Synthesis, Biological Evaluation, and QSAR Studies of 3-Iodochromone Derivatives as Potential Fungicides

Despite the emergence of novel biotechnological and biological solutions, agrochemicals continue to play an important role in crop protection. Fungicide resistance is becoming a major problem; numerous cases of fungicide resistance have occurred worldwide in the last decade, resulting in the loss of several fungicides. The discovery of new molecules has therefore assumed critical importance in crop protection. In our quest for biologically active molecules, we herein report the synthesis of a series of twenty-one 3-Iodochromone derivatives (4a–4u), in a two-step process by condensation of 2-hydroxyacetophenone derivatives (2a–2u) with N,N-dimethylformamidedimethylacetal yielding enaminones (3a–3u), followed by cyclization with iodine to corresponding 3-iodochromones. Characterization of these compounds was done by IR, 1H NMR, 13C NMR, and LC-HRMS techniques. All synthesized compounds were screened for their fungicidal activity against Sclerotium rolfsii. Among these 6,8-Dichloro-3-iodochromone 4r was found to be most active (ED50 = 8.43 mg L−1). 2D-Quantitative Structural Activity Relationship (2D-QSAR) analysis was also performed by generating three different models viz., Multiple Linear Regression (MLR, Model 1), Principal Component Regression (PCR, Model 2), and Partial Least Squares (PLS, Model 3). Predictive power and statistical significance of these models were assessed with external and internal validation and leave one-out cross-validation was used for verification. In QSAR study, MLR (Model 1) was found to be best having correlation coefficient (r2) 0.943, cross-validated correlation coefficient (q2) 0.911 and r2pred 0.837. It was observed that DeltaEpsilonC, T_2_Cl_6, T_2_F_6, T_T_F_3, and ZCompDipole are the major descriptors which influence the fungicidal activity of 3-Iodochromone derivatives. The physicochemical parameters were estimated by the VLifeMDS 4.6 software. The QSAR study results will be helpful for structure optimization to improve the activity.


INTRODUCTION
The growth of human civilization has been closely related to crop production, and plant diseases have been a concern for human being perhaps since plants were cultivated more than 10,000 years ago. As a consequence of plant diseases, world agriculture faces an estimated loss of 18% annually amounting to approximately 1,300 billion INR (Oerke, 2006). Sclerotium rolfsii Sacc. is a devastating soilborne fungus that infects more than five hundred agricultural and horticultural plant species around the world, causing root rot, stem rot, collar rot, willow, and foot rot diseases (Aycock, 1966;Punja, 1985). Crucifers, cucurbits, and legumes are its most common hosts. The fungus is of considerable economic significance because it causes 10-100 percent crop loss in different crops. Due to the formation of excessive sclerotia it may persist in soil for several years (Punja, 1985). Chemical crop protection measures continue to play an important role in agribusiness in spite of the emergence of novel biotechnological and biological solutions. Resistance to fungicides is becoming a major problem generating disease control problems in many crops. In the last decade, numerous cases of fungicide resistance have occurred worldwide, leading to loss of several fungicides (Hollomon, 2015). Therefore, the discovery of new molecules has assumed critical importance to combat the fungal infections.
Chromone is a group of naturally occurring compounds, reported, mainly in plants. The chromone moiety is a pharmacophore in a large number of natural and synthetic bioactive molecules. The chromone scaffold is present in plant's secondary metabolites: flavones and isoflavones. Chromones are reported to have anti-tumor, antiinflammatory and anti-fungal activities, and inhibitory activities, toward, phosphatases, kinases, cyclooxygenases, aromatases, acetylcholinesterases, and monoamine oxidases (Gasparova et al., 1997;Gaspar et al., 2011, Gaspar et al., 2014. Our group has been actively involved in developing new crop protection products Yadav et al., 2019). In our quest for biologically active molecules, we herein reported the synthesis of a series of iodochromones and their evaluation against S. rolfsii. A QSAR study was carried out with the objective to find the molecular properties which affect the fungicidal activity.

EXPERIMENTAL Chemicals and Instruments
Chemicals were purchased from industrial manufacturers and, unless otherwise specified, were used without any further purification. Precoated Merck-silica gel 60F 254 plates were used for thin layer chromatography (TLC); UV cabinet was used to detect developed plates. Column chromatography was performed with 100-200 mesh silica gels. Melting points were recorded by Buchi M-560 instrument and were uncorrected. The IR spectroscopy was done with PerkinElmer 2000 FT-IR spectrometer; KBr disc were used for samples preparation. The 1 H NMR and 13 C spectra were recorded on a Jeol alpha-400 and at 100.6 MHz, respectively, using TMS as an internal standard. The chemical shift values were on δ scale and the coupling constants (J) were in Hz. Signals from OH groups in 1 HNMR spectra were verified by removing them by shaking in D 2 O.
ED 50 values were estimated with the SPSS statistical package. The whole computational work was carried out by using VLifeMDS QSAR plus 4.6 software using the Lenovo PC having window 8.1 operating system and Intel (R) Celeron (R) processor.

Synthesis
Synthesis of Substituted 2-Hydroxyacetophenones (2a-2n) 2-hydroxyacetophenone and bromoalkanes or iodoalkanes of different chain lengths were taken in a molar ratio of 1:1.2 and stirred continuously for 6 h at 60°C in the presence of K 2 CO 3 and acetone. Reaction was supervised by thin layer chromatography (TLC) with ethyl acetate: hexane (3:7).

Spectral Analysis of Synthesized
It was obtained as a brown solid in 82% yield; m.            (72).

Test Fungus
The test fungus S. rolfsii ITCC 6474 was procured from Indian Type Culture Collection (ITCC) center, Division of Plant Pathology, ICAR-Indian Agricultural Research Institute, New Delhi-110012, India and kept at 27°C for at least 4-7 days on Potato Dextrose Agar (PDA) slant. The fungus was subcultured in Petri plates for further bioassay studies.

In vitro Fungicidal Activity
A stock solution (1, 000 mgL −1 ) of each synthesized compound was prepared in DMSO. Preliminary screening was carried out at different concentrations. A final bioassay was conducted at five different concentrations namely 1,000, 500, 250, 125, and 62.50 mg L −1 of 4i-4n, 4p, 4s, and 4u chromones, and all other chromones were tested at 100, 50, 25, 12.5, and 6.25 mg L −1 , respectively. All concentrations were tested in triplicates. Commercially available fungicide Mancozeb (technical) was taken as positive control. An in vitro antifungal bioassay was carried out on PDA medium by poisoned food technique (Nene and Thapliyal, 1979). Fungal growth (colony diameter) was measured and percentage inhibition was calculated by Abbott's formula (Abbott, 1925).
where C colony diameter (mm) of the control and T colony diameter (mm) of the test plate.
Corrected percentage inhibition (IC) was calculated by given formula.

IC
(I − CF) (100 − CF) × 100, where I Percentage inhibition, CF (90-C)/C × 100, 90 is the diameter (mm) of the Petri plate, and C is the growth of the fungus (mm) in control.

Quantitative Structure Activity Relationship
QSAR analysis was done by taking negative logarithm of observed ED 50 (mgL −1 ) [pED 50 −log (ED 50 )] as dependent variable and 2D descriptors ( Table 1) as independent variables. 2D Structures of compounds were drawn in Chemdraw Ultra 7.0 software and converted to 3D structures. A total of 239 2D descriptors were determined encoding different molecular structural characteristics consisting of electronic, spatial, thermodynamic, and structural descriptors, for example, element count, atomic valence connectivity index (chiV), path cluster, estate number, retention index (chi), chain path count, logP, semi-empirical path count, molecular cluster, molecular weight, topological index, and refractivity. Descriptors were calculated by geometry optimization and energy minimization carried out by the batch energy minimization method in the Merck molecular force field (MMFF) at RMS gradient (criteria for convergence) 0.01, distance dependent dielectric 1, and the number of cycles (max) 1, 000. Different Baumann alignment-independent (AI) descriptors were also calculated. All computational work was carried out with the help of VLifeMDS QSAR plus 4.6 software using the Lenovo PC with Windows 8.1 operating system and the Intel (R) Celeron (R) processor.

Training and Test Set
The entire data of 21 compounds were divided into a training set (14 compounds) and a test set (4 compounds), and 3 compounds were taken for validation with the help of the sphere exclusion method (Hudson et al., 1996). Unicolumn statistics were used to check the accuracy of selection of training and test sets, as the maximum value of the training set was greater than that of the test set and the minimum value of the training set was less than that of the test set ( Table 2).

Regression Analysis
Regression analysis was done with three model building methods, MLR, PCR, and PLS. Various 2D descriptors were taken as independent variables and pED 50 values as dependent variables by taking cross-correlation limit as 0.5; five variables in the final equation and r 2 as the term selection criteria, F-test "in" at 4 and "out" at 3.99, r 2 and F-test value. Values were fixed at 0 for variance cutoff, 10 for random iterations, and auto scaling for scaling. Developed QSAR models were assessed with the help of statistical parameters such as n total number of compounds used in regression, k total number of descriptors in a model, r 2 regression coefficient, q 2 cross-validated correlation coefficient, F F-test (Fisher test value) for statistical significance, pred_r 2 predictive squared correlation coefficients, pred_r 2 se coefficient of correlation of predicted data set, and r 2 se and q 2 se standard error (SE) of estimation.

Multiple Linear Regression Analysis
MLR defines linear relationship between a single response variable and a number of predictor variables. In the present work, pED 50 fungicidal activity was response variable and 2D descriptors were predictor variables. In this method, regression Frontiers in Chemistry | www.frontiersin.org April 2021 | Volume 9 | Article 636882 coefficients values (r 2 ) were calculated by the least squares curve fitting method. In regression analysis, conditional mean of the dependent variable (pED 50 ) Y depends on (descriptors) X (Eq. 1).

Principal Component Regression Method
In this method, the whole data were divided into principal components (PCs), smaller sets having major details of the large set. The main aim of PCR is to find out the values of a dependent variable with the help of selected PCs of independent variables. These PCs were not correlated, but had a basic linear relationship of original variables. The data were rotated into a new set of axes in such a way that first few axes showed greatest variability within the data. First PC (PC1) had maximum possible variation in the data, and each successive PC was taken perpendicular to preceding PCs and represent highest of the outstanding variance. The PC value is calculated by rotation of each point to a particular axis. A new group of axes for the data was chosen on the basis of a descending value data variance. Principal component analysis (PCA) also describes the fashion of similarity of the observations and the variables by exhibiting them as points in maps. PCR gives a mechanism for obtaining structure in datasets.

Partial Least Squares Regression Method
The partial least squares (PLS) test correlation between a set of dependent variables (Y) and a set of predictor variables (X). The main aim of PLS regression is helpful in describing the common structure by estimating the biological activity (dependent variables Y) from descriptors (X) (Huberty, 1994). PLS developed orthogonal components based on the relationship between predictors and respective outputs, while retaining highest variance of independent variables.

Validation of The QSAR Model
The QSAR model was validated with Leave-one-out (LOO) cross validation, by dividing training dataset into equal size subsets after eliminating one biological activity data (number of subsets number of data points). These subsets were used to develop the model for calculating predicted activity (value of response variable of excluded data). Since in LOO subset all the data points were serially considered as predicted, the mean value of predicted activity was similar for LOO q 2 and r 2 . After elimination of the next data point, the same procedure was repeated until all data points were removed. Thus, three statistically significant models were developed by LOO crossvalidation. (Kubyani, 1994). Eq. 2 was used for calculating q 2 .
where Y pred predicted, Y act actual, Y mean mean values of the pED 50 , and Σ (Y pred − Y act ) 2 predictive residual error sum of squares (PRESS). External validation has also been performed to verify model validity, which tests how well the equation generalizes. A training set was used to develop an adjustment model for predicting activities of test set members. The predictive performance of equations was determined by q 2 , and coefficients of predictive squared correlation (pred _r 2 ). pred_r 2 was calculated by Eq. 3. (3) where Y pred(Test) predicted activity and Y Test observed activity of test set compounds and Y Training mean activity value of the training set. Statistical significance of model was validated by the fitness plot ( Figure 1) and it was also supported by closeness of observed and predicted activity ( Table 3). The magnitude of different descriptors employed for developing QSAR models were present in contribution charts ( Figure 2).

Synthesis and Characterization
In this study, total 21 compounds (4a-4u) were synthesized out of which 10 compounds (4f, 4g, 4h, 4i, 4j, 4k, 4l, 4m, 4n, and 4o) were novel. The compounds synthesized by the above method were obtained in the yield ranging from 67 to 89%.  190.31-190.90 for C O were conspicuous for all the compounds. The higher chemical shifts values of H-3 and C-3 than H-2 and C-2 were due to carbonyl moiety, which polarizes the C C double bond. In IR spectra, stretching of (C O) at 1,628-1,647 and (C C) at 1,539-1,593 cm −1 supported the NMR data.

Model-3 (PLS)
pEd 50 13.5036 DeltaEpsilonC + 130.1390 ZcompDipole − 0.1607 (T T F 3) − 2.8393 (6) where n 14, DF 11, r 2 0.8006, q 2 0.6167, F_test 22.0866, r 2 _se 0.2191, q 2 _se 0.3038, pred_r 2 0.6186, and pred_r 2 se 0.2649. In above QSAR models, correlation coefficient (r 2 ) was used to calculate biological activity variance by multiplying with 100. The predictive ability (q 2 ) of generated QSAR models was assessed by LOO (Left-out-one) method. F is the ratio of variance of models and that of error in regression. Models with a higher F value and lower SE of estimation (r 2 se and q 2 se) were considered statistically significant. External validation with pred_r 2 > 0.3, established the predictive power of the QSAR model. Among these three models, the MLR model was found best as revealed by q 2 , r 2 , higher values of F-test, and pred_r 2 . The high q 2 value is the best indicator of 2D QSAR's reliability since only a high r 2 could be due to data overfitting. Quite often, a q 2 value of more than 0.5 is considered appropriate. (Golbraikh and Tropsha, 2002;Doweyko, 2004;Ponce et al., 2004).
The developed models showed that fungicidal activity was inversely related to descriptors, DeltaEpsilonC and AI descriptor, T_2_Cl_6, T_2_F_6, and T_T_F_3 and directly related to ZcompDipole. Two descriptors viz. DeltaEpsilonC and T_2_Cl_6 significantly (∼70%) impact the fungicidal activity of test compounds. Alignment Independent (AI) descriptors were estimated, as explained in Baumann's paper (Balaban, 1982), on the basis of molecular topology, type of bond, and atom. Every atom was given a minimum of one and a maximum of three attributes. Molecular topology (T) was designated as the first attribute, followed by atom symbol and atoms linked with multiple (double or triple) bonds as second and third attribute, respectively. Then, selective distance count statistics, which counts all the fragments between the first atom and the last atom isolated by a graph distance, for all combinations of various attributes were calculated. Graph distance is the least number of atoms across the path joining two atoms in molecular structure. For example, selective distance count statistic "AB2" (e.g., TOPO2N3) counts all the fragments between a start atom with attribute "A" (e.g., "2" a double bonded atom) and an end atom with attribute "B" (e.g., "N") separated by a graph distance 3. Topological indices are numerical values associated with chemical constitutions which establish correlation between biological activity and chemical structure. AI descriptors in this study were calculated with the help of attributes namely, 2 (atom with double bond), 3 (atom with double bond), C (Carbon), N (Nitrogen), O (Oxygen), S (Sulfur), H (Hydrogen), F (Fluorine), Cl (Chlorine), and Br (Bromine) with distance ranging from 0 to 7. DeltaEpsilonC is a measure of contribution of electronegativity. The result revealed that it is negatively correlated with fungicidal activity of the test compounds.

CONCLUSION
The study revealed that all test compounds showed fungicidal activity against S. rolfsii., but compound 4r showed the highest activity. The QSAR study determined quantitative correlation between fungicidal activity and structural/physicochemical properties of test compounds. The variables in developed model equations established that structural, molecular shape analysis, electronic, and thermodynamic descriptors played a major role in fungicidal activity of the compounds. In the case of MLR and PLS, the overall prediction was found to be around 94 and 80%, respectively. The 2D-QSAR study revealed that results of MLR analysis exhibited significant predictive power and reliability than the other two methods (PCR and PLS). Information and understanding of descriptors influencing fungicidal activity of these chromones could be used for structure optimization to improve activity.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material; further inquiries can be directed to the corresponding author.

AUTHOR CONTRIBUTIONS
PK and NS conceptualized the idea; VR did HRMS interpretation.

ACKNOWLEDGMENTS
The author (PK) is thankful to ICAR-IARI, New Delhi, India for providing all the facilities for research work.