Antibacterial Activities of Prenylated Isoflavones from Maclura tricuspidata against Fish Pathogenic Streptococcus: Their Structure-Activity Relationships and Extraction Optimization

Streptococcus zoonotic bacteria cause serious problems in aquaculture with clinical effects on humans. A structure-antibacterial activity relationships analysis of 22 isoflavones isolated from M. tricuspidata (leaves, ripe fruits, and unripe fruits) against S. iniae revealed that prenylation of the isoflavone skeleton was an important key for their antibacterial activities (minimum inhibitory concentrations: 1.95–500 μg/mL). Through principal component analysis, characteristic prenylated isoflavones such as 6,8-diprenlygenistein (4) were identified as pivotal compounds that largely determine each part’s antibacterial activities. M. tiricuspidata ripe fruits (MTF), which showed the highest antibacterial activity among the parts tested, were optimized for high antibacterial activity and low cytotoxicity on fathead minnow cells using Box–Behnken design. Optimized extraction conditions were deduced to be 50%/80 °C/7.5 h for ethanol concentration/extraction temperature/time, and OE-MTF showed contents of 6,8-diprenlygenistein (4), 2.09% with a MIC of 40 µg/mL. These results suggest that OE-MTF and its active isoflavones have promising potential as eco-friendly antibacterial agents against streptococcosis in aquaculture.


Introduction
Global fish production peaked at about 179 million tons in 2018, with aquaculture representing 46% of total fish production. Of produced fish, 52% were used for human consumption (excluding non-food uses). The total first sale value of aquaculture production in 2018 was estimated at USD 250 billion [1]. However, disease outbreaks are considered to be a significant constraint to the development of the aquaculture sector. Economic losses due to diseases globally have been estimated to be in the range of several billion US dollars per year [2]. In Korea, the olive flounder (Paralichthys olivaceus) is one of the most commercially important marine flatfish species for aquaculture. The production of olive flounder was 43,320 tons in 2019, which was 50.8% of the total production output, the highest proportion for a single fish aquaculture species in Korea (Korean Statistical Information Service, KOSIS, http://kosis.kr, accessed on 24 February 2021).
However, infectious bacterial diseases such as streptococcosis, vibriosis, and edwardsiellosis are major problems for olive flounder aquaculture in Korea [3]. Streptococcosis caused by bacteria in the genus Streptococcus is one of the major causes of mortality of farmed olive flounder in Korea [4]. Additionally, Streptococcus is an important zoonotic bacteria. It not only causes serious problems in aquaculture, but also has clinical effects on Sigma (St. Louis, MO, USA). Brain heart infusion agar (BHIA) and broth (BHIB) were bought from Difco (Sparks, MD, USA).

Plant Materials
Dried leaves of M. tricuspidata (MTL) were purchased from a local herbal market in Chungbuk, Korea in October 2013. They were identified by the herbarium of the College of Pharmacy at Chungbuk National University, where a voucher specimen was deposited (CBNU201310-MTL). Fresh unripe fruits (MTU) and fresh ripe fruits (MTF) of M. tricuspidata were purchased from a local herbal market in Hampyeong-gun, Korea in May 2015 and October 2015, respectively. Their voucher specimens (CBNU201505-CTUF and CBNU201510-CTRF) were deposited at the herbarium of the College of Pharmacy, Chungbuk National University.

LC-Q-TOF MS Analysis of Isoflavones in Extracts of M. tricuspidata Leaves, Ripe Fruits, and Unripe Fruits with Principal Component Analyses (PCA)
Analysis of isoflavones in extracts was performed with an AB Sciex (Framingham, MA, USA) ExionLC coupled to an X500R Q-TOF mass spectrometer equipped with an electrospray ionization (ESI) ion source. Each extract (methanolic extracts from each part, 1 mg/mL in methanol, 10 μL) was injected into a Kinetex C18 column (150 × 4.6 mm, 5 μm; Phenomenex, CA, USA) connected to a short pre-column. The column was operated at 40 °C. The mobile phase consisted of 0.1% formic acid water solution (A) and acetonitrile (B). The following elution gradient was used: 0 to 30 min, linear gradient from 5 to 100% B; 30 to 40 min isocratic at 100% B; 40 to 40.1 min, linear gradient from 100 to 5% B; 40.1 to 50 min, isocratic at 5% B. The optimized LC-Q-TOF MS conditions were as follows: curtain gas, 25 psi.; ion source gas 1 and gas 2, 50 psi; gas temperature, 400 °C; ion spray voltage, 5500 V; declustering potential, 80 V; and flow rate, 1 mL/min (using 1/5 splitter). Information-dependent acquisition (IDA) mode was used to automatically trigger the MS/MS spectra acquisition when a chromatographic signal exceeded a threshold of 100 counts per second. Mass spectra were acquired in the m/z range of 100 to 1000. Molecular masses of precursor ions were accurately detected using reference masses. Analysis of isoflavones in extracts was performed with an AB Sciex (Framingham, MA, USA) ExionLC coupled to an X500R Q-TOF mass spectrometer equipped with an electrospray ionization (ESI) ion source. Each extract (methanolic extracts from each part, 1 mg/mL in methanol, 10 µL) was injected into a Kinetex C 18 column (150 × 4.6 mm, 5 µm; Phenomenex, CA, USA) connected to a short pre-column. The column was operated at 40 • C. The mobile phase consisted of 0.1% formic acid water solution (A) and acetonitrile (B). The following elution gradient was used: 0 to 30 min, linear gradient from 5 to 100% B; 30 to 40 min isocratic at 100% B; 40 to 40.1 min, linear gradient from 100 to 5% B; 40.1 to 50 min, isocratic at 5% B. The optimized LC-Q-TOF MS conditions were as follows: curtain gas, 25 psi.; ion source gas 1 and gas 2, 50 psi; gas temperature, 400 • C; ion spray voltage, 5500 V; declustering potential, 80 V; and flow rate, 1 mL/min (using 1/5 splitter). Information-dependent acquisition (IDA) mode was used to automatically trigger the MS/MS spectra acquisition when a chromatographic signal exceeded a threshold of 100 counts per second. Mass spectra were acquired in the m/z range of 100 to 1000. Molecular masses of precursor ions were accurately detected using reference masses.
PCA was performed to differentiate extracts according to semi-quantitative data of identified isoflavones. This allowed identification of isoflavones that most significantly  Korea, 1999), and S. iniae BS9 (Tongyoung in Republic of Korea, 1998) were originally isolated from a diseased olive flounder and identified by 16s rRNA gene sequencing. After subculturing in brain heart infusion broth (BHIB) for cryopreservation, aliquots of all bacterial strains were kept frozen at −80 • C in BHIB containing 14% glycerol until use.

Antibacterial Susceptibility Test
Antibacterial activity was evaluated with a broth dilution method (approved guideline: M49-A) as described by the Clinical and Laboratory Standards Institute [25] with slight modifications [26,27]. Briefly, bacterial colonies taken directly from brain heart infusion agar (BHIA) plates were incubated into BHIB and cultured at 25 • C for 24 h. From this culture, a suspension equivalent to 0.5 McFarland standard in BHIB was prepared. Isoflavones dissolved in BHIB (including 5% of DMSO (v/v) or less) and an equal volume of 1 × 10 6 CFU/mL of bacteria were mixed in a 96-well plate and incubated at 25 • C for 24 h. Amoxicillin and oxytetracycline were used as reference controls. The lowest concentration of each antibiotic that visibly inhibited bacterial growth was considered the minimum inhibitory concentration (MIC). The minimum bactericidal concentration (MBC) was also determined. Briefly, the bacterial suspension at or above the MIC (20 µL of each well) was inoculated into a fresh broth (200 µL) and incubated for 25 • C for 24 h more. The lowest concentration where no growth was visually observed was considered the MBC. In the case of MBC/MIC ratio ≤ 4, the effect was considered to be bactericidal [28]. Each assay was repeated three times.

Time-Growth Curve and Scanning Electron Microscope (SEM) Analyses
Time-growth curve and SEM analyses were performed to investigate effects on bacteria of optimized extract from MTF (OE-MTF) against S. iniae DSJ19. Time-growth curves were made by the same method as in Section 2.6. Antibacterial susceptibility test. Absorbance was measured at 600 nm at 2 h intervals using the kinetic mode of a VERSA max microplate reader (Molecular Devices, CA, USA) at 25 • C. Amoxicillin, used as a reference control, was also tested. The same volume of BHIB (without any antibiotic and bacteria) was set as the blank.
The effect of OE-MTF on S. iniae was investigated by SEM following the method by Yun et al. [29]. with slight modifications. Briefly, 40 mL of S. iniae suspension (1 × 10 7 CFU/mL) was mixed with OE-MTF at MIC of 40 µg/mL for 20 h. The control was prepared by mixing equal volumes of bacterial suspension without OE-MTF. Following the incubation, the bacterial pellet was harvested by centrifugation at 3500× g for 10 min and thoroughly washed 3 times with phosphate-buffered saline (PBS) to eliminate the residue of OE-MTF. Both treated and control bacteria pellets were treated in 2.5% glutaraldehyde and 1% osmium tetroxide and then analyzed using a scanning electron microscope (JSM-7610F Plus, JEOL, Tokyo, Japan).

Cytotoxicity Assay
The fathead minnow (FHM) cell line (American Type Culture Collection No. CCL-42) was cultured in Dulbecco's modified Eagle medium (DMEM, Gibco, NY, USA) supplemented with 10% fetal bovine serum (FBS, Gibco, NY, USA), 50 IU/mL of penicillin, and 50 µg/mL of streptomycin (Gibco, NY, USA). This cell line was maintained at 20 • C. FHM cells were plated into 96-well plates at a density of 10 5 cells/well and cultured overnight. Cytotoxicity was evaluated with a neutral red (NR) uptake assay as described by Thompson [30] with slight modifications [31,32]. All samples were dissolved in DMSO and diluted in a medium to adjust the final concentration of DMSO to be 0.1% (v/v) or less. FHM cells (10 5 cells/well, 96-well plates) were treated with a medium containing 2-fold serially diluted samples and incubated at 20 • C for 96 h. Then, an NR working solution (50 µg/mL neutral red dye, Sigma, MO, USA) was added to each well and incubated at 20 • C for 2 h. Plates were washed with phosphate-buffered saline (PBS) twice. A solution containing 1% acetic acid in 50% ethanol was then added to each well to extract the dye for 10 min. The absorbance of the colored solution was measured at 540 nm and 690 nm with a VERSA max microplate reader (Molecular Devices, CA, USA). The growth rate as an index of cytotoxicity was calculated by dividing the absorbance of test cells by the absorbance of corresponding control cells. The 50% cytotoxic concentration (CC 50 ) was calculated using Microsoft Excel.

Box-Behnken Design (BBD)
Three-factor BBD was applied in this study to investigate individual and interactive effects of ethanol concentration (X 1 ), extraction temperature (X 2 ), and extraction time (X 3 ) as variables on antibacterial activity against S. iniae DSJ19 and cytotoxicity to FHM cells. These studied factors, along with their experimental levels, are presented in Table S1 (tables and figures marked with S found in Supplementary Materials). The whole design consisted of 15 experimental points carried out in a random order. A total of 3 replicates at the center of the design were used to allow for the estimation of a pure error sum of squares. Based on experimental data, regression analysis was performed and fitted into an empirical second-order polynomial model: Y = β 0 + β 1 X 1 + β 2 X 2 + β 3 X 3 + β 12 X 1 X 2 + β 13 X 1 X 3 + β 23 X 2 X 3 + β 11 X 1 2 + β 22 X 2 2 + β 33 X 3

2
(1) where Y was the predicted response; β 0 was the interception; β 1 , β 2 , and β 3 were linear coefficients of the ethanol concentration, extraction temperature, and solvent-to-sample ratio, respectively; β 12 , β 13 , and β 23 were interaction coefficients of the ethanol concentration, extraction temperature, and solvent-to-sample ratio, respectively; and β 11 , β 22 , and β 33 were squared coefficients of the ethanol concentration, extraction temperature, and solvent-to-sample ratio, respectively. Linear, interaction, and squared coefficients were determined by least square regression followed by analysis of variance (ANOVA) using Minitab 14 software (PA, USA). Statistical significance was considered at p < 0.05.

Quantification of 6,8-Diprenylgenistein (4) in OE-MTF and 15 Extracts from BBD
Stock solution for 6,8-diprenylgenistein (4) was prepared with HPLC methanol as the solvent. Working calibration solutions were prepared by successive serial dilution of the stock solution with methanol to yield final concentrations of 5, 10, 25, 50, 100, 250, and 400 ng/mL. OE-MTF, and 15 extracts from BBD (Table S1) were dissolved in 50% methanol to prepare stock solutions at 1 mg/mL. These stock solutions were diluted to make working solutions at 1, 2.5, and 5 µg/mL for OE-MTF and 10 µg/mL for 15 extracts from BBD. An Exion LC (AB Sciex, Framingham, MA, USA) coupled to an X500R Q-TOF mass spectrometer equipped with an ESI ion source was used for quantitative and qualitative analyses. Detailed LC and MS conditions are described in Supplementary Materials: 2. Quantification of 6,8-diprenlygenistein (4) in OE-MTF.

Method Validation
An external standard method was utilized for quantification. Linearity was studied by diluting stock solution to a series of at least five concentrations. Calibration plots were then constructed after triplicate analysis of the solution by plotting the mean integrated chromatographic peak area against the corresponding concentration. The limit of detection (LOD) and limit of quantification (LOQ) were calculated following International Conference on Harmonization (ICH) Q2B Guidelines.
The precision of the method was validated by determining intra-and inter-day variances. Intra-day precision was performed with three replications prepared from combined extract within one day. Inter-day precision was performed over three consecutive days. Relative standard deviation (RSD) was taken as a measure of precision.
Recovery was used to further evaluate the accuracy of the method. Known amounts of standard solutions were mixed with known amounts of samples. Resultant samples were then extracted and analyzed with the proposed method. Triplicate experiments were repeated at each level. Average recoveries were estimated with the following formula: recovery (%) = (amount found − original amount)/amount spiked × 100%; RSD (%) = (SD/mean) × 100%.
Specificity was achieved by analyzing multiple reaction monitoring (MRM) signals ( Figure S3). All peaks of target compounds in the OE-MTF sample were identified by comparing retention time, parent ions, and product ions with standards in MRM spectra. All results related to quantitative analysis are included in Supplementary Materials: 2. Quantification of 6,8-diprenlygenistein (4) in OE-MTF.

Statistical Analysis
Analysis of variance (ANOVA) was used to identify the main effect, curvature effect, and interaction effect of major factors by extraction conditions using MINITAB release 14 (Minitab Inc., State College, PA, USA). Statistical significance was considered at p < 0.05 for all analyses.

Results and Discussion
3.1. Antibacterial Activities of Isoflavones from M. tricuspidata against S. iniae and Their Structure-Activity Relationships (SAR)

Effects of Prenyl Group-Addition
The antibacterial activities of 22 isoflavones from M. tricuspidata were evaluated using S. iniae (Table 1). The MIC of compound 1, an isoflavone without any prenyl substituent, was >500 µg/mL. However, compounds 3, 4, 5, 6, 7, 8, and 9 as prenylated isoflavones showed excellent antibacterial activities against S. iniae, with MIC values of 1.95 to 62.5 µg/mL. Among them, di-prenylated compounds 4, 5, 7, and 9 had MICs of 1.95-15.63 µg/mL, showing much more enhanced antibacterial activities than monoprenylated compounds 3, 6, and 8. Cyclized compounds showed also similar results. Compound 10, in which the prenyl group of C-6 was cyclized with the hydroxyl group of C-7, showed an MIC > 500 µg/mL while compound 14 had an MIC of 62.5 µg/mL. The only difference between these two compounds was that compound 14 had a hydroxy-prenyl group at C-8. Recent studies have also reported that amphiphilic features of isoflavone skeletons due to the addition of a prenyl group play an important role in their antibacterial properties [33], and that di-prenylated flavanones exhibit higher antibacterial activities than mono-prenylated ones [19,34].

Effect of Cyclization of Prenyl Group
It was found that the cyclization of the prenyl group in prenylated isoflavones decreased their antibacterial activities against S. iniae. Compound 10, which was cyclized between the hydroxyl at C-7 and the prenyl group at C-6 of the A-ring of compound 8 (MIC of 7.81 µg/mL), showed markedly reduced antibacterial activity with an MIC > 500 µg/mL. Similar results were observed for compound 11 (MIC of 250 µg/mL) vs. compound 3 (MIC of 62.5 µg/mL) and compound 20 (MIC of 125 µg/mL) vs. compound 6 (MIC of 31.25 µg/mL).
In addition, even for cyclized ones, the existence of a linear prenyl group at C-6 (compound 22, MIC of 31.25 µg/mL) was more potent than the one with a linear prenyl group at Previous studies have also reported that the presence of a prenyl group at C-6 of the A-ring can improve antibacterial activities [18,33,35]. All samples were dissolved in DMSO and diluted in broth with the final concentration of DMSO to be 5% (v/v) or less. a molecular weight; b L, F, U = leaves, ripe fruits, and unripe fruits, respectively; c minimum inhibitory concentration; d minimum bactericidal concentration; OTC, oxytetracycline; AMX, amoxicillin; CFU, colony forming unit.

Effects of Other Substituents
The hydroxyl group in the B-ring influenced the antibacterial activities of prenylated isoflavones against S. iniae. The presence of a di-hydroxyl group at B-ring as in compounds 2 and 12 (MIC: 500 and 31.25 µg/mL, respectively) resulted in a stronger antibacterial activity against S. iniae than the presence of a mono-hydroxyl group, as in compounds 1 and 10 (MIC > 500 µg/mL). The difference in antibacterial activity with respect to the number of the hydroxyl groups in the B-ring might be related to the affinity of these hydroxyl groups to proteins. According to Wang et al. [36], bacterial neuraminidase inhibitory activity is increased when the number of the hydroxyl groups of the B-ring in prenylated isoflavones is increased from 1 to 2. Dhayakaran et al. [37] have also reported that hydroxyl groups in

Antibacterial Activities of Prenylated Isoflavones against Fish Pathogenic Clinical Strains of Streptococcus
The five most active prenylated isoflavones, 4, 5, 7, 8, and 21, also showed excellent antibacterial activities against fish pathogenic clinical strains of S. parauberis and S. iniae, with MICs ranging from 1.95 to 31.25 µg/mL. In particular, di-prenylated isoflavones 4 (6,8diprenylgenistein) and 7 (isoerysenegalensein E) exhibited bactericidal activities against Streptococcus, showing an MBC/MIC ratio ≤ 4. Furthermore, their bactericidal activities against S. parauberis were higher than those of amoxicillin, a clinically used antibiotic for streptococcosis in farmed fish (Table 2). Many studies have reported on the excellent antibacterial activities of prenylated (iso)flavonoids against methicillin-resistant Staphylococcus aureus (MRSA). However, they lack absolute specificity in the mechanism of action of this class [33]. Generally, cytoplasmic membrane disruption is known to be the dominant mechanism of action of these (iso)flavonoids [19]. Because of their relatively high hydrophobicity, prenylated isoflavones are expected to show high affinity for the cytoplasmic membrane [19,38]. Based on this evidence, the excellent antibacterial activities of prenylated isoflavones against Streptococcus strains shown in the present study might be due to their high hydrophobicity.

Antibacterial Activities of Extracts from Each Part of M. tricuspidata against Fish Pathogenic Bacteria and Principal Component Analysis (PCA) Using LC-Q-TOF MS
To evaluate the field applications of extracts containing prenylated isoflavones having excellent antibacterial activities, we investigated their antibacterial activities against Grampositive S. iniae and S. parauberis and Gram-negative E. tarda and A. salmonicida. Among MTL, MTF, and MTU extracts, MTF and MTL extracts only showed potent antibacterial activities against fish pathogenic Gram-positive S. iniae and S. parauberis, with MIC values of 62.5 to 1000 µg/mL. The MTF extract showed the highest antibacterial activities against fish pathogenic Gram-positive S. iniae and S. parauberis (MIC values of 62.5 and 250 µg/mL, respectively, Table 3). This suggests that it has a high potential to be used for further development. To better analyze and visualize similarities and differences among different parts for the antibacterial activities of isoflavones, multivariate data analyses were performed using data from LC-Q-TOF MS. PCA was carried out with a relative amount of 1000 peaks in order of the highest impact for M. tricuspidata extracts (Figure 2). In further analysis, 22 isolated isoflavones of M. tricuspidata extracts (leaves, ripe fruits, unripe fruits) were found in PCA data, and results are shown in Figure 2B. The first principal factor (PC1) explained 78.2% of the variation across samples, whereas the second principal factor (PC2) explained 21.4% of the variance. The cumulative variance contribution of PC1 and PC2 was 99.6%. In the PCA factor score plot (Figure 2A), MTF and MTL were negatively correlated with PC1, whereas MTU was positively correlated with PC1. In addition, MTL was negatively correlated while MTF was positively correlated with PC2 to distinguish these two extracts. In PC1 of the loading plot ( Figure. 2B), most of the isolated compounds showed a negative correlation, indicating that compound compositions of MTF and MTL were similar. However, compounds 5 and 8 (MTL extract), compounds 10 and 15 (MTU extract), and compound 4 (MTF extract) exhibited heavier factor loadings, indicating that these compounds had distinctive features from each extract. This result is similar to a previous report showing that flavonoids with a side chain of cyclization between hydroxyl and prenyl groups at the A-ring are predominant in unripe fruits, whereas flavonoids with a linear prenyl side chain are the main components in ripe fruits [11]. As shown in Table 1, compounds 4, 5, and 8 had MICs of 3.91 to 7.81 µg/mL against S. iniae, while compounds 10 and 15 had MICs > 500 µg/mL. Similar results were observed for MTF and MTL extracts. They showed much higher antibacterial activities against Gram-positive bacteria than the MTU extract (Table 3). These results indicate that each part's antibacterial activity and the antibacterial activity of isoflavones isolated from each part are highly correlated. Considering that the antibacterial activity of MTF against S. iniae was higher than that of MTL or MTU, MTF-containing compound 4 was considered to be of high value as a material. It could be used as a feed additive. Thus, it was selected for further optimization procedures in the present study. In addition, the use of this ripe fruit as an ingredient in dietary supplements and functional foods ingredients is being actively investigated in many fields [11]. These advantages might make it a promising candidate source for commercialization in aquaculture.
that the antibacterial activity of MTF against S. iniae was higher than that of MTL or MTU, MTF-containing compound 4 was considered to be of high value as a material. It could be used as a feed additive. Thus, it was selected for further optimization procedures in the present study. In addition, the use of this ripe fruit as an ingredient in dietary supplements and functional foods ingredients is being actively investigated in many fields [11]. These advantages might make it a promising candidate source for commercialization in aquaculture.

Statistical Analyses and Model Fitting of BBD
A common perception of natural product safety and high consumer acceptance with a belief that "natural" equals "safe" is not only false, but also misleading [39]. To maximize the antibacterial efficacy of the extract without causing toxicity to farmed fish, excellent antibacterial activity and low cytotoxicity were selected as important factors for optimization.
The BBD matrix applied both actual and predicted values of antibacterial activity against S. iniae DSJ19 and cytotoxicity to FHM cells, as shown in Table S1. When all cases were considered, experimental data were well fitted by second-order polynomial models (R 2 : 99%). In addition, since experimental and predicted responses were very similar to each other, unexplained variance among experimental data was considered irrelevant. According to ANOVA of a regression model, linear and quadratic terms were significant (p < 0.05), indicating that the relationship between the response variable and test variables was not simply a linear one (Table S3). All mathematical models obtained were found to be suitable for the analysis of the response surface, since there was no evidence of inadequacy based on the lack-of-fit test (p > 0.05).
Analysis of regression coefficients (Tables S3 and S4) showed that all linear terms (X1, X2, and X3) were negative in all evaluated responses, whereas all quadratic terms (X1 2 , X2 2 , and X3 2 ) were positive (p < 0.05). This meant that antibacterial activity could be enhanced by X1, X2, and X3 and that there was a curvature effect in the model. In addition, both antibacterial activity and cytotoxicity were positive in quadratic terms, meaning that cytotoxicity and antimicrobial activity showed the same pattern. Regarding the antibacterial activity, as the ethanol concentration and extraction temperature increased, the

Statistical Analyses and Model Fitting of BBD
A common perception of natural product safety and high consumer acceptance with a belief that "natural" equals "safe" is not only false, but also misleading [39]. To maximize the antibacterial efficacy of the extract without causing toxicity to farmed fish, excellent antibacterial activity and low cytotoxicity were selected as important factors for optimization.
The BBD matrix applied both actual and predicted values of antibacterial activity against S. iniae DSJ19 and cytotoxicity to FHM cells, as shown in Table S1. When all cases were considered, experimental data were well fitted by second-order polynomial models (R 2 : 99%). In addition, since experimental and predicted responses were very similar to each other, unexplained variance among experimental data was considered irrelevant. According to ANOVA of a regression model, linear and quadratic terms were significant (p < 0.05), indicating that the relationship between the response variable and test variables was not simply a linear one (Table S3). All mathematical models obtained were found to be suitable for the analysis of the response surface, since there was no evidence of inadequacy based on the lack-of-fit test (p > 0.05).
Analysis of regression coefficients (Tables S3 and S4) showed that all linear terms (X 1 , X 2 , and X 3 ) were negative in all evaluated responses, whereas all quadratic terms (X 1 2 , X 2 2 , and X 3 2 ) were positive (p < 0.05). This meant that antibacterial activity could be enhanced by X 1 , X 2 , and X 3 and that there was a curvature effect in the model. In addition, both antibacterial activity and cytotoxicity were positive in quadratic terms, meaning that cytotoxicity and antimicrobial activity showed the same pattern. Regarding the antibacterial activity, as the ethanol concentration and extraction temperature increased, the antibacterial activity also increased. Similarly, cytotoxicity increased with increasing ethanol concentration.
Phenolic compounds found in plants, including flavonoids, simple phenols, and phenolic derivatives, are generally water-soluble [40,41]. These phenolic compounds can also be extracted with organic solvents, particularly aqueous ethanol [42]. Although hot water extraction is a commonly used extraction method for functional ingredients in medicinal plants, previous studies have shown that using organic solvents can shorten the extraction time and increase the extraction efficiency and antibacterial activity [27]. In the present study, 6,8-diprenylgenistein (4), the active compound, is a hydrophobic molecule having a di-prenyl group. Ethanol could be more efficient than water when it is extracted from MTF [14,19]. Therefore, ethanol as an organic solvent was considered as the most suitable solvent during the extraction of bioactive substances such as antibacterial compounds from MTF. The relationship between dependent and independent variables was further elucidated by constructing a response surface plot. The effects of X 1 and X 2 with their interactions on antibacterial activity and cytotoxicity at a fixed level of X 3 (mid-level) are shown in Figure S2.

Multiple Responses Optimization of MTF Extracts and Predictive Capacities of Mathematical Models
The calculated optimal concentration of ethanol and extraction temperature for obtaining the maximal antibacterial activity against S. iniae and low cytotoxicity of MTF extracts toward FHM cells were 50% and 80 • C, respectively ( Figure S2B and Table 4). Regarding the extraction time, there was no significant difference in the extraction yield for each extraction condition. Thus, a central value (7.5 h) was used ( Figure S1). Under these conditions, the calculated desirability indices for antibacterial activity and cytotoxicity were 0.85 and 1, respectively ( Figure S2B). The composite desirability (D) value was calculated to be 0.92 (Table 4). After the extract was prepared with optimal conditions, verification experiments were performed. OE-MTF showed potent antibacterial activity against S. iniae with an MIC of 40 µg/mL (predictive capacity of 108.5%) and a CC 50 of 153.18 µg/mL (predictive capacity of 108.9%) in the cytotoxicity assay (Table 4). Therefore, a good agreement between the predicted and experimental values was obtained, indicating a satisfactory predictive capacity of the developed mathematical model of BBD. Interestingly, the MBC (80 µg/mL)/MIC (40 µg/mL) ratio of OE-MTF against S. iniae was 2, indicating that its antibacterial action was bactericidal, which is the same as those of 6,8-diprenylgenistein (4) and isoerysenegalensein E (7), the most active compounds in MTF (Table 2 and Figure 2). Although isoerysenegalensein E (7) also exhibited excellent antibacterial activity, it was confirmed that isoerysenegalensein E (7) was present at trace levels in MTF extract through comparison with a standard compound using LC-Q-TOF MS ( Figure S4). Therefore, these results suggested that the antibacterial activity of OE-MTF may be due to the active major compound 6,8-diprenylgenistein (4) or its synergistic effect with other active compounds. The study of the correlations between the content (%) of 6,8-diprenylgenistein (4) and the antibacterial activity against S. iniae ( Figure 3A) and cytotoxicity to FHM cells ( Figure 3B) showed that both antibacterial activity and cytotoxicity were highly proportional to the content of 6,8-diprenylgenistein (4), with R 2 values of 0.99 and 0.90, respectively. When the content of 6,8-diprenylgenistein was 2% or more, the MIC value remained almost constant, whereas the cytotoxicity increased as the content increased. Therefore, the optimized extraction conditions obtained through BBD resulted in a meaningful outcome that could satisfy both antibacterial activity and cytotoxicity (the content of 6,8-diprenlygenistein (4) was 2.09% in OE-MTF). These results suggest that the antibacterial activity and cytotoxicity of MTF extract are mainly due to the content of 6,8-diprenylgenistein (4). Taken together, these results indicate that the 6,8-diprenylgenistein (4) content of MTF extract can be used for quality control of products.

Cytotoxicity According to Extraction Conditions Based on BBD
The active compound, 6,8-diprenylgenistein (4), in 15 extracts from BBD was quantified using the quantification method established through method validation (see Supplementary Materials: 2. Quantification of 6,8-diprenlygenistein (4) in OE-MTF). The study of the correlations between the content (%) of 6,8-diprenylgenistein (4) and the antibacterial activity against S. iniae ( Figure 3A) and cytotoxicity to FHM cells ( Figure 3B) showed that both antibacterial activity and cytotoxicity were highly proportional to the content of 6,8-diprenylgenistein (4), with R 2 values of 0.99 and 0.90, respectively. When the content of 6,8-diprenylgenistein was 2% or more, the MIC value remained almost constant, whereas the cytotoxicity increased as the content increased. Therefore, the optimized extraction conditions obtained through BBD resulted in a meaningful outcome that could satisfy both antibacterial activity and cytotoxicity (the content of 6,8-diprenlygenistein (4) was 2.09% in OE-MTF). These results suggest that the antibacterial activity and cytotoxicity of MTF extract are mainly due to the content of 6,8-diprenylgenistein (4). Taken together, these results indicate that the 6,8-diprenylgenistein (4) content of MTF extract can be used for quality control of products.

Effects of OE-MTF on the Growth and the Morphology of S. iniae
To further confirm the antibacterial activity of OE-MTF against S. iniae, a time-growth curve of bacteria was plotted at an MIC of 40 μg/mL. As shown in Figure 4A, OE-MTF strongly inhibited the growth of S. iniae at 40 μg/mL, showing the same degree of inhibition as that of amoxicillin. To investigate the effects of OE-MTF on the morphology of S. iniae, SEM analysis was performed. As a result, untreated bacteria ( Figure 4B) displayed the regular cocci morphology of S. iniae. In contrast, OE-MTF induced distinct changes in the morphology of S. iniae, such as a lysed and indistinguishable cytoplasmic membrane structure ( Figure 4C). These results suggested that OE-MTF disrupted the cytoplasmic membrane of S. iniae, leading to their death and disintegration, and could act as a

Effects of OE-MTF on the Growth and the Morphology of S. iniae
To further confirm the antibacterial activity of OE-MTF against S. iniae, a time-growth curve of bacteria was plotted at an MIC of 40 µg/mL. As shown in Figure 4A, OE-MTF strongly inhibited the growth of S. iniae at 40 µg/mL, showing the same degree of inhibition as that of amoxicillin. To investigate the effects of OE-MTF on the morphology of S. iniae, SEM analysis was performed. As a result, untreated bacteria ( Figure 4B) displayed the regular cocci morphology of S. iniae. In contrast, OE-MTF induced distinct changes in the morphology of S. iniae, such as a lysed and indistinguishable cytoplasmic membrane structure ( Figure 4C). These results suggested that OE-MTF disrupted the cytoplasmic membrane of S. iniae, leading to their death and disintegration, and could act as a bactericidal agent against S. iniae. This bactericidal property of OE-MTF is probably due to bactericidal prenylated isoflavones such as 6,8-diprenylgenistein (4). This was confirmed through the correlation study, which exhibited that antibacterial activities of MTF extracts depended on the content of 6,8-diprenylgenistein (4). Similarly, Araya-Cloutier et al. [18]. Reported that extracts rich in prenylated isoflavonoids showed potent antibacterial and bactericidal activities against Gram-positive bacteria such as L. monocytogenes and MRSA.
to bactericidal prenylated isoflavones such as 6,8-diprenylgenistein (4). This was confirmed through the correlation study, which exhibited that antibacterial activities of MTF extracts depended on the content of 6,8-diprenylgenistein (4). Similarly, Araya-Cloutier et al. [18]. Reported that extracts rich in prenylated isoflavonoids showed potent antibacterial and bactericidal activities against Gram-positive bacteria such as L. monocytogenes and MRSA.

Conclusions
As a result of SAR analysis of 22 isoflavones from M. tricuspidata extracts (leaves, ripe fruits, and unripe fruits), it was confirmed that the prenyl group of the isoflavone skeleton

Conclusions
As a result of SAR analysis of 22 isoflavones from M. tricuspidata extracts (leaves, ripe fruits, and unripe fruits), it was confirmed that the prenyl group of the isoflavone skeleton was an important key for antibacterial activity against S. iniae. Some characteristic prenylated isoflavones containing 6,8-diprenylgenistein (4) (MTF), 6,8-diprenylorobol (5) (MTL), and alpinumisoflavone (10) (MTU) were identified as compounds that largely determined the antibacterial activity of the extract of M. tricuspidata from each part through PCA analysis. In addition, the content of 6,8-diprenlygenistein (4) in extracts of M. tricuspidata ripe fruits showed a positive correlation with antibacterial activity and cytotoxicity to FHM cells. The optimal extraction conditions of M. tricuspidata ripe fruits for antibacterial activity and cytotoxicity were successfully obtained using BBD. Based on the results of BBD, the optimal extraction conditions were an ethanol concentration of 50%, an extraction temperature of 80 • C, and an extraction time of 7.5 h. Furthermore, a good agreement between predicted and experimental responses showed a satisfactory predictive capacity of mathematical models developed. Under these optimized conditions, OE-MTF showed potent antibacterial activity, with an MBC/MIC ratio of 2, suggesting a bactericidal action. In addition, OE-MTF exhibited the disintegration of bacterial cytoplasmic membranes and cell disruption against S. iniae in the SEM analysis. Taken together, the results of the present study may provide a basic clue that the extract of M. tricuspidata ripe fruits and its active compounds can be applied to control Gram-positive fish pathogens. However, further in vivo efficacy and toxicity studies are needed to better understand the therapeutic efficacy and mechanism of action of OE-MTF and its active compounds in farmed fish, which could ultimately lead to their application as eco-friendly antibacterial agents against streptococcosis and aid food safety by reducing the use of antibiotics in the aquaculture industry.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

Conflicts of Interest:
The authors declare no conflict of interest.
Sample Availability: Some of samples are available from the authors.