**4-Hydroxy-7-Methoxycoumarin Inhibits Inflammation in LPS-activated RAW264.7 Macrophages by Suppressing NF-**κ**B and MAPK Activation**

#### **Jin Kyu Kang and Chang-Gu Hyun \***

Jeju Inside Agency and Cosmetic Science Center, Department of Chemistry and Cosmetics, Jeju National University, Jeju 63243, Korea; wlsrbtjsrb@naver.com

**\*** Correspondence: cghyun@jejunu.ac.kr; Tel.: +82-64-754-3542

Academic Editor: Maria João Matos Received: 14 August 2020; Accepted: 24 September 2020; Published: 26 September 2020

**Abstract:** Coumarins are natural products with promising pharmacological activities owing to their anti-inflammatory, antioxidant, antiviral, anti-diabetic, and antimicrobial effects. Coumarins are present in many plants and microorganisms and have been widely used as complementary and alternative medicines. To date, the pharmacological efficacy of 4-hydroxy-7-methoxycoumarin (4H-7MTC) has not been reported yet. Therefore, in this study, we investigated the anti-inflammatory effects of 4H-7MTC in LPS-stimulated RAW264.7 cells as well as its mechanisms of action. Cells were treated with various concentrations of 4H-7MTC (0.3, 0.6, 0.9, and 1.2 mM) and 40 µM L-N<sup>6</sup> -(1-iminoethyl)-L-lysine (L-NIL) were used as controls. LPS-stimulated RAW264.7 cells showed that 4H-7MTC significantly reduced nitric oxide (NO) and prostaglandin E<sup>2</sup> (PGE2) production without cytotoxic effects. In addition, 4H-7MTC strongly decreased the expression of inducible nitric oxide synthase (iNOS) and cyclooxygenase (COX-2). Furthermore, 4H-7MTC reduced the production of proinflammatory cytokines such as tumor necrosis factor (TNF)-α, interleukin (IL)-1β, and IL-6. We also found that 4H-7MTC strongly exerted its anti-inflammatory actions by downregulating nuclear factor kappa B (NF-κB) activation by suppressing inhibitor of nuclear factor kappa B alpha (IκBα) degradation in macrophages. Moreover, 4H-7MTC decreased phosphorylation of extracellular signal-regulated kinase (ERK1/2) and c-Jun N-terminal kinase/stress-activated protein kinase (JNK), but not that of p38 MAPK. These results suggest that 4H-7MTC may be a good candidate for the treatment or prevention of inflammatory diseases such as dermatitis, psoriasis, and arthritis. Ultimately, this is the first report describing the effective anti-inflammatory activity of 4H-7MTC.

**Keywords:** 4-hydroxy-7-methoxycoumarin; macrophage; inflammation; NF-κB; MAPK

#### **1. Introduction**

Coumarins (benzo-α-pyrones) are oxygen heterocycles that are naturally occurring benzopyrene derivatives which have been identified in plants, bacteria, and fungi [1]. Coumarins represent a broad family of secondary metabolites that are found naturally in over 1300 plant species. The main pathway of coumarin biosynthesis occurs through the shikimic acid pathway, which involves cinnamic acid and phenylalanine metabolism [2,3]. Natural coumarins are subdivided into several classes according to their chemical diversity and complexity, namely, simple coumarins, isocoumarins, furanocoumarins, pyranocoumarins (both angular and linear), biscoumarins, and phenylcoumarins [4].

Coumarins have several desirable features. First, they have a low molecular weight owing to their simple structures. Second, they have high solubility in most organic solvents. Third, they have high bioavailability and low toxicity. Fourth, they have various pharmacological effects such as anticoagulant,

antimicrobial, anti-inflammatory, neuroprotective, antidiabetic, anticonvulsant, and antiproliferative activities [4–6]. These characteristics and advantages support their roles as lead compounds in drug research and development [7]. Coumarins have diverse structures owing to the different types of substitutions in their underlying structures, which can affect biological activity. Thus, the structuresystem-activity-relationship of coumarin must be carefully studied [1].

During our ongoing screening program designed to identify modulators of skin inflammation and melanogenesis from coumarin and its derivatives, we reported that 8-methoxycoumarin increased melanogenesis via the MAPK signaling pathway [8]. In addition, we identified that auraptene, the most abundant naturally occurring geranyloxycoumarin, possesses anti-melanogenic activity through ERK-mediated MITF downregulation [9]. Furthermore, we reported that 7,8-dimethoxycoumarin stimulates melanogenesis via MAPK-mediated MITF upregulation and attenuates the expression of IL-6, IL-8, and CCL2/MCP-1 in TNF-α-treated HaCaT cells [10,11]. α

As an extension of this study, we investigated the anti-inflammatory effects of 4-hydroxy-7-methoxycoumarin (4H-7MTC, Figure 1). 4H-7MTC belongs to a class of organic compounds known as hydroxycoumarins. These are coumarins that contain one or more hydroxyl groups attached to the coumarin skeleton. 4H-7MTC can be found in plants such as coriander, artichoke, Tibetan hulless barley, and eggplant [12,13]. To the best of our knowledge, no studies have reported the pharmacological and biochemical properties and therapeutic applications of 4H-7MTC. Therefore, in this study, we investigated whether 4H-7MTC has anti-inflammatory effects; an initial step in the development of 4H-7MTC as a functional compound for use in human health applications.

**Figure 1.** Structures of 4-hydroxycoumarins: 4-hydroxy-7-methoxycoumarin (**a**), 4-hydroxy-6-methylcoumarin (**b**), and 4-hydroxy-7-methylcoumarin (**c**).

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

α β Macrophages, the main cells responsible for innate immunity, are activated by the invasion of foreign pathogens such as parasites, bacteria, and viruses, or by stimulation with external signals. In particular, lipopolysaccharide (LPS), an endotoxin produced by Gram-negative bacteria, stimulates macrophages, which in turn promotes secretion of proinflammatory cytokines, including tumor necrosis factor (TNF)-α, interleukin (IL)-1β, and IL-6, and induces the expression of inflammatory response factors such as nitric oxide (NO) and prostaglandin E<sup>2</sup> (PGE2) [14,15]. As such, regulation of the production of NO and proinflammatory cytokines in macrophages is a current research topic for the development of new anti-inflammatory agents, and there have been many attempts to derive new anti-inflammatory agents from natural compounds [16–18].

To demonstrate the anti-inflammatory activity of the three types of 4-hydroxycoumarin, including 4H-7MTC, we first assessed its ability to inhibit NO production in LPS-stimulated macrophage RAW264.7 cells (Figure 1). RAW264.7 cells were treated with various concentrations of 4-hydroxycoumarins, and cell viability was measured using the MTT assay. As shown in Figure 2, NO production increased by 3.43- to 15-fold in LPS-activated macrophages relative to untreated macrophages. Moreover, 4-hydroxycoumarins reduced LPS-induced NO production in a concentration-dependent manner. At 0.6 mM concentration of 4H-7MTC, the production of NO by LPS-treated macrophages decreased by 23.10%. At 0.5 mM concentration of 4H-6MC and 4H-7MC, the production of NO by LPS-treated macrophages decreased by 21.27% and 17.61%, respectively. These results show that the 4-hydroxy structure of coumarin influences the degree of inhibition of NO production, and the substituents on carbon 6 and 7 of the B-ring structure had little effect on the inhibition of NO production. No concentration of 4-hydroxycoumarins displayed significant cytotoxicity, indicating that the anti-inflammatory effects of 4-hydroxycoumarins were not attributable to cytotoxicity. Among them, we found that 4H-7MTC is a safe substance that does not induce cytotoxicity even at concentrations as high as 1.2 mM.

μ μ **Figure 2.** Effect of 4H-7MTC (**a**), 4H-6MC (**b**), and 4H-7MC (**c**), on nitric oxide production in LPS-stimulated RAW264.7 cells. The cells were plated in 24-well plates (1.5 × 10<sup>5</sup> cells/well), incubated for 24 h, and then pretreated with 4H-7MTC (0.3, 0.6, 0.9, 1.2, and 1.5 mM), 4H-6MC (100, 200, 300, 400, and 500 µM), and 4H-7MC (100, 200, 300, 400, and 500 µM) for 1 h, followed by LPS stimulation for 24 h. Cytotoxicity of 4H-7MTC, 4H-6MC, and 4H-7MC were evaluated using MTT assay. Nitric oxide production was determined by the Griess reagent method. L-N6-(1-Iminoethyl) lysine dihydrochloride (L-NIL) was used as a positive control. The data are presented as mean ± SD. Statistical significance was assessed by one-way analysis of variance (ANOVA), followed by Tukey's post-hoc test and represented as follows: # *p* < 0.05, ### *p* < 0.005, \*\* *p* < 0.01, \*\*\* *p* < 0.001 vs. LPS alone.

μ To investigate the additional functionalities of 4H-7MTC, which was confirmed to be safe at high concentrations, we aimed to evaluate its potential activity as an anticancer agent or as a preventive of gray hair. As shown in Figure 3a, 4H-7MTC upregulated melanin production in a concentration-dependent manner over a wide concentration range (25–200 µM), without any observed cytotoxicity. Additionally, 4H-7MTC showed no cytotoxicity up to 1.2 mM in normal macrophages, whereas it exhibited a cytotoxic effect on B16F10 melanoma cells at a low concentration of 0.3 mM (Figure 3b). This suggests that 4H-7MTC could be a potential anticancer agent.

α μ α α **Figure 3.** Effect of 4H-7MTC on the production of melanin (**a**) in α-MSH-stimulated B16F10 cells and Cytotoxicity of 4H-7MTC in B16F10 cells (**b**). Cells were plated in 60 mm cell culture dish (6.0 × 10<sup>4</sup> cells/dish), incubated for 24 h, and then treated with 4H-7MTC (25, 50, 100, 150 and 200 µM) for 72 h in the presence of α-MSH (100 nM). α-MSH was used as the negative control. Cytotoxicity of 4H-7MTC was evaluated using MTT assay. Cells were plated in 24-well plates (1.5 × 10<sup>4</sup> cells/well) for 24 h, and then treated with 4H-7MTC (0.3, 0.6, 0.9, 1.2, and 1.5 mM) for 72 h. The data are presented as mean ± SD. Statistical significance was assessed by one-way analysis of variance (ANOVA) followed by Tukey's post-hoc test and represented as follows: \* *p* < 0.05, \*\*\* *p* < 0.001 vs. LPS alone. α μ α α

β α β α To further elucidate the anti-inflammatory mechanisms of 4H-7MTC, we measured the levels of PGE2, IL-6, IL-1β, and TNF-α in culture supernatants using ELISA. Treatment of RAW264.7 cells with LPS alone resulted in a significant increase in cytokine production compared to that in the drug groups (Figure 4). However, NO, PGE2, IL-6, IL-1β, and TNF-α levels in the supernatants of LPS-stimulated cells pretreated with 0.3, 0.6, 0.9, and 1.2 mM 4H-7MTC were significantly reduced compared to those in the LPS group in a concentration-dependent manner (Figure 4). β α β α

β α β α **Figure 4.** The effect of 4-hydroxy-7-methoxycoumarin (4H-7MTC) on the LPS-induced production of proinflammatory cytokines in RAW264.7 cells. Cells were pretreated with 4H-7MTC (0.15, 0.3, 0.6, 0.9, and 1.2 mM) for 1 h and then stimulated for 20 h with LPS. The production of PGE<sup>2</sup> (**a**), IL-1β (**b**), IL-6 (**c**), and TNF-α (**d**) were determined using ELISA. The data are presented as the mean ± SD. Statistical significance was assessed by one-way analysis of variance (ANOVA) followed by Tukey's post-hoc test and represented as follows: Values are representative of three independent experiments. ### p < 0.005 vs. control cells. \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001 vs. LPS alone.

κ

μ

To further elucidate the mechanisms by which 4H-7MTC inhibited NO and PGE<sup>2</sup> production in LPS-activated macrophages, we analyzed the effects of 4H-7MTC on LPS-induced iNOS and COX-2 gene expression in macrophages. Under normal conditions, RAW264.7 cells expressed non-detectable levels of COX-2 expression, but iNOS and COX-2 protein levels markedly increased after 18 h of LPS stimulation (Figure 5). With the addition of 4H-7MTC (0.3, 0.6, 0.9, and 1.2 mM), concentration-dependent inhibition of iNOS and COX-2 expression was observed, indicating that 4H-7MTC modulates iNOS and COX-2 expression.

μ β β **Figure 5.** Effect of 4-hydroxy-7-methoxycoumarin (4H-7MTC) on the level of iNOS in LPS-induced RAW264.7 cells. Lysates were prepared from cells pretreated with 4H-7MTC (0.15, 0.3, 0.6, 0.9, and 1.2 mM) for 1 h and treated with LPS (1 µg/mL) for 18 h. β -actin was used as a loading control. Total cellular proteins were separated using SDS-PAGE, transferred to PVDF membranes, and detected using specific antibodies against iNOS and β-actin (**a**). Results are presented as representative of three independent experiments and summarized in the bar graphs (**b**,**c**). ### *p* < 0.005 vs. control cells. \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.005 vs. LPS-induced cells.

κ κ κ κ κ κ κ α μ A previous study revealed that NF-κB activation in response to pro-inflammatory stimuli involves the rapid phosphorylation of IκBs by the IKK signalosome complex. Free NF-κB produced by this process translocates to the nucleus where it binds to κB-binding sites in the promoter regions of target genes. It then induces the transcription of pro-inflammatory mediators such as iNOS and COX-2. Several studies have shown that anti-inflammatory agents inhibit NF-κB activation by preventing IκB degradation [19–21]. Thus, we attempted to determine whether 4H-7MTC inhibits IκB phosphorylation and degradation. Accordingly, RAW264.7 cells were pretreated for 1 h with 4H-7MTC, and IκB-α protein levels were determined after 20 min of LPS exposure (1 µg/mL). As shown in Figure 6, 4H-7MTC significantly suppressed LPS-induced phosphorylation and degradation of IκB-α. These results show that 4H-7MTC inhibits LPS-induced NF-κB activation by preventing the degradation of IκB-α phosphorylation.

κ α κ κ α MAPK plays a critical role in regulating cell growth and differentiation and controls cellular responses to cytokines and stress. In addition, three MAP kinases (JNK, p38 MAPK, and ERK 1/2) have been reported to be adjustable in LPS-induced pro-inflammatory cytokine production [22–25].

κ α κ α μ κ α κ α β **Figure 6.** Effect of 4-hydroxy-7-methoxycoumarin (4H-7MTC) on the level of phospho-IκB-α and IκB-α in LPS-induced RAW264.7 cells. Lysates were prepared from cells pretreated with 4H-7MTC (0.15, 0.3, 0.6, 0.9, and 1.2 mM) for 1 h and then treated with LPS (1 µg/mL) for 20 min. Western blotting was performed to detect the expression of IκBα and p-IκBα. β-actin was used as a loading control (**a**). Quantification of immunoreactive protein bands is shown via bar graphs (**b**,**c**).

To investigate the molecular mechanism of MAPK signaling by 4H-7MTC in LPS-stimulated RAW264.7 cells, we studied the inhibition of phosphorylation of ERK1/2, p-38, and JNK. RAW264.7 cells were pretreated with 4H-7MTC at the indicated concentrations for 1 h and then stimulated with 1 µg/mL LPS for 1 h. The total cell lysates were then probed with phosphospecific antibodies for ERK1/2 and JNK. Phosphorylation of ERK1/2 and JNK increased in cells treated with LPS alone. Pretreatment with 4H-7MTC inhibited the LPS-induced phosphorylation of JNK and ERK 1/2 in a concentration-dependent manner, but not that of p38 MAPK. The amount of non-phosphorylated MAPKs was not affected by either LPS or 4H-7MTC treatment (Figure 7). κ α κ α μ κ α κ α β

μ β **Figure 7.** Effect of 4-hydroxy-7-methoxycoumarin (4H-7MTC) on LPS-induced MAPK in RAW264.7 cells. Lysates were prepared from cells pretreated with 4H-7MTC (0.15, 0.3, 0.6, 0.9, and 1.2 mM) for 1 h and treated with LPS (1 µg/mL) for 15 min. Western blotting was performed to detect the expression of phospho-ERK, T-ERK, phospho-JNK, T-JKN, phospho-p38, and T-p38 (**a**). β-actin was used as a loading control. Quantification of immunoreactive protein bands is shown via bar graphs (**b**–**d**).

These results suggest that suppression of MAPK phosphorylation may be involved in the inhibitory effect of 4H-7MTC on LPS-stimulated inflammatory response factors and inflammatory cytokines via NF-κB signaling in RAW264.7 cells.

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

#### *3.1. Chemicals and Reagents*

4-Hydroxy-7-methoxycoumarin (4H-7MTC), 4-Hydroxy-6-methylcoumarin (4H-6MC), and 4-Hydroxy-7-methylcoumarin (4H-7MC) were obtained from Tokyo Chemical Industry (Tokyo, Kita-ku, Japan). Lipopolysaccharide (LPS) from *Escherichia coli*, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT), α-melanocyte-stimulating hormone (α-MSH), dimethyl sulfoxide (DMSO), Griess reagent, sodium nitrite, and protease inhibitor cocktail were obtained from Sigma-Aldrich (St Louis, MO, USA). Dulbecco's Modified Eagle Medium (DMEM), fetal bovine serum, and penicillin/ streptomycin were obtained from Thermo Fisher Scientific (Waltham, MA, USA). Radioimmunoprecipitation assay buffer, phosphate-buffered saline (PBS), enhanced chemiluminescence (ECL) kit, and tris-buffered saline (TBS) were obtained from Biosesang (Seongnam, Gyeonggi-do, Korea). *N*-[2-(Cyclohexyloxy)-4-nitrophenyl] methanesulfonamide (NS-398), and L-N<sup>6</sup> -(1-iminoethyl) lysine dihydrochloride (L-NIL) were obtained from Cayman Chemical Company (Ann Arbor, MI, USA). Prostaglandin E<sup>2</sup> (PGE2) ELISA kit, interleukin-1β (IL-1β) kit, IL-6 ELISA kit, and tumor necrosis factor (TNF-α) ELISA kits were obtained from R&D System Inc. (St. Louis, MO, USA). The following antibodies were used in this study: β-actin, anti-iNOS, anti-inhibitor of NF-κB (IκBα), Akt, p-Akt, p38, p-p38, JNK, p-JNK, ERK, and p-ERK were obtained from Cell Signaling Technology (Beverly, MA, USA). Anti-COX-2 was obtained from BD Biosciences (San Diego, CA, USA). All reagents used were of analytical grade.

#### *3.2. Cell Culture*

RAW264.7 mouse macrophages and B16F10 melanoma cells were obtained from the Korean Cell Line Bank (Seoul, Korea). RAW264.7 cells were subcultured at intervals of 2–3 days. The B16F10 melanoma cells were subcultured at 4-day intervals using DMEM with 10% FBS, 100 U/mL penicillin, and 100 µg/mL streptomycin at 37 ◦C in a humidified 5% CO<sup>2</sup> atmosphere.

#### *3.3. Cell Viability*

Cytotoxicity was determined using the MTT assay. RAW264.7 cells were cultured at a density of 1.5 × 10<sup>5</sup> cells/well in 24-well plates for 24 h. Cells were treated with various concentrations of 4H-7MTC (0.3, 0.6, 0.9, and 1.2 mM). RAW264.7 cells were incubated for 24 h and MTT solution (0.2 mg/mL) was added to the medium and incubated for 4 h. Next, the medium was removed and formazan crystals in each well were dissolved in DMSO for 20 min. Optical density (OD) was measured at 570 nm, and the percentage of cells showing cell viability relative to the control was determined.

#### *3.4. NO Production*

NO production in the cell culture was assayed by measuring the accumulated nitrite using Griess reagent. RAW264.7 cells were plated at a density of 1.5 × 10<sup>5</sup> cells/well in 24-well plates. Cells were pretreated with various concentrations of 4H-7MTC (0.3, 0.6, 0.9, 1.2 mM) for 1 h and treated with LPS (1 µg/mL) for 24 h. Then, the treated cell culture solution was mixed with the Griess reagent in a 1:1 ratio, reacted for 15 min, and the absorbance measured at 540 nm using a spectrophotometer. NO production in the sample was quantified from a standard curve constructed using sodium nitrite.

#### *3.5. Measurement of Cytokines*

RAW264.7 mouse cells were plated at a density of 1.5 × 10<sup>5</sup> cells/well in 24-well plates. Cells were pretreated with various concentrations of 4H-7MTC (0.3, 0.6, 0.9, and 1.2 mM) for 1 h and treated with LPS (1 µg/mL) for 24 h. Supernatants were harvested, and PGE2, IL-1β, IL-6, and TNF-α levels were measured using ELISA kits according to the manufacturer's protocols.

#### *3.6. Measurement of Melanin Content*

B16F10 melanoma cells were plated in 60 mm cell culture dishes (6.0 × 10<sup>4</sup> cells/dish), incubated for 24 h, and then treated with 4H-7MTC (25, 50, 100, 150, and 200 µM) for 72 h in the presence of α-MSH (100 nM). After incubation, the cells were washed with 1 × PBS and the pellets were solubilized in 1 N NaOH containing 10% DMSO at 70 ◦C for 1 h. Absorbance was measured at 405 nm with a spectrophotometer. The protein concentration was determined using a BCA protein analysis kit.

#### *3.7. Western Blot Analysis*

RAW264.7 mouse cells were plated at a density of 6.0 × 10<sup>5</sup> cells/dish in 60-mm cell culture dishes for 24 h. Cells were pretreated with various concentrations of 4H-7MTC (0.3, 0.6, 0.9, 1.2 mM) for 1 h and treated with LPS (1 µg/mL) for the indicated times. After incubation, cells were washed with 1 × PBS and lysed on ice with RIPA lysis buffer (150 mM NaCl, 50 mM Tris-HCl [pH 7.5], 2 mM EDTA, 1% Triton X-100, 0.1% SDS, and 1% protease inhibitor cocktail) for 30 min. The harvested cell lysates were centrifuged at −8 ◦C and 15,000 rpm for 20 min. A standard assay curve of bovine serum albumin (BSA) was prepared using the BCA Protein Assay Kit, and the protein contents of the extracted cell lysates were quantitatively determined. The protein concentration was determined using a BCA protein analysis kit. Whole-cell lysates (30 µg) were separated by SDS-polyacrylamide gel electrophoresis on a 10% gel (SDS-PAGE) and electroblotted onto polyvinylidene fluoride (PVDF) membranes. The membranes were then blocked with 5% skim milk and incubated for 2 h. The membrane was washed 6 times with TBS buffer containing 0.1% Tween 20 (TTBS) and then incubated with specific primary antibodies (1:2500) at 4 ◦C for 6 h. The membrane was washed 6 times with TTBS buffer and incubated with a peroxidase-conjugated secondary antibody (1:2000) at room temperature for 2 h. The membrane was then washed six times with TTBS buffer and the protein was detected using an ECL kit.

#### *3.8. Statistical Analysis*

All results are expressed as mean ± standard deviation (SD). Each value represents the mean of three independent experiments. Statistical analysis was performed using a one-way analysis of variance (ANOVA) followed by Tukey's post-hoc test, and survival rates between multiple groups were analyzed using the log-rank test. The significant difference was set at \* *p* < 0.05, \*\* *p* < 0.01, and \*\*\* *p* < 0.001.

#### **4. Conclusions**

This study is, to the best of our knowledge, the first to elucidate the anti-inflammatory properties of 4H-7MTC, which was mediated through the suppression of NO, PGE2, IL-6, IL-1β, and TNF-α production in LPS-stimulated RAW264.7 cells via the NF-κB and MAPK signaling pathways. Our findings indicate that 4H-7MTC may be a promising agent for the clinical prevention and treatment of inflammation-associated diseases in the future. Additionally, 4H-7MTC has also been shown to enhance melanin production and has a potential application as an anticancer agent.

**Author Contributions:** Conceptualization, C.-G.H.; validation and formal analysis, J.K.K.; C.-G.H.; writing—original draft preparation, review, and editing; C.-G.H.; C.-G.H.; funding acquisition, C.-G.H. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported by the Ministry of Trade, Industry & Energy (MOTIE), Korea Institute for Advancement of Technology (KIAT) through the Encouragement Program for the Industries of Economic Cooperation Region (P0006063).

**Acknowledgments:** The authors thank all the students in our research group for their helpful cooperation and discussions. English proofreading of this paper was supported by R&D Program of the Establishment Project of Industry-University Fusion District (N0002327).

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

#### **References**


**Sample Availability:** Samples of the compounds are not 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* **Comparison of Anticoagulation Quality between Acenocoumarol and Warfarin in Patients with Mechanical Prosthetic Heart Valves: Insights from the Nationwide PLECTRUM Study**

**Danilo Menichelli <sup>1</sup> , Daniela Poli <sup>2</sup> , Emilia Antonucci <sup>3</sup> , Vittoria Cammisotto <sup>4</sup> , Sophie Testa <sup>5</sup> , Pasquale Pignatelli <sup>1</sup> , Gualtiero Palareti <sup>3</sup> , Daniele Pastori 1,\* and the Italian Federation of Anticoagulation Clinics (FCSA) †**


**Abstract:** Vitamin K antagonists are indicated for the thromboprophylaxis in patients with mechanical prosthetic heart valves (MPHV). However, it is unclear whether some differences between acenocoumarol and warfarin in terms of anticoagulation quality do exist. We included 2111 MPHV patients included in the nationwide PLECTRUM registry. We evaluated anticoagulation quality by the time in therapeutic range (TiTR). Factors associated with acenocoumarol use and with low TiTR were investigated by multivariable logistic regression analysis. Mean age was 56.8 ± 12.3 years; 44.6% of patients were women and 395 patients were on acenocoumarol. A multivariable logistic regression analysis showed that patients on acenocoumarol had more comorbidities (i.e., ≥3, odds ratio (OR) 1.443, 95% confidence interval (CI) 1.081–1.927, *p* = 0.013). The mean TiTR was lower in the acenocoumarol than in the warfarin group (56.1 ± 19.2% vs. 61.6 ± 19.4%, *p* < 0.001). A higher prevalence of TiTR (<60%, <65%, or <70%) was found in acenocoumarol users than in warfarin ones (*p* < 0.001 for all comparisons). Acenocoumarol use was associated with low TiTR regardless of the cutoff used at multivariable analysis. A lower TiTR on acenocoumarol was found in all subgroups of patients analyzed according to sex, hypertension, diabetes, age, valve site, atrial fibrillation, and INR range. In conclusion, anticoagulation quality was consistently lower in MPHV patients on acenocoumarol compared to those on warfarin.

**Keywords:** warfarin; acenocoumarol; mechanical valve; time in therapeutic range; anticoagulation

#### **1. Introduction**

The burden of valvular heart disease (VHD) is still rising worldwide due to degenerative valve diseases. Although valve rheumatic disease is decreasing [1]. Implantation of mechanical prosthetic heart valves (MPHV) is associated with a reduction in valve-related morbidity compared to biological valves [2]. In MPHV, long-term antithrombotic treatment with only vitamin K antagonists (VKAs) is needed [3]. Consolidated evidence from studies including patients with atrial fibrillation (AF) showed that during VKA treatment, a poor anticoagulation quality, expressed as low time in therapeutic range (TiTR) (<65%–70%),

**Citation:** Menichelli, D.; Poli, D.; Antonucci, E.; Cammisotto, V.; Testa, S.; Pignatelli, P.; Palareti, G.; Pastori, D.; the Italian Federation of Anticoagulation Clinics (FCSA). Comparison of Anticoagulation Quality between Acenocoumarol and Warfarin in Patients with Mechanical Prosthetic Heart Valves: Insights from the Nationwide PLECTRUM Study. *Molecules* **2021**, *26*, 1425. https:// doi.org/10.3390/molecules26051425

Academic Editor: Mee Young Hong

Received: 8 February 2021 Accepted: 3 March 2021 Published: 6 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/).

was associated with an increased risk of thromboembolism [4], cardiovascular events [5], and mortality [6].

However, no specific indication regarding the type of VKAs, such as warfarin or acenocoumarol, is given by international guidelines or expert consensus documents. As a consequence, the use of different VKAs is highly variable among countries, with warfarin being more commonly used in the United Kingdom and Italy, acenocoumarol in Spain, phenprocoumon in Germany, and fluindione in France [7].

Few previous studies investigated potential differences in patients treated with different VKAs. A study including 498 patients with various indications of anticoagulant therapy (AF in 70% of cases) showed that acenocoumarol may be associated with a lower TiTR and a higher international normalized ratio (INR) variability, which improved after switching to phenprocoumon [8].

Previous evidence showed a generally low quality of anticoagulation with VKAs in patients implanted with MPHV [9], but the difference between warfarin and acenocoumarol in terms of clinical characteristics of patients and anticoagulation quality was not investigated in these patients.

The aims of our study were (1) to investigate the clinical characteristics of patients treated with acenocoumarol compared to those treated with warfarin, (2) to describe clinical determinants associated with acenocoumarol use, and (3) to report the proportion of suboptimal anticoagulation quality in acenocoumarol and warfarin use in patients enrolled in the multicenter PLECTRUM registry.

#### **2. Results**

The study enrolled 2111 patients with MPHV, of which 1716 (81.3%) were treated with warfarin and 395 (18.7%) with acenocoumarol. The mean age was 56.8 years and 44.6% of patients were women (Table 1).


**Table 1.** Characteristics of patients according to vitamin K antagonists.


**Table 1.** *Cont.*

INR: international normalized ratio. MPHV: mechanical prosthetic heart valve. TiTR: time in therapeutic range. \* Includes previous stroke/TIA/systemic embolism. \*\* Includes lower limb and carotid disease. ˆ Previous thromboembolism, previous ischemic heart disease, previous hemorrhage. § Includes hypertension, diabetes, heart failure, peripheral artery disease, atrial fibrillation.

> The MPHV site most represented in the whole cohort was aortic (60.7%) and 38.4% of patients had concomitant AF. Patients on acenocoumarol were more frequently affected by arterial hypertension, heart failure (HF), and peripheral artery disease (PAD) and had more comorbidities compared to those on warfarin (Table 1). There was no difference between anticoagulant treatment groups concerning age, sex, MPHV site, INR range, diabetes, previous ischemic heart disease, or thromboembolism at baseline (Table 1). Patients treated with acenocoumarol were affected by a higher number of comorbidities at baseline compared to those treated with warfarin (26.6% vs. 20.0%, respectively, *p* = 0.004).

> At univariable logistic regression analysis (Table 2), factors associated with acenocoumarol use were the number of comorbidities, in particular arterial hypertension, PAD, and HF. At multivariable logistic regression analysis, only the presence of three or more comorbidities (OR 1.443, 95%CI 1.081–1.927, *p* = 0.013) were associated with acenocoumarol use. In a second model using single comorbidities, we found that PAD (OR 1.536, 95%CI 1.085–2.174, *p* = 0.015) and arterial hypertension (OR 1.292, 95%CI 1.016–1.642, *p* = 0.036) were associated with acenocoumarol use.

#### *Anticoagulation Quality According to Treatment*

In the whole cohort, the mean TiTR was 60.6 ± 19.5%; anticoagulation quality was lower in patients treated with acenocoumarol compared to those on warfarin (61.6 ± 19.4% vs. 56.1 ± 19.2%, *p* < 0.001, Table 1).

Analyzing the proportion of suboptimal anticoagulation using different cutoffs of TiTR, we found that acenocoumarol users had a higher prevalence of TiTR < 60%, <65%, or <70% (*p* < 0.001 for all comparisons, Table 1).

Furthermore, after performing a multivariable regression analysis of factors associated with poor anticoagulation quality, acenocoumarol use was found to be associated with low TiTR regardless of the cutoff used: TTR < 60% (OR 1.590, 95%CI 1.262–2.002, *p* < 0.001), TTR < 65% (OR 1.747, 95%CI 1.368–2.232. *p* < 0.001), and TTR < 70% (OR 1.747, 95%CI 1.347–2.266, *p* < 0.001) (Figure 1).


**Table 2.** Univariable logistic regression analysis of clinical factors associated with acenocoumarol use.

PAD: peripheral artery disease. TE: Thromboembolism.

‐

‐

‐

≥

‐

‐

‐ **Figure 1.** Association among acenocoumarol and low time in therapeutic range using different cutoff values in multivariable regression analysis.

To better characterize the association between acenocoumarol and low TiTR, we performed a subgroup analysis according to sex, hypertension, diabetes, age (≥65 years), MPHV site, AF, and INR range (Table 3). A lower anticoagulation quality on acenocoumarol was found in all subgroups of patients analyzed (Table 3).


**Table 3.** Subgroup analysis of time in therapeutic range according to acenocoumarol or warfarin use.

MPHV: mechanical prosthetic heart valve; OAC: oral anticoagulant; TiTR: time in therapeutic range.

We repeated the analysis, excluding patients treated with antiplatelets, and found similar results in 1746 patients as follows: 56.2 ± 18.8 in acenocoumarol-treated vs. 61.7 ± 19.2 in warfarin-treated patients (*p* < 0.001).

#### **3. Material and Methods**

The FCSA-START Valve Study (PLECTRUM) is a retrospective multicenter observational study conducted within the Italian Survey on Anticoagulation Patient Records (START register) and conducted among 33 centers affiliated with the Italian Federation of Thrombosis Diagnosis Centers and Surveillance of Antithrombotic Therapies (FCSA) [10]. The centers were asked to select from their databases patients with a mechanical heart valve prosthesis that was implanted after 1990 and who were followed up on for the management of oral anticoagulant therapy. Patients with MPHV were treated with warfarin or acenocoumarol to prevent thromboembolic event according to European Society of Cardiology guidelines [11]. Each physician prescribed warfarin or acenocoumarol after individualized clinical evaluation. The patients followed by the FCSA centers for the management of oral anticoagulation received an adequate education on the purpose of the treatment, the risk of complications, the INR values, and the management of the dosage of the drugs. The centers performed periodic INR measurements based on INR value, prescribe daily VKA, dosage and scheduled the date for subsequent visits; they also monitored and recorded changes in patient habits, diet, co-medications, intercurrent illness, bleeding, and thrombotic complications during regular follow-up visits through patient interviews. All centers participated

in the specially designed external laboratory quality control program, which is performed 3 times a year and uses lyophilized plasma samples obtained from anticoagulated patients. For this reason, to standardize the quality of INR measurements, none of the patients were monitored at home.

Demographic information and clinical data were collected. Patients were classified as having high blood pressure if they were taking medicines to lower their blood pressure. Diabetes mellitus was defined according to the criteria of the American Diabetes Association. Coronary artery disease was defined on the basis of a history of myocardial infarction or stable and unstable angina. Heart failure was defined as the presence of signs and symptoms of right or left ventricular failure or both and confirmed by non-invasive or invasive measurements that demonstrated objective evidence of cardiac dysfunction. The quality of the anticoagulant control, calculated as TiTR using the linear interpolation method of Rosendaal et al. [12], was analyzed considering the INRs recorded in the last year of follow-up. The study protocol complied with the ethical guidelines of the 1975 Helsinki Declaration, as evidenced by the approval of the institution's human research committee, and informed consent was obtained from each patient. Authorization to set up the registry was obtained from the Ethical Committee of the University Hospital "S. Orsola-Malpighi," Bologna, Italy, in October 2011 (N = 142/2010/0/0ss"). The same institution is charged with deploying and upkeeping the registry central database.

#### *Statistical Analysis*

Continuous variables were reported as mean and standard deviation and compared by the Student t-test. Categorical variables were reported as count and percentage and compared by Pearson chi-squared test. A first descriptive analysis of clinical characteristics according to acenocoumarol or warfarin use was performed. Univariable and multivariable logistic regression analysis was used to calculate the relative odds ratio (OR) with a 95% confidence interval (95%CI) of factors associated with acenocoumarol use and low TiTR. Significance was set at a *p*-value < 0.05. All tests were two-tailed and analyses were performed using computer software packages (SPSS-25.0, SPSS Inc. IBM Corp, Armonk, NY, USA).

#### **4. Discussion**

The difference between acenocoumarol and warfarin effectiveness in terms of anticoagulation stability was never investigated in a large cohort of patients with MPHV. Findings from our study show that 18.7% of patients implanted with MPHV were treated with acenocoumarol in specialized outpatients' clinics for the management of antithrombotic therapies. Acenocoumarol prescription was more common in complex patients, as indicated by the higher number of comorbidities. Patients treated with acenocoumarol showed lower anticoagulation quality compared to those on warfarin. This difference was consistent in all thresholds of TiTR used and in all subgroups of patients regardless of sex, age, valve site, or INR range.

Acenocoumarol presents some pharmacokinetic and pharmacodynamic differences from warfarin that may turn useful in some patients, such a more rapid onset of action, a shorter half-life, and lower renal excretion. In our study, patients with a higher number of comorbidities and use of antiplatelet agents were more frequently prescribed acenocoumarol instead of warfarin. In this last context, the shorter half-life of acenocoumarol may be an advantage in the case of a major or life-threatening bleeding event in patients treated with dual therapy needing a rapid offset of action of the drug.

We found a generally lower anticoagulation quality in patients treated with acenocoumarol, which persisted after adjustment for confounders. Suboptimal anticoagulation with acenocoumarol compared to warfarin was also consistent in all subgroups of patients analyzed, such as sex, hypertension, diabetes (mostly for TiTR < 60%), AF, MPHV site, and INR range. This finding adds to previous evidence that female sex is associated with lower overall anticoagulation quality in the PLECTRUM registry [13].

In a study performed in Poland including 430 patients with mixed indications for VKAs therapy (65.8% AF, 22.6% venous thromboembolism, and 11.6% MPHV) and treated in most cases with acenocoumarol (78.8%), the mean TiTR was as low as 55%, with acenocoumarol use associated with a nearly threefold higher chance of having INR outside the therapeutic range [14].

A previous small study including patients with various indications of anticoagulation showed a significant improvement of TiTR in patients switched from acenocoumarol to warfarin (from 40.2% to 60.2%) [15].

Furthermore, in a population with similar age affected by venous thromboembolism enrolled within the EINSTEIN-DVT and EINSTEIN-PE studies, the use of acenocoumarol was a risk factor for long-term low TiTR (OR 1.81, 95%CI 1.49–2.20, *p* < 0.01) [16].

As patients treated with acenocoumarol were more frequently prescribed antiplatelet agents, which may lead to an increased risk of bleeding episodes and subsequently to a lower adherence to anticoagulant prescription and to a higher discontinuation rate [17], we also performed a subgroup analysis excluding patients on antiplatelets. In this group of patients, we found similar results than the overall cohort, suggesting that anticoagulation quality in MPHV patients is not affected by concomitant administration of antiplatelet drugs.

Our results may have implications for clinical practice. Prescribing acenocoumarol or switching from warfarin should be considered only in select patients in whom warfarin therapy is not successful, such as those with low TiTR or those with recurrent thrombotic events; in patients with a known or suspected warfarin resistance [18], such as antiphospholipid syndrome [19]; and in patients taking drugs interacting with warfarin metabolism.

Our study has limitations to be acknowledged. First, its retrospective observational design is an intrinsic limitation to establishing any inference on our observation. Second, some additional factors not considered in this study may affect both the choice of acenocoumarol use and TiTR; for instance, use of different VKAs may be affected by national guidelines in different countries. Furthermore, some drugs interacting with VKAs that were not considered in this study may influence the TiTR. Finally, we do not have data on concomitant hospitalizations and interruptions for diagnostic/therapeutic procedures that may lead to low anticoagulation quality. In addition, we included only Caucasian patients with a mean age of 60 years, and the reproducibility of our findings in elderly patients and in patients with different ethnic origins needs to be explored. Indeed, ethnic differences such as environmental factors and genetic variants of isoenzymes may affect pharmacokinetic features, hepatic metabolism, and renal elimination of warfarin [20]. Finally, the difference between acenocoumarol and warfarin in other settings such as AF and venous thrombosis needs to be explored, even if in these contexts the use of direct oral anticoagulation is replacing VKAs in many countries.

In conclusion, warfarin would be the first-choice treatment for thromboprophylaxis in patients with MPHV regardless of valve site and INR range. Switching from acenocoumarol to warfarin may improve TiTR in patients with unstable anticoagulation.

**Author Contributions:** Conceptualization, D.M. and D.P. (Daniela Poli); Formal analysis, D.P. (Daniela Poli); Investigation, D.M., D.P. (Daniela Poli), V.C., E.A., G.P., S.T., P.P., D.P. (Daniele Pastori); Data Curation, E.A., G.P.; Writing—Original Draft Preparation, D.M., D.P. (Daniela Poli), V.C., E.A., G.P., S.T., P.P., D.P. (Daniele Pastori); Writing—Review and Editing, D.M., D.P. (Daniela Poli), V.C., E.A., G.P., S.T., P.P., D.P. (Daniele Pastori). All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Institutional Review Board of the Coordinating Center at the University Hospital "S. Orsola-Malpighi," Bologna, Italy, in October 2011 (N = 142/2010/0/0ss") and by all participating centers.

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** The data presented in this study are available in this article.

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

**Sample Availability:** Not available.

#### **Appendix A**

Italian Federation of Anticoagulation Clinics (FCSA).

Sophie Testa, Oriana Paoletti; Dipartimento di Medicina di Laboratorio, Centro Emostasi e Trombosi ASST Cremona.

Corrado Lodigiani; Paola Ferrazzi; Ilaria Quaglia; Centro Trombosi e Malattie Emorragiche, Humanitas Research Hospital, IRCCS Rozzano-Milano.

Daniela Poli Centro Trombosi SOD Malattie Aterotrombotiche Azienda Ospedaliero Universitaria Careggi; Firenze.

Nadia Coffetti; Rosa Marotta; Varusca Brusegan, Orazio Bergamelli, Servizio di Immunoematologia e Medicina Trasfusionale Azienda Ospedaliera Bolognini, ASST Bergamo Est, Seriate e Ambulatorio TAO SIMT Ospedale Fernaroli, Alzano Lombardo.

Roberto Facchinetti Laboratorio Analisi Azienda Ospedaliera Universitaria Integrata Ospedale Civile Maggiore Di Borgo Trento; Verona.

Giuseppina Serricchio; Silvia Sarpau, Francesca Brevi; ASST Lariana Como.

Pietro Falco; Guarino Silverio; Poliambulatorio Specialistico MEDICAL PONTINO, Latina.

Catello Mangione; Giacomo Bellomo, Servizio Immunotrasfusionale Ospedale "Santa Caterina Novella" Galatina (Lecce).

Serena Masottini; Alessandra Cosenza; Centro per la prevenzione, diagnosi e trattamento dellemalattie tromboemboliche- Asl 8- Cagliari.

Lucia Ruocco; Paolo Chiarugi, Monica Casini; Ambulatorio Antitrombosi CAT-TAO AOU Pisana, Pisa.

Arturo Cafolla; Antonietta Ferretti; Ematologia, UOS Emostasi e Trombosi, Policlinico Umberto I◦ , Roma.

Giorgia Micucci; Serena Rupoli; Lucia Canafoglia; Azienda Ospedaliero-Universitaria Ospedali Riuniti Di Ancona.

Paolo Pedico; Rita Galasso; Rosa Rotunno; U.O. MedicinaTrasfusionale Ospedale Mons. Raffaele Dimiccoli Barletta.

Antonio Insana; Paolo Valesella; Servizio Di Patologia Clinica Dipartimento Dei Servizi—Ospedale S. Croce Moncalieri, Moncalieri, Torino.

Angelo Santoro; U.O.C. Patologia Clinica e Centro Trombosi, Presidio Ospedaliero "A. Perrino," ASL Brindisi.

Francesco Marongiu, Doris Barcellona; Centro Di Fisiopatologia dell'emostasi e Terapia Anticoagulante, Azienda Ospedaliera Universitaria di Monserrato, Cagliari.

Vittorio Pengo, Gentian Denas; Istituto di Cardiologia, Policlinico Universitario, Università di Padova.

Carmelo Paparo; Centro Anti Trombosi Ospedale Maggiore ASL TO5, Chieri (Torino).

Eugenio Bucherini; Flavia Tani; Enrico Carioli, Centro di Sorveglianza per la Terapia Anticoagulante, Angiolgia- Medicina Vascolare U.O. Cardiologia O.C. Per gli infermi— Faenza AUSL Romagna, (Ravenna).

Francesco Violi, Pasquale Pignatelli, Daniele Pastori, Mirella Saliola; Dipartimento di Scienze Cliniche, Internistiche, Anestesiologiche e Cardiovascolari, Sapienza Università di Roma.

Lucilla Masciocco, Pasquale Saracino, Angelo Benvenuto, UOC Medicina Interna, Centro Controllo Coagulazione, Ospedale Lastaria, Lucera (Foggia).

Anna Turrini; Stefano Ciaffone; Ospedale "SACRO CUORE" Laboratorio Analisi Cliniche E Medicina Trasfusionale Negrar, Verona.

Andrea Toma; Pietro Barbera UOC di Patologia Clinica Arzignano, Vicenza.

Paolo Gresele; Tiziana Fierro; Stefano Pasquino Department of Medicine, Section of Internal and Cardiovascular Medicine, University of Perugia.

Lucia La Rosa; Rino Morales Centro Trasfusionale e ambulatorio emostasi e trombosi; ASST Vimercate.

Francesco Ronchi, Giuseppe Isu Centro Tao Servizio Di Patologia Clinica Ospedale Ns Signora Di Bonaria Asl 6 Sanluri.

Teresa Lerede, Luca Barcella Centro Trombosi e Emostasi Immunoematologia eMedicina Trasfusionale ASST Papa Giovanni XXIII, Bergamo.

Luigi Ria, Centro Trombosi ed. Emostasi, U.O.C. di Medicina Interna, Ospedale "Sacro Cuore di Gesù" Gallipoli; ASL Lecce, Lecce.

Rosanna Crisantemo; Luciano Suriano, Luciano Lorusso; Mario De Sarlo Servizio di Immunoematologia e Medicina Trasfusionale Ospedale L.Bonomo, Andria.

Pasquale Carrato Istituto Polidiagnostico Santa Chiara, Agropoli, Salerno.

Carmine Oricchio U.O.S. Centro Trasfusionale del P.O. "Luigi Curto" di Polla— ASL Salerno.

Elvira Grandone, Donatella Colaizzo, Centro Trombosi, I.R.C.C.S. Casa Sollievo della Sofferenza, S. Giovanni Rotondo, Foggia.

Maurizio Molinatti Unità Funzionale di Ematologia Centro TAO Humanitas Cell, Torino.

#### **References**


### *Communication* **Chiral Tertiary Amine Catalyzed Asymmetric [4 + 2] Cyclization of 3-Aroylcoumarines with 2,3-Butadienoate**

**Jun-Lin Li <sup>1</sup> , Xiao-Hui Wang <sup>1</sup> , Jun-Chao Sun <sup>1</sup> , Yi-Yuan Peng <sup>1</sup> , Cong-Bin Ji <sup>2</sup> and Xing-Ping Zeng 1,\***


**Abstract:** Coumarins and 2*H*-pyran derivatives are among the most commonly found structural units in natural products. Therefore, the introduction of 2*H*-pyran moiety into the coumarin structural unit, i.e., dihydrocoumarin-fused dihydropyranones, is a potentially successful route for the identification of novel bioactive structures, and the synthesis of these structures has attracted continuing research interest. Herein, a chiral tertiary amine catalyzed [4 + 2] cyclization of 3-aroylcoumarines with benzyl 2,3-butadienoate was reported. In the presence of Kumar's 6'-(4-biphenyl)-β-iso-cinchonine, the desired dihydrocoumarin-fused dihydropyranone products could be obtained in up to 97% yield and 90% ee values.

**Keywords:** coumarins; dihydrocoumarin-fused dihydropyranones; 3-aroylcoumarines; benzyl 2,3 butadienoate; 6'-(4-biphenyl)-β-iso-cinchonine

**Citation:** Li, J.-L.; Wang, X.-H.; Sun, J.-C.; Peng, Y.-Y.; Ji, C.-B.; Zeng, X.-P. Chiral Tertiary Amine Catalyzed Asymmetric [4 + 2] Cyclization of 3-Aroylcoumarines with 2,3-Butadienoate. *Molecules* **2021**, *26*, 489. https://doi.org/10.3390/ molecules26020489

Received: 28 November 2020 Accepted: 14 January 2021 Published: 18 January 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/).

#### **1. Introduction**

Coumarin derivatives are among the most commonly found structural units in natural products, pharmaceuticals, and functional materials [1–5]. Therefore, numerous endeavors have been devoted to develop effective methods for the synthesis of coumarin based compounds [6–11]. On the other hand, 2*H*-pyran moieties also play a vital role in natural and unnatural bioactive compounds. Therefore, the introduction of 2*H*-pyran moiety into coumarin structural unit is a highly potential route for the identification of novel bioactive structures and the synthesis of these structures, i.e., dihydrocoumarin-fused dihydropyranones, have attracted continuing research interest. Among the developed methods, the [4 + 2] reaction of 3-aroylcoumarins are the most commonly used [12–14].

Early in 2012, Shi and co-worker described the first [4 + 2] cyclization of 3-aroylcoumarines (**1**) with ethyl 2,3-butadienoate (**2a**) to construct racemic dihydrocoumarin-fused dihydropyranones **3** in 79–95% yield using DABCO as the Lewis base catalyst (Figure 1a) [15]. This [4 + 2] process was initiated by the necleophilic attack of tertiary amine to 2,3-butadienoate to generate zwitterionic **I**. The γ-carbanion of **I** then attacks the β-carbon of enones **1** to give **II** with *Z* configuration to avoid the interaction of the ester group with the 3-position substituent. In the following, an intramolecular nucleophilic substitution of **II** could give cycloadduct **3** and regenerate the tertiary amine catalyst. Soon after that, the Ye group reported that chiral dihydrocoumarin-fused dihydropyranones could be accessed in a highly enantioselective manner via chiral NHC catalyzed [4 + 2] cycloaddition of ketenes and 3-aroylcoumarins [16]. In 2016, Lu et al. achieved a phosphine-catalyzed [4 + 2] annulation of allenones with 3-aroylcoumarins to afford chiral dihydrocoumarin-fused dihydropyrans [17]. Chen and co-workers reported the synthesis of chiral dihydrocoumarin-fused dihydropyrans through dienamine catalysis, but only two examples were explored [18]. Despite the above achievements, the identification of new protocols using easily available starting materials and chiral catalysts for the enantioselective construction of dihydrocoumarinfused dihydropyrans are still highly desirable.

**Figure 1.** (**a**) DABCO catalyzed [4 + 2] cyclization of 3-aroylcoumarines with 2,3-butadienoate; (**b**) Chiral tertiary amine catalyzed [4 + 2] cyclization of 3-aroylcoumarines with 2,3-butadienoate

Inspired by Shi's pioneering work and based on our interest in the synthesis of chiral coumarin derivatives, we envision that the replacement of DABCO with a suitable chiral tertiary amine catalyst to mediate the [4 + 2] cyclization of 3-aroylcoumarines with 2,3 butadienoate might offer a new method for the synthesis of chiral dihydrocoumarin-fused dihydropyrans (Figure 1b).

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

We started our investigations by carrying out the reaction between 3-benzoylcoumarine (**1a**) and benzyl 2,3-butadienoate (**2b**) in ClCH2CH2Cl at 25 ◦C (Table 1). Initially, a series of cinchona alkaloids, including quinine (**4a**), cinchonine (**4b**), and *C*2-symmetric (bis)cinchona alkaloid (**4c**–**f**) were screened (entries 1–6), and the corresponding chiral dihydrocoumarin-fused dihydropyran product (**3a**) was obtained in up to only 45% ee values when **4c** was used, but the yield of **3a** was pretty low even after 48 h (entry 3). To our delight, when the bifunctional β-isocupreidine (**4g**) was tested in our reaction, the reaction was greatly accelerated to complete with 8 h and delivered **3a** in almost quantitative yield with promising 52% ee value (entry 7) [19,20]. Based on this result, we turn our attention to modify β-isocupreidine, so as to improve the enantiocontrol of the reaction. According the methods reported by Kumar and co-workers, a series of 6'-aryl-β-iso-cinchonine (**4h**–**l**) were prepared and examined in the current reaction [21]. It was observed that 6'-phenyl-βiso-cinchonine **4h** could facilitate the model reaction to give 82% yield for **3a** with improved 60% ee (entry 8). Further variation of the phenyl into more steric aryl groups turned out to be ineffective, as is demonstrated by the 29–35% ee values obtained from catalyst **4i**–**k**. A slightly improved 63% ee was achieved when β-iso-cinchonine (**4l**) bearing a longer 4-biphenyl group at the 6' position was tried (entry 12), but no more improvement was obtained when further increase the length the substituent (entry 13).

326

**Table 1.** Condition optimization for the catalytic asymmetric [4 + 2] cyclization.

In the following, the solvent effects were examined using **4l** as the catalyst. The reaction was found to be more effective in solvent with moderate polarity (entries 14–21) and EtOAc was found to be the best, which afforded the desired product **3a** in 87% yield and 80% ee (entry 18). In the following, we tried to lower the reaction temperature to 0 ◦C to improve the ee value of **3a** (entry 22). To our surprise, the reaction finished within 1 h and gave an improved 93% yield, but no improvement of the ee value was observed. The observed higher reactivity at 0 ◦C than at 25 ◦C might be attributed to the competitive nucleophilic addition of quinoline nitrogen atoms of the catalyst to 2,3-butadienoate at higher temperature, which deactivate the catalyst and retard the reaction catalytic cycle.

Based on the above optimization, the scope of this tertiary amine catalyzed enantioselective [4 + 2] cyclization of 3-aroylcoumarines with benzyl 2,3-butadienoate (**2b**); we

then evaluated this using 10 mol% of **4l** as the catalyst in EtOAc at 0 ◦C (Figure 2). It was observed that the reaction outcome was significantly affected by the electronic properties of the substituents on the coumarin benzene ring. In general, substrates bearing electrondonating groups (EDG, such as Me, OMe) were relatively less reactive and afforded slightly higher ee values. As is shown by the 79–98% yields and 80–88% ee values for products **3b**–**e**. The more steric **1f** substrate could give the corresponding product **3f** in highest 90% ee under standard conditions. In contrast, the reactions of electron-withdrawing group (EWG) substituted substrates are found to be more reactive but delivered relatively lower enantioselectivities. For example, products **3g**–**i** were obtained in excellent yield but with only around 70% ee. Additionally, coumarins bearing both EDG and EWG on the benzyl ring of the aroyl group were also well tolerated under standard conditions, but the enantiomeric excess was significantly affected by the steric effect. The *para*-substituted products (**3k**,**n**) could be obtained in much higher ee values than the *meta*-substituted products (**3l**,**o**). Moreover, the current reaction is also suitable for the reaction of (1,1'-biphenyl)-4-carbonyl and thiophene-2-carbonyl substituted coumarins, which afforded the desired products **3o** and **3p** in 79% and 81% ee values, respectively.

The *Z*/*E* configuration of the products was determined by the converting product **3m** into the known compound **5** and comparing their NMR spectrum (Figure 3). Under the above optimized conditions, the reaction of 3-benzoyl coumarin **1m** with ethyl 2,3 butadienoate **2a** afforded product **5** with 45% ee value. The same product **5** could also be obtained via a 3-step sequence from **3m** and **2b** in 79% ee value. The NMR spectrum of these newly synthesized products **5** were identical with the previous report by Shi and co-workers. Thus, the configuration of the products **3** were assigned to be *E*. This process also highlighted the synthetic potential of product **1** to be elaborated into other dihydrocoumarin-fused dihydropyran derivatives. We also tried to recrystallize products **3** and determined their absolute configuration by X-ray crystallography analysis, but turned out to be unsuccessful.

In order to demonstrate the practicability of the current method, we conducted a gram-scale reaction of 3-benzoyl coumarin **1f** with benzyl 2,3-butadienoate **2b** (Figure 4). In the presence of only 2.5 mol% of **4l** as the catalyst, the reaction of 2.5 mmol of **1f** with 1.5 equivalents of **2b** could give rise to the desired dihydrocoumarin-fused dihydropyran **3f** in 91% yield (1.215 g) with slightly improved 93% ee.

In summary, a series of chiral dihydrocoumarin-fused dihydropyranones were sysnthesized via the enantioselective [4 + 2] cyclization of 3-aroylcoumarines with benzyl 2,3-butadienoate. In the presence of 10 mol% of Kumar's 6'-(4-biphenyl)-β-iso-cinchonine as the chiral tertiary amine catalyst, the desired products could be obtained in up to 97% yield and 90% ee values under mild conditions. The current method used an easily available chiral catalyst and starting materials and could be conducted on gram-scale without loss of enantiomeric excess. The thus obtained products are potential in the construction of other dihydrocoumarin-fused dihydropyran derivatives. Considering the wide existence of coumarins and 2*H*-pyran moieties in natural products and pharmaceuticals, the thus obtained optically active dihydrocoumarin-fused dihydropyranones should be of interest to medicinal chemists.

**Figure 2.** Substrate scope of tertiary amine catalyzed asymmetric [4 + 2] cyclization.

**Figure 3.** *Z*/*E* Configuration determination.

**Figure 4.** Gram-scale synthesis and product elaboration.

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

#### *3.1. General Information*

Reactions were monitored by thin layer chromatography using UV light or KMnO<sup>4</sup> to visualize the course of reaction. Purification of reaction products was carried out by flash chromatography on silica gel. Chemical yields refer to pure isolated substances. The [α]<sup>D</sup> was recorded using PolAAr 3005 High Accuracy Polarimeter (Optical Activity Ltd., Huntingdon, England). Infrared (IR) spectra were obtained using a Bruker tensor 27 infrared spectrometer (Bruker, Borken, Germany). <sup>1</sup>H, <sup>13</sup>C and <sup>19</sup>F NMR spectra were obtained using Bruker DPX-400 spectrometer (Bruker UK Limited, Coventry, UK). Chiral HPLC analyses were obtained using Agilent Technologies 1260 Infinity series (Agilent Technologies, Inc., Waldbronn, Germany) and DAICEL CHIRALPAK columns (CPI Company, Tokyo, Japan). Chemical shifts were reported in ppm from tetramethylsilane with the solvent resonance as the internal standard. The following abbreviations were used to designate chemical shift multiplicities: s = singlet, d = doublet, t = triplet, q = quartet, h = heptet, m = multiplet, br = broad.

#### *3.2. Tertiary Amine Catalyzed Asymmetric [4 + 2] Cyclization*

General procedure: To a 4 mL vial was sequentially added 3-aroylcoumarines **1** (0.2 mmol), catalyst **4l** (0.02 mmol, 10 mol%), and EtOAc (1.0 mL); the mixture was stirred at 0 ◦C for 15 min before benzyl buta-2,3-dienoate **2b** (0.3 mmol, 1.5 equiv.) was charged. The reaction was monitored by TLC analysis. After completion of the reaction, the solvent was removed by rotary evaporation and the residue was directly subjected to column chromatography using PE/EtOAc (20:1–15:1) as the eluent to afford product **3**.

Benzyl (*E*)-2-(5-oxo-4-phenyl-1,10b-dihydro-2H,5H-pyrano[3,4-c]chromen-2-ylidene) acetate (**3a**).

White solid (m.p. 121.4–122.1 ◦C); <sup>1</sup>H NMR (400 MHz, CDCl3) δ 7.51–7.31 (m, 12H), 7.20 (t, *J* = 7.2 Hz, 1H), 7.10 (d, *J* = 8.0 Hz, 1H), 5.90 (d, *J* = 1.6 Hz, 1H), 5.23 (s, 2H), 4.87 (dd, *J* = 14.8, 6.0 Hz, 1H), 4.01 (dd, *J* = 12.2, 5.6 Hz, 1H), 2.56–2.49 (m, 1H); <sup>13</sup>C NMR (101 MHz, CDCl3) δ 166.6, 164.2, 162.3, 161.5, 150.8, 136.0, 133.1, 130.9, 129.2, 129.1, 128.8, 128.5, 128.3, 128.3, 125.8, 124.9, 122.7, 117.3, 102.2, 101.5, 66.4, 30.3, 26.1; [α]<sup>D</sup> 26.0 = + 44.3 (c = 0.26, CHCl3); The enantiomeric purity of **3a** was determined by HPLC analysis (DAICEL CHIRALPAK AD–H (0.46 cmϕ × 25 cm), hexane:2-propanol = 60:40, flow rate = 0.75 mL/min, retention time: 11.2 min (major) and 14.5 min (minor)); HRMS (ESI): Exact mass calcd for C27H21O<sup>5</sup> [M+H]<sup>+</sup> : 425.1389, found: 425.1392.

Benzyl (*E*)-2-(9-methoxy-5-oxo-4-phenyl-1,10b-dihydro-2H,5H-pyrano[3,4-c]chromen-2-ylidene) acetate (**3b**).

White solid (m.p. 131.1–132.8 ◦C); <sup>1</sup>H NMR (400 MHz, CDCl3) δ 7.51–7.36 (m, 10H), 7.03 (d, *J* = 8.8 Hz, 1H), 6.91 (d, *J* = 2.0 Hz, 1H), 6.85 (dd, *J* = 8.8, 2.8 Hz, 1H), 5.91 (d, *J* = 1.6 Hz, 1H), 5.23 (s, 2H), 4.80 (dd, *J* = 14.6, 5.6 Hz, 1H), 3.98 (dd, *J* = 12.0, 5.6 Hz, 1H), 3.83 (s, 3H), 2.58–2.51(m,1H); <sup>13</sup>C NMR (101 MHz, CDCl3) δ 166.6, 164.1, 162.1, 161.9, 156.8, 144.7, 136.0, 133.0, 130.8, 129.1, 128.8, 128.5, 128.3, 128.3, 123.7, 118.0, 114.2, 111.1, 102.2, 101.4, 66.4 56.0, 30.5, 25.9; [α]<sup>D</sup> 26.0 = +242.2 (c = 0.49, CHCl3); The enantiomeric purity of **3b** was determined by HPLC analysis (DAICEL CHIRALPAK AD-H (0.46 cmϕ × 25 cm), hexane:2-propanol = 60:40, flow rate = 0.75 mL/min, retention time: 15.2 min (major) and 21.0 min (minor)); HRMS (ESI): Exact mass calcd for C28H23O<sup>6</sup> [M+H]<sup>+</sup> : 455.1495, found: 455.1496.

Benzyl (*E*)-2-(7-methoxy-5-oxo-4-phenyl-1,10b-dihydro-2H,5H-pyrano[3,4-c]chromen-2-ylidene)acetate (**3c**).

White solid (m.p. 108.7–110.2 ◦C); <sup>1</sup>H NMR (400 MHz, CDCl3) δ 7.53–7.34 (m, 10H), 7.14 (t, *J* = 8.4 Hz, 1H), 6.95 (dd, *J* = 13.0, 8.4 Hz, 2H), 5.91 (d, *J* = 1.2 Hz, 1H), 5.24 (s, 2H), 4.83 (dd, *J* = 14.2, 6.0 Hz, 1H), 4.00 (dd, *J* = 12.4, 6.0 Hz, 1H), 3.90 (s, 3H), 2.57–2.50 (m, 1H); <sup>13</sup>C NMR (101 MHz, CDCl3) δ 166.5, 164.2, 162.1, 161.0, 147.8, 140.2, 136.0, 132.9, 130.8, 129.1, 128.7, 128.4, 128.3, 128.2, 124.7, 123.8, 117.0, 111.7, 102.1, 101.2, 66.3, 56.2, 30.5, 26.0; [α]<sup>D</sup> 26.0 = +86.5 (c = 0.5, CHCl3); The enantiomeric purity of **3c** was determined by HPLC analysis (DAICEL CHIRALPAK AD-H (0.46 cmϕ × 25 cm), hexane:2-propanol = 60:40, flow rate = 0.75 mL/min, retention time: 13.0 min (major) and 17.0 min (minor)). HRMS (ESI): Exact mass calcd for C28H23O<sup>6</sup> [M+H]<sup>+</sup> : 455.1495, found: 455.1493.

Benzyl (*E*)-2-(8-methoxy-5-oxo-4-phenyl-1,10b-dihydro-2H,5H-pyrano[3,4-c]chromen-2-ylidene)acetate (**3d**).

White solid (m.p. 101.9–102.5 ◦C); <sup>1</sup>H NMR (400 MHz, CDCl3) δ 7.50–7.34 (m, 10H), 7.27–7.24 (m, 1H), 6.74 (dd, *J* = 8.6, 2.4 Hz, 1H), 6.63 (d, *J* = 2.4 Hz, 1H), 5.87 (d, *J* = 1.6 Hz, 1H), 5.21 (s, 2H), 4.81 (dd, *J* = 14.8, 5.6 Hz, 1H), 3.92 (dd, *J* = 12.4, 5.6 Hz, 1H), 3.80 (s, 3H), 2.47–2.44 (m, 1H); <sup>13</sup>C NMR (101 MHz, CDCl3) δ 166.6, 164.3, 162.2, 161.5, 160.3, 151.5, 136.0, 133.1, 130.8, 129.0, 128.8, 128.4, 128.3, 126.5, 114.5, 111.1, 102.5, 102.0, 101.7, 66.3, 55.7, 29.7, 26.4; [α]<sup>D</sup> 26.0 = −44.3 (c = 0.26, CHCl3); The enantiomeric purity of **3d** was determined by HPLC analysis (DAICEL CHIRALPAK AD-H (0.46 cmϕ × 25 cm), hexane:2-propanol = 60:40, flow rate = 0.75 mL/min, retention time: 12.2 min (major) and 17.7 min (minor)). HRMS (ESI): Exact mass calcd for C28H23O<sup>6</sup> [M+H]<sup>+</sup> : 455.1495, found: 455.1498.

Benzyl (*E*)-2-(9-methyl-5-oxo-4-phenyl-1,10b-dihydro-2H,5H-pyrano[3,4-c]chromen-2-ylidene)acetate (**3e**).

White solid (m.p. 127.6–128.9 ◦C); <sup>1</sup>H NMR (400 MHz, CDCl3) δ 7.51–7.35 (m, 10H), 7.18 (s, 1H), 7.12 (d, *J* = 8.4 Hz, 1H), 6.99 (d, *J* = 8.0 Hz, 1H), 5.91 (d, *J* = 1.6 Hz, 1H), 5.24 (s, 2H), 4.87 (dd, *J* = 14.8, 5.6 Hz, 1H), 3.97 (dd, *J* = 12.4, 5.6 Hz, 1H), 2.53–2.46 (m, 1H), 2.37 (s,3H); <sup>13</sup>CNMR (101 MHz, CDCl3) δ 166.7, 164.4, 162.1, 161.7, 148.7, 136.0, 134.6, 133.1, 130.8, 129.7, 129.0, 128.8, 128.5, 128.3, 128.2, 126.2, 122.2, 117.0, 102.0, 101.6, 66.4, 30.2, 26.1, 21.0; [α]<sup>D</sup> 26.0 = +155.3 (c = 0.50, CHCl3); The enantiomeric purity of **3e** was determined by HPLC analysis (DAICEL CHIRALPAK AD-H (0.46 cmϕ × 25 cm), hexane:2 propanol = 60:40, flow rate = 0.75 mL/min, retention time: 12.7 min (major) and 18.5 min (minor)). HRMS (ESI): Exact mass calcd for C28H23O<sup>5</sup> [M+H]<sup>+</sup> : 439.1545, found: 439.1548.

Benzyl (*E*)-2-(7,9-di-tert-butyl-5-oxo-4-phenyl-1,10b-dihydro-2H,5H-pyrano[3,4-c] chromen-2-ylidene)acetate (**3f**).

Yellow solid (m.p. 83.7–84.9 ◦C); <sup>1</sup>H NMR (400 MHz, CDCl3) δ 7.54–7.37 (m, 11H), 7.28 (s, 1H), 5.95 (s, 1H), 5.28 (s, 2H), 4.76 (dd, *J* = 14.8, 6.0 Hz, 1H), 4.01 (dd, *J* = 11.6, 6.0 Hz, 1H), 2.77 (t, *J* = 13.2 Hz, 1H), 1.49 (s, 9H), 1.39 (s, 9H); <sup>13</sup>C NMR (101 MHz, CDCl3) δ 166.6, 164.1, 162.1, 161.9, 156.8, 144.7, 136.0, 133.0, 130.9, 129.1, 128.8, 128.5, 128.3, 128.3 123.7, 118.0, 114.2, 111.1, 102.2, 101.4, 66.4, 56.0, 30.5, 25.9; [α]<sup>D</sup> 26.0 = +58.6 (c = 0.52, CHCl3); The enantiomeric purity of **3f** was determined by HPLC analysis (DAICEL CHIRALPAK AD-H (0.46 cmϕ × 25 cm), hexane:2-propanol = 60:40, flow rate = 0.75 mL/min, retention time: 6.0 min (major) and 5.1 min (minor)). HRMS (ESI): Exact mass calcd for C35H37O<sup>5</sup> [M+H]<sup>+</sup> : 537.2641, found: 537.2643.

Benzyl (*E*)-2-(9-fluoro-5-oxo-4-phenyl-1,10b-dihydro-2H,5H-pyrano[3,4-c]chromen-2-ylidene)acetate (**3g**).

Yellow solid (m.p. 118.4–119.7 ◦C); <sup>1</sup>H NMR (400 MHz, CDCl3) δ 7.53–7.39 (m, 10H), 7.11–7.00 (m, 3H), 5.93 (d, *J* = 1.6 Hz, 1H), 5.25 (s, 2H), 4.80 (dd, *J* = 14.6, 5.6 Hz, 1H), 3.97 (dd, *J* = 12.4, 5.6 Hz, 1H), 2.56–2.49 (m, 1H); <sup>13</sup>C NMR (101 MHz, CDCl3) δ 166.3, 163.5, 162.6, 161.1, 159.4 (d, <sup>1</sup> *J* = 243 Hz), 146.7, 135.9, 132.8, 130.9, 129.0, 128.7, 128.4, 128.2, 128.2, 124.3, 124.2, 118.5, 118.4, 115.8 (d, <sup>2</sup> *J* = 23 Hz), 112.6 (d, <sup>3</sup> *J* = 25 Hz), 102.4, 100.4, 66.4, 30.3, 25.7; [α]<sup>D</sup> 26.0 = −22.8 (c = 0.47, CHCl3); The enantiomeric purity of **3g** was determined by HPLC analysis (DAICEL CHIRALPAK AD-H (0.46 cmϕ × 25 cm), hexane:2-propanol = 60:40,

flow rate = 0.75 mL/min, retention time: 11.4 min (major) and 14.1 min (minor)). HRMS (ESI): Exact mass calcd for C27H20FO<sup>5</sup> [M+H]<sup>+</sup> : 443.1295, found: 443.1297.

Benzyl (*E*)-2-(9-chloro-5-oxo-4-phenyl-1,10b-dihydro-2H,5H-pyrano[3,4-c]chromen-2-ylidene)acetate (**3h**).

White solid (m.p. 127.4–128.9 ◦C); <sup>1</sup>H NMR (400 MHz, CDCl3) δ 7.50–7.47 (m, 3H), 7.43–7.40 (m, 8H), 7.39–7.28 (m, 1H), 7.03 (d, *J* = 8.4 Hz, 1H), 5.92 (d, *J* = 2.0 Hz, 1H), 5.24 (s, 2H), 4.83 (dd, *J* = 14.6, 5.6 Hz, 1H), 3.98 (dd, *J* = 12.4, 5.6 Hz, 1H), 2.55–2.47 (m, 1H); <sup>13</sup>C NMR (101 MHz, CDCl3) δ 166.4, 163.5, 162.9, 161.0, 149.3, 135.9, 132.8, 131.0, 130.1, 129.2, 129.0, 128.8, 128.5, 128.3, 128.3, 125.9, 124.3, 118.6, 102.6, 100.4, 66.5, 30.3, 25.8; [α]<sup>D</sup> 26.0 = −56.8 (c = 0.19, CHCl3); The enantiomeric purity of **3h** was determined by HPLC analysis (DAICEL CHIRALPAK AD-H (0.46 cmϕ × 25 cm), hexane:2-propanol = 60:40, flow rate = 0.75 mL/min, retention time: 13.7 min (major) and 16.4 min (minor)). HRMS (ESI): Exact mass calcd for C27H20ClO<sup>5</sup> [M+H]<sup>+</sup> : 459.0999, found: 459.0997.

Benzyl (*E*)-2-(9-bromo-5-oxo-4-phenyl-1,10b-dihydro-2H,5H-pyrano[3,4-c]chromen-2-ylidene)acetate (**3i**).

White solid (m.p. 131.4–132.7 ◦C); <sup>1</sup>H NMR (400 MHz, CDCl3) δ 7.51–7.35 (m, 12H), 6.98 (d, J = 8.8 Hz, 1H), 5.92 (d, J = 1.2 Hz, 1H), 5.24 (s, 2H), 4.83 (dd, J = 14.6, 6.0 Hz, 1H), 3.99 (dd, J = 12.2, 6.0 Hz, 1H), 2.55–2.48 (m, 1H); <sup>13</sup>C NMR (101 MHz, CDCl3) δ 166.4, 163.5, 162.9, 160.9, 149.9, 135.9, 132.8, 132.2, 131.0, 129.0, 128.8, 128.8, 128.5, 128.3, 128.3, 124.8, 119.0, 117.6, 102.6, 100.4, 66.5, 30.3, 25.8; [α]<sup>D</sup> 26.0 = −17.8 (c = 0.22, CHCl3); The enantiomeric purity of **3i** was determined by HPLC analysis (DAICEL CHIRALPAK AD-H (0.46 cmϕ × 25 cm), hexane:2-propanol = 60:40, flow rate = 0.75 mL/min, retention time: 17.1 min (major) and 14.2 min (minor)). HRMS (ESI): Exact mass calcd for C27H20BrO<sup>5</sup> [M+H]<sup>+</sup> : 503.0494, found: 503.0490.

Benzyl (*E*)-2-(8-bromo-5-oxo-4-phenyl-1,10b-dihydro-2H,5H-pyrano[3,4-c]chromen-2-ylidene)acetate (**3j**).

White solid (m.p. 119.2–120.7 ◦C); <sup>1</sup>H NMR (400 MHz, CDCl3) δ 7.49–7.46 (m, 3H), 7.45–7.29 (m, 8H),7.24–7.20 (m, 2H), 5.90 (d, *J* = 2.0 Hz, 1H), 5.21 (s, 2H), 4.80 (dd, *J* = 14.8, 5.6 Hz, 1H), 3.90 (dd, *J* = 12.2, 5.6 Hz, 1H), 2.51–2.44 (m, 1H); <sup>13</sup>C NMR (101 MHz, CDCl3) δ 166.4, 163.6, 162.8, 160.7, 151.2, 135.9, 132.7, 130.9, 129.0, 128.7, 128.4, 128.3, 127.8, 127.1, 122.0, 121.8, 120.4, 102.4, 100.5, 66.4, 30.0, 25.8; [α]<sup>D</sup> 26.0 = +67.4 (c = 0.52, CHCl3); The enantiomeric purity of **3j** was determined by HPLC analysis (DAICEL CHIRALPAK AD-H (0.46 cmϕ × 25 cm), hexane:2-propanol = 60:40, flow rate = 0.75 mL/min, retention time: 14.7 min (major) and 18.0 min (minor)). HRMS (ESI): Exact mass calcd for C27H20BrO<sup>5</sup> [M+H]<sup>+</sup> : 503.0494, found: 503.0497.

Benzyl (*E*)-2-(5-oxo-4-(p-tolyl)-1,10b-dihydro-2H,5H-pyrano[3,4-c]chromen-2-ylidene) acetate (**3k**).

Yellow solid (m.p. 133.2–133.9 ◦C); <sup>1</sup>H NMR (400 MHz, CDCl3) δ 7.44–7.32 (m, 9H), 7.24–7.19 (m, 3H), 7.11 (d, *J* = 8.0 Hz, 1H), 5.91 (d, *J* = 1.2 Hz, 1H), 5.24 (s, 2H), 4.87 (dd, *J* = 14.8, 5.6 Hz, 1H), 3.99 (dd, *J* = 12.2, 6.0 Hz, 1H), 2.55–2.48 (m, 1H), 2.41 (s, 3H); <sup>13</sup>C NMR (101 MHz, CDCl3) δ 166.5, 164.3, 162.4, 161.7, 150.8, 141.3, 136.0, 130.0, 129.1, 129.0, 128.7, 128.4, 128.3, 125.7, 124.8, 122.8, 117.2, 102.0, 100.9, 66.3, 30.3, 26.0, 21.7; [α]<sup>D</sup> 26.0 = +32.2 (c = 0.54, CHCl3); The enantiomeric purity of **3k** was determined by HPLC analysis (DAI-CEL CHIRALPAK AD-H (0.46 cmϕ × 25 cm), hexane:2-propanol = 60:40, flow rate = 0.75 mL/min, retention time: 12.1 min (major) and 18.0 min (minor)). HRMS (ESI): Exact mass calcd for C28H23O<sup>5</sup> [M+H]<sup>+</sup> : 439.1545, found: 439.1543.

Benzyl (*E*)-2-(5-oxo-4-(m-tolyl)-1,10b-dihydro-2H,5H-pyrano[3,4-c]chromen-2-ylidene) acetate (**3l**).

Yellow solid (m.p. 135.6–136.8 ◦C); <sup>1</sup>H NMR (400 MHz, CDCl3) δ 7.43–7.31 (m, 9H), 7.24–7.19 (m, 3H), 7.11 (d, *J* = 8.4 Hz, 1H), 5.91 (d, *J* = 1.2 Hz, 1H), 5.24 (s, 2H), 4.87 (dd, J = 14.8, 6.0 Hz, 1H), 3.99 (dd, *J* = 12.4, 5.6 Hz, 1H). 2.55–2.48 (m, 1H), 2.41 (s, 3H); <sup>13</sup>C NMR (101 MHz, CDCl3) δ 166.5, 164.3, 162.4, 161.7, 150.8, 141.3, 136.0, 130.0, 129.1, 129.0, 128.7, 128.4, 128.3, 125.7, 124.8, 122.8, 117.2, 102.0, 100.9, 66.3, 30.3, 26.0, 21.7; [α]<sup>D</sup> 26.0 = +42.2 (c = 0.52, CHCl3); The enantiomeric purity of **3l** was determined by HPLC

analysis (DAICEL CHIRALPAK AD-H (0.46 cmϕ × 25 cm), hexane:2-propanol = 60:40, flow rate = 0.75 mL/min, retention time: 9.9 min (major) and 13.2 min (minor)). HRMS (ESI): Exact mass calcd for C28H23O<sup>5</sup> [M+H]<sup>+</sup> : 439.1545, found: 439.1544.

Benzyl (*E*)-2-(4-(4-chlorophenyl)-5-oxo-1,10b-dihydro-2H,5H-pyrano[3,4-c]chromen-2-ylidene)acetate (**3m**).

Yellow solid (m.p. 122.5–123.7 ◦C); <sup>1</sup>H NMR (400 MHz, CDCl3) δ 7.46–7.32 (m, 11H), 7.23–7.19 (m, 1H), 7.10 (d, *J* = 8.0 Hz, 1H), 5.90 (d, *J* = 1.6 Hz, 1H), 5.23 (s, 2H), 4.86 (dd, *J* = 14.8, 6.0 Hz, 1H), 4.00 (dd, *J* = 12.4, 5.6 Hz, 1H), 2.55–2.48 (m, 1H); <sup>13</sup>C NMR (101 MHz, CDCl3) δ 166.4, 163.9, 161.4, 161.1, 150.7, 137.0, 136.0, 130.5, 129.3, 128.8, 128.6, 128.5, 128.3, 125.8, 125.0, 122.5, 117.3, 102.3, 101.9, 66.4, 30.3, 26.0; [α]<sup>D</sup> 26.0 = −148.3 (c = 0.49, CHCl3); The enantiomeric purity of **3m** was determined by HPLC analysis (DAICEL CHI-RALPAK AD-H (0.46 cmϕ × 25 cm), hexane:2-propanol = 60:40, flow rate = 0.75 mL/min, retention time: 13.9 min (major) and 21.5 min (minor)). HRMS (ESI): Exact mass calcd for C27H20ClO<sup>5</sup> [M+H]<sup>+</sup> : 459.0999, found: 459.0999.

Benzyl (*E*)-2-(4-(3-chlorophenyl)-5-oxo-1,10b-dihydro-2H,5H-pyrano[3,4-c]chromen-2-ylidene)acetate (**3n**).

Yellow solid (m.p. 124.5–125.3 ◦C); <sup>1</sup>H NMR (400 MHz, CDCl3) δ 7.49–7.34 (m, 11H), 7.23–7.21 (m, 1H), 7.11–7.09 (m, 1H), 5.92 (d, *J* = 2.0 Hz, 1H), 5.24 (d, *J* = 2.0 Hz, 2H), 4.90–4.85 (m, 1H), 4.01–3.98 (m, 1H), 2.52 (t, *J* = 12.8, 1H); <sup>13</sup>C NMR (101 MHz, CDCl3) δ 166.3, 163.8, 161.1, 160.6, 150.6, 135.9, 134.7, 134.2, 131.4, 130.8, 129.5, 129.2, 129.0, 128.7, 128.4, 128.3, 127.4, 125.8, 125.0, 122.3, 117.2, 102.4, 102.2, 66.4, 30.2, 25.9; [α]<sup>D</sup> 26.0 = +100.7 (c = 0.55, CHCl3); The enantiomeric purity of **3n** was determined by HPLC analysis (DAICEL CHI-RALPAK AD-H (0.46 cmϕ × 25 cm), hexane:2-propanol = 60:40, flow rate = 0.75 mL/min, retention time: 12.4 min (major) and 13.9 min (minor)). HRMS (ESI): Exact mass calcd for C27H20ClO<sup>5</sup> [M+H]<sup>+</sup> : 459.0999, found: 459.1003.

Benzyl (*E*)-2-(4-([1,1'-biphenyl]-4-yl)-5-oxo-1,10b-dihydro-2H,5H-pyrano[3,4-c]chromen-2-ylidene)acetate (**3o**).

White solid (m.p. 130.7–131.4 ◦C); <sup>1</sup>H NMR (400 MHz, CDCl3) δ 7.66–7.60(m, 6H), 7.49–7.33 (m, 10H), 7.22 (t, *J* = 7.6 Hz, 1H), 7.13 (d, *J* = 8.0 Hz, 1H), 5.95 (s, 1H), 5.26 (s, 2H), 4.89 (dd, *J* = 14.8, 5.6 Hz, 1H), 4.03 (dd, J = 12.0, 5.6 Hz, 1H), 2.55 (t, *J* = 14.0 Hz, 1H); <sup>13</sup>C NMR (101 MHz, CDCl3) δ 166.5, 164.2, 162.0, 161.6, 150.8, 143.7, 140.3, 136.0, 129.6, 129.1, 128.9, 128.7, 128.4, 128.3, 127.3, 126.9, 125.8, 117.2, 102.1, 101.4, 66.3, 30.3, 26.0; [α]<sup>D</sup> 26.0 = +91.1 (c = 0,34, CHCl3); The enantiomeric purity of **3o** was determined by HPLC analysis (DAICEL CHIRALPAK AD-H (0.46 cmϕ × 25 cm), hexane:2-propanol = 60:40, flow rate = 0.75 mL/min, retention time: 17.9 min (major) and 31.4 min (minor)). HRMS (ESI): Exact mass calcd for C33H25O<sup>5</sup> [M+H]<sup>+</sup> : 501.1702, found: 501.1708.

Benzyl (*E*)-2-(5-oxo-4-(thiophen-3-yl)-1,10b-dihydro-2H,5H-pyrano[3,4-c]chromen-2 ylidene)acetate (**3p**).

White solid (m.p. 135.7–136.2 ◦C); <sup>1</sup>H NMR (400 MHz, CDCl3) δ 7.83 (d, *J* = 3.6 Hz, 1H), 7.53 (d, *J* = 5.2 Hz, 1H), 7.41–7.30 (m, 7H), 7.18 (t, *J* = 7.6 Hz, 1H), 7.09 (dd, J = 8.4, 4.8 Hz, 2H), 5.92 (d, *J* = 1.2 Hz, 1H), 5.27–5.20 (m,2H), 4.70 (dd, *J* = 15.0, 6.0 Hz, 1H), 4.02 (dd, J = 12.6, 5.6 Hz, 1H), 2.69–2.62 (m,1H); <sup>13</sup>C NMR (101 MHz, CDCl3) δ 166.5, 164.0, 163.0, 161.6, 155.2, 150.6, 136.0, 133.6, 132.8, 130.7, 129.1, 128.8, 128.5, 128.3, 127.3, 125.7, 124.9, 122.6, 117.1, 101.8, 100.3, 66.4, 30.9, 29.8, 26.4; [α]<sup>D</sup> 26.0 = +123.7 (c = 0.22, CHCl3); The enantiomeric purity of **3p** was determined by HPLC analysis (DAICEL CHIRALPAK AD-H (0.46 cmϕ × 25 cm), hexane:2-propanol = 60:40, flow rate = 0.75 mL/min, retention time: 12.4 min (major) and 14.6 min (minor)). HRMS (ESI): Exact mass calcd for C25H19O5S [M+H]<sup>+</sup> : 459.0999, found: 459.1003.

#### *3.3. Synthesis of* **5** *from* **1m** *and* **2b**

To the reaction mixture obtained under standard condition using **1m** (1 mmol) and **2b** (1.25 mmol) was added Pd/C (wt. 10%), then the mixture was stirred under H<sup>2</sup> atmosphere (H<sup>2</sup> balloon) at room temperature for 48 h. The resulting reaction mixture was filtered through a pad of Celite and eiluted with EtOAc. The filtration was concentrated under

reduced pressure and the residue was purified by silica gel column chromatography (PE:EtOAc = 5:1–2:1) to afford the free acid intermediate, which was then dissolved in CH2Cl<sup>2</sup> (5 mL). This solution was cooled to 0 ◦C before DCC (2.0 equiv.), DMAP (2.0 equiv.) and EtOH (2 mL) were added. After that, the reaction mixture was moved to rt and stirred overnight. After completion of the reaction by TLC analysis, the solvent was removed by rotary evaporation and the residue was directly subjected to column chromatography using PE/EtOAc (15:1-9:1) as the eluent to afford product **5**.

Ethyl (E)-2-(4-(4-chlorophenyl)-5-oxo-1,10b-dihydro-2H,5H-pyrano[3,4-c]chromen-2 ylidene)acetate (**5**) [4].

<sup>1</sup>H NMR (400 MHz, CDCl3) δ 7.46–7.31 (m, 6H), 7.21 (t, *J* = 7.6 Hz, 1H), 7.09 (d, *J* = 8.0 Hz, 1H), 5.83 (s, 1H), 4.87 (dd, *J* = 14.8, 5.6 Hz, 1H), 4.24 (q, *J* = 7.2 Hz, 2H), 4.00 (dd, *J* = 12.0, 5.6 Hz, 1H), 2.50 (t, *J* = 14.4 Hz, 1H), 1.33 (t, *J* = 7.2 Hz, 3H); <sup>13</sup>C NMR (101 MHz, CDCl3) δ 166.6, 163.4, 161.5, 161.1, 150.6, 136.9, 131.4, 130.5, 129.2, 128.6, 125.8, 125.0, 122.5, 117.2, 102.6, 101.7, 60.6, 30.3, 25.8, 14.4; [α]<sup>D</sup> 26.0 = +78.6 (c = 0.38, CHCl3); The enantiomeric purity of **5** was determined by HPLC analysis (DAICEL CHIRALPAK OD-3 (4.6 mmϕ × 150 mml), hexane:2-propanol = 80:20, flow rate = 0.75 mL/min, retention time: 6.1 min (major) and 7.6 min (minor)).

**Supplementary Materials:** The following are available online: <sup>1</sup>H and <sup>13</sup>C NMR spectra and HPLC data of compounds **3a–p** and **5**.

**Author Contributions:** X.-P.Z. conceived and designed the project and wrote the paper after discussing with Y.-Y.P. and C.-B.J.; J.-L.L., X.-H.W. and J.-C.S. performed the experiments and analyzed the data. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by National Natural Science Foundation of China (21702080), Natural Science Foundation of Jiangxi Province of China (20181BAB213003), the Open Project Program of Key Laboratory of Functional Small Organic Molecule, Ministry of Education, Jiangxi Normal University (KLFS-KF-201603) and the foundation of Jiangxi Educational Committee (170223).

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

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Samples of the compounds are not available from the authors.

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

**Sample Availability:** Samples of the compounds are not available from the authors.

#### **References**


### *Article* **Coumarin's Anti-Quorum Sensing Activity Can Be Enhanced When Combined with Other Plant-Derived Small Molecules**

**Dmitry Deryabin, Kseniya Inchagova, Elena Rusakova \* and Galimzhan Duskaev**

Federal Research Centre of Biological Systems and Agro-technologies of the Russian Academy of Sciences, Orenburg 460000, Russia; dgderyabin@yandex.ru (D.D.); ksenia.inchagova@mail.ru (K.I.); gduskaev@mail.ru (G.D.)

**\*** Correspondence: elenka\_rs@mail.ru; Tel.: +7-919-860-24-78

**Abstract:** Coumarins are class of natural aromatic compounds based on benzopyrones (2H-1 benzopyran-2-ones). They are identified as secondary metabolites in about 150 different plant species. The ability of coumarins to inhibit cell-to-cell communication in bacterial communities (quorum sensing; QS) has been previously described. Coumarin and its derivatives in plant extracts are often found together with other small molecules that show anti-QS properties too. The aim of this study was to find the most effective combinations of coumarins and small plant-derived molecules identified in various plants extracts that inhibit QS in *Chromobacterium violaceum* ATCC 31532 violacein production bioassay. The coumarin and its derivatives: 7-hydroxycoumarin, 7.8-dihydroxy-4-methylcoumarin, were included in the study. Combinations of coumarins with gamma-octalactone, 4-hexyl-1.3-benzenediol, 3.4.5-trimethoxyphenol and vanillin, previously identified in oak bark (*Quercus cortex*), and eucalyptus leaves (*Eucalyptus viminalis*) extracts, were analyzed in a bioassay. When testing two-component compositions, it was shown that 7.8 dihydroxy-4-methylcoumarin, 4-hexyl-1.3-benzendiol, and gamma-octalactone showed a supra-additive anti-QS effect. Combinations of all three molecules resulted in a three- to five-fold reduction in the concentration of each compound needed to achieve EC<sup>50</sup> (half maximal effective concentration) against QS in *C. violaceum* ATCC 31532.

**Keywords:** coumarins; quorum sensing; QS inhibitors; plant-derived molecules; *Chromobacterium violaceum*

#### **1. Introduction**

Coumarins are a class of natural compounds based on benzopyrones (2H-1-benzopyran-2-ones) [1]. These compounds can be classified depending on the core's structure and the presence of substituents. There are the "simplest" coumarins (e.g., coumarin and dihydrocoumarin), followed by oxy-, meth-oxy-, and methylenedioxycoumarins with various substitutions in benzene/pyron rings (e.g., umbelliferon, 3-hydroxycoumarin, and scopoletin). The furancoumarins (e.g., bergamotin) contain an additional condensed furan core. Other, more structurally complex compounds are the result of coumarin condensation with pyran, benzene, and benzofuran rings. Most of the compounds of this class in plants are found in the free state, and only a small number are found in glycosides with D-glucose attached to the C6, C7, or C8 atoms of the coumarin nucleus [2].

Currently, coumarins are identified as secondary metabolites in about 150 different plant species distributed in almost 30 families, of which the most important are *Rutaceae, Umbelliferae, Clusiaceae, Guttiferae, Caprifoliaceae, Oleaceae, Nyctaginaceae,* and *Apiaceae* [3]. These substances are synthesized from phenylalanine via the shikimic acid formation pathway (hydroxylation, glycolysis, and cyclization of cinnamic acid) [4], and often, several different coumarins are found in the same plant.

Coumarins proposed for medical use due to their proven biological activity. They are showed anti-ulcerogenic [5], antiparasitic [6], anti-inflammatory [7–11], and other properties [12–14]. They are also antioxidant [15], and anticoagulant compounds [16–19]. As such, they can be defined as new pharmaceutical candidates [20].

**Citation:** Deryabin, D.; Inchagova, K.; Rusakova, E.; Duskaev, G. Coumarin's Anti-Quorum Sensing Activity Can Be Enhanced When Combined with Other Plant-Derived Small Molecules. *Molecules* **2021**, *26*, 208. https://doi.org/10.3390/ molecules26010208

Academic Editor: Maria João Matos Received: 30 November 2020 Accepted: 18 December 2020 Published: 3 January 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/).

The ability of coumarins to inhibit cell-to-cell communication in bacterial communities better known as "quorum sensing" (QS)—has been discovered relatively recently. Briefly: QS is a special type of regulator of bacterial gene expression that functions at a high microbial population density. Depending on the chemical nature of the autoinducer, QS can be divided into several types: 1) LuxI/LuxR type (autoinducers–acylated homoserin lactones); 2) type II QS systems (autoinducers–furanone derivatives); 3) QS systems with Gram-positive bacteria (autoinducers–short oligopeptides); 4) QS systems with autoinducers of various natures (e.g., epinephrine, norepinephrine). The first of the described and most common QS systems is a two-component system of the LuxI/LuxR type inherent in many bacterial pathogens [21] where it activates the synthesis of virulence factors and the biofilms formation.

Because the search for plant-derived molecules with anti-QS activity is very actual, the coumarins are interesting object for this screening. Experimental observations of the anti-QS activity of coumarin are mainly related to Gram-negative bacteria that use a LuxI/LuxR type communication system, e.g., *Pseudomonas. aeruginosa* (in which coumarin suppress of phenazine biosynthesis, and motility) and *Aliivibrio fischeri* (coumarin inhibits the bioluminescence) [22]. Another simple coumarin, i.e., dihydrocoumarin, effectively inhibited QS-dependent biosynthesis of violacein in *Chromobacterium violaceum* [23]. The subsequent comparative analysis of seven hydroxycoumarin derivatives in relation to the violacein biosynthesis in *C. violaceum* showed that the promising anti-QS effect is characteristic of 3-hydroxycoumarin [24]. The identification of other functional substitutions of coumarin core, which led to disruption of the bacterial biofilm formation and, at the same time, inhibits QS development, was done by Reen [4]. This variant of bioactivity was also characteristic of a larger group of compounds, including furanocumarins: bergamottin and 6.7-dihydroxybergamottin. Interestingly, the last two compounds found in citrus fruits (*Citrus bergamia, Citrus maxima*, and *Citrus* × *paradisi*) showed their activity against bacteria that use both LuxI/LuxR-type and type II QS system [25].

Significantly, coumarin and its derivatives are often found together with other plantderived small molecules that also have anti-QS properties. At the same time, our previous studies have shown that such molecules can act synergistically in a single plant [26,27]; however, until now, the possible combination of coumarins and small plant-derived molecules that are part of various plant extracts remains open.

The aim of this study was to find the most effective combinations of coumarins with small plant-derived molecules previously identified in extracts of oak bark (*Quercus cortex)*, and eucalyptus leaves (*Eucalyptus viminalis*) to inhibit LuxI/LuxR-type quorum sensing in *C. violaceum* ATCC (American Type Culture Collection) 31532.

#### **2. Results**

#### *2.1. Effect of Coumarin and Its Derivatives in Chromobacterium violaceum ATCC 31532 Violacein Production Bioassay*

Cultivation of *C. violaceum* ATCC 31532 with coumarin, 7-hydroxycoumarin, and 7.8 dihydroxy-4-methylcoumarin followed registration of the optical density values (OP450) of bacterial biomass and violacein production (OP600), allowed us to evaluate their effect on the growth and QS-dependent biosynthesis in the bioassay. All tested components showed antibacterial activity as follows: minimum inhibitory concentration (MIC50) = 2.689 mg/mL and MIC<sup>100</sup> = 3.650 mg/mL for coumarin; MIC<sup>50</sup> = 0.497 mg/mL and MIC<sup>100</sup> = 1.267 mg/mL for 7-hydroxycoumarin and MIC<sup>50</sup> = 0.325 mg/mL and MIC<sup>100</sup> = 2.400 mg/mL for 7.8-dihydroxy-4-methylcoumarin. Simultaneously, sub-inhibitory concentrations of these compounds provided an anti-QS effect evaluated by inhibition on pigment violacein production, which was expressed as follows: effective concentration (EC50) = 1.105 mg/mL and EC<sup>100</sup> = 3.650 mg/mL for coumarin; EC<sup>50</sup> = 0.199 mg/mL and EC<sup>100</sup> = 0.633 mg/mL for 7-hydroxycoumarin and EC<sup>50</sup> = 0.150 mg/mL and EC<sup>100</sup> = 1.200 mg/mL for 7.8-dihydroxy-4-methylcoumarin (Table 1). Thus, the highest anti-QS activity was demonstrated by 7.8-dihydroxy-4 methylcoumarin, and this coumarin derivative was taken for further studies on the combination of small plant-derived molecules.


**Table 1.** Effects of coumarin, 7-hydroxycoumarin and 7.8-dihydroxy-4-methylcoumarin (mg/mL) on growth and QS-controlled violacein pigment biosynthesis in *C. violaceum* ATCC 31532.


Results were obtained for combinations of 7.8-dihydroxy-4-methylcumarin (identified in *Baikal skullcap* extract) with small plant-derived molecules from oak bark (4-hexyl-1.3 benzenediol, 3.4.5-trimethoxyphenol, vanillin) or in eucalyptus leaves (gamma-octalactone), which have own anti-QS activity preliminary in *C. violaceum* ATCC 3153 bioassay. While most coumarins combinations showed simple additivity, some two-component mixtures led to pronounced mutual strengthening of anti-QS activity, which was evaluated as a synergetic (supra-additive) effect in 2D isobolographic analysis. The supra-additivity was revealed in combination of 7.8-dihydroxy-4-methylcoumarin with 4-hexyl-1.3-benzenediol (Figure 1a) as well as in 7.8-dihydroxy-4-methylcoumarin and gamma-octalactone mixtures (Figure 1b), where cultivation of *C. violaceum* ATCC 31532 in media enriched these paired molecular compositions showed a two-to four-fold decrease in the concentrations of each compound to achieve 50% inhibition of QS-controlled violacein biosynthesis. In turn, some combination of small molecules from oak bark and eucalyptus leaves showed the antagonistic (infra-additive) effect achieved in the compositions gamma-octalactone and vanillin, gamma-octalactone and 3.4.5-trimethoxyphenol, while the combinations gammaoctalactone and 4-hexyl-1.3-benzenediol have a supra-additive effect.

**Figure 1.** 2D isobolographic analysis of the combined use of 7.8-dihydroxy-4-methylcoumarin and 4-hexyl-1.3-benzenediol (**a**), 7.8-dihydroxy-4-methylcoumarin and gamma-octalactone (**b**) on the QS-controlled violacein biosynthesis in *C. violaceum* ATCC 31532. Isoboles are represented as straight lines connecting the EC<sup>50</sup> concentrations of each compounds. The points under the isoboles correspond to the supra-additive effect.

On this basis the further research on the combined use of the small molecules from Baikal skullcap, oak bark, and eucalyptus leaves included 7.8-dihydroxy-4-methylcoumarin, 4-hexyl-1.3-benzenediol, and gamma-octalactone, because all pairwise combination of these compounds showed supra-additive anti-QS effect in *C. violaceum* ATCC 31532 bioassay.

#### *2.3. Evaluation of the Effect of a Three-Component Composition of Small Plant-Derived Molecules on the Quorum Sensing in C. violaceum ATCC 31532*

The bioassay of small plant-derived molecules composition, which included various ratios of 7.8-dihydroxy-4-methylcoumarin, 4-hexyl-1.3-benzenediol and gamma-octalactone, confirmed supra-additive effect of two-component composition, and first showed synergy of three-component composition which manifested in the location of majority of the experimental points below the 3D isobole plane (Figure 2).

**Figure 2.** 3D isobolographic analysis of the combined use of 7.8-dihydroxy-4-methylcoumarin, 4-hexyl-1.3-benzenediol, and gamma-octalactone against the QS-controlled violacein biosynthesis in *C. violaceum* ATCC 31532: (**a**) front view; (**b**) back view. The 3D isobole is represented as a blue triangle, the vertices of which correspond to 50% violacein biosynthesis inhibition (EC50) for each compound; the "sail" plane show the supra-additive effect of the experimental samples under the 3D isobole plane.

Figure 2 shows a 3D isobole in the form of a triangle, the vertices of which connect the concentrations of each compounds that cause the same biological effect EC<sup>50</sup> (50% inhibition of violacein biosynthesis in *C. violaceum* ATCC 31532 bioassay). At the point of maximum supra-additive effect, with the ratio of tested compounds set to 0.6:1:0.8, the concentrations of each compound were three- to five- lower than the concentrations of individual compounds required to achieve EC50. Importantly, the supra-additive effect was detected in at least 85% of samples with various rations of 7.8-dihydroxy-4-methylcoumarin, 4-hexyl-1.3-benzenediol and gamma-octalactone.

Thus, the results of our study described original compositions of various in structure small plant-derived molecules of different origins: 7.8-dihydroxy-4-methylcoumarin from Baikal skullcap, 4-hexyl-1.3-benzenediol from oak bark, and gamma-octalactone from eucalyptus leaves, which enhance each other's anti-quorum activity. In such composition, the content of coumarin's derivative can be significantly decreased while maintaining the anti-QS effect, which makes it possible to avoid unfavorable manifestations of the bioactivity of this group of compounds.

#### **3. Discussion**

Coumarins have a wide range of biological properties including antiviral, antimicrobial, anti-inflammatory, and other bioactivities. Some coumarins are approved for use in the treatment of various diseases [28–31]. The most important are vitamin K antagonists, such as warfarin, phenprocumone, or acenocumarol, which are used as anticoagulants [32,33]. Numerous studies have also shown that these compounds do not exhibit significant toxicity to humans and animals (LD<sup>50</sup> = 275 mg/kg), and are only moderately toxic to the liver and kidneys [34]. In this study we used coumarin and its derivatives at concentrations significantly lower than their LD<sup>50</sup> for mammals [35–37].

The novel variant of coumarins bioactivity is anti-QS effect that disrupt cell-to-cell chemical communication in bacteria. In this study we continued this direction and followed the path of analyzing the coumarins compositions with other plant-derived molecules in order to enhance the anti-QS effect.

Using *C. violaceum* ATCC 31532 bioassay we found anti-QS effect at sub-inhibitory concentrations of coumarin, 7-hydroxycoumarin, and 7.8-dihydroxy-4-methylcoumarin, that is in good agreement with the same activity of other coumarin derivatives: esculetin (6.7-dihydroxycoumarin) [38,39], scopoletin (7-hydroxy-5-methoxycoumarin) [40], furanocoumarin [25], nodakenetin, fraxin [41], and fizetin [42]. This allows us to state the universality of this bioactivity variant for compounds in this group.

Important, that 7.8-dihydroxy-4-methylcoumarin was characterized as the most effective anti-QS compound in this study. This data has not been previously reported anywhere, whereas the described 7.8-dihydroxy-4-methylcoumarin bioactivity comprised its antioxidant properties only [43,44]. At the same time, its structural features, particularly the hydroxy groups positions, well-corresponded to all known anti-QS active coumarins [23–25]. Our results were consistent with those of Yang et al. which noted a significant increase in the antibacterial effect upon the hydroxylation of coumarins at positions 6, 7, or 8 [45]. Our data also partially agreed with the studies of Lee et al., who showed that hydroxylation at position 7 increased anti-QS activity, while dihydroxylation of coumarin at positions 6 and 7 decreased this activity in comparison to conventional coumarin [46].

The next step was to combine 7.8-dihydroxy-4-methylcoumarin with other small plant-derived molecules in order to get mutual potentiation of the final anti-QS effect. The originality of the proposed approach was that we combined molecules from different plant sources. When testing two-component compositions, it was shown for the first time that 7.8-dihydroxy-4-methylcoumarin, 4-hexyl-1.3-benzendiol, and gamma-octalactone demonstrated a synergetic (supra-additive) anti-Qs effect, and combining all three molecules together decreased the concentration of each compound required to achieve EC<sup>50</sup> in the composition by three-to-five-fold.

Discussing the mechanism of revealed super-additive effect we assumed that it based on the complementary bioactivity mechanisms for each compound (Figure 3). In this concept, gamma-octolactone is structurally close to LuxI/LuxR quorum sensing autoinducers (acylated homoserine lactones) and probably interferes with him for receptor binding. The 4-hexyl-1.3-benzenediol have a not fully identified mechanism, shown in one of our previous study [47,48], which repress the sensitivity of bacterial cells to autoinducers. Coumarins are characterized by a special mechanism through inhibition of the metabolism of cyclic 3',5'-diguanilate (c-di-GMP), an intracellular intermediate that is involved in the regulation of bacterial exopolysaccharide synthesis, biofilm formation, adhesion, and virulence [24]. Doing together, these three compounds block the quorum sensing development at different stages, which is manifested in their super-additive anti-QS effect (Figure 3).

**Figure 3.** Proposed mechanism of supra-additive anti-QS effect of molecular composition consists of 7.8-dihydroxy-4-methylcoumarin, 4-hexyl-1.3-benzendiol, and gamma-octalactone.

The practical aspect of these results assumes the combined use of coumarin derivatives and other small plant-derived molecules to combat bacterial pathogens of plants, animals, and humans that use quorum sensing systems for the induction of virulence factors and biofilm formation. The implementation of this approach is to use an artificial molecular composition consists of 7.8-dihydroxy-4-methylcoumarin, 4-hexyl-1.3-benzendiol, and gamma-octalactone or the plant materials mixtures with high content of these compounds: Baikal skullcap (*Scutellaria baicalensis*), oak bark (*Quercus cortex*), and eucalyptus leaves (*Eucalyptus viminalis*). Due to the high biological activity of these compositions, they can become a substitute for antibiotics in feeding of farm animals, and should also be considered as candidate pharmaceuticals for further preclinical and clinical studies.

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

#### *4.1. Chemical Compounds*

Coumarin and its derivatives were used to inhibit QS in *Chromobacterium violaceum* ATCC 31532: coumarin (2H-chromene-2OH; CAS 91-64-5) (Figure 4A), 7-hydroxycoumarin (7-hydroxy-2H-1-benzopyran-2-one; CAS 93-35-6) (Figure 4B), and 7.8-dihydroxy-4 methylcoumarin (4-methyldafnetin; CAS 2107-77-9) (Figure 4C).

**Figure 4.** Structural formula of coumarin (**A**), 7-hydroxycoumarin (**B**), and 7.8-dihydroxy-4-methylcoumarin (**C**).

Small plant-derived molecules with previously reported anti-QS activity that were identified in extracts of oak bark (*Quercus cortex*), and eucalyptus leaves (*Eucalyptus viminalis*), were tested in combinations with coumarin derivatives. The analysis included gammaoctalactone (2(3H)-furanone; CAS 147852-83-3), 4-hexyl-1.3-benzenediol (4-n-propylresorcinol; CAS 13331-19-6), 3.4.5-trimethoxyphenol (antiarol; CAS 642-71-7), vanillin (4-hydroxy-3 methoxy benzaldehyde; CAS 121-33-5).

С

Each of these compounds had a purity at least 99% and was purchased from Sigma-Aldrich (St. Louis, MO, USA).

#### *4.2. Bacterial Strain*

The wild strain of *C. violaceum* ATCC 31532 that possessed a two-component LuxI/LuxRtype QS system, was used in bioassay. In this strain CviI synthase (LuxI analog) produce autoinducer N-hexanoyl-L-homoserin lactone (C6-AHL) which bond CviR receptor protein (LuxR analog) and activate QS-controlled transcription of several target genes including *vioABEDC* operon [49]. The encoded VioA, VioB, VioE, VioD, and VioC proteins form a biosynthetic pathway for blue-violet pigment violacein with a maximum absorption at 585 nm. The amount of pigment in the bacterial culture allowed us to directly assess the QS activity.

#### *4.3. Methods for Investigating Anti-QS Activity of Coumarin Derivatives in C. violaceum ATCC 31532 Bioassay*

To determine the anti-QS activity of each compound in Luria-Bertani (LB) broth, double dilutions (n × 2) were prepared. The similar samples of LB-broth that did not contain tested compounds were used as positive (growth of the test strain) and negative (sterile) controls. Glass vessels containing 2 mL of experimental dilutions or control samples were inoculated 20 µl of a one-day *C. violaceum* ATCC 31532 culture and cultivated in a static mode at 27 ◦C. The results were evaluated using a multifunctional microplate reader Infinite 200 PRO (Tecan, Männedorf, Switzerland). The optical density at 450 ± 5 nm (OP450) measured the bacterial biomass and evaluated the effect of the studied compounds on bacterial growth, while the violacein pigment after its ethanol extraction was determined at 600 ± 5 nm (OP600), which was an indicator of the effect on the QS system. The absorption values of the negative control were subtracted. The antibacterial effect of the studied compounds was presented by the MIC<sup>100</sup> and MIC<sup>50</sup> values, which were minimal inhibitory concentrations that caused 100% and 50% growth suppression for the test strain relative to the positive control. The inhibition of quorum sensing was expressed as EC<sup>100</sup> and EC<sup>50</sup> values, which were equal to 100% and 50% inhibition of violacein pigment biosynthesis in grown culture, respectively.

#### *4.4. Evaluation of the Combined Use of Coumarins and Small Plant-Derived Molecules against in C. violaceum ATCC 31532*

To examine the combined effect, double dilutions of test compounds were introduced into plastic 96-well plates in perpendicular directions (connection X: connection Y), so that each well contained their individual ratio. Comparison samples were a series of dilutions containing only one of the tested compounds, as well as positive and negative controls. Further inoculation of *C. violaceum* ATCC 31532, cultivation, and recording of the study results were performed as described above. The effect of the paired compositions was evaluated using isobolographic analysis [49], which is based on the construction of 2D isoboles (i.e., lines connecting the EC<sup>50</sup> values for the studied compounds X and Y, on the abscissa and ordinate axes) followed by drawing points on this graph which corresponding to the combined effect of compounds X and Y at different concentration ratios. The location of such points on the isobole line corresponded to an additivity (summation effect), their placement above the isobole line described an infra-additive effect (antagonism), and below the isobole showed a supra-additive (synergetic) effect.

Studies of a three-component composition were beginning from the formation of a series of 96-well plates containing the double dilutions of compounds X and Y, as described above. On the next step the wells in each plate were filled with a certain concentration of compound Z (i.e., the total number of used 96-well plates was equal to the number of tested concentrations of compound Z). The control samples were a dilution series containing only compound X, only compound Y, or only compound Z, as well as positive and negative controls. Wells were inoculated with *C. violaceum* ATCC 31532, cultivated and analyzed as described above. The effect of the three-component compositions was evaluated using three-dimensional (3D) isoboles plotted based on the EC<sup>50</sup> values for each compound.

#### *4.5. Statistical Analysis*

All values presented a mean of the 5 experiments. The obtained results were processed using methods of statistical variance in Excel for Windows 10.

#### **5. Conclusions**

In this study, coumarin and its derivatives were tested against quorum sensing in *Chromobacterium violaceum* ATCC 31532, and promising activity was shown in 7.8-dihydroxy-4 methylcoumarin. This compound previously detected in Baikal skullcap (*Scutellaria baicalensis*) was combined with other small plant-derived molecules identified in extracts of oak bark (*Quericus cortex*), and eucalyptus leaves (*Eucalyptus viminalis*). It has been shown that 7.8 dihydroxy-4-methylcoumarin, 4-hexyl-1.3-benzenediol and gamma-octalactone exhibit a supra-additive anti-QS effect in two-component combinations, and the combination of all three molecules reduces the concentration of each of them required for reaching EC<sup>50</sup> against QS, by three- to five- times. It was proposed that the super-additive effect is based on various bioactivity mechanisms of tested molecules, which disrupt the QS development at different stages. The results provide a use for small plant-derived molecule compositions plant materials in the feeding of farm animals, replacing the similar use of prohibited feed antibiotics [50], and also determines the prospects for their testing against human pathogens that use QS to induce virulence factors and biofilm development.

**Author Contributions:** Conceptualization, D.D.; methodology, D.D. and K.I.; software, K.I.; formal analysis, K.I.; writing—original draft preparation, E.R.; writing—review and editing, D.D.; project administration, E.R. and G.D. All authors have read and agreed to the published version of the manuscript.

**Funding:** The research of the activity of coumarin and its derivatives was conducted with financial support from the research plan for 2019–2021 of the Federal state budgetary Research Center BST RAS within the framework of the thematic plan for state task No. 0526-2019-0002. The research of supra-additive effects of coumarins and plant-derived molecules was conducted with financial support from the Russian Science Foundation (Grant #21-16-00112).

**Informed Consent Statement:** Not applicable.

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

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

**Sample Availability:** The materials used in the study are commercially available and can be purchased from the relevant firms.

#### **References**


### *Article* **Synthesis of Ferulenol by Engineered** *Escherichia coli***: Structural Elucidation by Using the In Silico Tools**

**Anuwatchakij Klamrak <sup>1</sup> , Jaran Nabnueangsap <sup>2</sup> , Ploenthip Puthongking <sup>1</sup> and Natsajee Nualkaew 1,\***


**\*** Correspondence: nnatsa@kku.ac.th; Tel.: +66-43-202178

**Abstract:** 4-Hydroxycoumarin (4HC) has been used as a lead compound for the chemical synthesis of various bioactive substances and drugs. Its prenylated derivatives exhibit potent antibacterial, antitubercular, anticoagulant, and anti-cancer activities. In doing this, *E. coli* BL21(DE3)pLysS strain was engineered as the in vivo prenylation system to produce the farnesyl derivatives of 4HC by coexpressing the genes encoding *Aspergillus terreus* aromatic prenyltransferase (AtaPT) and truncated 1-deoxy-D-xylose 5-phosphate synthase of *Croton stellatopilosus* (CstDXS), where 4HC was the fed precursor. Based on the high-resolution LC-ESI(±)-QTOF-MS/MS with the use of in silico tools (e.g., MetFrag, SIRIUS (version 4.8.2), CSI:FingerID, and CANOPUS), the first major prenylated product (named compound-1) was detected and ultimately elucidated as ferulenol, in which information concerning the correct molecular formula, chemical structure, substructures, and classifications were obtained. The prenylated product (named compound-2) was also detected as the minor product, where this structure proposed to be the isomeric structure of ferulenol formed via the tautomerization. Note that both products were secreted into the culture medium of the recombinant *E. coli* and could be produced without the external supply of prenyl precursors. The results suggested the potential use of this engineered pathway for synthesizing the farnesylated-4HC derivatives, especially ferulenol.

**Keywords:** *Escherichia coli*; biotransformation; 4-hydroxycoumarin; ferulenol; structural annotation; in silico tools

#### **1. Introduction**

Prenylation is one of the post-structural modifications essential for the increasing biological activities of several natural products [1]. Transferring the prenyl moieties onto the aromatic acceptor molecules often leds to prenylated derivatives with greatly improved therapeutic potency [2]. This process has become a new frontier for developing novel drugs and lead compounds in the pharmaceutical industry, especially for antimicrobial, antioxidant, anti-inflammatory, and anti-cancer agents [3–5]. Nevertheless, plant-based production of these valuable products is limited by finite resources, low yields, slow growth rates, seasonal dependency, and rare, or completely absent in some regions [6,7]. Moreover, the chemoenzymatic and total synthesis of the prenylated products is rather difficult, as it is challenged by the structural complexity which requires multiple steps of the uncontrollable regio- and stereoselective prenylation, and expensive starting materials [8,9].

Ferulenol (2), a C-3 farnesylated 4-hydroxycoumarin, is a major constituent in *Ferula communis* (Giant fennel) which possesses many biological activities [5,10]. According to previous studies, this compound exhibits cytotoxicity against various lines of cancer cells including human breast (MCF-7), colon (Caco-2), ovarian (SK-OV-3), and leukemic (HL- 60) in a dose-dependent manner, with the mode of action resembling paclitaxel (taxol) [11,12]. Ferulenol (2) also exerts anti-cancer activity through the downregulation of Bcl2 protein along with upregulation of Bax protein in benzo[a]pyrene-induced lung cancer in a rat

**Citation:** Klamrak, A.; Nabnueangsap, J.; Puthongking, P.; Nualkaew, N. Synthesis of Ferulenol by Engineered *Escherichia coli*: Structural Elucidation by Using the In Silico Tools. *Molecules* **2021**, *26*, 6264. https://doi.org/10.3390/ molecules26206264

Academic Editors: Maria João Matos and Valeria Costantino

Received: 11 July 2021 Accepted: 13 October 2021 Published: 16 October 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/).

model [13]. This indicated ferulenol (2) as a pro-oxidant and chemotherapeutic agent and has been recognized as an interesting lead compound for anti-cancer semi-synthesis [13,14]. In searching for the novel anticoagulant warfarin derivatives, ferulenol (2) exhibits higher activity than the warfarin drug (approximately 22 times) with lower toxicity [15,16]. This compound also possesses antimycobacterial activity against fast-growing *Mycobacterium* species [17]. Although chemical synthesis of this compound could be achieved by the reaction between 4-hydroxycoumarin sodium salt (4HCNa) and all trans-farnesylchloride based on alkylation at the C-3 of coumarin [15], these processes are quite complicated, and the starting precursors are rather expensive.

Engineering microbial cells as the in vivo prenylation system provides several advantages over the two existing methods, as microbes can be grown very fast in the noncomplex medium and can be easily extended to large-scale production [9,18]. More importantly, microbes can supply various lengths of prenyl donors, e.g., dimethylallyl pyrophosphate (DMAPP, C5), geranyl pyrophosphate (GPP, C10), farnesyl pyrophosphate (FPP, C15), and geranylgeranyl pyrophosphate (GGPP, C20), through their inherent isoprenoid pathways [7,19]. *E. coli* solely synthesizes isoprenoids via the 1-deoxy-D-xylulose 5-phosphate (DXP) pathway (also known as non-mevalonate pathway), in which 1-deoxy-D-xylulose 5-phosphate synthase (DXS) is known as the first committing step enzyme that controls the metabolic flux of isoprenoids precursors [20–24]. In wild type *E. coli*, FPP serves as the key branching point in the synthesis of the vital molecules, i.e., heme O, ubiquinone, and peptidoglycan [25–28]. Consequently, the metabolic flux of this precursor in *E. coli* has been redirected towards taxadiene, carotenoids, and amorpha-4,11-diene through heterologous expression of various terpene synthases [29–34]. In the last few years, using microorganisms as the prenylation systems for the production of hybrid molecules containing prenyl moieties has focused mainly on yeasts (e.g., *Saccharomyces cerevisiae* and *Pichia pastoris*) and *Bacillus subtilis* based on the endogenous prenyl donors supplied via either the mevalonate (MVA) pathway or the DXP pathway [35–38]. However, an attempt at synthesizing ferulenol (2), the product of C-3 farnesylation of 4HC (1), through the genetic manipulation of the DXP pathway in *E. coli* has not been reported yet. Therefore, this gives rise to our interest in establishing a new synthesis pathway for producing this product by using the engineered microbe based on feeding of 4HC (1) to minimize chemical consumption.

Aromatic prenyltransferases (aPTs) are the enzymes that catalyze the regio-selective prenylation of the prenyl groups (so-called prenyl donors), incorporating the aromatic compounds (known as aromatic acceptors) that contain the electron-rich regions through a mechanism comparable to Friedel–Crafts aromatic electrophilic substitution [39–42]. Nowadays, the genes encoded for aPTs have been isolated from plants, fungi, and bacteria and have been found to exhibit broad substrates with specificity both in terms of aromatic acceptors and prenyl donors, creating a diverse range of prenylated products [1–3,43–45]. Of those enzymes characterized, *Aspergillus terreus* aromatic prenyltransferase (AtaPT) possesses the broad range substrate specificity towards various types of aromatic acceptors (e.g., coumarins, resveratrol, and naringenin) and prenyl donors, i.e., DMAPP, GPP, FPP, and GGPP, yielding prenylated products with structural diversity [46]. This enzyme also differs from formerly characterized aPTs, as it can generate mono-, di-, and/or triprenylated products via C-C- and/or C-O-bonded prenylation. Since 4-hydroxycoumarin (4HC) (1) contains the two promising regions for electrophilic alkylation, including the C3 and the oxygen atom at C4 positions [15,47], it was proposed to be utilized by AtaPT due to the unique catalytic activity of this enzyme. In wild-type *E. coli*, its native isoprenoid precursors are always insufficient for pathway engineering purposes, therefore driving metabolic fluxes via overexpression of the rate-limiting step enzymes in the DXP pathway, e.g., DXS, 1-deoxy-D-xylulose 5-phosphate reductoisomerase (DXR), and isopentenyl pyrophosphate isomerase (IDI) are required [19–24].. According to the previous findings, transcriptional profiling analysis revealed a positive correlation between CsDXS gene expression and plaunotol (acyclic diterpene alcohol) content in the young leaves of *Croton stellatopilosus*, and this enzyme has been suggested to control the metabolic flux of isoprenoid precursors

in that plant [48]. Therefore, the truncated CsDXS (namely CstDXS) was chosen to coexpress with the AtaPT in *E. coli* BL21(DE3)pLysS to establish a new synthetic route for the in vivo formation ferulenol (2), in which 4HC (1) was only a fed precursor. [ ]. ( ) - ( ) ( ) ( ) .

. . - - - - ( )

( ) [ – ]..

( )

In this study, *E. coli* was used as the bioconversion system of the newly designed pathway leading to the formation of farnesylated-4HC analogs, which includes two steps (Figure 1). The CstDXS was overexpressed for driving metabolic flux of isoprenoid precursors (e.g., GPP, FPP) in the DXP pathway [47]. Second, prenylation of those prenyls to the core structure of 4HC (1) through the catalytic function of AtaPT produced its corresponded products (2 and 3) [46]. This process requires magnesium ion (Mg2+) as a cofactor to facilitate the formation of an electrophilic prenyl carbocation, which subsequently binds to 4HC (1) via the Friedel-Crafts reaction [46]. - ( ). ( . . ) [ ]. ( ) ( ) [ ]. ( + ) ( ) - [ ].

**.** ( ) ( ) ( ) - - ( ) . ( . . ) . = = - = = = = = = . **Figure 1.** Proposed biosynthetic pathway of ferulenol (2) and its isomeric structure (3) reconstructed in *E. coli* BL21(DE3)pLysS carrying pCDFDuet-AtaPT-CstDXS, where 4HC (1) was the fed precursor. Note that both products can be formed without any supply of prenyl donors (i.e., FPP) and secreted into the culture medium. PYR = pyruvate; G3P = glyceraldehyde 3-phosphate; IPP = isopentenyl pyrophosphate; DMAPP = dimethylallyl pyrophosphate; GPP = geranyl pyrophosphate; FPP = farnesyl pyrophosphate; IDI = isopentenyl pyrophosphate isomerase; IspA = geranyl pyrophosphate synthase and farnesyl pyrophosphate synthase.


of the prenylated-4HC derivatives obtained from the bioconversion of 4HC (1) by the clones carrying pCDFDuet-AtaPT-CstDXS to verify the product formation.

Here, we report the newly artificial pathway for synthesizing ferulenol (2) established in the *E. coli* BL21(DE3)pLysS strain, where the chemical structures of this product are entirely elucidated by using in silico tools, included MetFrag [52], SIRIUS [54], CSI:FingerID web service [55], and CANOPUS [56]. The experimental mass spectra for the putative ferulenol (2) are further confirmed by the alignment to the mass spectra of the authentic ferulenol (2) established using the same LC-MS/MS condition, along with those established by Fourel and colleagues [57]. The putative mass peak corresponds to farnesylated derivative of 2-hydroxy-4-chromenone, which has been proposed to exist via the tautomerization of ferulenol (2) and is detected in this study. The results can be used to further establish the *E. coli* system as the microbial cell factory for producing the farnesylated-4HC derivatives to serve drug discovery purposes.

#### **2. Results**

] −

.

.

( ) . .

). ( )


( ) . . /

#### *2.1. Construction of Plasmid pCDFDuet-AtaPT-CstDXS*

Engineering *E. coli* as the in vivo prenylation system requires at least two steps: (1) overexpression of CstDXS as the rate-limiting step enzyme in the DXP pathway to increase the available pool of prenyl-donors; and (2) transferring the prenyl-donors into the core aromatic acceptor by the catalytic function of AtaPT. Therefore, the truncated DXS (CstDXS: GenBank accession no. AB354578.1) was used for enhancing the flux of isoprenoid precursors in *E. coli*. The gene encoding for AtaPT (GenBank accession no. KP893683) was chosen for incorporating the prenyl donors into the 4HC core structure. Construction of pCDFDuet-AtaPT-CstDXS was achieved via stepwise incorporation of AtaPT, followed by CstDXS into the multiple cloning site 1 (MCS-1) and MCS-2 of pCDFDuet-1 vector, respectively. DNA sequencing confirmed that both genes were inserted into the correct regions of the pCDFDuet-1 vector with no frame-shifted insertion. Each gene was independently controlled by their T7 promotor, and gene expression was driven by adding IPTG into the culture (Figures S1 and S2). The 5' end of AtaPT in the MCS-1 (*BamH*I/*Not*I) was joined with His-tag, and the 3′ end of CstDXS was fused with the S-tag. The map created using GenScript (https://www.genscript.com/gensmart-design/# (accessed on 26 September 2021)) is depicted in Figure 2.

**.** - - . **Figure 2.** Map of recombinant plasmid pCDFDuet-AtaPT-CstDXS.

/ ( )


( )

( )

 [ ]. ( - - ) . .

). / .

:

( ) - - -

( ). .

( ). ( - ) ( )

( ) - ( ) - - - ( ) ( ).

( ) - - - ( )



( ) - -


. . [ ]

(

( ) . [ −

" ( )"

−


#### *2.2. Identification of Prenylated 4HC Derivative Using LC-ESI(-)-QTOF-MS/MS*

The high-resolution LC-ESI-QTOF-MS/MS was used for the identification of the prenylated 4HC derivative produced by clones carrying pCDFDuet-AtaPT-CstDXS. The in vivo formation for the peak corresponding to the 4HC containing farnesyl moiety was detected on the extracted ion chromatogram (EIC) at the calculated *m/z* 366.211670 [M − H]−, in which the product was tentatively confirmed to be "ferulenol (2)" by comparing the MS/MS fragmentation profile against the authentic ferulenol (2) established by the previous work [57]. The results showed that there were two mass peaks (namely compound-1 and -2) which exhibited the identical *m/z* 365.21 eluted at the RT 21.3 min and 22.3 min, respectively (Figure 3A). No product formation was observed in the culture medium clones harboring pCDFDuet-AtaPT-CstDXS grown without supplementing 4HC (1) in the culture (Figure 3B). These products were therefore presumably farnesylated-4HC, where the farnesyl moiety was incorporated on the C3 position of 4HC (1) via C-C-bonded formation. Based on the direct comparison against previous mass spectra, only the compound-1 highly resembled the MS/MS profile to that of ferulenol (2) reported by Fourel et al. [57], which was characterized by the ion peak *m/z* 365 [M − H]−, 228 and 174 (Figure 3C). In the MS/MS spectrum of this product, the presence of product ions with *m/z* 228.08 (base peak), 214.06, and 174.03 illustrated the partial losses of prenyl side chain attached to the C3 position of 4HC (1), which was the core structure of this prenylated product (Figure 3D). This product (compound-1) is therefore believed to be ferulenol (2), where the obtained mass data (Figure S4) were further eluicidated by many aspects of in silico tools, including MetFrag web service, SIRIUS, CSI:Finger ID web service, and CANOPUS to entirely support the structural elucidation of this prenylated product. Abdou et al. [47] revealed that 4HC (1) is able to exist in tautomeric forms including 4-hydroxy-2-chromenone (a), 2,4-chromadione (b), and 2-hydroxy-4-chromenone (c) (Figure 3E). The compound-2 (*m/z* 365.21; Rt = 22.3 min) was thus presumably the isomeric form of ferulenol (2) proceeded via the tautomerization, where the 2-hydroxy-4-chromenone (c) served as the aromatic core structure (Figure 3F). The presence of fragmented ions with *m/z* 214.06 (base peak), 228.08, and 282.12 in the resulting MS/MS spectrum also indicated the partial losses of prenyl moiety that attached to the C3 position of this prenylated product (Figure 3G). Note that 4HC (1) as the fed substrate was also detected in the culture medium of the clones carrying pCDFDuet-AtaPT-CstDXS (Figure 3F).


#### - *2.3. Structural Annotation of Compound-1 by Using MetFrag Web Service*

 [ ]. - ( ) ( : ) . / - ( ) ( ). - ( ) . [ − ] - . The chemical structure of compound-1 was further annotated by using MetFrag web service, the in silico tool designed to elucidate chemical structure of query subjects from the experimental mass spectra [52]. Among the 2267 candidates retrieved from the PubChem database, compound-1 was best annotated as ferulenol (2), (compound CID: 54679300), with the highest score of 1.0 in which 8/11 peaks were matched with those of the in silico-generated fragmented ions of the candidated ferulenol (2) deposited in the PubChem database (Figure 4). We hence suggested that compound-1 is likely to be ferulenol (2), rather than the other prenylated 4HC derivatives exhibiting the identical mass value with *m/z* of 365.21 [M − H]−.


**.** - ( . [ − ] <sup>−</sup> = . ) ( ) . **Figure 4.** Structural annotation of compound-1 (*m/z* 365.21 [M − H]−; Rt = 21.3 min) using MetFrag web service, where the query subject was best annotated as ferulenol (2) among 2267 structures retrieved from the PubChem database.

#### *2.4. Computationally Assisted Identification of the Prenylated Product by Using SIRIUS (Version 4.8.2)*

 [ ]. . [ − ] − ( :// . . / / / / / / / / /) ( ) ( . − According to the user manual, SIRIUS requires high mass accuracy in which the ppm error is less than 20 ppm before conducting the annotation processes for the most reliable results [55]. The observed molecular ion *m/z* 365.21 [M − H]<sup>−</sup> was thus estimated by using the mass error calculation tool from the web service (https://warwick.ac.uk/fac/sci/ chemistry/research/barrow/barrowgroup/calculators/mass\_errors/ (accessed on 17 June 2021)), by comparing against its theoretical *m/z* (365.212218 [M − H]−), and showed an error of 8.762029 ppm, permitting the elucidation of the raw mass data by using this tool.

[ − ] ) . . - / . . % ( ) ( ) ( ) ( :// . . . . / / ) ( ). . . ( ) - - . ( - ) % ( ). / ( ) ( ). SIRIUS utilizes the high-resolution isotopic pattern analysis to locally annotate the correct molecular formula based on the experimentally acquired MS/MS data of the query subjects. Among the potential ten elemental formulas retrieved from the PubChem database, our query subject was annotated as C24H30O<sup>3</sup> with the highest Sirius score of 62.243% (Figure 5A), which was identical to the native elemental formula of ferulenol (2) (C24H30O3) deposited in the PubChem database (https://pubchem.ncbi.nlm.nih.gov/ compound/Ferulenol (accessed on 17 June 2021)). SIRIUS also offers a refined search option to explore the query subjects against biological databases, e.g., Natural Products, Collection of Natural Products (COCONUTS), and NORMAN databases to specify and narrow natural molecules at user-defined cut-offs. By searching the possible structure against the aforementioned databases, our query metabolite (compound-1) was annotated as C24H30O<sup>3</sup> with a score that greatly improved to 100% (Figure 5A). The annotated MS/MS spectra locally computed by the SIRIUS tool revealed that 8 of 11 peaks (indicated in the green spectra) matched with their local database (Figure 5B). The result was alsoconsistent with those fragmentation spectra predicted by MetFrag, where 8 out of 11 peaks matched with the in silico fragmented ions of the candidate ferulenol (2) in the PubChem database (Figure 4).

( ) ( ).



#### : *2.5. Structural Annotation and Compound Classification of Compound-1 by Using CSI:FIngerID and CANOPUS*

 [ ]. ( ) ( . . ). ( ) . % ( ). ( ) ( ). ( - ) ( ) ( ). : - " " ( ). ( ) ( . [ − ] − ). "[ ] ([ ]) ( ( ) )" % ( = . ) ( ). [ ]= :[ ]- :[ ]:[ ]- :[ ]( = - : - ) % ( = . ) . ( ) - - . . ( ) ( ). ( ) - / [ ]. : . [ − ] − ( ) / ( ). . ( ) SIRIUS has recently been integrated with CSI: FingerID web service to identify the chemical structure of the query subjects [55]. In this step, the annotated mass spectrum (Figure 5B) was compared against several compounds in the chosen molecular structure databases (e.g., PubChem, MeSH, and COCONUTS). Of more than 100 possible structures retrieved from all databases and locally predicted by the tool, ferulenol (2) as the target product was ranked as the tenth candidate structure with a percentage similarity of 43.62% (Figure 6A). Since ferulenol (2) is a natural product exclusively produced by *F. communis*, we narrowed the scope of the structural elucidation by using Natural Products and COCONUTS databases and found the rank of the candidate was substantially improved from tenth to fifth (Figure 6A). By searching against the structural compounds from the NORMAN database, our prenylated 4HC analog (compound-1) was perfectly matched to that of ferulenol (2) as the top ranked candidate (Figure 6A). Besides, CSI:FingerID can provide the crucial information of so-called "molecular fingerprints" to verify various substructures that can be found in the query subjects (Figure 6B). In this instance, several molecular fingerprints belonging to ferulenol (2) were predicted to be present in this prenylated product (*m/z* 365.21 [M − H]−). For example, substructures encoded by a SMARTS string "[#6]c1c([#8])ccc1 (Cc1c(O)ccc1)" with a score of 98% (F1 = 0.823) correspond to the benzene ring attached to the pyrone ring of 4HC (1). The basis pyrone ring of 4HC encoded by [#8]=,:[#6]-,:[#6]:[#6]-,:[#8](O=C-C:C-O) possessed 85% similarity (F1 = 0.815) and was verified to be present in same candidate structure. Biosynthetically, they were all obtained from the 4HC (1) fed in the culture medium of clones bearing pCDFDuet-AtaPT-CsTDXS. Equally important, there were several substructures belonging to the farnesyl moiety which originated from the DXP pathway, predicted to be present in the same candidate structure. Several substructures representing the basic benzene and pyrone rings along with the isoprene building block belonging to the query ferulenol (2) were predicted to be present in the trained structures of SIRIUS tool (Figure 6C). SIRIUS has also been developed to connect with CANOPUS (class assignment and ontology prediction using mass spectrometry) for logical classification of unknown metabolites based on the high-resolution MS/MS data [56]. According to the molecular properties annotated by CSI:FingerID web service, our query prenylated product *m/z* 365.21 [M − H]<sup>−</sup> was systematically classified as coumarins and derivatives (class), where phenylpropanoids/polyketides and organic compounds served as the superclass and kingdom, respectively (Figure 6D). CANOPUS also provides alternative classes of the query subjects. In this case, the putative ferulenol (2) was mainly classified as a coumarin derivative, but aromatic monoterpenoids, benzenoids, and lactones were recognized as the alternative classes of this product. Based on the

data acquired from the direct comparison against previously established mass spectra [56] along with the molecular formula and structural elucidation by means of the in silico tools included, MetFrag, SIRIUS, CSI:FingerID, and CANOPUS strongly supported that compound-1 was ferulenol (2), which is the product of C3-farnesylation of 4HC (1). . [ ] : - ( ) - ( ).

( ) ( . [ − ] − ) - - : ( . . ). ( ) ( ) ( ) . [ − ] − - - . ( ) ( ) ( ) ( . ) ( . ). ( ) ( ) . ( ) / ( ( %) . [ − ] − ) - ( - - - ). - - - / **Figure 6.** Structural annotation and classifications for the query ferulenol (2) (*m/z* 365.21 [M − H]−) synthesized by the recombinant *E. coli* carrying pCDFDuet-AtaPT-CstDXS based on CSI:FingerID and CANOPUS integrated in the SIRIUS (version 4.8.2). (**A**) In all included databases, ferulenol (2) as the target metabolite was ranked as the tenth potential candidate (highlighted in yellow) for the prenylated 4HC analog *m/z* 365.21 [M − H]−, meanwhile, the rank of this product significantly increased towards the fifth- and top-ranked candidates when structural identification was performed against the biological databases included Natural Products, COCONUTS, and NORMAN. (**B**) The examples of substructures (molecular fingerprints) predicted to exist in the query subject, including those belonging to the 4HC (1) core structure (No. 1284, 1222, and 1234), and the farnesyl moiety (No. 435, 411, and 502). (**C**) The examples of substructures belonging to ferulenol (2) found to exist in the training structures of the SIRIUS tool. (**D**) Based on the experimentally observed MS/MS data, CANOPUS showing the query compound (C24H30O<sup>3</sup> (SIRIUS score 100%), *m/z* 365.21 [M − H]−) was categorized as coumarins and derivatives, describing the polycyclic aromatic compounds that contains a 1-benzopyran moiety with a ketone group at the C2 carbon atom (1-benzopyran-2-one).

#### . [ ] ( :// . . / / / . = ) ( ). ( ) *2.6. Identification of Farnesylated-4HC Derivatives by Using LC-QTOF-MS/MS in Positive ESI Mode*

+

The high-resolution LC-QTOF-MS/MS in positive ion mode was thus implemented to corroborate both prenylated products based on monitoring of the molecular ion *m/z* 367.227320 [M + H]<sup>+</sup> computed by the web tool (https://www.sisweb.com/referenc/ tools/exactmass.htm?formula=C24H31O3 (accessed on 7 August 2021)). In the extracted ion chromatogram (EIC) of the medium extract of clones carrying pCDFDuet-AtaPT-CstDXS, there were two extracted ions (namely compound-1 and compound-2) eluted at the retention times of 21.3 and 22.3 min that exhibited identical *m/z* 367.22 [M + H]<sup>+</sup> , where they presumably farnesylated derivatives of 4HC (Figure 7A). The resulting MS/MS spectrum of compound-1 (*m/z* 367.22; Rt = 21.3 min) was clearly identical to those belonging to the authentic ferulenol (2) established under the same LC-MS/MS condition (Figure 7B). In the MS/MS spectrum of this prenylated product, the presence of proposed substructures

at *m/z* 175.04 (base peak), 217.08, 231.10, and 243.10 illustrated the partial losses of the prenyl side chain specifically attached to a C3 position of the 4HC core structure (1) (Figure 7C). The obtained evidence leds us to suggest that the compound-1 (*m/z* 367.22 [M + H]<sup>+</sup> ; Rt 21.3 min) was indeed ferulenol (2), a product derived from C3-farnesylation of 4HC (1). Hence, the compound-2 (*m/z* 367.22 [M + H]<sup>+</sup> ; Rt = 22.3 min) was presumably the tautomeric structure of ferulenol (2), where the 2-hydroxy-4-chromenone (c) served as the core structure (Figure 3E). In the MS/MS spectra along with proposed structures of this product, the existing product ions at *m/z* 177.08 (base peak), 215.06, 229.08, and 243.10 signify the partial elimination of prenyl moiety that incorporated the C3 region of 2-hydroxy-4-chromenone (Figure 7D). The existing of signals at *m/z* 189.05, which were presented in the MS/MS spectrum of both products as shown in Figure 7C,D, were presumably obtained from the prenyl moieties which were adjacent to the core structure of ferulenol (2) and its isomeric structure (3). Based on the acquired evidence, compound-2 was tentatively defined as the isomeric structure of ferulenol (2) which was formed via "tautomerization". Further elucidation (e.g., NMR) is required to verify the tentative confirmation of this prenylated product. The proposed reaction mechanisms illustrating the various chemical losses present in the MS/MS spectrum of compound-1 and compound-2 are shown in Figures S5 and S6. The postulated mechanisms underlying the formation of ions *m/z* 189.05 of compound-1 and compound-2 were shown in the Supplementary Materials (Figures S7 and S8).

**.** - - - - - - / - . ( ) . . [ ] + . ( - ) **Figure 7.** Identification of prenylated products produced by clones carrying pCDFDuet-AtaPT-CstDXS based on the highresolution LC-ESI-QTOF-MS/MS in the positive-ion mode. (**A**) The EIC for the molecular ion *m/z* 367.2268 ± 0.05 [M + H]<sup>+</sup>

)

( ) - - -

[ – ].

. **.** 

. . . ( ) /

( ) ( ) ( . .

( ). ( ) / -

showing two reaction products exhibited the equivalent *m/z* 367.22 (named compound-1 and 2) eluted from the HPLC colume at the retention times of 21.3 and 22.3 min, respectively. (**B**) The direct comparison between the MS/MS spectrum of compound-1 (*m/z* 367.22) and the authentic ferulenol (2) (*m/z* 367.22), which is characterized by the product ions with *m/z* 137.13, 163.04, 175.04, 217.09, 231.10, 243.10, and 285.15, respectively. (**C**) The MS/MS spectrum of compound-1 showing the partial losses of prenyl side chain from the 4HC core structure (1). (**D**) The MS/MS spectrum of compound-2 (proposed structure) signifying the partial elimination of prenyl moiety from the 2-hydroxy-4-chromenone, a core structure of the prenylated product.

#### **3. Discussion**

*E. coli* has the capability to supply various isoprenoid precursors through the native DXP pathway [19–24]. In the past few decades, this microorganism has been engineered to produce various bioactive terpenoids and valuable precursors such as limonene, taxadiene (taxol precursor), amorphadiene (artemisinin precursor), and carotenoids (e.g., lycopene, astaxanthin, carotenoids) by the heterologous expression of terpene synthases with the ratelimiting step enzymes in the DXP pathway [29–34]. However, few studies have focused on the use of this microorganism as the in vivo prenylation system of aromatic natural products. In wild-type *E. coli*, FPP is involved in the biosynthesis of various molecules (e.g., ubiquinone and peptidoglycan) and its level is tightly maintained, as it is essential for bacterial growth and viability [25–28]. A previous study also demonstrated that the deletion of genes encoded for FPP synthase (IspA) resulted in the observed growth retardation in the mutant strains of *E. coli* [25]. We hence speculated that FPP might become more readily available for the in vivo prenylation of 4HC (1) for ferulenol (2) than others (e.g., DMAPP, GPP, and GGPP). However, the native isoprenoid precursors of *E. coli* are always insufficient for metabolic engineering purposes, and the enhanced metabolic fluxes via overexpression of the rate-limiting step enzymes in the DXP pathway (e.g., DXS, DXR, and IDI) are required [7,19–24]. The transcriptional profile analysis in the young leaves of *C. stellatopilosus* revealed a positive association between CsDXS gene expression and plaunotol (acyclic diterpene alcohol) content, and this enzyme was hence suggested as one of the rate-limiting steps in the DXP pathway controlling the flux of isoprenoid precursors in the plant [48]. Since the active form of CsDXS could be achieved after the removal of the signaling peptide region [30,48], the truncated form of this enzyme, CstDXS, was used in this study to drive the metabolic flux of the DXP pathway in the *E. coli* BL21(DE3)pLysS strain. We demonstrated that *E. coli* BL21(DE3)pLysS harboring pCDFDuet-AtaPT-CstDXS is capable of producing the putative mass peak corresponding to ferulenol (2) from the fed 4HC (1) without relying on the external supply of prenyl-donors (e.g., DMAPP, GPP, and FPP). The synthesized product was found to be excreted into the culture medium. This means the step of breaking the cell during the downstream processes is not necessary. Furthermore, this expression system did not require the use of high-priced precursors. These reflected the potential economic system of the prenylated 4HC derivatives production by using recombinant *E. coli*. Besides, our results extend the findings of Chen et al. [46] in that AtaPT can utilize various substrates by showing that 4HC (1) could also be recognized as the aromatic acceptor, even though it has never been reported that this compound acted as the substrate of AtaPT.

Although NMR elucidation is required to verify the structures of the resulting prenylated product (compound-1), this technique is restricted to researchers in the fields of metabolic engineering and metabolomics, which have to deal with trace-level metabolites [49–51]. The structural elucidation of compound-1 was thus based on the use of in silico tools designed to annotate the structures of the query metabolites using the experimental mass spectra (MS2) (Figure S4). Having been confirmed to be highly consistent with the unique MS/MS fragmentation profile of ferulenol (2) from the previous work [56], MetFrag analysis clearly showed that the compound-1 was best annotated as "ferulenol (2)" out of the 2266 candidates presented from PubChem database. However, information regarding the molecular formula, chemical structure, substructures, and classifications of this product are still needed. SIRIUS is one of the in silico tools

developed to unravel various chemical features hidden in the MS/MS data of the query subjects [52]. This tool has typically been used in the field of metabolomics and has been shown to be helpful in the field of metabolic engineering, in which it was used to identify 2,4,6-trihydroxybenzophenone produced by the engineered *E. coli* [58]. The results from SIRIUS showed that our query product (compound-1) possessed the neutral molecular formula C24H30O3, which corresponded to that of ferulenol (2) deposited in the PubChem database (https://pubchem.ncbi.nlm.nih.gov/compound/Ferulenol (accessed on 26 September 2021)). SIRIUS also locally computed the relevant MS/MS spectra and fragmentation pathway which might be involved in the fragmentation process of compound-1 in which eight mass peaks were explained, meanwhile, the rest of the three peaks were considered as the noises. Based on the eight mass peaks explained by SIR-IUS, CSI:FingerID analysis suggested that the compound-1 was annotated as ferulenol (2) as the tenth-ranked candidate after searching against all databases provided by the SIRIUS tool. Although the expected ferulenol (2) was not perfectly categorized as the first candidate for the compound-1, it might be speculated that CSI:FingerID integrated in SIRUS exhibited extremely good performance in the case of the independent MS/MS being trained and deposited in the training data, and the rate of correct identification was substantially decreased when the trained data were removed from this tool [54]. This can be confirmed by the fact that the MS/MS data for the authentic ferulenol (2) (encoded by InChI=1S/C24H30O3/c1-17(2)9-7-10-18(3)11-8-12-19(4)15-16-21-23(25)20-13- 5-6-14-22(20)27-24(21)26/h5-6,9,11,13-15,25H,7-8,10,12,16H2,1-4H3/b18-11+,19-15+) has not yet been trained in the negative mode mass data of the SIRIUS tool (https://www.csifingerid.uni-jena.de/v1.6.0/api/fingerid/trainingstructures?predictor=2 (accessed on 23 June 2021)). Based on the LC-MS search (for the molecular ion *m/z* 365.21 [M − H]−), it might also be possible to exclude the rest of candidates (ranked first to ninth retrieved from all included databases, and ranked first to fifth from the Natural Products and COCONUTS databases) as incorrect structures since they are not natural occurring in *E. coli*, and there was only 3b-allotetrahydrocortisol showing the proximal *m/z* 365.23 [M − H]<sup>−</sup> found in the *E. coli* metabolomics database (ECMDB) (https://ecmdb.ca/spectra/ms/search (accessed on 23 June 2021)). This indicated that only ferulenol (2) as the product of pathway engineering could be accepted as the correct structure. In the case of the query metabolites acquired from the biological samples, SIRIUS also provides the biological databases as the choices to narrow the scope of natural molecules. When MS/MS spectra of compound-1 were annotated against biological databases, e.g., Natural Products, COCONUTS, and NORMAN, the rank annotated as "ferulenol (2)" was greatly improved as the fifth and first candidate, respectively. Based on the substructures predicted by CSI:Finger ID and CANO-PUS, it subsequently provided the vital information that the compound-1 was classified as coumarins and derivatives.

According to Abdou and colleaques [47], the 4HC (1) can exist in three different tautomeric structures: 4-hydroxy-2-chromenone (a), 2,4-chromadione (b), and 2-hydroxy-4-chromenone (c) (Figure 3E). We hence postulated that the compound-2 (eluted from the HPLC column at 22.3 min), obtained from the negative- and positive-ion mode analyses, is presumably the tautomeric form of ferulenol (2). The obtained MS/MS spectra (Figures 3G and 7D) were further elucidated to gain insight into the structural information of this prenylated product. In the negative ion mode analyses, the three chemical losses explained the daughter ions at *m/z* 214.06, 228.08, 282.12, signifying the prenyl side chain was partially removed from the 2-hydroxy-4-chromenone (c), the core structure of this prenylated product. A similar trend was observed in the positive ion mode analyses, where the partial loss of prenyl side chain attaching to the 2-hydroxy-4-chromenone could be characterized via the signals with *m/z* 177.08, 215.06, 229.08, and 243.10, respectively. The existence of product ions at *m/z* 189.05 [M + H]<sup>+</sup> in the MS/MS spectrum of both prenylated products (Figure 7C,D) were likely to be the outcome of non-enantioselective epoxidation of the double bond present in the farnesyl side chain of ferulenol (2), as recently described by Cortés et al. [59]. Based on the obtained evidence, the compound-2 (3) was interpreted

as the farnesylated derivative of 2-hydroxy-4-chromenone, however further elucidation (e.g., NMR) is required to fully support a chemical point of view.

The availability of prenyl precursors influences the patterns of isoprenoid-derived natural products produced by the bacterial systems [60]. Here, ferulenol (2) and its proposed isomeric structure (3) were exclusively detected as the predominant products produced by *E. coli*-carried pCDFDuet-AtaPT-CstDXS. Although AtaPT exhibits remarkable substrate promiscuity towards a variety of prenyl-donors such as DMAPP, GPP, FPP, and GGPP [46], the presence of two farnesylated-4HC analogs (compounds-1 and -2) clearly supports our hypothesis that the FPP accumulated inside the bacterial cells is more easily accessible for synthesizing the two prenylated products rather than the others (DMAPP, GPP, and GGPP). Our result was also consistent with the previous finding, where the farnesylated-menadione was the major product of the whole-cell catalysis by *P. pastoris*, harboring the gene encoding for aromatic prenyltransferase (NovQ) [37]. Studies have shown that overexpression of multiple rate-limiting enzymes in the DXP pathway, including 1-deoxy-D-xylulose 5-phosphate reductoisomerase (DXR), isopentenyl pyrophosphate isomerase (IDI), along with FPP synthase (e.g., IspA), causes the substantially increased production of terpenoids in their engineered microbes [23–25]. Furthermore, 4HC (1) as the fed precursor is unlikely and unsustainable from an economic standpoint, as a considerable amount of this expensive substrate must be supplied to the bacterial culture to make its desired prenylated analogs. This strategy might be impracticable for industrially scaled applications, where the newly designed strains capable of de novo manufacture of the two farnesylated-4HC analogs or the usage of the lower cost precursors (e.g., sodium salicylate) should be established. Previous research has shown that *E. coli* can be engineered to produce 4HC (1) via the inherent chorismate pathway, which is accomplished through a series of reactions catalyzed by isochorismate synthase (ICS), isochorismate pyruvate-lyase (IPL), salicylate-CoA ligase (SCL), and biphenyl synthase (BIS) [61]. Based on the broad substrate specificity of benzoate-CoA ligase (BadA) and benzophenone synthase (GmBPS), our group demonstrated that *E. coli* BL21(DE3)pLysS carrying pETDuet-BadA-GmBPS was able to synthesize 4HC (1) from the fed salicylate (sodium salt) as well (data unpublished). By considering these advantages, further establishments of *E. coli* systems capable of synthesizing 4HC (1) from the inexpensive precursors (e.g., glucose, sodium salicylate) to act as the ATaPT's substrate along with enhancing the flux of isoprenoid precursors would be beneficial for large-scale production of the two farnesylated-4HC analogs reported herein.

Although we demonstrated that the *E. coli* BL21(DE3)pLysS strain could be engineered to produce ferulenol (2) and its isomeric structure (3), further optimizations are needed to improve the yields of final product, which seems to be limited by the supply of isoprenoid precursors [7–9]. Several studies have demonstrated that overexpression of the multiple enzymes regulating the flux of isoprenoid precursors led to greatly improved terpenoidderived natural products in *E. coli* [19–24]. Future pathway engineering via overexpression of the other rate-controlling steps in the DXP pathway, such as DXR, IDI, and farnesyl pyrophosphate synthase (FPPS), will be carried out to improve the yields of the two prenylated-4HC derivatives reported herein. Since a large amount of 4HC (1) was also found to remain in the culture medium, optimizing substrate consumptions, i.e., a time course production is needed to enhance the yield of those prenyl derivatives. There are several factors affect the heterologous production of natural products in bacterial systems, such as temperature, codon usage, and plasmid copy number [62–64]. Thus, optimizing these parameters will be examined.

Our results illustrated the in vivo functional expression of AtaPT and CstDXS in the *E. coli* system, which could be seen from the formation of ferulenol (2) and its isomeric structure (3) secreted into the culture medium. These findings also shed new light on the use of the engineered *E. coli* system as the prenylation system to produce valuable secondary metabolites instead of isolation from plants and chemical synthesis. Since the AtaPT used in this study exhibits broad substrate specificity towards many types of aromatic acceptors, e.g., benzophenones, xanthones, stilbenes, and chalcones [46], the

future impact of clones carrying pCDFDuet-AtaPT-CstDXS on synthesizing the prenylated products might not be restricted to 4HC (1) but could be applied to the other types of aromatic acceptors.

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

#### *4.1. Reagents*

The general reagents were analytical grade and were purchased from Sigma-Adrich (St. Louis, MO, USA), Merck (Darmstadt, Germany), and Avantor (Center Valley, PA, USA).

#### *4.2. Construction of Plasmid pCDFDuet-AtaPT-CstDXS*

pCDFDuet-1 coexpression vector (Novagen, Darmstadt, Germany) was chosen to construct the recombinant plasmid containing the two genes encoding AtaPT (GenBank accession no. KP893683) and CstDXS (GenBank accession no. AB354578.1) based on the procedure reported by Toila and Joshua-Tor [65]. The initial insertion was performed by incorporating AtaPT in the MCS-1 (*BamHI* and *NotI*) followed by insertion of CstDXS into the MCS-2 (*Kpn*I and *Xho*I) of the pCDFDuet-1 vector to yield pCDFDuet-AtaPT-CstDXS. The details were as follows: The plasmid pUC57-AtaPT obtained from the gene synthesis technology (Invitrogen, Waltham, MA, USA) was used as the DNA template to provide AtaPT. The PCR reaction consisted of Phusion High-Fidelity DNA polymerase (NEB, Ipswich, MA, USA); the forward primer was AtaPT-F: 5′ -GGTGGATCCGATGCTCCCCCCATCAGACA-3 ′ , and the reverse primer was AtaPT-R 5′ -AAAGCGGCCGTCACACAGCTGCG-3′ (underlines are the recognition sites for *BamH*I and *Not*I, respectively). The PCR cycle included pre-denaturation at 98 ◦C for 1 min, followed by 30 cycles of 98 ◦C for 30 s, 60 ◦C for 30 s, and 72 ◦C for 30 s, and a final extension at 72 ◦C for 5 min. The obtained PCR product (~1275 bp) was purified by using a Gel Band Purification Kit (GE Healthcare, Chicago, IL, USA), digested with *BamH*I and *Not*I, and ligated into a pCDFDuet-1 vector which had been treated with the same restriction enzymes. The ligation mixture was transformed into *E*. *coli* DH5α and the positive clones were selected by spreading on the LB-agar-contained streptomycin (50 µg/mL). The resulting plasmids were extracted using PureYieldTM Plasmid Mini-prep System (Promega, Madison, WI, USA) and the gene insertion was confirmed by double digestion with *BamH*I and *Not*I. The resulting pCDFDuet-AtaPT was used as the DNA backbone in the next step.

Based on the previous finding, the gene-encoded 1-deoxy-D-xylose 5-phosphate synthase of *C*. *stellatopilosus* (CsDXS: 2163 bp) was predicted to contain the putative chloroplast transit peptide (cTP) that should be removed before the gene expression in *E*. *coli* systems [40]. Therefore, the truncated CsDXS (CstDXS) which was absent of the cTP coding sequence (171 bp) was cloned from the young leaves of *C*. *stellatopilosus*. The PCR reaction consisted of Pfu DNA Polymerase (ThermoScientific, Waltham, MA, USA), CstDXS-F: 5′ -TGCGGTACCATGGCATCACTTTCAGAAA-3′ , and CstDXS -R: 5′ - AGCCTCGAGTGCTGACATAATTTGCAGA -3′ (underlines are the restriction sites for *Kpn*I and *Xho*I, respectively). The PCR condition was as follows: Pre-denaturation at 95 ◦C for 1 min, followed by 30 cycles of 95 ◦C for 30 s, 60 ◦C for 30 s, and 72 ◦C for 2 min, and a final extension at 72 ◦C for 5 min. The PCR product (1992 bp) was purified from the agarose gel by using the gel purification kit, double digested with *Kpn*I and *Xho*I, and was ligated to the pCDFDuet-AtaPT which was digested with the same restriction enzymes. The double digestion by *Kpn*I and *Xho*I was performed to confirm the insertion of CstDXS in pCDFDuet-AtaPT (Figure S3). The nucleotide sequencing (IDT, Penang, Malaysia) was performed to verify the correct bases and the in-frame arrangement of both AtaPT and CstDXS in the pCDFDuet-AtaPT-CstDXS by using two pairs of primers: ACYCDuetUP1 primer 5′ -GGATCTCGACGCTCTCCT-3′ , and DuetDOWN1 primer 5′ -GATTATGCGGCCGTGTACAA-3′ for AtaPT in the MCS-1, and DuetUP2 primer 5 ′ -TTGTACACGGCCGCATAATC-3′ and T7term primer 5′ -GCTAGTTATTGCTCAGCGG-3 ′ for CstDXS in the MCS-2.

#### *4.3. Bioconversion of 4HC (1) into the Farnesylated-4HC Derivatives*

The pCDFDuet-AtaPT-CstDXS transformed into *E*. *coli* BL21(DE3)pLysS (Promega, Madison, WI, USA) by heat shock method. The engineered strains carrying pCDFDuet-AtaPT-CstDXS were cultured in the LB medium containing streptomycin (50 µg/mL) and chloramphenicol (34 µg/mL) at 37 ◦C, 200 rpm for 18 h. The 1.5 mL of culture was then inoculated into the 500 mL Erlenmeyer flask containing 150 mL of the same medium and was further cultivated at 37 ◦C (200 rpm) until the OD<sup>600</sup> reached 1.0. The gene expression was induced by adding 1 mM IPTG (final concentration); after that, cells were grown at 18 ◦C, 250 rpm for 5 h. To start the bioconversion process, 3 mM 4HC (1) (TCI, Tokyo, Japan) and 3 mM MgCl<sup>2</sup> were supplied to the induced culture and the cells were further cultivated at the same condition for 18 h. After that, the culture medium was harvested by centrifugation at 4 ◦C, 8000 rpm for 10 min. *E*. *coli* BL21(DE3)pLysS carried pCDFDuet-AtaPT-CstDXS that was grown in parallel at the same condition, except without the supplement of 4HC (1) used as the control in this study. Since ferulenol (2) is lightsensitive, the bioconversion experiment throughout this study took place in the darkness to minimize the product degradation.

#### *4.4. Extraction of Prenylated Products from the Culture Medium*

The extraction process was performed in the darkness. The culture medium (150 mL) was partitioned twice with 75 mL EtOAc in the 500 mL Erlenmeyer flask by shaking at 300 rpm, 25 ◦C for 30 min. The EtOAc layers were then harvested after centrifugation for 5 min, 6000 rpm, at 4 ◦C, and concentrated until dryness using N<sup>2</sup> gas. The dried residues were redissolved in MeOH (HPLC grade) before identification of the prenylated-4HC derivatives by using the high-resolution LC-ESI-QTOF-MS/MS in both negative- and positive-ion mode operations.

#### *4.5. Identification of Metabolites by LC-ESI-QTOF-MS/MS*

Identification of the prenylated products were carried out by HPLC Dionex (Thermo Scientific) connected with MS Bruker Maxis (esquire 4000 Daltonics, Bremen, Germany) and the RP-18 column (Acclaim RSCL 120 C18 column, 2.1 X 100 mm, 2.2 µM, Thermo Fisher Scientific, Waltham, MA, USA). The mobile phases consisted of H2O with 0.1% formic acid (solvent A) and acetonitrile with 0.1% formic acid (solvent B). The separation of prenylated products was achieved by using a linear gradient of solvent B as follows: 5% for 0–2 min, 5–95% for 15 min, 95% for 3 min, and back to 5% for 10 min, for a total running time of 30 min. The flow rate was set at 0.4 mL/min. The sample temperature was controlled to 10 ◦C. The column oven temperature was 40 ◦C. The injection volume was 10 µL. The product elucidation was conducted by tandem mass (MS/MS) with electrospray ionization (ESI) under the collision-induced dissociation (CID) energy 35 eV in negative ion mode analysis. Nebulizer pressure was set at 29 psi, the dry gas temperature was 180 ◦C, and the dry gas flow rate was 8.0 L/min. The masses were scanned over the *m/z* range of 100–1000 amu. The obtained MS/MS spectra (Figure S4) were directly compared against MS/MS spectra of ferulenol (2) [54], before being annotated by using the in silico tools to gain insight into the structure to ultimately confirm the obtained results. For positive ESI mode, the same chromatographic separation condition was applied for identifying both prenylated products (compound- 1 and 2), except the ion polarity was switched to the positive-ion mode with the CID energy of 25 eV (mass scan range 50–1500 amu). The obtained MS/MS spectra were then compared to the authentic ferulenol (2) (final concentration of 20 ppm) (Santa Cruz Biotechnology, Santa Cruz, CA, USA) run in parallel under the same positive ESI condition.

#### *4.6. Structural Annotation of Prenylated Products by Using MetFrag*

MetFrag, the freely accessible software (https://msbi.ipb-halle.de/MetFragBeta/ (accessed on 26 September 2021)), was used as a tool for structural annotation of metabolites from the high-quality MS/MS spectra of the query subjects, which plays a crucial role

as the initial step of structural identification. According to the user manual provided in https://ipb-halle.github.io/MetFrag/projects/metfragweb/ (accessed on 26 September 2021), structural annotation of the target metabolites is required for two steps of data processing, including "retrieving candidates" and "processing of candidates". Based on the experimental mass spectra of the compound-1, the neutral mass of 366.219495 (relative mass deviation of 5 ppm) with the calculated molecular ion (*m/z* 365.211670 [M − H]−) was defined as parameters to retrieve the candidate molecules from the PubChem database. The neutral molecular formula (C24H30O3) and data-specific identifiers (i.e., compound ID number) could also be defined in this step to perform the candidate search. Having retrieved a large number of candidates (2267 molecular structures, in our case) from PubChem, several parameters including the relative mass deviation (10 ppm), absolute mass deviation (0.001), the certain adduct type in ([M − H]−), and the MS/MS peak list (Figure S4) were then defined to match against "in silico generated fragments" of candidates from the PubChem database. After that, the score-ranked list of candidates was displayed. For each candidate in the row, the information regarding candidate image, identifier with linked database, exact mass, molecular formula, and several MS/MS peaks explained were also displayed. For a certain candidate of the list, "fragment view" also provided molecular formulas and their corresponding substructures for the matched peaks, which are highlighted in green.

#### *4.7. Structural Annotations of Prenylated 4HC Analog Using SIRIUS Tool*

SIRIUS is an in silico tool developed for the identification of molecular formula, mass spectra, and fragmentation tree annotation of the unknown metabolites based upon high-resolution raw MS/MS data [54]. This tool is also connected with CSI:FingerID web service for structural identification and prediction substructures (so-called molecular fingerprints) present in the query subject. SIRIUS can also systematically predict the classes of query metabolites via the computational tool (CANOPUS: class assignment and ontology prediction using mass spectrometry) [56]. Structural elucidation of the prenylated-4HC analogs were thus based on the prediction power of SIRIUS (version 4.8.2). According to the user manual, the raw MS/MS data in text format (Figure S4) was imported into the SIRIUS application window. The MS2 level and the collision energy (35 eV) were subsequently defined in the next dialogue. Then, two parameters, including the precursor mass ion (*m/z* 365.2154) and the specific adduct type ([M − H]−), were defined in the following application window. The annotation of prenylated-4HC derivatives was accomplished by selecting "compute option". In our case, the entire tools, including SIRIUS, CSI:FingerID web service, and CANOPUS, were selected to sufficiently detail structural elucidation of the prenylated product. Certain databases (e.g., Natural Products, COCONUTS, and NORMAN) were also chosen as biological databases to improve the rate of correct identification and rank of the query metabolite.

#### **5. Conclusions**

In this study, we demonstrated that the *E coli* BL21(DE3)pLysS strain could be engineered as the prenylation system of 4HC (1) via co-expression of the genes encoding for AtaPT and CstDXS. Based on the high-resolution LC-ESI(-)-QTOF-MS/MS, feeding 4HC (1) into the engineered strain resulted in the formation of the prenylated 4HC analog (namely compound-1) showing *m/z* 365.21 [M − H]−, which was detected from the culture medium. The mass fragmentation pattern of this product was highly identical to the authentic ferulenol (2) established by the previous work [66]. MetFrag analysis clearly showed that the compound was best explained as "ferulenol (2)" among 2267 candidated compounds from the PubChem database. Further annotation using SIRIUS integrated with CSI:FingerID and CANOPUS led to unraveling the chemical features hidden in compound-1, e.g., molecular formula (defined as C24H30O3), ferulenol (2) as a potential structure, and classes (defined as coumarin derivatives), indicating the prenylate-4HC derivative was indeed ferulenol (2). Meanwhile, compound-2 (*m/z* 365.21 [M − H]<sup>−</sup> and 367.22 [M + H]<sup>+</sup> ) was elucidated as the

farnesylated of 2-hydroxy-4-chromenone (c), which was proposed to be formed through the tautomerization of ferulenol (2). It is worthwhile to emphasize that both products could be formed without the external supply of prenyl donors (e.g., DMAPP, GPP, and FPP), which indicated the capacity of *E*. *coli* to supply the prenyl donors, particularly FPP, and verified the functional expression of AtaPT and CstDXS in this bacterial system. Equally important, they were secreted in the culture medium, providing a great benefit from an economic point of view since breaking cell pellets is not needed. Further experiments, e.g., overexpression of rate-limiting step enzymes in the DXP pathway along with the optimized culture conditions, are required to improve the yields of two products. Based on the substrate promiscuity of AtaPT, the in vivo prenylation by clones carrying pCDFDuet-AtaPT-CstDXS could be applied towards other types of natural products such as quercetin and mangiferin to generate diverse classes of prenylated derivatives.

**Supplementary Materials:** The following are available online, Figure S1: The insertion of AtaPT in MCS-1 of pCDFDuet-1 vector, Figure S2: The insertion of CstDXS in MCS-2 of pCDFDuet-1 vector, Figure S3: The verified insertion of CstDXS (~1992 bp) in pCDFDuet-AtaPT (~5020 bp) via *Kpn*I and *Xho*I digestion, Figure S4: The raw mass (MS2) data for putative ferulenol (2) (compound-1) obtained from negative-ESI mode, Figure S5: The proposed reaction mechanisms illustrating the various chemical losses present in the MS/MS spectrum of compound-1, Figure S6: The proposed reaction mechanisms illustrating the various chemical losses present in the MS/MS spectrum of compound-2, Figure S7: The postulated mechanisms underlying the formation of ion *m/z* 189.05 of compound-1, Figure S8: The postulated mechanisms underlying the formation of ion *m/z* 189.05 of compound-2.

**Author Contributions:** Conceptualization, N.N. and A.K.; methodology, A.K., J.N., P.P. and N.N.; validation, A.K., J.N., P.P. and N.N.; formal analysis, A.K., N.N., P.P. and J.N.; investigation, A.K. and J.N.; resources, N.N.; writing—original draft preparation, A.K. and N.N.; writing—review and editing, A.K., J.N., P.P. and N.N.; supervision, P.P. and N.N.; funding acquisition, N.N. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the National Research Council of Thailand-Khon Kaen University, 2018 (NRCT-KKU 2018): 610034; Faculty of Pharmaceutical Sciences, Khon Kaen University, Thailand.

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

**Informed Consent Statement:** Not applicable.

**Acknowledgments:** A.K. wishes to thank the Graduate School, Khon Kaen University for the overseas research scholarship. The authors would like to thank Assistant Professor Tiwatt Kuljanabhagavad, Suandusit University, Bangkok, Thailand for the valuable comments and advice on the mass spectrometry elucidation. We also thank Yutthakan Saengkun and Jatupong Sitsutheechananon for assistance.

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

**Sample Availability:** Samples of the compounds are not available from the authors.

#### **References**


## *Article* **New 3-Ethynylaryl Coumarin-Based Dyes for DSSC Applications: Synthesis, Spectroscopic Properties, and Theoretical Calculations**

**João Sarrato 1,† , Ana Lucia Pinto 1,† , Gabriela Malta 1,† , Eva G. Röck 2,3, João Pina <sup>2</sup> , João Carlos Lima <sup>1</sup> , A. Jorge Parola 1,\* and Paula S. Branco 1,\***


**Abstract:** A set of 3-ethynylaryl coumarin dyes with mono, bithiophenes and the fused variant, thieno [3,2-*b*] thiophene, as well as an alkylated benzotriazole unit were prepared and tested for dyesensitized solar cells (DSSCs). For comparison purposes, the variation of the substitution pattern at the coumarin unit was analyzed with the natural product 6,7-dihydroxycoumarin (Esculetin) as well as 5,7-dihydroxycomarin in the case of the bithiophene dye. Crucial steps for extension of the conjugated system involved Sonogashira reaction yielding highly fluorescent molecules. Spectroscopic characterization showed that the extension of conjugation via the alkynyl bridge resulted in a strong red-shift of absorption and emission spectra (in solution) of approximately 73–79 nm and 52–89 nm, respectively, relative to 6,7-dimethoxy-4-methylcoumarin (λabs = 341 nm and λem = 410 nm). Theoretical density functional theory (DFT) calculations show that the Lowest Unoccupied Molecular Orbital (LUMO) is mostly centered in the cyanoacrylic anchor unit, corroborating the high intramolecular charge transfer (ICT) character of the electronic transition. Photovoltaic performance evaluation reveals that the thieno [3,2-*b*] thiophene unit present in dye **8** leads to the best sensitizer of the set, with a conversion efficiency (η = 2.00%), best VOC (367 mV) and second best Jsc (9.28 mA·cm−<sup>2</sup> ), surpassed only by dye **9b** (Jsc = 10.19 mA·cm−<sup>2</sup> ). This high photocurrent value can be attributed to increased donor ability of the 5,7-dimethoxy unit when compared to the 6,7 equivalent (**9b**).

**Keywords:** dye-sensitized solar cells; coumarin dyes; thieno [3,2-*b*] thiophene; charge transfer; ethynylaryl

#### **1. Introduction**

Following O'Regan and Grätzel's seminal application [1] of an organic dye adsorbed on a mesoporous wide band gap semiconductor as a light-harvesting electrode, the field of DSSCs (Dye-Sensitized Solar Cells) has garnered much interest over the last three decades. Compared to alternative light-harvesting technologies, they possess reduced cost, ease of manufacture and low environmental impact, which combined with their wide array of possible colors and compatibility with flexible substrates, allows for a multitude of applications, such as integration into buildings [2,3] and interiors [4,5].

Although the original ruthenium(II)-polypyridyl chromophores used in DSSCs, such as N3 [6] and N719 [7], have mostly remained as gold standard dyes due to their large conversion efficiencies [8], their only moderate extinction coefficients and the use of a rare and expensive noble metal has led to the extensive search for highly efficient metal-free dyes. These metal-free dyes offer, in addition to the lower cost of production, easier and greater

**Citation:** Sarrato, J.; Pinto, A.L.; Malta, G.; Röck, E.G.; Pina, J.; Lima, J.C.; Parola, A.J.; Branco, P.S. New 3-Ethynylaryl Coumarin-Based Dyes for DSSC Applications: Synthesis, Spectroscopic Properties, and Theoretical Calculations. *Molecules* **2021**, *26*, 2934. https://doi.org/ 10.3390/molecules26102934

Academic Editor: Maria João Matos

Received: 31 March 2021 Accepted: 10 May 2021 Published: 14 May 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/).

synthetic versatility and tunable optical and electrochemical properties through structural modification, as demonstrated by the extremely wide array of compounds explored over the years [9–11]. Despite being mostly overshadowed by their inorganic counterparts, in recent years comparable or even superior performances have been achieved by metalfree chromophores, such as bulky indoline-quinoxaline dyes (10.65%) [12], tetrathioacene dyes (10.1%) [13] and more elaborate polycyclic aromatic push-pull dyes (12.6%) [14], all employing a Co(II/III) complex as redox shuttle. Additionally, the use of co-sensitization approaches has been employed to great success with organic dyes, managing to reach efficiencies of more than 14% [15,16].

To be adequately applied in DSSCs, organic dyes must present a donor-π-acceptor (D-π-A) structure. The push-pull effect in these D-π-A dyes leads to efficient intramolecular charge transfer (ICT) from the donor to the acceptor unit through the π-bridge upon light absorption. Among the many classes of organic compounds used, coumarins are of particular interest due to their wide use as fluorescent sensors [17–19], emitting layers in Organic Light-Emitting Diodes (OLEDs) [20–22] and in laser applications [23,24], owing to their large Stokes shift, high quantum yields and good solubility. Additionally, their photophysical properties can be easily tuned through the addition of substituents, namely electron-withdrawing substituents in position 3 and electron-donating substituents in position 7 [25].

This allows for a decrease in the energy gap between the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO), making coumarins great candidates as new sensitizers for DSSCs. Hara et al. [26–28] were some of the first to successfully design and employ dyes with coumarin donor units to achieve competitive efficiencies of up to 8.2% [29] when using deoxycholic acid (DCA) as a coadsorbent. In more recent work, Jiang et al. [30] and He et al. [31] used additional indoline and triphenylamine donors (respectively) attached to the coumarin unit, while Vekariya [32] investigated the effect of various *o*-halide phenylene spacers on dye structure and device performance.

Recently we have shown that the introduction of a linear ethynyl π-bridge into coumarin-based conjugated donor–acceptor systems resulted in redshifted absorption and emission spectra relatively to the styryl counterparts [33]. Electrochemical studies revealed that this derivatization resulted in a marked decrease in the HOMO energy levels, which influenced the overall conversion efficiency with a significantly superior performance. This revealed the importance of the alignment of the substituents with the direction of the intramolecular charge transfer. This is not completely surprising since the incident photon-to-current conversion efficiency (IPCE) reveals that the electron transfer yield (Φ(ν)ET) becomes larger with the introduction of a triple bond [34].

Following up on our previous experience with the synthesis of coumarin chromophores [33,35,36], several 3-ethynyl-6,7-dihydroxycoumarin-based dyes were prepared, with emphasis on the effect of varying the π-bridge on dye structure, photophysical properties and sensitizer efficiency. These groups include mono and bithiophenes, including a fused variant, thieno [3,2-*b*] thiophene, as well as a benzotriazole unit containing a long alkyl chain (Figure 1). Additionally, a 5,7-dihydroxycoumarin-based chromophore was prepared, to ascertain the effect of this alternative substitution pattern on device performance. All prepared dyes were then spectroscopically characterized, their optimized geometry was obtained from density functional theory (DFT) and their performance as sensitizers was evaluated.

**Figure 1.** Structure of target dyes prepared and characterized in this work.

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

#### *2.1. Synthesis and Characterization*

The synthetic approach to obtain the coumarin-based dyes, as described in Scheme 1, initiated with the brominated derivatives **1a** and **1b**. Compound **1a** was obtained as described by Martins et al. [36], while **1b** was prepared through condensation of ethyl propiolate and 1,3,5-trihydroxybenzene [37], followed by chloroformylation, bromination [38], hydrolysis and methylation. The 3-ethynyl coumarin derivatives (**3a** and **3b**) were then obtained through a high-yield Sonogashira coupling with ethynyltrimethylsilane, followed by the removal of the trimethylsilyl group, which at first was accomplished with tetrabutylammonium fluoride (TBAF). Since this method resulted in poor yields, partly due to the possible degradation of the unprotected ethynyl derivative, a method employing K2CO<sup>3</sup> in MeOH described by Wang et al. [39] was used instead, without further purification of the resulting product.

**Scheme 1.** Synthetic approach used for the preparation of the various chromophores: (**a**) **1a**/**1b** (1 eq.) ethynyltrimethylsilane (2 eq.), Pd(PPh<sup>3</sup> )4 (0.15 eq.), PPh<sup>3</sup> (0.06 eq.), CuI (0.12 eq.), (*i*-Pr) <sup>2</sup>NH (2 eq.), dry dioxane, sealed tube under N<sup>2</sup> , 40–45 ◦C, overnight; (**b**) **2a**/**2b** (1 eq.), K2CO<sup>3</sup> (0.15 eq.), dry MeOH, r.t., 4h; (**c**) ethynylcoumarin (1 eq.), aldehyde (1 eq.), Pd(PPh<sup>3</sup> )4 (0.15 eq.), PPh<sup>3</sup> (0.06 eq.), CuI (0.12 eq.), (*i*-Pr) <sup>2</sup>NH (2 eq.), dry dioxane, sealed tube under N<sup>2</sup> , 40–45 ◦C, overnight; (**d**) aldehyde (1 eq.), cyanoacetic acid (3 eq.), piperidine (2.7 eq.), dry acetonitrile, reflux, overnight.

The heterocyclic π-bridges, in the form of brominated aldehydes, were then coupled to the ethynyl moiety through another Sonogashira coupling. In the case of compounds **4** and **7** the aldehydes were not commercially available and as such were prepared by lithiation-formylation of 2,5-dibromothieno [3,2-*b*] thiophene in the case of compound **4**, and alkylation, double bromination [40] and lithiation-formylation of benzotriazole in the case of compound **7**.

With the aldehyde groups now present in the molecules, a Knoevanagel condensation is performed as described by Martins et al. [33] in order to insert the cyanoacrylic acid group that allows the binding to the TiO<sup>2</sup> surface. With this, the final cromophores **8**, **9a**, **9b**, **10** and **11** were obtained with overall yields of 20.2%, 20.3%, 1.1%, 23.6% and 3.6%, respectively. The coumarin derivative with substitution at the 5 and 7-positions (as in compound **9b**) proved to be more difficult to handle then their 6,7-disubstituted analogues, due to insolubility problems. This drawback, together with the increased synthetic complexity of the 5,7-dihydroxycoumarin dye system, led us to focus on the 5,7-disubstitued derivatives.

#### *2.2. Absorption and Fluorescence*

Figure 2 presents the UV-Vis absorption and fluorescence emission spectra of the investigated samples at room temperature in acetonitrile solution. The significant Stokes shift values (in the 2201–3998 cm−<sup>1</sup> range) point to a charge transfer (CT) character of the fluorescence emission band (Table 1). Indeed, when compared to 6,7-dimethoxy-4 methylcoumarin (λabs = 341 nm, λem = 410 nm in ethanol) the absorption and emission spectra of samples **8**–**11** are strongly redshifted [41]. This is associated with the increase in conjugation length of the chromophoric system due to overlapping of π-orbitals of the coumarin group with the π-orbitals of the cyanoacetic acid substituted benzotriazole or thienyl units. Moreover, the absorption spectra of samples **8**–**11** are also redshifted when comparison is made with 7-methoxy-3-acetylcoumarin (λabs = 341 nm) where intramolecular charge transfer (ICT) was found due to the introduction of the acetyl group in the 3-position of the coumarin moiety [42]. This behavior is explained by the increasing ICT character of compounds **8**–**11**, promoted by the strong electron-withdrawing character of the cyanoacetic acid substituted benzotriazole or thienyl units, thus leading to high ground-state dipole moments (values in the 14.560–24.230 D range, see Table 2).

**Figure 2.** Normalized absorption and fluorescence emission spectra for compounds **8**–**11** in acetonitrile solution and in the solid state (adsorbed in TiO<sup>2</sup> films) at 293K.


**Table 1.** Spectroscopic data for compounds **8**–**11** in acetonitrile solution (absorption and fluorescence emission maxima, molar extinction coefficients, ε, and Stokes shift, ∆SS) and absorbed in TiO<sup>2</sup> films (absorption and fluorescence emission maxima) at 293 K.

**Table 2.** Experimental absorption maxima obtained in acetonitrile solution together with the relevant computed absorption properties (predicted vertical excitation energies and associated orbitals transitions major contributions together with oscillator strengths, f, and band gap, Eg) for the investigated compounds obtained by TD-DFT at the CAM-B3LYP/6- 311G(d,p) level of theory after ground-state geometry optimization using the same functional and basis set.


a in brackets the experimental absorption band gap obtained from the intersection between the normalized absorption and fluorescence emission spectra.

> Porous TiO<sup>2</sup> films (about 1 µm thick) were made by spreading TiO<sup>2</sup> paste (ref. 30NR-D, from GreatcellSolar) onto electrically conductive Fluorine-Doped Tin Oxide (FTO) glass, using Scotch Magic tape as a spacer. The film paste was then gradually sintered up to 500 ◦C to reach the anatase TiO<sup>2</sup> phase. When the films had cooled to about 80 ◦C, they were placed for 1 min in concentrated solutions of the samples **8**–**11** in acetonitrile solution (0.1 mM) for adsorption. The absorption (here measured by the fluorescence excitation spectra) and fluorescence emission spectra are depicted in Figure 2. In general, the spectra of the investigated samples in the solid state are blue-shifted by 31–41 nm, with respect to the absorption spectra in MeCN solutions.

#### *2.3. Theoretical Calculations*

The ground state optimized geometry structures and the relevant HOMO and LUMO energy levels, together with their electron density distribution surface plots were obtained at the DFT/CAM-B3LYP/6-311G(d,p) level taking into account the bulk solvent effects of acetonitrile. Frequency analysis for each compound were also computed and did not yield any imaginary frequencies, indicating that the structure of each molecule corresponds to at least a local minimum on the potential energy surface. The geometry optimization for the investigated compound revealed that in general the thienyl and/or the benzotriazole moieties are mostly planar with the coumarin units.

The optimized ground-state molecular geometries found for the investigated compounds were used to obtain the vertical excitation energies, oscillator strengths (f) and excited state compositions in terms of excitations between the occupied and virtual orbitals

using the time-dependent density functional theory (TD-DFT) approach, see Table 2. For samples **8**–**11** the predicted S0→S<sup>1</sup> transitions are in good agreement with the observed lowest energy absorption bands in acetonitrile solution (Table 1). In general, for these transitions the major contribution arises from the HOMO→LUMO orbitals (contributions > 78%). It is worth mentioning that the calculated absorption band gap values agree with the experimental values (see Table 2), thus giving support for the predicted ground-state geometries.

The molecular orbital contours (Figure 3) show that the densities of the HOMO orbitals are, in general, spread over the entire molecules, while the LUMO shows a decrease in the electron density on the coumarin moieties and a concomitant increase in the benzotriazole or thienyl units. The electronic delocalization in the LUMO orbitals gives support for the occurrence of a charge transfer state (CT) in the singlet excited state.

**Figure 3.** DFT//CAM-B3LYP/6311G(d,p) optimized ground-state geometry together with the frontier molecular orbital energy levels and the relevant electronic density contours (calculated at B3LYP/6311G(d,p) level) for the investigated compounds. Additionally, displayed are the predicted optical band gap energy values, Eg.

#### *2.4. Electrochemical Characterization*

The electrochemical properties of the dyes were determined by differential pulse voltammetry (DPV), with the main results being presented in Table 3 (See Supplementary Materials for full DPV data). From the obtained onsets of oxidation and reduction peaks, the HOMO and LUMO energies were estimated. For this purpose, the following equation was used: E [eV] = −(Eonset (V vs. SCE) + 4.44) [43].

**Table 3.** Electrochemical properties in dimethylformamide (DMF) obtained from DPV measurements: HOMO energy level, determined from the onset of the oxidation peak (Eox); LUMO energy level, determined from the onset of the reduction peak (Ered). Gap energy (Eg), calculated with EHOMO—ELUMO.


All calculated band gap values are similar across the various dyes, arising from similar values in the HOMO (−5.74 to −5.60 eV) and in the LUMO energies (−3.62 to

−3.53eV) energies. These differences are within the error for electrochemical determination of the energy of frontier orbitals (~0.1 V) [43], and as such a comparison between the dyes cannot be confidently made. The determined electrochemical band gaps in DMF (~2.1 eV) are lower than the corresponding optical band gaps (~2.7 eV, Table 2), determined in acetonitrile. Solvation and coulombic effects are responsible for often observed differences between electrochemical and optical band gaps [44]. Additionally, the solvatochromic nature of substituted coumarins [45,46] leads to a shorter band gap in more polar solvent, which is the case with DMF.

When compared with other reported coumarin sensitizers (EHOMO: −5.2 eV; ELUMO: −2.4 eV) [30], a significant decrease in orbital energy is observed for the dihydroxycoumarin dyes. This is indicative of a lower electron-donating ability of the dihydroxy substitution pattern in comparison with the indoline moieties employed in the referenced work. On the other hand, this family of dyes presents a shorter band gap (2.0 V vs. 2.4 V), a direct consequence of the more extensive conjugation of the π-system.

#### *2.5. Photovoltaic Performance*

The prepared chromophores were tested in prototype devices and their performance was compared with reference dye N719 (in non-optimized conditions). These results are summarized in Figure 4 and Table 4.

**Figure 4.** (**a**) I–V curves of the test cells based on the synthesized dyes under 100 mW·cm−<sup>2</sup> under simulated AM 1.5 illumination. The results presented correspond to the best performing cell. (**b**) Pictures of the dyes adsorbed on the TiO<sup>2</sup> photoanodes.

**Table 4.** Performance values of the test cells based on the synthesized dyes and reference dye N719 under 100 mW·cm−<sup>2</sup> AM 1.5 illumination. The results presented correspond to the average values of at least two cells per dye, each cell measured 5 times.


Compound **8** was the best performing dye, with an efficiency of 2% and the highest values of VOC and Vmax, 367 and 256 mV, respectively. This marked difference from the other dyes can be attributed to the comparably higher ε value (6.4470 cm−1M−<sup>1</sup> ), as well as the more redshifted absorption when adsorbed on the TiO<sup>2</sup> surface (386 nm). Dye **9a**, which contains a 2,2′ -bithiophene group as π-bridge, obtained the worst efficiency value (0.95%), comparable to dye **10** (1.07%) containing only one thiophene ring. One possible explanation for this result may be the known torsion angles present between the two thiophene units [47], which are not present in the fused ring equivalent **8**, and result in hindered conjugation. This possibility seems to be further supported by the Stokes shift observed for this dye (∆SS = 2201 cm−<sup>1</sup> ), being the lowest of the group, which indicates a low ICT character for electronic transition.

Comparison of dyes **9a** and **9b** allows the evaluation of the effect of the position of the substituents on DSSC performance, and it is immediately apparent from the obtained efficiencies (0.95% vs. 1.78%, respectively) that position 5 (when compared to 6) is a superior choice for donor units in coumarin dyes. The superior efficiency is a consequence of the high photocurrent values (Jsc = 10.2 and Jmax = 7.8 mA·cm−<sup>2</sup> ), which even surpass the results for dye **8**, allowing it to have a good efficiency despite the relatively modest photovoltage (Voc = 339 and Vmax = 227 mV). The improved properties of position 5 as a donor are further supported by the obtained absorption properties, in particular the absorptivity (ε = 4.5480 cm−1M−<sup>1</sup> ) and Stokes shift (∆SS = 3998 cm−<sup>1</sup> ), which are indicative of the dye's effective light absorption and charge transfer character. Additionally, dye **9b** possesses a higher HOMO energy than dye **9a**, which once again points to position 5 of the coumarin unit being a more suitable choice for the inclusion of electron-donating groups.

Inversely, dye **11** shows a high VOC, comparable to dye **8** (359 vs. 367 mV, respectively), yet it has the worst Jsc (5.4 mA·cm−<sup>2</sup> ) of all the synthesized dyes, which can be due to its inferior ε (12,060 cm−1M−<sup>1</sup> ) and Stokes shift (∆SS = 2451 cm−<sup>1</sup> ). The photovoltage values can be attributed to the bulky alkyl chain present in the benzotriazole moiety, which will help suppress charge recombination between the semiconductor and the oxidized redox shuttle [48,49]. Another effect of the presence of this bulky group may be the distortion of the molecule's geometry [50], decreasing planarity and therefore leading to a less efficient electron delocalization and lower ICT character in the transition. Once again, this is reflected in the observed Stokes shift value (∆SS = 2451 cm−<sup>1</sup> ), which is on the lower end of the group, comparable to dye **9a**.

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

#### *3.1. General Information and Instruments*

All solvents and reagents were obtained commercially (Merck KGaA, Darmstadt, Germany) and used without further purification. The drying of the solvents was achieved with M2A molecular sieves (Merck KGaA), as described by Bradley et al. [51].

Thin-layer chromatography (TLC) was carried out on aluminum-backed Kieselgel 60 F254 silica gel plates (Merck KGaA). Plates were visualized with UV light (254 and 336 nm) and in certain cases chemical staining agents (Acidic solution of 2,4-dinitrophenylhidrazine). Preparative-layer chromatography (PLC) was performed on Keiselgel 60 F254 silica gel plates (Merck KGaA) with a thickness of 0.5 mm. Column chromatography was performed using Keiselgel 60 silica gel (Merck KGaA), 70–230 mesh and 230–400 mesh particle sizes as stationary phases, in the cases of regular and flash [52] normal-phase cromatographies, respectively.

The <sup>1</sup>H- and <sup>13</sup>C-NMR (nuclear magnetic spectroscopy) spectra were acquired with a Bruker Avance III 400 (Billerica, MA, USA), at 400 and 101 MHz, respectively. Absorption and fluorescence spectra were recorded on a Cary 5000 UV-Vis-NIR (Santa Clara, CA, USA) and Horiba–Jobin–Ivon Fluoromax4 spectrometers (Longjumeau, France), respectively. The fluorescence spectra were corrected for the wavelength response of the system.

Differential pulse voltammetry (DPV) measurements were performed on a µAutolab Type III potentiostat/galvanostat (Metrohm Autolab B. V., Utrecht, The Netherlands), controlled with GPES (General Purpose Electrochemical System) software version 4.9 (Eco-Chemie, B. V. Software, Utrecht, The Netherlands), using a cylindrical 5 mL threeelectrode cell. All measurements refer to a saturated calomel electrode (SCE, saturated KCl) reference electrode (Metrohm, Utrecht, The Netherlands). A Pt wire was used as

counter-electrode, a glassy carbon electrode (MF-2013, f = 1.6 mm, BAS inc., West Lafayette, IN, USA) was used as the working electrode. Prior to use, the working electrode was polished in aqueous suspensions of 1.0 and 0.3 mm alumina (Buehler, Esslingen, Germany) over 2–7/′′ micro-cloth (Buehler) polishing pads, then rinsed with water and ethanol. This cleaning procedure was systematically applied before any electrochemical measurement. The electrolyte composition was 0.1 M tetrabutylammonium tetrafluoroborate in DMF, with a dye concentration of 1.5 × 10−<sup>4</sup> M. Measurements were performed between 0 and +1.6 V for determination of oxidation potential and between 0 and −1.6 V for determination of reduction potential, with a scan rate of 10 mV/s in both cases. The samples in the electrochemical cell were de-aerated by purging with nitrogen for 10 min prior to, and during, the electrochemical measurements.

High-resolution mass spectra (HRMS) were obtained at the University of Porto, Mass Spectrometry Laboratory (LEM/CEMUP) using a mass spectrometer Linear Trap Quadrupole (LTQ) Orbitrap XLTM (Thermo Fischer Scientific, Bremen, Germany) controlled by a LTQ Tune Plus 2.5.5 and Xcalibur 2.1.0 and at the University of Salamanca (Spain), Elemental Analysis, Chromatography and Mass Spectrometry Service (NUCLEUS), using a High Performance Liquid Chromatography (HPLC) Agilent 1100 coupled to a QSTAR XL Hybrid qTOF (AB Sciex, Framingham, MA, USA) mass spectrometer.

#### *3.2. Synthesis*

3.2.1. Synthesis of 6,7-Dimethoxy-3-((Trimethylsilyl)Ethynyl)Coumarin (**2a**) and 5,7-Dimethoxy-3-((Trimethylsilyl)Ethynyl)Coumarin (**2b**)

To a sealed tube, 0.06 eq. of PPh3, 0.12 eq. of CuI, 0.15 eq. of Pd(PPh3)4, 1 eq. of 3-bromocoumarin (**2a**/**2b**) and 5 mL of dry dioxane were added under a N<sup>2</sup> atmosphere. After a few minutes, 2 eq. of ethynyltrimethylsilane and 0.35 mL 2 eq. of dry (*i*-Pr)2NH were added and the solution was stirred at 45 ◦C overnight under a N<sup>2</sup> atmosphere. Once the reaction was confirmed to be complete by TLC (hexane/AcOEt (7:3 *v*/*v*)), the solution was cooled to room temperature, the solvent was removed under reduced pressure and the solid residue dried in vacuo before being purified by flash chromatography with hexane/AcOEt (7:3 *v*/*v*) as eluent, affording the target compounds.

*6,7-dimethoxy-3-((trimethylsilyl)ethynyl)coumarin* (**2a**). Starting from 347.3 mg (1.22 mmol, 1 eq.) of 3-bromo-6,7-dimethoxycoumarin (**1a**), 343.4 mg (93.2%) of 6,7-dimethoxy-3- ((trimethylsilyl)ethynyl)coumarin (**2a**) were obtained. <sup>1</sup>H-NMR (400 MHz, CDCl3) δ (ppm) 7.83 (s, 1H, H4), 6.82 (s, 1H, H5/H8), 6.80 (s, 1H, H5/H8), 3.95 (s, 3H, H1′/H2′ ), 3.91 (s, 3H, H1′/H2′ ), 0.26 (s, 9H, H5′ ); <sup>13</sup>C-NMR (101 MHz, CDCl3) δ (ppm) 160.0 (C2), 153.5 (C7), 149.8 (C8a), 146.8 (C4/C6), 146.1 (C4/C6), 111.5 (C3/C4a/C5), 109.5 (C3/C4a/C5), 107.7 (C3/C4a/C5), 101.0 (C3′/C4′/C8), 99.9 (C3′/C4′/C8), 98.7 (C3′/C4′/C8), 56.6 (C1′/C2′ ), 56.5 (C1′/C2′ ), −0.1 (C5′ ); HRMS-ESI(+) Calculated for C16H19O4Si [M + H]<sup>+</sup> 303.1047; Found 303.1053.

*5,7-dimethoxy-3-((trimethylsilyl)ethynyl)coumarin* (**2b**). Starting from 147.1 mg (1.22 mmol, 1 eq.) of 3-bromo-5,7-dimethoxycoumarin (**1b**), 114.0 mg (72.9%) of 5,7-dimethoxy-3- ((trimethylsilyl)ethynyl)coumarin (**2b**) were obtained. <sup>1</sup>H-NMR (400 MHz, CDCl3) δ (ppm) 8.14 (s, 1H, H4), 6.37 (s, 1H, H6/H8), 6.25 (d, *J* = 2.4 Hz, 1H, H6/H8), 3.88 (s, 3H, H1′/H2′ ), 3.84 (s, 3H, H1′/H2′ ), 0.25 (s, 9H, H5′ ). HRMS-ESI(+) Calculated for C16H19O4Si [M + H]<sup>+</sup> 303.1047; Found 303.1041.

#### 3.2.2. General Method for the Synthesis of Coupled Aldehydes (**4**–**7**)

To a sealed tube, PPh<sup>3</sup> (0.06 eq), CuI (0.12 eq), Pd(PPh3)<sup>4</sup> (0.15), aldehyde (1 eq.) and 5 mL of dry dioxane were added under a N<sup>2</sup> atmosphere. After a few minutes ethynylcoumarin (**3**) (1 eq.) and dry (*i*-Pr)2NH (2 eq.) were added and the solution was stirred at 45◦C overnight under a N<sup>2</sup> atmosphere. Once the reaction was confirmed to be complete by TLC (hexane/AcOEt (7:3 *v*/*v*)), the solution was cooled to room temperature, the solvent was removed under reduced pressure and the solid residue dried in vacuo

before being purified by flash chromatography with DCM/MeOH (99.8:0.02 *v*/*v* and 99.5:0.05 *v*/*v*) as eluent, affording a bright yellow solid in all cases.

*5-((6,7-dimethoxy-2-oxo-2H-chromen-3-yl)ethynyl)thieno [3,2-b] tiophene-2-carbaldehyde* (**4**). Starting from 104.6 mg (0.42 mmol, 1 eq.) of 5-bromothieno [3,2-*b*] thiophene-2 carbaldehyde, 89.5 mg (51.5%) of 5-((6,7-dimethoxy-2-oxo-2*H*-chromen-3-yl)ethynyl)thieno [3,2-*b*] tiophene-2-carbaldehyde (**4**) were obtained. <sup>1</sup>H-NMR (400 MHz, CDCl3) δ (ppm) 9.99 (s, 1H, H7′ ), 7.92 (s, 1H, H4), 7.89 (s, 1H, H5′/H6′ ), 7.54 (s, 1H, H5′/H6′ ), 6.87 (s, 1H, H5/H8), 6.86 (s, 1H, H5/H8), 3.98 (s, 3H, H1′/H2′ ), 3.94 (s, 3H, H1′/H2′ ); HRMS-ESI(+) Calculated for C20H13O5S<sup>2</sup> [M + H]<sup>+</sup> 397.0199; Found 397.0197.

*5* ′ *-((6,7-dimethoxi-2-oxo-2H-chromen-3-yl)ethynyl)-[2,2*′ *-bithiophene]-5-carbaldehyde* (**5a**). Starting from 132.8 mg (0.49 mmol, 1 eq.) of 5-bromo-[2,2′ -bithiophene]-5-carbaldehyde, 95.2 mg (46.3%) of 5′ -((6,7-dimethoxi-2-oxo-2*H*-chromen-3-yl)ethynyl)-[2,2′ -bithiophene]-5 carbaldehyde (**5a**) were obtained. <sup>1</sup>H-NMR (400 MHz, CD2Cl2) δ (ppm) 9.85 (s, 1H, H9′ ), 7.89 (s, 1H, H4), 7.71 (d, *J* = 4.8 Hz, 1H, H5′/H6′/H7′/H8′ ), 7.30 (s, 3H, H5′/H6′/H7′/H8′ ), 6.87 (s, 1H, H5/H8), 6.85 (s, 1H, H5/H8), 3.92 (s, 3H, H1′/H2′ ), 3.87 (s, 3H, H1′/H2′ ); HRMS-ESI(+) Calculated for C22H15O5S<sup>2</sup> [M + H]<sup>+</sup> 423.0355; Found 423.0348.

*5* ′ *-((5,7-dimethoxi-2-oxo-2H-chromen-3-yl)ethynyl)-[2,2*′ *-bithiophene]-5-carbaldehyde* (**5b**). Starting from 154.0 mg (0.56 mmol, 1 eq.) of 5-bromo-[2,2′ -bithiophene]-5-carbaldehyde, 22.4 mg (14.1%) of 5′ -((5,7-dimethoxi-2-oxo-*2H*-chromen-3-yl)ethynyl)-[2,2′ -bithiophene]- 5-carbaldehyde (**5b**) were obtained. <sup>1</sup>H-NMR (400 MHz, DMF-d7) δ (ppm) 10.02 (s, 1H, H9′ ), 8.29 (s, 1H, H4), 8.08 (d, *J* = 4.03 Hz, 1H, H5′/H6′/H7′/H8′ ), 7.69–7.67 (m, 2H, H5′/H6′/H7′/H8′ ), 7.53 (d, *J* = 4.03 Hz 1H, H5′/H6′/H7′/H8′ ), 6.67 (s, 1H, H6/H8), 6.63 (s, 1H, H6/H8), 4.04 (s, 3H, H1′/H2′ ), 3.99 (s, 3H, H1′/H2′ ); HRMS-ESI(+) Calculated for C22H15O5S<sup>2</sup> [M + H]<sup>+</sup> 423.0355; Found 423.0349.

*5-((6,7-dimethoxy-2-oxo-2H-chromen-3-yl)ethynyl)-thiophene-2-carbaldehyde* (**6**). Starting from 64.8 mg (0.34 mmol, 1 eq.) of 5-bromothiophene-2-carbaldehyde, 58.7 mg (50.8%) of 5-((6,7-dimethoxy-2-oxo-2*H*-chromen-3-yl)ethynyl)-thiophene-2-carbaldehyde (**6**) were obtained. <sup>1</sup>H-NMR (400 MHz, CD2Cl2) δ (ppm) 9.87 (s, 1H, H7′ ), 7.95 (s, 1H, H4), 7.71 (d, *J* = 3.8 Hz, 1H, H5′/H6′ ), 7.41 (d, *J* = 3.7 Hz, 1H, H5′/H6′ ), 6.89 (s, 1H, H5/H8), 6.87 (s, 1H, H5/H8), 3.94 (s, 3H, H1′/H2′ ), 3.89 (s, 3H, H1′/H2′ ); HRMS-ESI(+) Calculated for C18H13O5S [M + H]<sup>+</sup> 341.0478; Found 341.0478.

*2-decyl-7-((6,7-dimethoxy-2-oxo-2H-chromen-3-yl)ethynyl)-2H-benzo[d][1,2,3]triazole-4- carbaldehyde* (**7**). Starting from 113 mg (0.31 mmol, 1 eq.) of 7-dibromo-2-decyl-2*H*-benzo[*d*] [1,2,3]triazole-4-carbaldehyde, 42.5 mg (50.8%) of 2-decyl-7-((6,7-dimethoxy-2-oxo-2*H*chromen-3-yl)ethynyl)-2*H*-benzo[*d*][1,2,3]triazole-4-carbaldehyde (**7**) were obtained. 1H-NMR (400 MHz, CD2Cl2) δ (ppm) 10.50 (s, 1H, H7′ ), 8.11 (s, 1H, H4), 7.99 (d, *J* = 7.3 Hz, 1H, H5′/H6′ ), 7.81 (d, *J* = 7.5 Hz, 1H, H5′/H6′ ), 6.96 (s, 1H, H5/H8), 6.93 (s, 1H, H5/H8), 4.90 (t, *J* = 7.4 Hz, 2H, H1′′), 3.99 (s, 3H, H1′/H2′ ), 3.94 (s, 3H, H1′/H2′ ), 2.26–2.18 (m, 2H, H2′′), 1.43–1.30 (m, 15H, H3′′-H9′′), 0.90 (t, *J* = 6.4 Hz, 3H, H10′′); HRMS-ESI(+) Calculated for C30H34N3O<sup>5</sup> [M + H]<sup>+</sup> 516.2493; Found 516.2509.

#### 3.2.3. General Method for the Synthesis of Final Chromophores (**8**–**11**)

To a round-bottom flask containing aldehyde (1 eq.), cyanoacetic acid (3 eq.), 5 mL of acetonitrile (ACN) and dry piperidine (2.7 eq.) were added and the resulting solution was stirred under reflux for 24 h. Once the reaction was confirmed to be complete by TLC (DCM/MeOH (9.5:0.5 *v*/*v*)), the solvent was evaporated under reduced pressure, the solid residue was washed 3–5 times with ACN, acidified with HCl (10%) and washed 3–5 times with distilled water. After each washing step the solvent used was centrifuged (4500 rpm, 10–30 min) to recover any lost product.

*2-cyano-3-(5-((6,7-dimethoxy-2-oxo-2H-chromen-3-yl)ethynyl)thieno[3,2-b]thiophen-2-yl)acrylic acid* (**8**). Starting from 15 mg (0.038 mmol, 1 eq.) of 5-((6,7-dimethoxy-2-oxo-2*H*-chromen-3-yl)ethynyl)thieno[3,2-*b*]tiophene-2-carbaldehyde (**4**), 7.8 mg (44.5%) of 2-cyano-3-(5-((6,7 dimethoxy-2-oxo-2*H*-chromen-3-yl)ethynyl)thieno[3,2-*b*]thiophen-2-yl)acrylic acid (**8**) were obtained. <sup>1</sup>H-NMR (400 MHz, DMSO-*d*6) δ (ppm) 8.75 (br s, 1H, OH), 8.37 (s, 1H, H4/H7′ ),

8.28 (s, 1H, H4/H7′ ), 8.09 (s, 1H, H5′/H6′ ), 7.89 (s, 1H, H5′/H6′ ), 7.26 (s, 1H, H5/H8), 7.14 (s, 1H, H5/H8), 3.90 (s, 3H, H1′/H2′ ), 3.82 (s, 3H, H1′/H2′ ). HRMS-ESI(+) Calculated for C23H14NO6S<sup>2</sup> [M + H]<sup>+</sup> 464.0257; Found 464.0249.

*2-cyano-3-(5*′ *-((6,7-dimethoxy-2-oxo-2H-chromen-3-yl)ethynyl)-[2,2*′ *-bithiophen]-5-yl)acrylic acid* (**9a**). Starting from 95 mg (0.225 mmol, 1 eq.) of 5′ -((6,7-dimethoxi-2-oxo-2*H*-chromen-3-yl)ethynyl)-[2,2′ -bithiophene]-5-carbaldehyde (**5a**), 54.7 mg (49.7%) of 2-cyano-3-(5′ -((6,7 dimethoxy-2-oxo-2*H*-chromen-3-yl)ethynyl)-[2,2′ -bithiophen]-5-yl)acrylic acid (**9a**) were obtained. <sup>1</sup>H-NMR (400 MHz, DMF-*d*7) δ (ppm) 8.57 (s, 1H, H4/H9′ ), 8.38 (s, 1H, H4/H9′ ), 8.08 (d, *J* = 3.5 Hz, 1H, H5′/H6′/H7′/H8′ ), 7.73 (d, *J* = 3.7 Hz, 1H, H5′/H6′/H7′/H8′ ), 7.70 (d, *J* = 4.0 Hz, 1H, H5′/H6′/H7′/H8′ ), 7.53 (d, *J* = 4.5 Hz, 1H, H5′/H6′/H7′/H8′ ), 7.37 (s, 1H, H5/H8), 7.15 (s, 1H, H5/H8), 4.02 (s, 3H, H1′/H2′ ), 3.92 (s, 3H, H1′/H2′ ). HRMS-ESI(+) Calculated for C25H16NO6S<sup>2</sup> [M + H]<sup>+</sup> 490.0414; Found 490.0408.

*2-cyano-3-(5*′ *-((5,7-dimethoxy-2-oxo-2H-chromen-3-yl)ethynyl)-[2,2*′ *-bithiophen]-5-yl)acrylic acid* (**9b**). Starting from 23.2 mg (0.055 mmol, 1 eq.) of 5′ -((5,7-dimethoxi-2-oxo-2*H*-chromen-3-yl)ethynyl)-[2,2′ -bithiophene]-5-carbaldehyde (**5b**), 14.1 mg (52.5%) of 2-cyano-3-(5′ - ((5,7-dimethoxy-2-oxo-2*H*-chromen-3-yl)ethynyl)-[2,2′ -bithiophen]-5-yl)acrylic acid (**9b**) were obtained. <sup>1</sup>H-NMR (400 MHz, DMF-*d*7) δ (ppm) 8.53 (s, 1H, H4/H9′ ), 8.27 (s, 1H, H4/H9′ ), 8.04 (s, 1H, H5′/H6′/H7′/H8′ ), 7.70 (br s, 1H, H5′/H6′/H7′/H8′ ), 7.67 (br s, 1H, H5′/H6′/H7′/H8′ ), 7.52 (br s, 1H, H5′/H6′/H7′/H8′ ), 6.65 (s, 1H, H6/H8), 6.61 (s, 1H, H6/H8), 4.04 (s, 3H, H1′/H2′ ), 3.99 (s, 3H, H1′/H2′ ). HRMS-ESI(+) Calculated for C25H16NO6S<sup>2</sup> [M + H]<sup>+</sup> 490.0414; Found 490.0406.

*2-cyano-3-(5-((6,7-dimethoxy-2-oxo-2H-chromen-3-yl)ethynyl)-thiophen-2-yl)acrylic acid* (**10**). Starting from 28.1 mg (0.083 mmol, 1 eq.) of 5-((6,7-dimethoxy-2-oxo-2*H*-chromen-3 yl)ethynyl)-thiophene-2-carbaldehyde (**6**), 17.7 mg (50.8%) of 2-cyano-3-(5-((6,7-dimethoxy-2-oxo-2*H*-chromen-3-yl)ethynyl)-thiophen-2-yl)acrylic acid (**10**) were obtained. <sup>1</sup>H-NMR (400 MHz, DMSO-*d*6) δ (ppm) 8.50 (s, 1H, H4/H7′ ), 8.36 (s, 1H, H4/H7′ ), 7.98 (d, *J* = 3.8 Hz, 1H, H5′/H6′ ), 7.57 (d, *J* = 3.9 Hz, 1H, H5′/H6′ ), 7.22 (s, 1H, H5/H8), 7.11 (s, 1H, H5/H8), 3.89 (s, 3H, H1′/H2′ ), 3.81 (s, 3H, H1′/H2′ ). HRMS-ESI(+) Calculated for C21H14NO6S [M + H]<sup>+</sup> 408.0536; Found 408.0529.

*2-cyano-3-(2-decyl-7-((6,7-dimethoxy-2-oxo-2H-chromen-3-yl)ethynyl)-2H-benzo[d][1,2,3]triazol-4-yl)acrylic acid* (**11**). Starting from 51,7 mg (0.10 mmol, 1 eq.) of 2-decyl-7-((6,7-dimethoxy-2-oxo-2*H*-chromen-3-yl)ethynyl)-2*H*-benzo[*d*][1,2,3]triazole-4-carbaldehyde (**7**), 8.9 mg (50.8%) of 2 cyano-3-(2-decyl-7-((6,7-dimethoxy-2-oxo-2*H*-chromen-3-yl)ethynyl)-2*H*-benzo[*d*][1,2,3]triazol-4-yl)acrylic acid (11) were obtained. 1H-NMR (400 MHz, DMF-*d*7) δ (ppm) 8.92 (s, 1H, H4/H7′ ), 8.55 (d, *J* = 8.0 Hz, 1H, H5′/H6′ ), 8.46 (s, 1H, H4/H7′ ), 7.97 (d, *J* = 7.8 Hz, 1H, H5′/H6′ ), 7.42 (s, 1H, H5/H8), 7.17 (s, 1H, H5/H8), 4.95 (t, *J* = 7.3 Hz, 2H, H1′′), 4.04 (s, 3H, H1′/H2′ ), 3.93 (s, 3H, H1′/H2′ ), 2.22–2.15 (m, 2H, H2′′), 1.44–1.14 (m, 20H, H3′ -H9′′), 0.85 (t, *J* = 6.4 Hz, 5H, H10′′). HRMS-ESI(+) Calculated for C33H35N4O<sup>6</sup> [M + H]<sup>+</sup> 583.2551; Found 583.2542.

#### *3.3. Theoretical Calculations*

The ground state molecular geometry was optimized using the density functional theory (DFT) by means of the Gaussian 09 program (Gaussian, Wallingford, CT, USA) [53], under CAM-B3LYP/6-311G(d,p) level [54,55] taking into account the bulk solvent effects of acetonitrile [56]. Optimal geometries were determined on isolated entities in acetonitrile and no conformation restrictions were imposed. For the resulting optimized geometries time-dependent DFT calculations (using the same functional and basis set as those in the previously calculations) were performed to predict the vertical electronic excitation energies. Molecular orbital contours were predicted at the B3LYP/6-311G(d,p) level of theory in vacuo and were plotted using GaussView 5.0.8 (Gaussian). The orbitals transitions percentage contributions of the predicted vertical excitation were calculated using GaussSum 2.2 (Dublin, Ireland) [57].

#### *3.4. DSSCs Fabrication and Photovoltaic Characterization*

The detailed procedure has been described elsewhere [58]. The conductive FTO-glass (TEC7, Greatcell Solar, Queanbeyan, Australia) used for the preparation of the transparent electrodes was first cleaned with detergent and then washed with water and ethanol. To prepare the anodes, the conductive glass plates (area: 15 cm × 4 cm) were immersed in a TiCl4/water solution (40 mM) at 70 ◦C for 30 min, washed with water and ethanol and sintered at 500 ◦C for 30 min. This procedure is essential to improve the adherence of the subsequently deposited nanocrystalline layers to the glass plates, as well as to serve as a 'blocking-layer', helping to block charge recombination between electrons in the FTO and holes in the I−/I<sup>3</sup> <sup>−</sup> redox couple. Afterwards, the TiO<sup>2</sup> nanocrystalline layers were deposited on these pre-treated FTO plates by screen-printing the transparent titania paste (18NR-T, Greatcell Solar) using a frame with polyester fibers with 43.80 mesh per cm<sup>2</sup> . This procedure, involving two steps (coating and drying at 125 ◦C), was repeated twice. The TiO2-coated plates were gradually heated up to 325 ◦C, then the temperature increased to 375 ◦C in 5 min, and afterwards to 500 ◦C. The plates were sintered at this temperature for 30 min, and finally cooled down to room temperature. A second treatment with the same TiCl4/water solution (40 mM) was performed, following the procedure described previously. This second TiCl<sup>4</sup> treatment is also an optimization step that enhances the surface roughness for dye adsorption, thus positively affecting the photocurrent produced by the cell under illumination. Finally, a coating of reflective titania paste (WER2-O, Greatcell Solar) was deposited by screen-printing and sintered at 500 ◦C. This layer of 150–200 nm sized anatase particles functions as a 'photon-trapping' layer that further improves the photocurrent. Each anode was cut into rectangular pieces (area: 2 cm × 1.5 cm) with a spot area of 0.196 cm<sup>2</sup> and a thickness of 15 µm. The prepared anodes were soaked for 16 h in a 0.5 mM solution of the dye in dichloromethane:methanol:H2O (65:20:2), at room temperature in the dark. The excess dye was removed by rinsing the photoanodes with the same solvent as that employed for the dye solution.

Each counter-electrode consisted of an FTO-glass plate (area: 2 cm × 2 cm) in which a hole (1.0 mm diameter) was drilled. The perforated substrates were washed and cleaned with water and ethanol to remove any residual glass powder and organic contaminants. The transparent Pt catalyst (PT1, Greatcell Solar) was deposited on the conductive face of the FTO-glass by doctor blade: one edge of the glass plate was covered with a strip of an adhesive tape (3 M Magic) both to control the thickness of the film and to mask an electric contact strip. The Pt paste was spread uniformly on the substrate by sliding a glass rod along the tape spacer. The adhesive tape strip was removed, and the glasses heated at 550 ◦C for 30 min. The photoanode and the Pt counter-electrode were assembled into a sandwich type arrangement and sealed (using a thermopress) with a hot melt gasket made of Surlyn ionomer (Meltonix 1170-25, Solaronix SA, Aubonne, Switzerland). The electrolyte was prepared by dissolving the redox couple, I−/I<sup>3</sup> <sup>−</sup> (0.8 M LiI and 0.05 M I2), in an acetonitrile/valeronitrile (85:15, % *v*/*v*) mixture. The electrolyte was introduced into the cell via backfilling under vacuum through the hole drilled in the back of the cathode. Finally, the hole was sealed with adhesive tape.

For each compound, at least two cells were assembled under the same conditions, and the efficiencies were measured 5 times for each cell resulting in a minimum of 10 measurements per compound.

Current-Voltage curves were recorded with a digital Keithley SourceMeter multimeter (PVIV-1A) (Newport, M. T. Brandão, Porto, Portugal) connected to a PC. Simulated sunlight irradiation was provided by an Oriel solar simulator (Model LCS-100 Small Area Sol1A, 300 W Xe Arc lamp equipped with AM 1.5 filter, 100 mW/cm<sup>2</sup> ) (Newport, M. T. Brandão). The thickness of the oxide film deposited on the photoanodes was measured using an Alpha-Step D600 Stylus Profiler (KLA-Tencor, Milpitas, CA, USA).

#### **4. Conclusions**

In the current work, five new 3-ethynylaryldimethoxycoumarin-based chromophores were synthesized by previously reported methods, with the aim of investigating not only the effect of varying the heterocyclic π-bridge on the device performance, but also assess the difference in properties between the 6,7- and 5,7-substitution patterns on the coumarin moiety. Through absorption and fluorescence emission spectroscopy, both in solution and adsorbed onto porous TiO2, it was determined that dyes **8** and **9b** present the most redshifted absorption, highest absorptivity, and most pronounced Stokes shift, which help explain their high photocurrent values. Additionally, the difference in the theoretically calculated HOMO orbitals of dyes **9a** and **9b** further support the more electrondonating nature of the 5,7-substitution pattern. Despite this, this set of dyes demonstrates comparatively lower conversion efficiencies than other coumarin dyes [26–32], which can be attributed to the superior donor ability of nitrogen-based donors there used. Overall, we can see that the increased planarity of the thieno [3,2-*b*] thiophene π-bridge (dye **8**) and the superior donor ability of the 5,7-disubstituted coumarin are promising avenues for the synthesis of more efficient coumarin-based donors.

**Supplementary Materials:** The following are available online, the detailed synthetic procedures, the NMR and HRMS spectral data and the full differential pulse voltammograms of dyes **8**–**11**.

**Author Contributions:** Conceptualization, P.S.B. and A.J.P.; methodology, P.S.B., A.J.P., J.C.L. and J.P.; validation, P.S.B., J.C.L., A.J.P. and J.P.; formal analysis, A.L.P., J.S., G.M. and E.G.R.; investigation, J.S., G.M., A.L.P. and E.G.R.; resources, P.S.B., A.J.P., J.C.L. and J.P.; data curation, P.S.B., A.J.P., J.C.L. and J.P.; writing—original draft preparation, J.S., P.S.B., J.P.; writing—review and editing, P.S.B., A.J.P., J.C.L. and J.P.; visualization, J.S., A.L.P. and J.P.; supervision, P.S.B., J.C.L., A.J.P. and J.P.; project administration, P.S.B.; funding acquisition, P.S.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was performed under the projects PTDC/QUI-QOR/7450/2020 "Organic Redox Mediators for Energy Conversion" through FCT—Fundação para a Ciência e a Tecnologia I. P. and POCI-01-0145-FEDER-016387 "SunStorage—Harvesting and storage of solar energy", funded by European Regional Development Fund (ERDF), through COMPETE 2020—Operational Programme for Competitiveness and Internationalisation (OPCI). This work was also supported by the Associate Laboratory for Green Chemistry—LAQV which is financed by national funds from FCT/MCTES (UIDB/50006/2020 and UIDP/50006/2020). FCT/MCTES is also acknowledged for the National NMR Facility (RECI/BBB-BQB/0230/2012 and RECI/BBB-BEP/0124/2012,) and PhD grants 2020.09047.BD (J.S.), PD/BD/135087/2017 (A.L.P.) and PD/BD/145324/2019/ (G.M.).

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

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data provided in this study is available in the article and Supplementary Material file submitted.

**Acknowledgments:** The authors acknowledge Hugo Cruz for his support with CV and DPV measurements. The authors acknowledge as well their respective, current institutions of affiliation: Universidade NOVA de Lisboa and University of Coimbra.

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

**Sample Availability:** Samples of the compounds are not available from the authors.

#### **References**


## *Article* **Structural Characterization of Mono and Dihydroxylated Umbelliferone Derivatives**

**Rubén Seoane-Rivero 1,\*, Estibaliz Ruiz-Bilbao <sup>2</sup> , Rodrigo Navarro 3,4,\* , José Manuel Laza <sup>5</sup> , José María Cuevas <sup>1</sup> , Beñat Artetxe <sup>2</sup> , Juan M. Gutiérrez-Zorrilla <sup>2</sup> , José Luis Vilas-Vilela <sup>5</sup> and Ángel Marcos-Fernandez 3,4**


Academic Editor: Maria João Matos Received: 23 June 2020; Accepted: 28 July 2020; Published: 31 July 2020

**Abstract:** Coumarin derivatives are a class of compounds with a pronounced wide range of applications, especially in biological activities, in the medicine, pharmacology, cosmetics, coatings and food industry. Their potential applications are highly dependent on the nature of the substituents attached to their nucleus. These substituents modulate their photochemical and photophysical properties, as well as their interactions in their crystalline form, which largely determines the final field of application. Therefore, in this work a series of mono and dihydroxylated coumarin derivatives with different chemical substituents were synthesized and characterized by UV-Visible spectroscopy, thermal analysis (differential scanning calorimetry (DSC) and TGA), <sup>1</sup>H NMR and X-Ray Diffraction to identify limitations and possibilities as a function of the molecular structure for expanding their applications in polymer science.

**Keywords:** coumarin; hydroxyl-modified coumarin; photophysical; thermal and structural characterization

#### **1. Introduction**

Coumarins (chromen-2-ones) are a family of benzopyrones widely distributed in nature. Since 1902, when Ciamician and Silber found that coumarin had the ability to be photoreactive, the photo-cyclodimerization and photo-cleavage of this product has received a lot of attention by several investigation groups [1–4]. Their structure are a class of lactones based on a benzene ring fused to α-pyrone ring, as can be seen in Scheme 1A [5,6]. Coumarins represent an important family of naturally occurring and/or synthetic oxygen-containing heterocycles, bearing a typical benzopyrone framework. One of the most important characteristics of coumarin derivatives is that they can undergo reversible photo-responsible reactions; depending on the type of irradiated wavelength, these moieties can yield a cyclobutane through dimerization or they can cleavage, reforming the double bond C=C. Thus, when irradiated at >300 nm, a [2 + 2] cycloaddition reaction takes place, forming a cyclobutane ring; in contrast, irradiating at 254 nm a photo-scission reaction leads to the original coumarin structures (Scheme 1B) [7–10].

**Scheme 1.** (**A**) Coumarin core structure and (**B**) photo-reversibility of coumarin moieties.

Umbelliferone is a benzopyrone (also known as 7-hydroxycoumarin, hydrangine, skimmetine, and beta-umbelliferone) and belongs to the Coumarin family which is commonly found in plants [11]. The word "Umbelliferone" was originated from the plant which belongs to the Umbelliferae family. It includes significant herbs such as celery, carrot, garden angelica, sanicle, parsley, cumin, alexanders, big leaf hydrangea, fennel, asafoetida, Justicia pectoralis and giant hogweed. The phenolic coumarins, which are derived from plants, have been supposed to play a vital role in our daily life, due to their antioxidant property and are taken in the human diet in the form of vegetables and fruit [12].

β Under acidic conditions and low temperatures, highly activated phenols, such as resorcinol and β-carbonyl ester, easily yield the desired coumarins; this synthetic approach is based on the biosynthesis of umbelliferone (Scheme 2). β

**Scheme 2.** Biosynthetic scheme of umbelliferone.

Moreover, coumarin and its derivatives are small molecular weight compounds that have demonstrated very interesting physical, chemical and pharmacological properties with broad applicability as biochemicals (drugs, cosmetics, dyes, antibacterials, etc.) [13]. The diverse oriented synthetic routes have led to very different derivatives with usefulness not only as biologically active agents [14–17] or optical materials [18–21], but in macromolecular chemistry they also can be seen as very diverse polymer backbones with photoreactive properties [7,22–24]. Owing to the strong demand imposed by various government organizations for the design of polymeric sustainable materials in a more circular economic model, nowadays coumarin derivatives have acquired relevant importance in this field.

To elicit photoactive polymers with self-healing properties, different coumarin derivatives have been previously incorporated in various polymeric backbones [25–28]. In the particular case of polyurethanes, some research groups have recently introduced different hydroxylated derivatives either within the hard segment (chain end or chain extender) or within the soft segment (coumarin functionalized polycaprolactone diols) [2,29–35]. More recently, we reported an outstanding three times increment increase in the tensile strength of polyurethanes with difunctional hydroxy-coumarins, which led to new irradiated polyurethanes having mechanical properties superior to any coumarin containing materials described in literature [35]. Motivated by this issue, other hydroxylated coumarins, with a high structural analogy, were also introduced into the polyurethane matrices [2,33]. Comparing the experimental data derived from those works, uneven behaviors were observed, for instance, some coumarin hydroxyl-derivatives could not be introduced into the soft segment or the photo-dimerization yields varied considerably, to name a few.

Owing to the coumarin contents in the polyurethane formulations being too low, a systematic study has been focused on the isolated coumarins. Therefore, in the present work those isolated hydroxy-coumarin compounds with different functionalities and chemical structure were synthesized and exhaustively characterized to shed light and understanding on those differences. Despite the structural similarity and considering the variability of substituents in the C7 position of the coumarins, an understanding at the atomistic level of the electronic interactions that determine the spectroscopic properties of coumarins has also been carried out. Single-crystal X-ray diffraction analyses revealed the key role of weak supramolecular forces in the self-assembly of molecular species within the crystal packing. The oxygen-rich molecular structures allow O–H···O hydrogen bonds and C–H···O type contacts to be established, which is demonstrated in their capacity to undergo edge-to-edge self-association [36]. Moreover, the presence of two fused aromatic rings in the coumarin structure contributes, with π-π stacking interactions, to the robustness of the system. These structural studies were essential to relate their different arrangements with the thermal behavior observed in DSC analyses. π π

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

#### *2.1. Synthetic Approach*

The chemical structures of the monohydroxy and dihydroxy-derived coumarins have been collected in Figure 1.

**Figure 1.** Chemical structures of functionalized coumarins bearing hydroxyl groups.

All studied coumarins studied are based on the umbelliferone core. Firstly, the HMC product was prepared and later different chemical reactions (etherification or esterification) were carried out on this product to achieve the rest of presented coumarins. These reactions were focused on the lateral group located at C7 carbon, which was varied to assess its impact on the performance and features of these photo-reactive systems. In essence, two types of coumarins have been proposed, monohydroxy and dihydroxy, these hydroxyl-functional groups could, in turn, be directly bound to the umbelliferone ring or separated by a spacer. The synthetic routes used for the preparation of these coumarin derivatives have been included in the Supporting Information. Likewise, a complete spectroscopic characterization of each coumarin has also been included in the Supporting Information. Additionally, it is important to note that the synthetic procedures were simple, easily scalable and yields were good in most cases. Furthermore, due to the absence of post-purification processes, this set of features are key elements for the application of these coumarins in Industry. Indeed, the quality of the obtained products following these synthetic routes were high enough in order to incorporate them directly into the polyurethane formulations, leading to polymer coatings with interesting performances [33].

On the other hand, despite the high purity of these products during their synthesis, additional recrystallization steps were required for structural analysis by X-ray diffraction. Crystals suitable were obtained by dissolving the final products in their corresponding hot solvents (ca. 90 ◦C) and leaving the resulting solutions to slowly evaporate in an open container. First attempts were carried out by using the solvents of the synthetic procedure, but some of the cases did not yield crystals of enough quality. Thus, mixtures of solvents with different polarities were employed for the recrystallization. The longer the aliphatic chain of the substituent, the lower the polarity of the molecule. Due to this fact, coumarins with shorter substituents (HMC and DHMC) were found to crystallize better in solvents with higher polarity than coumarins with longer substituents (HEOMC and DHEOMC). That is, single crystals of HMC were obtained in the most polar solvent mixture, EtOAc:EtOH (3:1), whereas those of DHEOMC were isolated from the most nonpolar solvent mixture, EtOAc:diethyl ether (1:1). The remaining coumarins were easily recrystallized from EtOAc.

#### *2.2. UV Experiments*

Owing to strong absorption of UV light shown by coumarins, in the first set of experiments, the UV spectra of prepared coumarins in aqueous solution were recorded for the first time.

In Figure 2, the absorption UV-spectra of pristine coumarin compounds (without irradiation) are shown. The concentration of these solutions ranged from 0.2 to 0.4 mM. In all cases, absorption of coumarins showed a π-π\* transition between 260 and 300 nm attributed to electrons of the conjugated benzene nucleus and another π-π\* transition between 310–340 nm assigned to pyrone nucleus [30]. Only in the case of DHMC, were these two transitions completely distinguishable, for the rest of coumarins the transition of the benzene ring appeared as a shoulder on the heterocycle transition band. However, the pyrone-associated transition is much more severely affected by UV radiation, while the transition of the benzene ring is practically unaltered. π π π π

**Figure 2.** Absorption spectra of the coumarin compounds.

It is well known that UV-radiation markedly affects the reversible dimerization process of coumarin (Scheme 1). Thus, depending on the wavelength used as the source of excitation, this equilibrium shifts to one side or the other. When the aqueous solutions were irradiated with a set of five lamps of 354 nm, photo-dimerization reaction was performed, however, with a set of five lamps of 254 nm, photo-cleavage reaction was induced.

The absorption UV-Vis spectra of HEOMC at 354 nm (A) and 250 nm (B) are depicted in Figure 3. In the photo-dimerization reaction (Figure 3A), the maximum transition at 320 nm gradually decreased in intensity with the irradiation time. As this band was associated with the heterocycle ring, the double bond of the pyrone core progressively disappeared, leading to the formation of a cyclobutene ring by cycloaddition [2 + 2]. In contrast, during photo-cleavage (Figure 3B), the double bond was restored and the absorption peak with the maximum recovered. Additionally, as the intensity of the band at 320 nm gradually decreased, the transition of the benzene ring was easy to detect, because it remained invariant with UV radiation.

**Figure 3.** Photo-dimerization (**A**) and photo-cleavage (**B**) spectra of HEOMC derivative.

The photoreactivity feature of characterized solutions can be quantitatively described by the time dependence of the maximum peak height [30]. Photo-dimerization degree was estimated from Equation (1), where A<sup>t</sup> shows the absorbance at maximum peak at time t, and A<sup>o</sup> the original absorbance at maximum wavelength prior to 354 nm exposure. When aqueous solutions were exposed to 254 nm, in order to characterize the recovery percentage, it used Equation (2); where A∞ denotes the absorbance after the solution exposed to 254 nm, A∞ shows the minimum absorbance at maximum peak after exposure to 350 nm UV light, and A<sup>o</sup> has the same meaning as that in Equation (1).

$$\% \text{ dimensionality degree} = \left(1 - \frac{\text{A}\_{\text{t}}}{\text{A}\_{0}}\right) \times 100 \tag{1}$$

$$\% \text{photlecule} \times \text{degrees} = \left(\frac{\text{A}\_{\text{os}}^{\prime} - \text{A}\_{\text{l}}}{\text{A}\_{0} - \text{A}\_{\text{os}}}\right) \times 100\tag{2}$$

Significant differences in photo-dimerization and photo-cleavage were found in the first cycle of irradiation between studied coumarins. In Figure 4, the variation of the UV-absorbance with irradiation time of the four coumarin derivatives is shown.

**Figure 4.** Photo-reversibility cycle of coumarin derivatives.

With respect to photo-dimerization, the DHMC coumarin product denoted a low dimerization degree: 13%. This could be based on the UV-spectrum of this derivative. As shown in Figure 2, the UV-spectrum of DHMC presented its absorption bands to higher wavelengths compared with its counterparts. Therefore, with an irradiation at 354 nm, the absorption of DHMC would be very weak and its dimerization would be hampered. Hence, to increase its dimerization degree, it would be necessary to irradiate at 313 nm (closer to the maximum peak of DHMC), but Seoane et al. demonstrated that irradiation at 313 nm gave a very strong irreversibility respect to photo-cleavage [2]. Optimum photoreversibility was achieved when irradiation was carried out with 354 and 254 nm sets of lamps.

Regarding the other molecules studied (HMC, HEOMC, DHEOMC), the irradiation at 354 nm led to photo-dimerization yields of about 70%. Subsequently, only the HEOMC and DHEOMC derivatives largely recovered the initial absorbance by irradiation at 254 nm. In contrast, the HMC dimer was not able to significantly cleave in aqueous solution, and the absorbance values of this solution practically did not vary with irradiation at 254 nm.

It is important to note that the HMC molecule has its functional group linked to the aromatic ring of coumarin, and this could be one reason to argue this behavior. With regard to how to improve the photoreactivity, there are some studies that denote that the addition of substituents to the coumarin dimer can improve the cleavage reaction efficiency, but Jiang et al. explained that it is not clear how the substituents modify the cleavage dynamics, or why they generally lead to enhanced efficiencies compared to the unsubstituted coumarin dimer [6].

One of the most important properties of coumarin derivatives is the photo-reversibility. This feature has been able to be studied in HEOMC and DHEOMC derivatives. In Table 1, the photo-dimerization and photo-scission yields for each cycle have been collected. Monohydroxy coumarin HEOMC progressively lost its photo-reversible capacity after the end of each cycle; in fact, after the third dimerization cycle (354 nm), only 30% of its coumarins had dimerized. In contrast, the dimers of dihydroxylated coumarin (DHEOMC) were easily cleaved after being irradiated with light at 254 nm. Although, its dimerization capacity was also depressed after each dimerization cycle, but the decrease was less pronounced than its monohydroxylated counterpart.


**Table 1.** Dimerization and cleavage conversions of coumarin derivatives.

The kinetic of photo-dimerization (irradiation at 354 nm) and photo-cleavage (irradiation at 254 nm) for the HEOMC derivative is depicted in Figure 5. At the end of the photo-cleavage cycle, the absorbance value was lower than the starting point, so that the coumarin photo-reversibility was not perfect. The same effect was shown after the next cycle of photo-cleavage (end of cycle 2), where the absorbance was still even lower. Our findings suggested that photo-dimerization and photo-cleavage lost efficiency cycle by cycle. Indeed, the dimerization degree was slightly reduced whilst the cleavage dropped substantially as the number of cycles were repeated. Some authors attributed that the decrease in the photo-reversibility of coumarins could be due to the existence of an equilibrium between coumarin and its dimers, as well as the formation of non-cleavable dimers, because the lactone ring had probably opened [37].

**Figure 5.** Photo-dimerization and photo-cleavage absorptions of HEOMC derivative cycle by cycle.

#### *2.3. Theory: Absorption UV-Vis Spectra*

As studied coumarins have exhibited disparate UV-vis behavior and been previously used as photochemical crosslinking agents to obtain high-performance coatings, the need has been raised to understand the impact of the substituent on the electron densities of coumarins.

Considering that the experimental UV-vis spectra were acquired in aqueous solution, the theoretical model should also include this effect. For this, the polarizable continuum model (PCM) was used, because this approach reproduced the experimental data with high precision at a low computational cost. Hence, we started with the optimized geometries of the ground state in the gas phase, which were then used as starting configurations for a geometric optimization of the molecules in aqueous solution using the PCM. Finally, these optimized settings served to determine the absorption properties of coumarins.

For all electrostatic potential surfaces (Figure S9), the carbonyl group (C=O) of the pyrone is the region with the highest concentration of electrons, making it the area with the lowest potential (in red). Furthermore, the coumarin family bearing two hydroxyl groups (DHMC, DHEOMC) presented a second region of low potential. This region was located in the other ester group and remained almost perpendicular to the coumarin ring. On the other hand, in all the coumarins studied, the regions with high potential correspond to hydroxyl groups, so that they will be the reactive centers during, for example, polymerization reactions.

In general, the maximum absorption for a molecule usually approximates the energy difference between the frontier orbitals (HOMO and LUMO). Eventually, the relative order of the calculated values of the HOMO–LUMO gap follows the same order with respect to the measured absorption peaks (Figure 6). Therefore, this observation could indicate that the dominant transition in UV-Vis spectra corresponds to the HOMO–LUMO transition for the C7-substituted coumarin family. However, in order to improve the description of the maximum UV-Vis absorption peaks, Time-dependent density functional theory (TD-DFT) calculations were performed for all coumarins, keeping the coumarin configuration frozen in the ground state according to the PCM. In this case, the theoretical results within the TD-DFT framework follow the same trend described above, but the maximum absorption peak is closer to the experimental one.

**Figure 6.** Comparison between experimental UV-Vis absorption peaks and calculated HOMO–LUMO gap at gas-phase and aqueous phase for C7-substituted coumarins.

Additionally, a deeper analysis of the TD-DFT transitions yields very interesting information (Table S1). For example, for all the cases studied, the first excited singlet state has the highest oscillator strength, except for coumarin DHMC, where the second excited singlet state also exhibits a significant contribution from the oscillator strength. As described above, the transition to the first singlet excited state corresponds mainly to the HOMO–LUMO transition. This transition has the same character for all C7-substituted coumarins except for DHMC, which also presents lower contributions than other transitions, such as HOMO-1 -> LUMO (2 -> 1) and HOMO-1 -> LUMO+1. (2 -> 2').

On the other hand, the second singlet excitation state (S2) presents a very similar nature between the coumarins HMC, HEOMC and DHEOMC. For these cases, the transition is mainly governed by the HOMO-1 -> LUMO (2 -> 1') transition and to a lesser extent by the HOMO -> LUMO +1 (1 -> 2') transition. However, for DHMC, additionally in this transition to S2 there is also a slight contribution from the HOMO–LUMO (1 -> 1') transition. Therefore, DHMC coumarin has a UV-Vis absorption spectrum with more significant differences compared to its counterparts, and the shape of this spectrum determines its behavior against UV radiation.

π π Through the morphological analysis of the frontier orbitals, which participate in the absorption processes, it should be possible to understand the optical properties of UV light absorption. Figure 7 depicts the frontier orbitals for the four C7-substituted coumarins. The two main frontier orbitals (HOMO and LUMO) of the coumarins studied are mainly extended along the heterocyclic ring defining delocalized π-orbitals. Only in the case of DHMC, is delocalization extended to the ester group located at C7; as for the rest of coumarin counterparts, the extension of the π-orbital is reduced to the C7 oxygen atom. This extension may be due to the aromatic ester nature of DHMC, while for the other ester coumarin (DHEOMC) it has a two-carbon spacer that breaks this conjugation. For the other two main orbitals (HOMO-1 and LUMO+1), they are preferentially concentrated in the benzene ring of coumarins, but only in the specific case of DHMC does the transition between these orbitals have a significant contribution.

**Figure 7.** Distribution and schematic representation of the frontier orbitals of the coumarins studied.

#### *2.4. Crystal Structures*

*σ*

*σ*

. Crystallographic data for compounds HMC, HEOMC, DHMC and DHEOMC are compiled in Table 2.

$$\mathbb{P}^{a}\mathbb{R}(F) = \sum \|F\_{0} - F\_{c}\| / \sum \left| F\_{0} \right| . \, ^{b}w \mathbb{R}(F^{2}) = \left\{ \sum \left[ w \left( F\_{0}^{2} - F\_{c}^{2} \right)^{2} \right] / \sum \left[ w \left( F\_{0}^{2} \right)^{2} \right] \right\}^{1/2}$$

− *λ α β γ μ* Thermal vibrations of all non-hydrogen atoms were refined anisotropically. Crystals of the DHMC derivative were systematically of much poorer quality than their analogues. Measurements on several crystals were made from several different batches in quest of a better set of crystallographic data but without any success. Although completeness was as low as 91%, and the used restrain/parameter ratio was considerably high, the best structural model was obtained solving the structure in the chiral Pc space group. The two coumarin molecules that form the asymmetric unit of DHMC were found to be involved in crystallographic disorder. The C and O atoms were refined with free population factors, resulting in a chemical occupancy of ca. 0.8 for the main form and 0.2 for the minor one. In order to model the disorder, several restrictions were applied for the thermal ellipsoids of C and O atoms belonging to the minor phase (ISOR). Some C–C and C–O bond lengths were also restricted to 1.54(2) and 1.43(2), respectively (DFIX). In all cases, hydrogen atoms of the organic molecules were placed in calculated sites using standard SHELXL parameters, whereas those from hydration water molecules were located in Fourier maps and restrained to O–H bond lengths of 0.84(2)Å. For DHMC, H atoms belonging to hydroxyl groups were placed in the Fourier map and bond lengths and angles were restricted using DFIX and DANG commands.


**Table 2.** Crystallographic data for HMC, HEOMC, DHMC and DHEOMC.

Single-crystal X-ray diffraction experiments revealed that molecular structures of the synthetized compounds are in good agreement with those proposed from <sup>1</sup>H-NMR studies. In all cases, the coumarin backbone consists of a completely planar benzolactone ring bearing a methyl group in the C4 position which displays different substituents at the C7 position (Figure S10). Our findings on the crystal structure of monohydrate HMC and the supramolecular interactions governing its crystal structure were aligned with previously reported works [38,39]. Using this system as a base, significant changes within supramolecular interactions in the crystal structure for the rest of studied coumarins were also observed, owing to the insertion of different substituents at position C7. Indeed, intermolecular forces involved in the crystal packing of HEOMC, DHMC and DHEOMC include π-π interactions established between aromatic ring, O–H···O hydrogen bonds and C–H···O-type contacts. Their geometrical parameters are compiled in Tables S2 and S3. All the bond lengths and angles are in concordance with those found in literature [36,40–42].

In a close analysis of the HEOMC crystal structure, we observed that HEOMC crystallized in the monoclinic space group *P*21/*n* and its asymmetric unit contained one HEOMC moiety and one hydration water molecule. As shown in Figure 8, the coumarin molecules were packed antiparallelly forming columns along the crystallographic z axis through π-π stacking. These arrangements were involved in an extensive three-dimensional network of OW–H···O and O–H···O<sup>W</sup> hydrogen bonds established between the hydroxyl group of this monohydroxy coumarin and the hydration water molecules. Additionally, weak C–H···O-type contacts linked adjacent columns along the [100] direction. ሺሻ ൌ ห| െ

π π



ሻ ൌ ሼሾሺ

<sup>ଶ</sup> െ ଶ ሻ ଶ

ሿ / ሾሺ

ଶ ሻ ଶ ሿሽ ଵ/ଶ

π π

**Figure 8.** View of the crystal packing of HEOMC along the crystallographic x-axis. Intermolecular interactions are depicted as dashed lines: O-H···O, red; C-H···O, green; π-π purple. π π

On the other hand, the crystal structure of DHMC belonged to the monoclinic Pc space group, including two coumarin moieties in the asymmetric unit. Both coumarins were involved in crystallographic disorder; therefore, the forms with higher occupancy were only represented in Figure 9. DHMC molecules were packed antiparallelly, forming columns along the crystallographic y axis and interacting through π-π stacking and O–H···O hydrogen bonds. These strong interactions involved the O atoms from the ester group and both hydroxyl moieties. Owing to these four O atoms within this dihydroxy-coumarin, one-dimensional arrangements were connected to each other by creating an extensive network of C–H···O contacts. π π

**Figure 9.** View of the crystal packing of DHMC along the crystallographic y axis (**left**) and detail of the supramolecular one-dimensional arrangement (**right**).

Finally, DHEOMC crystallized in the triclinic space group P-1; the lower symmetry in comparison to the other systems could come from the introduction of a flexible spacer between dihydroxyl groups and umbelliferone moiety. In the case of this dihydroxy-coumarin, the crystal structure showed a bidimensional character with ribbons that stacked along the [0–11] direction. These ribbons were constituted by double-chains of coumarins that interacted through O–H···O-type hydrogen bonds and involved O atoms from hydroxyl or ester groups of different DHEOMC moieties. Contiguous

π π

double-chains were linked to each other via π-π interactions along the crystallographic y-axis, and together with C–H···O contacts, completed the bidimensional arrangement (Figure 10).

**Figure 10.** View of the crystal packing of DHEOMC along the crystallographic X axis indicating its bidimensional character. Hydrogen-bonded double-chains are highlighted in green.

Ultimately, despite the symmetry of the space group of the coumarins studied, it should be highlighted that one molecule of water was included within the crystalline structures of monohydroxylated coumarins, whilst their dihydroxylated counterparts showed anhydrous structures. Probably, the two hydroxyl residues and the ester functional group were able to establish interactions similar to those arranged by the water molecules, in the monohydroxylates, to stabilize the crystal structure.

### *2.5. Thermal Analysis*

The crystal structure of coumarin derivatives was also studied by DSC. As shown in Figure 11A, initially, in the heating scan, DSC measurements for monohydroxy coumarins (HMC and HEOMC) showed a broad endothermic peak that was possibly due to the loss of the recrystallization solvents (ethyl acetate, ethanol) and hydration water that occurs at 25–80 ◦C and 60–110 ◦C, respectively. Nevertheless, the dihydroxy-coumarins (DHMC and DHEOMC) suffered this peak. This finding is in line with the crystallographic data discussed above.

**Figure 11.** Differential scanning calorimetry (DSC) measurements, (**A**) heating scan, (**B**) cooling scan.

Additionally, significant differences in the melting temperatures of each compound were found in the heating scan. Coumarins bearing one hydration molecule in their asymmetric unit were found to have higher melting points than their dihydroxy partners, being 188 ◦C for HMC and 150 ◦C for HEOMC. Indeed, anhydrous crystal structures (DHMC and DHEOMC) presented a similar melting point between them, but much lower than the other hydroxylated coumarin family (121 for DHMC and 123 ◦C for DHEOMC). Comparing these two dihydroxy coumarins, DHEOMC presented a slightly higher melting temperature, which could be due to the flexibility of the spacer between the hydroxyl groups and umbelliferone ring favoring intermolecular interactions slightly higher.

∆ ∆ As shown in Figure 11B, only monohydroxylated coumarins displayed an efficient crystal packing, and because of these compounds exhibited a crystallization peak during the cooling process. Moreover, this crystallization took place at lower temperatures than melting points (151 vs. 188 ◦C for HMC and 74 vs. 150 ◦C for HEOMC). This fact could be due to the kinetics of crystal nucleation and growth being very slow.

On the other hand, the enthalpies found for the melting and recrystallization processes are collected in Table S4. It can be observed that the HMC and HEOMC coumarins partially recrystallized in the cooling process as a result of the enthalpy values for this step (∆Hc) being lower than the melting process (∆Hm). Additionally, the recrystallized fraction of HMC was higher than HEOMC. This fact could be related, again, to the more efficient crystal packing shown by this monohydroxylated coumarin compared to its monohydroxy partner.

Apart from that, the study of the thermal stability of crystal structures was also completed by TGA. In Figure 12, all the TGA curves of the studied coumarins are shown. For monohydroxy coumarins (HMC and HEOMC) the thermal decomposition occurred in two main mass-loss steps, whilst for dihydroxy coumarins (DHMC and DHEOMC) the decomposition carried out in a unique mass-loss step. In HMC and HEOMC coumarins, the first decomposition step appeared around 60–100 ◦C, corresponding to the loss of recrystallization solvents and hydration water. This result concurs with the X-ray diffraction data and DSC measurements described above.

**Figure 12.** TGA curves for all the coumarin derivatives.

As shown in Table 3, an experimental mass loss of 2.7 and 7.5 wt % were observed for HMC and HEOMC, respectively. In the case of HEOMC, this TGA analysis agreed with the theoretically calculated value of 7.6 wt % for the loss of one water molecule within its crystal structure. Nevertheless, the theoretically calculated value for the loss of one water molecule in HMC was 9.3 wt %, which did not agree with the experimental value (2.7 wt %). This fact may be due to a weaker interaction between HMC and water that provoked a water loss at room temperature. To corroborate this issue, in a detailed analysis of the DSC curves (Figure 11), an endothermic peak was observed at room temperature for HMC, whilst for HEOMC at up 60 ◦C.



After this first mass-loss, the stability of all coumarins remained unaltered until 250 ◦C. Hence, this high thermal stability results in suitable unalterability for bulk polymerization reactive extrusion at temperatures above 150 ◦C, which is an efficient and very common industrial method for manufacturing thermoplastic polyurethane with incorporated target photoreactivity [43]. At 250 ◦C, a second mass-loss was observed, related to the thermal degradation of the coumarin ring. Depending on the system, the temperature of this second process appeared to vary. In fact, onset degradation temperatures (T0) and the weight loss corresponding to each degradation step are collected in Table 3 for all molecules. It should be noted that the two dihydroxylated coumarins had the highest and lowest decomposition temperatures for the second stage. In this sense, the DHMC product showed the lowest decomposition temperature (246.0 ◦C), probably due to the aromatic nature of its ester group. Through a transesterification reaction, this aromatic ester is very susceptible to attack by any hydroxyl group. This fact has already been described previously. The thermal stability of DHEOMC increased up to 323 ◦C, mainly due to the aliphatic nature of its ester group.

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

#### *3.1. Characterization*

Solution 1H NMR spectra were recorded at room temperature in a Varian Unity Plus 400 instrument using deuterated chloroform (CDCl3) or deuterated dimethylsulfoxide (DMSO-d<sup>6</sup> ) as solvent. Spectra were referenced to the residual solvent protons at 7.26 or 2.50 ppm, respectively.

To understand the habit of the reversible photoreaction of the coumarin derivatives, irradiations were carried out in an ultraviolet crosslinker supplied by Ultra-Violet Products equipped with two sets of 5 lamps for 354 or 254 nm irradiations (354 and 254 nm are the wavelengths on the maxima of the irradiation spectra of the lamps, respectively). The samples (from 0.2 to 0.4 mM concentration) were prepared by dissolving each coumarin derivative in distilled water. Firstly, each solution was exposed to 350 nm light in order to produce a crosslinked version via photo-dimerization. Afterwards, all the lamps were changed to give out 254 nm light and photo-cleavage of the crosslinked solution took place. In order to characterize these reversible UV kinetics, UV experiments were performed in a Perkin Elmer Lambda 35 UV-Vis spectrometer. Absorbance of the coumarin solutions were measured from 385 to 210 nm.

Thermal properties of all samples (5–10 mg) were measured (in duplicate) by differential scanning calorimetry (DSC, 822e from Mettler Toledo) in aluminum pans under constant nitrogen flow (20 mL·min−<sup>1</sup> ). Each sample was subjected to a heating/cooling cycle from 25 to 200 ◦C with a heating rate of 2 ◦C·min−<sup>1</sup> . Thermogravimetric analysis measurements were also performed twice using TGA/DSC1 from Metter Toledo, under nitrogen (50 mL min−<sup>1</sup> ) from room temperature to 600 ◦C with a heating rate of 10 ◦C min−<sup>1</sup> , where approximately 20 mg of sample was required.

#### *3.2. Single-Crystal X-Ray Di*ff*raction*

Intensity data were collected on an Agilent Technologies SuperNova diffractometer equipped with monochromated Mo Kα radiation (λ = 0.71073 Å) and an Eos Charge-couple device CCD detector in the case of HMC and HEOMC, and monochromated Cu Kα radiation (λ = 1.54184 Å) and an Atlas CCD detector for DHMC and DHEOMC. The collection was performed at 100(2) K for HMC and HEOMC and 150(2) K for DHMC and DHEOMC. Data frames were processed (unit cell determination, mathematical absorption correction, intensity data integration and correction for Lorentz and polarization effects) using the CrysAlis Pro software package [44]. The structures were solved using the OLEX2 program [45] and refined by full-matrix least-squares using SHELXL-2014/6 [46]. Final geometrical calculations were carried out with PLATON [47] as integrated in WinGX software package [48].

#### *3.3. Computational Methods*

Quantum chemical calculations were performed with Gaussian 09 software [49]. Initially, the ground states were obtained by geometrical optimizations with the hybrid functional B3LYP and 6-311+(2d,p) as the basis set chosen for all the atoms.

As the experimental UV-Vis spectra of coumarins have been obtained in aqueous solution, the solvent effects of water were considered with the conductor-like polarizable continuum model (PCM). The calculated absorption energy was defined as the energy difference between the ground state and the excited state at the optimized ground state geometry.

#### **4. Conclusions**

Coumarin derivatives are widely distributed within the plant families and/or synthetic analogs for different applications. In this work, two types of hydroxy-derivative coumarins have been extensively analyzed by different techniques. Firstly, it is important to note that the synthetic routes presented were simple, easily scalable and with high yields, which could arouse high interest in the industry.

DHMC coumarin showed an uneven behavior compared to its counterparts, showing two strong absorption bands by UV-Vis. This performance conditioned its photophysical behavior to the dimerization reaction at 365 nm. Through a detailed atomistic study, these two absorption bands were attributed to the transitions between the frontier orbitals (HOMO, LUMO) and the two closest (HOMO-1, LUMO+1). Nevertheless, for the rest of the coumarins, the main excitation strongly corresponds to a π-π transition between the frontier orbitals.

Single-crystal X-ray diffraction analysis also demonstrated how hydroxyl groups allowed weak supramolecular forces to be established within the crystal structure, which were key elements in describing the ability to experience edge-to-edge self-association. Additionally, despite the symmetry of the space group of each coumarin, monohydroxy-derivated coumarins (HMC and HEOMC) presented one water molecule within each of their crystal structures, while their dihydroxylated counterparts (DHMC and DHEOMC) showed anhydrous structures.

Moreover, significant differences in the melting temperatures of each compound were found in the heating scan, which were consistent with an efficient crystal packing described by X-ray analysis.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/1420-3049/25/15/3497/s1, Figure S1: A) 1H-NMR and B) 13C-NMR spectra of HMC, Figure S2: A) ATR-FTIR and B) Mass spectra of HMC, Figure S3: A) 1H-NMR and B) 13C-NMR spectra of HEOMC, Figure S4: A) ATR-FTIR and B) Mass spectra of HEOMC, Figure S5: A) 1H-NMR and B) 13C-NMR spectra of DHMC, Figure S6: A) ATR-FTIR and B) Mass spectra of DHMC, Figure S7: A) 1H-NMR and B) 13C-NMR spectra of DHEOMC, Figure S8: A) ATR-FTIR and B) Mass spectra of DHEOMC. Figure S9: Electrostatic potential surfaces for the coumarins studied. Contour values range from - 0.080 to 0.080 Hartree/e, Figure S10: ORTEP view of the asymmetric units in (A) HEOMC, (B) DHMC and (C) DHEOMC depicted at the 50% probability level, together with atom labelling. Colour code: C, black, O, red, H, white. Table S1: Experimental and theoretical vertical excitation energies and oscillator strength for the first two excited states. The labels 1, 2, 1' and 2' correspond to HOMO, LUMO, HOMO-1 and LUMO+1, respectively, Table S2: Geometrical parameters (Å, ◦ ) of intermolecular π-π interactions in HEOMC, DHMC and DHEOMC, Table S3: Geometrical parameters for O-H··O hydrogen bonds and C-H··O-type contacts in HEOMC, DHMC and DHEOMC, Table S4: DSC results. Scheme S1: Synthetic route for the preparation of HMC, Scheme S2: Synthetic route for the preparation of HEOMC, Scheme S3: Synthetic route for the preparation of DHMC, Scheme S4: Synthetic route for the preparation of DHEOMC.

**Author Contributions:** Methodology, R.S.-R. and R.N.; validation, J.M.C., B.A., J.M.G.-Z., J.L.V.-V., and Á.M.-F.; formal analysis, R.S.-R., E.R.-B., R.N. and J.M.L.; investigation, R.S.-R., E.R.-B., R.N. and J.M.L; resources, R.S.-R., J.M.C., R.N. and Á.M.-F.; writing—original draft preparation R.S.-R., E.R.-B., R.N. and J.M.L.; writing—review and editing, R.S.-R., E.R.-B., R.N. and J.M.L.; supervision, J.M.C., B.A., J.M.G.-Z., J.L.V.-V., and Á.M.-F. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Basque Government within the framework ELKARTEK through the research project KK-2018/00108 and KK-2019/00077. This work was also funded by the Ministry of Economy and Competitiveness—Spain (MINECO) through the research Projects RTC-2016-4887-4 and RTI2018-096636-J-100 within the framework of the National Programme for Research Aimed at the Challenges of Society.

**Acknowledgments:** The authors would like to thank the Basque Government for the financial support of this work within the framework ELKARTEK through the research project KK-2018/00108 and KK-2019/00077. Additionally, this work has been supported by the research Projects RTC-2016-4887-4 and RTI2018-096636-J-100 of the Ministry of Economy and Competitiveness—Spain (MINECO) within the framework of the National Programme for Research Aimed at the Challenges of Society. Authors thank Dr Leire San Felices (Molecules and Material unit of the General X-ray Service, SGIker, UPV/EHU) for technical and human support.

**Conflicts of Interest:** These authors have declared no conflict of interest.

#### **References**


© 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/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Molecules* Editorial Office E-mail: molecules@mdpi.com www.mdpi.com/journal/molecules

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18

www.mdpi.com ISBN 978-3-0365-2774-1