The sponge specimen was lyophilized, ground, and repeatedly extracted with a 1:1 CH₂Cl₂/MeOH mixture. The combined extracts were filtered, concentrated under vacuum and fractionated by reversed phase column liquid chromatography with solvents of decreasing polarities (H₂O, MeOH and CH₂Cl₂). The methanol fraction was further subjected to repeated HPLC purifications yielding the new ulososide F (1) and urabosides A (2) and B (3), together with the known compound uloloside A (4) (Figure 1). These compounds were identified by combined spectroscopic methods and comparisons of NMR data with literature [10,19,20,21].

Compound 1 was isolated as a white amorphous solid, with a molecular formula assigned as C₄₄H₇₁NO₁₆ on the basis of combined HRESIMS (m/z 892.46222 [M+Na]⁺) and ¹³C-NMR data. The saponin nature of this compound was evidenced in the ¹H-NMR spectrum of 1 by the presence of characteristic terpenoid methyl singlets (δ_H between 0.6 and 1.5) together with sugar residues (δ_H between 3.1 and 5.0). The portion of the ¹H-NMR spectrum corresponding to the aglycon evidenced seven methyl signals that were reminiscent of a 29-functionalized 24-methyl-30-norlanostane at δ_H 0.67 (s, H₃-18), 0.85 (d, H₃-24¹), 0.93 (d, H₃-21), 0.95 (d, H₃-26), 0.96 (d, H₃-27), 1.01 (s, H₃-19), 1.34 (s, H₃-28) (Table 1) [22]. The presence of a fused double bond at C-8/C-9 was mainly inferred from the key H₃-19/C-9 HMBC correlation. Two alcohols and a methyl were placed at C-22, C-23 and C-24 of the side-chain respectively after inspection of the COSY spectrum and characteristic signals at δ_H 3.51 (m, H-22), 3.68 (dd, H-23), 1.56 (m, H-24) and δ_C 73.4 (CH, C-22), 72.2 (CH, C-23), 41.0 (CH, C-24). The structure was then very similar to the previously isolated ulososides isolated from an Ulosa sp. sponge collected in Madagascar [19,20,21]. In fact, 1 shares exactly the same aglycon as ulososides A, C, D and E due to very close chemical shifts, even if all NMR spectra were reported in C₅D₅N in the literature. We just changed the numbering replacing the C-28 methyl described in the literature by a C-24¹ methyl at C-24, thus respecting the IUPAC recommendations for this family of triterpenoids. The relative configuration of the tetracycle was confirmed by NOESY correlations and ³J scalar coupling interpretation [23]. We decided to assess the relative configuration of the asymmetric centers at C-22, C-23 and C-24 of the side-chain using a close inspection of the NMR data reported for the closely related and well known brassinosteroids [23,24,25].

It appears that ¹H and ¹³C-NMR data are fully consistent with a 22R, 23S and 24R configurations assuming a usual absolute configuration at C-20 (Figure 2). Additionally, we then propose to change the configurations for ulososides A, C, D and E that were tentatively assigned as 22S, 23R and 24S.

NMR spectra indicated the disaccharide nature of compound 1. For the glycosidic part of the molecule, two sugar residues were evidenced by the signals corresponding to two anomeric carbons at δ_C 104.8 (CH, C-1′), and 105.1 (CH, C-1′′) and protons at δ_H 4.56 (d, J = 8.1 Hz, H-1′) and 4.49 (d, J = 7.8 Hz, H-1′′), with coupling constants at J = 8.1 and 7.8 Hz suggesting a β anomeric configuration (Table 2). The first sugar residue was identified as a N-acetylglucosamine on the basis of the ¹³C chemical shifts and coupling constant values, as well as a comparison with ulososide D-NMR data [19]. A H-1′/C-3 HMBC correlation placed this sugar at C-3. Compared to ulososide D an additional glucuronic acid was branched at C-6′ on the basis of a H-1′′/C-6′ HMBC correlation. The glucuronic acid was ascertained by interpretation of coupling constant values and the presence of a H-5′′/C-6′′ HMBC correlation. The D absolute configurations of both sugar residues were determined by GC after derivatization and comparison with standard sugars [26].

Compound 2 was isolated as a white, amorphous solid with a molecular formula of C₄₆H₇₄O₁₉ on the basis of the pseudomolecular ion at m/z 953.47165 [M + Na]⁺ by HRESIMS analysis. The aglycon of 2 differed from 1 in two main sites. First, changes in the ¹H and ¹³C-NMR chemical shifts of the signals corresponding to C-3, C-4 and C-5 indicated a change in the substitution pattern at C-4. In the ¹H-NMR spectrum, the methyl at C-28 for 1 was replaced by an AB system in 2 at δ_H 3.96 (d, J = 11.0 Hz, H-28a) and 3.89 (d, J = 11.0 Hz, H-28b) which was reminiscent of a methylene oxy at C-4 (Table 1). This assumption was confirmed by comparison with the corresponding signals found for ulososide B [20]. Comparison of the chemical shifts between both compounds and a clear H-29/H-3 NOESY correlation allowed us to assign the relative configuration at C-3 and C-4. The second modification was observed for the signals corresponding to the side-chain protons and carbons. Indeed, the signals corresponding to the vicinal diol of 1 were clearly absent as well as the one corresponding to the methyl at C-24. In contrast, a ketone was unambiguously placed at C-23 due to the ¹³C-NMR signal at δ_C 213.8 (C, C-23) and the key H-22a, H-22b, H-24/C-23 HMBC correlations. Interestingly, this side-chain was identical to the one of some pandarosides, steroidal saponins found in another Poecilosclerida sponge of the genus Pandaros [18,27].

The osidic part of this saponin was shown to contain three sugar residues as evidenced by the signals of three anomeric proton doublets at δ_H 4.56 (d, H-1′), 5.59 (d, H-1′′) and 4.50 (d, H-1′′′) and the corresponding anomeric carbons at δ_C 105.3 (CH, C-1′), 100.2 (CH, C-1′′) and 106.1 (CH, C-1′′′) observed in the HSQC spectrum. The three sugar residues were confidently located at C-3, C-2′ and C-3′ on the basis of H-1′/C-3, H-1′′/C-2′ and H-1′′′/C-3′ HMBC correlations. Two hexoses were deduced from the signals at δ_H 3.73 (m, H-6′), 3.70 (m, H-6′′′) and δ_C 62.6 (CH₂, C-6′ and C-6′′′). The third sugar residue was found to be a pentose due to the presence of the signals at δ_H 4.07 (d, H-5′′a), 3.42 (m, H-5′′b) and δ_C 64.9 (CH₂, C-5′′).

The ring sizes and identity of these sugar residues were determined by 2D-NMR experiments and by comparison of chemical shifts of protons and carbons of each monosaccharide with literature citations [28,29]. After the assignment of the individual NMR signals for these sugar units by extensive COSY, TOCSY, and HSQC analyses, HMBC experiments showed long-range correlations at H-1′/C-5′, H-5′/C-1′, H-5b′′/C-1′′, H-1′′′/C-5′′′, and H-5′′′/C-1′′′, thus revealing the pyranose nature of all three residues. The large coupling constant values for the doublets assigned to H-1′ and H-1′′′ (J = 7.9 and 7.8 Hz) implied that these hexoses were connected through β-glycosidic linkages, while the smaller coupling constant value for H-1′′ (J = 3.2 Hz) suggests also a β linkage for this arabinose residue. The absolute configurations of these three residues was ascertained after acidic hydrolysis, derivatization into chiral butylated derivatives and GC analysis which indicated a D-galactose at C-3, a D-arabinose at C-2′ and a second D-galactose at C-3′.

Compound 3 was obtained as a white, amorphous powder. The positive ion at m/z 849.38367 [M+Na]⁺ in HRESIMS and ¹³C-NMR data indicated an empirical molecular formula of C₄₁H₆₂O₁₇. The signals corresponding in the NMR spectra of 3 were closely related with the signals of 2 indicating a close relationship between these two compounds. The sole difference appears for the signals corresponding to the positions C-3, C-4 and C-5 (Table 1). The absence of the AB system assigned to the primary alcohol of 1 at C-4 suggested a diverse functionalization at this position. Even if the corresponding signal was not observed in the ¹³C neither in the HMBC NMR spectra, the molecular formula was only consistent with a presence of a second carboxylic acid at this position. This is the first example of such a substitution pattern at C-4 for a terpenoid and then no comparison with literature data was available. Nevertheless, NMR modeling with ChemDraw was fully consistent with the observed chemical shifts in ¹H-NMR and more importantly in ¹³C-NMR, especially for δ_C 63.8 (C, C-4), thus confirming this assumption.

Two hexose residues were observed in 3 which were identified as a first glucuronic acid linked at C-3 to the aglycon due to the presence of a H-1′/C-3 HMBC correlation. Coupling constant values and a clear H-5′/C-6′ HMBC correlation confirmed the β-glucuronic acid. The second sugar residue was linked at C-2′ to this first residue on the basis of a H-1′′/C-2′ HMBC correlation. This second residue was found to be a galactose unit due to the coupling constant values of the signal at δ_H 3.76 (d, J = 3.0 Hz, H-4′′). Furthermore, the value of J = 8.1 Hz for the coupling constant of H-1′′ ascertained a β linkage with the first residue. Using the same process used for 2 we confirmed the relative configurations and proposed a D absolute configuration for both residues.

No significant cytotoxicity against two cell lines (Jurkat and CHO cells) or hemolytic activity was detected below 50 µM for any of the isolated compounds. Only compound 3 exhibited a low cytotoxicity, with an IC₅₀ value of 100 µM. Other compounds were considered inactive, with IC₅₀ > 100 µM, and with cytotoxicities below 30% up to 100 µM. Interestingly, compounds 1 to 4 did not induce hemolysis, which is the major known adverse effect of this kind of molecules and the major obstacle for clinical trials progress. Previous studies on the bioactivities of triterpene glycosides from E. ferox reported moderate cytotoxicity against J774 (murine monocyte-macrophage), WEHI164 (murine fibrosarcoma), and P388 (murine leukemia) cell lines at IC₅₀ ranging from 8.5 to 19 μg/mL [9,10]. The biological activities of our new derivatives are significantly lower than those exhibited by the previously isolated ectyoplasides and feroxosides and consequently it suggests a key role of the aglycone in the cytotoxicity of these compounds. The assessment of detergent-like properties by the hemolysis assay demonstrated that there is no significant effect of our compounds.

Optical rotations were measured on Perkin Elmer 343 polarimeter equipped with a 10 cm microcell. IR spectra were obtained with a Perkin–Elmer Paragon 1000 FT–IR spectrometer. High resolution mass spectra (HRESIMS) were obtained from a LTQ Orbitrap mass spectrometer (Thermo Finnigan). NMR experiments were performed on a Bruker Advance 500 MHz spectrometer. Chemical shifts (δ in ppm) are referenced to the carbon (δ_C 49.0) and residual proton (δ_H 3.31) signals of CD₃OD, the solvent with multiplicity (s singlet, d doublet, t triplet, m multiplet). Column chromatography was performed using RP18 stationary phase (40–63 μm, Merck). HPLC separation and purification were carried out on an Agilent1100 system equipped with an Agilent UV detector and coupled with a Varian 385-ELSD. For GC-MS analysis of the derivatized sugar residues an Agilent 6890N GC interfaced to a 5975B MSD was used. TLC was performed with Kieselgel 60 F₂₅₄ (Merck glass support plates) and spots were detected after spraying with 10% H₂SO₄ in EtOH reagent and heating. The OD₅₇₀ nm absorbance for cytotoxicity evaluation was measured with a Thermo Scientific Multiskan® Spectrum instrument.

The marine sponge was collected off Caribbean Sea, Colombia, in October 2008 by SCUBA diving (Urabá Gulf 8°40′14′′N, 77°21′28′′W). A voucher specimen (INV-POR 0335) identified by Sven Zea, has been deposited in the sponge collection of Museo de Historia Natural Marina de Colombia, Invemar. The sponge was kept frozen from collection until the extraction process.

The frozen sponge (350 g wet) was cut into pieces of about 1 cm³ and extracted with 1:1 MeOH/CH₂Cl₂ (600 mL, 24 h) at room temperature yielding 3.6 g of crude extract after solvent evaporation. The crude extract was fractionated by RP-C₁₈ column chromatography (elution with a decreasing polarity gradient of H₂O/MeOH from 70:30 to 0:100, then MeOH/CH₂Cl₂ from 100:0 to 0:100). The third fraction (175.5 mg) from 10 collected (MeOH 100%) was then subjected to RP semi-preparative HPLC (Phenomenex, Gemini C₆-phenylhexyl 110 Å, 250 × 10 mm, 5 μm) with a gradient of H₂O/acetonitrile/TFA (flow 3.0 mL·min⁻¹ from 60:40:0.1 to 45:55:0.1) to yield pure compound 2 (tr: 23.1 min; 1.4 mg); whereas the second fraction (122.5 mg) was purified with an isocratic mobile phase of H₂O/Acetonitrile/TFA (flow 3.0 mL·min⁻¹, 55:45:0.1) to afford pure metabolites 1, 4 and 3 (tr: 15.2, 16.8 and 24.1 min; 4.5, 3.6 and 2.9 mg, respectively).

(22R,23S,24R)-3β-O-(β-D-Glucopyranosyluronic acid-(1→6)-2-acetamido-2-deoxy-β-D-gluco-pyranosido)-3,22,23-trihydroxy-24-methyl-30-norlanost-8(9)-en-29-oic acid (Ulososide F, 1). White amorphous solid; [α]²⁰_D −4.0 (c 0.036, MeOH); IR (thin film): γ_max 3392, 2956, 2874, 1721, 1669 cm⁻¹; ¹H-NMR and ¹³C-NMR see Table 1, Table 2; HRESIMS (+): m/z 892.46222 [M+Na]⁺ (calcd for C₄₄H₇₁NNaO₁₆, 892.46651, Δ −4.8 ppm).

3β-O-[β-D-Arabinopyranosyl-(1→2)-(β-D-galactopyranosyl-(1→3))-β-D-galactopyranosid]-3,28-dihydroxy-23-oxolanost-8(9)-en-29-oic acid (Uraboside A, 2). White amorphous solid; [α]²⁰_D −330 (c 0.013, MeOH); IR (thin film): γ_max3392, 2988,1625, 1072 cm⁻¹; ¹H-NMR and ¹³C-NMR see Table 1, Table 2; HRESIMS (+): m/z 953.47327 [M+Na]⁺ (calcd for C₄₆H₇₄NaO₁₉, 953.47165, Δ −1.7 ppm).

3β-O-[β-D-Galactopyranosyl-(1→2)-β-D-glucopyranosiduronic acid]-3-hydroxy-23-oxolanosta-8(9)-ene-28,29-dioic acid (Uraboside B, 3). White amorphous solid; [α]²⁰_D−130 (c 0.027, MeOH); IR (thin film): γ_max 3409, 2920, 1712, 1674 cm⁻¹; ¹H-NMR and ¹³C-NMR see Table 1, Table 2; HRESIMS (+): m/z 849.38367 [M+Na]⁺ (calcd for C₄₁H₆₂O₁₇Na, 849.38792, Δ −5.0 ppm).

(22R,23S,24R)-3β-O-(β-D-Glucuronopyranosyl-(1→6)-β-D-glucopyranosyl)-3,22,23-trihydroxy-24-methyl-30-norlanost-8(9)-en-29-oic acid (Ulososide A, 4). White amorphous solid; [α]²⁰_D −9.5 (c 0.23, MeOH); IR (thin film): γ_max 3425, 1715, 1648 cm⁻¹; ¹H-NMR (500 MHz, CD₃OD) 1.84 (m), 1.36 (m), 2.41 (m), 2.03 (m), 3.30 (m), 1.34 (m), 1.97 (m), 1.71 (m), 2.07 (m), 1.93 (m), 2.17 (m), 2.08 (m), 1.99 (m), 1.46 (m), 2.11 (m), 1.64 (m), 1.34 (m), 1.99 (m), 1.44 (m), 1.57 (m), 0.67 (s), 1.01 (s), 1.90 (m), 0.93 (d, 6.9 Hz), 3.51 (m), 3.66 (dd, 1.6 Hz, 9.7 Hz), 1.55 (m), 1.62 (m), 0.97 (d, 6.6 Hz), 0.95 (d, 6.6 Hz), 0.85 (d, 6.9 Hz), 1.44 (s), 4.32 (d, 7.8 Hz), 3.25 (m), 3.47 (m), 3.27 (m), 3.53 (m), 4.08 (s, b), 3.77 (m), 4.47 (d, 7.8 Hz), 3.25 (m), 3.39 (m), 3.51 (m), 3.81 (d, 9.6 Hz); ¹³C-NMR (125 MHz, CD₃OD) 37.7 (C-1), 28.5 (C-2), 89.4 (C-3), 50.7 (C-4), 53.6 (C-5), 21.3 (C-6), 29.9 (C-7), 129.3 (C-8), 137.1 (C-9), 23.2 (C-11), 38.5 (C-12), 43.3 (C-13), 53.3 (C-14), 24.8 (C-15), 29.1 (C-16), 52.5 (C-17), 11.9 (C-18), 18.6 (C-19), 38.0 (C-20), 12.3 (C-21), 73.2 (C-22), 72.2 (C-23), 41.0 (C-24), 32.2 (C-25), 21.6 (C-26), 21.2 (C-27), 10.0 (C-28), 179.0 (C-29), 24.3 (C-30), 107.2 (C-1′), 75.3 (C-2′), 77.0 (C-3′), 71.7 (C-4′), 73.2 (C-5′), 70.2 (C-6′), 105.1 (C-1′′), 74.9 (C-2′′), 77.9 (C-3′′), 73.5 (C-4′′), 76.8 (C-5′′), 171.2 (C=O); HRESIMS (+): m/z 851.43524 [M + Na]⁺ (calcd for C₄₂H₆₈NaO₁₆, 851.43996, Δ −5.6 ppm).

D- and L- determination was performed using combined gas chromatography/mass spectrometry (GC/MS) of the butylated derivatives of the monosaccharides produced from the compounds 1–4 by acidic hydrolysis. The sample materials were placed into individual test tubes. To this, aqueous 2 M TFA (400 µL) was added. The tubes were then placed at 121 °C for 1.5 h. Once cooled, the sample was dried under nitrogen. The dried samples were then re-N-acetylated with pyridine and acetic anhydride in methanol. After drying, the samples were butylated using S-(+)-2-butanol or R-(+)-2-butanol (Sigma) for 16 h at 80 °C. The samples were per-O-trimethylsilylated by treatment with Tri-Sil (Pierce) at 80 °C (0.5 h). GC/MS analysis of the butylated derivatives was performed using a Supelco EC-1 fused silica capillary column (30m × 0.25 mm ID). Individual standards of all the detected residues were analyzed in parallel with each sample. Comparison of retention times allowed for the elucidation of the specific conformation of each monosaccharide in the compounds. All residues were of D configuration.

Jurkat and CHO-K1 cells were cultured at 37 °C in a humidified 5% CO₂ atmosphere in 24 cm² cell culture flasks in RPMI 1640 medium supplemented with 5% heat inactivated fetal bovine serum, 2 mM L-glutamine, free of antibiotics.

The purity of all tested compounds was found to be higher than 96% based on HPLC-ELSD analysis. Compounds at final concentrations of 1 and 10 μM was added to 100 μL of the CHO and Jurkat cells suspension (1.0 × 10⁴/well for Jurkat cells and 8.0 × 10³/well for CHO cells in RPMI 1640 medium with 5% of FBS) onto wells. Cells were treated with each sample for 24 h at 37 °C in 5% CO₂. After treatment, the viability of cells was evaluated by the MTT assay. MTT reagent was added to each plate, and after 4 h of incubation, 100 µL of acidic isopropyl alcohol mixture (50 mL Triton X-100, 4 mL of 30% HCl and 446 mL of isopropyl alcohol) was added to dissolve the water-insoluble formazan salt for overnight. The OD₅₇₀ nm absorbance was measured. Each concentration was tested by triplicated and unexposed cells were regarded as 100% viable [30].

The method was adapted from van Duick et al. [31] and from Taniyama et al. [32]. Blood was obtained from healthy young male donors. The red blood cells (RBCs) were washed three times and resuspended in sterile RPMI 1640 medium (S&A) PBS to give about 15 × 10⁶ cells per mL and further processed [31]. The erythrocytes were incubated with compounds in a range 100–200 μM for 1 h at 37 °C. After centrifugation of the nonhemolysed erythrocytes (800 rpm, 5 min), the absorbance of the released hemoglobin in the wavelength 540 nm was measured. The percentage of hemolysis was determined by comparing the absorbance of hemoglobin at 540 nm released from the RBCs in the presence of each compound. The positive control (100% hemolysis) was determined by the amount of hemoglobin released from 15 × 10⁶ RBCs after 1h of incubation with 0.1% of Tween-20.