NF-κB inhibitors, unique γ-pyranol-γ-lactams with sulfide and sulfoxide moieties from Hawaiian plant Lycopodiella cernua derived fungus Paraphaeosphaeria neglecta FT462

LC-UV/MS-based metabolomic analysis of the Hawaiian endophytic fungus Paraphaeosphaeria neglecta FT462 led to the identification of four unique mercaptolactated γ-pyranol–γ-lactams, paraphaeosphaerides E–H (1–4) together with one γ-lactone (5) and the methyl ester of compound 2 (11). The structures of the new compounds (1–5 and 11) were elucidated through the analysis of HRMS and NMR spectroscopic data. The absolute configuration was determined by chemical reactions with sodium borohydride, hydrogen peroxide, α-methoxy-α-(trifluoromethyl)phenylacetyl chlorides (Mosher reagents), and DP4 + NMR calculations. All the compounds were tested against STAT3, A2780 and A2780cisR cancer cell lines, E. coli JW2496, and NF-κB. Compounds 1 and 3 strongly inhibited NF-κB with IC50 values of 7.1 and 1.5 μM, respectively.

Compound 5 was isolated as a colorless solid. Its molecular formula was determined to be C 14 1 H NMR signals in the 4-dihydropyranol (ring A) and the side chain at 8-position of 5 were almost the same as those of compound 6, while the N-methoxyl group and the olefinic protons at 14-position in ring B of 6 were absent in 5. Instead, a third methyl group at δ H 1.64 ppm was observed in the 1 H NMR spectrum of 5. The signal at δ H 1.64 ppm showed HMBC correlations to C-3 (δ C 102.5) and C-4 (δ C 174.7) indicating C-3 of 5 must be di-oxygenated, and 5 should be a lactone instead of a lactam. Hence, the structure of compound 5 was determined as shown.
To determine the configuration of 3-position of 1, 1 was treated with sodium borohydride 14 . We expected the double bond between A-and B-rings would be reduced so that we could run a ROESY NMR experiment, but instead the reaction yielded phaeosphaeride A (6) (Fig. 3), indicating that the configuration of ring A of 1 was the same as that of 6. In order to determine the configuration of C-2′, 1 and 2 were reacted with R-and S-Mosher reagents 15 . Results showed that both 1 and 2 should have an S configuration at 2′-position (Fig. 4). To determine   the absolute configuration unequivocally by the modified Mosher's method, the ∆ SR values on both left and right sides of a secondary alcohol should be calculated and the sign distributions should be consistent. Unfortunately, the ∆ SR value at the carboxylic acid proton for either compounds 8 and 7, or 10 and 9 couldn't be obtained. Fortunately, we isolated the methyl ester of compound 2 (11), which was reacted with both S-and R-Mosher reagents to yield two Mosher esters (12 and 13, Fig. 4). Result clearly showed negative ∆ SR values for both H-2′ and the methoxy at 3′ position, and positive ∆ SR values for the others (Fig. 4). Hence, the configuration of 1 at 2′-position was determined without any doubt. In our previous study, we determined the configuration at C-3 of paraphaeosphaeride A as S from experimental and calculated Electronic Circular Dichroism (ECD) analysis 2 . Given the close structural similarity of the former natural product with 1 and 2, we hypothesized that the absolute configuration at C-3 should be also S in both newly isolated compounds. In order to confirm our deduction, we next carried out DFT calculations of the NMR shifts using the GIAO method 16,17 . This approach has been extensively employed recently to settle the structure and stereochemistry of complex natural or synthetic products, and emerges as a suitable alternative to propose the most likely structure of an organic compound in a simple and affordable fashion [16][17][18][19][20][21] . Among the several strategies that have been developed to support (or reject) a given structural proposal 17, 22-26 , we decided  to use the DP4 + probability, the method of choice for assessing the most likely candidate when only one set of experimental data is available (as in this case) 23 . This probability is an updated and improved version of the DP4 method (developed by the Goodman group) 25 that includes the use of both scaled and unscaled chemical shifts computed at higher levels of theory. Thus, following the DP4 + methodology, we computed the NMR shifts at the PCM/mPW1PW91/6-31 + G**//PCM/B3LYP/6-31G* level of theory using methanol as solvent. Since the NMR data of compounds 1 and 2 were collected in CD 3 OD, here we also include the solvent effect during the geometry optimization stage. Confirming our hypothesis, the calculated values for the two isomers with the S configuration at C-3 (1-3S and 2-3S) (Tables 2 and SI) showed better agreement with the experimental data than those corresponding to the 3R counterparts (1-3R and 2-3R, respectively). For instance, the CMAE (corrected mean average error, defined as Σ n |δ scaled − δ exp |/n) computed for 1-3S and 2-3S from 13 C NMR data were 1.0 ppm and 1.1 ppm, respectively, lower than those calculated for 1-3R (1.6 ppm) and 2-3R (1.8 ppm). In a similar fashion, the CMAE values calculated from 1 H NMR data were lower for 1-3S (0.10 ppm) and 2-3S (0.08 ppm) than those of 1-3R (0.11 ppm) and 2-3R (0.10 ppm), respectively. The isomers with a 3R configuration (1-3R and 2-3R) also displayed higher outliers (CMaxErr, corrected maximum error, defined as max|δ scaled − δ exp |) values (7.3 ppm and 4.4 ppm, respectively, for 13 C data; 0.50 and 0.24 ppm, respectively, for 1 H data) in comparison with the corresponding values of 1-3S and 2-3S (3.3 ppm and 4.0 ppm, respectively, for 13 C data; 0.24 and 0.21 ppm, respectively, for 1 H data). The better fit of 1-3S and 2-3S was further corroborated with DP4 + calculations, where each isomer was identified as the most likely candidates in high probability (>99.9% in both cases).
Reduction of compound 3 with sodium borohydride also yielded compound 6. As proposed in our previous publication 2 , nucleophilic addition of the mercaptolactate thiol to C-14 of compound 6 could generate 1 and its 3-epimer (1′, which was not isolated from Paraphaeosphaeria neglecta FT462 in this study). Further oxidation of 1 and 1′ could yields four sulfoxides including 3 (3, 3′, 3″, and 3′″, Fig. 5). From a biosynthetic point of view, 3 should have the same configuration at 2′-position as 1 and 2. To confirm the above assumption, 1 was treated with  Table 2. Scaled (δ s ) 1 H and 13 C NMR chemical shifts of the more likely structures of 1-5 computed at the PCM/ mPW1PW91/6-31 + G**//PCM/B3LYP/6-31 G* level of theory (solvent = methanol). hydrogen peroxide and tert-butyldimethyl silyl chloride 27 , and compound 3 was detected (Fig. 5). However, the configuration at the sulfoxide in 3 was still not determined. Once again, the two possible isomers of 3 (3-SR and 3-SS) were submitted to NMR calculation at the PCM/mPW1PW91/6-31 + G**//PCM/B3LYP/6-31 G* level of theory (Tables 2 and SI). In this case, the 13 C and 1 H NMR data pointed toward different directions, which is a common situation in the field of quantum calculations of NMR shifts [16][17][18][19][20][21][22][23][24][25][26] . For instance, the CMAE and CMaxErr computed for 3-SR using 13 C NMR data (1.8 ppm and 5.4 ppm, respectively) were slightly lower than those calculated for 3-SS (1.9 ppm and 5.7 ppm, respectively), whereas using the experimental 1 H NMR shifts the trend was reversed (0.14 ppm and 0.56 ppm, respectively, for 3-SR; 0.11 and 0.44 ppm for 3-SS). In this case, proton data was the most conclusive one, and isomer 3-SS was identified as the correct structure of 3 in high probability (99.5%) after DP4 + calculations. Two more peaks, each of which has the same molecular weight as that of 3, were detected in the LC-MS of fraction B. Most likely, they are diastereoisomers of 3 with different configurations at the sulfoxide, 3-, and 6-positions since all the analogs that we isolated have the same configurations at 7-and 8-positions. We tried to isolate these compounds and determined their structures, but failed due to insufficient material and instability of these molecules. Nevertheless, it is worthy to investigate these compounds further.
In the case of compounds 4 and 5, we concluded that they should have the same configurations at C-6, C-7 and C-8 as those of 1 since minor differences in the 13 C NMR shifts were noted at those centers. Also, from a biosynthetic point of view, 4 should have the same configuration at 2′-position as 1-3. However, as throughout this study, we were not able to unequivocally determine the configuration at C-3 from experimental NMR basis, and we had to rely on the computational chemistry assistance to settle this issue. Thus, the NMR calculations of the two possible isomers of 4 (4-3E and 4-3Z) and 5 (5-3R and 5-3S) were carried out at the PCM/ mPW1PW91/6-31 + G**//PCM/B3LYP/6-31G* level of theory ( Table 2, Tables S4 and S5). In the case of 4, the 3Z isomer displayed much better agreement with the experimental values than the 3E candidate, mainly in the 13 C NMR region, in which lower CMAE (2.2 ppm vs 2.9 ppm) and CMaxErr (7.0 ppm vs 14.2 ppm) was noted. On the other hand, the computed 13 C NMR shifts of 5-3R and 5-3S showed similar match with the experimental shifts (CMAE 2.0 ppm in both cases), though the prediction of the 1 H NMR data was improved in the case of 5-3R (CMAE = 0.08 ppm vs 0.09 ppm; CMaxErr 0.22 ppm vs 0.25 ppm). The DP4 + calculations were in line with these findings, suggesting that the most likely configuration at C-3 of 4 and 5 is Z (>99.9%) and R (92.2%), respectively. ROESY correlations between H-14 and H-6, and H-14 and H-8 further confirmed the configuration of 5.
Since phaeosphaeride A showed STAT3 inhibition 2, 3 , and NF-κB and STAT3 act as two major transcriptional factors linking inflammation with cancer progression, and they functionally interact with each other at many different levels, we tested all compounds for their ability to inhibit NF-κB and iNOS. When evaluated in a mammalian cell-based assay designed to monitor TNF-α-induced NF-κB activity, compounds 1 and 3 were found to mediate inhibitory responses with IC 50 values of 7.1 and 1.5 μM (Table 3), respectively. When tested using the same conditions as the NF-κB assay, none showed toxicity (Table 3). In the absence of a cytotoxic response, inhibition of TNF-α-induced NF-κB activity suggests the potential of a cancer chemopreventative response. Compounds 1 and 3 also inhibited iNOS with IC 50 values of 47.9 and 16.1 μM (Table 3), respectively. According to the structures and NF-κB activity, it could be concluded that the sulfoxide (3) was more active than the sulfide derivatives (1 and 2), and the configuration at 6-position (6S) was also important for the NF-κB inhibition.
for their anti-proliferative, antibacterial, NF-κB and iNOS inhibitory activities. Since compounds 1 and 3 inhibited NF-κB with some specificity, they remain of interest as cancer chemopreventative agents.
Anti-proliferative, antibacterial, NF-kB and iNOS assays. Anti-proliferative Assays. Viability of normal mouse fibroblasts (NIH3T3) and two Stat3-activated cancer cell lines, MDA-MB-231 (breast cancer) and U251 MG (glioblastoma) cells was determined using the CyQuant assay according to the manufacturer's instructions (Life Technologies, CA, USA) 38,39 . Briefly, cells were cultured in 96-well plates at 1000 cells per well for 24 h and subsequently treated with compounds (20 μg/mL) for 72 h and analyzed. Relative viability of the treated cells was normalized to the DMSO-treated control cells 38,39 .
Antibacterial assay. MIC values were determined against E. coli JW2496 (∆bamB) and other bacteria using the standard microbroth dilution method exactly as previously described 40 , which is based on the methods by the Clinical and Laboratory Standards Institute 41,42 . The maximum test concentration used was 20 μg/mL. NF-κB assay. We employed human embryonic kidney cells 293, Panomic for monitoring changes occurring along the NF-κB pathway 43 . Stable constructed cells were seeded into 96-well plates at 20 × 10 3 cells per well. Cells were maintained in Dulbecco's modified Eagle's medium (DMEM) (Invitrogen Co.), supplemented with 10% FBS, 100 units/mL penicillin, 100 μg/mL streptomycin, and 2 mM L-glutamine. After 48 h incubation, the medium was replaced and the cells were treated with various concentrations of test substances. TNF-α (human, recombinant, E. coli, Calbiochem) was used as an activator at a concentration of 2 ng/mL (0.14 nM). The plate was incubated for 6 h. Spent medium was discarded, and the cells were washed once with PBS. Cells were lysed using 50 μL (for 96-well plate) of reporter lysis buffer from Promega by incubating for 5 min on a shaker, and stored at −80 °C. The luciferase assay was performed using the Luc assay system from Promega. The gene product, luciferase enzyme, reacts with luciferase substrate, emitting light, which was detected using a luminometer (LUMIstar Galaxy BMG). Data for NF-κB inhibition are expressed as IC 50 values (i.e., concentration required to inhibit TNFinduced NF-kB activity by 50%). The known NF-κB inhibitor TPCK was used as a positive control.
Nitric Oxide Assay. RAW 264.7 cells (1 × 10 4 cells/well) were seeded and incubated in 96-well culture plates at 37 °C, 5% CO 2 in humidified air for 24 h. Then, complete medium was replaced with phenol red-free medium containing various concentrations of the test compounds, followed by LPS stimulation (1 μg/mL) for 20 h. The nitrite released in the culture media was reacted with Griess reagent, and the absorbance was measured at 540 nm. The amount of nitrite was calculated using a standard curve of known nitrite concentration versus absorbance at 540 nm. TPCK was used as a positive control. IC 50 values were calculated using Computational details. All the quantum mechanical calculations were performed using Gaussian 09 45 . The conformational search was done using the MMFFaq force field (implemented in Spartan 08) 46 . Given the high conformational flexibility of this system, we used a cutoff energy of 5 kcal/mol, yielding an average of ~500, ~40, and ~400 different conformations per isomer in the cases of 1-3, 4 and 5, respectively. Next, each conformation was fully optimized at the PCM/B3LYP/6-31 G* level of theory in methanol as solvent using Gaussian 09, followed by frequency calculations at the same level to determine the nature of the stationary points and to compute the thermochemical properties (at 1.0 atm and 298.15 K). The most stable conformers (up to 1.5 kcal/mol from the global minima) were next subjected to NMR calculations. The magnetic shielding constants (σ) were computed using the gauge including atomic orbitals (GIAO) method [47][48][49][50] , the method of choice to solve the gauge origin problem 16,17 , at the PCM/mPW1PW91/6-31 + G** level of theory. The calculations in solution were carried out using the polarizable continuum model, PCM [51] , with methanol as the solvent). The unscaled chemical shifts (δ u ) were computed using TMS as reference standard according to δ u = σ 0 − σ x , where σ x is the Boltzmann averaged shielding tensor (over all significantly populated conformations) and σ 0 is the shielding tensor of TMS computed at the same level of theory employed for σ x . The Boltzmann averaging was done according to the eq. 1: (1) where σ i x is the shielding constant for nucleus x in conformer i, R is the molar gas constant (8.3145 J K −1 mol −1 ), T is the temperature (298 K), and Ei is the Gibbs free energy of conformer i (relative to the lowest energy conformer), obtained from the PCM/B3LYP/6-31 G* frequency calculations. The scaled chemical shifts (δ s ) were computed as δ s = (δ u − b)/m, where m and b are the slope and intercept, respectively, resulting from a linear regression calculation on a plot of δ u against δ exp . The DP4 calculations were carried out using the Applet from the Goodman group (at www-jmg.ch.cam.ac.uk/tools/nmr/DP4/). The DP4 + calculations were carried out using the Excel spreadsheet available for free at sarotti-nmr.weebly.com, or as part of the Supporting Information of the original paper 23 .