First Total Syntheses and Biological Evaluation of Both Antipodes of Several Hydroxylated Dimeric Nuphar Alkaloids

In 2006, Yoshikawa and co-workers reported that 6-hydroxythiobinupharidine (+)-1b induces apoptosis in U937 human leukemia cells within 1 h (2.5–10 μM).^[1] To their knowledge, this was the most rapid induction of apoptosis by a small molecule ever reported. Additional experiments based on incubation with 20 μM of specific caspase inhibitors implicated activation of caspases 8 and 3 (but not caspase 9). Gopas and co-workers subsequently reported that partially purified combination extracts of (+)-1b and (−)-2b inhibit NFκB signaling.^[2] Despite the unprecedented speed with which the dimeric nuphar alkaloids are able to induce apoptosis, it is surprising that no other studies on their biological mechanism of action have been published. Previous reports also show that (+)-1b is an effective anti-bacterial,^[3] anti-fungal,^[4] and immunosuppressant agent.^[5]

The dimeric nuphar alkaloids are structurally unique sulfur-containing triterpenoids isolated from the yellow water lilies, Nuphar pumilum, Nuphar japonicum, and Nuphar lutea, and were first described by Achmatowicz and co-workers (Figure 1).^[6] Three series (i.e. thiobinupharidines (+)-1a–d, thionuphlutines (−)-2a–d, and neothiobinupharidines 3b–d), whose structures differ in the relative stereochemical configurations at C7 and C7′, are known.^[7] However, neither (−)-3a from the neothiobinupharidine series nor any of the members of the neothionuphlutine series 4a–d have been found in Nature.

LaLonde proposed a biosynthetic mechanism for the formation of the gross structures of the dimeric nuphar alkaloids beginning from elimination/oxidation of (−)-nupharidine (Scheme 1).^[8] Attack of ene-iminium 6 with an ambiphilic sulfur reagent 7 would initiate dimerization through formation of the central thiaspirane ring. Hydration or reduction of bis-iminium 10 would result in compounds 1–4. Shenvi and co-workers proved the feasibility of LaLonde’s proposal by completing the first and only total synthesis of a member of the dimeric nuphar alkaloids, (−)-neothiobinupharidine (−)-3d,^[9] which unfortunately is not active in apoptosis assays.^[1] Yoshikawa reported that at least one hemiaminal is required for there to be substantial apoptotic activity.

Considering 1) the unprecedented speed at which the dimeric nuphar alkaloids are able to induce apoptosis, 2) the lack of information regarding their biological mechanism of action, and 3) their complex and exquisite chemical structure, we decided to pursue the total syntheses of the biologically active hydroxylated dimeric nuphar alkaloids. In the absence of any catalysts or promoters, Shenvi reports that dimerization favors the formation of the stereochemical relationship corresponding to the neothiobinupharidine architecture (4:1 series 3 versus all others).^[9] Thus, the principle challenge we faced was how to override the inherent diastereochemical outcome at C7 and C7′ during the formation of the thiaspirane ring.

Although the requisite ene-iminium 6 (Scheme 1) can be derived from hemiaminal 11 (Scheme 2), this is a hydrolytically labile compound that cannot be purified by chromatography and must be freshly made before each use. We therefore elected to pursue a strategy employing nitrile (−)-12 as a “masked” hemiaminal that is both chromatographically and hydrolytically stable.^[10–12] We further reasoned that promoting the dimerization by application of chiral Brønsted or Lewis acids to (−)-12 would generate a chiral ion pair intermediate 13 in which the stereochemical information of counterion A*⁻ could assist in the control of facial selectivity (Scheme 2).

We prepared both enantiomers of 14 using known methods.^[9,13,14] We then converted (−)-14 to the nitrile (−)-12 by oxidation to the N-oxide, followed by Polonovski rearrangement and quenching with KCN (Scheme 3).

With (−)-12 in hand, we surveyed an extensive set of chiral and achiral Brønsted acids and Lewis acid complexes for their ability to alter the intrinsic selectivity at C7/C7′ in dimerization reactions with Na₂S₄. These included chiral phosphoric acids and numerous permutations of Lewis acids and ligands. Table 1 is an abbreviated summary of these studies (for a complete list, see SI). The products obtained are bis-nitriles 15–18. The assignment of stereochemical configuration at C7/C7′ was accomplished by converting them to the corresponding fully reduced compounds 1d, 2d, 3d, and 4d, then comparing with literature data. After considerable experimentation, we identified In(OTf)₃/(3R,8S)-22 as the optimal Lewis acid–ligand^[15–17] combination for generating (+)-15 (corresponds to thiobinupharidine configuration) (entry 16), while Cu(OTf)₂ alone gave predominantly (−)-17 (corresponds to neothiobinupharidine configuration) (entry 1). Under these conditions, we could obtain synthetically useful amounts of all stereoisomers 15–18. The use of chiral phosphoric acids or catalytic amounts of Lewis acids was not optimal. That there was no pronounced matched-mismatched effect with different antipodes of ligands suggests that the mechanism of dimerization may be more complex than simple substrate-ligand interaction.

Table 1.

Optimization of dimerization

Entry[a]	Metal	Ligand[b]	15:16:17:18[c]
1	Cu(OTf)2	none	0.2 : 0.2 : 1.0 : 0.3
			7% : 6% : 51% : 10%
2	Sc(OTf)3	none	0.3 : 0.3 : 1.0 : 0.2
3	In(OTf)3	none	0.3 : 0.2 : 1.0 : 0.2
4	Cu(OTf)2	(S)-19a	0.4 : 0.2 : 1.0 : 0.3
5	Sc(OTf)3	(S)-19a	0.4 : 0.2 : 1.0 : 0.2
6	In(OTf)3	(S)-19a	0.3 : 0.2 : 1.0 : 0.2
7	Cu(OTf)2	(R)-19a	0.4 : 0.2 : 1.0 : 0.2
8	In(OTf)3	(R)-19a	0.9 : 0.2 : 1.0 : 0.2
9	In(OTf)3	(R)-19b	0.8 : 0.3 : 1.0 : 0.2
10	In(OTf)3	(R)-19c	0.9 : 0.3 : 1.0 : 0.3
11	In(OTf)3	(R)-19d	1.1 : 0.3 : 1.0 : 0.2
12	In(OTf)3	(4R,5S)-20a	1.1 : 0.3 : 1.0 : 0.2
13	In(OTf)3	(4R,5S)-20c	1.0 : 0.3 : 1.0 : 0.3
14	In(OTf)3	(4R,5R)-21	1.0 : 0.3 : 1.0 : 0.2
15	Cu(OTf)2	none	0.2 : 0.2 : 1.0 : 0.3
16	In(OTf)3	(3R,8S)-22	1.9 : 0.4 : 1.0 : 0.3
			42% : 8% : 23% : 6%
17	In(OTf)3	(3S,8R)-22	1.6 : 0.3 : 1.0 : 0.3

Open in a new tab

^[a]

Isolated yields for selected entries provided in cells below.

We then converted bis-nitrile (+)-15 to the corresponding bis-hemiaminal (+)-1a by treatment with AgNO₃ in MeCN/H₂O at 45 °C (85% yield, Scheme 4).^[18] The preparation of 2a–4a was carried out in a similar fashion and in comparable yields. Selective mono-reduction of the C6′ hemiaminal over the one at C6 in 1a could be accomplished to give 1b.^[19] This type of chemoselective reduction appears limited to 1a as similar attempts on 2a–4a resulted in numerous reduction products. The absolute stereochemical configurations as depicted in Figure 1 are the naturally occurring ones, but we also completed the syntheses of the non-natural (or predicted non-natural) antipodes (−)-1a–b, (+)-2a, (+)-3a, and (−)-4a.

The ¹H and ¹³C NMR spectra of synthetic (+)-1a–1b in both CD₃OD and CDCl₃ were identical to both reported literature values and to the spectra of materials isolated from Nuphar lutea carried out by our group (see SI for isolation/purification details).^[20–23] However, if the data acquisition for the NMR spectrum in CD₃OD preceded that of the experiment in CDCl₃, then noticeable differences between synthetic and literature values were apparent in the CDCl₃ ¹H NMR spectra. No differences were observed in the CD₃OD spectrum. Under these circumstances, if we removed the CD₃OD in vacuo and allowed the residue to age in air, the material slowly converted back to (+)-1a–1b as judged by ¹H NMR spectroscopy in CDCl₃. The conversion was greatly facilitated if the residue was washed with saturated aq. NaHCO₃. We speculated that in CD₃OD, (+)-1a–1b is converted to the CD₃O– adducts 23a and 24a (Figure 2).

Indeed, when we dissolved (+)-1a–b in CH₃OH, then removed the excess methanol and obtained the ¹H NMR spectra in CDCl₃, new singlet resonances integrating to 3H each were present. We assigned these resonances to the methoxy protons in 23b–24b. From these data, we believe it necessary to revise those literature structures of (+)-1a–b—in which NMR experiments was carried out in CD₃OD—to the corresponding CD₃O– adducts 23a and 24a. It is likely that an analogous revision of 2a is warranted. The ethoxylated versions of some dimeric nuphar alkaloids has been noted, but there is some disagreement as to whether these were artifacts of the isolation/purification process.^[24,25]

We then assayed both enantiomers of 1a–1b, 2a, 3a, 4a, as well as bis-nitriles 15–18 for their apoptotic properties against the human U937 cell line as judged by caspase cleavage of poly(ADP-ribose) polymerase (PARP), a marker of apoptosis (Table 2 and Figure 3). The activity of synthetic (+)-1a–1b and (−)-2a were similar to that of the corresponding materials we isolated from Nuphar lutea (see SI for full data set). Moreover, the mono- and dihydroxylated dimeric nuphar alkaloids all exhibited nearly the same activity, regardless of their absolute or relative stereochemical configurations. And, it appears that the (−)-3a not found in nature seems to be the most potent at the 1 h time point (entry 7). Bis-nitriles 16–18 exhibited no apoptotic activity.

In conclusion, we have successfully completed the first syntheses of several hydroxylated nuphar alkaloids. This was accomplished by means of dimerization of nitrile (−)-12, promoted by chiral Lewis acid complexes. Chemoselective mono-reduction of (+)-6,6′-dihydroxythiobinupharidine (+)-1a furnished (+)-6-hydroxythiobinupharidine (+)-1b. Moreover, for the first time, apoptosis data is available for these compounds. PARP cleavage assays confirm their ability to induce very rapid apoptosis in human U937 cells (within 1 h). We anticipate that these compounds will serve as useful tools for dissecting an important, but as yet undefined step in the regulation of apoptosis. Studies to clarify the biological mechanism by which they operate are ongoing.