Glycosylated and Succinylated Macrocyclic Lactones with Amyloid-β-Aggregation-Regulating Activity from a Marine Bacillus sp.

Two new glycosylated and succinylated macrocyclic lactones, succinyl glyco-oxydifficidin (1) and succinyl macrolactin O (2), were isolated from a Bacillus strain collected from an intertidal mudflat on Anmyeon Island in Korea. The planar structures of 1 and 2 were proposed using mass spectrometric analysis and NMR spectroscopic data. The absolute configurations of 1 and 2 were determined by optical rotation, J-based configuration analysis, chemical derivatizations, including the modified Mosher’s method, and quantum-mechanics-based calculation. Biological evaluation of 1 and 2 revealed that succinyl glyco-oxydifficidin (1) inhibited/dissociated amyloid β (Aβ) aggregation, whereas succinyl macrolactin O (2) inhibited Aβ aggregation, indicating their therapeutic potential for disassembling and removing Aβ aggregation.


Introduction
Glycosylation is an important biological process in enhancing the structural and biological diversity of metabolites [1,2]. In many cases, the biological function of natural products is altered after glycosylation; glycosylation is thus a frequently adopted route to the functional modification of molecules in nature [3]. For example, a study on the effects of glycosylation on the biological activity of the microbial immunosuppressive drug rapamycin revealed that glycosylation improved water solubility and reduced cytotoxicity depending on its positions [4]. A chemical investigation of Bacillus sp. reported glycosylated macrolactins, which displayed inhibitory activity against Staphylococcus aureus peptide deformylase along with antibacterial activity differing from that of previously reported aglycone macrolactins, indicating that the discovery of glycosylated natural products would lead to the diversification of their biological functions [5].
Marine bacteria are also fruitful sources of structurally and biologically diverse natural product discovery [6]. Since the 21st century, more and more drug candidates were discovered from Gram-positive bacteria such as Streptomyces sp., Bacillus sp. and so on [7]. Our chemical studies on marine bacteria discovered pulvomycins B-D, new macrolides incorporating sugar, and pulvomycin D showed potent cytotoxic effects against cancer cell lines [8]. Suncheonosides A-D, hexasubstituted benzothioate glycosides, were reported
The double-bond geometry configurations of 1 were determined by 1 H-1 H coupling constants and ROESY correlations. The 3 J H7H8 value (11.0 Hz) determined the 7Z configuration, which was also supported by the H-7 ( 3.420) assigned these three protons in the same plain, and thus the sugar was determined to be β-glucose. In addition, 3 J H32H33a (2.2 Hz), 3 J H32H33b (6.5 Hz), and 3 J H33aH33b (12.0 Hz) supported β-glucose ( Figure 3). The relative configuration between C-4 and C-5 was revealed by J-based configuration analysis [17]. Long-range 13 C− 1 H coupling constants were measured by hetero-halffiltered TOCSY (HETLOC) NMR experiments [18]. The large coupling constant of 3  Once the assignment of β-glucose was established, chemical derivatization was conducted for the absolute configuration of the glucose [19]. We used acid hydrolysis to break the connection between the corresponding aglycone with glucopyranose. β-glucopyranose was assigned as β-D-glucopyranose by chiral derivatization with L-cysteine methyl ester hydrochloride and σ-tolyl isothiocyanate and subsequent LC/MS analysis ( Figure S16).
Succinyl macrolactin O (2) was isolated as a yellow powder. The molecular formula of 2 was deduced as C 34 H 48 O 13 based on its HRESIMS data. Its molecular formula revealed 11 degrees of unsaturation. The UV spectrum (λ max : 236 and 260 nm) of 2 indicated at least two chromophores in the structure. By careful analysis of 1 H, 13    Analogously to 1, the sugar moiety in 2 was assigned as β-glucose by 1 H-1 H coupling constants and ROESY correlations. Its absolute configuration was revealed as β-D-glucopyranose by the chiral derivatization and LC/MS analysis ( Figure S28). As many macrolactin derivatives were discovered and their absolute configurations are conserved in the family, we compared the

Biological Activity
According to the previous studies, the antimicrobial activities of difficidin [23] and macrolactin [11] families were reported. However, we could not find any activity of 1 and 2 in our antimicrobial assays (Tables S4-S6). Therefore, we searched for unreported activity of the difficidin and macrolactin families and targeted amyloid-β-aggregation-regulating activity for these two compounds. Amyloid-β (Aβ) aggregates in the brain of patients with Alzheimer's disease (AD) are considered the pathological and biological hallmarks of this neurodegenerative disorder [24]. Thus, drug candidates have been discovered to inhibit and reverse the Aβ aggregation process [25]. To investigate whether our compounds, succinyl glyco-oxydifficidin (1) and succinyl macrolactin O (2), inhibit and/or reverse Aβ aggregation, we performed two different types of assays, inhibition and dissociation, utilizing a high throughput screening platform we recently developed [26]. For both assays, we immobilized Aβ 1-42 with an additional C-terminal cysteine on the maleimidecoated 96-well plate and added fluorescent Aβ 1-42 to induce on-plate oligomer and fibril formation of Aβ ( Figure 6A). First, we assessed 1 and 2, in Aβ aggregation inhibition assay ( Figure 6A). Each compound, at concentrations of 5 and 50 µM, was added to the Aβ 1-42immobilized plate with the fluorescent Aβ 1-42 (10 µM) and incubated for 24 h at room temperature (RT) to observe in situ aggregation inhibition. We used a previously reported Aβ aggregation inhibitor, (1r,2r,3r,4r,5r,6r)-cyclohexane-1,2,3,4,5,6-hexol (scyllo-inositol), as a positive control [27]. The plate was washed after the incubation step, and we measured the levels of the remaining fluorescent Aβ 1-42 . Data were then normalized to the signal of fluorescent Aβ 1-42 -free wells as 100% inhibition of Aβ aggregation, and the inhibition rate (%) was analyzed as previously reported [26]. As a result, both compounds inhibited Aβ aggregation significantly; 1 by 35.54% (at 5 µM) and 54.35% (at 50 µM), and 2 by 18.88% (at 5 µM) and 40.08% (at 50 µM), when the control compound was inhibited by 43.00% (at 5 µM) and 52.22% (at 50 µM) ( Figure 6B).
As a result, both compounds reversed Aβ aggregation significantly; 1 by 28.46% (at 5 µM) and 49.27% (at 50 µM), and 2 by 30.04% (at 5 µM) and 31.02% (at 50 µM), when the control compounds dissociated pre-formed aggregates by 46.42% (EPPS at 5 µM) and 59.94% (Nec-1 at 5 µM) ( Figure 6C). Next, we docked compounds 1 and 2 to the U-shaped oligomeric Aβ 1-42 structure (PDB ID: 2BEG) to predict the potential binding interactions between them (Figure 7). The docking score of 1 (−9.9 kcal/mol) was slightly better than that of 2 (−9.3 kcal/mol). The docking models suggested that the branched carbon chain of 1, which is not present in 2, contributes to forming extensive hydrophobic contacts with the core of Aβ aggregate. Also, the additional contacts of 1 to the edge strand of Aβ aggregate seem to be a primary factor for inhibiting or dissociating Aβ aggregation. In contrast, 2 showed similar but relatively unfavorable interactions with the Aβ hydrophobic core through the sugar moiety ( Figure 7). Overall, we observed that succinyl glyco-oxydifficidin (1) dose-dependently inhibited/dissociated Aβ aggregation, whereas succinyl macrolactin O (2) dose-dependently inhibited Aβ aggregation. In these assays, we assumed that our scaffold had the therapeutic potential to disassemble and remove Aβ aggregation.

General Experimental Procedures
Optical rotations were measured by a Jasco P-2000 polarimeter with a 1.0 cm cell (Tokyo, Japan). UV and CD spectra were recorded using an applied photophysics Chirascan plus spectrometer (Leatherhead, UK). IR spectra were acquired with a JASCO FT/IR-4200 spectrometer (Tokyo, Japan). 1 H, 13 C, and 2D NMR spectra were obtained on Bruker Avance 800 MHz NMR spectrometers (Billerica, MA, USA), all the signals being referenced to 13 C (49.045 ppm) and 1 H (3.306 ppm) signals of CD 3 OD [30]. Electrospray ionization source (ESI) low-resolution LC/MS data were collected on an Agilent Technologies 6130 quadrupole mass spectrometer (Santa Clara, CA, USA) coupled with an Agilent Technologies 1200 series HPLC using a reversed-phase C 18 (2) column (Phenomenex Luna, 5 µm, 4.6 × 100 mm, Torrance, CA, USA). High-resolution electrospray ionization source (ESI) LC/MS data were collected on an AB SCIEX Q-TOF 5600 high-resolution mass spectrometer at the National Instrumentation Center for Environmental Management (NICEM, Seoul, Republic of Korea).

Isolation and Identification of the Bacterial Strain Bacillus sp. AMD05
The strain, AMD05, was isolated from a mudflat sample collected from the intertidal mudflat on Anmyeon Island, Republic of Korea, using a sterilized 40 mL plastic tube. Various strain isolation media were applied for single-strain isolation, while AMD05 was isolated on a YEME-based agar medium (10 g/L of malt extract, 4 g/L of yeast extract, 4 g/L of glucose, and 18 g/L of agar) incubated at 25 • C for 7 days. The AMD05 strain was most closely related to Bacillus velezensis F-30 (97% identity, accession # MF988699) according to 16S rDNA sequence analysis (AMD05 16S rDNA GenBank deposit #OM319625).

Cultivation and Extraction
The spores of the bacterial strain Bacillus sp. AMD05 were inoculated into 50 mL of YEME liquid medium in a 125 mL flask. The culture was incubated at 200 rpm at 30 • C for two days. After incubation, 10 mL of the AMD05 liquid culture was inoculated into a 500 mL Erlenmeyer flask containing 200 mL of YEME medium and shaken at 170 rpm and 30 • C for two days. Then, 15 mL of the medium culture was transferred into 1 L of YEME medium in a 2.8 L Fernbach flask for four days at 170 rpm and 30 • C (24 ea × 1 L, total volume 24 L). The entire culture was extracted with 36 L of ethyl acetate (EtOAc). The EtOAc layer was separated using a separation funnel (capacity 3 L), and the residual water in the EtOAc layer was removed by adding anhydrous sodium sulfate. The extract was concentrated in vacuo to yield dry material. This procedure was repeated 3 times (72 L of culture in total) to yield extracted material.

Determination of the Configuration of the Sugar in 1 and 2
Succinyl glyco-oxydifficidin was hydrolyzed with 6 N HCl at 115 • C for 1 h to yield the free glucopyranose. After drying in vacuo, the acid hydrolysate was reacted with L-cysteine methyl ester hydrochloride and σ-tolyl isothiocyanate at 60 • C, each for 1 h. The authentic β-L-glucose and β-D-glucose were also reacted with L-cysteine methyl ester hydrochloride and σ-tolyl isothiocyanate at 60 • C, each for 1 h. The β-glucopyranose reaction product from 1 was co-injected with the products of the authentic β-L-glucose and β-D-glucose using LC/MS analysis (gradient solvent conditions: 10-100% aqueous CH 3 CN (0.1% formic acid) for 20 min). The reaction products of β-glucopyranose from 1 had the same retention time as the reaction product of authentic β-D-glucose (only one peak in the LC/MS spectrum). This method was also applied for succinyl macrolactin O (2), identifying β-D-glucose ( Figure S28).

MTPA Esterification of Succinyl Glyco-Oxydifficidin (1)
Succinyl glyco-oxydifficidin was transferred to two 40 mL vials (2 mg of 1 in each vial) and dried completely under high vacuum overnight. A total of 1 mL of distilled anhydrous pyridine was added to each vial under argon gas. The mixtures were stirred at room temperature for approximately 5 min. Then, Rand S-α-methoxy-α-(trifluoromethyl) phenylacetyl chloride (MTPA-Cl) (50 µL) were added into one of the two vials. The reactions were terminated after 30 min by adding 50 µL of MeOH. The reaction mixtures were dried in vacuo and subjected to reversed-phase HPLC (Kromasil C 18 column, 5 µm, 10 × 250 mm). An isocratic solvent system (94% aqueous CH 3 CN for 40 min, flow rate: 2 mL/min, detection: UV 230 nm) was used. The S-MTPA ester (1a) and the R-MTPA ester (1b) of 1 were both eluted at a retention time of 30 min. Low-resolution LC/MS analysis was carried out for 30 min using from 70% to 95% aqueous CH 3 CN (0.1% formic acid) with an IB chiral column. The ∆δ S-R values around the stereogenic centers were assigned by analyzing their 1 H NMR and 1 H-1 H COSY NMR spectra (Figures S8-S11). S

Conformational Search and DP4 Calculations
For the determination of the configurations of C-4, C-5, and C-21, 1c (4R, 5S, and 21S), 1d (4R, 5S, and 21R), 1e (4S, 5R, and 21S), and 1f (4S, 5R, and 21R) were generated by Avogadro 1.2.0. A conformational search for these diastereomers was performed by MacroModel with the Merck Molecular Force Field to find the stable conformers (with 10 kJ/mol energy limit) of the diastereomers: twelve conformers for 1c (4R, 5S, and 21S), nine conformers for 1d (4R, 5S, and 21R), six conformers for 1e (4S, 5R, and 21S), and twenty-nine conformers for 1f (4S, 5R, and 21R) ( Table S1). The Boltzmann populations of the conformers were also calculated by MacroModel. Ground-state geometry optimization was performed by density functional theory (DFT) modeling of Turbomole X 4.3.2. All calculations were performed at the B3LYP/def-SV(P) level in the gas phase. This basis set, taken from the work [31], was used for all atoms. Calculated chemical shifts were calculated based on this equation: δ X calc. = σ 0 −σ X 1−σ 0 /10 6 . δ X calc. is the calculated chemical shifts of nucleus x (e.g., 1 H or 13 C), while σ X and σ 0 are the calculated isotropic constants of nucleus x and tetramethylsilane (TMS) [33]. The calculated NMR chemical shifts of each conformer were averaged by the Boltzmann populations. By comparing these Boltzmann-populationaveraged chemical shifts with the experimental chemical shifts of 1 (Table S2), the DP4 calculation result proposed 1c (4R, 5S, and 21S) configurations with 100.0% probability using both carbon and proton data (71.0% probability using only the carbon data and 100.0% probability using only the proton data) ( Figure S17).

Aβ Aggregation Assay Plate Preparation
A maleimide-activated microplate was used to immobilize the peptide. The bovine serum albumin coated on the maleimide-activated microplate was removed after washing the peptide three times with 200 µL of wash buffer (0.1 M sodium phosphate, 0.15 M sodium chloride, 0.05% Tween-20; pH 7.2) in each well on the plate. A full-length Aβ 1-42cys solution (50 µg/mL, 5% DMSO) was made in binding buffer (0.1 M sodium phosphate, 0.15 M sodium chloride, 10mM EDTA; pH 7.2). A total of 100 µL of 10 µM peptide solution was added to each well and reacted with maleimide for 24 h at RT. After peptide immobilization, the unbounded peptides were washed three times with 200 µL of wash buffer. To inactivate additional maleimide groups, 200 µL of cysteine solution was added to each well and incubated at RT for 1 h. After cysteine capping, all the wells in the plate were washed three times with 200 µL of wash buffer.

Aβ Aggregation Assay
To examine the inhibition effect of succinyl glyco-oxydifficidin and succinyl macrolactin O, 50 µL of succinyl macrolactin and succinyl glyco-oxydifficidin was prepared in binding buffer (1% DMSO), and 50 µL of Flmma 552-labeled full-length Aβ 1-42 peptide solution (20 µM) was prepared in binding buffer (1% DMSO). Each solution was made with two concentrations: 5 µM and 50 µM. The solution containing two compounds and Flamma-552-labeled full-length Aβ 1-42 peptide was added to the plate and incubated for 24 h at RT. After the incubation, all the wells were washed three times with 200 µL of wash buffer. For the comparison, we only added 100 µL of Flamma-labeled full-length Aβ 1-42 peptide solution (10 µM) on the other wells and incubated them for 24 h at RT. The intensity of the dug-treated well and control well was measured by the microplate reader.

Aβ Disociationn Assay
Before treating succinyl glyco-oxydifficidin and succinyl macrolactin O, 100 µL of fluorescent Aβ 1-42 peptide solution (10 µM) was added to the wells and incubated for 8 h at 37 • C. After the incubation, all the wells were washed three times with 200 µL of the wash buffer. Succinyl macrolactin (5, 50 µM) and succinyl glyco-oxydifficidin (5, 50 µM) were prepared in a binding buffer (1% DMSO). EPPS and necrostatin-1 (5 µM each) were prepared as positive controls. We added 100 µL of each compound solution to the wells and incubated them for 24 h at RT. After the incubation, all the wells were washed three times with 200 µL of the wash buffer, and 100 µL of binding buffer was added to each well prior to reading. Fluorescent scanning was carried out with a microplate reader (555/580 nm, ex/em).

Docking Model Generation
The 500 three-dimensional conformers were generated using RDKit (https://www. rdkit.org/, accessed on 15 December 2022) with 0.2 Å RMSD threshold for succinyl glycooxydifficidin and succinyl macrolactin O, respectively. Each conformer was used to define potential binding sites by applying global docking with PatchDock program [34] without a predefined binding region. The U-shaped Aβ 1-42 (PDB ID: 2BEG) was used as a receptor structure. The docking search space for the receptor was confined to edge strands of β-sheets to reflect the experimental results (Figure 7). The top 10 docking models for each conformer were retrieved to infer binding site information (center_x, center_y, and center_z parameters) for the subsequent docking refinement by Autodock Vina [35]. The final docking model was selected based on the Autodock Vina score.

Statistical Data Evaluation
All results were given as a mean ± standard error of the mean (SEM). All data were analyzed through GraphPad Prism 9.0 software and compared using One-way ANOVA analysis followed by Bonferroni's post-hoc comparison. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001.

Conclusions
Succinyl glyco-oxydifficidin (1), a new glycosylated and succinylated member of the difficidin family, was discovered in an intertidal mudflat-derived Bacillus sp. AMD05. Compared its structure with that of oxydifficidin, succinyl glyco-oxydifficidin (1) was modified through glycosylation and succinylation, along with a double-bond migration from C-6 to C-7. Although difficidin was discovered more than 35 years ago in 1986, its absolute configuration had remained undetermined until now. By using combinational tools of spectroscopic analysis, the modified Mosher's method, and DP4 computational calculation, we elucidated the absolute configuration of succinyl glyco-oxydifficidin (1) for the first time among the difficidin family compounds. Interestingly, succinyl macrolactin O (2) is the only compound in the macrolactin family to bear both glucose and succinyl acid.
Furthermore, succinyl glyco-oxydifficidin (1) inhibited/dissociated Alzheimer-diseaserelated Aβ aggregation and succinyl macrolactin O (2) inhibited Aβ aggregation, indicating their therapeutic potential to regulate Aβ aggregation. Even though the difficidin and macrolactin families were discovered in the 1980s with anti-microbial bioactivity [11,23], their Aβ-regulating activities were observed for the first time here. Succination and glycosylation may distribute to bioactive modification. Our discovery of the new glycosylated and succinylated macrocyclic lactones with Aβ-regulating activity from marine Bacillus sp. highlights that marine bacteria are prolific sources of natural products diversified by glycosylation, an important biological process for changing the structures and bioactivity of compounds.