Synthesis and Stability Studies of a Simplified, Thiazole-containing Macrocycle of the Anticancer Agent Salarin C

Abstract

While SalaC is a potent marine cytotoxin, Kashman demonstrated that congeners which had undergone Wasserman rearrangement exhibit little to no cytotoxicity. Given that thiazoles are known to undergo Wasserman rearrangement at a significantly reduced rate, we hypothesized that a thiazole-containing SalaC would exhibit greater stability without significantly altering the macrocyclic conformation. Herein, we describe the synthesis of a simplified, thiazole-containing macrocycle which demonstrates significantly improved stability under identical aerobic conditions.

Graphical Abstract

The salarins are a family of macrocyclic, marine natural products isolated in 2007 by Kashman and coworkers from the Fascaplysinopsis sp. sponge collected in Salary Bay, Madagascar.^1-6 In addition, three other families of nitrogenous macrolides, tulearins⁵, taumycins³, and tausalarins, have also been isolated from this sponge.⁶ Bioactivity-guided purification of the CHCl₃-MeOH extract using a brine shrimp assay led to the initial isolation of two salarins, A (SalaA) and B (SalaB).¹ These natural products possess a highly unsaturated N-acetyl imide within a macrocycle and one or two epoxides, including an allylic epoxide in the case of SalaA, and an N-acetyl carbamate moiety. Another collection of this sponge revealed the presence of a related compound salarin C (SalaC),² which showed very potent inhibition of cell proliferation. SalaC contains an oxazole ring in place of the N-acetyl imide found in SalaA (1). Finally, Kashman also isolated seven additional salarins (D-J) in 2010,⁴ which were related but distinct congeners of the previously isolated family members (Figure 1).

Figure 1.

Structures of salarins A-J (SalaA-J) with SalaD-J represented as partial structures. The relative stereochemistry of diols and chlorohydrins in SalaG, H, F, and I is unknown.

SalaC (2) exhibits the most potent cytotoxic activity of all salarins with an EC₅₀ of 0.1 μM and 1 μM toward a chronic myelogenous leukemia (K562) and an acute myeloid leukemia (UT-7) cell line, respectively. SalaC induces cell cycle arrest in the G2/M phase and leads to induction of apoptosis.⁷ All other members of the salarin family devoid of either the allylic C16/17-epoxide, such as SalaB, F, I, or the oxazole-containing macrocycle, such as the rearranged N-acetylimide SalaA, D, E, exhibit greatly reduced or no cytotoxicity. Furthermore, Kashman also conducted a series of derivatization experiments of SalaC and found that all derivatives involving C16/17 epoxide cleavage led to loss of cytotoxicity.⁸ Taken together, this preliminary SAR data suggest that the presence of the allylic C16/17-epoxide and the overall conformation of the oxazole-containing macrocycle are essential for bioactivity.

SalaC exhibits a pronounced instability to ambient singlet oxygen (¹O₂) which leads to conversion of the SalaC oxazole to the N-acetyl imide found in SalaA (Scheme 1).^9,10 Interestingly, the concentration of ambient ¹O₂ can be as high as 20 ppb during the daytime in the outdoors.¹¹ While we briefly considered the possibility that SalaC could itself be a photosensitizer, we concluded that this was unlikely due to SalaC’s reported λ_max (329 nm, MeOH)² being much lower than most common photosensitizers (~560-660 nm, e.g. rose bengal and methylene blue).¹² Kashman proposed that this occurs through a Wasserman rearrangement,² a transformation enabling the use of oxazoles as latent activated carboxyl groups.¹⁰ Lindel synthesized a linear conjugated oxazole fragment structurally related to that found in the macrocycle of SalaC and determined that it was stable to aerobic conditions.¹³ However, upon exposure to photochemically-generated ¹O₂, the corresponding N-acetyl imide was obtained. Subsequently, Altmann reported synthetic studies towards SalaC and observed that synthetic intermediates became reactive toward ¹O₂ once the macrocycle is formed. However, they were able to synthesize dideoxy-SalaC devoid of the allylic C16/17-epoxide.^14,15 Molecular modeling, cross-ring nOe’s observed by Kashman² (H4→H10, H12→H16, H4→H16) and the expected planarity of the dienoate to maximize conjugation, all point to a strained macrocycle in SalaC. We propose that torsional strain induced by the rigid SalaC macrocycle leads to a higher energy HOMO for the conjugated oxazole and greater strain release¹⁶ resulting in a rate acceleration of the Wasserman rearrangement.

Scheme 1.

Aerobic instability of SalaC (2) toward ¹O₂ leading to SalaA (1) via a Wasserman rearrangement.

McNeill demonstrated that peptidic thiazoles are much less susceptible to the Wasserman rearrangement compared to their oxazole counterparts.¹⁷ The increased stability might be expected in part due to increased aromaticity of thiazoles compared to oxazoles, as evidenced by isotropic magnetic shielding distributions¹⁸ and chemical shift differences. We reasoned that a thiazole-containing SalaC macrocycle would be more stable toward ¹O₂. We thus set out to synthesize a simplified macrocycle bearing a thiazole and compare its stability to the corresponding simplified oxazole SalaC macrocycle. The C12/13-trans-epoxide would be replaced with an equivalent, conformationally-biasing trans-alkene. We also reasoned that removal of the SalaC side-chain and C14-oxygenation would have minimal effect on the conformation of the macrocycle. Indeed, computational studies demonstrated that the low energy conformations of the two macrocycles 3a, 3b are quite similar (Figure 2b) with negligible changes in planarity but expected changes in bond lengths and angles between these heterocycles (Figure 2c). This is a common heterocycle interchange performed in medicinal chemistry, thus we anticipated minimal effect on bioactivity of an eventual fully elaborated thia-SalaC variant.¹⁹ While our work was in progress, Lindel reported the synthesis of the identical simplified oxazole macrocycle 3b.²⁰ However, they only studied its reactivity to ¹O₂ under photolytic conditions with rose bengal. Herein, we describe the synthesis of both a simplified, thiazole-containing SalaC macrocycle 3a and the analogous oxazole-containing macrocycle 3b for comparison. The intrinsic reactivity of these macrocycles toward ambient ¹O₂ was also studied under various conditions as a prelude to further synthetic and mechanism of action studies.

Figure 2.

(a) Structures of simplified thiazole- and oxazole-containing macrocycles 3a, b. (b) Overlay of minimized lowest energy conformations (B3LYP/6-31G) of the two macrocycles (oxazole 3b, green; thiazole 3a, gold) showing minimal change in lowest energy conformations. (c) Close-up views of the thiazole and oxazole overlays with two different perspectives.

Our proposed syntheses of the two, simplified macrocycles 3a, 3b employed similar disconnections for simplicity with an eye towards scalable routes that could potentially be modified to access derivatives of thia-SalaC (Scheme 2). Key steps in our synthesis ultimately included formation of the thiazole and oxazole through cyclodehydration and oxidation, a sp²-sp³ Suzuki-Miyaura coupling, successfully utilized by Altmann toward SalaC,¹¹ a cationic Heck reaction to install the required Z,E-dienoate, and a final macrolactonization.

Scheme 2.

Simplification of SalaC and retrosynthesis of thiazole and oxazole-SalaC macrocycles 3a and 3b.

The synthesis of the simplified thiazole macrocycle 3a commenced with (Z)-vinyl iodide 9 (Scheme 3), available in one step with high selectivity (>19:1 Z/E) from commercially available 2-butynoic acid (Scheme 3).²¹ Carboxylic acid 9 was then coupled with threonine benzyl ester by in situ conversion to the acid chloride to deliver amide 10 in 87% yield (>19:1 Z/E). Subsequent treatment with Lawesson’s reagent gave the corresponding thioamide.²² Cyclodehydration to the thiazoline was accomplished by conditions first described by Williams employing Deoxo-Fluor®²³ and later converted to flow format by Ley.²⁴ Flow conditions were employed to access the intermediate thiazoline. Typical conditions to oxidize the thiazoline (BrCCl₃, DBU)²⁵ to the thiazole were unsuccessful due to the base sensitivity of the vinyl iodide, resulting in elimination of HI to reform the alkyne. We instead attempted a much milder radical oxidation developed by Meyers based on the Kharasch-Sosnovsky reaction^26,27 which delivered the desired thiazole vinyl iodide 8a with minimal isomerization of the C8,9-alkene over the 3-step sequence integrity (13:1 Z/E). The known alkyl iodide 7, available in four steps from commercially available β-hydromuconic acid,²⁸ was subjected to lithium-halogen exchange and subsequent borylation with 9-MeO-9-BBN enabling a one-pot Suzuki sp²-sp³ cross-coupling²⁹ with vinyl iodide 8a.^14,30 This coupling proceeded efficiently and without alkene isomerization but was most easily purified following subsequent reduction of the benzyl ester with DIBAl-H and oxidation to deliver aldehyde 12 in 49% overall yield (3 steps).

Scheme 3.

Synthesis of a simplified thiazole-containing SalaC macrocycle 3a.

Wittig olefination and subsequent cationic Heck coupling with known Z-3-iodoacrylic methyl ester 6³¹ delivered the (2Z, 4E)-dienoate 13 in 77% yield proceeding with high stereoselectivity (>19:1, 2Z).³² However, significant isomerization of the C8,C9 alkene was observed under Heck coupling conditions (8:1 Z/E). Careful control of stoichiometry was also essential to preventing additional Heck coupling to the C12, 13-alkene. This side-product could be completely avoided by syringe pump addition of the β-iodo acrylate 6 as a solution in degassed MeCN over 1 h. Following desilylation with TBAF, we serendipitously found that during attempts to saponify the methyl ester with TMSOK,³³ the desired macrocycle could be detected by NMR and thus we set out to optimize this macrocyclization. Altmann also utilized TMSOK for hydrolysis of their seco ester but did not observe macrocyclization, perhaps because their substrate bore a pendant secondary rather than primary alcohol as in our seco ester.¹⁴ We anticipated that the (Z)-C8,9-alkene isomer might undergo macrocyclization at a faster rate than the undesired (E)- alkene. However, following optimization involving use of a large excess of TMSOK, high dilution, and shorter reaction times, two isomeric macrocycles (3a) were obtained in the same ratio of C8,C9-geometrical isomers as the seco ester substrates. Separation of the alkene isomers was accomplished by preparative HPLC for characterization and subsequent stability studies.

For comparative stability studies, we also synthesized a simplified, oxazole-containing SalaC macrocycle 3b (Scheme 4). The synthesis mirrored that of the thiazole variant with a few notable challenges including more facile C8, C9-alkene (SalaC numbering) isomerization. Substantial isomerization of the C8,9-alkene in vinyl iodide 8b was observed during the Karasch-Sosnovsky oxidation of the oxazoline to oxazole (from >10:1 to 2:1 Z/E). We thus studied a number of known oxidation methods for this transformation (MnO₂;³⁴ DDQ; NBS/hν;³⁵ CuBr/peroxide²⁶) which offered no improvement. Several studies were undertaken to minimize isomerization without success thus the inseparable mixture of C8,C9-alkene geometrical isomers (Z/E, 2:1) was taken through the remainder of the sequence with the expectation that separation would be possible upon macrocyclization as with the thiazole macrocycle (vide supra). One outcome of the optimization studies was development of a serviceable, albeit very slow, in situ hydroboration-Suzuki-Miyaura coupling sequence. which, following direct reduction, afforded alcohol 16 in 39% yield (2 steps). Alcohol 16 was oxidized to the corresponding aldehyde with Dess-Martin reagent³⁷ rather than MnO₂ since the latter oxidation conditions led to reduced yields and further C8,9-alkene isomerization. Wittig reaction and Heck coupling of the derived alkene 17 with (Z)-vinyl iodide 6 delivered ester 18 with high selectivity for the (Z)-C4,5-alkene (>19:1). Silyl group deprotection and treatment with TMSOK also led to serviceable yields of the macrocyclic oxazole 3b as a 3:1 mixture of C8,C9-isomeric alkenes. Enrichment of the Z- alkene isomer was achieved through preparative HPLC purification to provide macrocycle 3b (15:1 Z/E; C8,C9) which correlated well to the same macrocycle described by Lindel.²⁰

Scheme 4.

Synthesis of a simplified oxazole-containing, SalaC macrocycle 3b as a comparator.

Toward understanding the stability of these macrocycles when collecting NMR data of synthetic intermediates following macrocyclization and diluting samples for biological studies, we measured half-lives of macrocycles 3a, 3b, and the linear oxazole seco ester 4b as a comparator, under various conditions in CDCl₃ and DMSO-d₆. We also studied the stability of macrocycles in CD₃OD under one set of conditions to approximate water. The rate of degradation of macrocycles 3a,3b and seco ester 4b to both Wasserman rearrangement and isomerization was monitored by ¹H NMR through integration of H_C4 and H_C15 in comparison to an internal standard (1,3,5-trimethoxybenzene). As expected, in an argon flushed (AF), 5 mm regular glass NMR tube (RGT), with storage between NMR analyses on the lab bench (LB), no reaction (NR) was observed for macrocycles 3a, 3b or seco ester 4b in DMSO-d₆ after 21 days. (Table 1, Conditions #1). These conditions were thus not further explored in CDCl₃ or CD₃OD. Likewise, rearrangement was not observed with any compounds in 5 mm amber tubes (AT) in CDCl₃, open to ambient air (OA), with storage between NMR analyses at the lab bench (LB) in DMSO-d₆ (Conditions #2). Wasserman rearrangement, accompanied by alkene isomerization, was first observed when samples were kept in regular glass tubes (RGT), open to the air (OA), and stored at the lab bench (LB) (Conditions #3). In CDCl₃, this occurred only with macrocycles 3a,b with t_1/2 values of ~35 and~42 d, respectively, but not seco ester 4b. On the other hand, in DMSO-d₆, t_1/2 was almost the same for macrocycles 3a and 3b (~21 vs ~24 d, respectively). The difference observed with Kashman’s report of a ~3 day half-life for SalaC, may be due to absence of stirring as performed by Kashman,² which may impact the rate of ambient ¹O₂ absorption however differences in structure cannot be excluded. Interestingly, initial use of 3 mm NMR tubes also prevented Wasserman rearrangement from occurring after 3 days likely due to reduced solvent surface area for absorption of ambient ¹O₂. In CD₃OD, again no rection was observed with the seco ester 4b however the thiazole macrocycle 3a displayed a significantly increased half-life relative to the oxazole 3b (~3 vs ~1 d, respectively). We speculate that the increased life-time of ¹O₂ in deuterated methanol³⁷ leads to increased rates of Wasserman rearrangement which occurs faster with the oxazole. Finally, storage in a regular glass tube, open to air, and on a window sill (not in direct sunlight) unsurprisingly led to the fastest rates of rearrangement/isomerization for the macrocycles 3a and 3b in DMSO-d₆ (t_1/2 ~6 vs ~1 h) and CDCl₃ (t_1/2 ~13 vs 7 h). However, in the case of seco ester 4b, the t_1/2 was ~23 h in CDCl₃ and ~8 h in DMSO-d₆ supporting the notion that macrocyclic strain increases the rate of rearrangement.

Table 1.

Half-lives of a simplified SalaC thiazole-containing macrocyclic thiazole 3a in comparison to an analogous oxazole variant 3b and seco ester precursor 4b^a

	Solvent	t1/2 (days or hours)b
		Conds. #1	Conds. #2	Conds. #3	Conds. #4
		RGT, AF, LB	AT, OA, LB	RGT, OA, LB	RGT, OA, WS
3a	DMSO-d6	NRc	NR	~21 d	~6 h
	CDCl3	NDd	ND	~35 d	~13 h
	CD3ODe	ND	ND	~3 d	ND
3b	DMSO-d6	NR	NR	~24 d	~1 h
	CDCl3	ND	ND	~42 d	~7 h
	CD3ODd	ND	ND	~1 d	ND
4b	DMSO-d6	NR	NR	NR	~8 h
	CDCl3	ND	ND	NR	~23 h

^aSamples of 3a, 3b, and 4b were stored at 22-24 °C between NMR analyses either in a regular (RGT) or amber NMR tube (AT) and either open to air (OA) or flushed with argon and capped (AF). The location of the NMR tube was either on the lab bench (LB) or the windowsill (WS). RGT = regular glass NMR tube; AF = argon flushed; WS = window sill (time on the windowsill during daylight hours); AT = amber tube; OA = open to air; LB = lab bench ~20 ft. from a window.

^bHalf-lives were determined by comparative integration of two ¹H NMR resonances (600 MHz, values provided for d₆-DMSO), namely H_C4 (δ 7.86) and H_C15 (δ 4.37), relative to an internal standard (1,3,5-trimethoxybenzene, δ 6.09). NR = no change after 21 days. ND = not determined.

^cIn CD₃OD, other degradation pathways were observed in addition to Wasserman rearrangement and alkene isomerizations.

In conclusion, substitution of the oxazole for a thiazole in simplified SalaC macrocycles 3a,b leads to significantly reduced rate of Wasserman rearrangement with ambient ¹O₂ in both DMSO-d₆ and CDCl₃ (roughly 6 times and 2 times greater stability, respectively) when exposed to sunlight. Our results support the notion that replacement of the oxazole with a thiazole will decrease the rate of Wasserman rearrangement of SalaC derivatives in these solvents. Conformational studies (DFT) suggest that introduction of sulfur does not perturb the macrocyclic conformation nor planarity around the thiazole and thus should not dramatically impact bioactivity. Our results, along with those previously reported by Lindel, reveal some of the structural features required for the observed reactivity of SalaC toward ¹O₂ and support the proposal that the strained macrocycle and extended conformation of the oxazole likely promotes Wasserman rearrangement with ambient ¹O₂. Studies disclosed herein also suggest ideal ways to store both synthetic samples following macrocyclization and samples for biological assays. A thia-salarin C analog may provide greater aerobic stability facilitating mode of action studies which are currently underway.