(+)-epi-Epoformin, a Phytotoxic Fungal Cyclohexenepoxide: Structure Activity Relationships

(+)-epi-Epoformin (1), is a fungal cyclohexene epoxide isolated together with diplopimarane and sphaeropsidins A and C, a nor-ent-pimarane and two pimaranes, from the culture filtrates of Diplodia quercivora, a fungal pathogen for cork oak in Sardinia, Italy. Compound 1 possesses a plethora of biological activities, including antifungal, zootoxic and phytotoxic activity. The last activity and the peculiar structural feature of 1 suggested to carry out a structure activity relationship study, preparing eight key hemisynthetic derivatives and the phytotoxicity was assayed. The complete spectroscopic characterization and the activity in the etiolated wheat coleoptile bioassay of all the compounds is reported. Most of the compounds inhibited growth and some of them had comparable or higher activity than the natural product and the reference herbicide Logran. As regards the structure-activity relationship, the carbonyl proved to be essential for their activity of 1, as well as the conjugated double bond, while the epoxide could be altered with no significant loss.


Introduction
(+)-epi-Epoformin (1, Figure 1) is a cyclohexene epoxide, isolated together with some nor-ent-pimaranes and pimarane diterpenes, namely diplopimarane and sphaeropsidins A and C, from culture filtrates of the fungus Diplodia quercivora, isolated from declining Quercus canariensis trees in Tunisia as a canker-causing agent [1]. The fungi belonging to Diplodia genus are well known as pathogens of forest plants and as producers of several bioactive metabolites belonging to different classes of natural compounds [2].
Compound 1 was also previously reported from an unidentified fungus isolated from a diseased leaf of crepe myrtle (Lagerstroemia indica L. Pers.) [3] and showed inhibition of lettuce seed germination. Its total enantioselective and asymmetric synthesis [4][5][6][7] was also developed and the assignment of its absolute configuration was determined by time-dependent density together with that of other close naturally occurring cyclohexene oxides [8].
When isolated from D. quercivora, 1 was tested for its phytotoxic activity using different bioassays. In leaf-puncture tests on holm oak (Quercus ilex L.), cork oak (Quercus suber L.), and tomato (Lycopersicum esculentum Mill.) leaves, 1 caused necrotic lesions on holm and cork oak (area lesion sizes of 6.7 and 9.5 mm 2 , respectively), while inducing irregular necrotic lesions (13.5 mm 2 area) on tomato. In a tomato cutting bioassay, 1 caused, to different extents, stewing on stem at 0.2 and 0.1 mg/mL. It also showed activities against phytopathogenic fungi and oomycetes Phytophthora cinnamomi and Phytophthora plurivora were the most sensitive target organisms followed by Athelia rolfsii, whereas Diplodia corticola was the most resistant as its mycelial growth was only slightly inhibited (39.7% inhibition) [1]. Finally, when assayed on brine shrimps (Artemia salina L.) 1 caused more than 50% larval mortality at 200 µg/ mL after 36 h to metabolites exposure [1].
More recently, 1 was also used in a preliminary screening to evaluate the potential of plant and fungal metabolites, belonging to different classes of natural compounds, for the biological control of several rusts species belonging to the genera Puccinia and Uromyces, which are pathogens of very important crops including legumes. Compound 1 showed significant inhibition of all the species tested also exhibiting an effective penetration [9]. The potential fungicide activity of 1 also in plants against U. pisi and P. triticina was successively confirmed [10].
In this manuscript, 1 was isolated and tested with eight hemisynthetic derivatives (2)(3)(4)(5)(6)(7)(8)(9) in order to obtain clues about the structure-activity (SAR) requirements of these compounds, where we acetylated and oxidized the hydroxy group, open the epoxide, reduce the carbonyl group or saturated the double bond.
Molecules 2018, 23, x FOR PEER REVIEW 2 of 12 sizes of 6.7 and 9.5 mm 2 , respectively), while inducing irregular necrotic lesions (13.5 mm 2 area) on tomato. In a tomato cutting bioassay, 1 caused, to different extents, stewing on stem at 0.2 and 0.1 mg/mL. It also showed activities against phytopathogenic fungi and oomycetes Phytophthora cinnamomi and Phytophthora plurivora were the most sensitive target organisms followed by Athelia rolfsii, whereas Diplodia corticola was the most resistant as its mycelial growth was only slightly inhibited (39.7% inhibition) [1]. Finally, when assayed on brine shrimps (Artemia salina L.) 1 caused more than 50% larval mortality at 200 μg/ mL after 36 h to metabolites exposure [1]. More recently, 1 was also used in a preliminary screening to evaluate the potential of plant and fungal metabolites, belonging to different classes of natural compounds, for the biological control of several rusts species belonging to the genera Puccinia and Uromyces, which are pathogens of very important crops including legumes. Compound 1 showed significant inhibition of all the species tested also exhibiting an effective penetration [9]. The potential fungicide activity of 1 also in plants against U. pisi and P. triticina was successively confirmed [10].
In this manuscript, 1 was isolated and tested with eight hemisynthetic derivatives (2)(3)(4)(5)(6)(7)(8)(9) in order to obtain clues about the structure-activity (SAR) requirements of these compounds, where we acetylated and oxidized the hydroxy group, open the epoxide, reduce the carbonyl group or saturated the double bond.

Results and Discussion
The structure-activity relationship study of (+)-epi-epoformin (1) was carried out preparing eight of its hemisynthetic derivatives. Although 1 was previously described in the literature as a fungal phytotoxic compound [11], there is no information available regarding its activity in etiolated wheat coleoptiles. Thus, the phytotoxic activity of the parent compound (1) and its eight hemisynthetic derivatives was assayed using this test.

Results and Discussion
The structure-activity relationship study of (+)-epi-epoformin (1) was carried out preparing eight of its hemisynthetic derivatives. Although 1 was previously described in the literature as a fungal phytotoxic compound [11], there is no information available regarding its activity in etiolated wheat coleoptiles. Thus, the phytotoxic activity of the parent compound (1) and its eight hemisynthetic derivatives was assayed using this test.
The culture filtrates of D. quercivora were extracted and purified as reported in detail in the Section 3.2. Compound 1 was obtained as a white solid and identified comparing its physical ([α] 20 D ) and spectroscopic ( 1 H and 13 C-NMR and ESI-MS) properties with those previously reported [1,3].
The acetylation of 1 to obtain 2, in a moderate yield (67%), was successfully carried out using acetic anhydride [12]. Its NMR spectra (Section 3.4) showed that an acetyl group was incorporated to the compound by the new singlet at δ 2.13 (H-9) in the 1 H-NMR and the corresponding carbons of the acetyl group in the 13 C-NMR [13], with a carbonyl at δ 193.5 (C-8) and the carbon of the methyl at δ 20.7. A significant downfield shift (from δ 4.67 to δ 5.72) was also observed for H-4 when the 1 H-NMR spectrum of 1 was compared to that of 2 [14]. In addition, the OH broad band of 1 disappeared in the FTIR spectrum. 2D NMR experiments [15] allowed confirming the structure of 2. HSQC spectrum was employed to assign each carbon to their corresponding hydrogens and the HMBC showed the correlations to elucidate the position of the acetyl group: the signal at δ 2.13 (H-9) correlated with that at δ 64.5 (C-4), while the signal at δ 5.72 (H-4) correlated with that at δ 169.8 (C-8).
On the other hand, the oxidation of the hydroxyl group with Dess-Martin periodinane [16] was repeated several times changing the reaction times. The highest yield was 44% for 3, which was obtained after 16 hours, recovering a 51% of the starting material (1). A longer reaction time did not significantly increase the yield and less starting material would be recovered. The success on synthesizing 3 was confirmed firstly by the absence of any hydroxyl group and the presence of two narrow bands at 1737 and 1715 cm −1 corresponding to the two conjugated carbonyls in the FTIR spectrum [17]. The NMR experiments showed two signals of conjugated carbonyl carbons in the 13 C-NMR at δ 192.3 (C-1) and 191.2 (C-4) and the signal of the H-4 of 1 disappeared in the 1 H-NMR.
In the HMBC the correlation of the signal at δ 192.3 (C-1) with the methyl signal at δ 2.02 (H-7) was observed allowing to assign that signal to the C-1. Due to the proximity of the signals at δ 54.1 (C-5) and δ 53.7 (C-6) the HSQC experiment was indispensable to assign them to their corresponding signals in the 1 H-NMR at δ 3.79 (H-5) and δ 3.84 (H-6). These signals were assigned to their positions in the compounds thanks to the correlations in the COSY experiment, where the signal at δ 3.79 (H-5) correlated with the signal at δ 6.43 (H-3).
Following the procedure described by Nicolaou et al. [18] we attempted the opening of the epoxide to obtain a dihydroxylated compound in C-4 and C-5. Unexpectedly, the integrals of the 1 H-NMR of the product showed that there was only one hydrogen at C-6 at δ 4.77 (H-6), too high for hydrogens in α position to the carbonyl at C-1. After studying all the NMR spectra we determined that the structure was that of 4, obtained with a high yield (83%), but the compound did not ionize properly to observe the mass corresponding to the product. The presence of iodine in the compound was confirmed by the presence of an I − ion in the MS (ES − ), with m/z 126.9045, acting as a leaving group in an intramolecular nucleophilic substitution, which leads to the corresponding neutral epoxide. The stereochemistry of the hydroxyl at C-5 was confirmed by the coupling constants value of H-5 at δ 2.80 with δ 4.77 (H-6) and δ 4.21 (H-4) being J 5,6 = 4.1 Hz (equatorial-equatorial) and J 4,5 = 7.7 Hz (axial-axial).
In order to study the influence of the double bond at C-2 in the bioactivity of 1, this compound was treated with carbon supported palladium. After several tries, only traces of the expected compound with the epoxide at C-5 and the reduced double bond at C-2 were found. Instead, the major compounds were, unexpectedly, 5 and 6, with different stereochemistry at C-5 and derived from the reductive opening of the epoxide that occurred at the same time that the reduction of the double bond. By using carbon supported platinum 5 and 6 were again the major products. Interestingly, by using Pd/C the yields were slightly higher for obtaining compounds 5 and 6 (29 and 33%, respectively) than by using Pt/C (19% and 21%, respectively). Only by direct infusion and solving the compounds in MeOH (0.1% formic acid) was possible to ionize these compounds, obtaining the [M + Na] + ions (m/z 167.3 for C 7 H 12 O 3 Na). The lack of conjugation with a double bond in 5 and 6 moved the carbonyl bands to higher wavenumbers in the FTIR spectrum, from the starting 1674 cm −1 in 1 to 1704 and 1711 cm −1 in 5 and 6, respectively. The 1 H-NMR spectra of both compounds showed that the typical signal of H 2 -3 moved to higher field, being located at δ 1.99 in 5 and δ 2.22 and 1.35 in 6. Although the stereochemistry of the hydroxyl at H-5 could be assigned by using the values of coupling constants of H-5 with H-4 in both compounds (J 4,5 = 4.3 Hz in 5 and J 4,5 = 8.8 Hz in 6 indicating equatorial-axial and axial-axial couplings, respectively), NOE-and ROE-difference experiments were needed to assign the relative position of the methyl at C-2. By irradiating H-2, H-4 and H-5 in 5 with a ROE-1D experiment, it was confirmed that H-2 (δ 2.78) was close in space to H-4 (δ 4.02) and H-5 (δ 4.18). In addition, H-4 was close in the space to H-6a (δ 2.43). On the other hand, H-4, H-5 and H-3b were irradiated in 6 with a NOE-1D experiment. In this case, it was observed that the signal for H-3b (δ 1.35) was close in space with H-5 (δ 3.65). The latter was close to H-6b (δ 2.75) and H-4 (δ 3.88) was close in space to H-2 (δ 2.46) and H-3a (δ 2.22), thus confirming the opposite stereochemistry of H-5 in 6. There was one signal for each hydrogen at C-6 in the 1 H-NMR of both compounds (δ 2.91 and δ 2.43 for 5, and δ 2.75 and δ 2.46 for compound 6). In the case of 6, this was also observed for the hydrogens at C-3 (δ 2.22 and δ 1.35). The COSY and the HSQC experiments were used to assign each pair of signals to their corresponding carbons, while the HMBC allowed differentiating the close carbons at δ 74.0 (C-5, correlated with both H-6) and δ 73.8 (C-4 correlated with both H-3) in 6.
The opening of the epoxide was performed using basic and acidic media (0.1 M NaOH, 0.1 M HCl and 0.05 M H 2 SO 4 ) and none of them worked properly. Surprisingly, while using acyl chlorides with DMAP to acetylate the hydroxyl at C-4 using the Yamaguchi esterification [19], the epoxide opened spontaneously, yielding both compounds 7 and 8 with high yields (39% and 45%, respectively) and no amount of the desired esters. At first, we checked the literature to compare the spectra of our trihydroxylated compounds with the already reported gabosines A and N [20], but the signals were different. Therefore, the spectra were analyzed to determine their structures. Firstly, it was observed the opening of the epoxide in the shift of the signals to lower fields, from δ 3.78 (H-5) and δ 3.53 (H-6) to δ 4.05 and 4.57 (7) or δ 3.84 and 4.38 (8). H-5 and H-6 were assigned by observing the COSY, where H-5 and H-6 were correlated among themselves, and the HMBC where H-5 correlated with δ 69.3 (C-4), and H-6 correlated with δ 190.6 (C-1) and δ 74.6 (C-5) in the spectrum of 7. The same correlations were found in 8, H-5 and H-6 were correlated in the COSY, and in the HMBC H-5 correlated with δ 71.4 (C-4), and H-6 correlated with δ 189.7 (C-1) and δ 78.3 (C-5). Regarding the stereochemistry, the analysis of the NOE-difference experiments and the coupling constants [21] indicated that H-5 and H-6 of 7 were close in the space, H-5 had a high coupling constant with H-4 (J = 7.1 Hz) and a lower constant with H-6 (J = 3.5 Hz), indicating the 4S,5R,6R stereochemistry. Regarding 8, the coupling constants of H-5 were with H-4 (J = 8.2 Hz) and H-6 (J = 11.3 Hz), both high, indicating axial-axial coupling in both cases and the 4S,5R,6S stereochemistry.
Finally, 1 was treated with NaBH 4 and CeCl 3 in MeOH to reduce the carbonyl without affecting the double bond or the epoxide, obtaining 9 in a moderate 39% yield. The FTIR showed clearly the presence of an OH broad band at 3234 cm −1 and the missing narrow band of the carbonyl. Regarding the 1 H-NMR, all the signals shifted to a higher field. The lack of the carbonyl was clearly observed in the 13

SAR Study
The results from the etiolated wheat coleoptile bioassay of compounds 1-9 are shown in Figure 2. In the case of 1 and 2, a stimulatory activity was observed at 30 and 10 µM. It has been reported that allelopathic compounds can stimulate growth at lower doses and inhibit growth at higher doses [22,23]. According to their IC 50 values (Table 1), the most active compounds were 3 and then 4. Compound 3 was one order of magnitude more active than 4 and comparable to the synthetic herbicide Logran ® , however, the activity profile was inconsistent with the increasing concentration. The IC 50 for 3 was found at the lowest concentration tested (10 µM) and this value did not change significantly at higher concentrations. The test solutions of 3 were clear and no precipitate was observed; however, this profile is typically linked with a problem of solubility at the highest concentration. It is unknown if this or the nature of the compound caused the strange profile. that allelopathic compounds can stimulate growth at lower doses and inhibit growth at higher doses [22,23]. According to their IC50 values (Table 1), the most active compounds were 3 and then 4. Compound 3 was one order of magnitude more active than 4 and comparable to the synthetic herbicide Logran ® , however, the activity profile was inconsistent with the increasing concentration. The IC50 for 3 was found at the lowest concentration tested (10 μM) and this value did not change significantly at higher concentrations. The test solutions of 3 were clear and no precipitate was observed; however, this profile is typically linked with a problem of solubility at the highest concentration. It is unknown if this or the nature of the compound caused the strange profile.  Compound 4, containing iodine in the structure was the most active and three times more active than the next compounds 2 and 7. The halogen in this structure seems to have a key role in these results, since the other dihydroxylated compound with the same stereochemistry (6) and another one with different stereochemistry (5) were much less active or not active at all. However, the lack of a  Compound 4, containing iodine in the structure was the most active and three times more active than the next compounds 2 and 7. The halogen in this structure seems to have a key role in these results, since the other dihydroxylated compound with the same stereochemistry (6) and another one with different stereochemistry (5) were much less active or not active at all. However, the lack of a double bond removing the α-β-unsaturated system could also play a role. Since the reaction for removing the iodine did not work, we could not explore further this hypothesis.
AlogP values varied between −0.88 and 1. Only compounds 2 and 4 followed the Lipinski's rule of five [24,25], having the appropriate molecular weight and AlogP. As predicted by Lipinski, these two compounds were part of the most active compounds of the collection. However, 7, violating both requirements, also exhibited a remarkable activity. The stereochemistry seems to play an important role in this compound, since its epimer 8 had almost half the activity.
The least active compounds were 5, 6 and 9, showcasing that the double bond and the carbonyl are both indispensable for the observed activity. The epoxide, on the other hand, proved to be superfluous, since the opening of the epoxide did not result in a drop in activity (compounds 4, 7 and 8) and the compound preserving the epoxide and lacking the carbonyl had nil activity (9).
In summary, the addition of iodine in the molecule (compound 4) and an extra hydroxyl with the appropriate stereochemistry (compound 7) had a positive effect in the bioactivity, while the acetylation (compound 2) of the hydroxyl naturally present had little effect and the reduction of the carbonyl (compound 9) or the double bond (compounds 5 and 6) had a negative effect.
A possible explanation of the higher activity of derivatives 3 and 4 was that 3 has a hemiquinone structure and derivative 4 by elimination of HI via E2 could be converted into the corresponding hemiquinone derivative and thus in the corresponding quinone. This could also occur in derivative 7 but the leaving group from C-6 is OH − with respect to the good one I − in 4. The hemiquinone generated from 7, being also an enolic compound could be stabilized from the hydrogen bond between OH-6 and O=C-1. Because of the low oxidation level of derivatives 2, 5, 6, 8 and 9, as well as 1, they would be more difficult to convert in the corresponding hemiquinone.
This hypothesis on the mode of action of epi-epoformin is in full agreement with the SAR study carried out using some derivatives of sphaeropsidones, two epimeric phytotoxic cyclohexene epoxides isolated from Diplodia cupressi (a cypress pathogen in the Mediterranean basin) [26] and the testing their phytotoxic and antifungal activities on non-host plants and on five fungal pathogenic species belonging to the genus Phytophthora [27]. Successively, when sphaeropsidones and the same and new other derivatives were tested by haustorium-inducing activity in Orobanche cumana, O. crenata and Striga hermonthica, the results obtained suggested that the activity is due to the possibility to convert the natural sphaeropsidones and natural and hemisynthetic derivatives into the corresponding 3-methoxyquinone [28].
This hypothesis on the mode of action of sphaeropsidone, that as above reported could work also for epi-epoformin, is in full agreement with the results obtained using natural and synthetic quinones as sorghum xenognosin and dimethoxybenzoquinones, in studies carried out on haustoria and the chemistry in host recognition parasitic angiosperms. Quinone/hydroquinone structures serve as cofactors in many metabolic pathways, playing critical chemical roles in oxidation/reduction processes [29,30].

General Experimental Procedures
The purities of the compounds were determined by 1 H-NMR spectroscopy and every compound was purified in the HPLC prior bioassays. 1 H-NMR and 13 C-NMR spectra were recorded on Agilent™ (Santa Clara, CA, USA) 400 and 500 MHz spectrometers using CDCl 3

Fungal Strain
The fungal strain of D. quercivora used in this study was originally isolated from a symptomatic branch of Q. canariensis Willd. collected in a natural area in Tunisia as previously described [1].

Etiolated Wheat Coleoptile Bioassay
The compounds 1-9 were tested for their bioactivity in an etiolated wheat coleoptile bioassay. The conditions for this bioassay were reported previously [31,32] and replicated in this study, using the same herbicide (Logran ® ) as positive control and the buffer solution as negative control. Wheat (Triticum aestivum) was the 'catervo' variety. All the samples were solved in a 0.5% of dimethyl sulfoxide and gave clear solutions at all the tested concentrations (10 −3 -10 −5 M). The results are shown in Figure 2.

Etiolated Wheat Coleoptile Bioassay
The compounds 1-9 were tested for their bioactivity in an etiolated wheat coleoptile bioassay. The conditions for this bioassay were reported previously [31,32] and replicated in this study, using the same herbicide (Logran ® ) as positive control and the buffer solution as negative control. Wheat (Triticum aestivum) was the 'catervo' variety. All the samples were solved in a 0.5% of dimethyl sulfoxide and gave clear solutions at all the tested concentrations (10 −3 -10 −5 M). The results are shown in Figure 2.

Calculation of IC 50 and logP
The bioactivity data were fitted to a sigmoidal dose-response model using the GraphPad Prism v.5.00 software package (GraphPad Software, San Diego, CA, USA) [33] to obtain the IC 50 values shown in Table 1. The AlogP was calculated using the ALOGPS v.

Conclusions
(+)-epi-Epoformin (1) was successfully isolated from extracts of D. quercivora following the procedures in the literature. Then, eight derivatives of 1 were synthesized and tested altogether for their bioactivity in the etiolated wheat coleoptile bioassay. The derivatives were compared with the starting material to look for structure-activity relationships. The α-β-unsaturated system in 1 proved to be essential to obtain active derivatives and the inclusion of iodine and an extra hydroxyl in the molecules increased the activity, while the acetylation of the hydroxyl did not have a significant effect. The results described here illustrate that natural products, especially those provided by fungi, are valuable resources which need to be investigated further as natural herbicides against weeds.