Development of α-Selective Glycosylation with L-Oleandral and Its Application to the Total Synthesis of Oleandrin

Abstract

This article describes the development of an α-selective glycosylation using L-oleandrose, a 2-deoxysugar that is frequently found in natural products, and its application to the total synthesis of natural cardiotonic steroids oleandrin and beaumontoside. To improve the reaction diastereoselectivity and to minimize side-product formation, an extensive evaluation and optimization of the conditions leading to α-selective glycosylation of digitoxigenin with L-oleandrose-based donors was conducted. These studies led to the exploration of 8 different phosphine•acid complexes or salts and yielded HBr•PPh₃ as the optimal catalyst, which provided in the cleanest α-glycosylation and produced protected beaumontoside in 67% yield. Subsequent application of these conditions to synthetic oleandrigenin afforded the desired α-product in 69% isolated yield-enabling the completion of the first synthesis of oleandrin in 17 steps (1.2% yield) from testosterone.

Graphical Abstract

It is well-established that glycosylation present in natural products may greatly impact the biological properties and stability of such compounds.¹ While installing or altering the carbohydrate structure on a natural product may greatly improve its bioactivity and bioavailability, such studies usually face significant challenges from a synthetic standpoint.² This is particularly problematic in the case of the natural products containing 2-deoxyglycosides (Figure 1A) due to the notoriously low α:β diastereoselectivities and side-reactions such as the Ferrier rearrangement.³ Not surprisingly, numerous studies have focused on developing stereoselective methods for the installation of 2-deoxyglycosides; however, despite many promising recent developments in this area, achieving such reactions with complex aglycone donors is still challenging.⁴

Figure 1.

Our group has long-standing interests in developing and exploring new methods for the selective synthesis and glycosylation of natural products.⁵ Recently, we achieved a concise 15 step synthesis of C16-oxidized aglycone oleandrigenin that is present in a variety of cardiotonic steroids including natural cardenolide oleandrin (Figure 1B).⁶ Oleandrin is the primary ingredient in the extracts of the Mediterranean plant Nerium oleander that is responsible for the medicinal properties of this shrub.⁷ The hot (Anvirzel)⁸ or cold (PBI-05204)⁹ extracts of Nerium oleander containing oleandrin have been recently explored in phase I and II clinical trials for the treatment of cancer, and antiviral diseases including SARS-CoV-2.⁹ In addition, oleandrin holds potential for the treatment of neurodegenerative diseases due to its unique ability to pass the blood brain barrier and reduce the amount of cellular prion protein (PrP^C) in the brain.¹⁰ This improved brain bioavailability was attributed to the presence of the C16-acetate and α-L-oleandrose at the C3 position of oleandrin. It must be noted that α-oleandrose-containing motifs are frequently found not only in cardiotonic steroids (cf. Figure 1B, compounds 4–8), but also in many other natural products including oleandomycin (1), avermectin A1a (2) or amplexicoside B (3).

While the synthetic installation of α-oleandrose is central to the medicinal chemistry exploration of these compounds, the direct methods for α-glycosylation with various oleandrose donors suffer from low yields and α:β-selectivity.¹¹ This manuscript describes exploration and optimization of various α-glycosylation conditions of digitoxigenin and synthetic oleandrigenin (10) that enabled the completion of the first synthesis of oleandrin.

Our prior work focused on glycosylation of 10 with rhamnose-based donors uncovered significant sensitivity of 10 to Lewis acids and bases.⁶ Consequently, acid labile protection on trichloroacetimidate rhamnose donor was used to generate natural cardenolide rhodexin B. With these prior observations in mind, our initial studies focused on exploring oleandrose donors 11 with acid-labile protecting groups (Table 1).

Table 1.

Glycosylation with oleandrose donors 10a–10c using digitoxigenin acceptor.^a

^aAll of the reactions were carried on 27 μmol scale in 267 μL of the solvent. The conversion and selectivity were determined by ¹H NMR (cf. SI Section 3).

The evaluation of various standard oleandrose-based donors 11a–11c using digitoxigenin aglycone as the model acceptor indicated that the desired α-oleandroside 12 was formed with low selectivities along with the hardly separable side-products such as β-oleandroside 13 and Ferrier rearrangement product 14. In addition, the glycosylation conditions resulted in the elimination of the C14-hydroxyl of digitoxigenin producing 15. Using an anomeric acetate donor 11a, led to moderate conversion and selectivity (entry 1).¹² Detection of 14 in this case suggests that the acetate moiety undergoes Lewis-acid promoted elimination to produce oleandral. This intermediate may still undergo TMSOTf promoted glycosylation or undergo an S_N2’ reaction to produce 4% of 14. Using the more labile and less basic trichloroacetimidate 11b results in higher conversion to 12 and 13 with no detectable rearrangement product 14, but with diminished selectivity ~1:1 (entry 2).¹³ In attempts to improve the diastereoselectivity for the formation of 12, picolyl containing donor 11c was synthesized and evaluated (entry 3). Picolyl swere developed by Demchenko and coworkers as a directing group that engages in hydrogen bonding with the acceptor.¹⁴ However, to our dismay, subjecting 11c and digitoxigenin to DMTST, did not lead to significant improvement in selectivity.¹⁵

With consistently low observed yield for the formation of 12 with donors 11a-11c, we turned our attention to exploring a protected oleandral as a donor (Table 2). Recent advances in catalysis resulted in several valuable catalytic methods for the α-selective formation of 2-deoxyglycosides, and our subsequent studies were focused on exploring these options. Based on the observations made during the synthesis of rhodexin B, TBS protected oleandral 16a^11e was selected as the donor for these studies. Our initial studies commenced with exploring various Lewis acids ascatalysts (entries 1–7). Au(I)^17a and Cu(I)^17b complexes were previously effective for the α-selective formation of 2-deoxyglycosides, but the use of these compounds with 16a and digitoxigenin resulted in 14 as the major product (entries 1, and 2). Other Lewis acid-based methods that are reported to minimize the Ferrier rearrangement were also ineffective for our system (entries 3–6). The use of catalytic B(C₆F₅)₃ at 50 °C^16c resulted in low conversion to 12 (6%), and predominant formation of 13 (10%), and 14 (11%) along with the elimination product 10 (6%). The reaction with catalytic Cu(OTf)₂^17d led to the elimination of the C14-hydroxyl and formation of 15 as the only observed product (entry 4) while the use of Bi(OTf)₃^16e resulted in low conversion (39%) and significant amounts of 14 (entry 5). Although the glycosylation with [Ir(cod)Cl]₂ as the catalyst should not involve the formation of strongly acidic medium,^16f the exclusive formation of 15 was observed when this catalyst was used (entry 6). At the same time, the use of the Pd(MeCN)₂Cl₂ complex with N-phenyl-2-(di-tert-butylphosphino)pyrrole under conditions developed by the Galan group^16g resulted in the most promising selectivity (entry 7). Thus, the desired α-product 12 was formed in 55% conversion along with 11% of 13, 12% of 14 and 11% of 15. This route was subsequently tested on a larger scale; however, the presence of hard-to-separate Ferrier product 14 in the reaction mixture prompted us to continue the optimization studies. Therefore, our further attempts were focused on identifying even milder conditions relying on hydrogen-bond donors such as thiouracil or thiourea (entries 8–10). The reactions catalyzed by thiouracil^16h and Schreiner’s thiourea¹⁶ⁱ (10 mol%) produced products 12–14 in low conversions (27% and 22%, correspondingly). However, the use of Schreiner’s thiourea in combination with (R)-TRIFP^16j (10 mol%) at 45 °C resulted in 90% conversion leading to 43:15:28:4 mixture of 12:13:14:15 (entry 10). With these failed attempts in hand, our optimization studies continued with examining the photo- and electrochemical conditions that have been developed to accomplish α-selective formation of 2-deoxyglycosides with impressive selectivity without Ferrier side-product formation (entries 11–13). Surprisingly, the electrochemical^16k and photochemical^16l methods by Ye and coworkers, were not as effective even though they were applied to the similar acceptor: acetyldigoxigenin (entries 11 & 12) highlighting that the protected oleandral is the culprit for the challenges in this glycosylation. However, using the Eosin Y photocatalysis developed by Wang (entry 13),^16m resulted in only trace quantities of hard-to-separate 14 marking this method as one of the key contenders.

Table 2.

α-Selective digitoxigenin glycosylation with oleandrals 16ab.

entry	donor	catalyst	T, (°C)	conversion (%)b	α-12	: β–13	: 14	: 15
1	16a	[CF3C6H4)3P]AuCI	rt	30	8	-	22	-
2	16a	Cu(OTf) • C6H6	45	75	31	-	40	4
3	16a	B(C6F5)3	50	32	6	10	11	6
4	16a	Cu(OTf)2	50	>99	-	-	-	>99
5	16a	Bi(OTf)3	rt	39	15	-	24	-
6	16a	[lr(cod)CI]2	rt	>99	-	-	-	>99
7	16a	Pd(MeCN)2CI2/PR3b	50	89	55	11	12	11
8	16a	thiouracil	40	27	18	5	3	-
g	16a	Schreiner’s Cat.	40	22	10	5	7	-
10	16a	Schreiner’s Cat/ (R)-TRIFP	45	90	43	15	28	4
11	16a	C(+) || Pt(—) (2 mA)	rt	>99	50	38	12	-
12	16a	faclr(ppy)3, blue LED	rt	22	19	-	-	3
13	16a	Eosin Y, PhSSPh blue LED	rt	79	51	24	4	-
14	16a	TMSI/PPh3	40	>99	59	24	17	-
15	16a	NIS/P(OPh)3	rt	>99	61	8	31	-
16	16a	HBr • PPh3	rt	>99	73	22	5	-
17	16a	HBr • PPh3	-78	98	68	17	8	4
18	16b	HBr-PPh3	rt	>99	50	-	49	-

^aAll of the reactions in entries 1–18 were evaluated on 27 μmol scale (cf. SI Section 3 for more details). The conversion and selectivity were determined by ¹H NMR.

^bPR₃ = N-phenyl-2-(di-tert-butylphosphino)pyrrole.

In addition, we evaluated several transformations that are based on the reaction of phosphonium salts (entries 14–16). The reaction with TMSI/PPh₃¹⁶ⁿ proceeded with excellent conversion and led to the mixture containing 59% of 12, 24% of 13 and 17% of 14 (entry 14). The use of less Lewis basic triphenylphosphite in combination with NIS previously described by Toshima^16o significantly increased the α:β-selectivity (entry 15); however, resulted in 31% of the Ferrier rearrangement product 14. Finally, a common method relying on catalytic quantities of HBr•PPh₃ by Mioskowski and Falck^16p was evaluated (entry 16). This method was found to be the best in terms of producing highest quantities of 12 (73%) and only minor quantities of 14 (5%) and side-product 15 (0%) were observed. This transformation was also carried at –78 °C without improvement in selectivity for 12 (entry 17), and, therefore, the conditions in entry 16 were selected for the further exploration. Surmising that the trans-orientation of the methoxy and tert-butyldimethylsilyl protected alcohol promotes a ¹C₄ to ⁴C₁ conformational switch, which favors the formation of the Ferrier rearrangement product 14, we investigated the glycosylation with the TBDPS-containing donor 16b (entry 18). Indeed, the use of bulkier TBDPS-protected oleandral 16b lead to the enhanced formation of TBDPS-protected 14 (49%) and exclusive formation of the TBDPS-protected α-glycoside 12.

While HBr•PPh₃ provided the best selectivity for the formation of 12, further optimization of the phosphine salt was attempted. To our surprise, despite the widespread applications of catalytic HBr•PPh₃ for the formation of 2-deoxyglycosides,⁴ very little information can be found on the use of the alternative phosphonium salts. It is, however, known that the reaction of such salts with dihydropyrans may lead to the formation of the anomeric phosphonium salts.¹⁷

If such salts are involved in the glycosylation catalytic cycle, the structure of the phosphine may significantly impact the reactivity and selectivity of this reaction. Therefore, our further attempts to improve the selectivity were focused on exploring phosphine salts of various strong acids (Table 3). The selectivities and conversions observed for more acidic complex HBr•P(C₆H₄F-p)₃ (entry 2) and less acidic salt HBr•P(C₆H₄OMe-p)₃ (entry 3) were similar to the parameters observed for HBr•PPh₃ (entry 1). Surprisingly, HBF₄•PPh₃ (entry 4) and HOTf•PPh₃ (entry 5) catalysts promoted the reactions at slower rate (57% and 53% conversion, correspondingly); however, a significantly higher α-selectivity and more side-product 14 were observed in each case (12:13:14:15 = 27:2:28:0 and 23:0:30:0). The salts of other sulfonic acids such as p-TsOH•PPh₃ and (+)-CSA•PPh₃ (entries 6–7) were found to be more catalytically active (>99% conversion), but led to lower levels of α:β selectivities (~2:1). These two catalysts contrasted with MsOH•PPh₃ that promoted reaction in 79% conversion and resulted in 36:32:11 ratio of 12:13:14 (entry 8). These results suggest that the basicity of the arylphosphine does not play a significant role; however, the nature of the acid may play a significant impact on both the reactivity and selectivity of this reaction. While HBr•PPh₃ and HBF₄•PPh₃ salts are known to provide anomeric phosphonium salts such as 17 (Scheme 1),¹⁷ the lack of dependence on the basicity of phosphine is not completely consistent with 17 reacting directly to provide 12 or 13. These results are more consistent with a complex mechanistic scenario involving multiple reactive species 17–18 or covalent variants of 18 such as glycosyl bromides or sulfonates, distribution and reactivity of which is dependent on the counterion.

Table 3.

Exploration of the phosphine salts as the catalysts for glycosylation with 16a

entry	catalyst	conversion (%)b	α-12	: β-13	: 14 :	15
1	HBr-PPh3	98	73	22	5	-
2	HBr • P(C6H4F-p)3	96	62	24	6	-
3	HBr • P(C6H4OMe-p)3	98	70	22	7	-
4	HBF4 • PPh3	57	27	2	28	-
5	HOTf • PPh3	53	23	-	30	-
6	p- TsOH • PPh3	99	61	30	8	-
7	(+)-CSA-PPh3	99	62	30	7	-
8	MsOH•PPh3	79	36	32	11	-

^aThe reactions were run on 27 μmol scale using 5 mol% of the phosphine salt in CH₂Cl₂ for 18 h (cf. Section 3 in SI).

^bThe conversion and selectivity were determined by ¹H NMR.

Scheme 1.

Potential reaction intermediates

With the optimized glycosylation protocol in hand, the completion of the syntheses of oleandrin 4 beaumontoside 5 were accomplished (Scheme 2). The glycosylation with digitoxigenin was run on 0.13 mmol scale yielding 12 in similar isolated yields of 64% using Eosin Y and 67% using HBr•PPh₃. Deprotection of 12 using HF•Py in THF buffered with pyridine proceeded smoothly and resulted in the formation of beaumontoside in 66% yield over two steps (Scheme 2). The subsequent application of the same steps to oleandrigenin (10), gratuitously, resulted in the desired α-glycosylated oleandrin product (61% yield for the photochemical protocol and 69% yield for the HBr•PPh₃ protocol). The following TBS-group deprotection using the same protocol as for beaumontoside resulted in the completion of the synthesis of oleandrin 4 in 17 steps (LLS) and 1.2 % yield from testosterone.

Scheme 2.

Synthesis of oleandrin (4) and beaumontoside (5)

In summary, the exploration of different oleandrose-based donors uncovered challenges with the α:β-selectivity and formation of hard-to-separate side-products. To address these challenges, an extensive evaluation and optimization of the conditions with L-oleandral-based donors helped to identify HBr•PPh₃ as the most optimal catalyst resulting in the cleanest α-glycosylation profile. Subsequent exploration of 7 different phosphonium•acid complexes indicated little dependence on the aryl phosphine, but strong dependence of both selectivity and reactivity on the counterion. Finally, the application of these conditions to the glycosylation of synthetic oleandrigenin aglycone afforded the desired α-product in 69% yield and enabled the completion of the first total synthesis of oleandrin in 16 steps (1.2% yield) from testosterone.

Data Availability Statement.

The data underlying this study are available in the published article and its supporting information.