Introduction

Sirtuins (SIRTs) are enzymes belonging to the NAD+-dependent lysine deacetylase (KDAC) family that are highly conserved from yeast to human. There are seven isoforms of SIRTs (SIRT1–7) in mammals, and SIRT1 is mainly localized in the nucleus, whereas SIRT2 is found in the cytoplasm to control cellular cytoskeleton dynamics. SIRT3–5 are mainly localized in mitochondria and regulate mitochondrial functions.1 SIRTs remove acetyl groups from various proteins, including histones H3 and H4, PGC-1α, p53, NF-κB, FOXO and so on, thereby regulating energy metabolism, genome stability, longevity and aging as well as age-associated pathologies.2 SIRT2 deacetylates receptor-interacting protein 1 (RIP1) and modulates RIP1–RIP3 complex formation and tumor necrosis factor-α-stimulated necrosis. Deletion or knockdown of SIRT2 prevents formation of RIP1–RIP3 complex in mice.3 Also, the SIRT2 inhibitor AK-7 acts as a neuroprotective agent in Huntington’s disease mouse model.4 Thus SIRTs have been one of the most promising targets of antiaging treatments.

In this study, we have attempted to discover new SIRT inhibitors from a marine sponge Halichondria sp. collected at Hachijo-jima Island. As a result, halistanol sulfate (1)5 as well as two novel analogs were identified (Figure 1). Highly sulfated steroids such as halistanol sulfates are often found as ‘HIT’ in wide variety of screenings as enzyme inhibitory,6 antimicrobial7 or antiviral activities.8, 9, 10 Therefore, they are often regarded as ‘non-specific nuisance compounds’.11 However, no study has shown a mechanism for enzymatic inhibitory activities of halistanol sulfate (1) at the molecular/atomic level. We could, for the first time, reveal the binding mode of halistanol sulfate (1) to SIRT3 protein at the atomic level on the basis of crystal structure study. This finding may be a cue for understanding the mechanisms of the wide variety of bioactivities of halistanol sulfate (1) and designing a new drug motif with sulfate groups as the pharmacophore.

Figure 1
figure 1

Structure of halistanol sulfate (1), halistanol sulfates I (2) and J (3).

Results and discussion

The frozen specimen of Halichondria sp. (695 g wet weight), collected by hand using SCUBA at Hachijo-jima Island, was extracted with MeOH. The combined methanolic extract was evaporated in vacuo and subjected to solvent partitioning and ODS flash chromatography, followed by repetitive reversed-phase HPLC. Final purification was carried out by recycling reversed-phase HPLC on C18 column to afford 6.6 and 3.0 mg of halistanol sulfates I (2) and J (3), respectively (9.5 × 10−3% and 5.5 × 10−3% yield based on wet weights). Halistanol sulfate (1) was also identified as the major component (301.4 mg) based on the HRESIMS (m/z 731.2194 [M-Na], calcd for C29H49Na2O12S3 m/z 731.2187) and NMR data (see Table 1).

Table 1 NMR spectral data for halistanol sulfate (1) and halistanol sulfates I (2) and J (3) in CD3OD (400 MHz)

Halistanol sulfate I (2) had a molecular formula of C29H47Na3O12S3, which was determined by the negative mode HRESIMS analysis (m/z 729.2027 [M-Na], calcd for C29H47Na2O12S3 m/z 729.2031). The ESIMS showed sequencial ion peaks at m/z 729.2 [M-Na], 609.3 [729.2-NaSO4H] and 489.3 [609.3-NaSO3H], indicating that 2 have three sulfate groups in the molecule. IR absorption at 1248 and 1220 cm−1 supported the presence of the sulfate groups. 1H and 13C NMR spectra of 2 (Table 1) indicated the similarity between two structural features of 1 and 2; that is, two singlet methyl protons at C-18 and 19 (δH 0.69 and 1.06, respectively) and three oxymethine protons at C-2, C-3, and C-6 (δH 4.80/4.82, 4.75/4.74 and 4.19/4.15, respectively, for 1/2). Based on the COSY and HMQC correlations, as well as HMBC cross peaks from the H3-18, H3-19, and H3-21, the same steroidal core (C-1 to C-22) as 1 was deduced for 2. The obvious differences between 1 and 2 lay in the terminal parts of the side chain; two down-field shifted doublets for methylene protons (δH 0.06 and 0.11) in 2, instead of characteristic singlet methyl protons of t-butyl group at δ 0.86 in 1, indicating that one of the methyl carbon directly bind to C-24 to form a cyclopropane ring. This terminal structure was secured by the HMBC cross peaks among H3-29/C-23, C-24, C-25 and C-28, as well as H3-26 and H3-27/C-24, C-25 and C-28 (Figure 2). Although connectivity between C-22 and C-23 was not confirmed by COSY or HMBC analysis because of the ambiguous cross peaks caused by overlap of proton signals, the only possible structure fulfilling the molecular formula of 2 was as that shown in Figure 2. Comparison of 1H (J value) and 13C NMR signals of 2 with those of 1 indicated that the relative configuration of the steroidal part of 2 was the same as 1, including 2β, 3α, 6α-hydroxy groups (δC 75.7, 75.7 and 78.9). The 20R* configuration in 2 was also inferred by the chemical shift and the J value (δC 0.92 d, JH=6.5 Hz) of H-20. Although configuration of the same side chain was determined for the triterpene isolated from leaves of Pandanus boninensis,12 we could only obtain the NMR data in CDCl3. Therefore, we could not compare the NMR data of 2 that was not dissolved in CDCl3, and as a result, the configuration at C-24 remains to be determined.

Figure 2
figure 2

Key HMBC cross peaks of halistanol sulfates I (2) and J (3).

In the ESIMS of halistanol sulfate J (3), three characteristic sequential ion peaks were observed at m/z 743.3 [M-Na], 623.3 [743.3-NaSO4H] and 503.3 [623.3-NaSO3H], indicating that compound 3 contains three sulfate groups as was the case for compound 2. IR absorption at 1244 and 1220 cm−1 again supported the presence of sulfate groups. The molecular formula of compound 3 was determined as C30H49Na3O12S3, which includes a CH2 unit in addition to that of 2, based on the HRESIMS analysis (m/z 743.2200 [M-Na], calcd for C30H49Na2O12S3 m/z 729.2187). 1H and 13C NMR spectra of 3 showed that it is also an analog of halistanol sulfate (1) bearing a cyclopropyl ring in the side chain (Table 1). 1H-1H COSY analysis revealed the presence of three spin systems (Figure 2). Along with HMQC and HMBC analysis, the core part (C-1 to C-23) was deduced to be identical to those of 1 and 2. Different cyclization pattern and an additional more carbon on this ring were secured by HMBC cross peaks among H-29, H-30/C-24 and H-30/C-28. Attachment of an isopropyl group (C-25 to C-27) to C-24 was also confirmed by HMBC cross peaks among H-26 and H-27/C-24. Relative configuration of the core part was again assigned to be the same as that of 1 on the basis of superposition of their NMR data. Although the side chain structure of 3 was shared by weinbersterol disulfate B13 and ibisterol sulfate,9 the configuration at C-24 and C-28 was not determined for either compounds.

Halistanol sulfate (1) and halistanol sulfates I (2) and J (3) showed no cytotoxicity against HeLa and P388 cells at the concentration of 100 μM, while they showed SIRT1–3 inhibitory activities with IC50 values of 45.9–67.9, 18.9–21.1 and 21.8–37.5 μM, respectively (Table 2). However, halistanol sulfate (1) did not show the increase of the acetylation level of eIf5A, a physiological substrate of SIRT2 in the whole-cell assay.

Table 2 SIRT1–3 inhibitory activity of compounds 13

Poly-sulfated steroids, including halistanol sulfate (1), are remarkable for their wide variety of bioactivities, such as antimicrobial, anti-HIV14, and antifouling activities,15 as well as enzyme inhibitory activities against protein kinases,6 HIV reverse transcriptase,16 HIV integrase,17 glycosidases or proteases (unpublished data). However, inhibitory activities against KDACs or SIRTs have not yet been known. This is the first report of inhibitory activity of halistanol sulfates against SIRTs.

One of the possible mechanisms for the poly-sulfated steroids to interfere with the function of enzymes have been through the detergent-like nature of these type of compounds. However, the crystal structure of SIRT3–halistanol sulfate (1) complex (Figure 3) successfully obtained by sitting drop vapor diffusion method, suggests the possibility for a different mode of action of this compound (Table 3). The conformational change of the SIRT3 structure was shown to be induced by substrate binding.18 In the crystal, two halistanol sulfate (1) molecules bind to one SIRT3 molecule (Figure 3). Two molecules of 1 are stacked through the steroid moieties on α-helix (Q300-M311) with hydrophobic surface formed by residues L303, L304, V306 and V307. Hydrogen bonding interactions are formed between sulfate ions of compound 1 and residues R224, K243, Q300 and R301. The binding sites are far from the active site H248, the substrate-binding sites or the co-factor NAD+-binding site. The crystal structure suggests similar allosteric effects on antithrombin caused by heparin. Heparin, a sulfated polysaccharide, binding to the exosite of the molecule induces conformational change allowing complex formation with factor Xa.19 Thus we found an allosteric site for SIRT3 inhibition, which is different from the resveratrol-binding allosteric sites.20 This finding reveals the possibility that polysulfated steroids may mimic the binding mode to various proteins by sulfated polysaccharides, such as heparin or haparan sulfate. Therefore, steroid sulfates can be a good motif to allosteric controller of enzymes.

Table 3 Crystallographic statistics of SIRT3–halistanol sulfates (1)–AceCS2-K(Ac) peptide)
Figure 3
figure 3

Crystal structure of SIRT3–halistanol sulfate (1) complex. (Left panel) Molecular surface of SIRT3 in complex with two halistanol sulfate (1) molecules and substrate acetyl lysine peptide. The arrow indicates the view point of the left panel. (Right panel) Close-up view of the halistanol sulfate-binding sites of SIRT3. Hydrogen bonding interactions are drawn in yellow dotted lines.

Materials and methods

General experimental procedures

NMR spectra were recorded on Avance 400 MHz spectrometers (Bruker, Billerica, MA, USA). 1H and 13C NMR chemical shifts were referenced to the solvent peaks, δH 3.31 and δC 49.15 for CD3OD. ESIMS were measured on a AccuTOF CS mass spectrometers (JEOL, Tokyo, Japan). HRESIMS were measured on a SCIENTIFIC Exactive Plus mass spectrometers (Thermo Fisher, Waltham, MA, USA). Optical rotation was determined on a JASCO DIP-1000 digital polarimeter in MeOH. UV spectra were recorded using a UV-1800 spectrophotometer (SHIMADZU Corporation, Kyoto, Japan). IR spectra were measured on a Nicolet 6700 spectrometer (Thermo Fisher, Waltham, MA, USA).

Biological material

Halichondria sp. was collected by hand using SCUBA at the depth of 15 m, Sokodo, Hachijo-jima Island (N 33° 07.16’, E 139° 49.07’), Tokyo, Japan in December 1994. The sample was immediately frozen and kept at −25 °C until extraction. The specimen was deposited in the Naturalis Biodiversity Center (RMNH POR 7354).

Isolation and identification

The sponge specimen (695 g w.w.) was extracted with MeOH (2 l × 7) and the combined extract was evaporated in vacuo. The concentrated extract was suspended in H2O and extracted with CHCl3 and then n-BuOH. The CHCl3 and n-BuOH layers were combined and subjected to the Kupchan procedure21 yielding n-hexane, CHCl3 and aqueous MeOH layers. The aqueous MeOH layer was concentrated to dryness and then separated by ODS flash chromatography (H2O/MeOH=100/0, 80/20, 50/50, 30/70 and 0/100 and CHCl3/MeOH/H2O =6/4/1) to give six fractions (fr. A–F). Inhibitory activity against SIRT3 was found in the fraction eluting with H2O/MeOH=50/50 (fr. C). A part (500 mg) of this fr. C was further separated by the reversed-phase HPLC (column: COSMOSIL 5C18-AR-II (250 × 20 mm2), flow rate; 8 ml min−1, detection; UV 220 nm and RI, solvent system: H2O/MeOH=25/75 with 200 mM NaClO4) to yield 22 fractions (fr.1–22). Of these, fr.18 contained halistanol sulfate (1, 301.4 mg, 3.5 × 10−2%, yield based on wet weight) and fr.9 contained 11.4 mg of crude halistanol sulfate I (2). Crude fr. 9 was purified by HPLC (column; COSMOSIL 5C18-AR-II (250 × 10 mm2), flow rate; 2 ml min−1, detection; UV 220 nm and RI, solvent system; H2O/MeOH=20/80 with 200 mM NaClO4) to afford pure compound 2 (6.6 mg, 9.5 × 10−3%, yield based on wet weight).

Fr.22 contained crude halistanol sulfate J (3), which was purified by three-step reversed-phase HPLC; repetitive separation under the same condition (column: COSMOSIL 5C18-AR-II (250 × 20 mm2), flow rate; 8 ml min−1, detection; UV 220 nm and RI, solvent system: H2O/MeOH=20/80 with 200 mM NaClO4), followed by final purification under the different condition (column: COSMOSIL 5C18-MS-II (250 × 10 mm2), flow rate; 2 ml min−1, detection; UV 220 nm and RI, solvent system: H2O/MeOH=25/75 with 200 mM NaClO4), yielding 0.7 mg of halistanol sulfate J (3). To gain more amount of compound 3, additional 560 mg of the fraction eluting with H2O/MeOH=50/50 (fr. C) was separated by HPLC (column: COSMOSIL 5C18-AR-II (250 × 20 mm2), flow rate; 8 ml min−1, detection; UV 220 nm and RI, solvent system: H2O/MeOH=10/90 with 200 mM NaClO4), followed by HPLC with the same column and with the different solvent system of H2O/MeOH=15/85 with 200 mM NaClO4. Final purification was carried out by two-step recycling reversed-phase HPLC (column: COSMOSIL 5C18-MS-II (250 × 10 mm2), flow rate; 2 ml min−1, detection; UV 220 nm and RI, solvent system: H2O/MeOH=10/90 with 200 mM NaClO4) to afford halistanol sulfate J (3, 3.0 mg, 5.5 × 10−3%, yield based on wet weight).

Halistanol sulfate (1)

Colorless amorphous solid; ESIMS m/z 731.3 [M-Na], 611.3 [M-Na-NaSO4H], 509.4 [M-Na-NaSO4H-NaSO3], 491.3 [M-Na-2(NaSO4H)], 354.1 [M-2Na]2−; HRESIMS m/z 731.2194 [M-Na] (calcd for C29H49Na2O12S3 m/z 731.2187); for 1H and 13C NMR data, see Table 1.

Halistanol sulfate I (2)

Colorless amorphous solid; [α]D25.4 +25.0° (c 0.01, MeOH); UV (CH3OH) λmax (logɛ) 279(2.61) nm; IR (KBr) νmax 2939, 2868, 1642, 1472, 1445, 1386, 1248, 1122, 1064, 968, 941, 930, 897, 826, 799, 754, 731, 628 cm−1; ESIMS m/z 729.2 [M-Na], 609.3 [M-Na-NaSO4H], 507.3 [M-Na-NaSO4H-NaSO3], 489.3 [M-Na-2(NaSO4H)], 353.1 [M-2Na]2−; HRESIMS analysis (m/z 729.2027 [M-Na], calcd for C29H47Na2O12S3 m/z 729.2031); for 1H and 13C NMR data, see Table 1.

Halistanol sulfate J (3)

Colorless amorphous solid; [α]D25.4 +32.0° (c 0.01, MeOH); UV (CH3OH) λmax (logɛ) 278 (2.67) nm; IR (KBr) νmax 2934, 2870, 1633, 1472, 1454, 1382, 1244, 1126, 1972, 969, 930, 893, 844, 826, 799, 749, 732, 624 cm−1; ESIMS m/z 743.3 [M-Na], 623.3 [M-Na-NaSO4H], 521.4 [M-Na-NaSO4H-NaSO3], 503.3 [M-Na-2(NaSO4H)], 360.1 [M-2Na]2−; HRESIMS analysis (m/z 743.2200 [M-Na], calcd for C30H49Na2O12S3 m/z 729.2187); for 1H and 13C NMR data, see Table 1.

Cytotoxicity test

HeLa cells in Dulbecco’s modified Eagle’s medium (low glucose, Wako, Osaka, Japan), containing 10% of fetal bovine serum (Biosera, Ringmer, UK), 2 μg ml−1 of gentamicin reagent solution, 10 μg ml−1 of antibiotic–antimycotic were cultured at 37 °C under an atmosphere of 5% CO2 with various concentrations of halistanol sulfates (1, 2 and 3). P388 cells in RPMI-1640 medium (Wako), containing 10% of fetal bovine serum, HRDS solution (2, 2’-dithiobisethanol) and kanamycin sulfate were cultured at 37 °C under an atmosphere of 5% CO2 with various concentrations of 1, 2 and 3. After 72 h cultivation, to each cell was added to 3-(4,5-dimethyl-2-thiazoyl)-2,5-diphenyl-2H tetrazolium bromide (MTT) saline solution and incubated. After 4 h, medium was removed by aspiration and dimethyl sulfoxide was added to lyse cells. Concentration of the reduced MTT was quantified by measuring the absorbance at 650 nm to estimate IC50 values.

SIRT inhibition assays

In vitro SIRT1–3 inhibitory tests were performed by an electrophoretic mobility shift assay as described previously.22 Briefly, recombinant SIRT proteins were incubated with a carboxyfluorescein (FAM)-labeled acetylated lysine peptide (1.5 μM FAM-RHKK(Ac)LM for SIRT1 and SIRT2; 1.5 μM FAM-QPKK(Ac)KPL for SIRT3) and 1 mM NAD in 50 μl of assay buffer (25 mM Tris-HCl (pH 9.0), 137 mM NaCl, 2.7 mM KCl, 1 mM MgCl2, 0.1 mg ml−1 bovine serum albumin) in 384-well plates for 60 min at 37 °C. After the reaction was stopped by adding in 50 μl of the stop buffer (100 mM HEPES (pH 7.5), 10 mM EDTA, 0.25% CR-3) containing nicotinamide (final concentration 10 mM), the samples were analyzed using a LabChip EZ Reader II (PerkinElmer, Waltham, MA, USA). Percentge of conversion is defined as 100 × P/(P+S), where P and S are peak heights of product and peptide substrate, respectively.

Crystal structure of SIRT3 in complex with halistanol sulfate (1)

SIRT3 (residues:118–399, final concentration of 5 mg ml−1), an acetylated lysine peptide (TRSGK(Ac)VMRRLLR: AceCS2-K(Ac), 1 mM) and halistanol sulfate (1, 1 mM) were mixed and crystallized. Crystals were obtained in 1.2 M sodium phosphate monobasic and potassium phosphate dibasic buffer, pH 5.4, at 20 °C, and frozen with liquid nitrogen, using 18% glycerol as a cryoprotectant. X-ray diffraction data were collected at 100 K in a nitrogen gas stream at the synchrotron beamlines, PF-AR NW12A and PF-17 A at Photon Factory, KEK. Data were processed and scaled with the HKL2000 program.23 The crystal structures were determined by the molecular replacement method with MOLREP,24 using the structure of SIRT3/AceCS2-K(Ac) complex18 (PDB id 3GLU). Refinement and model building were performed with REFMAC525 and Coot.26 The geometric quality of the model was assessed with MolProbity.27 Data collection and refinement statistics are listed in Table 3. Structural illustrations were generated using PyMol (Schrödinger).

Dedication

This work is dedicated to Professor K. C. Nicolaou celebrating his great scientific contribution to total synthesis of 27 highly complex and biologically important natural products.