Enhancement of Cyclopamine via Conjugation with Non-metabolic Sugars

Abstract

The Veratrum alkaloid cyclopamine, an inhibitor of cancer stem cell growth, was used as a representative scaffold to evaluate the inhibitory impact of glycosylation with a group of non-metabolic saccharides, such as D-threose. In a five-step divergent process, a 32-member glycoside library was created and assayed to determine that glycosides of such sugars notably improved the GI₅₀ value of cyclopamine while metabolic sugars, such as D-glucose, did not.

Glycosylation is an important contributor to the biological activity of macromolecules (e.g., proteins and lipids)¹ and pharmaceutically-active agents (e.g., natural products and synthetic drugs). In the latter case, sugars attached to small molecule-based drugs can dramatically influence the corresponding mechanism, pharmacodynamics, pharmacokinetics and even patent life of the parental structure.² Given this precedent, the development of convenient strategies to synthesize glycoconjugate libraries are anticipated to facilitate the application of glycodiversification in drug lead identification and optimization.³ Neoglycorandomization, a divergent chemoselective glycosylation method, is advantageous in this regard as it employs free reducing sugars and thereby avoids the need for subsequent post-glycosylation modification/deprotection to produce the desired glycoconjugate.⁴ A survey of neoglycosylation studies to date – in the context of natural products,⁵ synthetic drugs,⁶ aglycon-linking strategies,⁷ and chemoselective handle variation⁶ – surprisingly reveals a core set of non-metabolic sugars (i.e., sugars not known to be involved in human metabolic processes) that consistently improve the latent anticancer activity of a range of target scaffolds. This core set includes both D- and L-entantiomers of the aldopentoses, D-glucuorono-6,3-lactone, and D-threose. While the underlying mechanism of anticancer enhancement in some of the cases investigated to date remains unclear, this intriguing trend suggests a focused library biased toward inclusion of these core sugars may offer the greatest chance for anticancer improvement of a given target scaffold.

To test this hypothesis, we selected the Veratrum alkaloid cyclopamine (Figure 1, 1) as a representative model. Identified in the 1960s as a teratogen capable of inducing cyclopia and holoprosencephaly in sheep embryos,⁸ cyclopamine disrupts natural embryonic development by binding to Smoothened, a key regulatory component of the Hedgehog (Hh) pathway,⁹ which is conserved in various organisms including humans.¹⁰ Because the Hh pathway is also responsible for proper division of adult stem cells, abnormal constitutive activation of Smoothened ultimately leads to an increase in the cancer stem cell population. Small molecule inhibitors such as 1 act to reverse this detrimental effect and thus function as anticancer agents without being directly cytotoxic.¹⁰ Tremblay et al. reported that small changes at the C3 position of D-ring-expanded cyclopamine, including functional group attachment (e.g., sulfonamide, acetamide), lowered Hh inhibitory EC₅₀ values to nanomolar concentrations.¹¹ The most potent analog, IPI-926 (Figure 1, 3),¹² is presently in six clinical trials for assessment in single- and combination-drug therapies against various types of cancer.¹³ As support for glycoconjugation, numerous veratramine, cevanine, and jervine (e.g., 1) alkaloids from the common family of Liliaceae have also been isolated as β-D-glucosides,¹⁴ some of which invoked increased cAMP phosphodiesterase activity over their aglycon counterparts.¹⁵ One such compound, cycloposine (Figure 1, 2), the 3-O-β-D-glucoside of cyclopamine, is known to impact embryonic development via Hh inhibition.^8b,16 Glycosylation at the F-ring amine, via Huisgen 1,3-dipolar cycloaddition-mediated installation of anomeric azido sugars,¹⁷ or installation of hydrophobic moieties at the same site, also led to 1 improvements.^9b,11a Herein we report the synthesis of a suitable cyclopamine neoaglycon in four steps and the subsequent creation of a 32-member library of cyclopamine neoglycosides. Evaluation of tumor cell growth inhibition revealed that the core group of non-metabolic sugar derived glycosides induced a greater cancer cell growth inhibition than those deriving from typical metabolic sugars. Neoglycosylation also increased the solubility of cyclopamine. Consistent with prior observations, this study supports the idea that particular glycosides offer distinct advantages in glycoconjugation directed toward improving anticancer activity, the general mechanism of which warrants further investigation.

Figure 1.

Structures of cyclopamine (1), cycloposine (2), and IPI-926 (3).

To prepare the cyclopamine neoaglycon, oxidation at the C3 position was found to best occur under Oppenauer conditions as previously reported^11a due to product decomposition using TPAP/NMO and the potential for opening the spirofuran ring under Jones oxidation conditions (see Scheme 1). Upon oxidation, a 1,2-vinyl shift of the native double bond to C4 and C5 led to the newly-formed α,β-unsaturated ketone (4). Due to the lack of susceptibility of this conjugated system to reductive amination, the C4-C5 alkene was subsequently reduced by hydrogenolysis using a palladium-carbon catalyst. Rather than a mixture of diastereomers, this reaction proceeded stereospecifically, providing the cis-decalone 5 as the only observable product. In support of this assignment, spectroscopic data of 5 were consistent with the similarly-reduced D-ring-expanded analog described by Tremblay et al.^11b Methoxyimine 6 was subsequently generated by reacting 5 with methoxylamine HCl salt in the presence of a weak organic base and the corresponding imine reduced using sodium cyanoborohydride with acetic acid catalytic amounts. These reduction conditions, which were much milder than borane-amine complexes requiring HCl activation, were also found to open the spirofuran ring when not quenched properly prior to chromatographic purification. This was avoided by vigorously washing the reaction with an equal volume of saturated aqueous NaHCO₃ for at least five minutes. Unlike stereoselective C4-C5 alkene reduction, imine reduction led to a separable 1:1 mixture of C3 diastereomers (7 and 8). The stereochemical assignment at C3 for each diastereomer was based upon a comparison of chemical shifts of the C3 protons (3S: 3.25–3.18 ppm, multiplet; 3R: 2.81–2.73 ppm, multiplet) and carbons (3S: 55.23 ppm; 3R: 60.43 ppm), which were consistent with previously-reported values.^11b,18 Attempts to induce stereoselective reduction of the imine through alternate means (e.g., Na(AcO)₃BH; NaBH₄/CeCl₃•7 H₂O; K-selectride) were unsuccessful. However, having access to both diastereomers also provided the means to assess the impact of the C3 stereocenter upon activity in the context of the corresponding neoglycosides. It is important to note that replacing EtOAc with CH₂Cl₂ as an extraction solvent during purification steps led to a dramatic increase in the overall isolated yield of aglycons 7 and 8 over the four step pathway from 23% to 72%.

Scheme 1.

Synthesis of Cyclopamine Neoaglycons (7 and 8) and Neoglycosides (CYC1-CYC32)

With suitable neoaglycons in hand, neoglycosylation was performed under mild conditions (i.e., MeOH/HOAc 8/1, 40 °C)^6,7 to facilitate the chemoselective ligation in the presence of the spirofuran ring. To compare with the historically-active set of glycosides (i.e., pentoses, D-threose, and D-glucurono-6,3-lactone) to alternative variants, carbohydrates that either had not previously induced increased cytotoxicity as neoglycosides or had not consistently done so, notably unmodified hexoses and deoxyhexoses, were also included within the library synthesis. A 32-member library (CYC1–CYC32) with 16 unique sugars was thus created (see Figure S1, Supporting Information) with an average isolated yield of 49%. Interestingly, when acetic acid was used in great excess (e.g., 25 equivalents) of aglycon, the yields typically doubled over the same reaction time (4 h to 28 h) without detriment to spirofuran opening. NMR revealed anomeric stereoselectivity to be dictated by the sugar employed as previously observed.^5–7 Notably, the cyclopamine C3 configuration had no impact upon anomeric specificity, a phenomenon not previously observed in the context of other natural product-neoglycoside conjugates. In general, neoglycoside anomeric configuration was biased toward a 1,2-trans relationship, with the exception of D-threosides CYC1 & CYC17, D-lyxosides CYC5 & CYC21, and L-lyxosides CYC6 & CYC22 (see Table S1, Supporting Information for characterization data).

To evaluate the impact of cyclopamine C3 neoglycosylation on biological activity, all 32 library members, along with aglycons 7 and 8 and parent compound 1, were evaluated for growth inhibition using the representative lung cancer cell line NCI-H460.¹⁹ A preliminary two-dose growth inhibitory assay (10 and 100 μM) revealed 5 of the 32 cyclopamine neoglycosides to display ≥ 20% inhibition at 10 μM (see Table S2, Supporting Information). Intriguingly, there was a disparity of activity between the two diastereomeric groups at this concentration that appeared to be independent of the cyclopamine C3 configuration. Specifically, with the exception of the similarly potent D-fucosides CYC11 and CYC27, active compounds were identified from both the (3S)-7 and (3R)-8 aglycons for which existed notably less active diastereomeric counterparts. Based upon this preliminary comparison, seven neoglycosides were subsequently used for a full dose study guided by the following rationale: D-threosides CYC1/CYC17 and D-arabinosides CYC3/CYC19 to compare the impact of the C3 stereocenter in the context of the most active sugars; L-arabinosides CYC4/CYC20) to compare to the corresponding enantiomers CYC3/CYC19; and D-fucoside (CYC27) as the most active hexose representative (see Table 1). As parental comparators, (3S)-aglycon 7, (3R)-aglycon 8, and 1 were also included.

Table 1.

GI₅₀ Data of Cyclopamine Neoglycosides Against NCI-H460 Lung Cancer Cell Line

entry	neoglycoside	C3a	GI50b
CYC1	D-threoside	S	8.5 ± 0.4
CYC3	D-arabinoside	S	21 ± 1
CYC4	L-arabinoside	S	17 ± 1
7	(3S)-aglycon	S	5.5 ± 0.3
CYC17	D-threoside	R	10.1 ± 0.9
CYC19	D-arabinoside	R	6.4 ± 0.5
CYC20	L-arabinoside	R	16 ± 1
CYC27	D-fucoside	R	9.9 ± 0.4
8	(3R)-aglycon	R	4.1 ± 0.1
1	cyclopamine	S	76 ± 5

^aConfiguration at the aglycon C3 position.

^bGI₅₀ values in μM.

The full dose study revealed all neoglycosides to be markedly improved over the parent 1, having growth inhibition of 50% of the cellular population (i.e., GI₅₀) 3.5 to 12 times lower with D-arabinoside CYC19 as the most active neoglycoside (6.4 μM). Consistent with our hypothesis, sugars previously found to consistently increase the anticancer activity of unrelated scaffolds^5–7 led to the best cyclopamine improvements and little or no improvement was found via conjugation to representative metabolic sugars (e.g., D-Gal and D-Glc had no observable inhibition at 10 μM). Only in the case of the most active sugar did the configuration at C3 effect notable impact on GI₅₀ values (CYC3/CYC19 - 21 μM vs. 6.4 μM, respectively). Surprisingly, both neoaglycons 7 and 8 displayed improvements similar to that of the best neoglycoside CYC19. Deletion of the C5-C6 double bond likely did not have a deleterious effect consistent with previously reported 5,6-dihydro analogs of 1.¹¹ While neoglycosides and neoaglycons were equipotent, glycosylation greatly improved solubility of the target scaffold. In DMSO, the solubility of neoglycosides CYC1–32 was ≥ 200 mM, aglycons 7 and 8 was < 50 mM, and 1 was < 4 mM. As cyclopamines modified at the A-ring with small molecular appendages (i.e., 3) have been reported to have comparatively lower clearance and higher oral bioavailability,^11b the alterations in physical properties via neoglycosylation (as reflected by solubility differences) may offer added benefit in the context of cyclopamine-based therapeutic development.

In summary, this study highlights a rapid strategy for differential glycosylation of cyclopamine, a naturally-occuring inhibitor of the Hh signaling pathway. Conjugation of cyclopamine with non-metabolic sugars previously found to improve the potency of a range of structurally and mechanistically unrelated anticancer agents led to glycosides with greater overall improvements in both anticancer activity and solubility and suggest the improvements invoked by such sugars may in part be due to a general (e.g. transport) rather than target-specific mechanism. This work extends the trend of cyclopamine improvement via small-molecule conjugation at C3 to include neoglycosylation and may point to a benefit of similar C3 glycosylation in the context of the acid-stable expanded D-ring (e.g., 3) cyclopamine series.