Mandelalides A-D, cytotoxic macrolides from a new Lissoclinum species of South African tunicate

Abstract

Mandelalides A-D are variously glycosylated, unusual polyketide macrolides isolated from a new species of Lissoclinum ascidian collected from South Africa, Algoa Bay near Port Elizabeth and the surrounding Nelson Mandela Metropole. Their planar structures were elucidated on sub-milligram samples by comprehensive analysis of 1D and 2D NMR data, supported by mass spectrometry. The assignment of relative configuration was accomplished by consideration of homonuclear and heteronuclear coupling constants in tandem with ROESY data. The absolute configuration was assigned for mandelalide A after chiral GC-MS analysis of the hydrolyzed monosaccharide (2-O-methyl-α-L-rhamnose) and consideration of ROESY correlations between the monosaccharide and aglycone in the intact natural product. The resultant absolute configuration of the mandelalide A macrolide was extrapolated to propose the absolute configurations of mandelalides B-D. Remarkably, mandelalide B contained the C-4′ epimeric 2-O-methyl-6-dehydro-α-L-talose. Mandelalides A and B showed potent cytotoxicity to human NCI-H460 lung cancer cells (IC₅₀, 12 and 44 nM, respectively) and mouse Neuro-2A neuroblastoma cells (IC₅₀, 29 and 84 nM, respectively).

Introduction

Over the last 50 years, ascidians have been shown to be a prolific source of natural products with promising biomedical potential.¹ Indeed, ascidian-derived natural products have yielded promising drug leads, among which ecteinascidin 743 (Yondelis®) and dehydrodidemnin B (Aplidin®) are in clinical use for the treatment of specific cancers.² The diverse chemotypes reported from Lissoclinum species collected globally have important biological properties that range from cancer cell toxicity to antifungal and antibacterial activities,¹ and include peptides, alkaloids, chlorinated diterpenes, polyether amides, lactones, and macrolides such as haterumalide B³ and patellazoles B and C.^4,5 It has been noted that many secondary metabolites isolated from Lissoclinum, as well as other ascidians species, are present in bacteria, sponges and mollusks.^6-8 This diversity of ascidian metabolites can be explained by the fact that tunicates are often hosts to cyanobacterial and heterotrophic bacterial symbionts, as well as being fed upon by predatory mollusks.^7-10 Therefore, identification of the biogenetic origin of ascidian natural products is very often challenging.¹¹ This is highlighted by the recent isolation of didemnin B (an analog of Aplidin®) from the α-proteobacterium Tistrella mobilis.^8,12 This compound was originally identified from the ascidian Aplidium albicans, and is biosynthetically similar to peptides from the marine cyanobacterium Lyngbya majuscula⁸ (now Moorea producta).

In our search for new potential anticancer compounds, we encountered the highly cytotoxic organic extract (IC₅₀ = 0.7 μg/mL in NCI-H460 lung cancer cells) of a new Lissoclinum species from Algoa Bay, South Africa. Bioassay-guided fractionation of this extract, yielded a series of new macrolides, named mandelalides A-D (1-4), two of which were tested and show good cytotoxicity to mouse Neuro-2A neuroblastoma and human NCI-H460 lung cancer cells. While the relative configuration of the compounds could be assigned from analysis of ROESY data in tandem with homonuclear (³J_HH) and heteronuclear (^2,3J_CH) coupling constants, the assignment of absolute configuration relied on hydrolysis of the glycosylated mandelalide A and chiral GC-MS analysis of the released monosaccharide. Remarkably, mandelalide A contains 2-O-methyl-α-L-rhamnose, while mandelalide B contains the C-4′ epimer 2-O-methyl-6-dehydro-α-L-talose.

Results and Discussion

A new Lissoclinum ascidian species was collected from White Sands Reef in Algoa Bay, South Africa. The organic extract (1.45 g) was subjected to bioassay-guided fractionation through consecutive LH-20 and RP-HPLC chromatography to yield mandelalides A-D (1 - 4), of which mandelalide D degraded to deacylmandelalide D (4b).

The HR-ESI-MS data for mandelalide A (1) gave a pseudomolecular ion [M+Na]⁺ at m/z 647.3394, which is consistent with a molecular formula of C₃₃H₅₂O₁₁, and implies 8 degrees of unsaturation. The ¹³C and multiplicity-edited HSQC NMR spectra for compound 1 (Table 1) indicated the presence of an ester carbonyl (δ_C 167.4), fourteen sp³ methines, twelve of which were oxygen-bearing (δ_C 94.2–68.1, 37.3 and 34.2), six olefinic methines (δ_C 147.1, 141.5, 131.3, 126.9, 123.9 and 123.1), eight methylenes (δ_C 66.1, 43.1, 39.7, 38.8, 37.6, 36.8, 34.1 and 31.1) and four methyl groups (δ_C 59.1, 18.3, 17.7 and 14.5). Interpretation of COSY and TOCSY correlations delineated two spin systems, one comprising 25 carbons from C-2 to C-26 (Figure 2a, fragment A), and the other of 6 carbons from C-1′ to C-6′ (Figure 1a, fragment B). The acquisition of a semi-phase sensitive HMBC¹³ optimized for a 4 Hz coupling constant (Figure S6), revealed a correlation between δ_H 5.23 (H-23) and δ_C 167.4 (C-1) that indicated cyclization of fragment A into a 23-carbon macrolactone. The C-1 carbonyl also showed HMBC correlations from the olefinic H-2 and H-3 multiplets, for which a ³J_HH of 15.5 Hz indicated an E double bond geometry. An HMBC correlation between H-9 (δ 3.32) and C-5 (δ 73.9) delineated a tetrahydropyran ring, while a correlation between H-17 (δ 3.98) and C-20 (δ 83.2) was consistent with a tetrahydrofuran ring, both within fragment A (Table 1, Figure 1a). The four contiguous (TOCSY-coupled) olefinic methines CH-12 to CH-15 (δ_H 5.45, 6.28, 6.05 and 5.28 ppm) were consistent with a conjugated diene, situated β to the tetrahydrofuran and γ to the tetrahydropyran on the basis of HMBC correlations from H-15 to C-17 and H-12 to C-10, respectively. From homonuclear coupling constant values (³J_HH = 14.8 and 10.8 Hz, respectively) it was evident that the geometry of H-12 and H-13 is trans, while H-14 and H-15 are cis. Knowing that ³J_HH values are not sufficiently characteristic to assign the orientation between H-13 and H-14, ROESY data were examined.¹⁴ Signals confirming NOE enhancements between H-12 and H-10a, as well as between H-13 and H-16b, suggested an anti orientation for H-13 and H-14 (Figure 1b). In the case of fragment B, the downfield shift of C-1′ (δ_C 94.2), the presence of four midfield oxymethine ¹³C resonances, an upfield methyl doublet (CH₃-6′, δ_C 17.7, δ_H 1.27) and an HMBC correlation between H-1′ (δ 5.02) and C-5′ (δ 68.1) suggested that fragment B is a 6-dehydro monosaccharide. An HMBC correlation between H-2′ (δ 3.40) and deshielded methyl C-7′ (δ 59.1) indicated methylation of the C-2′ hydroxyl group. An HMBC correlation from H-7 (δ 3.82) to C-1′ placed the monosaccharide (fragment B) at C-7 of fragment A.

Table 1.

NMR data for Mandelalide A (1) in CDCl₃

No.	δC (mult)	δH (J in Hz)	COSY	HMBC	TOCSY	ROESY
1	167.4 (s)
2	123.1 (d)	6.01 (dd, 15.5, 1.2)	3, 4a	1, 4	3, 4a, 4b, 5, 6ax, 6eq	3b, 4a, 4b, 5
3	147.1 (d)	6.97 (ddd, 15.2, 10.4, 4.6)	2, 4a, 4b	1, 2, 4, 5	2, 4a, 4b, 5, 6ax, 6eq	3b, 4ab, 4bb, 5
4a	38.8 (t)	2.36 (m)	2, 3	2, 3, 5, 6	2, 3, 5, 6ax, 6eq, 7	2, 3b, 5b, 6axb, 6eq
4b		2.39 (ddd, 14.1, 10.6, 10.6)	3, 5	2, 3, 5, 6	2, 3, 5, 6ax, 6eq, 7	2, 3b, 5b, 6axb, 6eq
5	73.9 (d)	3.36 (dddd, 11.4, 11.4, 2.3, 2.3)	4a, 4b, 6ax, 6eq	3, 4	2, 3, 4a, 4b, 6ax, 6eq, 7, 8ax, 8eq	2, 3, 4a, 4b, 6eq, 7
6ax	37.6 (t)	1.20 (m)	5, 6eq, 7	4, 5, 8	3, 4a, 4b, 5, 6eq, 7	4bb, 7
6eq		2.02 (dddd, 12.6, 4.4, 2.3, 1.6)a	5, 6ax, 7, 8eq	7, 8	4a, 4b, 5, 6ax, 7, 8ax,	6axb, 4a, 4b, 5, 7, 1′
7	73.1 (d)	3.82 (dddd, 11.1, 10.5, 4.4, 4.4)a	6ax, 6eq, 8ax, 8eq	8, 9, 1′	4a, 4b, 5, 6ax, 6eq, 8ax, 8eq, 9	5, 6axb, 6eq, 8axb, 8eq, 9, 1′
8ax	39.7 (t)	1.22 (m)	7, 8eq, 9	6, 7, 9, 10	6eq, 7, 8eq, 9, 10b	6eqb, 7, 8eqb
8eq		1.87 (m)	6eq, 7, 8ax, 9	6, 7	6eq, 7, 8ax, 9, 10a, 10b	7, 8axb, 9, 10b, 5′
9	72.5 (d)	3.32 (dddd, 11.2, 11.2. 2.2, 2.2)a	8ax, 8eq, 10a, 10b	5, 7	6eq, 7, 8ax, 8eq, 10a, 10b, 11, 12, 13, 25	7, 8eq, 10a, 10bb, 25
10a	43.1 (t)	1.21 (ddd, 15.2, 9.6, 2.2)a	9, 10b, 11	11, 12, 25	8eq, 9, 10b, 12, 13, 14, 25	8eq, 9, 10bb, 11b, 12, 25
10b		1.51 (ddd, 15.2, 11.2, 3.7)a	9, 10a, 11	8, 9, 11, 12, 25	8ax, 8eq, 9, 10a, 11, 12, 13, 25	8eq, 9b, 10ab, 11, 25
11	34.2 (d)	2.37 (dqd, 9.6, 6.5, 3.7)a	10a, 10b, 12, 25	10, 12, 25	8ax, 8eq, 9, 10a, 10b, 11, 12, 13, 14, 15	10ab, 10b, 12b, 13, 25b
12	141.5 (d)	5.45 (dd, 14.8, 9.7)	11, 13	10, 11, 14, 25	9, 10a, 10b, 11, 13, 14, 15, 16a, 16b, 25	11b, 10a, 13b, 14b, 25
13	123.9 (d)	6.28 (dd, 14.8, 11.0)	12, 14	10, 11, 14, 15	10a, 10b, 11,12, 14, 15, 16a, 16b, 17, 25	11, 12b, 14b, 15b, 16b, 21
14	131.3 (d)	6.05 (dd, 10.9, 10.9)	13, 15	12, 13,16, 17	10a, 11, 12, 13, 15, 16a, 16b, 17, 25	12b, 13b, 15b, 16a
15	126.9 (d)	5.28 (ddd, 10.8, 10.8, 5.6)	14, 16a, 16b	13, 16, 17	12, 13, 14, 16a, 16b, 17, 18, 26	13b, 14b, 16a, 16bb, 17b
16a	31.1 (t)	1.88 (m)	14, 15, 16b, 17	14, 15, 17	12, 13, 14, 15, 16b, 17, 18, 19a, 26	14, 15, 16b, 17b, 26
16b		2.28 (ddd, 13.1, 11.4, 11.4)	15, 16a, 17	14, 15, 17, 18	12, 13, 14, 15, 16a, 17, 18, 19a, 19b, 26	13,15b,16a,17b,19a,21,26
17	81.0 (d)	3.98 (ddd, 11.1, 8.1, 1.8)	16a, 16b, 18	15, 19, 20	14, 15,16a, 16b, 18, 19a, 19b, 20, 21, 26	15, 16ab, 16bb,18, 26
18	37.3 (d)	2.52 (dddq, 12.0, 7.0, 7.0, 7.0)a	17, 19a, 19b, 26	16, 17, 19, 26	15, 16a, 16b,17, 19a, 19b, 20,21, 22a, 26	17, 19ab, 19b, 20, 26b
19a	36.8 (t)	1.17 (ddd, 11.9, 11.9, 10.3)	18, 19b, 20	18, 20, 21, 26	17, 18, 19a, 20, 21, 22a, 22b, 26	16b, 18b, 21, 22b, 26
19b		2.01 (ddd, 12.2, 7.0, 5.6)a	18, 19a, 20	17, 18	16b, 17, 18, 19a, 20, 21, 22a, 26	18, 19a, 20, 22a, 22b, 26
20	83.2 (d)	3.63 (m)	19a, 19b, 21	21, 22	17, 18, 19a, 19b, 21, 22a, 22b, 26	18, 19b, 21b, 22a, 22b
21	73.0 (d)	3.42 (ddd, 11.1, 8.8, 1.8)	20, 22a, 22b	20, 22, 23	17, 18, 19a, 19b, 20, 22a, 22b, 23, 24a, 24b, 26	13, 16b, 19a, 22ab, 22b, 23
22a	34.1 (t)	1.46 (ddd, 14.1, 11.1, 1.9)	21, 22b, 23	20, 21	18, 19a, 19b, 20, 21, 22b, 23, 24a, 24b, 26	19b, 21b, 22bb, 23, 24a, 24b
22b		1.76 (ddd, 13.9, 11.7, 1.8)	21, 22a, 23	23, 24	19a, 20, 21, 22a, 23,24a, 24b	19a, 19b, 21, 22ab,23b, 24a, 24b
23	72.3 (d)	5.23 (dddd, 11.7, 4.9, 2.9, 1.9)	22a, 22b, 24a, 24b	1	21, 22a, 22b, 24a, 24b	21, 22a, 22bb, 24a, 24b
24a	66.1 (t)	3.61 (m)	23, 24b	22, 23	23, 24b	22a, 22b, 23, 24b
24b		3.81 (dd, 12.2, 2.9)a	23, 24a	22, 23	22b, 23, 24a	22a, 22b, 23, 24a
25	18.3 (q)	0.85 (d, 6.6)	11	10, 11, 12	9, 10a, 10b, 11, 12, 13, 14	9, 10a, 10b, 11b, 12
26	14.5 (q)	1.03 (d, 6.9)	18	17, 18, 19	15, 16a, 16b, 17, 18, 19a, 19b, 20 21	16a, 16b, 18b, 19a, 19b
1′	94.2 (d)	5.02 (d, 1.1)	2′	7, 2′, 3′, 5′	2′, 6′	6eq, 7, 8eq, 2′, 7′
2′	80.8 (d)	3.40 (dd, 3.8, 1.4)	1′, 3′	3′, 4′, 7′	1′, 3′, 4′, 5′, 6′, 8′, 9′	1′
3′	71.7 (d)	3.68 (m)	2′, 4′, 8′		2′, 4′, 5′, 6′, 8′, 9′	2′, 4′b
4′	74.3 (d)	3.34 (dd, 9.4, 9.4)	3′, 5′	3′, 5′, 6′	3′, 5′, 6′, 8′, 9′	3′b, 5′b, 6′
5′	68.1 (d)	3.62 (m)	4′, 6′	3′, 4′, 6′	4′, 6′, 8′, 9′	8eq, 4′b, 6′
6′	17.7 (q)	1.27 (d, 6.3)		4′, 5′	1′, 3′, 4′, 5′, 8′	4′, 5′b
7′	59.1 (q)	3.45 (s)		2′		1′
	OH-3′	2.24 (s)	2, 4a, 4b		2′, 3′, 4′, 5′, OH-4′
	OH-4′	1.54 (s)	2, 3		3′, 4′, 5′, OH-3′

HMBC correlations are presented from proton to indicated carbon;

^a3J_HH values obtained from DQFCOSY and/or NMRSim;

^bCOSY artifacts observed in ROESY spectrum.

Figure 2.

a) Key ROESY correlations indicated on the C-5/C-15 fragment of 1; b) Two possible rotamers around the C-9/C-10 bond; c) Most feasible rotamers around the C-10/C-11 bond of 1, with anticipated large/small heteronuclear coupling constant indicated on each. For b) and c) the J_HH and J_HC values (CDCl₃) are listed below the Newman projections and the rotamer of best fit is indicated by a dashed outline.

Figure 1.

a) Planar structure of 1 showing TOCSY correlations from H-2, H-6, H-8, H-10b, H-13, H-17, H-21, H-1′, H-2′ (black circles) indicated as bolded lines and key HMBC correlations represented by single-headed arrows. b) The assignment of geometry around the C-13/C-14 bond in 1 using key ROESY correlations indicated by double-headed arrows.

The second compound characterized, mandelalide B (2), was assigned a molecular formula of C₃₇H₅₈O₁₃, based on HR-ESI-MS data for [M+Na]⁺ m/z 733.3730, which is consistent with 9 degrees of unsaturation. Inspection of the ¹H and ¹³C NMR spectra for mandelalide B (2) suggested that it was structurally related to mandelalide A (1). However, the ¹H NMR spectrum for 2 lacked the coupled olefinic signals at δ 6.01 (H-2) and 6.97 (H-3) and the relatively deshielded diastereotopic H₂-24 signals at δ 3.81 and 3.61. Instead an additional oxymethine double doublet at δ 5.48 was evident, as well as a methyl triplet (δ_H 0.95) indicative of an aliphatic chain. Correspondingly, comparison of the ¹³C NMR spectra for 1 and 2 revealed the absence of olefinic ¹³C resonances at δ 123.1 (C-2) and 147.1 (C-3) from the spectrum for 2, and the presence of two additional midfield ¹³C resonances (δ 79.3 and 69.5), as well as a second carbonyl resonance (δ_C 173.6). These data accounted for 8 of the 9 degrees of unsaturation, implying the presence of an additional cycle in 2. The analysis of HSQC and HMBC data for 2 permitted assignment of an oxygenated quaternary C-2 (δ 79.3) and an oxymethine CH-3 (δ 69.5, δ 5.48). While the same C-1/C-23 macrolactone linkage was apparent in both compounds, two- and three-bond HMBC correlations from H₂-24 (δ 1.93, 2.39) to C-2 and C-3, respectively, indicated a C-2/C-24 bond in 2. These data described a γ-butyrolactone ring around the ester linkage of the mandelalide B macrocycle. Placement of the second carbonyl carbon (δ_C 173.6) in a butyrate substituent at C-3 was facilitated by HMBC correlations to δ_C 173.6 from δ_H 5.48 (H-3), as well as 2.35 (H₂-2″) and 1.67 (H₂-3″). Further analysis of NMR data confirmed the presence of a 2-O-methyl-6-dehydro sugar in 2. However, differences in both ¹H and ¹³C shifts for CH-3′ to CH-6′ compared to compound 1 suggested a different relative configuration for the monosaccharide in 2 (Table 2).

Table 2.

¹H and ¹³C NMR data for mandelalides B-D (2-4) in CDCl₃

	Mandelalide B (2)		Mandelalide C (3)		Mandelalide D (4)
No.	δC (mult)	δH (J in Hz)	δC (mult)	δH (J in Hz)	δC (mult)	δH (J in Hz)
1	175.6 (s)	-	174.7 (s)	-	172.9 (s)	-
2	79.3 (s)	-	82.0 (s)	-	81.3 (s)	-
3	69.5 (d)	5.48 (dd, 6.0, 1.7)	68.3 (d)	5.51 (dd, 6.6, 0.9)	68.6 (d)	5.56 (dd, 6.6, 0.7)
4a	36.3 (t)	1.59 (m)	36.4 (t)	1.65 (m)	36.2 (t)	1.68 (m)
4b		2.14 (ddd, 15.0, 11.7, 1.5)		2.14 (ddd, 15.6, 11.7, 1.0)		2.18 (ddd, 15.4, 11.5, 0.9)
5	72.7 (d)	3.30 (dddd, 11.4, 11.4, 1.5, 1.5)	72.3 (d)	3.25 (dddd, 11.4, 11.4, 1.8, 1.8)	72.2 (d)	3.21 (dddd, 11.2, 11.2, 1.3, 1.3)
6ax	37.4 (t)	1.10 (ddd, 11.7, 11.7, 11.7)	41.2 (t)	1.11 (ddd, 12.5, 11.4, 11.4)	41.2 (t)	1.12 (ddd, 11.3, 11.3, 11.3)
6eq		1.89 (m)		1.86 (m)		1.85 (dddd, 12.2, 4.6, 1.5, 1.5)
7	73.3 (d)	3.75 (m)	68.3 (d)	3.76 (dddd, 10.9, 10.9, 4.6, 4.1)	68.4 (d)	3.75 (m)
8ax	39.6 (t)	1.23 (m)	41.8 (t)	1.13 (ddd, 12.5, 11.2, 11.2)	41.8 (t)	1.12 (ddd, 11.3, 11.3, 11.3)
8eq		1.81 (dddd, 12.4, 4.8, 1.8, 1.8)		1.83 (m)		1.83 (dddd, 12.2, 4.6, 1.5, 1.5)
9	72.4 (d)	3.39 (dddd, 11.0, 11.0, 2.2, 2.2)	72.3 (d)	3.37 (dddd, 10.9, 10.9, 2.1, 2.1)	72.2 (d)	3.36 (dddd, 11.0, 11.0, 2.2, 2.2)
10a	42.2 (t)	1.18 (ddd, 14.1, 12.3, 2.8)	42.1 (t)	1.19 (ddd, 14.2, 12.1, 2.8)	42.1 (t)	1.19 (ddd, 13.9, 11.5, 2.8)
10b		1.57 (m)		1.57 (m)		1.57 (ddd, 14.0, 11.5, 4.0)
11	33.9 (d)	2.50 (m)	34.0 (t)	2.49 (m)	34.2 (d)	2.48 (m)
12	142.1 (d)	5.50 (dd, 15.1, 9.8)	142.2 (d)	5.50 (dd, 14.6, 9.6)	142.3 (d)	5.50 (dd, 14.8, 9.7)
13	123.3 (d)	6.40 (dd, 14.7, 11.3)	123.2 (d)	6.39 (dd, 14.6, 11.4)	123.1 (d)	6.38 (dd, 15.0, 11.3)
14	131.0 (d)	6.10 (dd, 11.1, 11.1)	131.1 (d)	6.10 (dd, 11.0, 11.0)	131.1 (d)	6.10 (dd, 11.0, 11.0)
15	127.1 (d)	5.28 (ddd, 11.1, 11.1, 5.2)	127.1 (d)	5.28 (ddd, 11.0, 11.0, 5.4)	127.1 (d)	5.28 (ddd, 11.3, 11.3, 5.4)
16a	30.7 (t)	1.90 (m)	30.7 (t)	1.90 (ddd, 13.6, 5.2, 1.1)	30.6 (t)	1.90 (ddd, 13.6, 5.3, 1.0)
16b		2.31 (ddd, 13.6, 11.6, 11.5,)		2.30 (ddd, 13.8, 11.9, 11.9)		2.29 (ddd, 13.1, 11.3, 11.3)
17	81.5 (d)	3.95 (ddd, 11.6, 7.46, 1.6)	81.6 (d)	3.95 (ddd, 11.5, 7.5, 1.1)	81.6 (d)	3.94 (ddd, 12.0, 7.5, 1.1)
18	38.3 (d)	2.52 (m)	38.3 (d)	2.53 (m)	38.3 (d)	2.54 (dddq, 12.3, 7.5, 7.02, 6.8)
19a	35.6 (t)	1.33 (ddd, 12.2, 12.2, 8.2)	35.7 (t)	1.33 (ddd, 12.2, 12.2, 9.1)	35.7 (t)	1.34 (ddd, 12.3, 12.3, 9.2)
19b		2.05 (ddd, 12.8, 6.9, 6.9)		2.10 (ddd, 12.5, 7.0, 6.8)		2.08 (ddd, 12.2, 7.02, 7.02)
20	82.1 (t)	3.77 (m)	82.4 (d)	3.82 (ddd, 9.2, 9.2, 7.0)	82.2 (d)	3.79 (ddd, 9.3, 9.3, 7.0)
21	74.5 (d)	3.76 (m)	74.4 (d)	3.73 (ddd, 10.8, 9.5, 1.2)	74.3 (d)	3.73 (ddd, 10.0, 10.0, 1.6)
22a	37.9 (t)	1.51 (m)	32.1 (t)	1.59 (m)	32.5 (t)	1.47 (ddd, 14.2, 10.5, 1.0)
22b		1.64 (m)		1.82 (ddd, 14.4, 11.0, 1.0)		1.65 (m)
23	74.9 (d)	4.84 (dddd, 10.4, 10.4, 4.9, 1.0)	78.9 (d)	5.01 (ddd, 11.2, 2.0, 2.0)	76.7 (d)	5.13 (ddd, 10.8, 3.3, 0.8)
24a	39.7 (t)	1.93 (dd, 12.7, 10.7)	72.2 (d)	3.98 (d, 2.8)	74.0 (d)	5.17 (d, 3.3)
24b		2.39 (dd, 12.8, 4.9)	-	-	-	-
25	18.5 (q)	1.07 (d, 6.6)	18.4 (q)	1.06 (d, 6.2)	18.3 (q)	1.07 (d, 6.6)
26	14.2 (q)	1.03 (d, 6.9)	14.2 (q)	1.03 (d, 6.9)	14.3 (q)	1.04 (d, 6.8)
	OH-21	2.70 (br s)	OH	2.80 (br s)	OH	2.71 (br s)
	OH-2	2.87 (s)	-	-	OH	2.78 (br s)
1′	94.8 (d)	5.04 (s)	173.4 (s)	-	173.6 (s)	-
2′a	80.9 (d)	3.33 (ddd, 3.0, 1.5. 1.3)	36.3 (t)	2.34 (td, 7.5, 1.2)	36.3 (t)	2.40 (dt, 15.5, 7.3)
2′b	-	-	-	-		2.43 (dt, 15.5, 7.3)
3′	66.6 (d)	3.74 (m)	18.7 (t)	1.65 (m)	18.6 (t)	1.70 (m)
4′	73.3 (d)	3.49 (ddt, 11.8, 3.3, 1.2, 1.2)	13.9 (q)	0.94 (t, 7.5)	13.9 (q)	0.96 (t, 7.4)
5′	67.4 (d)	3.85(br q, 6.6)	-	-	-	-
6′	16.6 (q)	1.23 (d, 6.5)	-	-	-	-
7′	59.6 (q)	3.45 (s)	-	-	-	-
	OH-3′	2.86 (d, 11.6)	-	-	-	-
	OH-4′	2.74 (d, 11.6)	-	-	-	-
1″	173.6 (s)	-	-	-	173.1 (s)	-
2″a	36.3 (t)	2.35 (t, 7.5)	-	-	36.0 (t)	2.31 (dt, 12.8, 7.5)
2″b	-	-	-	-		2.33 (dt, 12.8, 7.5)
3″	18.8 (t)	1.67 (tq, 7.4, 7.4)	-	-	18.6 (t)	1.63 (m)
4″	14.0 (q)	0.95 (t, 7.4)	-	-	13.8 (q)	0.93 (t, 7.3)

The HR-ESI-MS data ([M+Na]⁺ m/z 589.2970) for mandelalide C (3) supported a molecular formula of C₃₀H₄₆O₁₀, for 8 degrees of unsaturation. The ¹H NMR spectrum for 3 was reminiscent of that for 2. However, a careful comparison of the two spectra revealed the absence of midfield glycosidic signals in the ¹H spectrum for 3. Similarly, inspection of the ¹³C NMR spectrum for 3 showed a lack of midfield resonances for a methoxy methyl (δ 59.6, C-7′ in 2), anomeric carbon (δ 94.8, C-1′ in 2), and other glycosidic carbons. Together, the MS and 1D NMR data for compound 3 indicated that the structure of 3 is related to the aglycone of mandelalide B (2, Table 2). Indeed, the only difference between the ¹³C NMR signals for the aglycones of compounds 2 and 3 was a relatively downfield methine resonance at δ 72.2 in the spectrum for 3, compared to the methylene signal at δ 37.9 (C-24) in the spectrum for 2. This methine was HSQC-correlated to a 1H doublet at δ_H 3.98, and could be assigned to an oxymethine CH-24 on the basis of HMBC correlations from the latter ¹H doublet to C-1, C-2, C-3 and C-23. Thus, considering its MS and NMR data, mandelalide C (3) could be assigned as the C-24 hydroxylated aglycone of mandelalide B (2). Unfortunately, an attempt to crystallize mandelalide C (0.5 mg) was unsuccessful and degradation of the compound over several months prevented the acquisition of additional spectroscopic or biological data.

The molecular formula of mandelalide D (4) was assigned by HR-ESI-MS ([M+Na]⁺ m/z 659.3326) as C₃₄H₅₂O₁₁, implying 9 degrees of unsaturation. Comparison of the similar ¹H NMR spectra (Table 2) for mandelalides C (3) and D (4) indicated a second aglycone structure. A key difference between the ¹H spectra for the two compounds was the further deshielded H-24 doublet (δ_H 5.17) for 4 relative to that for 3 (δ_H 3.98). In the HMBC spectrum for 4, the H-24 doublet was correlated to a carbonyl ¹³C resonance (δ_C 173.1, C-1″), which also showed correlations from H-2a”, H-2b” and H-3″, consistent with an additional butyrate substituent at C-24. Thus, mandelalide D (4) is 24-butyro mandelalide C (3). Over a period of 12 months, we observed changes in the ¹H NMR spectrum of 4 that were consistent with the loss of both butyrate moieties, resulting in the new compound deacylmandelalide D (4b, m/z 519.3 [M+Na]⁺, Table S6).

Consideration of the differences in the planar macrocyclic structures, and available amounts, of mandelalides A-D (1-4) lead to the conclusion that the assignment of relative configuration should be performed on mandelalide A and one of mandelalides B-D. Similarities in coupling constant (³J_HH) values for mandelalides B-D suggested retention of the relative configuration of the macrocycle between these structures (Table 2). Therefore besides mandelalide A, mandelalide B (2) was selected for further analysis because it also possesses a monosaccharide moiety, although it lacks the C-24 chiral center present in 3 and 4. For both mandelalides A (1) and B (2), the relative configuration of the macrocycle could be assigned fully using a combination of ROESY and J-based configuration analysis of data acquired in CDCl₃ and pyridine-d₅, which were consistent between the two solvents.

Considering the tetrahydropyran ring of 1, ROESY correlations between H-5, H-7 and H-9 supported their axial orientation, consistent with a chair conformation and equatorial C-7 glycosidic bond (Figure 2a). The relative configuration for fragment C-9 to C-11 (-CHCH₂CH-) was assigned on the basis of coupling constants obtained from the ¹H spectrum and a DQFCOSY. In CDCl₃, ³J_HH values of 2.2 and 11.2 Hz for H-9/H-10a and H-9/H-10b, respectively, localized H-10a gauche and H-10b anti to H-9 in two possible rotamers (Figure 2b), the correct one of which should be distinguishable from ROESY data. ¹H signal overlap in the CDCl₃ ROESY spectrum for mandelalide A (1) obscured ROE interactions in the region of interest. However, acquisition of NMR data for 1 in py-d₅ provided sufficient resolution (0.16 ppm, Figure S20) between H-8_ax and H-10a to reveal ROESY correlations between H-8_ax and H-10b, and H-8_eq and H-10a. This lead to the conclusion that rotamer A is correct, orienting H-10a to the outside of the macrocycle, (Figure 2). The assignment of relative configuration around the C-10/C-11 bond required direct measurement of ³J_HH values from DQFCOSY, again due to significant overlap. In CDCl₃, ³J_HH values of 9.6 Hz for H-10a/H-11 and 3.7 Hz for H-10b/H-11 indicated anti and gauche orientations, respectively (Figure 2c). Finally, ROESY correlations between H-9 and H₃-25, and H-11 and H-13 were consistent with the relative configuration shown in Figure 2a. This assignment was supported by HETLOC data in CDCl₃ (Figures S11), which provided very weak, but measurable ³J_HC between H-10a and C-25 (0.9 Hz), and between H-10b and C-25 (8.9 Hz).

For the tetrahydrofuran ring of 1, H-17 and H-20 could be in a cis or trans configuration. There are two close conformational minima (envelope and half-chair) for THF rings.¹⁵ However, in the envelope conformation CH₃-26, H-18, H-19a and H-19b in 1 would be eclipsed, while these same moieties would be gauche in the half-chair conformation and there would be no eclipsed interaction along the C-18/19 bond.¹⁵ Therefore we decided to analyze ROESY correlations in the context of a half-chair THF ring to assign the relationship of H-17 and H-20. The relatively large ³J_HH couplings between vicinal proton pairs H-16b/H-17 (11.4 Hz) and H-20/H-21 (8.8 Hz) supported their anti orientation in each case (Figure 3a). Moreover, ROESY correlations between H-16b and H-19a, H-16b and H-21, and H-19a and H-21 were consistent with a cis localization of H-17 and H-20 (Figure 3b), despite that no direct ROESY correlation was observed between H-17 and H-20. Instead, the proposed conformation, consistent with ROESY correlations between H₃-26/H₂-16 and H-16b/H-21, brings ROE-correlated H-18 and H-20 into close proximity (Figure 3b). For the C-21 chiral center, the large values of ³J_HH and ²J_HC (8.8 and -4.0 Hz, respectively), in parallel with the lack of NOE contact between H-20 and H-21 (resolved in py-d5), are consistent with their anti orientation (Figure 3a), as stated above. Analysis of ³J_HH between H-21/H-22a (11.1 Hz), and H-21/H-22b (1.8 Hz), placed H-22a anti and H-22b gauche with respect to H-21 (Figure 3c). ROESY correlations between H-19a and H-22b, H-19b and H-22a, and H-19b and H-22b suggested that H₂-22 are directed away from the center of macrocycle, with the C-21 hydroxyl pointing towards the center. Analogous reasoning was used to resolve the relative configuration at C-23. The respective large and small ³J_HH values between H-22b/H-23 (11.7 Hz) and H-22a/H-23 (1.9 Hz) suggested anti and gauche orientations, respectively (Figure 3d).

Figure 3.

a) Newman projections along the C-16/C-17 and C-20/C-21 bonds in 1; b) The most feasible cis orientation of the THF ring with double-headed arrows indicating ROESY correlations; c) Newman projection along the C-21/C-22 bond with applicable J_HH and J_HC values (CDCl₃) indicated on the right. d) Newman projection along the C-22/C-23 bond and relevant coupling constants.

Configurational assignment of the mandelalide A monosaccharide (C-1′ to C-6′) relied on ³J_HH values, obtained from the ¹H NMR and assigned by DQFCOSY, given the overlap for H-3′ to H-5′ signals in both CDCl₃ and py-d₅ ROESY spectra. Although the small value of ³J_{H-1′/H-2′} = 1.1 Hz was inconclusive, H-2′ was assigned as equatorial based on ³J_{H-2′/H-3′} = 3.8 Hz. The H-4′ multiplet indicated couplings of 9.4 Hz with both H-3′ and H-5′, indicative of axial hydrogens, although the direct analysis of H-3′ and H-5′ multiplets was impeded by their overlap with other ¹H shifts. Overall, these data for CH-2′ to CH-5′ were consistent with the relative configuration of 2-O-methylrhamnose. To confirm this and to assign the monosaccharide as α- or β-2-O- methylrhamnose, we decided to analyze the magnetization transfer pattern of TOCSY signals originating from the anomeric center (C-1′), following the method proposed by Gheysen et al.¹⁶ This “TOCSY matching” approach takes advantage of the fact that the size of the ³J_HH scalar couplings affect the rate of magnetization transfer through a ¹H spin system during the TOCSY spin-lock period. The acquisition of a TOCSY with a spin-lock of 100 ms (Figure S9) affords a differential presence or absence of the CH₃-6′ signal in α- and β-rhamnose, respectively. In the TOCSY acquired for mandelalide A (1), the intensity of peaks fit to previously reported data for α-rhamnose (Figure S9).¹⁶ Additional confirmation that the monosaccharide attached to 1 is 2-O- methyl-α-rhamnose was provided by the measurement of ¹J_HC = 167.2 Hz for the anomeric CH- 1′ from a HETLOC experiment (Figure S12). This value is in agreement with that reported previously for 2-O-methyl-α-rhamnose, although is smaller than in the case of nonmethylated common sugars (~170 Hz).^17,18

As in the case of 1, the coupling constants values and ROESY correlations measured from both CDCl₃ and py-d₅ data for mandelalide B (2) were comparable, indicating similar conformations in both solvents. The chair conformation of the THP ring in 2 (C-5 to C-9) was confirmed by the presence of diaxial ROESY correlations between H-5, H-7 and H-9 (Figure 4a). For the C-9/C-11 segment, coupling constants of 2.2 and 11 Hz, between H-9/H-10a, and H-9/H- 10b confirmed gauche and anti relationships, respectively. Similarly for H-10a/H-11 and H- 10b/H-11, large (12.3 Hz) and small (4.1 Hz in CDCl₃, 3.8 Hz in py-d₅) coupling constants assigned from DQFCOSY supported anti and gauche relationships, respectively, overall leading to the same relative configuration in this region as for 1. For the C-3/C-5 fragment of 2, ³J_HH values of 1.5 and 11.7 Hz between H-4a/H-5 and H-4b/H-5 supported gauche and anti orientations, respectively. A small coupling of 1.6 Hz between H-3 and H-4b indicated their gauche relationship (Figure 4b), and was in agreement with large H-4a/C-3 (-5.6 Hz) and small H-4b/C-3 (0 Hz) heteronuclear couplings. However, medium couplings for H-3/H-4a (6.0 Hz) and H-5/C-3 (4.3 Hz) suggested that the staggered orientation along C-3/4 may be distorted or that more than one conformation is present along this bond. Clear ROESY correlations between H-3 and each of H-4b, H-5, H-9, H-11, H-23 and H-24b, but not H-4a (apparent COSY artifact only), support a distorted rotamer in which 138° and -102° angles are present between H-3/H-4a and H-3/H-4b, as calculated using MestReJ for ³J_HH 6.0 and 1.6 Hz respectively (Figure 4b).

Figure 4.

a) The most probable rotamer along the C-3/C-4 bond highlighted by the dashed outline and alternative equilibrium of two rotamers explaining medium size coupling constants. b) Key ROESY correlations for the C-3/C-5 fragment are indicated by double-headed arrows on the partial structure. But = butyrate.

The presence of a ROESY correlation between H-18 and H-20 on the THF ring in mandelalide B (2) placed CH₃-26 quasi-equatorial, consistent with ROESY correlations from the latter methyl to H₂-16, (as for 1, Figure 3). ROESY correlations (Table S2) and comparative analysis of coupling constant values (Table 2) around the C-16/C-21 fragment confirmed that a cis conformation of the THF ring is conserved in mandelalides A (1) and B (2).

Given extensively overlapped CDCl₃ ¹H chemical shifts for the C-20/C-23 fragment in 2, analysis of ¹H NMR and DQFCOSY in py-d₅ was used to assign the relative configuration of this region. The ³J_HH values for 2 were similar to those for mandelalide A (1), despite the presence of the γ-lactone ring in 2. However, in contrast to 1, in which the more shielded H-22a (δ 1.46) is anti to H-21 and gauche to H-23 (Figure 3), for 2 the more shielded H-22a (δ 1.65) is gauche to H-21 (³J_HH <1 Hz) and H-22b (δ 1.77) is anti to H-21 (³J_HH = 9.4 Hz), while H-20 and H-21 remain antiperiplanar (³J_HH = 9.4 Hz) (Figure 5a). Coupling constant values obtained directly from the oxymethine H-23 multiplet, and assigned by DQFCOSY, indicate anti relationships with H-22a (J_HH = 11.2 Hz), and a gauche relationship with H-22b (J_HH = 1.0 Hz). The orientation of the ester bond in the butyrolactone ring could be assigned knowing the localization of H-23 anti to H-22a. The ester oxygen must then be positioned gauche or anti to H-22b. The ROEs observed between H-24a, H-24b and H-22b in py-d₅, in parallel with the absence of ROEs between H₂-24 and H-21 suggested that the ester bond is anti to H-22b (Figure 5a). Finally, knowing the relative configuration of the C-3/C-11 and C-20/C-23 fragments, the chirality of the quaternary C-2 in the butyrolactone ring was considered. ROESY correlations between H-24b and both H-3 and H-23 were consistent with an α-oriented H-24b, while H-24a, showing ROESY correlations to H₂-22, is oriented above the butyrolactone ring in a β configuration. Similarly, the C-2 hydroxyl could be oriented to the outside of the macrocycle due to the presence of weak ROESY signals between this OH-2 and H-4b and H-24a, suggesting a 2R configuration (Figure 5b).

Figure 5.

a) Assigned rotamers along the C-21/C-22 and C-22/C-23 bonds in 2. b) ROESY correlations around the γ-butyrolactone of 2. But = butyrate.

The relative configuration of the 2′-O-methyl, 6′-deoxy monosaccharide moiety (C1′-C7′) in 2 was established analogously to that in 1. A very weak ROESY correlation between H-1′ and H-5′ together with ¹J_HC = 167 Hz for anomeric CH-1′ suggested the presence of an α-sugar. A ROESY correlation between H-3′ and H-5′ localized these protons axial, while the small ³J_HH between H-2′ and H-3′ (2.8 Hz) localized H-2′ equatorial (Figure 6). The presence of a W coupling (⁴J_HH = 0.98 Hz) between H-2′ and H-4′, as well as ³J_HH’ = 3.2 Hz between H-3′ and H- 4′ localized H-4′ equatorial, overall leading to the conclusion that mandelalide B (2) contains 2- O-methyl-6-dehydro-α-L-talose, the C-4′ epimer of the monosaccharide in mandelalide A.

Figure 6.

a) Key ROE correlations (CDCl₃) between the monosaccharide and the macrolide THP moiety in mandelalide A; b) Key ROE correlations between the monosaccharide and the macrolide, as well as through ring correlations for mandelalide B.

Given the assignment of relative configuration of the macrocycle and monosaccharide units of mandelalides A (1) and B (2), it remained to establish the absolute configuration of these glycosidic macrolides. Despite our desire to conserve the limited sample quantities available for further biological investigations, a chemical degradation/derivatization approach to the absolute configuration would be most rigorous and was facilitated by the contiguous nature of the stereogenic fragments throughout the molecular framework. Importantly, inspection of the CDCl₃ and py-d₅ ROESY spectra for 1 and 2 revealed clear correlations between protons of the monosaccharide and THP of the macrocycle (Figure 6a, 6b). Given the well-defined solution structures of monosaccharides and glycosidic bond conformations,¹⁹ assignment of the rhamnose absolute configuration would permit subsequent relay of configurational assignments around the macrocycle, as shown for mandelalide B (Figure 6b). Therefore, a portion (100 μg) of mandelalide A (1) was sacrificed for hydrolysis of the glycosidic bond and chiral GC-MS comparison of the liberated derivatized monosaccharide with permethylated and silylated D- and L-rhamnose standards. During co-injection of each standard with the derivatized natural product hydrolysate, the natural product sugar coeluted with the L (97.4 min) and not the D (96.4 min) synthetic standard, confirming that the sugar substituent of 1 is 2-O-methyl-α-L-rhamnose. The absolute configuration of this sugar could then be extrapolated to the aglycone of 1 based on the ROESY correlations between H-1′ and H-7, H-1′ and H-6_eq, and H-5′ and H-8_eq to provide an assignment of the mandelalide A (1) aglycone as 2E, 5S, 7S, 9R, 11R, 12E, 14Z, 17R, 18R, 20R, 21R, 23R. Given retention of the macrocycle configuration (C-5 to C-21) between 1 and 2, the absolute configuration of 2 could be assigned similarly by considering ROESY correlations between the monosaccharide and THP moiety, and key correlations across this more rigid macrocycle (Figure 6b). Thus the aglycone of mandelalide B (2) could be assigned as 2R, 3R, 5R, 7S, 9R, 11R, 12E, 14Z, 17R, 18R, 20R, 21R, 23R.

Computational modeling of mandelalide B was used to examine possible mandelalide B conformations (Figure 7). An attempt to investigate the fit of calculated conformations to experimental data was based on measurements of ROE distances across the macrocycle in py- d₅.²⁰ However, as could be expected for a flexible macrolide, for the ten lowest energy conformations of mandelalide B (2), the differences between the calculated and experimental average distances between all analyzed protons except H-21/H-11 and H-21/H-13 were significantly different (0.2 Å or more). Nevertheless, the lowest energy computational model assists in visualizing the through ring ROE contacts that facilitated assignment of the absolute configuration (Figure 7).

Figure 7.

Computational model of the lowest energy conformer for mandelalide B (2).

The ³J_HH values acquired in CDCl₃ for mandelalides B-D (2-4) were all similar, suggesting that the relative configuration of the macrocycle incorporating a γ-butyrolactone is retained. The configuration at the remaining C-24 stereocenter present in mandelalides C (3) and D (4) was assigned from ROESY data. In each case, the H-24 doublet (δ_H 3.98 and 5.17 in 3 and 4, respectively) is correlated to the H-3 double doublet, localizing the C-24 hydroxyl/butyrate moiety antiperiplanar to H-3 and H-23, in an S configuration.

Cytotoxicity of the organic extract from Lissoclinum sp. was examined against mouse Neuro-2A neuroblastoma, and human MDA-MB-231 breast and NCI-H460 lung cancer cell lines following 48 h exposure. In all cases low μg/mL IC₅₀ values were obtained (Table 3). The Neuro-2A cell line was chosen to perform activity-guided fractionation leading to the purification of mandelalides A–D (1-4). The pure compounds mandelalides A and B yielded nanomolar IC₅₀ values against Neuro-2A (44.0 and 83.8, respectively) and NCI-H460 (12.0 and 29.4 respectively) cell lines (Table 3, Figure S57). The potent cytotoxicities of these mandelalides are somewhat surprising given the reported minimal cytotoxicity of the related glycosylated polyketides madeirolides A and B.²¹ The latter metabolites caused less than 50% inhibition of AsPC-1 and PANC-1 pancreatic cancer cells at 10 μg/mL, but showed potent fungicidal activity against Candida albicans. Madeirolides A and B share essentially the same western hemisphere of the mandelalides (C-5 to C-22), including a THP (alternatively substituted), diene, THF and neighboring hydroxymethine. The two series of compounds vary only in the closure of the macrocycle: the position of the lactone and additional cycle. Given that the madeirolides were isolated from a deep-water lithistid Leiodermatium sponge, their structural relatedness is consistent with a microbial biogenetic origin. Side-by-side evaluation of the biological properties of all four mandelalides and the two madeirolides would provide insight into the structure-activity relationships that result in potent cytotoxicity and fungicidal activity, respectively. However, the inaccessible supply of both source organisms means that further investigation of these metabolites will likely await their total syntheses.

Table 3.

Cytotoxicity of Lissoclinum sp. organic extract and mandelalides A (1) and B (2) to mouse Neuro-2A neuroblastoma and human NCI-H460 lung and MDA-MB-231 breast cancer cells.

	aIC50 (at 48 h) in Cancer Cell Lines:
	NCI-H460	Neuro-2A	MDA-MB-231
organic extract	0.7 μg/mL	5.6 μg/mL	22.1μg/mL
mandelalide A (1)	12 nM	44 nM	-
mandelalide B (2)	29 nM	84 nM	-

^aCell viability was assessed by MTT assay and IC₅₀ values were derived using nonlinear regression analysis.

Experimental Section

General Experimental Methods

were as described previously.²² Additionally NMR data in CDCl₃ were acquired at 700 (¹H) and 175 (¹³C) MHz on a 5 mm inverse cryogenic probe. The NMR data in py-d₅, TOCSY (mixing time 100 ms), HETLOCs and gated-decoupled ¹³C NMR experiments were obtained at 700 (¹H) and 175 (¹³C) MHz on a 5 mm ¹³C cryogenic probe. The spectra were referenced to internal residual solvent signals in ppm (¹H NMR: CDCl₃, 7.24, py-d₅, 8.74, CD₃OD, 3.31; ¹³C NMR: CDCl₃, 77.23, py-d₅, 150.35, CD₃OD, 49.15). High-resolution MS data were acquired using an orthogonal acceleration time of flight (oa-TOF) mass analyzer and electrospray ionization (ESI).

Extraction and Isolation

The ascidian, Lissoclinum sp. (Ascidiacea, Aplousobranchia, Didemnidae) was collected by hand using SCUBA at a depth of 18 m (July 20, 2004) from White Sands Reef in Algoa Bay, Eastern Cape Province, South Africa (33:59.916S, 25:42.573W). The type specimen for this new ascidian species is housed at the South African Institute for Aquatic Biodiversity (SAIAB), Grahamstown, South Africa.

South Africa

The freeze-dried organism (15.1 g) was extracted with 2:1 CH₂Cl₂-MeOH yielding 1.45 g of organic extract. This organic extract was fractionated on Sephadex LH-20 (CH₂Cl₂- MeOH, 1:3) to give eight fractions, of which fractions six and seven were subjected to reversed phase C₁₈ solid phase extraction (RP-SPE) using a stepped gradient of 50-100% MeOH in H₂O. The 75% MeOH-H₂O and 100% MeOH fractions were further separated by RP-HPLC (C₁₈ column, 250 mm x 10 mm, 7:3 MeOH-0.1% FA in H₂O) to yield mandelalides A (1, 0.8 mg), B (2, 0.5 mg), C (3, 0.8 mg) and D (4, 0.6 mg).

Mandelalide A (1)

Amorphous solid; [α]²³_D –9 (c = 0.25, MeOH); UV (MeOH) λ_max (log ε) 279 (2.7), 217 (4.1); LR-ESI-MS m/z [M+Na]⁺ 647.4; HR-ESI-MS m/z [M+Na]⁺ 647.3394, (calcd for C₃₃H₅₂O₁₁Na, 647.3407); ¹H and ¹³C NMR, COSY, HMBC, TOCSY, ROESY (Tables 1, S1).

Mandelalide B (2)

Amorphous solid; [α]²⁴_D –13 (c = 0.5, MeOH); UV (MeOH) λ_max (log ε) 279 (3.0), 229 (4.1); LR-ESI-MS m/z [M+Na]⁺ 733.5; HR-ESI-MS m/z [M+Na]⁺ 733.3773, (calcd for C₃₇H₅₈O₁₃Na, 733.3775); ¹H and ¹³C NMR, COSY, HMBC, ROESY (Tables 2, S2, S3).

Mandelalide C (3)

Amorphous solid; UV (MeOH) λ_max (log ε) 280 (2.7), 229 (4.0); LR-ESI- MS m/z [M+Na]⁺ 589.5; HR-ESI-MS m/z [M+Na]⁺ 589.2970, (calcd for C30H46O10Na, 589.2989); ¹H and ¹³C NMR, COSY, HMBC, TOCSY, ROESY (Tables 2, S4).

Mandelalide D (4)

Amorphous solid; [α]²⁵_D –50 (c = 0.2, MeOH); UV (MeOH) λ_max (log ε) 280 (2.7), 229 (4.0); LR-ESI-MS m/z [M+Na]⁺ 659.5; HR-ESI-MS m/z [M+Na]⁺ 659.3424, (calcd for C₃₄H₅₂O₁₁Na, 659.3407); ¹H and ¹³C NMR, COSY, HMBC, TOCSY, ROESY (Tables 2, S5).

Deacylmandelalide D (4b)

Amorphous solid; [α]²⁷_D –10 (c = 0.2, MeOH); UV (MeOH) λ_max (log ε) 280 (2.8), 227 (3.8); LR-ESI-MS m/z [M+Na]⁺ 519.3; HR-ESI-MS m/z [M+H]⁺ 497.2752, (calcd for C₂₆H₄₁O₉, 497.2751); ¹H and ¹³C NMR, COSY, HMBC (Table S6).

Measurement of ^1,2J_HC coupling constants

The sensitivity and gradient-enhanced HETLOC (ω1-hetero half-filtered TOCSY) experiment was employed to measure J_HC coupling constants, with DIPSI-2 spin-lock set to 60 ms.²³ Spectral widths of 6229 Hz and 5597 Hz, with a data matrix of 4K (F2) x 128 (F1), and 146 or 96 scans were employed in py-d₅ and CDCl₃, respectively for mandelalide A. In the case of mandelalide B, a spectral width of 5597 Hz with a data matrix of 4K (F2) x 128 (F1) and 136 scans were implemented in CDCl_3. The 1D spectra that were obtained after extraction of F2/ F1 slices were subjected to inverse Fourier transform. The resulting FIDs were multiplied by the exponential window function prior to linear prediction processing.²⁴

TOCSY data acquisition

To determine the magnetization transfer pattern in 2-O-methyl-L-rhamnose a 2D TOCSY was acquired with a DIPSI-2 spin-lock sequence in CDCl₃. The spectra were recorded with a spin-lock mixing time of 100 ms, 5597 Hz spectral width at 2K (F2) x 512 (F1) data matrix for 40 scans. Data were processed with a sine-bell squared function with 1.5 Hz (F2) and 0.3 Hz (F1) line broadening before measurement of the absolute volumes of peaks. The following designation was applied to depict intensity of peaks; grey ovals indicate intensity or signal greater than 1.5% of intensity measured for the signal of the anomeric proton, white ovals indicate intensity of at least 0.5% but no more than 1.5% intensity measured for the anomeric ¹H signal.¹⁶

Calculation of Dihedral Bond Angles in MestReJ

²⁵ The calculation of angles between H-3/H- 4a and H-3/H-4b was performed using the Altona equation with substituents at C-3 defined as CH₂OR and OCOR, and substituents at C-4 as H and CH₂OR. Resulting ³J_HH values for H-3/H- 4a (138.7°) and H-3/H-4b (-102.2°) were 6.0 and 1.6 Hz, respectively.

Computational Modeling of Mandelalide B (2)

Computational modeling was performed using the 2009 version of a contemporary software package. Minimization using the Amber* force field with PRCG algorithm, in pyridine (ε = 12.9, 10000 steps, maximum derivative less than 0.05 kcal/mol) and constrained torsion angles H-3/H-4a (138°), H-3/H-4b (-102.2°), H-4b/H-5, H-9/H-10b, H-20/H-21, H-21/H-22b, H-22a/H-23 and distance H-4b/H-24b = 5 Å was first performed. All torsions were restrained based on the ³J_HH coupling constants values. The minimized structure was subjected to conformational search using Amber* force field via the low mode sampling method with an energy cut off of 21 kJ/mol and 1000 steps (100 steps per rotatable bond).^26,27 Torsion restraints H-3/H-4a (138°), H-3/H-4b (-102.2°), H-4b/H-5, H-9/H- 10b, H-20/H-21, H-21/H-22b, H-22a/H-23 were applied to obtain ground state conformations. Optimization of the ten lowest energy conformations using DFT with the B3LYP functional and 6-31G** basis set in the gas phase resulted in the structure presented in Figure 7.

Methanolysis of Mandelalide A (1)

Compound 1 (0.1 mg) was treated with 1 N methanolic HCl (1.0 mL) and heated for 24 h at 70 °C with stirring. The mixture was concentrated in vacuo, redissolved in 0.75 mL of 1-(TMS)-imidazole/py (1:4) and the reaction was continued for 40 min at 70 °C with stirring.²⁸ The solution was concentrated, the final residue partitioned between CH₂Cl₂ and H₂O (1:1), and the organic fraction used for GC-MS analysis.

Preparation of 1,2-di-O-methyl-3,4-di-O-TMS-α-L-rhamnose and 1,2-di-O-methyl-3,4-di-O- TMS-α-D-rhamnose standards

L-rhamnose (50 mg) was dissolved in 1N methanolic HCl, heated for 24 h at 70 °C with stirring, and concentrated in vacuo. The residue was redissolved in dioxane (0.6 mL) and 2 M TMS-diazomethane in diethyl ether (1.5 mL) was added, with H₃BO₃ as catalyst. The solution was maintained at room temperature with stirring for 5 h and then evaporated to dryness in vacuo.²⁹ Column chromatography on Si gel GF (15 μm) with 5% MeOH in CH₂Cl₂ as mobile phase afforded a 1:3 mixture of 1,2-di-O-methyl-α-L-rhamnose and 1,3-di-O-methyl-α-L-rhamnose iñ65% yield from this rate limiting reaction: equivalent to ~16% yield of the desired 1,2-di-O-methyl-α-L-rhamnose to be carried forward for silylation. This 1:3 mixture of methylation products was treated with 1-(TMS)-imidazole/py (1:4) for 40 min at 60 °C and concentrated in vacuo.²⁸ Finally, the residue was dissolved in CH₂Cl₂, loaded on a Si gel column, washed with 100% hexanes, and eluted with 3% EtOAc in hexanes to obtain pure 1,2-di-O-methyl-3,4-di-O-TMS-α-L-rhamnose in a final yield of 5.4%. The same procedure was applied to produce 1,2-di-O-methyl-3,4-di-O-TMS -α-D-rhamnose in 6.4% yield. ¹H NMR (700 MHz, CDCl₃): δ 0.12 (s, 9H), 0.15 (s, 9 H), 3.31 (s, 3H), 3.32 (dd, J = 1.7, 3.2 Hz, 1H), 3.48 (s, 3H), 3.52 (m, 1H), 3.53 (m, 1H), 3.78 (dd, J = 3.6, 8.4 Hz, 1H), 4.61 (d, J = 1.3, 1H); ¹³C NMR (175 MHz, CDCl₃): δ 0.7, 1.2, 18.5, 54.8, 60.0, 69.1, 73.1, 74.3, 81.6, 98.8, ppm; CI-LR- MS: m/z 305 calculated for oxonium ion C₁₃H₂₉O₄Si₂.

Absolute configuration of 2-O-methyl-α-rhamnose from mandelalide A (1)

Analyses of the synthetic standards and permethylated and silylated mandelalide A hydrolysate were performed by GC-MS using a 30 m × 0.25 mm i.d. Cyclosil-B column, and electron impact (EI) ionization. The injector and detector were operated at 250 °C, while the temperature gradient was set to 75 – 175 °C at 0.5 °/min. The retention times for 1,2-di-O-methyl-3,4-di-O-TMS-α-D-rhamnose and 1,2-di-O-methyl-3,4-di-O-TMS-α-L-rhamnose injected separately were 96.5 and 98.3 min, respectively. The retention time for the monosaccharide in the derivatized mandelalide A hydrolysate injected separately was 97.5 min. (Figure S56). Therefore co-injection of the natural product hydrolysate with each standard was performed, yielding two peaks with retention times of 96.4 and 97.4 min, which is indicative of the L glycoside in mandelalide A (Figure S56).

Cell viability assays

Cytotoxicity of the organic extract and crude fractions was evaluated in mouse Neuro-2a neuroblastoma, and human MDA-MB-231 breast and NCI-H460 lung cancer cells (ATCC, Manassas, VA) using a previously described protocol subjected to slight modifications.²² Cells were seeded into 96-well plates (20,000 cells per well for MDA-MB-231 and NCI-H460, 25,000 cells per well for Neuro-2A) in 50 μL of medium twelve hours before treatment. Each test sample was added in a 25 μL aliquot generated by serial dilution in serum-free medium on the day of the experiment, after prior removal of 25 μL of media from the treated well. Aliquots were generated from stock solutions of 6 mg/mL compound in 100% DMSO. Pure compounds 1 and 2 were evaluated in Neuro-2A and NCI-H460 cells after 48 h treatment, as described above. Each compound was tested at final concentrations ranging from 0.001 μg/mL to 10 μg/mL. In all cases, cell viability was determined after 48 h treatment using a standard 3-(4,5-dimethylthiazol-2-yl)-2,5,diphenyl tetrazolium bromide (MTT) assay.²² The cytotoxicity of each pure compound was assessed in at least three independent cultures with the viability of vehicle-treated control cells defined as 100% in all experiments. Dose response curves (Figure S57) were plotted using contemporary biostatistics and curve fitting software.