**Adenosine Receptor Ligands: Coumarin–Chalcone Hybrids as Modulating Agents on the Activity of** *h***ARs**

#### **Saleta Vazquez-Rodriguez 1,\* ,**† **, Santiago Vilar <sup>1</sup> , Sonja Kachler <sup>2</sup> , Karl-Norbert Klotz <sup>2</sup> , Eugenio Uriarte 1,3, Fernanda Borges <sup>4</sup> and Maria João Matos 1,4,\***


Academic Editor: Pascal Richomme Received: 27 August 2020; Accepted: 18 September 2020; Published: 19 September 2020

**Abstract:** Adenosine receptors (ARs) play an important role in neurological and psychiatric disorders such as Alzheimer's disease, Parkinson's disease, epilepsy and schizophrenia. The different subtypes of ARs and the knowledge on their densities and status are important for understanding the mechanisms underlying the pathogenesis of diseases and for developing new therapeutics. Looking for new scaffolds for selective AR ligands, coumarin–chalcone hybrids were synthesized (compounds **1**–**8**) and screened in radioligand binding (*h*A1, *h*A2A and *h*A3) and adenylyl cyclase (*h*A2B) assays in order to evaluate their affinity for the four human AR subtypes (*h*ARs). Coumarin–chalcone hybrid has been established as a new scaffold suitable for the development of potent and selective ligands for *h*A<sup>1</sup> or *h*A<sup>3</sup> subtypes. In general, hydroxy-substituted hybrids showed some affinity for the *h*A1, while the methoxy counterparts were selective for the *h*A3. The most potent *h*A<sup>1</sup> ligand was compound **7** (*K*<sup>i</sup> = 17.7 µM), whereas compound **4** was the most potent ligand for *h*A<sup>3</sup> (*K*<sup>i</sup> = 2.49 µM). In addition, docking studies with *h*A<sup>1</sup> and *h*A<sup>3</sup> homology models were established to analyze the structure–function relationships. Results showed that the different residues located on the protein binding pocket could play an important role in ligand selectivity.

**Keywords:** coumarin; chalcone; neurodegenerative diseases; adenosine receptors; binding affinity; docking

#### **1. Introduction**

Adenosine receptors (ARs) are cell membrane receptors, belonging to the G protein-coupled receptor (GPCRs) superfamily. ARs comprised of four different subtypes: A1, A2A, A2B and A<sup>3</sup> [1]. Adenosine is a purine nucleoside and an endogenous modulator of several physiological processes [1–4]. Extracellular adenosine activates the G<sup>i</sup> -coupled receptors of the A<sup>1</sup> and A<sup>3</sup> subtypes, depressing the action of the brain, heart, kidneys, and the immune system, amongst other systems, as a consequence of the inhibition of adenylyl cyclase [5]. The A<sup>3</sup> subtype of AR has been cloned [6,7], making it possible to establish its pharmacological [8–11] and regulatory features [12].

Due to their widespread presence in cells, ARs proved to be promising targets in drug discovery. During the last decade, the search for selective ligands has been raised [13–15]. Several AR antagonists appeared as promising drug candidates for different pathological processes such as inflammation (A3) [14], heart and renal failure (A1) [16], or neurological disorders including Parkinson [17,18] and Alzheimer's diseases (A2A and/or A1) [19]. ARs can work as targets for various diseases and can open a new window for new therapeutic approaches.

In particular, A<sup>1</sup> antagonists are effective as diuretic agents [20,21] and also show neuroprotective activity in animal models of in vivo ischemia [22]. On the other hand, A<sup>3</sup> antagonists are being investigated as potential agents against renal injury [23] and also as neuroprotective agents [24,25], while A<sup>3</sup> agonists are also under consideration for treating conditions of the central nervous system (CNS) and peripheral nervous system [26,27].

From the arsenal of molecules presenting high potency and selectivity on ARs, the xanthine scaffold was the first to be used to develop the so-called classical AR antagonists [28,29]. In the search for non-xanthine AR ligands, numerous structures were discovered over the years. Flavones and isoflavones have played a remarkable role. As an example, genistein, was described as a competitive antagonist at A<sup>1</sup> in FRTL (thyroid) cells [30], and galangin was found to bind to the three subtypes of ARs displaying micromolar affinity for the A<sup>3</sup> [31]. The affinity of flavonoids and other phytochemicals to ARs brings about the hypothesis that probably other types of natural substances, namely those present in the diet, can interact with this type of receptor.

Coumarins (chromone isosteres) and chalcones (a flavonoid precursor) are naturally occurring benzopyran-related molecules presenting a variety of pharmacological activities [32–34]. Having in mind that both the coumarin and chalcone nuclei are structurally close to flavonoids, the design of novel AR ligands based on their scaffolds emerged as an interesting idea. Our study was also motivated by the structural similarity between the coumarin and the chromone scaffolds, which were previously described as AR ligands [35,36], and by the similarities with some coumarin derivatives previously described in our group [37–42]. In this context, we focused our attention on the 3-benzoylcoumarin core, considered as a hybrid scaffold in which the chalcone is fixed in a *trans* conformation through the double bond of the pyrone ring of the coumarin skeleton (Figure 1), presenting a more restricted conformation compared to the previously described coumarin–chalcone hybrids [36].

**Figure 1.** Rational design of coumarin–chalcone hybrids.

‐ ‐ ‐ ‐

Therefore, based on the structural similarities between flavones, chalcones and coumarins, we decided to design and synthesize a novel family of coumarin–chalcone hybrid derivatives and study their activity on the different subtypes of human AR.

#### **2. Results and Discussion**

#### *2.1. Chemistry*

Two sets of coumarin–chalcone hybrids have been synthesized: one decorated with methoxy substituents (**1**–**4**) and another with hydroxy substituents (**5**–**8**). An efficient and versatile Knoevenagel reaction, treating a commercially available salicylaldehyde and the corresponding methoxylated ethyl benzoylacetate with piperidine in ethanol (EtOH) at reflux for 2–6 h, allowed the desired methoxy-3-benzoylcoumarins **1**–**4** with 85–97% yield. The hydroxy-3-benzoylcoumarins **5**–**8** were obtained by hydrolysis of the corresponding methoxy derivatives, with 75–94% yield, by employing boron tribromide (BBr3) as deprotecting reagent in dichloromethane (DCM) at 80 ◦C in a Schlenk tube for 48 h [43]. The synthetic approach is illustrated in Scheme 1 and described in the methods and materials section.

‐ **Scheme 1.** Synthetic route to obtain the coumarin-chalcone hybrids. *Reagents and conditions:* (**a**) piperidine, EtOH, reflux, 2–6 h; (**b**) BBr<sup>3</sup> , DCM, 80 ◦C, 48 h.

#### *2.2. Pharmacology*

#### Adenosine Receptor Binding Affinity Assays

The adenosine binding affinity of derivatives **1**–**8** for the human AR subtypes *h*A1, *h*A2A and *h*A3, expressed in Chinese Hamster Ovary (CHO) cells, was determined in radioligand

‐ ‐ ‐ ′‐

‐

‐

‐

competition experiments [43,44]. In the binding affinity assay, it is measured the competition of ligands for specific binding of [3H]CCPA (2-chloro-*N*<sup>6</sup> -cyclopentyladenosine) to *h*A1; specific binding of [3H]NECA (5′ -*N*-ethylcarboxamidoadenosine) to *h*A2A; and specific binding of [ <sup>3</sup>H]HEMADO (2-(1-hexynyl)-*N*<sup>6</sup> -methyladenosine) to *h*A3. The results are expressed as *K*<sup>i</sup> (dissociation constants), which were calculated with the program SCTFIT, and given as geometric means of at least three experiments, including 95% confidence intervals. Due to the lack of a suitable radioligand for the *h*A2B receptor, the potency of antagonists at the *h*A2B receptor was determined by inhibition of NECA-stimulated adenylyl cyclase activity with increasing concentrations of antagonist [43,44]. As a result, cAMP (cyclic adenosine monophosphate) production was inhibited in a concentration-dependent fashion, and K<sup>i</sup> values were calculated from the measured IC<sup>50</sup> values [45].

Derivatives **1**–**8** were efficiently synthesized and their in vitro binding affinity for human AR subtypes *h*A1, *h*A2A, *h*A2B and *h*A3, expressed in CHO cells, was evaluated. In the present communication, the studies were focused on the inspection of the effect on the binding affinity of different number and position of methoxy or hydroxy substituents on the 3-benzoylcoumarin scaffold. Data obtained for the binding affinity for *h*A<sup>1</sup> and *h*A<sup>3</sup> is summarized in Table 1. For all the tested compounds, no significant affinity was detected for the *h*A2A (*K*i > 100 µM) or *h*A2B (*K*<sup>i</sup> > 10 µM).


**Table 1.** Binding affinity (*K<sup>i</sup>* ) of compounds **1**–**8** on *h*A<sup>1</sup> and *h*A<sup>3</sup> AR.

<sup>a</sup> Results are geometric means of 3 experiments and given with 95% confidence intervals (in parentheses). <sup>b</sup> At higher concentrations, the compounds precipitate.

The binding affinity results show that derivatives **1** and **2**, without substitutions on the coumarin scaffold or with a single methoxy group at the position 6 of the coumarin core, respectively, display no detectable binding affinity for the evaluated receptors (*K*<sup>i</sup> > 100 µM). However, the presence of two methoxy groups at positions 5 and 7 (compounds **3** and **4**, respectively) lead to an increment on both the potency and selectivity for the *h*A3. Compound **3**, presenting three methoxy groups at positions 5, 7 and 4′ proved to be *h*A<sup>3</sup> selective, displaying a *K*<sup>i</sup> = 9.03 µM, whereas compound **4**, presenting an extra methoxy groups at position 3′ is not only selective for *h*A3, but also displays a increase in potency (*K*<sup>i</sup> = 2.49 µM). Compared to theophylline, classically used as a reference compound, we would like to highlight that both compounds **3** and **4** are more potent and *h*A<sup>3</sup> selective molecules.

Based on this data, it can be concluded that both nature and position of the substitution patterns on the coumarin–chalcone scaffold play a key role in the interaction with the *h*A3. It can be highlighted that positions 5 and 7 of the studied scaffold seem to be relevant for the observed selectivity and potency. Analyzing the methoxylated derivatives **1**–**4**, only the molecules presenting substituents at these two positions (compounds **3** and **4**) are *h*A<sup>3</sup> active and selective ligands.

Interestingly, a similar tendency was observed for *h*A<sup>1</sup> binding of the hydroxylated derivatives (**5**–**8**), which bear hydroxy groups instead of methoxy groups at positions 5 and 7 (compounds **7** and **8**). Derivatives **7** and **8** display the highest potency and selectivity of the studied series towards *h*A1, but their activity towards this receptor is still low with *K*<sup>i</sup> = 17.7 µM and *K*<sup>i</sup> = 29.1 µM, respectively.

#### *2.3. Theoretical Evaluation of ADME Properties*

In order to explore the drug-like properties of compounds **1**–**8**, the lipophilicity, expressed as the octanol/water partition coefficient and herein named clogP, as well as other theoretical calculations such as number of hydrogen acceptors and number of hydrogen bond donors, and topological polar surface area (TPSA), were calculated using the Molinspiration software [46]. Theoretical prediction of absorption, distribution, metabolism and excretion (ADME) properties of all derivatives is summarized in Table 2.


**Table 2.** Theoretical evaluation of the ADME properties of coumarin–chalcone hybrids.a.

<sup>a</sup> TPSA, topological polar surface area; n-OH, number of hydrogen acceptors; n-OHNH, number of hydrogen bond donors.

Based on this theoretical data, it can be concluded that the study molecules **1**–**8** do not violate any of Lipinski's rules (namely molecular weight, clogP, number of hydrogen donors and acceptors). In addition, TPSA, described as an indicator of membrane permeability, was favorable for the studied compounds.

#### *2.4. Molecular Modeling*

*h*A<sup>1</sup> and *h*A<sup>3</sup> homology models were successfully constructed (Materials and methods section). A selection of models obtained from Induce Fit calculations were tested based on their ability to discriminate between known ligands, decoys and between subtype-selective compounds. The models selected for the docking calculations showed excellent results in both tests. A dataset of 200 randomly selected decoys from the ZINC database [47] were mixed up with 22 known ligands of each adenosine receptor subtype [48] Glide SP precision was used to dock the database to the *h*A<sup>1</sup> and *h*A<sup>3</sup> models [49]. Table 3 presents the area under the receiver operating characteristic (ROC) curve (AUROC) for both systems. To differentiate between subtype-selective ligands, a second and more challenging test was performed. As in a previous study [48], 66 subtype-selective molecules (22 *h*A1, 22 *h*A2A and 22 *h*A<sup>3</sup> compounds) were docked to the *h*A<sup>1</sup> model (22 true positives vs. 44 false positives) and to the *h*A<sup>3</sup> (22 true positives vs. 44 false positives). Results corroborate those previously published by Katritch et al. [50] and proved that the developed homology models are able to discriminate between subtype-selective compounds (Table 3).

**Table 3.** Area under the ROC curve (AUROC) for the two homology models.


<sup>a</sup> 22 *h*A<sup>1</sup> or 22 *h*A<sup>3</sup> ligands as true positives (TP) and 200 random decoys as false positives (FP) were considered. <sup>b</sup> For *h*A1, 22 *h*A<sup>1</sup> selective compounds as TP and 22 *h*A2A + 22 *h*A<sup>3</sup> compounds as FP were considered. For *h*A3, 22 *h*A<sup>3</sup> compounds as TP and 22 *h*A2A + 22 *h*A<sup>1</sup> compounds as FP were considered.

Glide SP molecular docking simulations were run with our data using the *h*A<sup>1</sup> and *h*A<sup>3</sup> selected homology models as protein structures to detect the hypothetical binding mode of the new synthesized compounds [51]. The Prime module was used to optimize the protein structure for each binding mode [52]. Molecular docking simulations are represented in Figure 2.

‐ **Figure 2.** (**a**) Comparative study of the co-crystallized ligands (green carbons) in the *h*A2A [3EML (left) and 3UZC (right)] with the pose of compound **3** extracted from the *h*A<sup>3</sup> docking calculations (grey carbons). Binding pockets in *h*A2A and *h*A<sup>3</sup> were superposed. (**b**) Pose extracted for compound **3** inside the *h*A<sup>3</sup> after docking. Hydrogen bonds are represented in yellow color. (**c**) Hypothetical binding mode for compound **5** (pink carbons) in the *h*A<sup>3</sup> . (**d**) Pose obtained through docking simulations for compound **7** (green carbons) in the *h*A<sup>1</sup> protein pocket.

Docking calculations and the established homology models for the *h*A<sup>1</sup> and *h*A<sup>3</sup> identified the hypothetical binding mode and rationalized the interaction of these derivatives with their respective ARs binding sites.

The calculations showed a high level of variability since all the synthetized derivatives yielded different possible binding modes inside the pockets. Selection of the hypothetical binding pose was accomplished considering the number of similar poses extracted from the simulations and geometrical correspondence to crystallized ligands in the *h*A2A (Figure 2a).

‐ Docking results disclosed important data about the binding mode: the oxygens presented in the benzopyrone system are oriented towards the Asn250 residue and the benzoyl moiety was buried in the *h*A<sup>3</sup> pocket. This hypothetical binding mode corroborates the conformations shown by the co-crystallized ligands in the *h*A2A (PDB: 3EML and 3UZC) [48,53] (Figure 2a,b). The pose of compound **3** produced effective hydrogen bonds with Gln167, Asn250 and His272 residues.

 μ ′ ′ Interestingly, when methoxy substituents were demethylated and changed into hydroxy equivalents (compounds **5**–**8**) a modification in the profile of the studied derivatives was noticed: a loss of affinity for *h*A<sup>3</sup> and a tendency for interaction with *h*A1. The only compound that discloses some affinity for both receptors was compound **5** (*h*A<sup>1</sup> *K*<sup>i</sup> = 39.5 and *h*A<sup>3</sup> *K*<sup>i</sup> = 34.5 µM), which presents a catechol at positions 3′ and 4′ and no substitutions in the coumarin fragment. The hypothetical binding mode for compound **5** in the *h*A<sup>3</sup> pocket is represented in Figure 2c. The compound can establish hydrogen bonds with Ala69, Asn250 and His272 residues. As observed in the *h*A2A crystallized

structure and previously published studies [54,55], the corresponding Asn250 residue seems to play an important role in ligand recognition. The compound **5** pose inside the *h*A<sup>1</sup> pocket is likewise the described pose in the *h*A<sup>3</sup> one. However, the position was slightly shifted, and calculations were not able to retrieve a hydrogen bond with the Asn250 residue. The introduction of an additional hydroxy group at position 6 of the coumarin scaffold (compound **6**), resulted in a loss of measurable *h*A<sup>3</sup> binding affinity. The most noticeable binding affinities were found for derivatives with hydroxy substitutions at positions 5 and 7 of the coumarin core, as stated for methoxy equivalents. Thereby, compound **7**, with the same substitution pattern as quercetin (Figure 1), that is, hydroxy groups at positions 5, 7, 3′ and 4′ , displays *h*A<sup>1</sup> selectivity, and the best binding affinity (*K*<sup>i</sup> = 17.7 µM). Compound **8**, with the same substitution pattern as genistein (Figure 1, hydroxy substituents at positions 5, 7 and 4′ ) shows a similar *h*A<sup>1</sup> selectivity (*K*<sup>i</sup> = 29.1 µM). The pose obtained through docking calculations for compound **7** in the *h*A<sup>1</sup> protein pocket showed the possibility of establishment of hydrogen bonds with Glu172, Asn254 and Thr277 residues (Figure 2d). ′ ′ μ ′ μ

Moreover, we calculated the interaction energy contributions of the residues in *h*A<sup>3</sup> and *h*A<sup>1</sup> pockets with compounds **3** and **7**, respectively (Figure 3). The sum of different individual contributions, such as Coulomb, *van der Waals* and hydrogen bond energies, was taken into account in the calculation of the interaction energies for each residue.

**Figure 3.** Interaction energy contribution (sum of Coulomb, *van der Waals* and hydrogen bond energies) between the residues in the (**a**) *h*A<sup>3</sup> and (**b**) *h*A<sup>1</sup> and the respective derivatives **3** and **7** (residues in a distance of 3 Å from the ligand).

In addition, Figure 4 shows the molecular surface around the two residues in the *h*A<sup>1</sup> and *h*A<sup>3</sup> that could be responsible for the observed selectivity.

Regarding the interaction energy contributions (Figure 4), calculations showed that the molecular surface around the two residues in the *h*A<sup>1</sup> and *h*A<sup>3</sup> could be responsible for the observed selectivity. Phe168, Asn250, Ile268 and His272 are important residues in the interaction between compound **3** and the *h*A3. Residues with important contributions in the stabilization of compound **7** inside the *h*A<sup>1</sup> are Phe171, Gln172, Asn254, Ile274 and Thr277.

surface showing favored interaction areas generated in ‐ **Figure 4.** Molecular surface showing favored interaction areas generated in the (**a**) *h*A<sup>1</sup> and (**b**) *h*A<sup>3</sup> . Red color represents hydrogen-bond areas, green color shows hydrophobic areas, and blue represents mildly polar interfaces. Protein structures are viewed from the extracellular side.

There are different residues in both *h*A<sup>1</sup> and *h*A<sup>3</sup> with different hydrophobic/hydrophilic characteristics, which may be important to understand the observed selectivity. Hydrophobic residues in the *h*A3, such as Val169 and Leu264, could establish hydrophobic interactions and contribute towards stabilizing the ligand when the derivatives present hydrophobic substituents, like methoxy groups (i.e., **3** and **4**) (Figure 4). However, in the case of *h*A1, the corresponding residues are Glu172 and Thr270. They have hydrophilic characteristics and so can stabilize the binding of derivatives with polar substituents, such as the hybrids with hydroxy groups (compounds **6**–**8**). Yet, compound **5,** with no substituents in the coumarin ring, can be stabilized in the pocket of both proteins.

#### **3. Materials and Methods**

#### *3.1. Chemistry*

#### 3.1.1. General Methods

‐ ‐ ‐ δ Starting materials and reagents were obtained from commercial suppliers and were used without purification. Melting points (mp) were determined using a Reichert Kofler thermopan or in capillary tubes on a Büchi 510 (Flawil, Switzerland) apparatus and were uncorrected. <sup>1</sup>H-NMR (300 MHz) and <sup>13</sup>C-NMR (75.4 MHz) spectra were recorded with a Bruker AMX spectrometer (Bruker Daltonics Inc., Fremont, CA, USA) using DMSO-*d<sup>6</sup>* or CDCl<sup>3</sup> as solvent. Chemical shifts (δ) are expressed in parts per million (ppm) using TMS as an internal standard. Coupling constants *J* are expressed in hertz (Hz). Spin multiplicities are given as s (singlet), bs (broad singlet), d (doublet), dd (doublet of doublets) and m (multiplet). Mass spectrometry was carried out with a Kratos MS-50 or a Varian MAT-711 spectrometer (Thermo Fisher Scientific, Waltham, MA, USA). Elemental analyses were performed by a Perkin–Elmer 240B microanalyzer (Thermo Fisher Scientific, Waltham, MA, USA) and were within ±0.4% of the calculated values in all cases. The analytical results were ≥95% purity for all compounds. Flash Chromatography (FC) was performed on silica gel (Merck 60, 230–400 mesh,

‐

Kenilworth, NJ, USA) and analytical TLC on precoated silica gel plates (Merck 60 F254, Kenilworth, NJ, USA). Organic solutions were dried over anhydrous sodium sulfate. Concentration and evaporation of the solvent after reaction or extraction was carried out on a Büchi rotavapor (BÜCHI Labortechnik AG, Switzerland) operating at reduced pressure. The purity of compounds was assessed by high performance liquid chromatography (HPLC) coupled at diode array detector (DAD) on a Thermo Quest Spectrasystem (Thermo Fisher Scientific, Waltham, MA, USA) equipped with a P4000 pump, an UV6000 UV-Vis diode array detector, and a SN4000 interface to be operated via a personal computer. The instrument software ChromQuest 5.0 (Thermo Fisher Scientific, Waltham, MA, USA) was used for data acquisition. Different analytical columns and mobile phases (all solvents were HPLC grade) were tested. The mobile phase was H2O:CH3CN = 70:30 and an Eclipse xdb C18 column (5 µm particle size, 0.46 mm i.d., 25 cm length; Agilent Technologies, CA, USA) was used. The purity of the compounds was found to be higher than 95%.

#### 3.1.2. Synthetic Protocol to Obtain the Methoxy-3-benzoylcoumarins **1**–**4**

To a solution of the appropriate β-ketoester (1 mmol) and the corresponding salicylaldehyde (1 mmol) in ethanol (5 mL) piperidine in catalytic amount (0.10 mL) was added. The reaction mixture was refluxed for 2–6 h and, after completion (followed by TLC), the reaction was cooled, and the precipitate was filtered and washed with cold ethanol and ether. The obtained solid was recrystallized in DCM to afford the corresponding methoxy-3-benzoylcoumarin compounds.

#### 3.1.3. Synthetic Protocol to Obtain the Hydroxy-3-benzoylcoumarins **5**–**8**

In a Schlenk tube, the appropriate methoxy derivative compound **1**–**4** (1 mmol) was dissolved in DCM (1 mL), and BBr<sup>3</sup> (20 mmol, 1M) was added dropwise. The tube was sealed, and the reaction mixture was heated at 80 ◦C for 48 h. The resulting crude product was treated with MeOH and rotated to dryness. The obtained crude solid was recrystallized in MeOH or purified by flash chromatography using hexane/ethyl acetate mixtures as eluent, to afford the desired hydroxy derivatives.

3-(3′ ,4′ -Dimethoxybenzoyl)coumarin (**1**): 85% yield; white solid; mp 190–191 ◦C; <sup>1</sup>H-NMR (300 MHz, CDCl3) δ ppm 8.01 (s, 1H, H-4), 7.71–7.52 (m, 3H, 3x Ar-H), 7.50–7.29 (m, 3H, 3x Ar-H), 6.87 (d, *J* = 8.4 Hz, 1H, H-5′ ), 3.95 (s, 6H, 2x OCH3); <sup>13</sup>C-NMR (75 MHz, CDCl3) δ ppm 190.3, 154.8, 154.4, 149.5, 144.6, 133.6, 129.3, 129.2, 127.8, 125.7, 125.2, 118.5, 117.2, 111.2, 110.2, 56.4, 56.3; EI-MS *m*/*z* (%): 311 ([M + 1]+, 59), 310 (M+, 100), 173 (41), 166 (25), 165 (99), 79 (22), 77 (22); Anal. Calcd. For C18H14O5: C 69.67, H 4.55. Found: C 69.69, H 4.58.

6-Methoxy-3-(3′ ,4′ -dimethoxybenzoyl)coumarin (**2**): 97% yield; beige solid; mp 202–203 ◦C; <sup>1</sup>H-NMR (300 MHz, CDCl3) δ ppm 7.78 (s, 1H, H-4), 7.38 (d, *J* = 1.9 Hz, 1H, H-2′ ), 7.26 (dd, *J* = 8.4, 2.0 Hz, 1H, H-6′ ), 7.16 (d, *J* = 9.1 Hz, 1H, H-8), 7.04 (dd, *J* = 9.1, 2.9 Hz, 1H, H-7), 6.82 (d, *J* = 2.9 Hz, 1H, H-5), 6.70 (d, *J* = 8.4 Hz, 1H, H-5′ ), 3.78 (s, 6H, 2x OCH3), 3.69 (s, 3H, OCH3); <sup>13</sup>C-NMR (75 MHz, CDCl3) δ ppm 190.4, 156.6, 154.4, 149.5, 149.3, 144.4, 129.4, 128.0, 125.7, 121.6, 118.8, 118.2, 111.2, 110.8, 110.2, 56.4, 56.3, 56.1; EI-MS *m*/*z* (%): 341 ([M + 1]+, 58), 340 ([M]+, 94), 165 (100), 77 (22); Anal. Calcd. For C19H16O6: C 67.05, H 4.74. Found: C 67.09, H 4.75.

5,7-Dimethoxy-3-(3′ ,4′ -dimethoxybenzoyl)coumarin (**3**): 91% yield; pale yellow solid; mp 210–211 ◦C; <sup>1</sup>H-NMR (300 MHz, CDCl3) δ ppm 8.19 (s, 1H, H-4), 7.34 (d, *J* = 1.9 Hz, 1H, H-2′ ), 7.26 (dd, *J* = 8.4, 2.0 Hz, 1H, H-6′ ), 6.70 (d, *J* = 8.4 Hz, 1H, H-5′ ), 6.29 (d, *J* = 2.0 Hz, 1H, H-6), 6.14 (d, *J* = 2.0 Hz, 1H, H-8), 3.77 (bs, 6H, 2x OCH3), 3.72 (bs, 6H, 2x OCH3); <sup>13</sup>C-NMR (75 MHz, CDCl3) δ ppm 190.9, 165.8, 159.4, 158.4, 158.0, 153.9, 149.3, 141.5, 130.0, 125.4, 121.3, 111.5, 110.1, 103.9, 95.4, 93.0, 56.3; EI-MS *m*/*z* (%): 371 ([M + 1]+, 24), 370 (M+, 100), 339 (21), 233 (30), 165 (63); Anal. Calcd. For C20H18O7: C 64.86, H 4.90. Found: C 64.88, H 4.93.

5,7-Dimethoxy-3-(4′ -methoxybenzoyl)coumarin (**4**): 97% yield; pale yellow solid; mp 174–175 ◦C; <sup>1</sup>H-NMR (300 MHz, CDCl3) δ ppm 8.21 (s, 1H, H-4), 7.69 (d, *J* = 8.8 Hz, 2H, H-2′ , H-6′ ), 6.77 (d, *J* = 8.8 Hz, 2H, H-3′ , H-5′ ), 6.29 (d, *J* = 2.2 Hz, 1H, H-6), 6.13 (d, *J* = 2.2 Hz, 1H, H-8), 3.72 (2s, 3H + 3H, 2x OCH3), 3.70 (s, 3H, OCH3); <sup>13</sup>C-NMR (75 MHz, CDCl3) δ ppm 190.7, 165.6, 163.8, 159.1, 158.2, 157.8, 141.5, 132.1, 129.8, 121.2, 113.7, 103.8, 95.2, 92.8, 56.1, 56.0, 55.5; EI-MS *m*/*z* (%): 341 ([M + 1]+, 33), 340 (M+, 88), 325 (28) 312 (30), 309 (45), 297 (20), 233 (48), 135 (100), 92 (27), 77 (38). Anal. Calcd. For C19H16O6: C 67.05, H 4.74. Found: C 67.08, H 4.76.

5,7-Dihydroxy-3-(4′ -hydroxybenzoyl)coumarin (**8**): 88% yield; pale green solid; mp 290–292 ◦C; <sup>1</sup>H-NMR (300 MHz, DMSO-*d*6) δ ppm 11.10 (s, 1H), 10.85 (s, 1H), 10.53 (s, 1H), 8.09 (d, *J* = 1.4 Hz, 1H), 7.67 (d, *J* = 8.6 Hz, 2H), 6.80 (d, *J* = 8.7 Hz, 2H), 6.25 (d, *J* = 2.0 Hz, 1H), 6.22 (d, *J* = 1.8 Hz, 1H); <sup>13</sup>C-NMR (75 MHz, DMSO-*d*6) δ ppm 190.4, 164.2, 162.4, 158.8, 157.5, 157.2, 141.1, 132.3, 128.3, 119.0, 115.3, 101.5, 98.5, 94.3. EI-MS *m*/*z* (%): 299 ([M + 1]+, 9), 298 (M+, 31), 283 (16), 218 (20), 121 (100), 93 (26), 65 (27). Anal. Calcd. For C16H10O6: C 64.43, H 3.38. Found: C 64.39, H 3.37.

#### *3.2. Biological Assays*

#### 3.2.1. Binding Affinity Assays

The binding affinity for *h*A1, *h*A2A, *h*A<sup>3</sup> of the synthetized compounds was evaluated using radioligand competition experiments in CHO cells that were stably transfected with the individual receptor subtypes [44,45]. The radioligands used were 1 nM [3H]CCPA for *h*A<sup>1</sup> (*K*<sup>D</sup> = 0.61 nM), 10 nM [3H]NECA for *h*A2A (*K*<sup>D</sup> = 20.1 nM), and 1 nM [3H]HEMADO for *h*A<sup>3</sup> (*K*<sup>D</sup> = 1.2 nM) receptors. Due to the lack of a suitable radioligand for the *h*A2B receptor, the potency of antagonists at the *h*A2B receptor (expressed on CHO cells) was determined by inhibition of NECA-stimulated adenylyl cyclase activity [44,45]. The IC<sup>50</sup> for inhibition of cAMP (cyclic adenosine monophosphate) production was determined and converted to K<sup>i</sup> values using the Cheng and Prusoff equation [56]. For all the tested compounds, no measurable activity for the *h*A2B (K<sup>i</sup> > 10 µM) was detected.

#### 3.2.2. Statistical Methods

*K*<sup>i</sup> values (dissociation constants) were determined in radioligand competition experiments with 7–8 different concentrations of test compound and each concentration was tested in duplicate. *K*<sup>i</sup> values are given as geometric means of three independent experiments with 95% confidence intervals. The program Prism 6 (GraphPad Software) was used for the analysis of the competition curves.

### *3.3. Theoretical Evaluation of ADME Properties*

cLogP was calculated by the methodology developed by Molinspiration as a sum of fragment-based contributions and correction factors. Topological Polar Surface Area (TPSA) was calculated based on the methodology published by Ertl et al. as a sum of fragment contributions [57]. Oxygen- and nitrogen-centered polar fragments are considered. TPSA has been shown to be a very good descriptor characterizing drug absorption, including intestinal absorption, bioavailability, Caco-2 permeability and blood–brain barrier penetration. The method for calculation of molecule volume developed at Molinspiration is based on group contributions. These have been obtained by fitting the sum of fragment contributions to "real" 3D volume for a training set of about twelve thousand, mostly drug-like molecules. Three-dimensional molecular geometries for a training set were fully optimized by the semiempirical AM1 method.

### *3.4. Molecular Modeling*

Homology modeling was carried out using the Molecular Operating Environment (MOE) suite [49]. Molecular docking simulations were performed with the Schrodinger package [51,52].

### 3.4.1. Homology Models of *h*A<sup>1</sup> and *h*A<sup>3</sup>

Homology models of the *h*A1 and *h*A<sup>3</sup> were constructed. The crystallized structure of the *h*A2A receptor (PDB: 3EML) was used as a template [48]. Protein sequence alignment of the 3 receptors (*h*A1, *h*A2A and *h*A3) used to generate the homology models was performed as previously described by Katritch et al. [50]. The alignment was made considering the highly conserved residues in the different TM helices. MOE software was used to generate the homology models [49]. Protein geometry was evaluated for the models taking into account Phi–Psi plots, rotamers, bond angles, bond lengths, atom clashes, dihedrals and contact energies. The presence of different conserved disulfide bridges was manually checked, such as the bridge between the corresponding Cys77 and Cys166 residues in the *h*A2A. Induce Fit Docking Workflow in the Schrodinger package was used to optimize the final models [58]. Selective high affinity ligands (compounds coll\_11 and jaco\_mre3008\_f20) [50] were used to adapt the protein pocket for the *h*A<sup>1</sup> and *h*A3, respectively. This procedure involved three steps: 1) Glide-based docking of the ligands using SP mode (standard-precision); 2) Protein pocket optimization using Prime and considering the residues within 5Å from the ligand poses; 3) Glide-based docking of the ligands in the refined pocket using XP mode (Extra-precision). As previously described [50], homology models were tested for their capability to discriminate ligands from decoys and between known subtype-selective compounds. ROC curves were performed, and the best models were selected for further molecular docking studies.

#### 3.4.2. Molecular Docking of *h*A<sup>1</sup> and *h*A<sup>3</sup> ARs

Molecular docking studies using the *h*A<sup>1</sup> and *h*A<sup>3</sup> homology models, selected in the previous step, were carried out. Compounds were docked using Glide SP mode [52]. Ten poses for each ligand were collected and optimized using MM-GBSA in Prime [53], taking into account a flexible protein region defined by 5 Å from the ligand. Final binding modes were selected, taking into account the number of similar poses extracted from the calculations and geometrical correspondence to co-crystallized ligands in the *h*A2A.

#### **4. Conclusions**

The current study was focused on the synthesis and the evaluation of binding affinity towards the four subtypes of human ARs of a selected series of methoxy and hydroxy coumarin–chalcone hybrids. Structure–activity relationship (SAR) studies of the new molecules highlighted that, in general, methoxy substitutions, as in the example of compounds **3** and **4,** allow a superior *h*A<sup>3</sup> binding affinity and selectivity, whereas the hydroxy substitutions, as in the example of compounds **5**–**8**, allow a modest *h*A<sup>1</sup> binding affinity. Substitutions at positions 5 and 7 of the coumarin scaffold proved to be essential for the potency and selectivity in both series of compounds. Compound **4**, a methoxy derivative, and compound **7**, a hydroxy derivative, proved to be the most potent compounds of the studied series, displaying a *h*A<sup>3</sup> *K*<sup>i</sup> = 2.49 µM and a *h*A<sup>1</sup> *K*<sup>i</sup> = 17.7 µM, respectively. Docking calculations allow an understanding the binding preference of the studied molecules. Finally, the theoretical values for the ADME properties show that all the coumarin–chalcone hybrids **1**–**8** do not break any of Lipinski's rules, being promising scaffolds for further compound optimization.

**Author Contributions:** Conceptualization, S.V.-R. and M.J.M.; methodology, S.V.-R., S.V. and S.K.; software, S.V.; validation, K.-N.K., F.B. and E.U.; formal analysis, S.V.-R., S.K. and S.V.; investigation, S.V.-R., S.K., S.V. and M.J.M.; resources, S.V.-R., S.K., S.V., K.-N.K., E.U., F.B. and M.J.M.; data curation, S.V.-R., S.K., S.V.; writing—original draft preparation and editing, S.V.-R.; writing—review and editing, M.J.M.; visualization, K.-N.K.; supervision, F.B. and E.U.; project administration, S.V.-R. and M.J.M.; funding acquisition, S.V.-R., S.K., S.V., K.-N.K., E.U., F.B. and M.J.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Xunta de Galicia (Galician Plan of Research, Innovation and Growth 2011–2015, Plan I2C, ED481B 2014/027-0, ED481B 2014/086–0 and ED481B 2018/007), Angeles Alvariño program from Xunta de Galicia, European Social Fund (ESF) and Fundação para a Ciência e Tecnologia (FCT, CEECIND/02423/2018 and UIDB/00081/2020).

**Acknowledgments:** The authors would like to thank Lourdes Santana for her scientific support. The authors would like to thank the use of RIAIDT-USC analytical facilities.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Sample Availability:** Samples of the compounds are available from the authors.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Coumarin-Chalcone Hybrids as Inhibitors of MAO-B: Biological Activity and In Silico Studies**

**Guillermo Moya-Alvarado 1,†, Osvaldo Yañez 2,3,† , Nicole Morales <sup>4</sup> , Angélica González-González <sup>5</sup> , Carlos Areche <sup>6</sup> , Marco Tulio Núñez <sup>7</sup> , Angélica Fierro 8,\* and Olimpo García-Beltrán 9,10,\***


**Abstract:** Fourteen coumarin-derived compounds modified at the C3 carbon of coumarin with an α,β-unsaturated ketone were synthesized. These compounds may be designated as chalcocoumarins (3-cinnamoyl-2*H*-chromen-2-ones). Both chalcones and coumarins are recognized scaffolds in medicinal chemistry, showing diverse biological and pharmacological properties among which neuroprotective activities and multiple enzyme inhibition, including mitochondrial enzyme systems, stand out. The evaluation of monoamine oxidase B (MAO-B) inhibitors has aroused considerable interest as therapeutic agents for neurodegenerative diseases such as Parkinson's. Of the fourteen chalcocumarins evaluated here against MAO-B, **ChC4** showed the strongest activity in vitro, with IC<sup>50</sup> = 0.76 ± 0.08 µM. Computational docking, molecular dynamics and MM/GBSA studies, confirm that **ChC4** binds very stably to the active rMAO-B site, explaining the experimental inhibition data.

**Keywords:** chalcocoumarin; MAO-B; molecular dynamics; in silico studies; neurodegenerative diseases

### **1. Introduction**

Coumarins (α-benzopyrones, 2*H*-chromen-2-ones) are a large family of compounds, of natural and synthetic origin, that show numerous biological [1–6] and medicinal chemistry activities, such as anticoagulant, anticancer, antioxidant, antiviral, anti-diabetic, antiinflammatory, antibacterial, antifungal and anti-neurodegerative properties [7–9], among which recent studies have paid special attention to enzyme inhibition. With regard to monoamine oxidase (MAO) inhibition, recent findings have revealed that MAO affinity and selectivity can be efficiently modulated by appropriate substitutions on the coumarin ring system [1,10–13].

MAOs (EC 1.4.3.4) are flavoproteins located in the outer mitochondrial membrane and involved in the oxidative deamination of endogenous and exogenous monoamines

**Citation:** Moya-Alvarado, G.; Yañez, O.; Morales, N.; González-González, A.; Areche, C.; Núñez, M.T.; Fierro, A.; García-Beltrán, O. Coumarin-Chalcone Hybrids as Inhibitors of MAO-B: Biological Activity and In Silico Studies. *Molecules* **2021**, *26*, 2430. https:// doi.org/10.3390/molecules26092430

Academic Editor: Maria João Matos

Received: 1 April 2021 Accepted: 18 April 2021 Published: 22 April 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

using oxygen (O2) as electron acceptor. In humans they exist in two isoforms called MAO-A and MAO-B. The high resolution crystal structures of both human isoforms A and B (hMAO) rat MAO-A (rMAO-A) have made it possible to analyze binding modes of ligands inside these macromolecules [14]. While the active site is formed by the common FAD cofactor and similar amino acid residues in the different forms, these are distinguished by their selectivity for substrates and inhibitors [15]. Thus, serotonin and noradrenaline are substrates of MAO-A which is selectively inhibited by clorgyline, while MAO-B oxidizes β-phenylethylamines and benzylamines and is selectively inhibited by l-deprenyl. MAO genes are expressed in various tissues. However, in the brain, although both isoforms are widely distributed, MAO-B is expressed in high concentrations in the hypothalamus, striatum, globus pallidus and thalamus, and mainly in serotonergic cells while the A isoform is rather evenly distributed, mainly in the cortex, and in nuclei containing preferably catecholaminergic and glial cells [16–21].

Although knowledge about MAO inhibition by compounds containing coumarin scaffolds is scarce, publications of articles describing new inhibitors of this class of compounds are increasing. The variety of substitutions on the coumarin ring provide insight into the influence on the activity-structure relationship. Among the most reported modifications of the coumarin ring with MAO activity are on C3 and the steric effect of the substituent appears to be important in modulating MAO-B inhibitory activity [11]. In addition, it has been reported that the introduction of various substituents at the *para* position of the 3-phenyl ring is a good strategy for improving the desired MAO-B inhibitory activity [22] and when the 3-phenyl skeleton is replaced by a 3-benzoyl group, the activity is strongly diminished [20]. It has also been observed that coumarins substituted with 3-indolyl and 3-thiophenyl shows greatest selective inhibition was against MAO-B [11,23,24].

In this work a merger of the coumarin scaffold and a 3-cinnamyl group led to new hybrid (chalcocoumarin) derivatives (Scheme 1) that preserve structural characteristics of compounds with the ability to interact with MAO. The synthetic strategy chosen allowed a large variety of substituents on the cinnamyl benzene ring to be accessed using different readily accessible benzaldehydes. Thus, the quantity and/or type of interactions with the enzyme were explored involving some bulky groups to determine their possible contribution to the biological activity as MAOIs. Our new compounds were screened versus both MAO isoforms, and in silico studies were carried out to rationalize their main interactions in the MAO active cavity. The computational biochemistry tools were used considering the geometrical restrictions and most probable positions in the formation of the ligand-receptor complex. The chalcocoumarin molecules were subjected to theoretical studies in which binding energies were estimated using docking and MM/GBSA analysis. In addition, physicochemical parameters that are responsible for governing the pharmacokinetic properties of drug molecules were determined.

**Scheme 1.** Hybrids of chalcocoumarin.

In the present study, a series of coumarin-chalcone hybrid compounds were synthesized and tested on the 2 MAO isoforms. The activity shown was selective for MAO-B and in particular, compound **ChC4** showed the highest inhibitory activity on rMAO-B at submicromolar concentrations. The results obtained will be useful to understand the mode of inhibition of chalcocoumarins against rMAO-B, and to help predict the activities of these new inhibitors that could be promising as therapeutics to treat neurodegenerative diseases such as Parkinson's disease.

#### **2. Results and Discussion**

#### *2.1. Chemistry*

The route employed to synthesize the compounds is summarized in Scheme 1. The compounds were obtained starting from resorcinol (**1**), which was formylated using the Vilsmaier-Haack reaction [25]. Knoevenagel condensation of the aldehyde intermediate with ethyl acetoacetate afforded hydroxycoumarin **2**. The 7-hydroxycoumarin obtained was methylated using a Williamson reaction using methyl sulfate as methylating agent, obtaining the compound **3**, finally the compounds derived the 3-cinnamoyl-2H-chromen-2-one (**ChC1**–**ChC14**) (Table 1) were prepared in moderate yields (25–47%, unoptimized) by Claisen-Schmidt condensation with the respective aldehyde (Supplementary Materials; Scheme S1) [26]. Coumarin-chalcone hybrids have been studied and are currently still being synthesized for various uses and their spectroscopy is well known, however, we will detail some signals that are key to their identification. The <sup>1</sup>H-NMR spectra of the compounds **ChC1**–**ChC14** present very similar chemical shift patterns with a particular signal that identifies this type of molecules, the neighboring vinyl protons of the α,β-unsaturated ketone appear at very close low field from the aromatic proton region. These protons present signals corresponding to two doublets with variable δ between 8.5 and 7.0 and with Jab = 16 Hz on average. this high constant corresponds to a trans isomer [27–29]. As for the <sup>13</sup>C-NMR spectrum, we will mention typical signals such as carbonyl shifts. first of all, we will detail that the carbon of the α,β-unsaturated ketone has a δ 190–180 ppm and carbonyl carbons of α-pyrone δ 165–155 pmm on average [27–29], the compounds were characterized by <sup>1</sup>H and <sup>13</sup>C NMR (Supplementary Materials; Figures S1–S15).

**Table 1.** Structures of the synthetized chalcocoumarin hybrids.


‐

‐

μ

μ

‐ ‐

‐ ‐

μ

‐ ‐

‐

‐

‐


**Table 1.** *Cont.*

#### *2.2. Biological Analysis in Rat MAO*

Fourteen derivatives differing in the substitution pattern of the cinnamyl benzene ring were studied, these compounds were tested on rat MAO-A and B to determine their inhibitory activity MAO. A general screening was carried out at 10 µM finding moderate activity for some of the compounds against rMAO-B but none against rMAO-A. Thus, five molecules were identified as possibly selective IMAO-B.

**ChC4**, **ChC5**, **ChC6**, **ChC9** and **ChC11** in MAO-B exhibited micromolar or submicromolar in vitro potencies, all below 10 µM (Table 2). Out of these **ChC4**, substituted with a hydroxyl group on the meta position of the variable ring, displayed the highest rMAO-B inhibitory activity (IC<sup>50</sup> = 0.76 µM). Interestingly, changing the position of the hydroxyl group from meta to ortho or para (**ChC2** and **ChC10** respectively) led to loss of the inhibitory activity. An approximately 12-fold lower IMAO activity was observed when the hydroxyl group (in **ChC4**) was methylated (**ChC5**). This might be attributed to steric hindrance and/or to the loss of hydrogen bonding donor quality which could be crucial for some interaction in the binding site. Moving the methoxyl group from the meta to the para position (**ChC6** vs. **ChC5**) slightly increased potency.


**Table 2.** IC<sup>50</sup> of the compounds in rMAO-A and rMAO-B.

Each IC<sup>50</sup> value was obtained from an average of three evaluations (*<sup>n</sup>* = 3). Replacing the methoxy group of **ChC6** with a bulkier, less electronegative and more polarisable methylthio group (**ChC11**) only produced **ChC6** is less potent than **ChC11**. The second most potent molecule was **ChC9**, with a methylenedioxy group bridging the meta and para carbons. The methylenedioxy group increases the rigidity of the molecule, possibly stabilizing the complex protein-ligand interaction. The same effect, extending the rigidity, has been observed, on other derivatives as IMAO [30,31].

#### *2.3. Molecular Docking and Ligand Efficiency Analysis*

To analyze the changes in potency of the coumarin-chalcone hybrids, docking studies were carried out using the crystal structure of rMAO-A and the homology model of rMAO-B (Supplementary Materials; Figures S16 and S17), analyzing the possibility that each one of them has to form a stable complex with each of the 14 molecules synthesized by us. Table 3 shows that most of these molecules present better interactions with the rMAO-B binding site, since the corresponding energies are at least 2.5 kcal mol more negative in all but one of the cases. This difference could be due to the substitutes present in the molecules. The results of the molecular docking experiments showed more favorable interactions (more negative ∆*Ebinding*) for the complexes in rMAO-B than in rMAO-A, with average values around −9.26 kcal·mol−<sup>1</sup> vs. −6.57 kcal·mol−<sup>1</sup> respectively) which are in accord with the experimental data for the whole series. In the rMAO-B, although no major differences were observed in the binding modes of the active compounds, subtle energy differences were identified. The results of this molecular docking study point to strong interactions of the chalcocoumarins in the binding pocket of rMAO-B, but considerably weaker in rMAO-A.

**Table 3.** Molecular docking results for **ChC1**–**ChC14** in the rMAO-A/rMAO-B models. Intermolecular docking energy values (∆*Ebinding*), *K<sup>d</sup>* values and calculated Ligand Efficiency (*LE*) for the rMAO-A and rMAO-B complexes.


*a* In each site, the energy was calculated to see which site bound more strongly to the ligand. In bold ChC4 displayed the highest rMAO-B inhibitory activity.

> When analyzing the docking results for rMAO-B from the conformational viewpoint, it is necessary to consider the residues that constitute the substrate-binding site of rMAO-B, which is composed of the FAD cofactor, two flanking residues, Tyr398 and Tyr435, that form an "aromatic box", and a number of others, particularly Cys172, Tyr326, Met341, Ser200 Gln206 and Thr314 [32,33]. The results show that all the chalcocoumarins settle in the active site of rMAO-B (Supplementary Materials; Figure S18), with the benzene ring of the coumarin moiety close to the FAD, more specifically the central N-5, at a distance of about 4.0 Å. The benzene ring of the cinnamyl moiety extends into the generally hydrophobic entrance cavity adjoining the substrate-binding site. The mere length of the ChC molecules indicates that to bind in the active site of MAO-A the latter must undergo a rearrangement of the residues separating the entrance and the substrate cavities, which may explain their general preference for MAO-B.

> **ChC4** was located inside the cavity interacting with Tyr435, Tyr398, Tyr60, Phe343, Asn83, Arg307, Thr314, and Leu328. Two hydrogen bonds where generated with Asn83 and, via its C-3′ hydroxyl group, Thr316. **ChC2** actually when interact with the amino acids into the pocket adopt a planar conformation because the hydrogen bond confirms our discussion that could be responsible for none activity of **ChC2** in rMAO-B. A quantum geometric optimization of **ChC2** and **ChC4** at the M05-2X-D3/6-31G(d,p) level of theory, showed their C-2′ and C-3′ hydroxyl groups pointing in opposite directions, suggesting

different preferred intramolecular interactions (Figure 1). Both the different electronic potential distribution and the resulting preferred intermolecular interaction might be responsible for the difference in in IC<sup>50</sup> values.

The best three ligands obtained from the docking exhibit low *K<sup>d</sup>* values, these ligands are **ChC4**, **ChC9** and **ChC13**, which means that these ligand/rMAO-B complexes are the most stable in the series. These results are consistent with those obtained in the docking experiments in which these complexes were the most stable according to their ∆*Ebinding* values. The proposed tolerable values of *LE* for inhibitor candidates are *LE* > 0.3 kcal·mol−<sup>1</sup> [34–36]. According to this reference value, **ChC4** is a good prospect for development as an rMAO-B inhibitor with a *LE* value of 0.408. Although **ChC2** and **ChC4** have similar *K<sup>d</sup>* , *LE* and ∆*Ebinding* values, the **ChC2** molecule does not show in vitro activity against rMAO-B, on the other hand, ChC4 has a good inhibitory activity against rMAO-B, since micromolar concentrations are needed to inhibit it, which is consistent with the values obtained for the *K<sup>d</sup>* . ChC1 would appear to be almost as good, with *LE* = 0.404, but again its activity, if any, is worse than our cutoff value. The low micromolar-active **ChC5**, **ChC6**, **ChC9** and **ChC11** have *LE* values of 0.384, 0.372, 0.380 and 0.356, respectively, and the less (or in-) active **ChC13** and **ChC14** have *LE* values of 0.380 and 0.391.

**Figure 1.** (**A**) NCIplot of the non-covalent interaction regions with isosurface gradient (0.6 au) for **ChC2** (left) and **ChC4** (right). (**B**) Electrostatic potential (in a.u.) of **ChC2** (left) and **ChC4** (right) mapped on the 0.001 a.u. isodensity surface for the selected structure computed at the M05-2X-D3/6-31G(d,p) level of theory.

#### *2.4. Analysis of Molecular Dynamics Simulations*

Molecular dynamics simulations were performed for 100 ns to analyze the conformational stability of the rMAO-B/**ChC2** and rMAO-B/**ChC4** complexes. The RMSD, a quantitative parameter, was used to estimate the stability of the protein-ligand systems and the apoprotein. The RMSD in Figure 2A shows that the rMAO-B/**ChC2** and rMAO-B/**ChC4** complexes remain highly stable throughout the simulation time. We can see that the structures of the complexes does not change significantly. The RMSD values for the **ChC4** complex are remarkably constant about 1.5 Å, with a very slight instability and increase near the end of the simulation. The **ChC2** complex shows similar, somewhat less stable behavior for almost 40 ns, and then its RMSD value falls abruptly to about 1.0 Å and rises slowly with appreciable fluctuations to about 1.2 Å at 100 ns, indicative of weaker

binding in the rMAO-B site. However, a maximum difference of 3.0 Å in the RMSD is taken to indicate that a system is in equilibrium [37], so this condition is fulfilled by both compounds. To complement the analysis carried out calculating the RMSD, the Radius of Gyration (RGyr) was analyzed for the same runs. From this analysis (Figure 2B), we can conclude that the RGyr of **ChC2** and **ChC4** oscillate in a narrow interval between 4.3–4.8 Å. These stable values during the 100 ns simulation indicate again that ligand binding does not induce major conformational changes in the protein structure.

**Figure 2.** (**A**) Root Mean Square Deviation (RMSD) and (**B**) Radius of gyration (RGyr) as a function of simulation times for the complexes formed between rMAO-B and **ChC2** and **ChC4**.

Structural studies in MAO have shown that two residues Tyr398 and Tyr435 in MAO-B located in the active site approximately perpendicular to the FAD play a functional role in this enzyme, acting as a cofactor stabilizing the active site, forming an aromatic box whose function is to stabilize the ligand [13]. Molecular simulation results show a difference in the interaction of the compounds **ChC2** and **ChC4** with the FAD cofactor, see Figure 3A. Compound **ChC2** shows a spacing that fluctuates between 17.0 Å and 20.0 Å from its original position, signifying a null interaction with the FAD cofactor. On the other hand, the compound **ChC4** is within the range of interaction with the FAD cofactor. This distance was measured between the nitrogen atom of the alloxazine planar ring of FAD and the center of the benzaldehyde aromatic ring of compounds, Figure 3B.

Molecular dynamics simulations showed of rMAO-B that residues that interact with the ligands ChC2 and ChC4, see Figure 4. The most frequent residues in rMAO-B/**ChC2** were Ile164, Ile199, Leu167, Leu171, Phe168, Pro104, Trp119, Val316, Phe103, Pro102, Tyr115 and Thr196. In contrast, the most frequent residues in rMAO-B/**ChC4** were Ile164, Ile199, Leu171, Phe168, Pro104, Trp119, Tyr326, Val316, Cys172 and Tyr115 with van der Waals and hydrogen bonds interactions. Highlighting residues Cys172 and Tyr326, which

are important for the active site of the rMAO-B flavoprotein. Tyr326 and Cys172 are key residues that determines substrate and inhibitor specificity, also exhibits conformational changes on the inhibitor binding and restricts the binding of certain inhibitors (e.g., harmine) to human MAO-B [38]. These results documents that ChC4 is a reversible inhibitor of rMAO-B.

**Figure 3.** (**A**) Last frame of the molecular simulation showing the positions between the FAD molecule and **ChC2**–**ChC4** compounds interacting with rMAO-B. (**B**) Distance as a function of simulation time, between the nitrogen atom of the aloxazine planar ring of FAD and the center of the benzaldehyde aromatic ring, for compounds **ChC2** and **ChC4**. Dashed lines represent the position of the nitrogen atom of the aloxazine planar ring.

**Figure 4.** Frequency of the appearance of residues at a distance of 3.0 Å or closer from ligands (**A**) **ChC2** and (**B**) **ChC4** calculated using MD procedures.

The analyses of trajectories indicate that during most of the simulation the ligand **ChC4** maintain hydrogen bonds with residues of the active site of rMAO-B. However, the number of hydrogen bonds formed was different for **ChC2** and **ChC4** (Figure 5). **ChC2** formed two hydrogen bonds between the residues Glu483 and Tyr115, highlighting the participation of the residues Val316, Ala325, Ile164 and Leu167. Finally, **ChC4** formed two hydrogen bonds with the Phe168, Cys172, Ile164 and Tyr115, highlighting the participation of the residues Ile199, Trp119 and Tyr326. These residues, see Figure 6, are

consistent with previous theoretical-experimental studies carried out [39,40] where they detail the interaction that some of the synthesized compounds have with the active site of the rMAO-B. This difference in the formation of hydrogen bonds with key residues in rMAO-B could be explained the difference in experimental activity between the **ChC2** and **ChC4** compounds.

**Figure 5.** Fraction (in %) of intermolecular hydrogen bonds for rMAO-B interacting with (**A**) **ChC2** and (**B**) **ChC4**. The graph bar shows the most common hydrogen bonds formed between the residues on the pocket and the inhibitors.

**Figure 6.** Schematic representations at the end (100 ns) of their respective production runs for ligands (**A**) **ChC2** and (**B**) **ChC4** bound to rMAO-B. (**I**) The surrounding amino acid residues in the binding pocket of rMAO-B within 4Å from ligands. (**II**) Two-dimensional interaction map of **ChC2** and **ChC4** and rMAO-B. The arrows indicate potential interactions between amino acid residues and the ligands.

Finally, the binding free energy (MM/GBSA) was computed after the MD simulation; the last 70 ns for all the complexes and the results are given in Table 4. The compound **ChC2** has a binding free energy of -29.06 kcal·mol−<sup>1</sup> with rMAO-B enzyme, while the compound **ChC4** showed relatively binding free energy of –25.87 kcal·mol−<sup>1</sup> . The results

obtained from MM/GBSA show a slight difference in their binding free energy between **ChC2** and **ChC4** compounds bound to rMAO-B. This slight difference is due to the R1 to R2 position of the hydroxyl group in benzaldehyde aromatic ring. In particular, the **ChC4** compound has a better activity at the experimental and in silico level.

**Table 4.** Predicted binding free energies (kcal·mol−<sup>1</sup> ) and individual energy terms calculated from molecular dynamics simulation through the MM-GBSA protocol for rMAO-B complexes.


#### *2.5. In Silico Pharmacokinetic Prediction*

A good drug candidate is absorbed in required time and well distributed throughout the system for its effective metabolism and action. Toxicity is another very important factor that often overshadows the ADME behaviour. SwissADME explorer online was used for in silico prediction of drug likeness of the synthesized compounds (**ChC1**–**ChC14**) based on various molecular descriptors and the results are presented in Table 5.


**Table 5.** In silico predicted physicochemical properties of all compounds **ChC1**–**ChC14**.

MW = 150–500 g/mol; *TPSA* = 20 Å2–130 Å<sup>2</sup> ; HBA = Nº of H-bond acceptors ≤ 10; HBD = Nº of H-bond donor ≤ 5; *RB* = 0–9; Log S = Insoluble < −10 < Poorly < −6 < Moderately < −4 < Soluble < −2; Log P ≤ 5; log K<sup>p</sup> ≥ −2.5 considered to be permeable; Nº Violations of Lipinski, Ghose, Veber, Egan and Muegge rules. **\*** Violation of Ghose and Muegge rules.

> The most potent compound ChC4 in biological experiment data having logP value of 2.97, it's clear that it doesn't violate of five Lipinski rules, while the other molecules have logP values in the range of 2.90–4.48 and are expected to be orally active. In addition, the logS values for **ChC4** have a value of −4.94 indicating proper solubility, which is an indication of favorable drug like property, makes compound **ChC4** promising drug candidate for further research and development. Thirteen of fourteen synthesized molecules do not break the rules of Lipinski, Ghose, Veber, Egan and Muegge, since the molecule **ChC12** breaks the rules of Ghose and Muegge.

> The Boiled-egg model is proposed as an accurate predictive model that works by computing the lipophilicity and polarity of small molecules. The Boiled-egg analysis of the fourteen molecules (Figure 7) has shown that compounds **ChC1**, **ChC3**, **ChC5**, **ChC6**, **ChC12**, **ChC13** and **ChC14** are highly absorbable at the brain barrier, whereas compounds **ChC2**, **ChC4**, **ChC7**, **ChC8**, **ChC9**, **ChC10** and ChC11 are highly absorbable in the gastrointestinal tract.

**Figure 7.** Predictive human intestinal absorption (HIA) model and blood-brain barrier permeation (BBB) method (boiled-egg plot) of the 14 compounds.

The ADMET properties showed much similarity among the thirteen molecules that can be used for advanced clinical trials.

#### **3. Materials and Methods**

#### *3.1. Solvents and Reagents*

Solvents and reagents (analytical grade and spectroscopic grade) were obtained from Sigma-Aldrich (St. Louis, MO, USA) and Merck (Darmstadt, Germany). Melting points were determined on a Galen III hot-plate microscope (Reichert-Jung, St. Louis, MO, USA) equipped with a thermocouple. <sup>1</sup>H- and <sup>13</sup>C-NMR spectra were recorded on a 400 MHz multidimensional spectrometer (Bruker Corporation, Billerica, MA, USA) using the solvent or the TMS signal as an internal standard.

#### *3.2. Synthesis*

*3-Cinnamoyl-7-methoxy-2H-chromen-2-one* (**ChC1**). 3-Acetyl-7-methoxy-2H-chromen-2-one (0.44 g, 2.0 mmol) and benzaldehyde (0.21 g, 2.0 mmol) were dissolved in 25 mL of DCM and to this solution 0.5 mL of piperidine were added. The mixture was kept at reflux temperature, monitoring the reaction by TLC for 10 h. The solution was concentrated under reduced pressure and dissolved in a small aliquot of DCM and then MeOH was added in excess to induce precipitation. This procedure was performed twice. The precipitate was finally purified by column chromatography on silica gel eluting with DCM: 0.25 g yellow solid, 40.8%, m.p. 190–192 ºC; <sup>1</sup>H NMR (CDCl3): δ 8.59 (s, 1H, Ar-H), 8.01 (d, 1H, *J* = 15.8 Hz, Ar-CH), 7.85 (d, 1H, *J* = 15.8 Hz, CO-CH=), 7.67 (s, 2H, Ar-H), 7.56 (d, 1H, *J* = 8.6 Hz, Ar-H), 7.40 (s, 3H, Ar-H,), 6.90 (dd, 1H, *J* = 8.6, 1.0 Hz, Ar-H), 6.85 (s, 1H, Ar-H), 3.91 (s, 3H, OCH3). <sup>13</sup>C-NMR (CDCl3): δ 56.1, 100.2, 112.4, 113.8, 124.0, 129.0, 130.6, 131.5, 135.3, 144.3, 148.5, 157.8, 160.0, 165.0, 186.3.

*(E)-3-(3-(2-Hydroxyphenyl)acryloyl)-7-methoxy-2H-chromen-2-one* (**ChC2**). 3-Acetyl-7 methoxy-2*H*-chromen-2-one (0.44 g, 2.0 mmol) and *o*-hydroxybenzaldehyde (0.24 g, 2.0 mmol) were reacted and worked up according to the previous procedure: 1.75 g, pale white solid, 95.6%, m.p.; 188–190 ◦C. <sup>1</sup>H-NMR (DMSO-d6), δ 10.43 (s, 1H, OH), 8.71 (s, 1H, Ar-H), 8.04 (d, 1H, *J* = 15.9 Hz, Ar-CH=), 7.92 (d, 1H, *J* = 8.8 Hz, Ar-H), 7.89 (d, 1H, *J* = 15.9 Hz), 7.69 (dd, 1H, *J* = 7.7, 1.0 Hz, Ar-H), 7.35 (t, 1H, *J* = 7.0 Hz, Ar-H), 7.13 (d, 1H, *J* = 2.2 Hz, Ar-H),

7.08 (dd, 1H, *J* = 8.6, 2.2 Hz, Ar-H), 7.00 (d, 1H, *J* = 8.0 Hz, Ar-H), 6.94 (t, 1H, *J* = 7.5 Hz, Ar-H), 3.96 (s, 3H, OCH3), <sup>13</sup>C-NMR (DMSO-*d6*): δ 56.7, 100.8, 112.5, 113.9, 116.8, 120.0, 121.8, 121.9, 124.3, 129.2, 132.3, 132.6, 139.5, 148.2, 157.4, 157.8, 159.4, 165.1, 187.1.

*(E)-7-methoxy-3-(3-(2-methoxyphenyl)acryloyl)-2H-chromen-2-one*(**ChC3**). 3-Acetyl-7-methoxy-2*H*-chromen-2-one (0.44 g, 2.0 mmol) and *o*-methoxybenzaldehide (0.27 g, 2.0 mmol), were reacted and worked up according to the previous procedure: 0.330 g, pale white solid, 49%, m.p.. 184–186 ◦C; <sup>1</sup>H-NMR (CDCl3): δ 8.55 (s, 1H, =C-H), 8.21 (d, 1H, Ar-CH=, *J* = 15.8 Hz), 8.04 (d, 1H, CO-CH=, *J* = 15.8 Hz), 7.71 (d, 1H, Ar-H, *J* = 7.6 Hz), 7.56 (d, 1H, Ar-H, *J* = 8.4 Hz), 7.37 (t, 1H, Ar-H, *J* = 7.9), 6.98 (t, 1H, Ar-H, *J* = 7.7 Hz), 6.92 (d, 1H, Ar-H, *J* = 8.2 Hz), 6.90 (d, 1H, Ar-H, *J* = 8.5 Hz), 6.85 (s, 1H, Ar-H), 3.91 (s, 6H, 2 OCH3); <sup>13</sup>C-NMR (DMSO-d6): δ 55.6, 56.0, 100.4, 111.2, 112.5, 113.7, 120.8, 121.9, 124.1, 124.6, 129.3, 131.3, 132.0, 139.9, 148.2, 157.6, 159.0, 159.8, 165.0, 186.8.

*(E)-3-(3-(3-hydroxyphenyl)acryloyl)-7-methoxy-2H-chromen-2-one* (**ChC4**). 3-Acetyl-7 methoxy-2*H*-chromen-2-one (0.44 g, 2.0 mmol) and *m*-hydroxybenzaldehyde (0.24 g, 2.0 mmol) were reacted and worked up according to the previous procedure: 0.195 g, white solid, 30%, m.p. 184–186 ◦C; <sup>1</sup>H NMR (DMSO-d6): δ 9.77 (sbr, 1H), 8.76 (s, 1H, =C-H), 7.94 (d, 1H, Ar-H, *J* = 8.0 Hz), 7.82 (d, 1H, Ar-CH=, *J* = 15.9 Hz), 7.71 (d, 1H, CO-CH=, *J* = 15.9 Hz), 7.62 (d, 1H, Ar-H, *J* = 8.6 Hz), 7.33 (m, 1H, Ar-H), 7.25-7.17 (m, 2H, Ar-H), 7.15 (s, 1H, Ar-H), 7.09 (dd, 1H, Ar-H, *J* = 8.0, 1.0 Hz), 6.94 (dd, 1H, Ar-H, *J* = 8.0, 1.0 Hz), 3.97 (s, 3H, OCH3); <sup>13</sup>C NMR (DMSO-d6): 56.7, 100.9, 112.5, 114.0, 114.8, 118.5, 120.5, 121.6, 124.9, 130.6, 132.5, 136.3, 143.9, 148.6, 157.5, 158.2, 159.4, 165.3, 186.7.

*(E)-7-methoxy-3-(3-(3-methoxyphenyl)acryloyl)-2H-chromen-2-one* (**ChC5**). 3-Acetyl-7-methoxy-2*H*-chromen-2-one (0.44 g, 2.0 mmol) and *m*-methoxybenzaldehide (0,27 g, 2.0 mmol), were reacted and worked up according to the previous procedure: 0,280 g, pale white solid, 42%, m.p. 164–166 ◦C; <sup>1</sup>H-NMR (CDCl3): δ 8.58 (s, 1H, =C-H), 7.99 (d, 1H, Ar-CH=, *J* = 15.9 Hz), 7.82 (d, 1H, CO-CH=, *J* = 15.9 Hz), 7.57 (d, 1H, Ar-H, *J* = 8.0 Hz), 7.35-7.25 (m, 2H, Ar-H), 7.18 (s, 1H, Ar-H) 6.96 (dd, 1H, Ar-H, *J* = 8.0, 2.0 Hz), 6.91 (dd, 1H, Ar-H, *J* = 8.8, 2.0 Hz), 6.85 (d, 1H, Ar-H, *J* = 2.0 Hz), 3.83 (s, 3H, OCH3), 3.76 (s, 3H, OCH3); <sup>13</sup>C-NMR (CDCl3): δ 55.8, 56.5, 100.8, 112.8, 113.9, 114.3, 117.1, 121.7, 122.0, 124.9, 130.3, 131.8, 136.8, 144.8, 149.0, 158.1, 160.2, 160.3, 165.8, 186.8.

*(E)-7-methoxy-3-(3-(4-methoxyphenyl)acryloyl)-2H-chromen-2-one*(**ChC6**). 3-Acetyl-7-methoxy-2*H*-chromen-2-one (0.44 g, 2.0 mmol) and *p*-methoxybenzaldehide (0.27 g, 2.0 mmol), were reacted and worked up according to the previous procedure: 0.26 g, pale white solid, 39%, m.p. 158–160 ◦C; <sup>1</sup>H-NMR (CDCl3): δ 8.58 (s, 1H, =C-H), 7.91 (d, 1H, Ar-CH=, *J* = 15.9 Hz), 7.84 (d, 1H, CO-CH=, *J* = 15.9 Hz), 7.64 (d, 2H, Ar-H, *J* = 8.0 Hz), 7.56 (d, 1H, Ar-H, *J* = 8.8 Hz), 6.93 (d, 2H, Ar-H, *J* = 8.0 Hz), 6.91 (dd, 1H, Ar-H, *J* = 8.8, 2.0 Hz), 6.85 (d, 1H, Ar-H, *J* = 2.0 Hz), 3.92 (s, 3H, OCH3), 3.86 (s, 3H, OCH3); <sup>13</sup>C-NMR (CDCl3): δ 57.2, 57.8, 102.0, 114.2, 115.5, 116.1, 123.4, 123.6, 129.6, 132.5, 133.0, 146.3, 150.0, 159.4, 161.6, 163.5, 166.8, 188.0.

*(E)-3-(3-(3,4-dimethoxyphenyl)acryloyl)-7-methoxy-2H-chromen-2-one* (**ChC7**). 3-Acetyl-7-methoxy-2*H*-chromen-2-one (0.44 g, 2.0 mmol) and 3,4-dimethoxybenzaldehide (0.33 g, 2.0 mmol) were reacted and worked up according to the previous procedure: 0.42 g, bright yellow solid, 57%, m.p. 182–184 ◦C; <sup>1</sup>H-NMR (CDCl3): δ 8.55 (s, 1H, =C-H), 8.16 (d, 1H, Ar-CH=, *J* = 15.9 Hz), 8.00 (d, 1H, CO-CH=, *J* = 15.9 Hz), 7.55 (d, 1H, Ar-H, *J* = 8.6 Hz), 7.34 (d, 1H, Ar-H, *J* = 1.0 Hz), 7.19 (dd, 1H, AR-H, *J* = 8.0, 8.0 Hz) 7.07 (d, 1H, Ar-H, *J* = 8.0 Hz), 7.00 (dd, 1H, Ar-H, *J* = 8.0, 2.0 Hz), 6.94(d, 1H, Ar-H, *J* = 2.0 Hz), 3.90 (s, 3H, OCH3), 3.89 (s, 3H, OCH3), 3.87 (s, 3H, OCH3); <sup>13</sup>C-NMR (CDCl3): δ 55.9, 56.0, 61.5, 100.4, 112.4, 113.8, 114.4, 119.9, 121.6, 124.2, 125.5, 129.2, 131.3, 139.1, 148.4, 149.2, 153.2, 157.7, 159.7, 165.1, 186.7.

*(E)-3-(3-(2,5-dimethoxyphenyl)acryloyl)-7-methoxy-2H-chromen-2-one* (**ChC8**). 3-Acetyl-7-methoxy-2*H*-chromen-2-one (0.44 g, 2.0 mmol) and 2,5-dimethoxybenzaldehide (0.33 g, 2.0 mmol), were reacted and worked up according to the previous procedure: 0,45 g, bright yellow solid, 62%, m.p. 174–176 ◦C; <sup>1</sup>H-NMR (CDCl3): δ 8.55 (s, 1H, =C-H), 8.17 (d, 1H, Ar-CH=, *J* = 15.6 Hz), 8.0 (d, 1H, CO-CH=, *J* = 15.6 Hz), 7.6 (dd, 1H, Ar-H, *J* = 8.0, 9.0 Hz), 7.2 (d, 1H, Ar-H, *J* = 2.0 Hz), 6.92-6.97 (m, 4H, Ar-H), 3.91 (s, 3H, OCH3), 3.87 (s, 3H, OCH3),

3.82 (s, 3H, OCH3); <sup>13</sup>C-NMR (CDCl3): δ 56.2, 56.4, 56.6, 100.7, 100.7, 112.8, 112.9, 113.7, 114.1, 114.3, 118.3, 125.1, 131.7, 131.9, 139.9, 148.6, 153.9, 154.0, 158.0, 165.4, 187.1.

*(E)-3-(3-(benzo[d][1,3]dioxol-5-yl)acryloyl)-7-methoxy-2H-chromen-2-one*(**ChC9**). 3-Acetyl-7-methoxy-2*H*-chromen-2-one (0.44 g, 2.0 mmol) and piperonal (0.30 g, 2.0 mmol) were reacted and worked up according to the previous procedure: 0.29 g, yellow solid, 41%, m.p. 178–180 ◦C; <sup>1</sup>H-NMR (DMSO-d6): δ 8.62 (s, 1H, =C-H), 7.84 (d, 1H, *J* = 12.0 Hz), 7.63 (d, 1H, Ar-CH=, *J* = 15.7 Hz), 7.56 (d, 1H, CO-CH= *J* = 15.7 Hz), 7.68 (d, 1H, Ar-H, *J* = 8.0 Hz), 7.33 (s, 1H, Ar-H), 7.25 (d, 1H, Ar-H, *J* = 8.0 Hz), 7.06 (s, 1H, Ar-H), 7.03-6.94 (m, 3H, Ar-H), 6.07 (s, 2H, OCH2O), 3.87 (s, 3H, OCH3); <sup>13</sup>C-NMR (CDCl3): δ 55.8, 100.6, 101.2, 106.7, 108.4, 111.5, 122.8, 125.7, 128.1, 130.8, 134.6, 142.2, 147.3, 147.9, 153.6, 160.9, 163.5, 187.1.

*(E)-3-(3-(4-hydroxy-3-methoxyphenylacryloyl)-7-methoxy-2H-chromen-2-one* (**ChC10**). 3- Acetyl-7-methoxy-2*H*-chromen-2-one (0.44 g, 2.0 mmol) and vainillin (0.30 g, 2.0 mmol), were reacted and worked up according to the previous procedure: 0,340 g, yellow solid, 48.3%, m.p. 210–212 ◦C; <sup>1</sup>H-NMR (CDCl3): δ 8.65 (s, 1H, =C-H), 8.27 (d, 1H, Ar-CH=, *J* = 16 Hz), 8.10 (d, 1H, CO-CH=, *J* = 16 Hz), 7.65 (d, 1H, Ar-H, *J* = 8.0 Hz), 7.32 (d, 1H, Ar-H, *J* = 2.0 Hz), 7.45 (dd, 1H, Ar-H, *J* = 8.8, 2.0 Hz) 7.00 (dd, 1H, Ar-H, *J* = 8.8, 2.0 Hz), 6.91 (d, 1H, Ar-H, *J* = 8.8 Hz), 6.93 (d, 1H, Ar-H, *J* = 2.0 Hz), 3.97 (s, 3H, OCH3), 3.92 (s, 3H, OCH3); <sup>13</sup>C-NMR (CDCl3): δ 55.9, 56.0, 100.4, 112.4, 112.5, 113.3, 113.7, 117.9, 121.7, 124.6, 124.8, 131.3, 139.6, 148.2, 153.5, 153.6, 157.6, 159.7, 165.7, 186.7.

*(E)-3-(3-(4-mercaptophenyl)acryloyl)-7-methoxy-2H-chromen-2-one* (**ChC11**). 3-Acetyl-7 methoxy-2*H*-chromen-2-one (0.44 g, 2.0 mmol) and methyl(4-vinylphenyl)sulfane (0.27g, 2.0 mmol), were reacted and worked up according to the previous procedure: 0,305 g, yellow solid, 43%.m.p. 196–198 ◦C; <sup>1</sup>H-NMR (CDCl3): δ 8.58 (s, 1H, =C-H), 7.99 (d, 1H, Ar-CH=, *J* = 15.7 Hz), 7.82 (d, 1H, CO-CH=, *J* = 15.7 Hz), 7.58 (m, 3H, Ar-H), 7.26 (m, 3H, Ar-H 6.91 (dd, 1H, Ar-H, *J* = (8.0, 2.0), 6.86 (d, 1H, Ar-H, *J* = 2.0 Hz), 3.91 (s, 3H, OCH3), 2.50 (s, 3H, SCH3); <sup>13</sup>C-NMR (CDCl3): δ 15.6, 56.4, 100.8, 112.9, 114.3, 121.9, 123.6, 126.3, 129.6, 131.7, 131.9, 142.9, 144.4, 148.8, 158.1, 160.2, 165.5, 165.6, 186.7.

*(E)-3-(3-(3,5-dibromo-4-methoxyphenyl)acryloyl)-7-methoxy-2H-chromen-2-one* (**ChC12**). 3-Acetyl-7-methoxy-2*H*-chromen-2-one (0,44 g, 2.0 mmol) and 3,5-dibromo-4-methoxybenzaldehyde (0.58 g, 2.0 mmol), were reacted and worked up according to the previous procedure: 0.21 g, yellow solid, 21.4%, m.p. 228–230 ◦C; <sup>1</sup>H-NMR (CDCl3): δ 8.57 (s, 1H, =C-H), 8.18. (s, 2H, Ar-H), 7.95 (d, 1H, Ar-CH=, *J* = 16 Hz), 7.81 (d, 1H, CO-CH=, *J* = 16 Hz), 7.63 (d, 1H, Ar-H, *J* = 8.0 Hz), 7.22 (dd, 1H, Ar-H, *J* = 8.0, 2.0 Hz), 6.98 (d, 1H, Ar-H, *J* = 2.0 Hz) 4.01 (s, 3H, OCH3), 3.95 (s, 3H, OCH3); <sup>13</sup>C-NMR (CDCl3): δ 60.9, 56.5, 100.5, 110.8, 111.7, 117.4, 123.7, 126.9, 132.3, 133.6, 136.1, 141.9, 147.7, 154.0, 154.5, 157.6, 159.4, 163.2, 188.1.

*(E)-3-(3-(4-(dimethylamino)phenyl)acryloyl)-7-methoxy-2H-chromen-2-one* (**ChC13**). 3-Acetyl-7-methoxy-2*H*-chromen-2-one (0.44 g, 2.0 mmol) and 4-(dimethylamino)benzaldehyde (0.27 g, 2.0 mmol), were reacted and worked up according to the previous procedure: 0.15 g, red solid, 22%, m.p. 220–222 ◦C; <sup>1</sup>H-NMR (CDCl3): δ 8.56 (s, 1H, =C-H), 7.97 (d, 1H, Ar-CH=, *J* = 15.7 Hz), 7.81 (d, 1H, CO-CH=, *J* = 15.7 Hz), 7.58 (d, 1H, Ar-H, *J* = 8.8 Hz), 7.54 (d, 2H, Ar-H, *J* = 8.6 Hz), 6.88 (d, 1H, Ar-H, *J* = 8.0, 2.0 Hz), 6.85 (d, 1H, Ar-H, *J* = 2.0 Hz), 6.68 (d, 2H, Ar-H, *J* = 8.8 Hz), 3.90 (s, 3H, OCH3), 3.04 (s, 6H, N(CH3)2); <sup>13</sup>C-NMR (CDCl3): δ 40.5, 56.4, 64.1, 100.7, 112.2, 113.0, 114.0, 119.3, 122.6, 123.3, 131.4, 131.5, 146.4, 148.2, 152.6, 158.0, 160.3, 165.2, 186.4.

*(E)-3-(3-(4-bromophenyl)acryloyl)-7-methoxy-2H-chromen-2-one* (**ChC14**). 3-Acetyl-7-methoxy-2*H*-chromen-2-one (0.44 g, 2.0 mmol) and *p*-methoxybenzaldehide (0.27 g, 2.0 mmol), were reacted and worked up according to the previous procedure: 0.27 g, pale white solid, 35%, m.p. 158–160 ◦C; <sup>1</sup>H-NMR (CDCl3): δ 8.76 (s, 1H, =C-H), 7.94 (d, 1H, Ar-H, *J* = 8.0, 8.0 Hz), 7.82 (d, 1H, Ar-CH=, *J* = 16 Hz), 7.71 (d, 1H, Ar-CH=, *J* = 16 Hz), 7.33 (d, 1, Ar-H, *J* = 8.0, 8.0 Hz), 7.25-7.18 (m, 2H, Ar-H), 7.15 (s, 1H, Ar-H), 7.09 (dd, 1H Ar-H, *J* = 8.0, 2.0 Hz), 6.94 (dd, 1H Ar-H, *J* = 8.0, 2.0 Hz), 3.97 (s, 3H, OCH3); <sup>13</sup>C-NMR (CDCl3): δ 57.2, 60.8, 110.2, 112.4, 118.5, 126.1, 129.6, 133.0, 134.2, 136.5, 142.0, 147.3, 154.1, 154.4, 159.6, 160.5, 186.7.

#### *3.3. Biological Assessment*

The effect of coumarin derivatives on MAO-A and MAO-B were measured using a suspension of crude rat brain mitochondria as enzyme source. 4-Dimethylaminophenethylamine (4-DMAPEA, 2.5 µM) and 5-hydroxytryptamine (5-HT, 100 mM) were used as substrates selective of MAO-B or MAO-A, respectively. Evaluation of the test compounds on rMAO activity was executed by measuring their effects on the production of 4-dimethylaminophenylacetic acid (DMAPAA) by rMAO-B and 5-hydroxyindoleacetic acid (5-HIAA) by rMAO-A with O<sup>2</sup> using HPLC-ED (L-7110 LaChrom and amperometric detector L-3500 LaChrom Recipe, Hitachi, (Tokyo, Japan) (for more detail see methodological references [41,42]). The IC<sup>50</sup> values (average ± SD was measured in two independent experiments each in triplicate) were assessed representing percentage of inhibition in function of the negative logarithm of different inhibitor concentrations (10−<sup>4</sup> to 10−<sup>8</sup> ) using the GraphPad Prism software [43].

#### *3.4. Computational Analysis*

#### 3.4.1. Homology Modeling

Human monoamine oxidase B (hMAO-B) at 1.6Å resolution was used as template (PDB code 1OJ9) to obtain a 3D structure of rat MAO-B (rMAO-B) using homology modeling. The amino acid sequence and crystal structure of the protein was extracted from NCBI and PDB databases [44,45] considering the high level of amino acid identity (around 90%) the target protein and template were aligned through a single alignment using MultAlin interface [46]. MODELLER9v6 program [47] was used and 100 structures were prepared using standard parameters and the outcomes were ranked on the basis of the internal scoring function of the program (DOPE score). The best model was chosen as the target model. The cofactor FAD was placed inside of MAO using the corresponding crystal coordinates. To analyze the rMAO-B model, VMD program [48] was used to evaluate the 3D distribution and general physical chemistry characteristics. Then, stereochemical and energetic quality of the homology models was evaluated using PROSAII server [49], ANOLEA server [50] and Procheck program [51]. The crystal structure of rMAO-A (PDB code 1O5W [52]) and model of rMAO-B isoform were submited to H++ server [53,54] to computes pK values of ionizable groups and adds missing hydrogen atoms according to the specified pH of the environment as is described in H++ server.

#### 3.4.2. Molecular Docking

Coumarin-chalcone hybrids were docked in the binding cavity of rMAO-A (PDB code 1O5W) and homology model for rMAO-B using AutoDock 4.012 suite. In general, the grid maps were calculated using the AutoGrid 4.0 option and were centered on the sites described before. The volume chosen for the grid maps were made up of 60 × 60 × 60 points, with a grid-point spacing of 0.375 Å. The author's option was used to define the rotating bond in the ligand. In the Lamarckian genetic algorithm (LGA) dockings, an initial population of random individuals with a population size of 150 individuals, a maximum number of 2.5 × 10<sup>7</sup> energy evaluations, a maximum number generation of 27,000, a mutation rate of 0.02 and crossover rate of 0.80 were employed. Each complex was built using the lowest docked-energy binding positions. Van der Waals interaction cutoff distances were set at 12 Å and dielectric constant was 10. The partial charges of each ligand were determined with PM6-D3H4 semi-empirical method [55,56] implemented in the MOPAC2016 [57] software. PM6-D3H4 [56] introduces dispersion and hydrogen-bonded corrections to the PM6 method.

#### 3.4.3. Ligand Efficiency Approach

Ligand efficiency (LE) calculations were performed using one parameter *K<sup>d</sup>* . The *K<sup>d</sup>* parameter corresponds to the dissociation constant between a ligand/protein, and their value indicates the bond strength between the ligand/protein [34–36]. Low values indicate strong binding of the molecule to the protein. *K<sup>d</sup>* calculations were done using the following Equations (1) and (2):

$$
\Delta G^0 = -2.303RT\log(K\_d) \tag{1}
$$

$$K\_d = 10^{\frac{\Lambda G^0}{2.303RT}} \tag{2}$$

where ∆*G* 0 is the binding energy (kcal·mol−<sup>1</sup> ) obtained from docking experiments, *R* is the gas constant, and *T* is the temperature in Kelvin. In standard conditions of aqueous solution at 298.15 K, neutral pH and remaining concentrations of 1 M. The LE allows us to compare molecules according to their average binding energy [36,58]. Thus, it determined as the ratio of binding energy per non-hydrogen atom, as follows (Equation (3)) [34–36,59]:

$$\text{LLE} = -\frac{2.303 \text{RT}}{\text{HAC}} \log(\text{K}\_d) \tag{3}$$

where *K<sup>d</sup>* is obtained from Equation (2) and HAC denotes the heavy atom count (i.e., number of non-hydrogen atoms) in a ligand.

#### 3.4.4. Molecular Dynamic Simulations

Two complexes were built for each modeled **ChC2**/rMAO-B and **ChC4**/rMAO-B, and each model was confined inside a periodic simulation box. Water model TIP3P [60] with 20.459 molecules was used as solvent. Furthermore, Na<sup>+</sup> and Cl<sup>−</sup> ions were added to neutralize the systems and maintain an ionic concentration of 0.15 mol·L −1 .The full geometry optimizations of the two molecules were carried out with the density functional theory method by a M05-2X [61]-D3 [62] in conjunction with the 6-31G(d,p) basis set. **ChC2**, **ChC4** and FAD compounds were parametrized using LigParGen web server and implementing the OPLS-AA/1.14\*CM1A(-LBCC) force field parameters for organic ligands [63–65]. The partial charges of each ligand were determined with generated by the restrained electrostatic potential (RESP) model [66]. MD simulations were carried out using the modeled CHARMM22 and CHARMM36 force fields [67,68] within the NAMD software [69]. First, each system included 20,000 steps of conjugate-gradient energy minimization followed by 10 ns of simulation with the protein backbone atoms fixed and gradually releasing the backbone over 50,000 ps with 10 to 0.001 kcal·mol−1Å−<sup>2</sup> restraints. The total duration of simulation was approximately 100 ns for each system. During the MD simulations, motion equations were integrated with a 1 fs time step in the NPT ensemble at a pressure of 1 atm. The SHAKE algorithm was applied to all hydrogen atoms, and the van der Waals cutoff was set to 12 Å. The temperature was maintained at 310 K, employing the Nosée-Hoover thermostat method with a relaxation time of 1 ps. The Nosée-Hoover-Langevin piston was used to control the pressure at 1 atm. Long-range electrostatic forces were taken into account by means of the particle-mesh Ewald approach. Data were collected every 1 ps during the MD runs. Molecular visualization of the systems and MD trajectory analysis were carried out with the VMD software package [48].

#### 3.4.5. Free Energy Calculation

The molecular MM/GBSA method was employed to estimate the binding free energy of the rMAO-B/ligand complexes. For calculations from a total of 100 ns of MD, the last 70 ns were extracted for analysis, and the explicit water molecules and ions were removed. The MM/GBSA analysis was performed on three subsets of each system: the protein alone, the ligand alone, and the complex (protein-ligand). For each of these subsets, the total free energy (∆*Gtot*) was calculated as follows (Equation (4)):

$$
\Delta G\_{\text{tot}} = E\_{\text{MM}} + G\_{\text{solv}} - T\Delta S\_{\text{conf}} \tag{4}
$$

where *EMM* is the bonded and Lennard–Jones energy terms; *Gsolv* is the polar contribution of solvation energy and non-polar contribution to the solvation energy; T is the temperature; and ∆*Sconf* corresponds to the conformational entropy [70]. Both *EMM* and *Gsolv* were calculated using NAMD software with the generalized Born implicit solvent model [71,72]. ∆*Gtot* was calculated as a linear function of the solvent-accessible surface area, which was calculated with a probe radius of 1.4 Å [73]. The binding free energy of rMAO-B and ligand complexes (∆*Gbind*) were calculated by the difference where ∆*Sconf* values are the averages over the simulation (Equation (5)):

$$
\Delta \mathbf{G}\_{\text{bind}} = \Delta \mathbf{G}\_{\text{tot}}(\text{complex}) - \Delta \mathbf{G}\_{\text{tot}}(\text{protein}) - \Delta \mathbf{G}\_{\text{tot}}(\text{ligand}) \tag{5}
$$

#### 3.4.6. ADMET Prediction

The ADMET properties of a compound deal with its absorption, distribution, metabolism, excretion, and toxicity in and through the human body. ADMET, which constitutes the pharmacokinetic profile of a drug molecule, is very essential in evaluating its pharmacodynamic activities. In this study for all molecules, we have used the SwissADME [74] prediction tool, for in silico physicochemical properties such as molecular hydrogen bond acceptor (*HBA*), hydrogen bond donor (*HBD*), weight (*MW*), topological polar surface area (*TPSA*), rotatable bond count (*RB*), octanol/water partition coefficient (*LogP*), water solubility (*LogS*) and skin permeation (*logKp*). Further the ligands were analyzed for Bioavailability property using Boiled Egg analysis [75].

#### **4. Conclusions**

Fourteen compounds derived from chalcocoumarins were synthesized and evaluated against monoamine oxidase enzyme isoforms. The experimental results obtained against MAO-A and MAO-B show that the compounds **ChC4**, **ChC5**, **ChC6**, **ChC9** and **ChC11** exhibit MAO-B affinity at micro and sub-micromolar concentrations, in particular **ChC4** which shows an IC<sup>50</sup> value of 0 0.76 ± 0.08 µM. Where compound **ChC4** is highlighted in molecular modeling, ADMET predictions, docking and MM/GBSA calculations, these results suggest that compound **ChC4** has the appropriate interactions with the active site of rMAO-B. Furthermore, the ADMETox values obtained for the compound **ChC4** indicate adequate solubility in the gastrointestinal tract, which is a favourable indication for it to be a promising drug candidate for further research and development. This compound complies with the interactions described for the active site of rMAO-, fitting into a distance close enough to the nitrogen atom of the aloxazine planar ring of FAD to form an interaction necessary for the inhibition of rMAO-B. These analyses may be important initial steps towards the development of new drugs in the fight against depressive disorder and Parkinson's disease.

**Supplementary Materials:** The following are available online. **Scheme S1**. Synthetic route to compounds **ChC1**–**ChC14**; **Figure S1**–**S15**: <sup>1</sup>H and <sup>13</sup>C NMR spectrum of **ChC1**–**ChC14**: **Figure S16**. Homology modeling analysis; **Figure S17**. Ramachandran plot generated via PROCHECK for the rMAO-B model; **Figure S18**. Alignment of all the ChC ligands docked in complex with rMAO-B.

**Author Contributions:** O.Y., M.T.N., C.A., A.F. and O.G.-B. contributed to the conception and design of the study; O.Y., A.F. and A.G.-G. preformed the theoretical calculations; O.Y., A.F., and O.G.-B. organized the database; A.F., M.T.N., G.M.-A. and N.M. design and performance of biological assay; O.Y., A.F. and O.G.-B. wrote the first draft of the manuscript. All authors contributed to manuscript revision, read and approved the submitted version. All authors have read and agreed to the published version of the manuscript.

**Funding:** Ministry of Science, Technology and Innovation, the Ministry of Education, the Ministry of Industry, Commerce and Tourism, and ICETEX, Programme Ecosistema Científico-Colombia Científica, from the Francisco José de Caldas Fund, Grand RC-FP44842-212-2018.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Conflicts of Interest:** The authors declare that there is no conflict of interest.

**Sample Availability:** Samples of compounds **ChC1**–**ChC14** are available from the authors.

#### **References**


### *Article* **Identification and Quantification of Coumarins by UHPLC-MS in** *Arabidopsis thaliana* **Natural Populations**

**Izabela Perkowska <sup>1</sup> , Joanna Siwinska <sup>1</sup> , Alexandre Olry <sup>2</sup> , Jérémy Grosjean <sup>2</sup> , Alain Hehn <sup>2</sup> , Frédéric Bourgaud <sup>3</sup> , Ewa Lojkowska <sup>1</sup> and Anna Ihnatowicz 1,\***


**Abstract:** Coumarins are phytochemicals occurring in the plant kingdom, which biosynthesis is induced under various stress factors. They belong to the wide class of specialized metabolites well known for their beneficial properties. Due to their high and wide biological activities, coumarins are important not only for the survival of plants in changing environmental conditions, but are of great importance in the pharmaceutical industry and are an active source for drug development. The identification of coumarins from natural sources has been reported for different plant species including a model plant *Arabidopsis thaliana*. In our previous work, we demonstrated a presence of naturally occurring intraspecies variation in the concentrations of scopoletin and its glycoside, scopolin, the major coumarins accumulating in Arabidopsis roots. Here, we expanded this work by examining a larger group of 28 Arabidopsis natural populations (called accessions) and by extracting and analysing coumarins from two different types of tissues–roots and leaves. In the current work, by quantifying the coumarin content in plant extracts with ultra-high-performance liquid chromatography coupled with a mass spectrometry analysis (UHPLC-MS), we detected a significant natural variation in the content of simple coumarins like scopoletin, umbelliferone and esculetin together with their glycosides: scopolin, skimmin and esculin, respectively. Increasing our knowledge of coumarin accumulation in Arabidopsis natural populations, might be beneficial for the future discovery of physiological mechanisms of action of various alleles involved in their biosynthesis. A better understanding of biosynthetic pathways of biologically active compounds is the prerequisite step in undertaking a metabolic engineering research.

**Keywords:** analytical methods; model plant; natural genetic variation; natural products; simple coumarins

#### **1. Introduction**

Coumarins are secondary metabolites widely distributed throughout the plant kingdom. They are synthetized via the phenylpropanoid biosynthesis pathway. We can distinguish several simple coumarins like coumarin, scopoletin (7-hydroxy-6-methoxycoumarin), esculetin (6,7-dihydroxycoumarin), umbelliferone (7-hydroxycoumarin), fraxetin (7,8 dihydroxy-6-methoxycoumarin), sideretin (5,7,8-trihydroxy-6-methoxycoumarin) and their respective glycosylated forms–scopolin, esculetin, skimmin, fraxin and sideretin-glycoside, respectively [1]. Figure 1 presents the semi-developed formula of simple coumarins and their glycosides derivatives identified in this research.

**Citation:** Perkowska, I.; Siwinska, J.; Olry, A.; Grosjean, J.; Hehn, A.; Bourgaud, F.; Lojkowska, E.; Ihnatowicz, A. Identification and Quantification of Coumarins by UHPLC-MS in *Arabidopsis thaliana* Natural Populations. *Molecules* **2021**, *26*, 1804. https://doi.org/10.3390/ molecules26061804

Academic Editor: Maria João Matos

Received: 27 February 2021 Accepted: 18 March 2021 Published: 23 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

**Figure 1.** Chemical structures of simple coumarins and their glycosides analysed in this work (www.chem-space.com (accessed on 20 January 2021)).

– – Coumarins have been recognized for many years as an important class of pharmacologically active compounds. They have anticoagulant, anticancer, antiviral and antiinflammatory properties [2]. In addition to the listed medicinal benefits, it was shown recently in numerous studies that coumarins play an important role in iron (Fe) homeostasis, oxidative stress response, plant-microbe interactions and that they can act as signalling molecules in plants [3–8]. In the last few years, an increasing number of reports concern the analysis of root extracts and root exudates that are rich in phenolic compounds, such as simple coumarins, which mediate multiple interactions in the rhizosphere. Coumarins were shown to have a strong impact on the plant interactions with microorganisms and play a crucial role in nutrient acquisition [6,9–18]. Moreover, the root-secreted scopoletin was proved to exert a selective antimicrobial action in the rhizosphere [8]. These numerous reports examining the biochemical and physiological functions of coumarins, make this class of specialized metabolites extremely interesting from a scientific and commercial point of view. The vast majority of these studies were performed using a reference accession, Col-0, of the model plant *Arabidopsis thaliana* (hereinafter Arabidopsis), and its mutants defective at various steps of coumarin biosynthesis.

– – Here, we conducted the qualitative and quantitative assessment of coumarin content in leaf and root tissue of a set of Arabidopsis natural populations (accessions). Numerous studies on primary and specialized metabolites profiling were conducted using the Arabidopsis model system [19–28]. Previously, due to the importance of coumarins for human health, most research on their metabolic profiling were carried out on plants of economic importance, such as e.g., sweet potato (*Ipomoea batatas* L.), rue (*Ruta graveolens* L.) or lettuce (*Lactuca sativa* L.) [29–34]. One of the first metabolic profiling of root exudates using Arabidopsis natural populations (Col-0, C24, Cvi-0, Ler) was made by Micallef et al. [35] who attempted to correlate them with the compositions of rhizobacterial communities. However, the authors of this work did not undertake the qualitative and quantitative evaluation of the isolated compounds. Consequently, they could only conclude that there are differences between Arabidopsis accessions in terms of the quality and quantity of released substances, which may have an impact on the composition of the rhizobacterial communities.

So far, only a few more studies focusing on the accumulation of coumarins in Arabidopsis natural populations have been published. As shown by our group [36], a significant natural variation in the accumulation of coumarins is present among the roots of Arabidopsis accessions. Using HPLC and GC-MS analytical methods, we identified and quantified coumarins in the roots of selected seven accessions-Antwerpen (An-1, Belgium),

Columbia (Col-0, Germany), Estland (Est-1, Estonia), Kashmir (Kas-2, India), Kondara (Kond, Tajikistan), Landsberg *erecta* (Ler, Poland) and Tsu (Tsu-1, Japan). Subsequently, we conducted a QTL mapping and identified new loci possibly underlying the observed variation in scopoletin and scopolin accumulation. Thereby, we demonstrated that Arabidopsis natural variation is an attractive tool for elucidating the basis of coumarin biosynthesis. Other studies focusing on differential accumulation of coumarins between Arabidopsis accessions were conducted by Mönchgesang et al. [14]. A non-targeted metabolite profiling of root exudates revealed the existence of distinct metabolic phenotypes for 19 Arabidopsis accessions. Scopoletin and its glycosides were among phenylpropanoids that differed in the exudates of tested accessions. This research group also focused on the plant-to-plant variability in root metabolite profiles of 19 Arabidopsis accessions [15]. In the current study, a larger set of 28 accessions was chosen, that represent a wide genetic variation existing in Arabidopsis. To increase the scope of this work, we extracted and quantified coumarins from two different types of tissue–roots and leaves. The latter one, in the light of our best knowledge, have never been tested for coumarin content using Arabidopsis natural variation. We believe our results will be beneficial for further studies focusing on a better understanding of coumarin physiological functions and the exact role of enzymes involved in their biosynthesis.

#### **2. Results**

#### *2.1. UHPLC-MS Targeted Metabolite Profiling of Root and Leaf Tissues Reveals Distinct Metabolic Phenotypes for 28 Arabidopsis Accessions*

The average content of each tested coumarins (Table 1) grouped by the 28 Arabidopsis accessions (Table 2), type of tissue (extracts from roots and leaves) and method of preparing extracts for analysis (without and after hydrolysis) were depicted through a general heatmap. For each compound, we quantified both the non-glycosylated coumarins scopoletin (Figure 2A), umbelliferone (Figure 3A), esculetin (Figure 4A), and their respective glycosylated forms—scopolin (Figure 2B), skimmin (Figure 3B), and esculin (Figure 4B), respectively. The concentration µM was based on the fresh weight (FW).


**Table 1.** Coumarins and their glycosides identified in this study.

Our analyses made evidence a significant variation in accumulation of all tested compounds between Arabidopsis accessions, both in roots and leaves. In accordance with the current state of knowledge [6,8,36–39], we identified the coumarin scopoletin and its glycoside, scopolin, to be the major metabolites that accumulate in Arabidopsis roots (Figure 2A,B).


**Table 2.** Basic information on the Arabidopsis accessions used in this study.

Scopoletin was the most abundant compound in each of the 28 Arabidopsis accessions studied, especially in the roots (from 2.61 to 151.90 µM), but interestingly this phytochemical was also detected in the leaf extracts (Table S1). The highest amount of scopoletin was detected in Bay-0, Br-0 and Kondara, respectively (Figure 2A), in samples prepared from the roots and subjected to hydrolysis. In non-hydrolyzed root samples, the highest content of scopoletin was detected for the same accessions–Kondara, Bay-0 and Br-0. As expected, scopoletin content in the leaf extracts was several dozen times smaller (from 0.03 to 2.6 µM) when compared with extracts prepared from the root tissue. Amount of scopoletin in the leaf sample was the highest in Br-0, Est-1 and Bay-0 when subjected to hydrolysis and in Est-1, Br-0, Col-0 and Bay-0 when not hydrolyzed.

Relatively large amounts of scopolin (from 2.94 to 67.26 µM) were found in almost all root extracts that were not subjected to hydrolysis (Figure 2B), the highest in Br-0, Fei-0, Ga-0, Leb 3/4, Ri-0, C24 and Bay-0 accessions (Table S1). We also identified some accessions (Fuk-1, Bay-0, Ri-0, Sha-1 and Eri-1) with relatively high content of scopolin in root extracts after hydrolysis (from 2.04 to 8.09 µM), most probably due to non-effective enzymatic reaction. Interestingly, another set of accessions (Est-1, Sorbo, Bay-0, Kyo-0, No-0, Ga-0, Fei-0, Ws-0) with relatively high scopolin concentration (from 2.58 to 6.48 µM) was also detected in leaf extracts not subjected to hydrolysis. As could be expected, in hydrolysed leaf samples in which sugar residues were cut off and most of scopolin was transformed into scopoletin, the amounts of scopolin were quite low or close to the LOQ.

**Figure 2.** Heat maps based on the average (**A**) scopoletin and (**B**) scopolin concentration (µM/FW) in Arabidopsis tissue extracts from roots and leaves, without and after hydrolysis. The values used in the plots (https://app.displayr.com (accessed on 20 January 2021)) are the mean of 3 biological replicates. The mean values and standard deviations (±SD) are gathered in the supplementary materials (Table S1).

Interestingly, in this study, we identified small amounts of umbelliferone (from 0.02 to 1.64 µM) in Arabidopsis plants (Figure 3A). Importantly, we detected this phytochemical in all of the hydrolyzed root extracts (Table S2). The highest levels of umbelliferone were found in Bay-0, Ri-0, Est-1, Col-0, Br-0, C24, Sorbo, Fuk-1 and Leb 3/4, respectively. The quantity of umbelliferone in all leaf extract samples was below LOQ.

Skimmin, which is a glucoside of umbelliferone, was detected and quantified (from 0.69 to 19.80 µM) mostly in root samples of Arabidopsis accessions that were not subjected to enzymatic hydrolysis (Figure 3B). The highest levels were detected in extracts originating from Ga-0, C24, Ws-0, Ri-0, Est-1, Kyo-0 and Eri-0 accessions. It should be noted that skimmin could also be quantified (concentration from 0.04 to 18.76 µM) in all hydrolyzed root extract samples (Table S2), which needs further investigation.

Most of the results obtained for the leaf tissues were very low and near the LOQ, however in Eri-1, An-1, C24, Col-0, Van-0, Kondara, Ws-0, Ga-0, Fuk-1, Can-0 and Tsu-1 accessions, we observed values slightly above the limit.

**Figure 3.** Heat maps based on the average (**A**) umbelliferone and (**B**) skimmin concentration (µM/ FW) in Arabidopsis tissue extracts from roots and leaves, without and after hydrolysis. The values used in the plots (https://app.displayr.com (accessed on 20 January 2021)) are the mean of 3 biological replicates. The mean values and standard deviations (±SD) are gathered in the supplementary materials (Table S2).

Small amounts of esculetin were detected only in a few of root extracts (max. concentration 0.29 µM) and leaf samples (max. concentration 0.16 µM) (Figure 4A). In root non-hydrolysed samples, esculetin was present in Bay-0, Br-0 and Can-0 accessions, while in hydrolysed extracts it was detected in Can-0, Bay-0, Col-0, Ri-0 and Tsu-1 (Table S1). It may be puzzling that in some accessions, esculetin was only detected in samples which were not subjected to hydrolysis but not in the hydrolysed ones. This is the case for the root extract of Br-0 (0.07 µM), and leaf samples of C24, Br-0, An-1, Col-0, No-0, Ws-0 and Ri-0 accessions (from 0.01 to 0.16 µM, Table S3). In leaf samples after hydrolysis, only trace amount of esculetin was detected in Ws-0.

Esculin, which is a glycoside form of esculetin, was not found in any root extract (Figure 4B), except Col-0 sample with quantity near to LOQ (0.01 µM). Trace amounts of esculin were detected in some leaf extracts without hydrolysis (from 0.01 to 0.15 µM) with the highest content in Col-0 accession, and in the leaf samples subjected to hydrolysis (from 0.01 to 0.36 uM). Here, the highest esculin content was detected in Ws-0 accession (Table S3).

#### *2.2. Principal Component Analysis (PCA) for 28 Arabidopsis Accessions Using Coumarin Quantification by UHPLC-MS in Selected Geographic and under Diverse Climatic Factors*

In order to compare and visualize the possible relationship between coumarin content variability present among 28 Arabidopsis accessions in selected geographic and in various climatic factors (maximal altitude [m], average winter minimal temperature [ ◦C], average summer maximal temperature [ ◦C] and average annual precipitation [mm]), we performed Principal Component Analysis (PCA). About half of the variance of used dataset was covered by the first two principal components, explaining 49% of the overall data variance (27.1% and 21.9% for PC1 and PC2, respectively) (Figure 5A).

According to the results presented on the Variables-PCA plot (Figure 5B), we assumed that there is a positive correlation between scopoletin, umbelliferone and scopolin concentration in root samples before hydrolysis. Despite the fact that scopolin content has relatively small contribution in explaining the variability between tested accessions, it can be also positively correlated with skimmin and umbelliferone concentration. Skimmin content is positively correlated with annual precipitation data.

**Figure 5.** (**A**) Principal component analysis (PCA) for 28 Arabidopsis accessions using the concentration of umbelliferone, scopoletin and their corresponding glycosides (skimmin and scopolin, respectively) in root samples without hydrolysis, and four geographic and climatic factors (maximal altitude [m], average winter minimal temperature [ ◦C], average summer maximal temperature [ ◦C] and average annual precipitation [mm]; Table S4). Factor coordinates are marked with arrows. Observations indicated by blue accession names represent European locations (n = 14), green represent Asian locations (n = 9), violet represent North American locations (n = 3) and red represent African locations (n = 2). The abbreviations indicate the accessions according to Table 2. Component one and two explain 49% of the point variability. (**B**) The Variables-PCA contribution plot shows the correlation of the variables used in PCA with the respective contribution of each factor (contrib) indicated with a colour gradient. (**C**) The scree plot/graph of variables demonstrate the percentage of variability explained by each dimension (PC). Principal Component 1 and 2 explain 27.1% and 21.9% of the variance respectively.

A negative correlation is highlighted between the following variables: (1) umbelliferone concentration and temperatures (average winter minimal temperature and average summer maximal temperature); (2) scopolin concentration and temperatures (average

winter minimal temperature and average summer maximal temperature); (3) skimmin concentration and average summer maximal temperature, as well as skimmin concentration and maximal altitude; (4) scopoletin concentration and annual precipitation, as well as scopoletin concentration and average winter minimal temperature (Figure 5B).

On the Figure 5C we demonstrate the graph of variables (scree plot) which indicates the percentage of variability explained by each dimension (PC). Principal Component 1 and 2 explain 27.1% and 21.9% of the variance respectively, while the other 6 dimensions account for the total remaining variability between each accession (PC3 = 14.7%, PC4 = 13.8%, PC5 = 9.4%, PC6 = 6.2%, PC7 = 3.8% and PC8 = 3%).

#### **3. Discussion**

Our previous study strongly suggest that Arabidopsis is an excellent model for elucidating the basis of natural variation in coumarin accumulation [36]. Here, we identified and quantified a set of coumarin compounds in the root and leaf methanol extracts prepared from 28 Arabidopsis accessions. In the light of our best knowledge, it is the largest set of Arabidopsis natural populations used in the coumarin profiling analysis that should well represent a wide genetic variation existing in this model plant. It is assumed that these accessions reflect genetic adaptation to local environmental factors [40]. As a result of evolutionary pressure differentially acting on the studied accessions originating from various geographical locations, a large number of genetic polymorphisms is present that have led to different levels of expression of genes involved in the biosynthesis, transport and metabolism of coumarins, and ultimately to different levels of their accumulation. In the current work, we detected a significant natural variation in the content of simple coumarins present in the root and leaf extracts of 28 Arabidopsis accessions. Among tested compounds, scopoletin and its glycosylated form, scopolin, were the most abundant, which is in line with the current state of knowledge [6,14,15,36–38].

The previous study on differential accumulation of coumarins between 19 Arabidopsis accessions belonging to the MAGIC lines characterized by a high genetic variability [14], confirm the hypothesis that the composition of Arabidopsis accessions root exudates is genetically determined. They revealed the existence of distinct root metabolic phenotypes among tested natural populations, including variation in the accumulation of scopoletin and its glycoside. Another study focused on extensive profiling of specialized metabolites in root exudates of Arabidopsis reference accession, Col-0, by non-targeted metabolite profiling using reversed-phase UPLC/ESI-QTOFMS [16]. As many as 103 compounds were detected in exudates of hydroponically grown Col-0 plants. Among them, 42 were identified by authenticated standards, including the following coumarins: esculetin, scopoletin, and their glucosides esculin and scopolin. In addition to these coumarins, further esculetin and scopoletin conjugates were initially identified in the root exudates based on their mass spectral fragmentation pattern [16].

It has to be noted that among other coumarin compounds identified in our study, we detected umbelliferone for the first time in Arabidopsis model plant. Authors of the first publication on the accumulation of coumarins in Arabidopsis [37], in which various type of tissues (roots, shoots and callus) were tested, detected trace amounts of skimmin (umbelliferone glucoside) in the wild type plants and slightly increased skimmin level in mutants of CYP98A3. No umbelliferone was detected in that study, or in any other work with Arabidopsis to date, in the light of our best knowledge. It cannot be excluded that we were able to detect umbelliferone due to the sample types tested. We conducted coumarin profiling of extracts prepared from the plant tissues grown in in vitro liquid cultures. Moreover, umbelliferone was detected in root methanol extracts additionally subjected to enzymatic hydrolysis prior to quantification done by UHPLC-MS in order to hydrolyze the glycoside forms of coumarins, while its glycosylated form, skimmin, was detected in samples without enzymatic treatment. It should be highlighted that low amounts of umbelliferone were also detected in non-hydrolysed extracts.

Coumarins have become important players both in optimizing Fe uptake and shaping the root microbiome, thus affecting plant health [5,41]. The link between plant specialized metabolites, in particular coumarins, nutrient deficiencies and microbiome composition that was discovered in recent studies [7,8,42,43], could provide a new set of tools for rationally manipulating the plant microbiome [44]. The selection of underground tissue was an obvious choice in such analyses, considering that coumarins are essentially synthesized in roots where optimization of Fe uptake is coordinated with plant requirements and interaction with soil microorganisms. Therefore, most of the previous coumarin metabolic profiling analysis, including functional characterization of Arabidopsis mutants defective in genes encoding enzymes involved in coumarin biosynthesis or transport, were performed using the root exudates and root tissue [3,6,12,17,38,44,45]. It was also the case in research conducted on the effects of Fe, phosphorus (Pi) or both deficiencies on coumarin profiles in the root tissue of several T-DNA insertional mutants defective in genes involved in Pi or Fe homeostasis [11]. Importantly, in the current study we detected variation in accumulation of esculetin and esculin. The latter one was identified in Arabidopsis leaf extracts, both with and without enzymatic hydrolysis. This requires further research and is of particular interest in the light of recent research findings on coumarin cellular localization, trafficking and signalling [5,7]. Coumarins were found to be involved in the plant response to pathogens in aerial tissues [41,46] and proposed to play an important signalling role in bidirectional chemical communication along the microbiome-root-shoot axis [7].

The study of natural variation in coumarin content present among Arabidopsis accessions is a starting point in elucidating direct links between metabolic phenotypes and genotypes. In the presented research, we also checked whether the climatic and geographic data on the regions from which particular accessions originate, are correlated with the concentrations of tested coumarins. The conducted PCA showed a number of positive and negative correlations between climatic factors and coumarin content. Further investigation is needed to draw a more precise conclusion about possible relationship between the accumulation of coumarins and habitat data. Taking into account, the recent studies showing an important role of coumarins in plant interactions with soil microorganisms and nutrient acquisition, a more in-depth analysis, including data on soil parameters at the origin sites of a given accession, would explain the greater variance and give us more information on the potential correlations. It will be beneficial for the future discovery of physiological mechanisms of action of various alleles involved in the coumarin biosynthesis and can help to select biosynthetic enzymes for further metabolic engineering research.

#### **4. Materials and Methods**

#### *4.1. Chemicals and Reagents*

The coumarins standards umbelliferone (purity ≥ 99%), coumarin (>99% purity), esculin (≥98% purity) were purchased from Sigma-Aldrich (St. Louis, MO, USA), scopoletin (>95% purity) and esculetin (>98% purity) from Extrasynthese (Genay, France), skimmin (98% purity) from Biopurify Phytochemicals (Chengdu, China), scopolin (>98% purity) from Chemicals Aktin Inc. (Chengdu, China). Stock solutions of each standard at a concentration of 10 mmol/L were prepared by diluting the powder in dimethyl sulfoxide (Fisher scientific, Illkirch, France) and kept at −18 ◦C until use. HPLC-grade methanol was purchased from CarloErba Reagents (Val de Reuil, France), formic acid was purchased from Fisher Scientific (Illkirch, France). Water was purified by a PURELAB Ultra system (Veolia Water S.T.I., Antony, France).

#### *4.2. Plant Material*

All seeds of the 28 Arabidopsis accessions (Table 2) from various habitats which were used in this study were obtained courtesy of prof. Maarten Koornneef.

#### *4.3. In Vitro Plant Culture*

All the Arabidopsis accessions seeds were surface sterilized with 70% ethanol for 2 min, 5% calcium hypochlorite solution for 8 min and then washed 3 times with sterile ultrapure water. The seeds were placed in Petri dishes with <sup>1</sup> <sup>2</sup> Murashige-Skoog (MS) medium solidified with agar (Sigma-Aldrich) for in vitro plant culture and incubated for 72 h in the dark at 4 ◦C. Then the plates were transferred to a growth chamber (daily cycle: 16 h light 35 µmol m−<sup>2</sup> s −1 temperature 20 ◦C and 8 h dark temperature 18 ◦C) for 10 days. After that time, seedlings were transferred from agar plates into 200 mL flasks (three individuals per flask) containing 5 mL of <sup>1</sup> <sup>2</sup> MS liquid medium containing 1% sucrose, MS salts, 100 mg/L myo-inositol, 1 mg/L thiamine hydrochloride, 0.5 mg/L pyridoxine hydrochloride and 0.5 mg/L nicotinic acid (Sigma-Aldrich/Merck KGaA, Darmstadt, Germany). Plants were grown in the growth chamber on a rotary platform with shaking 120 rpm. After one week, 3 mL of fresh <sup>1</sup> <sup>2</sup> MS medium was added. Plants were grown for 17 supplementary days and after that time were rinsed with demineralized water, dried on paper towels. Roots and leaves samples were weighted (50 ± 2 mg fresh weight (FW)) and frozen in liquid nitrogen. The plant material was stored in a freezer at −80 ◦C until extraction process. All accessions were grown in three biological replicates (in three independent flasks, three seedlings per flask).

#### *4.4. Metabolites Extraction*

For the metabolites extraction, plant tissue frozen in liquid nitrogen was grinded by the usage of 5 mm diameter stainless steel beads (Qiagen, Hilden, Germany). To the 2 mL microtubes, 2 clean beads were added and samples were frozen in liquid nitrogen. Then, using vortex, samples were mixed. For the better performance, the freezing and vortexing procedure was repeated several times until all tissue was powdered. To the powdered tissues 0.5 mL of 80% methanol containing 5 µM 4-methylumbelliferone as an internal standard was added. After that, samples were sonicated for 30 min with ultrasonic cleaner (Proclean 3.0DSP, Ulsonix, Expondo, Berlin, Germany) (70% frequency, sweep function) and incubated in 4 ◦C in darkness for 24 h. Next day, all samples were vortexed, centrifuged at 13,000× *g* for 10 min and the supernatant was transferred into new microtubes. Centrifugation was repeated in order to get rid of any sediment. The extracts were firstly dried for 2 h in an incubator at 45 ◦C and then, for the next 2 h in a vacuum centrifuge (Savant SpeedVac vacuum concentrator, Thermo Fisher Scientific, Waltham, MA, USA). To the dried extracts 100 µL of 80% methanol was added to dissolve samples during the night at 4 ◦C. Then the extracts were vortexed for 10 min and separated by 50 µL. One of the replicates was subjected to enzymatic hydrolysis, and the second one was stored at −20 ◦C until UHPLC-MS analysis (Shimadzu Corp., Kyoto, Japan).

#### *4.5. Enzymatic Hydrolysis*

The enzymatic hydrolysis was performed according to Nguyen et al. [47]. Methanolic extracts were subjected to enzymatic hydrolysis with a β-glucosidase (Fluka Chemie GmbH, Buchs, Switzerland) in 0.1 M acetate buffer at a concentration of 0.5 mg/mL in order to determine the amounts of glycosylated compounds (o-glycosides). 50 µL of acetate buffer with β-glucosidase at pH 5.0 (0.1 M sodium acetate, 0.1 M acetic acid and 0.5 mg/mL β-glucosidase buffer) was added to 50 µL of the prepared extract and incubated for 22 h at 37 ◦C. The reaction was stopped by adding 100 µL of 96% ethanol to the reaction mixture. The extracts were dried in an incubator at 45 ◦C for 2 h and then for about 1 h in a vacuum centrifuge (Savant SpeedVac vacuum concentrator). The obtained extract was dissolved in 50 µL of 80% methanol overnight and stored at −20 ◦C until UHPLC-MS analysis.

#### *4.6. UHPLC Separation*

The coumarins analyses were performed using a NEXERA UPLC-MS system (Shimadzu Corp., Kyoto, Japan) equipped with two UHPLC pumps (LC-30AD), an automatic sampler (SIL-30AC), a photodiode array detector (PDA, SPDM-20A) and combined with

a mass spectrometer (single quadrupole, LCMS-2020). Coumarins separation was done on a C18 reversed phase column (ZORBAX Eclipse Plus), 150 × 2.1 mm, 1.8 µm (Agilent Technologies, Santa Clara, CA, USA) protected with an Agilent Technologies 1290 Infinity filter. The column was kept at 40 ◦C in a column oven (Shimadzu CTO-20AC). Mobile phase consisted of 0.1% formic acid in ultrapure water (buffer A) and 0.1% formic acid in methanol (buffer B) at a constant flow rate of 200 µL/min. The linear gradient solvent system was set as follows: 0 min, 10% B; 16 min, 70% B; 18 min, 99% B; 18.01 min, 10% B; 20 min, 10% B. The total analysis duration was 20 min. The injection volume was 5 µL.

#### *4.7. MS Detection*

The UHPLC system was connected to the MS by an electrospray ionization source (ESI), operating in positive mode (ESI+) and scanning in single ion monitoring mode (SIM). The inlet, desolvation line and heating block temperatures were set at 350 ◦C, 250 ◦C, and 400 ◦C, respectively. The capillary voltage was set at 4.5 kV. Dry gas flow was set at 15 L/min and nebulizing gas at 1.5 L/min. The instrument was operated and data were processed using LabSolution software version 5.52 sp2 (Shimadzu Corp., Kyoto, Japan).

#### *4.8. Peak Identification and Quantitation*

Each standard molecule was individually injected in the UHPLC-MS in full scan mode to determine retention time and *m/z* ratio for the analysis. The quantitation of each molecule (Table 2) was based on the signal obtained from the MS detection, using angelicin, as an analytical internal standard. Angelicin was added at the same concentration (5 µM) in all the samples before injection as well as in 7 calibration solutions. The calibration solutions contained all of the standard molecules at the same concentrations ranging from 0.1 to 10 µM (0.1, 0.2, 0.5, 1, 2, 5 and 10 µM). Calibration curves were drawn for each compound by linking its relative peak area (compound area divided by the angelicin area) and its concentration. Each curve fit type was linear. The limit of quantitation (LOQ) was calculated as the analyte concentration giving signal to signal to noise ratios (S/N) of 10. Three measurements were assessed per accession.

#### *4.9. Principal Component Analysis (PCA)*

Principal Component Analysis (PCA) were performed using *prcomp()* package and visualize with the *factoextra* 1.0.7 version package in the R 4.0.4 environment developed by the R Core Team (2020). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria (www.R-project.org (accessed on 20 January 2021)) and the RStudio Team (2019). RStudio: Integrated Development for R. RStudio, Inc., Boston, MA, USA (www.rstudio.com (accessed on 20 January 2021)). All the variables were standardized before analysis. Data used for the analysis are presented in Table S4.

#### **5. Conclusions**

Multi-pharmacological properties of coumarins that are widely used in medical applications, make the study of coumarin biosynthesis attractive from the commercial point of view. Considering that all medicinal plants currently used in studying the biosynthesis of coumarins are non-model organisms and many approaches are not available in those species, makes a model plant Arabidopsis, with its extensive genetic variation and numerous publicly accessible web-based databases, an excellent model to study accumulation of coumarins in natural populations. The presented results focusing on qualitative and quantitative characterization of natural resources provide a basis for further research on identification of genetic variants involved in coumarin biosynthesis in plants, which is the first step in metabolic engineering for the production of natural compounds. We identified scopoletin, and its glycosylated form, scopolin, to be the most abundant coumarins in Arabidopsis tissues. It should be emphasized that among other coumarin compounds identified in this study, we detected umbelliferone for the first time in Arabidopsis. In

view of the considerable importance of umbelliferone in synthesis and its pharmacological properties, this is a significant step in the study of biosynthesis of coumarins using this model plant.

**Supplementary Materials:** The following are available online, Table S1: The mean values (scopoletin and scopolin concentrations) and standard deviations (±SD) that were used in making heat maps shown in Figure 2, Table S2: The mean values (scopoletin and scopolin concentrations) and standard deviations (±SD) that were used in making heat maps shown in Figure 3, Table S3: The mean values (scopoletin and scopolin concentrations) and standard deviations (±SD) that were used in making heat maps shown in Figure 4, Table S4. Coumarin concentrations in root samples before hydrolysis and four geographic and climatic factors used in principal component analysis (PCA).

**Author Contributions:** Conceptualization, A.O., A.H. and A.I.; methodology, A.O., A.H. and A.I.; validation, I.P., J.S. and A.O.; investigation, I.P., J.S., A.O. and J.G.; writing—original draft preparation, I.P., E.L. and A.I.; writing—review and editing, E.L., A.O., A.H. and A.I.; supervision, E.L., F.B. and A.I.; funding acquisition, A.I. and A.O. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Narodowe Centrum Nauki (Polish National Science Centre) grant number 2014/15/B/NZ2/01073 and Narodowa Agencja Wymiany Akademickiej (Polish National Agency for Academic Exchange) grant number PPN/BFR/2019/1/00050 to A.I.; and by Campus France (PHC POLONIUM) grant number 45026ZE to A.O.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study are available within the article and supplementary materials [Tables S1–S4].

**Acknowledgments:** We thank Maarten Koornneef for providing all Arabidopsis seeds used in this study and Thibaut Duval for technical support.

**Conflicts of Interest:** The authors declare no conflict of interest.

**Sample Availability:** Coumarin standards used in the study and seeds corresponding to each accessions are available from the authors.

#### **References**


#### *Article*
