The experiment with cotton (Gossypium hirsutum L.) was conducted between January and May 2019 in a greenhouse at the São Paulo State University (UNESP), Jaboticabal, Brazil. Seeds of the cultivar 954 GLT were used, commercially acquired from BASF, as it is a cultivar that presents consistent results in different regions and is highly adaptable to producing regions in Brazil. Seeds of cultivar 954 GLT, commercially obtained from BASF, were used as it is a cultivar that presents consistent results in different regions and is highly adaptable to cotton producing regions in Brazil. Plants were grown in a soilless cultivation system. Relative humidity (78.4 ± 4.4%), maximum (35.1 ± 6.3°C) and minimum temperature (22.0 ± 3.2°C) were recorded with a thermometer hydrometer in the greenhouse during the experimental period (Figure 6).
Cotton seeds were sown into trays containing sand previously washed with water and HCl solution (0.1 M). After germination, four seedlings were transplanted to plastic pots with a 7 dm³ capacity (upper diameter: 16 cm; lower diameter: 11 cm; height: 33 cm), filled with 6 dm³ of sand previously washed. Thinning was performed with the plants in the vegetative growth stage F2 (two fully developed leaves), maintaining one plant per pot, which was considered as the experimental unit.
A complete nutrient solution was applied to the plants in the vegetative stage F4 (four fully developed leaves) . The nutrient solution was prepared with Fe-EDDHMA as the iron source; pH between 5.5 and 6.5, adjusted with NaOH (1 M) or HCl (1 M) solution, and with a reduction only in B concentration (from 46.2 to 33.7 µmol L-1), in order to cause moderate B deficiency in cotton plants grown in a soilless system .
The nutrient solution was applied for seven days at 20% of its total concentration, as indicated by Hoagland and Arnon . After this period, the concentration was increased to 40% for one week and then to 60%, which was maintained until the end of the experimental period.
The substrate was drained once a week to eliminate excess of salts, with 700 mL of deionized water being added to the substrate of each pot to drain the nutrient solution, which was then discarded. After 2 hours, a new nutrient solution was provided to the plants and during the rest of the growth cycle the nutrients required for plant development were supplied regularly.
Experimental design and treatments
The experiment was carried out under a completely randomized blocks design (5 x 2 + 1) with four repetitions. The following treatments were applied: five foliar B concentrations (0.0; 0.5; 1.0; 1.5 and 2.5 g L-1 of B), absence and presence (1.00 g L-1) of Si; and one control treatment with no micronutrient deficiency, adding 46.2 µmol L-1 of B to the nutrient solution during the entire experimental period.
B concentrations in the spray solution were 0%; 33%; 66%; 100% and 166% of the concentration recommended by Görmüs , where the author indicates foliar spraying of 1.5 g L-1 of B for cotton plants grown in a micronutrient-deficient environment.
The Si concentration to be mixed with the B solution was defined based on the visual evaluation of the final solution to avoid using it with evidence of polymerization, which was measured by using the turbidity index.
Turbidity index of the B+Si spray solution
To prepare the solution, boric acid (B: 175 g kg-1; density: 1.43 g cm-3; solubility in water at 20°C: 47.2 g L-1) was used as the B source, with the pH of the B spray solution adjusted to 9.0 using an NaOH solution (1M). The Si source consisted of potassium silicate stabilized with sorbitol (SiKE; Si: 115 g L-1; K2O: 113.85 g L-1; sorbitol: 100 mL L-1; pH:12.0). A potassium solution was also prepared with 47 mg L-1 of K, in the form of KCl, to balance the macronutrient between treatments.
Two turbidity index assessments of the B+Si spray solution were performed to determine the homogeneity of the Si solution, since an increase in its value would indicate advancing Si polymerization. For this test, the B concentration of the solution was fixed at 2.5 g L-1, the highest concentration studied in this experiment. Assessment I consisted of adding Si to the B solution at concentrations of 1.00; 1.25 and 1.50 g L-1 of Si; assessment II used a B mixture (2.5 g L-1) with smaller intervals between Si concentrations of 1.00; 1.05; 1.10; 1.15; 1.20 and 1.25 g L-1 in order to increase the accuracy of the concentration in the spray solution. In both assessments, SiKE was used as Si source and H3BO4 as B source in the solution with a final volume of 50 mL, pH adjusted to 9.05 ± 0.02 using a solution of HCl (1 M) or NaOH (1 M) at a temperature of 20°C, with three repetitions. After the spray solution was prepared, the turbidity index was measured with a Tecnopon® microprocessed turbidity meter (model TB1000) at 0; 30; 60; 90; 120; 180; 240; 300 and 360 minutes, and photographs were taken using a camera with 9238 x 6928-pixel resolution.
All beakers with the different solutions were placed in front of a completely black background and were then photographed. The images were adjusted using Adobe Photoshop® in order to improve the contrast with the black background without changing B+Si solution color.
Foliar application in the treatments
The solution was prepared according to the different treatments and immediately applied to the plants with a manual sprayer until run-off, in order to cover all the leaves of the plants. A 5.0 mL volume of the solution was applied to each pot. Sprayings occurred between 7 and 8 am, starting with cotton plants in the reproductive stage B1 (first completely developed flower bud), with four foliar applications four days apart. Temperature and relative humidity were measured by a thermometer hydrometer during applications, resulting in a temperature between 20 and 23°C, and relative humidity above 85%, conditions favorable for foliar spraying .
During foliar applications, all pots were covered with cotton to avoid any dropping or spillage from the plant shoot after being sprayed with the B+Si solution to the substrate, in order to guarantee that the elements absorption was totally foliar.
Non-enzymatic antioxidant system (proline and glycine-betaine)
Leaves were collected three weeks after foliar application of the B+Si mixture, and immediately frozen in liquid nitrogen, then placed in a freezer at -80°C to assess the oxidative stress.
The rest of the shoot (branches and stem) was also collected, washed with water, detergent solution (0.1%), HCl solution (0.1 %) and deionized water, being then placed in a forced air circulation oven at 65 ± 5°C until constant mass. Next, the branches and stems were weighed, ground in a Wiley mill and stored for subsequent assessments.
In order to assess the oxidative stress, malondialdehyde (MDA) and hydrogen peroxide (H2O2) content were determined on the leaves.
Initially, 0.4 grams of the frozen plant material was weighed, maintained at -80°C, and ground with 20% (m/v) polyvinylpyrrolidone and 0.1% trichloroacetic acid (TCA). The material was centrifuged at 11,000 rotations per minute at 4°C for 10 minutes. The supernatant was separated into Eppendorf tubes containing a 20% TCA solution and 5% thiobarbituric (TBA) acid. The samples were incubated in a water bath for 30 minutes at 95°C, transferred to an ice bath for 10 minutes to stop the reactions, and then centrifuged at 11,000 rotations per minute at 4°C for 10 minutes. Next, the samples were read in a spectrophotometer at wavelengths between 535 and 600 nanometers and MDA calculated using an extinction coefficient of 1.55 10-5 mol-1 cm-1 .
H2O2 content was determined by homogenizing the frozen ground plant tissues with 0.1% TBA, followed by centrifugation at 11,000 rotations per minute at 4°C for 10 minutes. The supernatant was also transferred to Eppendorf tubes containing a buffer solution (pH 7.5) and potassium iodine and then incubated for one hour in an ice-filled container. Next, spectrophotometric readings were conducted at a wavelength of 390 nm, in line with the methodology described by Alexieva et al. .
Proline content was determined using the method proposed by Bates et al. . Plant material was defrosted at ambient temperature and 0.5 g of fresh matter was macerated in liquid nitrogen and subsequently 2 mL of sulfosalicylic acid was added, followed by adding more 8 mL of the same reagent. The grinded material was doubly filtered in a glass funnel using filter paper. After filtering, 1 mL of glacial acetic acid, 1 mL of ninhydrin acid and 1 mL of plant extract were pipetted into a glass test tube. The tubes were then agitated and placed in a water bath at 100°C for 1 hour and then into an ice bath to stop the reaction. Toluene was added (2 mL), followed by a 20-second agitation. Spectrophotometric readings were performed at 520 nm, adding 0.5 mL of the supernatant in a quartz cuvette.
Glycine-betaine content was determined according to the methodology proposed by Grieve and Grattan . For that, 1 g of frozen plant material was placed in paper bags and dried in a forced air circulation oven at 80ºC for four days, then manually grinded in a crucible. Extracts were prepared by adding 5 mL of deionized water to 0.125 g of macerated material, which remained under agitation for 24 hours at 25ºC. The extracts were mixed at 1:1 with H2SO4 (2 M), then maintained in an ice bath for 1 hour. Next, 0.1 mL of KI-I2 was added to the tubes, which were then agitated and kept at 4º C for 16 hours. The KI-I2 solution was previously prepared by diluting 15.7g of iodine and 20g of K in 100 mL of distilled water. The tubes were then centrifuged at 3500 rotations per minute for 15 minutes at 0ºC. The supernatant was discarded, leaving periodate crystals, which were dissolved in 4.5 mL of 1,2-dichloroethane. Two hours and 30 minutes later absorbance was read at a wavelength of 365 nm in a Beckman DU 640 spectrophotometer and the glycine-betaine content was calculated.
Nutritional analysis of B, Si accumulation and plant shoot dry matter production
After analysis of oxidative stress and nonenzymatic antioxidant compounds, leaves were defrosted at ambient temperature and then washed with water, neutral detergent solution (0.1%), HCl solution (0.1%) and deionized water. After decontamination, leaves were placed in a forced air circulation oven at 65 ± 5°C until constant mass and then weighed. Next, the leaves were manually grinded in a crucible and mixed with the previously collected macerated plant shoot samples (branches and stems).
Shoot dry matter consisted of the leaf dry matter left over from dry matter analysis of branches and stems collected initially. After the plant material was mixed, B content was determined by a dry digestion of the samples, burning in a muffle furnace at 400°C, followed by colorimetric reaction with azomethine-H and colorimetric spectrophotometric reading . Si content was determined from an alkaline digestion of the plant material with H2O2 and NaOH in an oven at 90°C for 4 h , followed by colorimetric reaction with ammonium molybdate in acid medium (oxalic acid and hydrochloric acid), being then determined by colorimetric spectrophotometric readings .
B and Si accumulation were calculated as the product of B or Si content and shoot dry matter. In addition, B use efficiency was calculated as the ratio between the square of plant shoot dry matter values and plant shoot B content, as described by Siddiqi and Glass .
The data obtained were submitted to analysis of variance (F-test), and when significant, to polynomial regression or exponential growth.
The exponential growth model was used to study polymerization. Singular models with one or two parameters were tested, and the best fit models applied, that is, those with the highest regression coefficient at 5% using the T-test.
The other variables were submitted to polynomial regression, the linear and quadratic mathematical models tested, and those with the best fit were applied. The model selection criterion established was the magnitude of significant polynomial regression coefficients at 5% probability using the T-test. When significant, the maximum and minimum points were obtained by the derivation of equations.
Statistical analyses were conducted with Sisvar® software  and the graphs formulated with Sigmaplot®.