Embellicines C-E: Macrocyclic Alkaloids with a Cyclopenta[b]fluorene Ring System from the Fungus Sarocladium sp.

Macrocyclic alkaloids with a cyclopenta[b]fluorene ring system are a relatively young structural class of fungal metabolites, with the first members reported in 2013. Bioassay-guided fractionation of a Sarocladium sp. (fungal strain MSX6737) led to a series of both known and new members of this structural class (1–5), including the known embellicine A (1), three new embellicine analogues (2, 4, and 5), and a semisynthetic acetylated analogue (3). The structures were identified by examining both high-resolution electrospray ionization mass spectrometry data and one-dimensional and two-dimensional NMR spectra. The relative configurations of these molecules were established via 1H–1H coupling constants and nuclear Overhauser effect spectroscopy, while comparisons of the experimental electronic circular dichroism (ECD) spectra with the time-dependent density functional theory ECD calculations were utilized to assign their absolute configurations, which were in good agreement with the literature. These alkaloids (1–5) showed cytotoxic activity against a human breast cancer cell line (MDA-MB-231) that ranged from 0.4 to 4.8 μM. Compounds 1 and 5 were also cytotoxic against human ovarian (OVCAR3) and melanoma (MDA-MB-435) cancer cell lines.

C ancer is still the second leading cause of death worldwide despite recent improvements in its diagnosis and treatment. 1,2 The prevalence of drug-resistant malignancies requires continual efforts to discover and develop new anticancer leads. 3−5 Filamentous fungi are a promising source for potential anticancer leads, 6−8 as they have yielded compounds with novel structures, new mechanisms of actions, and/or selective biological activities. 6,9,10 Alkaloids, in particular, are a highly diverse class of natural products that exhibit a wide range of pharmacological properties. 11−13 The Dictionary of Natural Products reports the isolation of more than 30 000 alkaloids from various natural sources. 14 However, alkaloids with a cyclopenta[b]fluorene (6/5/6/5) ring system are relatively new to the literature. To date, only ten have been reported, and all of those have been isolated from endophytic fungi. 14 Embellicines A and B, reported in 2013, have a tetracyclic core fused with a 13-membered macrocycle, 15 and their cytotoxic properties were purported to be due to TNFα-induced NF-κB transcriptional activity. Phomapyrrolidones A−C were reported concurrently and found to exhibit weak antitubercular activity. 16 Recently, ascomylactams A−C 17 and didymellanosine 18 were characterized and reported to exhibit cytotoxic and antimicrobial activities. Didymellanosine is the first 13membered macrocyclic alkaloid with a cyclopenta[b]fluorene ring system conjugated with adenosine. 18 A structurally related group of compounds, hirsutellones, 19−21 pyrrocidines, 22−24 and pyrrospirones, 25,26 have a decahydrofluorene skeleton and gained considerable attention 21,27 for their unique core and pharmacological activities as antimicrobial and cytotoxic leads.
In the course of ongoing studies to discover fungal metabolites with potent cytotoxic activities, 28−32 we report the characterization of a series of embellicines (1−5), including the isolation of the known embellicine A (1), three new embellicines C, D, and E (2, 4, and 5, respectively), and the semisynthetic generation of the acetylated analogue, 2′-Oacetyl-embellicine C (3). These macrocyclic alkaloids with a cyclopenta[b]fluorene ring system were obtained from the organic extract of a Sarocladium sp. (fungal strain MSX6737), which exhibited cytotoxic activity against human melanoma, breast, and ovarian cancer cells when tested at a concentration of 20 μg/mL (i.e., less than 40% cancer cell survival relative to a negative control). 32 The structures were determined by examining their one-dimensional (1D) and two-dimensional (2D) NMR data. In addition, nuclear Overhauser effect spectroscopy (NOESY) spectra, along with the experimental and calculated ECD spectra, were used to assign their relative and absolute configurations.

■ RESULTS AND DISCUSSION
Fungal strain MSX6737 was cultured on a solid rice medium for 3 weeks, and extraction and bioactivity-directed fractionation afforded compounds 1, 2, 4, and 5. The taxonomy of this strain was examined using both morphological and molecular methods, indicating that it is likely a new Sarocladium sp., and although it was obtained as a saprobe from leaf litter, many other members of this genus are reported as plant pathogens, opportunistic human pathogens, endophytes, and mycoparasites. 33 As it may represent a new fungal species, 34 further taxonomic studies on strain MSX6737 are ongoing.
The molecular formula of compound 1 was deduced as C 34 H 41 NO 4 via high-resolution electrospray ionization mass spectrometry (HRESIMS), indicating an index of hydrogen deficiency of 15. 1 H and 13 C NMR data ( Figure S2 and Table  S1) indicated the presence of six methyl, two carbonyl, three methylene, six aromatic, and six olefinic carbons. Searching the Dictionary of Natural Products for compounds with a similar molecular formula, compound 1 was identified as embellicine A based on the matched NMR data (Table S1) and ECD spectrum ( Figure S27). Embellicine A was reported in 2013 by Ebrahim et al. as an alkaloid with a cyclopenta[b]fluorene (6/ 5/6/5) ring system fused with a 13-membered macrocycle mediated by the presence of a γ-lactam moiety and a parasubstituted benzene with restricted free rotation. 15 Embellicine A exhibits a double bond between C-1′ and C-18, as suggested by the 13 C signals at δ C 157.8 and 131.1 and the olefinic proton at δ H 7.14. The absolute configuration of embellicine A (1) was previously reported as 1R, 4S, 7R, 8S, 10R, 12S, 13R, 14R, 15S, 16S, 2′R as established via time-dependent density functional theory (TDDFT) ECD calculations. 15 Later, ascomylactam B was isolated from Didymella sp. and found to share the same planar structure of embellicine A (1); however, ascomylactam B differs from 1 in the configurations at C-1, C-4, C-7, C-14, and C-16. 17 Comparison of the ECD spectra and optical rotations between embellicine A and ascomylactam B were reported previously. 15,17 The molecular formula of compound 2, which was obtained as a white amorphous powder, was C 34 H 41 NO 5 , as deduced via HRESIMS, indicating an index of hydrogen deficiency of 15 ( Figure S1). The 1 H and 13 C NMR data (Tables 1 and 2) presupposed structural similarities with embellicine A (1). Correlation spectroscopy (COSY) and heteronuclear multiplebond coherence (HMBC) correlations of the four doublet of doublet protons resonating at δ H 6.96, 6.99, 7.16, and 7.24 suggested a para-substituted aromatic ring with restricted rotation (Figures S4−S6). The presence of a tetracyclic cyclopenta[b]fluorene (6/5/6/5) ring system was deduced by three proton spin systems (Figure 1). The first spin system extended from H-7 to H-16, the second was observed between H 3 -20 and H-1, and the third consisted of the allylic coupling between H 3 -21 and the olefinic proton at C-3. The six methyl groups attached to the tetracyclic ring system were positioned based on their COSY and HMBC correlations (Figure 1). Carbons resonating at δ C 63.3 for C-1′, 60.0 for C-18, 84.9 for C-2′, and 166.5 for the amide carbonyl (C-19) suggested the presence of a γ-lactam moiety, similar to that reported in ascomylactam C and phomapyrrolidone C. 16,17 This was confirmed by the HMBC correlations of H-1′ with C-2′ and C-19 ( Figure 1). The chemical shifts of the two adjacent carbons, C-1′ and C-18, were more shielded as compared to the hydroxylated carbon at C-2′, indicating they were part of an oxirane. 16,17 The HMBC correlations of H 2 -3′ with the aromatic carbons (C-4′, C-5′, and C-9′) and the hemiaminal carbon (C-2′) suggested that C-3′ intermediates the connection between the para-substituted benzene and the γlactam ( Figure 1). On the other hand, the HMBC correlation of H-14 with the oxygenated aromatic carbon (C-7′) confirmed the ether linkage between the para-substituted benzene and the tetracyclic ring system. The linkage between the γ-lactam moiety and the tetracyclic ring system through the C-17 keto group was established based on the HMBC correlations of H-15 and H-16 with C-17 ( Figure 1). Both phomapyrrolidone C and ascomylactam C were found to share the same planar structure as 2. 16,17 However, for both of these compounds neither the NMR data (Tables S2 and S3) nor the ECD spectra ( Figure S9) matched those of 2, suggesting a diastereomeric relationship between these three compounds. The absolute configuration of phomapyrrolidone C, deduced via analysis of NOESY spectra and ECD calculations, was recently reported as 1S, 4R, 7S, 8S, 10R, 12S, 13R, 14R, 15S, 16S, 18R, 1′R, 2′R and was shown to be the 16S-epimer of ascomylactam C. 17 The NOESY spectrum, along with the 1 H− 1 H coupling constants, were analyzed to assign the relative configuration of the asymmetric centers in 2 ( Figure 2). The large diaxial coupling constant of H-9b with both H-8 and H-10 ( 3 J H-9b, H-8 = 3 J H-9b, H-10 = 11.9 Hz) and the coupling of H-11b with H-10 and H-12 ( 3 J H-11b,H-10 = 3 J H-11b,H-12 = 12.0 Hz) indicated that H-8, H-10, and H-12 were cofacial and in axial orientations, while CH 3 -24 and CH 3 -25 were in equatorial orientations (Table 1) and syn to each other ( Figure S7). This was further confirmed by the NOESY cross-peak of H-10 with H-12 ( Figure 2). The relative configuration of the tetracyclic ring system was assigned based on the observed NOESY correlations of H-7/H 3 S8). The NOESY correlations between H-13/ H 3 -25/H-14 indicated their cofacial relationship, which was presumed to be opposite to those stated above. The configuration at C-1 was challenging to assign due to the close chemical shift values of the two methyls CH 3 -20 and CH 3 -22. Accordingly, NOESY correlation between H 3 -20 and H 3 -22 could not be recognized within the high noise level around the diagonal axis in the spectrum ( Figure S7). However, these could be cofacial based on biogenetic considerations, as all previously described fungal metabolites with a cyclopenta[b]fluorene (6/5/6/5) ring system have CH 3 -20 and CH 3 -22 oriented on the same face, including embellicine A (1). 16−18 The relative configuration of H-1′ was assigned based on its cross-peak with H-16 and H-9′. The restricted rotation of the benzene ring inside the 13-membered macrocycle was further supported by the NOESY correlation of H-15 with H-8′, while H 3 -25 and H-14 correlated with H-6′. The orientation of the γ-lactam was established based on the NOESY correlation of the NH proton with H-3a′ ( Figure 2). Therefore, the relative configuration of 2 was assigned as 1R*, 4S*, 7R*, 8S*, 10R*, 12S*, 13R*, 14R*, 15S*, 16S*, 18R*, 1′R*, 2′R*.
Acylation of 2 afforded the acetyl analogue at the 2′−OH position (3) with a molecular formula of C 36 H 43 NO 6 as deduced via HRESIMS; this analogue was generated to further explore the structure−activity relationship (SAR) of these compounds. The two extra carbons at δ C 170.5 and 21.5 and the singlet methyl at δ H 2.17 supported this assignment (Tables 1 and 2). The NOESY spectrum of 3 showed similar correlations to those observed for 2, which further supported the interpretations of the NOESY correlations observed in 2.
Their ECD spectra were also nearly identical ( Figure S20), and given the relationship to 2, this compound was ascribed the trivial name 2′-O-acetyl-embellicine C (3).
Compound 4 was found to have the same molecular formula as 2, as indicated by HRESIMS (C 34 H 41 NO 5 ), and 1 H and 13 C NMR data suggested the same tetracyclic system and aromatic ring as 2 (Tables 1 and 2). However, 4 showed an extra signal at δ C 201.3 and was missing a signal at δ C 84.9 (i.e., C-2′ in 2), which suggested opening of the γ-lactam and the formation of a ketone at C-2′. COSY correlations between the two exchangeable protons at δ H 5.40 and 5.68 supported the presence of an NH 2 group and the lack of a bond between the amide nitrogen and C-2′. 1 H− 1 H coupling constants and the NOESY spectrum of 4 suggested the same relative configuration as 2 (Figure 2). Major similarities in their experimental ECD spectra were also observed ( Figure S20), and partial conversion of 2 into 4 was observed upon storage as a dry compound at room temperature, and vice versa ( Figure S21). Thus, 4 shared the same absolute configuration as 2, which was 1R, 4S, 7R, 8S, 10R, 12S, 13R, 14R, 15S, 16S, 18R, 1′R, and it was ascribed the trivial name embellicine D. Interestingly, the opening of the γ-lactam was not reported previously for the two diastereoisomers of 2, namely, phomapyrrolidone C and ascomylactam C. 16,17 Compound 5 was isolated as a white amorphous powder with a molecular formula of C 35 H 45 NO 5 as determined via HRESIMS. Comparing the 1D and 2D NMR data of 5 with those observed for embellicine A (1) indicated the presence of an extra methoxy group attached to C-1′ (δ C 59.5), while lacking the double bond between C-1′ and C-18. The HMBC    15 which is structurally similar to 5. 15 However, the methoxy group attached to C-1′ in 5 is a hydroxy in embellicine B. The NOESY spectrum of 5 indicated the same relative configuration for the tetracyclic rings observed in 1−4 ( Figure 2). The orientation of H-1′ was established by its NOESY correlation with H-16 ( Figure 2). The enol functional group in both 5 and embellicine B limited their stability, as conversion to 1 was noted both in the literature 15 and in this study. Moreover, 5 is likely an artifact that results from exposing 1 to MeOH, which was used through the isolation and purification process. 38 Accordingly, the absolute configuration of 5 was suggested as 1R, 4S, 7R, 8S, 10R, 12S, 13R, 14R, 15S, 16S, 1′R, 2′R, and it was ascribed the trivial name embellicine E. The calculated and experimental ECD data for 5 were in good agreement, further supporting the absolute configuration of this compound ( Figure 3). Ascomylactam A is a diastereoisomer of 5, as they share the same planar structure; however, the absolute configurations at C-1, C-4, and C-7 were different between these compounds. 17 The purity of compounds 1−5 was >95%, as assessed by UPLC ( Figure S28), prior to evaluation of cytotoxic activities against MDA-MB-231 (human breast cancer), OVCAR3 (human ovarian cancer), and MDA-MB-435 (human melanoma cancer) cell lines. Their half-maximal inhibitory concentration (IC 50 ) values against breast cancer cells ranged between 0.4 and 4.8 μM (Table 3). Compounds 1 and 5 were also cytotoxic against both ovarian and melanoma cell lines (IC 50 1.0−1.8 μM), while compounds 2−4 showed less toxicity against these two cell lines (IC 50 > 10 μM) ( Table 3). Compounds 1 and 5 were more potent against the three cancer cell lines, and both lack the oxirane ring exhibited by 2−4. Cytotoxic activities against various cancer cell lines were reported previously for the structurally related analogues, ascomylactams A−C, phomapyrrolidone A, and phomapyrrolidone C. 17 The biosyntheses of embellicines, and similar analogues with a cyclopenta[b]fluorene (6/5/6/5) ring system, have been reported via polyketide synthase−nonribosomal peptide synthetase (PKS−NRPS) complexes. 21,27 Unlike hirsutellones and pyrrocidines, which are derived from the polycyclization of a Tyr-C 18 linear precursor, alkaloids with an additional fivemembered cycle fused with ring A are hypothesized to derive from a Tyr-C 20 intermediate ( Figure S29). 21,27,39 The five methyl moieties along the linear C 18 chain could then be introduced from S-adenosylmethionine. It is suggested that the 13-membered ring, along with ring C of the cyclopenta[b]fluorene (6/5/6/5) ring system, results from connecting the tyrosine phenol with the acyclic polyketide chain. A subsequent Diels−Alder cyclization results in the formation of rings A and B. The five-membered cycle (i.e., ring D) is probably formed via electrophilic cyclization of the diene intermediate. 21 Interestingly, embellicines (1−5) share the same core in terms of structure and configurations, and these coisolated compounds differ only in the structure of the γlactam, where embellicine A (1) possesses a double bond between C-18 and C-1′ and, thus, may be the parent molecule of the embellicine class ( Figure S29).

■ EXPERIMENTAL SECTION
General Experimental Procedures. Optical rotations were obtained using a Rudolph Research Autopol III polarimeter (Rudolph Research Analytical). UV and ECD spectra were collected using a Varian Cary 100 Bio UV−Vis spectrophotometer and an Olis DSM 17 ECD spectrophotometer, respectively. 1D and 2D NMR experiments were conducted using a JEOL ECA−500 NMR spectrometer operating at 500 MHz or a JEOL ECS-400 NMR spectrometer, operating at 400 MHz and equipped with a highsensitivity JEOL Royal probe and a 24-slot autosampler. Residual solvent signals were used as an internal standard (For CDCl 3 δ H /δ C 7.26/77.16, acetone-d 6 δ H /δ C 2.05/29.8, and for deuterated dimethyl sulfoxide (DMSO-d 6 ) δ H /δ C 2.50/39.52). UPLC-HRESIMS data were collected via a Thermo Fisher Scientifc Q Exactive Plus mass spectrometer that is connected to Waters Acquity UPLC system. BEH Shield RP18 column (Waters, 1.7 μm; 50 × 2.1 mm) was used and heated to 40°C. The mobile phase consisted of CH 3 CN-H 2 O (1% formic acid) in a gradient system of 15:85 to 100:0 over 10 min and at  Journal of Natural Products pubs.acs.org/jnp Article a flow rate of 0.3 mL/min. MS data were collected from 150 to 2000 m/z while alternating between positive and negative modes. The purity of compounds 1−5 was evaluated using a Waters Acquity UPLC system (Waters Corp.) utilizing a BEH Shield RP18 column (Waters, 1.7 μm; 50 × 2.1 mm). Data were collected and analyzed using Empower 3 software (Waters). All analytical and preparative high-performance liquid chromatography (HPLC) runs were carried out using a Varian Prostar HPLC system equipped with ProStar 210 pumps and a Prostar 335 photodiode array detector (PDA). HPLC data was collected and analyzed using Galaxie Chromatography Workstation software (version 1.9.3.2, Varian Inc.). For preparative HPLC, a Phenomenex Gemini-NX C 18 (5 μm; 250 × 21.2 mm) was used, while a Phenomenex Luna-PFP (5 μm; 250 × 10 mm) was utilized for semipreparative HPLC. Flash chromatography was carried out using a Teledyne ISCO CombiFlash Rf 200 that was equipped with UV and evaporative light-scattering detectors. Fungal Strain Isolation and Identification. The fungal strain MSX6737 was isolated from tropical terrestrial leaf litter by Dr. Barry  Katz (MYCOsearch, Inc.). To identify the fungal strain, both morphological and molecular analyses were employed. The micromorphological characters, such as narrowly cylindrical phialides arising from vegetative hyphae, tapering toward the apexes, producing abundant, one-celled, cylindrical, hyaline conidia in slimy heads ( Figure S30), agree with the concept of the genus Sarocladium. 33,40−42 To corroborate morphological identification and to add an additional means of orthogonal data, the entire internal transcribed spacer region (ITS1, 5.8S, ITS2) as well as the partial large subunit (nrLSU-26S or 28S) were both polymerase chain reaction (PCR) amplified and sequenced using primer combinations ITS1F and ITS4 43,44 for the former and LROR and LR6 for the latter; 45,46 previously outlined protocols were utilized. 47 The Sanger sequenced ITS region was BLAST searched using NCBI GenBank, showing ≥90% sequence homology with Sarocladium, while LSU BLAST search showed ≥98% sequence homology with members of Sarocladium. To better understand the phylogenetic affiliation of strain MSX6737 within the genus Sarocladium, a maximum likelihood phylogeny using combined ITS and LSU was inferred using IQ-TREE 48 in the program PhyloSuite v.1.2.1. 49 All previously described Sarocladium sequences were obtained from recent molecular phylogenetic studies. 33,[40][41][42]50,51 ModelFinder 52 predicted GTR+I+G as the best fitting substitution model for both ITS and LSU regions according to the Akaike Information Criterion. 53 The trimmed nucleotide alignment after removing ambiguous nucleotide positions with GBlocks 54,55 was then used to run the maximum likelihood analysis using IQ-Tree using Ultrafast bootstrapping; 48,49,56 only values ≥95% for the clades were considered strongly supported. The maximum likelihood analysis showed a distinct lineage within Sarocladium, suggesting that strain MSX6737 is a putative new species of Sarocladium, Sarocladiacae, Hypocreales, Ascomycota ( Figure S31). Thus, based on morphology and molecular data, we refer to strain MSX6737 as Sarocladium sp. pending new species description in a mycology journal. 34 The sequence data were deposited in GenBank with accession numbers: ITS: OP650545, OP650546; LSU: OP650548, OP650549.
Extraction and Isolation. EtOAc (900 mL) was added to a largescale solid fermentation culture of MSX6737 before being chopped with a spatula and shaken for ∼18 h at ∼100 rpm at rt. The sample was then filtered using vacuum, and the filtrate was evaporated to dryness. The dried extract was reconstituted in 300 mL of CH 3 CN and 200 mL of hexane. The biphasic solution was stirred for 30 min and then transferred to a separatory funnel. The CH 3 CN layer was drawn off and evaporated to dryness under vacuum. This defatted material (575 mg) was dissolved in CHCl 3 , adsorbed onto Celite 545, and then fractionated via flash chromatography using a gradient solvent system of hexane-EtOAc at a flow rate of 35 mL/min and 28 column volumes over 27 min to afford three fractions. The second fraction of flash chromatography was subjected to a preparative reverse-phase HPLC over a Phenomenex Gemini C 18 column using a gradient system of 60:40 to 80:20 CH 3 CN-H 2 O (0.1% formic acid) over 15 min at a flow rate of 21.2 mL/min to yield five subfractions. Subfractions 3 and 4 were subjected to semipreparative reverse-phase HPLC to yield compound 1 (11.7 mg) and compound 2 (26.0 mg), respectively. Subfraction 5 was subjected to semiprepative HPLC to yield both compounds 4 (3.2 mg) and 5 (3.7 mg). UPLC was used to evaluate the purity of compounds 1−5 using a gradient solvent system of 15:85 CH 3 CN−H 2 O (0.1% formic acid) to 100% CH 3 CN over 3 min; all compounds were >95% pure (Supporting Information, Figure  S28). Acetylation of Embellicine C (2). Pyridine was used to dissolve 5 mg of 2, which was placed in an ice bath before gradually adding 100 μL of acetic anhydride. The reaction was left stirring at room temperature for 5 h and then evaporated to dryness. The reaction mixture was subjected to semipreparative HPLC over a Phenomenex Synergi C 12 column using a gradient system of 40:60 to 90:10 CH 3 CN-H 2 O (0.1% formic acid) over 15 min at a flow rate of 4.7 mL/min to afford 2 mg of the acetylated product, 2′-O-acetylembellicine C (3).
Cytotoxicity Assay. To evaluate the cytotoxic activity of compounds 1−5, human melanoma cancer cells MDA-MB-435, human breast cancer cells MDA-MB-231, and human ovarian cancer cells OVCAR3 were purchased from the American Type Culture Collection. The cell lines were propagated at 37°C in 5% CO 2 in RPMI 1640 medium, supplemented with fetal bovine serum (10%), penicillin (100 units/mL), and streptomycin (100 μg/mL). Cells in log phase growth were harvested by trypsinization followed by two washings to remove all traces of enzyme. A total of 5000 cells was seeded per well of a 96-well clear, flat-bottom plate (Microtest 96, Falcon) and incubated overnight (37°C in 5% CO 2 ). Samples dissolved in DMSO were then diluted and added to the appropriate wells. The cells were incubated in the presence of a test substance for 72 h at 37°C and evaluated for viability with a commercial absorbance assay (CellTiter-Blue Cell Viability Assay, Promega Corp) that measured viable cells. IC 50 values are expressed in μM relative to the solvent (DMSO) control; taxol (paclitaxel) was used as a positive control.
Computational Methods. Molecular Merck force field (MMFF) calculations were carried out with Spartan'10 (Wave function Inc.) to obtain the minimum energy conformers of 2 and 5. These calculations resulted in one major conformer, each, for both 2 and 5. For the ECD prediction, the resulting conformers were optimized using a timedependent DFT (TDDFT) method at the B3LYP/6-311G+(2d,p) level in CH 3 CN. Then, the TDDFT method at the B3LYP/6-31G +(d) level of theory was employed for ECD calculations. All calculations were performed via the Gaussian'09 program package (Gaussian Inc.), 57,58 while the calculated ECD spectra were plotted using Specdis software and compared to the experimental ECD data. 59 ■ ASSOCIATED CONTENT