Substitution of a triazole for the central olefin in biologically active stilbenes

Abstract

The stilbene moiety is commonly found in natural products and these compounds display an extraordinary range of biological activity. Efforts to derive useful drugs from stilbenes must address the potential liabilities of this structure, including a propensity for cis/trans isomerization. To identify olefin replacements that address this limitation while preserving biological activity we have prepared analogues of two bioactive stilbenes, a pawhuskin and a schweinfurthin, where a 1,2,3-triazole ring formally replaces the stilbene double bond. The new schweinfurthin analogue (23) has been tested for anti-proliferative activity against 60 cell lines, and shows a strong correlation of bioactivity when compared to the compound that inspired its synthesis (22).

Graphical Abstract

The stilbene substructure is a common component in a variety of natural products, many of which display intriguing biological activities.¹ Arguably two of the best known natural stilbenes are resveratrol^2-3 (Figure 1, 1) and combretastatin A4 (2).^4-5 Both of these natural products display significant biological activity, and so they have become the target of numerous synthetic efforts aimed at the natural products themselves or at analogues that might have even more attractive biological properties.

Figure 1.

Some biologically active natural stilbenes.

Our own synthetic efforts have been focused on two other stilbene-containing families of natural products. The pawhuskins (e.g. pawhuskin A, 3), isolated from the purple prairie clover (Dalea purpurea) collected in Oklahoma,⁶ have demonstrated opioid receptor activity, and some synthetic analogues have favored activity at either the delta⁷ or the kappa⁸ opioid receptor. The schweinfurthins (e.g. schweinfurthin A, 4), first isolated from Macaranga schweinfurthii collected in Cameroon,⁹ have significant anti-cancer activity as measured by the NCI 60 cell line screen. The activity that many natural schweinfurthins and their analogues have against CNS-derived cell lines, especially glioblastomas such as SF295, is especially intriguing (cf SI for representative bioassay data of natural schweinfurthin A (4)⁹ and synthetic (+)-3-deoxyschweinfurthin B).¹⁰

In an effort to increase the potency, selectivity, and pharmaceutical utility of representatives of both families of natural products, we have prepared a variety of analogues to identify the functionality important for biological activity. Most of these structures preserved the central stilbene core and decorated or redecorated its periphery. For example, through synthesis and bioassay of seven natural schweinfurthins (A,¹¹ B,¹² C,¹³ E,¹² F (as both enantiomers),¹⁴ G,¹⁵ and vedelianin¹⁶) and ~80 analogues, we have: 1) established the natural absolute stereochemistry; 2) determined that the C-3 hydroxyl group is not essential for significant activity; 3) found that a trans-fused A-B ring system is essential; 4) discovered that a substituent at C-5 capable of hydrogen bonding is important; and 5) observed that a wide variety of modifications can be made on the D-ring without significant loss of activity (Figure 2).

Figure 2.

Exploration of structure-activity correlations on natural and synthetic schweinfurthins.

Our efforts have helped to identify the essential features of the pharmacophores, but for the most part they have overlooked potential problems arising from the stilbene olefin. Those concerns include the possibility of olefin cis/trans isomerization (where the cis olefin is known to have a negative impact on schweinfurthin activity¹⁷), the potential for dimerization, and a variety of others.¹ Recently we reported that replacing the stilbene olefin with an amide unit gave rise to pawhuskin¹⁸ and schweinfurthin¹⁹ analogues with favorable biological activity. For example, the amide analogue of 3-deoxyschweinfurthin B with the carbonyl group attached to the C-ring showed a mean value of 0.21 μM for 50% growth inhibition (GI₅₀) across the 60 cell lines of the NCI bioassay.¹⁹ This was approximately 3-fold more potent than the parent stilbene 3-deoxyschwinfurthin B (0.62 μM), while the isomeric amide with the carbonyl group positioned on the D-ring displayed a 1.1 μM mean GI₅₀ value, corresponding to ~1.8-fold less potency. Because a triazole ring can be viewed as a hydrolytically stable amide replacement favoring either cis²⁰ or trans orientations,^21-23 we now have prepared new variations on these natural products containing a central triazole ring.

To test the hypothesis that a triazole substitution for the stilbene olefin would preserve biological activity, we set our sights initially on a triazole-linked pawhuskin analogue of the same ring substitution pattern as the pawhuskin amide 5 (Figure 3).¹⁸ The shorter synthetic sequences needed to access the two components necessary for a convergent synthesis of the triazole 6 made a pawhuskin-based triazole significantly more accessible as an initial target than a schweinfurthin triazole (vide infra, Schemes 1 and 2).

Figure 3.

An amide-linked pawhuskin analogue 5 and the triazole-linked target 6.

Scheme 1.

Regiospecific route to amine 11 and formation of the azide 12

Scheme 2.

Synthesis of alkyne fragment 20.

While the amine component of amide 5 was a known compound,¹⁸ the previous alkylation leading to this prenylated catechol was not regiospecific, which resulted in a decreased yield of either isomer and a challenging separation. In an effort to make the desired isomer more accessible, and to avoid risk of contamination by another isomer, a regiospecific sequence was pursued to produce amine 7 as a precursor to the A-ring fragment of triazole 6. For this new route, acidic bromination²⁴ of commercial 5-nitroguaiacol led to the desired brominated regioisomer 7 (Scheme 1). While the product was obtained in low yield (28%), the regiochemistry was clean and clear from interpretation of the ¹H NMR spectrum. Specifically, the spectrum shows ortho-coupling of the two hydrogens in the aromatic region which only fits the desired regioisomer.

Deprotection of the methyl ether of compound 7 using aluminum trichloride and pyridine²⁵ furnished the nitrocatechol 8 in moderate yield. Protection of both phenols as MOM ethers through reaction with MOMCl and diisopropylethylamine (DIPEA) provided the expected di-acetal 9. A lithium halogen exchange on compound 9 followed by lithium copper exchange (CuBr·DMS) and exposure to prenyl bromide generated the desired prenylated intermediate 10 in moderate yield. Finally, a mild reduction of the nitro group of compound 10 using trichlorosilane and DIPEA furnished the desired aniline 11 in moderate yield.²⁶ Even though this pathway has not yet been optimized, this sequence compares favorably with earlier preparations of amine 11 because it is completely regiospecific. Finally, our earlier work had shown that amides could be converted to amines by a Curtius rearrangement with DPPA in the presence of isoprenoid groups. In a similar sense, the aromatic amine 11 was converted to the corresponding azide 12 with preservation of the prenyl group through reaction with t-BuONO and TMSN₃ (Scheme 1).²⁷

With the azide precursor to the pawhuskin triazole 6 complete, an alkyne representing a B-ring fragment was pursued through a sequence parallel to that used to prepare pawhuskin A (Scheme 2).²⁸ Reduction of the methyl ester of 3,5-dimethoxybenzoic acid (13) with LiAlH₄ furnished the expected benzylic alcohol in good yield and subsequent bromination with NBS gave bromide 14. Protection of the benzylic alcohol through reaction with TBSCl and imidazole smoothly provided the silyl ether 15. The geranyl chain was installed using a lithium-halogen exchange, lithium-copper exchange sequence, followed by exposure of the presumed cuprate intermediate to geranyl bromide to form the desired geranylated product 16. After deprotection of the silyl ether 16 through reaction with TBAF, the resulting benzylic alcohol 17 was oxidized with the Dess-Martin periodinane (DMP) to provide the aldehyde 18. Finally a Corey-Fuchs protocol was used to convert aldehyde 18 to alkyne 20 through the intermediate vinyldibromide 19.²⁹

After both azide 12 and alkyne 20 were in hand, investigation of the click reaction was initiated. In the presence of CuSO₄ and sodium ascorbate, the azide and alkyne reacted smoothly in a click reaction³⁰ to form the desired triazole-linked intermediate 21 in good yield (Scheme 3). The sequence was completed through the acid-catalyzed hydrolysis of the MOM acetal protecting groups of compound 21 providing the desired triazole-linked pawhuskin analogue 6.

Scheme 3.

Completion of triazole-linked pawhuskin analogue 6

Triazole 6 behaves much like the active amide analogue (5) in the OGFR assay of SKOV3 cells,^{18, 31} with exposure to a 0.5 μM concentration of the triazole producing an effect equivalent to that 10 μM naltrexone. These results encouraged us to pursue the more challenging preparation of a schweinfurthin-based triazole. Even though the schweinfurthin analogue required a significantly longer synthetic sequence, the opportunity to determine its biological activity in assays against 60 cell lines and compare its activity to data for our many schweinfurthin analogues was compelling. Therefore, after the central reactions were established through preparation of the pawhuskin analogue 6 attention was turned to synthesis of a schweinfurthin-based triazole by a parallel strategy

As a result of the studies noted above (cf Figure 2), it has become clear that a hexahydroxanthene ring system with the natural R,R,R-stereochemistry¹⁴ is required for significant and differential activity. Modifications of the D-ring are better tolerated, which suggested that the landscape shifted by incorporation of the larger triazole ring for the stilbene olefin might preserve biological activity. Given the importance of the tricyclic system to their biological activity, it was once surprising that compounds such as (+)-3-deoxyschweinfurthin B¹⁰ ((+)-3dSB, 22, Figure 4) have nearly the same activity as the natural product schweinfurthin B.¹² Deletion of the C-3 hydroxyl group significantly shortens the synthetic sequence, and so the triazole analogue of (+)-3dSB became our target compound (i.e. triazole 23). This allowed us to use the powerful NCI 60 cell line bioassay to test whether replacement of the stilbene olefin with a triazole would preserve antiproliferative activity.

Figure 4.

(+)-3-Deoxyschweinfurthin B (22) and its triazole analogue 23.

Use of a click reaction to prepare the triazole 23 required preparation of an appropriate azide and an acetylene. Synthesis of a hexahydroxanthene azide could take advantage of the amine 24 prepared to access the schweinfurthin amide.¹⁹ Even though we already had reported this reaction sequence,¹⁹ it still required an 11-step series of reactions from commercial vanillin and geraniol to prepare the amine. Fortunately, formation of the azide required just one added step, treatment of amine 24 with t-BuONO and TMSN₃.²⁷ The final reaction gave the desired azide 25 in low yield, and limited amounts of the starting amine precluded efforts to optimize the transformation (Scheme 4).

Scheme 4.

Preparation of the tricyclic azide 25.

The next step toward a triazole-linked schweinfurthin analogue was synthesis of the resorcinol “D-ring” fragment 29 (Scheme 5). Reduction of the known ester 26¹⁹ with LiAlH₄ provided the corresponding benzylic alcohol in good yield, and subsequent oxidation with the Dess-Martin periodinane (DMP) furnished aldehyde 27. A modified Corey-Fuchs³² protocol then was used to convert aldehyde 27 first to the vinyl dibromide 28, and then to the aryl alkyne 29.

Scheme 5.

Synthesis of the D-ring alkyne 29.

With both the requisite azide (25) and alkyne (29) fragments in hand, the click reaction was examined. Treatment of a solution of the azide and alkyne with CuSO₄ and sodium ascorbate furnished the desired triazole-linked schweinfurthin analogue 30 (Scheme 6). A final acid-catalyzed hydrolysis of the MOM ethers then provided the desired schweinfurthin analogue 23.

Scheme 6.

Assembly of the schweinfurthin triazole 23.

The National Cancer Institute established the 60 cell line bioassay as a broad screen to test compounds for selective anti-cancer activity. While this screen can be used to identify compounds of different scaffolds that act through a similar mechanism of action,³³ it also provides a wealth of information that can be used to identify structure-activity relationships within a structural class. When the new schweinfurthin analogue 23 was tested in this assay, it displayed significant biological activity. As shown in Table 1, the triazole 23 has a mean GI₅₀ value slightly higher, and differential activity somewhat lower, than schweinfurthin A (4). However when compared to the more appropriate model compound, (+)-3dSB (22), the mean GI₅₀ values differ only by a factor of 1.6, and the differential activity differs only by 0.41 log units (a factor of ~2.5). The potency of the triazole against the glioblastoma-derived SF295 cell line actually is somewhat greater than that of (+)-3dSB (35 nM versus 62 nM). This suggests that the triazole component is able to maintain the electronic communication between the C and D-rings, and/or the flat structure, that is required for schweinfurthin-like activity. Further evidence for the utility of the triazole ring can be found by comparison with the saturated analogue 31¹⁷ (Figure 5), where the mean GI₅₀ potency value is decreased by a factor of ~7 relative to the triazole 23, the differential activity has decreased by a factor of ~5, and the potency in the SF295 cell line has decreased to ~2500 nM or by a factor of ~70. Through a COMPARE analysis, compound 23 showed a strong correlation to (+)-3dSB (22) the model that inspired its preparation.

Table 1.

Comparison of the activity (mean GI₅₀, differential activity, and against the glioblastoma-derived cell line SF295) of compound 23 to representative schweinfurthins. The GI₅₀ value is the concentration necessary for 50% growth inhibition.

Compound # (NSC #)	Mean GI50 (μM)a	Log10 differential activity	SF-295 (μM)	Pearson Crltn to 4	Pearson Crltn to 22
4 (696119)	0.36	3.11	0.011	1.0	0.76
22 (735927)	0.62	2.88	0.062	0.78	1.0
23 (799728)	0.71	2.47	0.035	0.74	0.79
31 (750657)b	4.9	1.76	2.5	0.71	0.72

^bFigure 5.

Figure 5.

A reduced analogue of 3dSB.

In conclusion, these studies have developed synthetic strategies that can be used to install a 1,2,3-triazole ring system in place of the central olefin of a stilbenoid natural product. They also have shown that this substitution essentially preserves the potency and differential activity of the pawhuskin and schweinfurthin analogues in comparison to their stilbene-containing models. These findings encourage further exploration of this substitution in other bioactive natural products, including compounds such as resveratrol^34-35 and its myriad derivatives,³ combretastatin A4,⁴ the chiricanines,³⁶ the arachidins and arahypins,³⁷ and many others.¹