Poecillastrosides, Steroidal Saponins from the Mediterranean Deep-Sea Sponge Poecillastra compressa (Bowerbank, 1866)

The first chemical investigation of the Mediterranean deep-sea sponge Poecillastra compressa (Bowerbank, 1866) led to the identification of seven new steroidal saponins named poecillastrosides A–G (1–7). All saponins feature an oxidized methyl at C-18 into a primary alcohol or a carboxylic acid. While poecillastrosides A–D (1–4) all contain an exo double bond at C-24 of the side-chain and two osidic residues connected at O-2′, poecillastrosides E–G (5–7) are characterized by a cyclopropane on the side-chain and a connection at O-3′ between both sugar units. The chemical structures were elucidated through extensive spectroscopic analysis (High-Resolution Mass Spectrometry (HRESIMS), 1D and 2D NMR) and the absolute configurations of the sugar residues were assigned after acidic hydrolysis and cysteine derivatization followed by LC-HRMS analyses. Poecillastrosides D and E, bearing a carboxylic acid at C-18, were shown to exhibit antifungal activity against Aspergillus fumigatus.

In our continuous efforts to describe the chemical diversity of marine sponges from the Mediterranean, we undertook the first chemical study of the deep-sea Tetractinellid sponge Poecillastra compressa (Bowerbank, 1866). The genus Poecillastra is known to produce a broad range of secondary metabolites such as macrolactams [38,39], nitrosohydroxyalkylamines [40], sesquiterpenes, and steroids [41,42]. We report herein the isolation and structure elucidation of seven new steroidal glycosides named poecillastrosides A-G (1-7) from the deep-sea sponge P. compressa ( Figure 1). Their structures were deduced from spectroscopic data including 1D-and 2D-NMR experiments as well as high-resolution mass spectra (HRESIMS) analyses. Three different aglycone moieties were identified, and oxidation at the C-18 position is a common feature among all isolated saponins. Poecillastroside A (1) contains an ergostane aglycone, whereas poecillastrosides B-D (2-4) contain a poriferastane, and poecillastrosides E-G (5-7) a cholestane with a cyclopropyl ring on the side-chain.

Results and Discussion
The freeze-dried sponge sample (43.1 g) was macerated and repeatedly extracted with a mixture of CH2Cl2/CH3OH (1:1) under sonication. The extract (7.9 g) was fractionated by Reversed Phase C18 Vacuum Liquid Chromatography with solvent mixtures of decreasing polarity. The methanolic fraction was then purified by successive RP-Phenylhexyl and C18 HPLC yielding pure compounds 1-7.

Results and Discussion
The freeze-dried sponge sample (43.1 g) was macerated and repeatedly extracted with a mixture of CH 2 Cl 2 /CH 3 OH (1:1) under sonication. The extract (7.9 g) was fractionated by Reversed Phase C18 Vacuum Liquid Chromatography with solvent mixtures of decreasing polarity. The methanolic fraction was then purified by successive RP-Phenylhexyl and C18 HPLC yielding pure compounds 1-7.
Compound 1 was isolated as a yellowish amorphous solid. Its molecular formula C 40 H 68 O 13 was determined by HRESIMS. The 1 H NMR spectrum of 1 suggested a steroidal saponin (Table 1). First, the characteristic anomeric signals at δ H 4.49 (d, J = 7.6 Hz, 1H, H-1 ), 4.56 (d, J = 7.9 Hz, 1H, H-1"), and δ C 101.8 (C-1 ), 105.2 (C-1") evidenced the presence of two sugar residues. The 1 H NMR data of the steroid revealed one methyl singlet at δ H 0.88 (s, 3H, H 3 -19), three methyl doublets at δ H 1.02 (d, J = 6.8 Hz, 3H, H 3 -21) and 1.03 (d, J = 6.8 Hz, 6H, H 3 -26 and -27), ten methylene groups, an oxygenated methylene with the AB system at δ H 3.95 and 3.59, a 1,1-disubstituted olefin at δ H 4.70 and 4.71 (H 2 -24 1 ), seven methine groups, two oxygenated methines at δ H 3.72 (m, 1H, H-3), 4.26 (td, J = 7.7, 3.7 Hz, 1H, H-16), and three quaternary carbons at C-10, C-13 and C-24. When compared to usual steroids, this aglycone lacks one characteristic methyl signal for C-18. A hydroxylation was proposed at this position based on the presence of an AB system at δ H 3.59 (d, J = 11.5 Hz, 1H, H-18b) and 3.95 (d, J = 11.5 Hz, 1H, H-18a) and further key H-12b, H-14, H-17/C-18, and H 2 -18/C-13, C-14, C-17 HMBC correlations. Another unusual feature for the steroid moiety was evidenced in the HSQC spectrum with signals of an oxygenated methine at δ H 4.26 (td, J = 7.7, 3.7 Hz, 1H, H-16) and δ C 72.8 (CH, C-16). The location of this hydroxyl group at C-16 was confirmed after interpretation of key H-16/H-17 and H-16/H-15a COSY and TOCSY correlations. While most of the relative configurations were in accordance with a common steroid core, the relative configuration at C-16 was established after examination of the NOESY spectrum. Absence of clear nuclear Overhauser effect (nOe) between H-16 and H-14 but also H-18 together with some overlap between H-17 and H-22 did not allow a straightforward determination of the relative configuration at this position. However, H-16/H-15a and H-8/H-15b nOes suggested a β orientation for the hydroxyl group at C-16. As a confirmation of this orientation, the coupling constant values of H-16 were in perfect accordance with those observed for the same signal of a closely related analogue weinbergsterol B, isolated from the sponge Petrosia weinbergi [43]. NMR signals of the sugar residues were assigned by extensive COSY, TOCSY, and HSQC interpretation. HMBC experiment evidenced H-5 /C-1 , H-1"/C-2 , H-5"/C-1" long-range correlations, thus revealing the pyranose nature of these two sugars and their connection at C-2 . Finally, the connectivity of the sugar with the aglycone at C-3 was confirmed through the key HMBC H-1 /C-3 correlation. Moving to the relative configuration of the residues, the large coupling constants between H-1 /H-2 and H-1"/H-2" (7.9 and 7.6 Hz, respectively) were consistent with a β configuration for both anomeric centers. This interpretation was confirmed with the one-bond coupling constant 1 J CH ≈ 160 Hz for the two anomeric positions [44]. In addition, the coupling constant values of 3 J H3 -H4 3.2 Hz and 3 J H5 -H4 close to zero suggested an axial position for the hydroxyl at C-4 and, therefore, a β-galactopyranosyl residue attached at C-3 of the aglycone [45]. For the second sugar residue, all coupling constants were measured with values between 7 and 9 Hz which implies equatorial positions for all oxygen atoms and, therefore, a β-glucopyranosyl residue connected at C-2 of the first residue.
Assuming a usual absolute configuration for the aglycone, we turned towards the pyranose moieties. After hydrolysis of the acetal bonds, the resulting monosaccharides were derivatized with L-cysteine methyl ester and phenylisothiocyanate in pyridine [46]. By comparison with standards, a D absolute configuration was assigned for both glucose and galactose monosaccharides.
Compound 2 was isolated as a yellowish amorphous solid. The molecular formula of 2 was determined by HRESIMS as C 41 H 70 O 13 . The spectroscopic data were very similar to those of 1, thereby suggesting that both compounds were close analogues. Examination of the 1 H NMR spectrum revealed the presence of an additional methyl group at δ H 1.59 (d, J = 6.3 Hz, 3H, H 3 -24 2 ) placed on the double bond at C-24 1 , therefore, leading to a poriferastane skeleton. The relative configuration of 2 was found to be the same as that of poecillastroside A based on nOe correlations. A key H 3 -24 2 /H 2 -23 nOe led us to assign the configuration of the double bond as E.
Compound 3 was isolated as a pale yellowish amorphous solid with the same molecular formula C 41 H 70 O 13 . Both compounds 2 and 3 are, therefore, isomers. The 1 H NMR spectra were almost identical except for a deshielding of the signal corresponding to H-25, from δ H 2.24 in 2 to δ H 2.85 for 3. We first supposed that a change in the configuration of the double had occurred. Due to the low amount of compound available, the corresponding carbons were not visible neither in the 13 C NMR spectrum nor in the HSQC, HMBC spectra. We, therefore, decided to enhance the sensitivity of the HSQC spectrum using the recently developed Pure Shift HSQC experiment [47]. Gratifyingly, we were then able to observe both HSQC spots corresponding to C-24 1 and C-25 ( Figure S24). The shielding of the C-25 signal from δ C 36.0 for 2 to δ C 29.8 for 3 clearly confirmed a Z configuration for the double bond of 3.  H-12a). The only explanation consistent with all these observations, including the molecular formula, was the replacement of the hydroxyl group at C-18 by a carboxylic acid. This interpretation was further supported by a key H-17/C-18 HMBC correlation. Based on the chemical shift of the signal H-25 the configuration of the double bond was found to be the same as in 2.
Compound 5 was isolated as a white amorphous solid with a molecular formula of C 43 H 66 O 15 . Despite strong differences when compared with 1-4, the NMR data of 5 evidenced that the molecule was a steroidal saponin ( Table 2). The aglycone exhibited an unusual skeleton with the presence of a terminal methylated cyclopropyl ring on the lateral chain. This assumption was based on the shielded signals of H-25 and H-26 but also by COSY, HSQC, and HMBC data analyses with the key H-27/C-24, H-27/C-26 HMBC correlations. Further analysis of 1 H NMR data revealed the E geometry of the olefinic bond (J H-22,-23 = 15.2 Hz). No clear nOe correlations were observed for assessing the relative configuration around the cyclopropane ring. Gratifyingly, comparison with literature data and synthetic analogues of sterols with an identical side-chain led us to propose a trans configuration for the substituents at C-24 and C-25 of this ring [48][49][50][51]. To confirm this configuration in our case, we decided to look further into the coupling constants of the signals corresponding to the cyclopropane protons. Only the signals of the methylene and their multiplicity were clearly identified in the 1 H NMR spectrum (Figure 2). In the case of a trans configuration of the two substituents around the cyclopropane, H a and H b would have the same splitting pattern as they would have in the presence of a pseudo C2 axial symmetry perpendicular to the cyclopropane plane. The 3 J coupling constants between protons in a cis configuration are known to be between 8 and 10 Hz while values below 7 Hz are always observed when placed in a trans configuration. The multiplicity for both signals is observed as a doublet or triplet with coupling constants around 8 and 4 Hz, respectively. This same splitting pattern for both signals is only consistent for a trans configuration. Indeed, for a cis configuration, one of the two gem protons H b would exhibit two large 3 J coupling constants of 8 Hz. We, therefore, confirm a trans configuration for the two substituents and estimate the gem 2 J coupling constants between H a and H b to be around 4 Hz. The presence of a carboxyl group at C-18 was inferred first from the HRESIMS data and then from the deshielding of H-12a, exactly in the same manner as for compound 4. Another difference with 4 arose from the absence of the signal corresponding to the oxygenated methine at C-16. This feature was confirmed by COSY, HSQC, and HMBC correlations. Looking at the glycosidic part of the saponin, the relative configuration was similar to those of 1-4, therefore, confirming one galactose linked to the aglycone and one glucose linked to the galactose. HMBC showed long-range correlations between H-1"/C-3 , H-2 /C Ac (δ C 172.2), and H-6"/C Ac (δ C 172.8), thereby indicating the presence of two acetyl groups at C-2 and C-6". Unlike compounds 1-4, the glycosidic link between both sugar residues was placed at C-3 of the galactose. Deshielding of the signal of C-3 at δ C 82.4 in the 13 C NMR spectrum confirmed this new substitution pattern. was a steroidal saponin ( Table 2). The aglycone exhibited an unusual skeleton with the presence of a terminal methylated cyclopropyl ring on the lateral chain. This assumption was based on the shielded signals of H-25 and H-26 but also by COSY, HSQC, and HMBC data analyses with the key H-27/C-24, H-27/C-26 HMBC correlations. Further analysis of 1 H NMR data revealed the E geometry of the olefinic bond (JH-22,-23 = 15.2 Hz). No clear nOe correlations were observed for assessing the relative configuration around the cyclopropane ring. Gratifyingly, comparison with literature data and synthetic analogues of sterols with an identical side-chain led us to propose a trans configuration for the substituents at C-24 and C-25 of this ring [48][49][50][51]. To confirm this configuration in our case, we decided to look further into the coupling constants of the signals corresponding to the cyclopropane protons. Only the signals of the methylene and their multiplicity were clearly identified in the 1 H NMR spectrum (Figure 2). In the case of a trans configuration of the two substituents around the cyclopropane, Ha and Hb would have the same splitting pattern as they would have in the presence of a pseudo C2 axial symmetry perpendicular to the cyclopropane plane. The 3 J coupling constants between protons in a cis configuration are known to be between 8 and 10 Hz while values below 7 Hz are always observed when placed in a trans configuration. The multiplicity for both signals is observed as a doublet or triplet with coupling constants around 8 and 4 Hz, respectively. This same splitting pattern for both signals is only consistent for a trans configuration. Indeed, for a cis configuration, one of the two gem protons Hb would exhibit two large 3 J coupling constants of 8 Hz. We, therefore, confirm a trans configuration for the two substituents and estimate the gem 2 J coupling constants between Ha and Hb to be around 4 Hz. The presence of a carboxyl group at C-18 was inferred first from the HRESIMS data and then from the deshielding of H-12a, exactly in the same manner as for compound 4. Another difference with 4 arose from the absence of the signal corresponding to the oxygenated methine at C-16. This feature was confirmed by COSY, HSQC, and HMBC correlations. Looking at the glycosidic part of the saponin, the relative configuration was similar to those of 1-4, therefore, confirming one galactose linked to the aglycone and one glucose linked to the galactose. HMBC showed long-range correlations between H-1′′/C-3′, H-2′/CAc (δC 172.2), and H-6′′/CAc (δC 172.8), thereby indicating the presence of two acetyl groups at C-2′ and C-6′′. Unlike compounds 1-4, the glycosidic link between both sugar residues was placed at C-3′ of the galactose. Deshielding of the signal of C-3′ at δC 82.4 in the 13 C NMR spectrum confirmed this new substitution pattern.    13 . The spectroscopic data were very similar to those of 5, thereby suggesting a close aglycone moiety. However, some changes were noticed by HSQC and HMBC analyses. Indeed, in the aglycone moiety, we observed the same AB system for H 2 -18 as that present in compounds 1-3. The long-range H-17/C-18 HMBC correlation confirmed the presence of an oxygenated methylene at C-13. In the D-β-glucose residue, the chemical shifts, and the COSY data were consistent with a terminal primary alcohol at C-6", thereby implying the loss of the acetate at this position.
Compound 7 was isolated as a white amorphous solid with a molecular formula C 43 H 68 O 14 . The 1 H NMR spectrum evidenced the fact that 7 is a close analogue of 6. The long-range H-6"/C Ac (δ C 172.8) HMBC correlation revealed the presence of an acetate group linked at O-6" as in compound 5. The relative configuration of 7 was the same as those of 5 and 6.
Poecillastrosides A-G were tested in a panel of antimicrobial and cytotoxicity assays, including antibacterial activity against Gram positive (methicillin resistant (MRSA) and methicillin sensitive (MSSA) Staphylococcus aureus), and Gram negative bacteria (Escherichia coli, Klebsiella pneumoniae, Pseudomonas aeruginosa, and Acinetobacter baumannii), antifungal activity against Aspergillus fumigatus, and cytotoxicity against the hepatic tumoral cell line hep_G2. Poecillastrosides D (4) (MIC 90 = 6 µg/mL) and E (5) (MIC 90 = 24 µg/mL) were the only two molecules active in the assay against A. fumigatus, revealing a key role of the carboxylic acid functionality at C-18 in the antifungal activity of this structural class. On the other hand, cytotoxicity assays also revealed weak activity of some members of the family against the hep_G2 human cell line, with IC 50 values of 38, 28, and 89 µg/mL for poecillastrosides B, C, and D (2-4), respectively. None of the compounds of this family displayed activity against the bacterial pathogens at the highest concentration tested (96 µg/mL for compound 1-5, and 64 µg/mL for compounds 6 and 7).

General Experimental Procedures
Optical rotations were recorded with a PerkinElmer 343 polarimeter equipped with a 10 cm microcell and a sodium lamp. UV measurements were obtained by extraction of the Diode Array Detector (DAD) signal of the Ultra-High Pressure Liquid Chromatography (UHPLC) Dionex Ultimate 3000 (Thermo Scientific, Waltham, MA, USA). NMR experiments were performed on a 500 MHz (Advance, Bruker, Billerica, MA, USA) or a 600 MHz (Agilent, Santa Clara, CA, USA) spectrometer. Chemical shifts (δ in ppm) are referenced to the carbon (δ C 49.0) and residual proton (δ H 3.31) signals of CD 3 OD. High-resolution mass spectra (HRESIMS) were obtained from a mass spectrometer Agilent 6540. HPLC separation and purification were carried out on a Jasco LC-2000 series equipped with a UV detector coupled with an Evaporative Light Scattering Detector, ELSD (Sedere, Alfortville, France).

Biological Material
Poecillastra compressa (Bowerbank, 1866) was collected in the Mediterranean Sea, off the French coasts, on 15 October 2014 at 200 m depth using a Remotely Operated Vehicle (Super Achille, COMEX S.A., Marseille, France). The voucher specimen "CS2ACHP09_ECH04" is kept at the Marine Station of Endoume (OSU Institut Pythéas, Marseille, France).