Plant material and explant preparation
The seeds of T. polium were obtained from the Research Center of Agriculture (Jahad-e Keshavarzi) of Mashhad-Iran. After surface sterilization with aqueous ethanol (70% v/v, for 30 sec) and sodium hypochlorite solution (5% v/v, for 3 to 5 min), the seeds were kept in Petri dishes with sterilized filter paper and gibberellic acid (1000 ppm) under 10 ± 2 °C and dark condition. After the appearance of the radicle and the top of the cotyledon, the samples were exposed under standard condition (2000 lux light, 16 hr light/ 8 hr dark, 25 ± 2 °C, and 50% relative humidity) for one week. Then the little seedlings were transferred into pots containing Hoagland's complete nutrient solution (Hoagland and Arnon 1950) and kept in a growth chamber under the condition as mentioned earlier. Detached leaf explants (as the best explant reported in our previous research) were immersed in sodium hypochlorite (5% v/v) for 3 min and aqueous ethanol (70% v/v) for 30 sec and then rinsed three timeswith sterile distilled water.
Establishment of callus induction
The disinfected leaf pieces (approximately 5–5 mm2) were placed on MS media supplemented with 12 treatments of various concentrations of BAP (0, 0.5, 1, and 1.5 mg L-1) and NAA (0, 0.5, and 1 mg L-1). Hormone-free MS medium was used as a control. Each treatment had three replication with five leaf explants (for each jar) per replicate. The cultures were maintained in the darkness for two weeks and then transferred to a condition with a 16 hr light/ 8 hr dark for four weeks. The first subculture was carried out after six weeks on MS media augmented with the same hormones. At the end of the eight weeks, the percentage of callogenesis was counted as the number of callus / total number of explants × 100. The weight of fresh and dried calli [dried in an oven (Memmert, 5-1486) at 40 °C for 24 hr] was expressed in gram mass. According to callus induction with 100% of callogenesis and callus fresh weight, the green calli of the best sources were selected for further studies. The whole process was repeated twice.
Elicitors' preparation and treatment
The elicitation effect of different concentrations of MeJA and nano-sized TiO2 was analyzed. Nano-sized TiO2 was provided by the Nanomaterials Pioneers Co, NANOSANY (Mashhad, IRAN). Shows a transmission electron microscopy (TEM) of nano-sized TiO2 powder with the particles exhibiting an approximate equi-axes shape and a size of fewer than 25 nm (Fig. 2). In brief, a certain amount of MeJA and nano-sized TiO2 were dissolved in absolute ethanol (96% v/v) and sterile distilled water, respectively (Shoja et al. 2022). After using ultrasonic vibration by providing a 40 kHz wave at 40 °C for 25 min for nano-sized TiO2 stock, both elicitors were filtered and sterilized through a membrane filter (0.22 μm) and used in the present study. In the continuation of the research and after selecting the best MS culture, the calli of media containing 1 mg L-1 BAP, 1.5 mg L-1 BAP, and 1.5 mg L-1 BAP plus0.5mg L-1 NAA were transferred to fresh MS media supplemented with MeJA (0, 50, 100, and 200 μM) and nano-sized TiO2 (0, 10, 60, and 120 mg L-1) for two weeks.
Extract preparation
To quantify the set of phytochemical compounds, calli cultures were extracted by ultrasound-assisted extraction according to the protocol of Annegowda et al. (2012), under the following conditions; methanol concentration 80% (v/v), extraction time and temperature of 30 °C, and 30 min, respectively. After filtration, with filter paper, the solvent of each extract was evaporated under air and dark condition for 24 hr. Eventually, the extract powders were prepared in 5 mL of methanol solvent (80% v/v) and used as hydromethanolic extracts for phytochemical analysis.
Secondary metabolite quantification
Total Polyphenols
An aliquot (100 µL) of methanolic extract was mixed with 2.5 mL of Folin Ciocalteu reagent 10% (v/v), and after 5 min of rest in the dark, 2 mL of sodium carbonate (7.5% v/v) was added. Before reading the absorbance at 765 nm, the samples were placed at 25 ± 2 °C for 1.5 hr. The content of phenolic compound was calculated based on the standard of gallic acid equivalents (GAE), and the result was calculated in terms of mg GAE per 100 g DW of the callus sample (Marinova et al. 2005).
Total Ortho-diphenols
The content of total Ortho-diphenol compound was based on the method of Carrasco-Pancorbo et al. (2005). Accurately, 100 μL of methanolic extract was mixed with 2 mL of methanol (50% v/v) and 0.5 mL of sodium molybdate (5% w/v) and subsequently placed in a dark room for 15 min. The absorbance was read at 370 nm, and the total content of compounds was calculated as a gallic acid equivalent, in terms of mg GAE per 100 g DW of the callus sample.
Total flavonoids
Based on the method described by Zhishen et al. (1999), 3.3 mL of methanol (30% v/v), 0.5 μL of 0.5 M sodium nitrite, and 150 μL of 0.3 M aluminum chloride were added to 300 μL of methanolic extract. After incubating for 5 min in the dark, 1 mL of 1 M sodium hydroxide was added to the samples and the absorbance was measured at 510 nm. The flavonoid content in extracts was expressed in terms of catechin equivalents (mg CATE) in 100 g DW.
Total flavones
The flavone content was expressed as quercetin equivalent (QE)(Kosalec et al. 2004). A sample volume (50 μL) of methanolic extract was mixed with 50 μL of aluminum-chloride (10%, w/v), 50 μL of 1 M potassium acetate, 750 μL of methanol (80% v/v), and 1.4 mL of distilled water. After incubation for 30 min under darkness, the absorbance was measured at 415 nm. The flavone content was expressed as mg QE for 100 g DW.
Total phenolic acids
Using a test tube, 250 μL of methanolic extract, 1.25 mL of distilled water, 250 μL of HCL, and 250 μL of Arnow's reagent [sodium-molybdate 10% (w/v) and sodium nitrite 10% (w/v)] was mixed. After 30 min of incubation in the dark, 250 μL of sodium hydroxide and 250 μL of distilled water were added, and then, the absorbance was read at 490 nm using a spectrophotometer (Jasco7800, Germany). The contents of total phenolic acid content in the extracts were determined based on an equation attained from the standard caffeic acid curve (Matkowski et al. 2008).
Rosmarinic acid
By using the method of Öztürk et al. (2010), 920 μL of ethanol (96% v/v) and 40 μL of the extract (1 mg dry extract in 1 mL-1 absolute methanol) were introduced to 40 μL of 0.5 M zirconium oxide chloride. After 5 min, the absorbance was defined at 362 nm. The content of rosmarinic acid in the extract was determined based on the rosmarinic acid curve (mg RA for 100 g DW).
Antioxidant activity
DPPH (2,2-diphenyl-1-picrylhydrazyl) radical scavenging assay
To measure the antioxidant assay of the extracts, 100 μL of different concentrations of methanolic extract was mixed with 100 μL of methanolic DPPH (0.2 mM) solution as a sample. The methanolic extract (100 μL, 50:50 v/v%) and methanolic DPPH solution (100 μL, 50:50 v/v%) were used as blank and control samples, respectively. After 30 min of incubation in the dark, the absorbance was recorded by the ELISA Reader (Gentaur, ST2100) at 490 nm. In order to make a comparison, butylated hydroxytoluene (BHT) was tested under the same condition as a positive control. The total antioxidant activity in the extracts were determined based on the equation scavenging activity (%) = [1- (AS – AB) / (AS – AC)] × 100
Here, AS is mixture absorbance with the sample, AB is mixture absorbance with blank, and AC is mixture absorbance with control. The extract concentration providing 50% inhibition (IC50) was from the graph of scavenging effect percentage against extract concentration (de Torre et al. 2019).
Ferric antioxidant power assay (FRAP)
To measure the antioxidant capacity, 20 μL of methanolic extract was added to 130 μL of FRAP solution. After 30 min of rest, the reduction of ferric iron (Fe3+) to ferrous iron (Fe2+) in the presence of antioxidants was recorded by ELISA Reader at 630 nm. The antioxidant content of the sample was expressed in mg of ferrous sulfate per 100 g DW of callus (Li et al. 2017).
LC-MS/MS analysis
Standards
The following standards apigenin (4′,5,7-trihydroxyflavone) and luteolin (3′,4′,5,7-tetrahydroxyflavone) purchased from Sigma Co and prepared in HPLC grade methanol. The samples were recognized by comparing the standard calibration of apigenin and luteolin.
Sample preparation
Form the calli with the highest content of phenolic derivatives; approximately 0.5 g of dry matter was suspended in 5 mL of absolute methanol (96% v/v) and incubated in an ultrasonic bath for 2 hr at 30 ± 2 °C and 40 kHz frequency. After filtration by a 0.45 μm filter, the remaining solvent (methanol) was evaporated at room temperature. To extract flavones, the residue (0.5 mg of each extract) was dissolved in 1 mL of methanol 75% (v/v) and formic acid 0.1% (v/v) solvent according to a procedure described by Gómez et al. (2018).
LC-MS/MS
To provide a brief review of LC-MS/MS system model based on Shoja et al. (2022), the separation of flavones was accomplished with an Atlantis T3 column, 100 Å, 3 µm, 2.1 mm X 150mm at a flow rate of 0.2 mL min-1. The solvent system was a mixture of acetonitrile and water containing formic acid 0.1% (v/v) that were combined in a gradient. The mobile phase was sonicated and degassed before injection into HPLC. The temperature of 40 ± 2 °C was held for 5 µL sample injection. The chromatographic peaks of the flavones mentioned above, in the sample solutions were confirmed by comparing their retention time and UV spectrum with those of the reference standard.
Data analysis
Statistical analysis of data was evaluated by analysis of variance (ANOVA) followed by using the SPSS data analysis software package (version 25 for windows). All results presented are mean ± standard error and a probability of P ≤ 0.01 and P ≤ 0.05 was considered significant.
ANN modeling
ANN model with feed-forward multiplayer perceptron and K-fold cross-validation was designed by MATLAB software 2020a. The units of this model comprise input/output, weights, hidden layer, and activation function. The mathematical explanation of the fundamental ANN is as follows:
O (Outputs) = f (wx+bias); W (Weights) = w1, w2, . . . ,wn; X (inputs) = x1, x2, . . ., xn.
In the present study, the layouts of ANN were two inputs (with ten levels), one output, and a single hidden layer. Inputs consist of the selected concentrations of BAP and NAA, as well as MeJA and nano-sized TiO2 and outputs comprise total phenolic, Ortho-diphenols, flavonoids, flavones, phenolic acids, rosmarinic acid, DPPH, and FRAP of leaf-derived calli. In order to simplify the analysis, each of the output variables was separately analyzed.
Presents the flow chart of the neural network optimized by K-fold cross-validation. Based on this, the data is randomly broken up into K groups, after which, for each group, the following operations are performed (Fig.3) (Lyu et al., 2022): 1) Select one of the training folds as the testing dataset, 2) The remaining K-1 groups are used as the training set, and 3) Use the selected training dataset to predict the model. Figure 4 shows the final network architecture of ANN used in this study for secondary metabolites prediction.
In a small sample dataset of the present work, the data was classified into two categories to develop the ANN model: training (70%) and testing (30%). Correlation coefficient (R2), sum squared error (SSE), and relative error for training and testing (RE) were used as the main indicators to determine the best layout of ANN (Ramezannejad et al. 2022). The normal range of R2 values is 0-1, and the closer it is to 1, there is a perfect correlation between inputs and outputs. In addition, the lower the values of SSE and RE, the better the perdition of data.