Derivatives of Salarin A, Salarin C and Tulearin A—Fascaplysinopsis sp. Metabolites

Derivatives of salarin A, salarin C and tulearin A, three new cytotoxic sponge derived nitrogenous macrolides, were prepared and bio-evaluated as inhibitors of K562 leukemia cells. Interesting preliminary SAR (structure activity relationship) information was obtained from the products. The most sensitive functionalities were the 16,17-vinyl epoxide in both salarins, the triacylamino group in salarin A and the oxazole in salarin C (less sensitive). Regioselectivity of reactions was also found for tulearin A.


Introduction
Four groups of nitrogenous macrolides ( Figure 1) were isolated from the Madagascar Fascaplysinopsis sp. sponge collected in Salary Bay, ca. 100 km north of Tulear [1][2][3]. Among these OPEN ACCESS metabolites the most abundant and active compound was salarin C (1), a potent inhibitor of proliferation of K562 leukemia cells (in concentration of 0.0005-0.5 μg/mL). The K562 cells underwent apoptotic death, as monitored by cell cycle analysis, annexin V/propidium iodine staining and caspase 3 and caspase 9 cleavage [1][2][3]. Interestingly, a remarkable change in the amounts of the different compounds was observed in various collections of the sponge. This, together with the resemblance of the functional groups to functionalities found in microorganism metabolites [1][2][3][4][5][6][7], suggested that the four groups originate from guest microorganisms rather than from the sponge itself.
The salarins contain seven functional groups which complicate the chemistry of these compounds. Among the currently identified ten naturally occurring salarins (A-J) the most active one is salarin C (1) which changes under light and air to salarin A (2) [2,8]. The suggested mechanism for this transformation is a singlet oxygen cleavage of the oxazole via a cycloaddition adduct [9,10]. The latter change, which occasionally occurs while performing chemical reactions with salarin C, complicated the chemistry even more. It was therefore advisable to compare reactions of salarin C with those of salarin A, vide infra. Among the more reactive moieties of 1 is the 16,17-vinyl epoxide which, as expected, is more reactive than the 12,13-epoxide, thus enabling regioselective reactions. Different openings of the vinyl epoxide were undertaken for: (a) A preparation of α-methoxy-αtrifluoromethylphenylacetic acid (MTPA) derivatives for determination of the absolute stereochemistry of 1; (b) Linking a spacer for immobilization of the molecule; (c) Preparation of more polar derivatives of 1 for SAR studies; and (d) Supporting the suggested biogenesis of tausalarin C (Figure 1) [4].
In five of the natural salarins the 16,17-epoxide is replaced by the corresponding vicinal 16,17-or 16,19-diol obtained, most likely, due to an allylic rearrangement.

Results and Discussion
Selective acid catalyzed openings of the vinyl epoxide were disclosed with catalytic amounts of HClO 4 in acetone or HCl in MeOH. The former provided the major compounds 16,17-diol 5 and its acetonide derivative 6, and the latter gave the 16-hydroxy-17-methoxy derivative 7 (Scheme 1). Other tested conditions afforded complex mixtures. According to the NMR data of the dieneoate moiety, which is very sensitive to the stereochemistry of the double bonds (e.g., the 8.31 ppm chemical shift of H-4) it was evident that this functional site stayed intact [11]. The same was the case with other moieties that did not take part in the reaction whose NMR data remained unchanged (for all intact sites Δδ H ± 0.1, Δδ C ± 0.5, the same is the case for the other derivatives) [11]. Scheme 1. Acid catalyzed openings of the vinyl epoxide of salarin C (1).
Characteristic in the proton NMR spectra of compounds 5-7 were the disappearance of the 16,17-epoxide protons around 3 ppm and the appearance of methinoxy proton signals around 4 ppm (see Experimental and Supporting Information). All above reactions were very sensitive to the conditions and afforded low yields. It can be assumed that C-16 in compounds 5-7 maintains its (16R*) stereochemistry (protonation of the epoxide followed by opening of the O-C17 bond), which is doubtful for C-17, as the intermediate allylic C-17 carbocation can give two epimers due to a nucleophilic attack from both sides of the molecule and/or allylic rearrangement as shown, for e.g., in 9 (Scheme 1).
The 17 methoxy location was determined from COSY (H-15 to -18) and HMBC data from the methoxyl to C-17. A 16,17-threo configuration was deduced for acetonide 6 from the very close chemical shifts of the two acetonide methyl groups (28.7q and 1.44s for Me-29, 27.3q and 1.43s for Me-30, and 101.7s for C-28 of the acetonide), as well as the J H16,H17 = 7.8 Hz coupling constant [12], thus determining the 17S*-stereochemistry for both 5 and 6.
Treating salarin C with CAN (cerium ammonium nitrate) in acetonitrile (Scheme 1), a reaction suggested for opening radically vinyl epoxides to the corresponding diols [13], provided diol 8 and the allylic rearranged diol 9. As CAN is highly acidic it is not clear if the reagent or the accompanying HNO 3 acid are responsible for the epoxide opening (Scheme 1). A single diastereoisomer was isolated; this was also the case with the other epoxide openings.
Another interesting opening of the 16,17-epoxide was with the freshly prepared Lewis acid MgBr 2 in ether, conditions known to afford from vinyl epoxides the corresponding bromohydrins (Scheme 2) [14]. In the event, the expected bromohydrin (10) was initially indeed obtained as inferred from the MS. However, the compound was not stable and rearranged overnight to a stereoisomer of N-desacetyl salarin J (11) as concluded from the NMR data [6]. Namely, H-16 moved to 5.04 ppm (dd, J = 7.0, 5.1 Hz) and H-17 to 4.40 ppm (t, J = 6. 8 Hz). C-16 was found to resonate at 75.4 and C-17 at 80.9 ppm. The chemical shifts of the intact moieties did not change. The main 2D correlations are depicted in Figure 2. A suggested mechanism for the rearrangement is given in Scheme 2. Because of the small J values in the THF ring, the stereochemistry around the ring could not be determined.  Worth mentioning is another opening of the 16,17-epoxide, namely, hydrogenation of salarin C which saturated the four double bonds, without affecting the oxazole, and opened up the 16,17-vinyl epoxide to the 17-alcohol (12, Scheme 3 and Figure 3). The 17-hydroxy location is suggested on the basis of COSY correlations from H-13 to H 2 -18 (see Experimental Section) and supported by HMBC correlations (Figure 3). Furthermore, the trans J 12,13 = 2.2 Hz value confirmed that the 12,13-epoxide remained intact. Therefore, the configuration of C-17 could not be determined.  The next studied reaction was an amine/Zn(ClO 4 )· 6H 2 O opening of the 16,17-vinyl epoxide [15], which provided, as expected, the 16-hydroxy-17-amino-derivatives (13)(14)(15)(16)(17)(18) (Scheme 4) [16]. The reaction was performed both on salarin C and salarin A. Unexpectedly, the two compounds behaved differently. While salarin C gave the expected 16-hydroxy-17-amino derivatives, via infra, (13)(14)(15)(16)(17)(18), in the case of salarin A the epoxide remained intact while the triacyl moiety of the macrocycle of 2 (compound 19) was found to be more reactive and opened up as depicted in Scheme 5. The only changed site in the NMR was that of CH-16 and CH-17; e.g., for compound 14, H-16 moved to 3.60 ppm (t, J = 7.7 Hz), H-17 to 2.83 ppm (t, J = 7.7 Hz), C-16 resonated at 71.6 ppm and C-17 at 64.8 ppm.  Carbonyls C-6 and C-7 of compound 19 moved up-field in the NMR spectrum from 171.9 to 164.0 and from 167.9 to 163.6, respectively. Evidence for change in the molecule was further obtained by FABMS m/z 796.1, (C 39 H 55 N 3 O 13 Na, M + Na + ).
Observed NOEs between H-4 and H-5, between H-5 and H-28, and between NH-30 and H-8, as well as HMBC correlations from NH-30 to C-27, and from H-4 to C-6 differentiated between three possible cleavages of the triacylamine moiety (i.e., C-6/C-7; C-30/C-6 or C-30/C-7) determining the structure of 19 ( Figure 5). In addition to preliminary SAR studies of salarin C derivatives, the amination of salarin C was also directed to obtain a congener possessing a spacer for immobilization of salarin C (compounds 17 and 18, Scheme 4). However, as compounds 17 and 18 already lost activity, no further reactions were undertaken. Also, opening of the vinyl epoxide with an amine supported the earlier suggested biogenesis of tausalarin C, i.e., coupling between salarin A (or C) and pre-taumycin that, by a different route, affords taumycin A [4].
Next the preparation of a more polar derivative of salarin C was undertaken to try to improve its solubility in water. As was demonstrated in acidic conditions, salarins were also found highly sensitive in basic conditions. Short treatment of salarin C with different bases first gave the N-desacetyl derivative 20 (Scheme 6) characterized by the disappearance of the acetyl resonances in the 1 H-NMR spectrum, while prolonging the reaction (2 h), using a stronger base, also led to hydrolysis of the octanoate ester (21, Scheme 6). Further elongation of the hydrolysis resulted in an unidentified complex mixture.
As shown above, in different reactions, salarin A behaved differently from salarin C. The same was the case with the hydrolysis. Again, in the first hydrolysis step the N-acetyl group split off to give compound 22, known naturally as salarin E (most likely together with methyl acetate which was not isolated) [6], while the second step was the opening of the azamacrocycle, providing the C-6 methyl ester-C-7 amide (23, Scheme 7).
The 1 H-NMR spectrum of 23 showed a three proton singlet at 3.78 ppm, indicating the presence of a methyl ester, accompanied by the disappearance of the two acetates (C-25 and C-27 of 2). Scheme 6. Treatment of salarin C with base.

Scheme 7. Treatment of salarin A with base.
Two structures are possible for 23, i.e., in one C-6 is the ester and C-7 the amide, or vice versa. HMBC correlations from H-4, H-5 and the methoxyl to carbonyl C-6 and from H-8 to CH 3 -22 and to the amide carbonyl C-7 clarified the structure (23, Scheme 7); key CH-correlations are depicted in Figure 6.
After obtaining compound 21, its attachment to glucosamine (Scheme 8) was achieved via activation of the primary C(21)H 2 OH group through the p-nitrophenyl carbonate group (24), which was then converted to the sugar carbamate 25 (Scheme 8). The HRESI-MS and NMR data of 23 confirmed its structure. Downfield shifts of H-2′ of the glucosamine (∆δ H = +0.9 ppm), compared to the starting free glucosamine (δ H 2.67 m), and also of methylene H 2 -21 (δ H 4.12 m) and HN-2′ (δ H 5.16 s), as well as 3 J CH correlations to the carbamate carbonyl (δ C 155.9 s), evidenced the connection of the sugar 21.
Tulearin A (3) belongs to the second group of the Fascaplysinopsis sp. metabolites (tulearins A-C) [5] their cytotoxicity to K562 leukemia cells was lower than that of salarin C (72 h, 1 μM, ~60% inhibition of proliferation), while salarin C inhibited completely in this concentration [3,5]. However, it was decided to check the bioactivity of these derivatives.
In our previous report, Diels Alder reactions of the 18,20-diene of tulearin A, as well as a modification of the 8-carbamate to a crystalline carbonate, were reported [5]. For the absolute configuration determination of the tulearins we prepared regioselectively the 9-MTPA esters to apply the modified Mosher method [17]. The latter 9-ester was obtained exclusively without the 3-isomer [5]. Thus, it was interesting to find out where the other esterifications would take place. Acetylation, mesylation, tosylation, benzoylation and p-bromobenzoylation were performed. All first gave the 9-mono derivative (26-34), followed by a slower esterification of the 3-hydroxyl. The 9-hydroxyl esterification location became clear from the H-9 chemical shift to a lower field ( Figure 7 and Experimental Section). The only exception was the 3-carbonate obtained from the reaction of 3 with p-nitrophenyl chloroformate (PNPCl, 35, Table 1). To exclude/diminish the possibility that an initially obtained 9-isomer rearranges to the 3-isomer via a 3,9-bridge, reactions with oxalyl chloride and thionyl chloride were undertaken-no transannular bridge could be revealed.
In addition, the acylation of the carbamate was examined (33 and 34, Table 1). Next, dimesyl 29 was transformed to the 3,9-diazido derivative (36) which was found, in low yields, to slowly transfer to the cycloadditon pyrolidine product 37 (Scheme 9) [18] -3 and N 4 , most likely because of too small a polarization transfer, thus the carbamate nitrogen was not disclosed (Figure 8) [18]. Furthermore, C-12 resonances, were as expected from an imine at δ C 179.2 ppm, CH-3 at δ C 61.8 and δ H 3.78 ppm, and CH-9 at δ C 74.8 and δ H 4.20 ppm.

Salarin Derivatives
The NMR data of the salarin derivatives is given only for the transformed sites, for the rest of the functional moieties of the molecules the chemical shift changes of the various atoms were minimal (Δδ H ± 0.1, Δδ C ± 0.5). Thus for example the chemical shift of H-4 is very characteristic for the dienoate moiety and its surroundings (8.31 ppm). Full representative NMR spectra are given in the Supporting Information. The NMR data of the changed sites are given for clarity in order of the atom numbers.

Compounds 5 and 6
To a mixture of salarin C (20 mg, 0.03 mmol) in acetone (5 mL), was added HClO 4 (7%, 0.1 mL) at −20 C and the reaction was stirred for 10 min. The reaction mixture was neutralized with aqueous NaHCO 3 solution, evaporated and the residue diluted with water and extracted with DCM. The combined organic extract was dried over anhydrous MgSO 4 and evaporated. The residue was purified by VLC (vacuum liquid chromatography, petroleum ether/ethyl acetate, 8:2) to afford 6, as a colorless oil, 3 mg (20%) and 5, a second colorless oil, 4.5 mg (21%). 6: [α]

Compound 7
A solution of salarin C (12 mg, 0.02 mmol) in MeOH (10 mL) was treated with methanolic HCl solution (0.05% v/v, 0.1 mL). The mixture was stirred at rt in the dark for 1 h. The reaction mixture was then neutralized with aqueous NaHCO 3 solution, the solvent was evaporated and the residue diluted with water and extracted with DCM. The combined organic extract was dried over anhydrous MgSO 4 and evaporated. The residue was purified by VLC (vacuum liquid chromatography, petroleum ether/ethyl acetate, 7:3) to afford 7, as a colorless oil, 8 mg (65%). [

Compound 11
MgBr 2 was prepared from magnesium turnings (100 mg) in dry diethyl ether (10 mL) and dibromoethane (1.5 mL). A solution of salarin A (21 mg, 0.03 mmol) in dry diethyl ether (1 mL), was added dropwise to the 0 °C mixture of the MgBr 2 . The solution was stirred for 2.5 h, then water added and the solution extracted with diethyl ether. The organic layer was washed with brine, dried over anhydrous MgSO 4 and the solvents were evaporated. The residue was purified by VLC (petroleum ether/ethyl acetate, 1:1) to afford a bromohydrin product which was unstable; ESIMS m/z 789.

Compound 22
To a solution of salarin A (11 mg, 0.016 mmol) in EtOH (5 mL) was added NH 4 Cl (2.5 mg, 0.048 mmol) and NaN 3 (3.1 mg, 0.048 mmol). The solution was slowly warmed up to 60 °C for 1 h, and then allowed to cool down to room temperature. The reaction mixture was filtered off, the solid residue washed with EtOH, and the solvent was evaporated. The residue was dissolved in DCM (50 mL) and washed with H 2 O (30 mL). The organic layer was dried over anhydrous Na 2 SO 4 and then evaporated to afford 22, as a yellow oil, 8 mg (80%). The NMR-data is identical to natural salarin E [6].

Compound 28
To a solution of tulearin A (10 mg, 0.025 mmol) in DCM (2 mL) were added Et 3 N (5.2 mg, 0.05 mmol) and p-bromobenzoylchloride (11.3 mg, 0.05 mmol). The mixture was stirred at room temperature for 24 h. DCM was then added (5 mL) and the mixture was washed with saturated NH 4 Cl, water and brine, dried over anhydrous MgSO 4 and evaporated. The residue was purified by VLC (petroleum ether/ethyl acetate, 1:1) to afford 28, as a colorless oil, 7.2 mg (40%).