 m/z) and the following acquisition settings: a 0.2 sec/scan time, acquisition interval of 10 sec, 3 scans to average, and a mass window of ± 0.5 Da.
All raw data were processed and analyzed directly using Progenesis QI (v 2.4, Nonlinear Dynamics) for pick picking, feature retention time (RT) alignment, feature quanti cation, and metabolite identi cation. All raw les were imported using the "High resolution mass spectrometer" settings with pro le data and positive polarity. RT alignment was achieved using the "Assess all runs in the experiment for suitability" option, which resulted in successful alignment across the dataset (scores >87%). Peak picking used "Automatic" for Ion Sensitivity, included no RT limits, used "0.2% of base peak" for the high energy fragment ion intensity limit, and included the following adducts: (M+H + ) + , (M+2H + ) 2+ , (M+Na + ) + , and (M+K + ) + . Experiment-wide normalization was completed using the "Normalize to all compounds" option. The entire molecular feature list was searched against the METLIN spectral library database using a 10-ppm tolerance for precursor masses, a 20-ppm tolerance for fragment ion masses, and only "empirical" fragmentation type. Quanti ed feature information and METLIN-derived molecular identi cations were exported from Progenesis QI for additional analysis. Statistical analysis and pathway analysis of the metabolites was performed using MetaboAnalyst 5.0 For multivariate analysis, log10(x) transformation was performed for the normalization by sum 34,35 . One-way ANOVA followed by Fisher's LSD test (p ≤ 0.05) was performed to screen for metabolites. A supervised partial least-squaresdiscriminant analysis (PLS-DA) was applied to get a global overview of the variation of metabolic pro les. The library used in pathway enrichment and pathway topology analysis is the Arabidopsis thaliana KEGG library. The impact value threshold >0.1 was considered as signi cantly affected. The features with a VIP greater than 0.1 are regarded as responsible for separation 36 .

Rhizosphere bacteria characterization
Rhizosphere soils were de ned as the soil tightly adhering to the root tissues and were collected as described previously 37 . At harvest, ten-cm sections of tomato roots were collected from each plant and shaken vigorously to remove loosely adhering soil. The root sample was then transferred to a centrifuge tube containing 25 ml of sterile phosphate-buffered saline. The rhizosphere soil was removed from the roots by vortexing the samples for 2 min at full speed. The roots were then removed with ethanolsterilized forceps, and soils were concentrated by centrifugation at 14,000 rpm for 10 min at 4°C. The concentrated rhizosphere soils were stored at -80°C until DNA extraction.
DNA was extracted from 0.25 g of soil from the resulting pellet using the DNeasy PowerSoil kit (Qiagen). DNA extractions were veri ed by gel electrophoresis in a 1% agar gel. Bacterial 16S rRNA genes were ampli ed with the primer pair 515F (GTGYCAGCMGCCGCGGTAA) and 806R (GGACTACNVGGGTWTCTAAT) 38 . Extracts were ampli ed with 10 μl Platinum SuperFi II DNA polymerase (Invitrogen), which also included 7.5 μM both the mPNA and pPNA peptide nucleic acid (PNA) clamps (mPNA, GGCAAGTGTTCTTCGGA; pPNA, GGCTCAACCCTGGACAG) to block the ampli cation of host plant mitochondria and plastid rRNA genes, respectively. PCR conditions consisted of 94°C for 2 min followed by 30 cycles of 94°C for 15 s, 60°C for 15 s, 68°C for 15 s, and 4°C for an in nite hold. The resulting ampli cation products were veri ed by gel electrophoresis, and cleaning and normalization of individual PCR products were performed with a SequalPrep normalization plate (Invitrogen). The normalized PCR amplicons were mixed, and the quantity and quality of the DNA pool were veri ed using an Agilent TapeStation. The resulting 16S rRNA gene amplicons were sequenced on the Illumina iSeq 100 system employing 150 base pair chemistry.
16S rRNA gene sequences were initially processed using the mothur software package (v. 39 . Because the 150 bp reads were too short to generate contigs, only the forward reads were retained for analysis. Quality ltering was done and selected for sequences of at least 150 bp in length. Chimeric sequences were identi ed with the VSEARCH algorithm as implemented in mothur, using the most abundant sequences as a reference for chimera detection 40 . All putative chimeric sequences were removed from the data sets. A total of 36,523 high-quality 16S rRNA sequences were retained, with an average of 1,523 per sample. The 16S rRNA sequences were classi ed against the SILVA v138 reference database using the RDP naive Bayesian classi er as implemented in mothur, and sequences identi ed as belonging to eukaryotes were removed 41 . The resulting sets of sequences were assigned to amplicon sequence variants, employing a 100% sequence similarity threshold. The mothur output les were imported into the microeco R package for descriptive and statistical analyses. Biomarkers distinguishing the different treatments were identi ed with linear discriminant analysis (LDA) effect size (LEfSe) method 42 .

Statistical analysis
A Shapiro-Wilk's W Test of homogeneity was performed on all data before further analysis. A one-way ANOVA followed by a Tukey-Kramer multiple comparison test was conducted (SPSS, IBM Corp.) with the S treatments as xed factors. Experiment two was similarly analyzed with S treatment and dose as xed factors. A student's t-test was also used to compare the differences between speci c treatments.
Statistical signi cance was determined at p ≤ 0.05. Data are expressed as the mean ± standard error (SE).

S material Characterization
Images from the SEM-EDS, TEM, and XRD patterns are shown in Fig. S1-3. Both uncoated and coated nanoscale S showed a uniform spheroidal shape with average particle sizes of 65 and 38 nm, respectively, and average hydrodynamic sizes of 851.2 ± 28.7 nm and 982.5 ± 86.6 nm (pH = 7), respectively ( Fig. S2, S4, and Table S1). The zeta potential of nS, cS, and bS ( Fig. S4 and Table S1) at pH 7 was -23.6 ± 0.4, -33.5 ± 0.3, and -13.9 ± 4.3, respectively. The crystal diffraction pattern of nS, cS and bS from XRD analysis ( Fig. S3) matched well with the orthorhombic structure of elemental sulfur (S8) 43 . The EDS analysis con rmed the elemental compositions of nS and bS as pure S, and cS as a combination of S and C (Fig. S1). The dissolution of all the three S based compounds in DI was less than 0.4% at 15 days ( Fig. S5 and Table S2).

Two-photon microscopy images
Detection of S by two-photon microscopy is both size-and coating-dependent ( Fig. 1, S6, and S7). In DI suspensions at 500 mg/L, much less uorescence signal was detected from nS suspensions that were not sonicated, since particle aggregation signi cantly reduced signal intensity (Fig. 1e); after thorough sonication, the uorescence signal from nS was clearly observed (Fig. 1k). At the same concentration, neither bS suspension with or without sonication, or ionic S solutions, had a uorescence signal (Fig. 1f, g, l, m, j, and p). The nding that smaller size enables a stronger uorescence has been reported previously for nanoscale ZnO 44 . Importantly, steric acid-coated nanoscale S did not uoresce (Fig. 1h, i, n, and o), likely due to the change in surface chemical composition; an ionic SO 4 2solution also exhibited no uorescence. This unique uorescence response of uncoated nanoscale S will be important to understand the nanomaterial movement in the plant. The root uptake of S based (200 mg/L) compounds and translocation to leaf tissues as determined by two-photon microscopy is shown in Fig. 1c, d, and S6. Representative uorescent signals from S are indicated by the arrows. These distinguishable response points are similar to the pure nanoscale S suspended in DI ( Fig. 1e and k). In nS treated roots, a S uorescent signal was evident, indicating nS accumulation; clear signals are also evident in the corresponding leaf tissue, demonstrating translocation of nanoscale S. For cS, a weak S signal was found in cS treated roots. Given the increased S content observed by ICP-OES (below) and the lack of cS response in solution, this suggests that cS was accumulated and translocated through tomato either in the surface coated form or as dissolved ions. No detectable signal S was found in bS exposed roots, indicating the absence of nanoscale S; interestingly, signi cant S uorescence was shown in the corresponding leaves of bS treated plants, suggesting that active in planta transformation of bS was occurring, including likely reduction of sulfate ions to nanoscale S in leaf tissues. Spatially, S was observed near or in the vascular tissues and the intercellular spaces, indicating apoplastic route of uptake. In the leaf tissues, S resulting from nS and bS exposure was distributed similarly and primarily around the stomata, indicating a xylem-based transport pathway via the transpiration stream. Additional time resolved videos of the 3D biodistribution pattern of S within the leaves and roots and representative images taken at different depths from the surface are available (Videos 1-6 and Fig. S7) and temporally con rm the above ndings.

Disease progress and plant growth
As shown in Fig. 2 and S8-9, at 4 d, there was no impact on biomass as a function of treatment or disease. However, at 8 and 16 d, disease reduced leaf biomass by 35.3% and 79.5% and root biomass by 20.1% and 73.5%, respectively. The positive impacts of S amendment were evident by day 8; cS and bS signi cantly increased shoot biomass by 51.2% and 43.6%, and root biomass by 26.7 and 48.2%, respectively. By day 16, the impact of bS became non-signi cant, but nS and cS increased shoot biomass by 264% and 378% and root biomass by 200% and 175%, respectively.
Results from the 35 d greenhouse experiment were similar ( Fig. S10 and S11). Treatment with the cS and nS decreased disease progress by 54 and 56%, respectively; bS decreased disease by 32% but ionic S had no impact on disease progression. Disease decreased plant shoot mass by 87.4% but nS at 100 and 200 mg/L increased shoot mass by 1,160%-1,750% relative to the disease controls; cS increases were 819%-1,180%, respectively. Similarly, for the root mass, the increases were 263 and 412% for nS, and 219 and 221% for cS, respectively. Treatment with bS and iS had no impact on biomass under diseased conditions, regardless of concentration. Both nS and cS increased relative chlorophyll content compared to the infested control and bS in the 16 d greenhouse experiment; in the 35 d greenhouse study, cS increased linear electron ow (LEF) relative to the infested control and bS, and had thicker leaves than bS (Fig. S12). This may be related to the increment of sucrose in corresponding leaves. cS increased leaf sucrose content by 1.4-1.5-fold over controls, nS, and bS at 8 d; at 16 d, both nS and cS had higher amount of sucrose than the infested control and bS by ~50-68%. Sucrose is the end product of photosynthesis and the primary sugar transported in the phloem of most plants. It also serves as both a source of carbon skeletons and energy for plant organs unable to perform photosynthesis.
A number of nanoscale amendments have been shown to reduce disease progression and increase biomass, although these have largely been with foliar applications. Elmer and White 45 showed that foliar treatment of CuO signi cantly suppressed Fusarium infection and increased biomass of tomato and eggplant by 37.5% and 10.6%, respectively, under eld conditions. Similar fungal disease suppression capability has been reported for foliar application of different nanoscale forms of silica to watermelon 46,47 . Borgatta et al. 16 and Ma et al. 48 both demonstrated the importance of material properties, reporting that foliar application of 2 ml of 10-1000 mg/: copper phosphate nanosheets suppressed Fusarium infection and increased the biomass of watermelon and soybean at concentrations as low as 10 mg/L, whereas > 100 mg/L of spherical nanoscale CuO was needed to exert similar bene t.
Less data is available on nanoscale sulfur amendments. Shang et al. 49 reported that seed and foliar treatment of rice with CuS nanoparticles signi cantly suppressed Bakanae disease by 35.1-45.9%. Importantly, results with the nanoscale materials were more effective than with Kocide (conventional fungicide). The current ndings align with a previous study from our group that focused on seed treatment or foliar application of nanoscale S in tomato 50 . Speci cally, foliar application of 1 mg/plant significantly decreased disease by 47.6% and increased shoot biomass by 55.6%. Importantly, the disease control efficacy was 3-fold greater than hymexazol (conventional fungicide).
The mechanism of action for disease suppression clearly depends on the material used. Nanoscale Cu is known to uniquely stimulate plant defense systems and secondary metabolism 48 , although under su cient concentrations, it can also directly act as an antimicrobial agent. Conversely, nanoscale silica can stimulate the deposition of cell water material in the roots, boosting the physical barrier protection against invading pathogens 46,47 . Nanoscale S is somewhat similar to Cu; it is known to stimulate plant defense and the glutathione system, but also directly acts as an antimicrobial agent at su cient doses 49 .
For example, Cao et al. 24 reported that the expression of select pathogenesis and oxidative stress genes were signi cantly upregulated by 11−352%, although TEM con rmed the presence of nanoscale S in the plant stem, potentially yielding direct antifungal effects.

S accumulation
In the 16-day experiment, disease had no impact on root S content at day 4 and 8, but by harvest, the diseased root S content was 40% greater than the healthy controls (Fig. S13a). All S treatments increased the root S content at 16 d (range 25.2%-47.0%), although the nanoscale S treatments increased those levels more rapidly. Similar to the roots, the presence of disease signi cantly increased stem S content in the untreated plants by day 16 (Fig S13b). Interestingly, all S treatments increased stem S content at 4 d (10.0%-14.4%), had no impact at 8 d and signi cantly reduced S content at 16 d (range 11.9%-24.8%).
Similar to the roots and stems, disease signi cantly increased leaf S content at 4 d (Fig. 2c). All S treatments further increased leaf S content (6.1%-27.0%) at 4 d; these effects were more marginal at 8 d but were again statistically signi cant at 16 d.
The dynamics of tissue S content as a function of time, disease and treatment are complex. It is clear that changes in tissue S content directly correlate with infection, often not evident at 4 or even 8 d but signi cantly greater by 16 d. Even in the absence of treatment, this is a clear indication of increased S demand as a function of biotic stress and likely activation of secondary defense pathways that utilize S.
Interestingly, S amendment did impact S tissue content, although this varied signi cantly with time, disease status, and treatment. In treated diseased plants, the impact of particle size and coating seemed to be minimal, suggesting that S demand as a function of disease was the driving force for changes in the utilization pattern of this nutrient. ) and yields adenosine-5′-phosphosulfate (APS) that is reduced to sul de (S 2-) and incorporated into cysteine (Cys) 53 .
At 4 d, ATP sulfurylase 1 (ATPSA1) expression was signi cantly downregulated with bS treatment in infected plants (Fig. 3b). No signi cant changes in nS and cS treatments occurred at any time points, indicating no S deprivation and the S uptake through this pathway was not signi cantly affected 54 .
Conversely, at 8 d the expression of thiosulfate transferase (TST), which is a gene related to the elemental sulfur accumulation pathway, is signi cantly upregulated by both nanoscale treatments (Fig. 3a). Importantly, TST expression is unaffected by bS treatment. These ndings indicate different pathways of sulfur accumulation and assimilation as a function of particle size. In plants, Str are found in the mitochondria and chloroplasts, as well as the cytoplasm and plastids. Niu et al. 51 isolated a Str gene similar to AtStr1 from wheat that was resistant to the powdery mildew fungus Erysiphe graminis 55 .
Similarly, Walz et al. 56 isolated a rhodanese-like protein displaying similarity to AtStr17 from the phloem exudates of Curcubita maxima and reported its involvement in stress and defense response by acting as a phytohormone and/or in a signaling pathway.

Tissue Nutritional Content
The effect of S treatment on the uptake of select nutrients was also assessed (Fig. S14-16, S28-29, and Table S17). Stem accumulation of Cu, Fe and Zn in the diseased plants were greater than in the healthy plants (Fig. S14). Treatment with S did not further in uence stem content of these nutrients, except for Zn content with cS. Although stem Mn content was not affected by disease, exposure to the nS and cS signi cantly increased Mn concentration in this tissue. All S amendments, particularly cS, further increased leaf Fe content, whereas Ca content was signi cantly reduced with nS and cS (Fig. S15).
Conversely, exposure to S did impact Cu and Zn leaf content. The translocation of Mn and Si was not affected by disease. However, under S amendment, levels of these nutrients were increased, particularly with nS and cS. In roots (Fig. S16) to 16 d. Although nutrient level changes were evident with treatment, consistent trends were not evident, suggesting that the dynamics of these processes is complex and the impact on overall plant health is di cult to ascertain. The amount of bioavailable nutrient elements in soil was evaluated to examine whether the changes in root element accumulation were affected the S based exposure (Fig. S17). In the DTPA extract, higher Zn content was found in cS treated soil than nS, bS, and the controls; cS resulted in higher soluble P content than bS in the same DTPA extract. However, such increments were not observed in the extracts by DI or CaCl 2 solution. However, in general, the bioavailable nutrient element content in soil was only marginally affected by treatment.

Gene expression
Several genes related to the S assimilation pathway were evaluated as a function of treatment and time indicating that nanoscale treatments had alleviated the need for response 17,48 . The time-dependent nature by which nanoscale treatments inhibit disease progress is notable and supports previous ndings suggesting a relatively narrow window of physiological opportunity where these crop protection strategies can be successful.
WRKY6 codes for a plant resistance protein that provides speci c immunity by recognizing F. oxysporum f. sp. lycopersici effectors 59 . This recognition results in effector-triggered immunity (ETI), a rapid plant defense response that inhibits successful infection. ETI occurs only between speci c plant cultivars and pathogen strains based on the presence of corresponding R and Avr proteins. As shown in Fig. 3, at 8 d, nS and cS signi cantly upregulated WRKY6 expression in infected tomato plants compared to control and bS by 280-365% and 212-282%, respectively. Again, expression was signi cantly greater at 8 d than at 4 or 16 d (Fig. 3c).
A hierarchical clustering analysis was performed to differentiate the impacts of S-based materials on the expression of S assimilation and disease-related genes in leaves harvested at 4, 8, and 16 d. The metadata heatmap (Fig. S21a) shows the overall effect of exposure time and treatment on gene expression. A two-way ANOVA ( Fig. S21b and Table S4) shows the effect of the two main factors, as well as their interaction effect, on the regulation of the 13 genes. The overall correlation heatmap and correlation analysis is shown in Fig. S22-23 and Table S5-7. CS was the gene with the greatest correlation with treatment; for exposure time, TST had the highest correlation. A correlation analysis performed with WRKL6 and γGCS revealed closely related expression to SRLK4 and ERF4, respectively.
To probe the effect of particle size and surface coating, the above analysis was performed within each S material group. The correlation heatmap ( Fig. S24 and Table S8-10) displays different patterns for nS, cS, and bS. A Debiased Sparse Partial Correlation (DSPC) network analysis ( Fig. 4 and Table S11-12) reveals the differential effect of the S materials on the relationship between the analyzed genes. The edges represent the association measures between the two ends. Interestingly, in both nS and cS treatments, TST expression was strongly correlated with WRKL6; however, with bS, this relationship was much weaker. This highlights the importance of elemental S uptake in nanoscale treatments yielding disease suppression. In addition, TST expression was correlated with TSRF1 and SAMS2 in both nS and cS treatments, but no such relationship was found for bS. These results indicate different mechanisms by which nS, cS, and bS effect plant metabolism and impact disease. A correlation analysis of the differentially affected relationship between all genes as a function of S type is shown in Fig. S25. The PLS-DA analysis ( Fig. S26 and Table S13) shows a clear separation of nS and cS, revealing a signi cant impact of exposure time on gene expression in these two treatments; this time-dependent effect is not evident in bS treatment.
An interesting time-dependent effect was evident after further data analysis at different time points. The correlation heatmap (Fig. S27 and Table S14-16) shows different patterns at 4 d, 8 d, and 16 d. The score plot obtained from PLS-DA (Fig. 5) shows that the separation between each group (nS, cS, bS, and control) was much clearer at 8 d than at 4 d or 16 d. Importantly, according to the VIP score analysis ( Fig.   5 and Table S18), the rank of TST among all the genes increased from the 9 th at 4 d to the 1 st at 8 d, and then decreased to 6 th at 16 d. Similarly, the VIP score of WRKY6 increased from 8 th at 4 d to 3 rd at 8 d, and then declined to 13 th at 16 d. These data suggest a time sensitive physiological window whereby nanoscale S treatments can signi cantly impact disease course 17 . A correlation analysis and the DSPC network of gene expression within each treatment was conducted at 4 d, 8 d, and 16 d, respectively (Fig.  S30 and Table S19). Interestingly, the expression of TST gene starts to strongly correlate to WRKY6 at 8d, and that remains through 16 d. This correlation does not occur at 4 d. This result is consistent with the biomass data, which again support a nanoscale-speci c mechanism of disease suppression through the enhanced elemental S uptake in nS and cS treatments.

Metabolomics
A total of 229 metabolites in tomato leaves were identi ed and semi-quanti ed. The relationship between all the metabolites and the enrichment analysis is shown in Fig. S31 and S32. The score plot obtained from the multivariate partial least-squares-discriminant analysis (PLS-DA) provides a general overview of the clustering information between groups and highlights good separation of nS and cS treatments from the bS and control groups at 16 d along the second principal component (PC2); however, bS and control are not well separated with each other (Fig. S33 and Table S20). There was no noticeable separation between the control, nS, cS, and bS groups at 8 d. A clear separation was found within each S based treatment by 16 d. This time-dependent response is in line with the gene expression and S accumulation data discussed above.
The heatmap (Fig. S35) shows the abundance of representative metabolites that affected the metabolic pathways related to disease responses. This systematic positive modulation of metabolic processes in tomato leaves upon nS and cS treatment suggests a generally bene cial impact on tomato metabolism. Speci c metabolic pathways in the leaves were compared between nS and bS and also between nS and cS. At 8 d, nS enhanced 4 metabolic pathways and downregulated 3 pathways relative to bS (Fig. S36). Uridine monophosphate (UMP), L-glutamine, and sphinganine were the metabolites that were present at increased concentration with nS. The enhanced sphingolipid metabolism and aminoacyl-tRNA biosynthesis pathway was found at both 8 d and 16 d but the downregulation of anthocyanin biosynthesis, starch/sucrose metabolism, and avone/ avonol biosynthesis was only evident at 8 d. At 16d, nS resulted in signi cantly different metabolism of pyrimidine, glycerophospholipid, the folatederived single carbon pool, phenylalanine, phenylpropanoid biosynthesis, sphingolipid, lysine degradation, and aminoacyl-tRNA biosynthesis as compared to bS (Fig. S37). Speci cally, nS signi cantly increased the level of tetrahydrofolate, sphinganine, and L-pipecolate compared to bS by 110.3%, 56.9%, and 123.6%, respectively, which likely enhanced the metabolic pathways of single carbon pool compounds (folate), aminoacyl-tRNA biosynthesis, sphingolipid, and lysine degradation. Conversely, the phenylalanine pathway was downregulated by nS compared to bS (by 119.0%). Other important metabolites associated with the above affected pathways, such as choline phosphate and ferulate, were all greater with nS treatment than bS. Among the 8 affected pathways, 4 were enhanced and 1 was inhibited by nS as compared to bS.
The phenylpropanoid pathway results in the production of compounds responsive to pathogen infection and is a complex network regulated by multiple gene families that exhibit regulatory mechanisms that are involved in the production of anti-microbial compounds and signaling molecules. Manipulation of this pathway enhanced defensive systems, including salicylic acid and antimicrobial compounds 60 .
Separately, sphingolipids are structural components of membranes and endomembrane systems and contribute to uidity and other cellular functions, including defense against both abiotic and biotic stressors. The folate metabolic pathway in uences a salicylic acid-independent interaction with plant immunity. Taken together, a nS-speci c upregulation of these important metabolites is clearly indicative of enhanced plant defensive activity across a range of pathways and aligns well with the gene expression data.
In comparing pristine and coated nanoscale sulfur treated tomato leaves at 8 d, cS enhanced 3 metabolic pathways and downregulated 3 pathways relative to nS (Fig. 38). D-glucose 6-phosphate, D-fructose 6phosphate, pyridoxal, and 12-oxophytodienoic acid were present at greater concentrations with cS; the decreased pathways included lower levels of UMP, L-glutamine, and L-proline with cS. Interestingly, the modulated regulation of these 6 pathways was observed only at 8d; these changes were not evident at 16 d, highlighting a time and surface coating-dependent effect on cellular metabolism. At 16 d, signi cant differences were found in 5 metabolic pathways (Fig. S39). Treatment with cS resulted in signi cant enhancement of isoquinoline alkaloid biosynthesis, tyrosine metabolism, and phenylalanine metabolism relative to nS; this was mainly due to the increased levels of tyramine and phenylacetaldehyde. Conversely, nS increased sphingolipid metabolism more than cS, as evident by the increased amount of sphinganine by 56.9%.
There were additional individual metabolites in tomato leaves that were uniquely modi ed by disease or Pipecolic acid, a non-proteinaceous product of lysine catabolism, is an important regulator of immunity in plants and accumulates upon infection, enhances resistance, and has been associated with systemic acquired resistance 68,69 cS also increased delphinidin content by 36% relative to bS, which is an anthocyanidin and an antioxidant. Fatty acids are associated in the early interactions between plants and pathogens, triggering a form of immunity that may help resist infection and colonization by pathogens 70 . cS increased palmitic acid by 1.4, 2.0 and 2.1-fold over the control, nS, and bS, and increased palmitic amide, a primary fatty acid amide derived from palmitic acid that has been reported to alleviate Fusarium disease in watermelon and tomato, by 1.7-fold over bS 71,72 . Although jasmonic acid and methyl jasmonate were unaffected treatment, a signi cant increase in the precursor 12-oxo-Phytodienoic acid (12-OPDA) was induced by cS (2.9-fold) compared to controls. In addition to the link with jasmonic acid activity, 12-OPDA plays an independent role in signaling and mediating resistance to pathogens and pests. 73 Accumulation of high 12-OPDA levels correlated with reduced ROS and elevated GSH 74 .
Interestingly, no such increase was found with nS and bS. Tyrosine serves as a precursor of numerous specialized metabolites that have diverse physiological roles as electron carriers, antioxidants, attractants, and defense compounds 76 . Also, tyrosine-derived metabolites, such as tocopherols (vitamin E), plastoquinone, cyanogenic glycosides and suberin, have crucial roles in plant tness. nS and cS also increased p-coumaric acid, a phenolic acid, by 1.5 and 1.2fold over bS and the controls. p-Coumaric acid exerts bene cial effects against several diseases due to its high free radical scavenging, pathogen suppression 77 , and antibacterial activities 78 .

Rhizosphere microbiome analysis
There were no signi cant differences in the number of amplicon sequence variants (ASVs) recovered between treatments, indicating minimal in uence on the rhizosphere bacterial diversity, although the mean for the disease control was lower than the healthy or sulfur treated soils (Fig. S42). Similarly, the Shannon's diversity index returned no signi cant differences between the treatments, suggesting that the diversity of the community was generally resilient to the treatments. The taxonomic composition of the bacterial communities was also characterized at the phylum level (Fig. S42). Generally, the composition of the community, including the phyla Proteobacteria, Actinobacteriota, and Bacteroidota, was similar across all treatments. A PCA was performed (Fig. S42) which shows that the healthy controls clustered independently and signi cantly from the other groups (Table S21), suggesting the fungal pathogen exerted some effect on the rhizosphere community, independent of the form of sulfur added. Finally, a biomarker analysis was performed to identify if any bacterial taxa were enriched. Among the sulfur treatments, the enrichment of several taxa, particularly the Gammaproteobacteria and Proteobacteria, distinguished the bS treatment (Fig. S42), indicating that there were some taxa that differed in their response to the form of the sulfur amendments. Thus, taken together, these data point to small but measurable shifts in the tomato rhizosphere bacterial communities in response to the sulfur treatments, with some slight differences between the different sulfur types employed.
In conclusion, this study demonstrates that soil amended nS and cS signi cantly suppressed Fusarium disease in tomato plants as compared to bS. The time-dependent gene expression and metabolomic pro le highlight 8d as a critical period for NP-plant-pathogen interaction and disease suppression.
Speci cally, the expression of many defense and stress-related genes was signi cantly enhanced at 8d, whereas many of these alterations were reduced at 16d. Similarly, unique metabolomic pro les were evident 8d and 16d but not earlier. The potential mechanisms of nS and cS mediated disease suppression was investigated at the molecular level and was shown to be closely correlated with the enhanced expression of S bioassimilation and disease defense-related genes, increased disease resistance and plant immune system related metabolites, and more importantly, the unique S assimilation pathway in which elemental S was directly transferred into plant tissue. This unique accumulation pathway allowed more e cient utilization of S 0 NPs by the plants and avoided the excessive formation of SO 3 2and S 2which could be harmful to plant cells. This nding is supported by the upregulation of gene expression related to the element S assimilation pathway in nS/cS treatments, the increased S content in plant tissues and the unique S translocation ratio in nS/cS treated samples relative to bS, as well as by two-photon microscopic images and videos. The ndings demonstrate that soil application of nS and cS at an appropriate time offers great potential as a novel crop defense management strategy for disease suppression and signi cantly advances efforts to develop sustainable nanoscale treatments to achieve food security.   Relative expression of tomato S bio-assimilation and disease resistance-related genes (TST, ATPSA1, WRKY6; a-c) in leaves of Fusarium-infected tomato upon soil exposure to nano S (nS), coated nano S (cS), and bulk S (bS) at 200 mg/L at 4d, 8d, and 16d after transplanting to infested media normalized to

Declarations
Page 28/30 the healthy controls. Error bars correspond to the s.e.m. (n=9). Values shown by different letters are signi cantly different at p ≤ 0.05 (one-way ANOVA with a Tukey post hoc test). "*" represents signi cant difference between the diseased control and each S treatment at p ≤ 0.05 (Student t-test). "^" indicates signi cant difference between bS and each S-based NPs treatment at p ≤ 0.05 (Student t-test).