Antifungal Activity against Fusarium oxysporum of Botanical End-Products: An Integration of Chemical Composition and Antifungal Activity Datasets to Identify Antifungal Bioactives

Plants produce various compounds as defensive barriers to naturally control fungal diseases. Among them, vascular wilt caused by Fusarium oxysporum is one of the most destructive diseases in crops, causing relevant economic losses. The application of synthetic fungicides is the most used management for this disease. However, this kind of method also involves adverse environmental impacts. Therefore, alternative methods are continuously being developed as a strategy to be involved in integrated pest management programs. Thus, as part of our research on antifungals of plant origin, a group of botanical extracts was assessed for the respective inhibitory effect on mycelium and conidia of F. oxysporum. Mycelial growth inhibition was measured in 12-well plates containing amended semi-solid medium, whereas conidial susceptibility was determined through microdilution. The identification of the bioactive compounds among test extracts was performed using an indirect approach, consisting of the integration of chemical composition and antifungal activity datasets through single-Y orthogonal partial least squares (OPLS) regression. Results showed that Piper aduncum extract was the most potent mycelial growth inhibitor whereas P. elongatum exhibited the best effect on conidia susceptibility. The active compounds identified through statistical integration and subsequent isolation were piperaduncin C, asebogenin and (−)-methyllinderatin. These findings indicated that the integrative, indirect approach is useful for the identification of bioactive metabolites from botanical extracts to be further used as biological protective agents against this phytopathogen.


Introduction
Plants are well-known factories and huge reservoirs of naturally-occurring compounds, many of them used by plants in ecological relationships to be able to relate and subsist [1]. These compounds, usually comprising specialized metabolites (formerly called secondary), are characterized by being biologically active and employed in defensive roles, among others, to maintain a natural balance with the environment, pathogens and predators [2]. Indeed, some of these metabolites are defensive barriers against microbial diseases, especially those caused by soil-borne fungal pathogens [3]. Among them, different Ascomycota fungi, particularly Fusarium species, are phytopathogens that affect several economically-important crops. In this group, Fusarium oxysporum is one of the most widespread pathogens globally, comprising more than 120 reported forma specialis [4]. This fungal phytopathogen can cause many losses through the vascular disease so-called 'Fusarium Wilt'. Consequently, agronomic practices have been developed for controlling this disease, such as biological control, cultural practices and chemical control. The most common agricultural practice is oriented to chemical control by using synthetic fungicides since such a strategy is versatile and easily adopted [5]. Currently, the most used compounds to control the presence of this phytopathogen are mainly nitrogen-containing heterocyclic environment, and genetic factors have been revealed for some Piper plants, influencing the biological activity of a particular plant source [20] and, consequently, expanding the scope of Piper-based products as a source of antifungals.
The microscopic effect of the most active extract (i.e., P. aduncum leaves, PaL) on F. oxysporum mycelium was also recorded. In the control treatment, normal mycelial growth and shape was observed, including a normal spore production ( Figure 1A). In contrast, dithane caused visible damage in fungal structures ( Figure 1C), particularly related to hyphal damage and cytoplasm disintegration [30]. However, dithane also stimulated the overproduction of viable spores as a response to the fungicide in comparison to the control [31]. This outcome coincided with previous reports since the active principle mancozeb has been reported as a sporulation promoter on F. oxysporum [32]. Thus, despite the important inhibitory activity on mycelial growth, the sporulation stimulation by dithane could be considered an adverse effect on controlling the pathogen proliferation at the field level. growth (MGI > 68%). Contrarily, extracts obtained from Cedrela odorata, Senna viarum, Tithonia diversifolia, and Virola elongata exhibited low inhibitory activity against F. oxysporum (MGI < 30%). The microscopic effect of the most active extract (i.e., P. aduncum leaves, PaL) on F. oxysporum mycelium was also recorded. In the control treatment, normal mycelial growth and shape was observed, including a normal spore production ( Figure 1A). In contrast, dithane caused visible damage in fungal structures ( Figure 1C), particularly related to hyphal damage and cytoplasm disintegration [30]. However, dithane also stimulated the overproduction of viable spores as a response to the fungicide in comparison to the control [31]. This outcome coincided with previous reports since the active principle mancozeb has been reported as a sporulation promoter on F. oxysporum [32]. Thus, despite the important inhibitory activity on mycelial growth, the sporulation stimulation by dithane could be considered an adverse effect on controlling the pathogen proliferation at the field level.
In the case of P. aduncum, the effect was more evident even in comparison to the other test extracts and positive control, since several damages on fungal structures were observed. PaL-derived extracts caused loss of cytoplasm, cell reduction and lysis ( Figure 1B), which suggested that the chemical components of this extract would target cell wall or cell membrane. Other effects were also detected, such as agglomeration, hyphal crosslinking, and considerable increase of vacuoles [33] and, finally, the spore production was reduced and the effect was found to be fungicide. These effects can be rationalized due to this active extract could contain lignans and flavonoids, common phytoconstituents reported for P. aduncum [34]. Such phenolic compounds can disrupt the cell wall of pathogenic fungi, promoting its osmotic lysis [35,36], and other subsequent fungal malfunctions and anomalies, as reported previously for phenolic-rich extracts and individual phenols against phytopathogenic fungi [37,38]. On the other hand, the conidial susceptibility on non-germinated F. oxysporum conidia suspensions for the entire set of extracts at 0.1% (w/v) was also evaluated. Since the assessment of conidial susceptibility through microdilution implies an indirect extent of conidial germination inhibition (CGI) of molds [39], it represented another fungal structure to observe the effect of test extracts. However, most extracts (86.4%, n = 38) exhibited no CGI at the test dose and they were, therefore, discarded. The remaining six extracts exhibited inhibitory activity, whose results are presented in Table 2, ranging from 15 to 84% GI. The most active extracts were P. elongatum leaves (PeL) (CGI = 83.5%) and T. diversifolia stems (TdS) (CGI = 74%), being significantly different (p < 0.05) to the other test extracts. Interestingly, the extracts with high inhibitory capacity on mycelial growth (i.e., PaL and PeS) exhibited low activity on conidial germination (CGI < 25). TdF-derived extract exhibited low activity on both mycelial and conidial susceptibility assays (<20%).   In the case of P. aduncum, the effect was more evident even in comparison to the other test extracts and positive control, since several damages on fungal structures were observed. PaL-derived extracts caused loss of cytoplasm, cell reduction and lysis ( Figure 1B), which suggested that the chemical components of this extract would target cell wall or cell membrane. Other effects were also detected, such as agglomeration, hyphal cross-linking, and considerable increase of vacuoles [33] and, finally, the spore production was reduced and the effect was found to be fungicide. These effects can be rationalized due to this active extract could contain lignans and flavonoids, common phytoconstituents reported for P. aduncum [34]. Such phenolic compounds can disrupt the cell wall of pathogenic fungi, promoting its osmotic lysis [35,36], and other subsequent fungal malfunctions and anomalies, as reported previously for phenolic-rich extracts and individual phenols against phytopathogenic fungi [37,38].
On the other hand, the conidial susceptibility on non-germinated F. oxysporum conidia suspensions for the entire set of extracts at 0.1% (w/v) was also evaluated. Since the assessment of conidial susceptibility through microdilution implies an indirect extent of conidial germination inhibition (CGI) of molds [39], it represented another fungal structure to observe the effect of test extracts. However, most extracts (86.4%, n = 38) exhibited no CGI at the test dose and they were, therefore, discarded. The remaining six extracts exhibited inhibitory activity, whose results are presented in Table 2, ranging from 15 to 84% GI. The most active extracts were P. elongatum leaves (PeL) (CGI = 83.5%) and T. diversifolia stems (TdS) (CGI = 74%), being significantly different (p < 0.05) to the other test extracts. Interestingly, the extracts with high inhibitory capacity on mycelial growth (i.e., PaL and PeS) exhibited low activity on conidial germination (CGI < 25). TdF-derived extract exhibited low activity on both mycelial and conidial susceptibility assays (<20%). The outcome from both methods (i.e., mycelial growth inhibition and conidial susceptibility) cannot be compared, due to two factors, i.e., the characteristics (e.g., cell wall composition or shape) of each fungal form [40] and exposition manner of extracts in each assay (i.e., submerged cultures vs. agar surface) [41], promoting considerable differences between them. However, F. oxysporum mycelium appeared to be more susceptible to most extracts under the conditions of the in-vitro test on amended solid medium, although only six extracts exhibited an in-vitro effect on conidia susceptibility under the submerged culture characteristics, possibly by solubility issues and/or selectivity to this fungal form. In this sense, the activity evaluation on these two fungal forms is very important to be evaluated as a combined/complementary information, considering the corrective aim of mycelial growth inhibition and the preventive scope of conidial germination [42]. In addition, owing to the homogeneity and quality of inoculum the major challenge of the microdilution assay for filamentous fungi, microscopic scrutiny of conidial/spore germination is recommended to verify the information provided by the conidial susceptibility indirectly measured by optical dispersion [41]. Accordingly, all six extracts effectively

Integration of Chemical and Activity Datasets: Indirect Recognition of Bioactives
Chemical composition (based on chromatographic profiles) and antifungal activity (based on MGI) datasets were integrated through supervised statistics to indirectly recognize bioactives within test botanical extracts. They exhibited different profiles as expected, involving 125 differentiable signals. Some extracts shared particular signals having different intensities, but other signals were exclusive for a particular extract. The chromatographic profiles were decomposed into absorbance at 270 nm as a function of time (each 0.64 s) and a data matrix was then built (5064 × 44). The resulting data matrix was employed for the subsequent procedure. Thus, a supervised statistical analysis, by means of single-Y orthogonal partial-least squares (OPLS) regression (using MGI as continuous Y-variable), was used to integrate the chemical composition and biological activity of test plant extracts. The resulting statistical model (consisted of one predictive and one orthogonal component) could separate extracts, revealing an adequate explained variance (R 2 X = 0.78, R 2 Y = 0.82) and predictability (Q 2 Y = 0.46). The resulting scores plot (Figure 2A) showed the discrimination depending on the chemical composition but also according to the MGI data as a continuous Y-variable, comprising the first predictive component, t(1), versus the first orthogonal component, to (1), that explain the variations according to the MGI activity (37.2%) and chemical composition (40.8%), respectively. Such Single-Y-depending discrimination is intuitively visualized as a heat map representation (0-100% scaling) between high (red) and low (blue) MGI values. This two-factor discrimination was very important for our purpose since the scores plot revealed those extracts discriminated by chemical composition, but the chemical composition related to the MGI activity as a statistically important trend. Once the pattern recognition between datasets was found, the loadings were investigated according to their discriminating influence, using an S-line transformation. This S-line plot is a tailor-made visualization and interpretation tool for point-to-point decomposed 2D data (e.g., chromatographic profiles under ultraviolet-visible (UV-Vis) detection, infrared (IR) spectra or nuclear magnetic resonance (NMR) spectra) [43,44]. It is very useful to show the covariance loading (p(ctr)) along the chromatographic signals (i.e., variables expressed as time values) and colored according to the absolute value of the correlation loading, p(corr) [45], providing a linear indication of the correlation intensity. Hence, the single-Y OPLS-derived S-line plot visualized differences between least active (negative) and most active (positive) extracts ( Figure 2B). Consequently, three metabolites (1-3) were found to influence (p(corr) > 0.5) the OPLS-based discrimination between most and least active extracts and, consequently, they were identified as the significant metabolites to the observed MGI. Compound 1 was the most significant metabolite (p(corr) = 0.75), and occurred in P. aduncum and P. elongatum extracts, whereas compound 4 exhibited good reliability but poor correlation loading (p(corr) = 0.34). Therefore, compound 4 was not considered a good candidate in the pattern recognition of the inhibitory activity on mycelial growth.
A similar procedure was employed to the SGI results. The corresponding single-Y OPLS model also separated extracts with suitably explained variance (R 2 X = 0.76, R 2 Y = 0.93) and predictability (Q 2 Y = 0.51) and led to the plots presented in Figure 3. Similarly, integration of conidial susceptibility and chemical datasets resulted in an adequate relationship along the t(1) (39.5%) versus to(1) (36.5%). Additionally, the respective S-line plot revealed that only compound 3 showed the best correlation (p(corr) = 0.70), being the most influencing variable for this SGI assay, whereas compounds 2 and 4 were more related to the least active extracts. influence (p(corr) > 0.5) the OPLS-based discrimination between most and least active extracts and, consequently, they were identified as the significant metabolites to the observed MGI. Compound 1 was the most significant metabolite (p(corr) = 0.75), and occurred in P. aduncum and P. elongatum extracts, whereas compound 4 exhibited good reliability but poor correlation loading (p(corr) = 0.34). Therefore, compound 4 was not considered a good candidate in the pattern recognition of the inhibitory activity on mycelial growth.  Table 1. A similar procedure was employed to the SGI results. The corresponding single-Y OPLS model also separated extracts with suitably explained variance (R 2 X = 0.76, R 2 Y = 0.93) and predictability (Q 2 Y = 0.51) and led to the plots presented in Figure 3. Similarly, integration of conidial susceptibility and chemical datasets resulted in an adequate relationship along the t[1] (39.5%) versus to [1] (36.5%). Additionally, the respective S-line plot revealed that only compound 3 showed the best correlation (p(corr) = 0.70), being the most influencing variable for this SGI assay, whereas compounds 2 and 4 were more related to the least active extracts.
Our findings indicate that a correlation between plant metabolite profiles and antifungal activity datasets can be effectively done through single-Y OPLS to recognize patterns under the decomposition of the spectral data into uncorrelated latent variables. OPLS (and even PLS) differs from principal component analysis (PCA) in the maximization of the covariance of independent variables (i.e., spectral data from IR, UV, or MS) as  Table 1.
Plants 2021, 10, x FOR PEER REVIEW 7 of 14 a function of a dependent variable (i.e., biological activity), either continuous or categorical [46]. Consequently, there is an increasing number of studies focused on the application of untargeted metabolomics to integrate the chemical composition and the activity of natural product mixtures [19,47]. Metabolomics-based approaches are actually employed to profile multiple mixture components simultaneously, typically through the application of chromatographic and/or spectroscopical analyses [46]. The use of metabolomics as the source of independent variables to be integrated with a dependent variable is the major advantage of this approach [19]. Indeed, it can also recognize unstable compounds that would be lost upon purification, which is the main challenge during fractionation procedures [48,49]. However, an innate restraint of chemical and biological dataset association is the recognition of false positives, since this integration does not have causal but correlative essence [50]. Therefore, statistically-identified bioactives are required to be fractionated or isolated to validate the correlative recognition [51]. In our case, the process of isolating the compounds with the highest statistical influence on the discrimination of the most active and least active extracts was undertaken in order to validate the correlative finding, since compounds 1-3 represented the best correlation values (p(corr) > 0.6).  Table 1.

Isolation and Identification of Antifungal Metabolites
The OPLS-based integration recognized compounds 1-3 as the most influencing metabolites in the antifungal activity from both assays (i.e., mycelial growth inhibition and conidial susceptibility of F. oxysporum). To validate this indirect recognition of potential bioactives, these compounds were purified by semipreparative HPLC separation from PaL and PeL-derived extracts and structurally elucidated by nuclear magnetic resonance (NMR). This analysis supported the unequivocal identification of these compounds (i.e., 1-3) as known metabolites since their 13 C NMR profiles were identical to those reported for piperaduncin C (1) [52], asebogenin (2) [53] and methyllinderatin (3) [54], respectively. In the case of 3, its optical rotation showed the same sign (i.e., levorotatory) as that reported previously for (−)-methyllinderatin, isolated from leaves of P. aduncum [55]. Hence, the structures of these isolated, identified compounds are presented in Figure 4.  Our findings indicate that a correlation between plant metabolite profiles and antifungal activity datasets can be effectively done through single-Y OPLS to recognize patterns under the decomposition of the spectral data into uncorrelated latent variables. OPLS (and even PLS) differs from principal component analysis (PCA) in the maximization of the covariance of independent variables (i.e., spectral data from IR, UV, or MS) as a function of a dependent variable (i.e., biological activity), either continuous or categorical [46]. Consequently, there is an increasing number of studies focused on the application of untargeted metabolomics to integrate the chemical composition and the activity of natural product mixtures [19,47]. Metabolomics-based approaches are actually employed to profile multiple mixture components simultaneously, typically through the application of chromatographic and/or spectroscopical analyses [46]. The use of metabolomics as the source of independent variables to be integrated with a dependent variable is the major advantage of this approach [19]. Indeed, it can also recognize unstable compounds that would be lost upon purification, which is the main challenge during fractionation procedures [48,49]. However, an innate restraint of chemical and biological dataset association is the recognition of false positives, since this integration does not have causal but correlative essence [50]. Therefore, statistically-identified bioactives are required to be fractionated or isolated to validate the correlative recognition [51]. In our case, the process of isolating the compounds with the highest statistical influence on the discrimination of the most active and least active extracts was undertaken in order to validate the correlative finding, since compounds 1-3 represented the best correlation values (p(corr) > 0.6).

Isolation and Identification of Antifungal Metabolites
The OPLS-based integration recognized compounds 1-3 as the most influencing metabolites in the antifungal activity from both assays (i.e., mycelial growth inhibition and conidial susceptibility of F. oxysporum). To validate this indirect recognition of potential bioactives, these compounds were purified by semipreparative HPLC separation from PaL and PeL-derived extracts and structurally elucidated by nuclear magnetic resonance (NMR). This analysis supported the unequivocal identification of these compounds (i.e., 1-3) as known metabolites since their 13 C NMR profiles were identical to those reported for piperaduncin C (1) [52], asebogenin (2) [53] and methyllinderatin (3) [54], respectively. In the case of 3, its optical rotation showed the same sign (i.e., levorotatory) as that reported previously for (−)-methyllinderatin, isolated from leaves of P. aduncum [55]. Hence, the structures of these isolated, identified compounds are presented in Figure 4. Table 1.

Isolation and Identification of Antifungal Metabolites
The OPLS-based integration recognized compounds 1-3 as the most influencing metabolites in the antifungal activity from both assays (i.e., mycelial growth inhibition and conidial susceptibility of F. oxysporum). To validate this indirect recognition of potential bioactives, these compounds were purified by semipreparative HPLC separation from PaL and PeL-derived extracts and structurally elucidated by nuclear magnetic resonance (NMR). This analysis supported the unequivocal identification of these compounds (i.e., 1-3) as known metabolites since their 13 C NMR profiles were identical to those reported for piperaduncin C (1) [52], asebogenin (2) [53] and methyllinderatin (3) [54], respectively. In the case of 3, its optical rotation showed the same sign (i.e., levorotatory) as that reported previously for (−)-methyllinderatin, isolated from leaves of P. aduncum [55]. Hence, the structures of these isolated, identified compounds are presented in Figure 4.  The IC 50 values for the three purified compounds are presented in Table 3. Piperaduncin C (1) and asebogenin (2)   The three isolated compounds 1-3 shared the dihydrochalcone moiety but involved structural differences related to the hydroxyl substituent at ring B (in 2), and the dimeric and adduct forms, having the particular structural variation on ring A (in 1 and 3, respectively) ( Figure 4). Piperaduncin C (1) is a dimeric dihydrochalcone metabolite which exhibited moderate antibacterial activity against Bacillus subtilis and Micrococcus luteus [52], whereas asebogenin (2) (4 -methoxy-2 ,4,6 trihydroxydihydrochalcone) exhibited higher activity against these bacteria [52] but also against Staphylococcus aureus and methicillin resistant S. aureus [56], and antifungal activity against Candida albicans, C. krusei, C. parapsilosis and Cryptococcus neoformans (MIC < 15.6 µg/mL) [57]. Finally, (−)-methyllinderatin (3) is an interesting monoterpene-dihydrochalcone adduct and exhibited antibacterial activity against M. luteus [55]. However, the antifungal activity against Fusarium species of these compounds 1-3 has not been reported yet, so they are evaluated for the first time in the present study against mycelial growth and conidial susceptibility of F. oxysporum. The discovery of these compounds as antifungals was made possible by statistical integration of HPLC-based chemical composition and antifungal activity datasets, confirming that this approach is a valuable tool for identifying antifungals against F. oxysporum from a reasonable number of botanical end-products.

Preparation of Botanical Extracts
From an in-house collection of native and exotic plants sampled in different places of Colombia (n = 34), forty-four botanical extracts were prepared. Plants were collected in Meta, Casanare, Cundinamarca and Guaviare departments (coordinates 3.543, −73.669; 5.190, −72.547; 4.931, −74.008; 2.332, −72.616, respectively) between 2014 and 2017. Voucher specimens are kept at Colombian National Herbarium. The selected healthy plant materials (without visible damage) were separately air-dried and extracted using 96% ethanol at constant shaking speed (120 rpm) using a Heidolph Rotamax 120 platform orbital shaker (Heidolph Instruments GmbH & Co.KG, Schwabach, Germany). The extraction lasted one week with daily removal of the extract-containing solvent and replaced by fresh 96% ethanol. These mixtures were separated by filtration and the resulting solution was concentrated by distillation under reduced pressure at 40 • C using an IKA RV 10 Control rotary evaporator (IKA ® RV 10, IKA ® Werke GmbH & Co. KG, Staufen, Germany) to afford the raw extracts. Resulting raw extracts per plant from each daily extraction were collected, dried and subsequently stored at −20 • C until use.

In-Vitro Assays against F. oxysporum
F. oxysporum LQB-03, a virulent isolate obtained from wilting Physalis peruviana plants [58] was used as the test phytopathogen. This isolate has been preserved on Whatman paper # 1 at −20 • C, and further reactivated in potato dextrose agar (PDA) to be used in the antifungal assays. The in-vitro antifungal activity against F. oxysporum was evaluated through the inhibition of mycelial growth (12-well plate amended semi-solid medium assay) and conidial susceptibility (96-well plate microdilution assay). Both assays were performed using the respective monosporic culture of the test fungal isolate, developed in PDA (potato-dextrose-agar) medium for 8 days.

Mycelial Growth Inhibition Assay
A micro-scale amended medium protocol was performed following the previously reported procedure [59]. Briefly, 12-well plates (autoclavable glass) and semi-solid medium (1.2 g of PDB (potato-dextrose broth) and 0.5 g of Agar per 100 mL of distilled water) were used. The final volume (for treatments and controls) was 200 µL per well (1-cm diameter). The test concentration (i.e., 0.2% w/v) was prepared by supplementing fresh semisolidmedium with the required amount of each test dry, raw extract (n = 44) (i.e., the residue after extraction without solvent). A stock dispersion was reached by a direct mixture between the semi-solid medium (5 mL) and the required extract (10 mg). This mixture was vigorously stirred until achieving a homogenous dispersion before solidifying. Tween-20 (5%) was used to assist the dispersion of the extracts. The resulting homogeneously amended medium was added into three wells of the 12-well plate. Subsequently, a 1-mm agar-mycelial plug from 5-days actively growing cultures of F. oxysporum was inoculated onto the center of each well. Each 12-well plate was placed into a 17-mm Petri dish, under the appropriate conditions of humidity and sterility, and sealed with plastic films. Each trial comprised a randomized design with three replicates for each extract compared to an absolute control (untreated semi-solid medium). Sportak (prochloraz) and Dithane (mancozeb) were employed as positive controls using the same concentration (i.e., 0.2% w/v). After inoculation and sealing, this assembly was incubated at 25 • C. The assay was concluded once the colony on the negative control covered the whole well (after ca. 48 h). Mean colony area (mm 2 ) was, therefore, measured for treated and untreated wells, under processing in the software Image J ® using the respective photographic records. The percent mycelial growth inhibition (MGI) was then calculated for each replicate.
This calculation was made employing this equation: % MGI = [(absolute control area − treatment area)/absolute control area)] × 100. Finally, a portion of the mycelium after being treated with botanical extracts was removed and stained with Lactophenol Blue. The stained fungal structures were visualized microscopically at 2000× using the Motic BA210 microscope (Motic, Carlsbad, CA, USA).

Conidial Susceptibility Assay
A microdilution protocol was implemented following the previously reported procedure [60]. The assays were carried out in 96-well plates by homogeneously mixing the test extracts in PDB liquid medium and DMSO (>0.2%) at 0.1% (w/v) as the final dose. Four replicates were adopted for each test extract. To prepare the non-germinated conidial suspensions, 1 mL of sterile distilled water (SDW) was added to a 9-mm Petri dish containing a 5-days actively growing culture of F. oxysporum. After that, an aliquot (100 µL) was placed onto the Neubauer chamber to perform the conidia counting. The final inoculum (1 × 10 6 conidia/mL) was adjusted using the appropriate amount of SDW. The experiment comprised the following mixtures: (1) Plant extract + conidia + PDB medium (treatment); (2) conidia + PDB medium (absolute control); (3) plant extract + SDW (blank 1); (4) PDB medium + SDW (blank 2). Treatment (180 µL) and previously prepared non-germinated conidia dispersion (20 µL) were combined to start the trial. SDW (20 µL) was added instead conidia dispersion for blanks 1 and 2. Prochloraz TM (iprodione) and Dithane FMB TM (mancozeb) were used as positive controls. The resulting optical dispersion (OD) values were taken at 0 and 8 h, using an Elisa EZ Read 2000 analyzer, recording at 600 and 680 nm. Due to the assessment of conidial susceptibility through 8-h microdilution from non-germinated conidial suspension assuming an indirect extent of conidial germination inhibition (CGI) of filamentous fungi [39], we opted to employ this indirect measurement. The calculation of CGI was carried out employing this equation: % CGI = [(OD cor absolute control − OD cor treatment)/OD cor absolute control)] × 100, where OD cor means OD values corrected with blanks 1 and 2. Finally, a portion of the conidia submerged cultures after treated with botanical extracts was removed and stained with Lactophenol Blue. The stained conidial structures were visualized microscopically at 2000× using the Motic BA210 microscope (Motic, Carlsbad, CA, USA) to verify the conidial germination and absence of hyphal structures.

Statistical Analysis
A Shapiro-Wilk test was performed to determine the data normality of the inhibition percentage data. Once normality was verified, ANOVA tests were conducted on all samples (type I error, significance value p < 0.05) followed by post-hoc Tukey tests to determine the significant differences between data. These statistical tests were performed software R version 3.4.1 (R Foundation, Vienna, Austria).
On the other hand, chromatographic profiles of the test extracts were exported to an ASCII 2D format and a matrix containing point-to-point HPLC data per extract sample was built (5428 × 44). This dataset was aligned in the Matlab ® software (Vr2013a) (The Mathworks Inc., Natick, MA, USA). The aligned data were also normalized and autoscaled.
These pre-treated chemical data (i.e., chromatographic profiles) were combined with the respective antifungal data (i.e., %MGC or %CGI) to assemble the integrated dataset. The resulting matrix was then imported into the SIMCA software (v 14.0) (Umetrics, Umeå, Sweden) to build the respective models by single-Y orthogonal partial least squares (OPLS). The obtained results were visualized by means of the scores and S-line plots.

Conclusions
Among test botanicals (n = 44), four Piper-derived extracts showed the most potent inhibitory activity on mycelial growth of F. oxysporum, whereas conidia were more susceptible to P. elongatum and T. diversifolia extracts, having moderate activity. This fact indicated a differential effect of extract components on each fungal form. Extract from leaves of P. aduncum exhibited the highest activity on mycelial growth, possibly targeting the cell wall or cell membrane of the fungus. The indirect detection of plausible bioactives through untargeted metabolic profiling, integrating the chemical and antifungal activity against F. oxysporum datasets, led to the recognition of three antifungal compounds piperaduncin C (1), asebogenin (2) and (−)-methyllinderatin (3) as the most active compounds. This finding was adequately validated after the isolation of these metabolites and evaluation of the pure isolates against F. oxysporum through both assays. This is the first report for these dihydrochalcone-type metabolites against F. oxysporum. Therefore, the usefulness of this indirect approach based on HPLC profiles for the identification of antifungals from botanical extracts was confirmed, which extends a rationale for less-time consuming antifungal recognition and lead finding from nature. Additionally, our findings constitute the basis for the development of antifungal agents based on active dihydrochalcone-containing extracts and individual compounds and, consequently, for outlining their potential as promising biorationals to protect plants against this phytopathogen.

Data Availability Statement:
The data that support the findings of this study are available from the corresponding author upon request.