De Novo Synthesis of an l-Lemonose Thioglycoside Donor from d-Threonine

Abstract

A short de novo synthesis of an l-lemonose thioglycoside is described starting from d-threonine. The synthesis leverages a Dieckmann condensation and Stork-Danheiser transposition to arrive at a key vinylogous ester intermediate on gram scale. Ensuing 1,2-addition diastereoselectively establishes the C3 tetra-substituted center and subsequent glycal hydration allows for anomeric functionalization to the thioglycoside. ¹H and NOESY NMR analyses reveal that the major α-anomer of thioglycoside deviates from the expected ¹C₄ conformation.

Graphical Abstract

Introduction

l-Lemonose (1) is a rare 4-amino-2,4,6-trideoxy-sugar that features a tetrasubstituted C3 branched center (Fig. 1). It occupies the non-reducing position of the arimetamycin A (AMA, 2) glycan¹ and serves as the sole glycosidic residue of the tetrahydroisoquinoline natural product lemonomycin (3).² Additionally, it is incorporated into various members of the thiazolyl peptide class of natural products (e.g., 4).³ Our lab has taken a recent interest in the arimetamycin A disaccharide as cytotoxicity studies from the Brady lab indicate that it imparts its steffimycin aglycone with nanomolar potency (IC₅₀) as compared to the micromolar activity (IC₅₀) of congeners bearing 2-deoxy-l-rhamnose (l-olivose) based monosaccharide glycans.¹ In particular, we want to investigate the role of C3 branching and amine dimethylation as AMA is the only known anthracycline with two branched N,N-dimethylated sugars (Fig. S1). As such, we required reliable access to a suitable l-lemonose glycosyl donor. Herein we describe our initial efforts toward this goal which culminate in the synthesis of an l-lemonose thioglycoside donor.

Figure 1.

Structure of the l-lemonose reducing sugar (1) and several l-lemonose containing natural products (2 – 4).

Results and Discussion

As we began our retrosynthetic analysis, we first took stock of prior l-lemonose syntheses. Stoltz reported the first synthesis of l-lemonose in 2003 wherein derivative 8 served as the aldehydic component for a late stage Pictet-Spengler cyclization in his group’s total synthesis of lemonomycin (Scheme 1a, S1).⁴ Their de novo synthesis of 8 from d-threonine (5) featured a stereoselective polar Felkin-Ahn type addition to establish the C3 branched center and a ring closing transesterification to forge the pyranose ring. In similar fashion, Masson and Zhu initiated their 2011 de novo synthesis of the l-lemonose reducing sugar (1) from d-threonine, using a chelation controlled Grignard addition to form the tetrasubstituted C3 center and a hemiacetalization to close the six membered ring (Scheme 1b, S2).⁵

Scheme 1.

Retrosynthetic Analysis for L-Lemonose Thioglycoside 22. ^aSeeberger did not characterize 19 as it was taken on crude to the next step.

In 2012, Fukuyama and Kan synthesized and employed l-lemonose glycosyl fluoride donor 13 in their total synthesis of lemonomycin 3 (Scheme 1c, S3).⁶ Though their de novo synthesis of intermediate lactone 12 closely paralleled that of Stoltz, their judicious choice of the N-Cbz protecting group allowed for more efficient N-deprotection/dimethylation. Finally, in 2015, Giuliano published the first approach to l-lemonose that began from a carbohydrate precursor rather than d-threonine (Scheme 1d, S4).⁷ Beginning from l-rhamnose (14), their synthesis featured an α-ketone alkylation to install the C3 methyl group, a reductive amination to establish the C4 amine, and a deoxygenation to excise the superfluous C2 alcohol.

While each of these routes successfully provided access to l-lemonose or a derivative thereof, we noted that the three de novo routes relied upon Felkin-Ahn or chelation control to install the C3 stereocenter prior to cyclization. In contrast, Nicolaou/Koskinen⁸ and Seeberger⁹ capitalized on conformationally constrained vinylogous esters 18 and 19 to achieve diastereoselective 1,2-addition at C3 in their syntheses of d-callipeltose (20) and d-ATT donor 21, respectively. Nicolaou and Koskinen attributed the diastereoselectivity of their methyl addition to 18 to the steric and electronic influences of the angular C4-N₃ (Scheme 1f).⁸ One can apply an identical argument to justify Seeberger’s selective C3 reduction of 19.

Turning to our goal of synthesizing and studying the AMA disaccharide, we envisioned that our ideal route to an l-lemonose donor would accomplish the following: (1) enable α-selective glycosylation, (2) achieve C3 functionalization in a diastereoselectively robust manner to facilitate the introduction of alternative C3 substituents without the need for chemical reoptimization, and (3) allow for late stage variation of the amine substitution with an eye to understanding the biological role of N,N-dimethylation. Taking these considerations in turn, we selected l-lemonose thioglycoside 22 as our synthetic target as activation of 2-deoxy thioglycoside donors under Hirama’s α-selective conditions (AgPF₆) has proven effective with C3 branched donors.¹⁰ We envisioned that 22 could result from functionalization of glycal intermediate 23 which we would access through a 1,2-addition to cyclic vinylogous ester 24 in manner analogous to Nicolaou and Koskinen.⁸ The potent nucleophile (CH₃Li) employed by these authors to achieve this alkylation would necessitate the use of an inert amine protecting group; we anticipated that an N,N-dibenzyl protected amine would prove would prove capable to the task. Finally, we would ultimately derive 24 from d-threonine (5) via adaption of Seeberger’s synthesis of 19⁹ and a related approach to l-kedarosamine by Myers.¹¹

Synthesis of thioglycoside 22 started from d-threonine (5). Fisher esterification, dibenzylation of the amine, and conversion of the alcohol to its acetate gave 25 on decagram scale (Scheme 2).^11–12 Exposure of 25 to LiHMDS facilitated the desired Dieckmann condensation.^{9a, 11} Quenching of the intermediate vinylogous carboxylate (26) with dimethyl sulfate gave vinylogous methyl ester 27.¹¹ We terminated the reaction by the addition of methanolic NH₃ to destroy excess dimethyl sulfate as orienting studies showed that this potent carcinogen persisted through aqueous workup and column chromatography. Next, DIBAL-H reduction of the crude lactone^9a followed by protonation (aq. H₂SO₄)¹³ and warming of the solution facilitated a Stork-Danheiser like transposition that produced vinylogous ester 24 in 71% yield over the two-pot, four-transformation sequence.

Scheme 2.

Synthesis of Glycal 23 from d-Threonine (5).

Reverse addition of vinylogous ester 24 to a solution of CH₃Li yielded allylic alcohol 28 with high diastereoselectivity (≥ 20:1); ¹H NMR analysis of the crude product indicated the absence of the epimeric addition product (Scheme 2). Subsequent silyl protection gave fully protected glycal 23. NOESY and single crystal X-ray diffraction (XRD) analyses of 23 verified that addition had occurred from the re face of the carbonyl as desired.¹⁴ The observed selectivity is analogous to that reported by Nicolaou and Koskinen for the addition of CH₃Li to azide 18 (Scheme 1e).⁸ They too observed complete stereocontrol and posited that the steric and electronic influences of the angular C4-N₃ (in our case C4-NBn₂) directed methyl addition to the opposite face of the carbonyl. Synthetically, we noted that telescoping these steps (i.e., quenching the intermediate lithiate with TBSOTf) gave 23 in lower and less reproducible yields (34–66%), demonstrating the superiority of the two-step method.

Having established the full carbon skeleton of lemonose donor 22, our next task was hydration of glycal 23 to provide a handle for further C1 functionalization. Unexpectedly, 23 resisted hydrolysis with HCl, H₂SO₄, and TsOH as TLC analysis indicated only the presence of starting material after 18 h at room temperature and a mixture of starting material and desired hydrolysis product 29 after an additional 5.5 h at 60 °C (Table 1, Entries 1–3). Switching to PPh₃•HBr, an acid commonly used for glycal hydrolysis,¹⁵ allowed us to isolate the desired hemiacetal α/β−29 in a modest 38% yield (Table 1, Entry 4).

Table 1.

Optimization of the Hydration of Glycal 23.

Entry	Acid (Eq)	Temperature	Time/h	Yield (%)a
1	HCl (2.6)	rt → 65 °Cb	25	---c
2	H2SO4 (2.9)	rt → 65 °Cb	25	---c
3	TsOH•H2O (2.6)	rt → 65 °Cb	25	---c
4	PPh3•HBr (2.1)	40 °C	44	38
5	Hg(OAc)2 (1.5)	rt	8d	58
6	Hg(OCOCF3)2 (1.5)	rt	5.5d	60
7e	Hg(OAc)2 (1.5)	rt	6d	51f

^aIsolated yield of α/β−29 after purification.

^bTemperature increased to 65 °C after 19.5 h at rt.

^cα/β−29 not isolated; 23 present by TLC after 25 h.

^dAfter the stated amount of time, the reaction mixture was treated with aq. NaOH followed by NaBH₄.

^eReaction ran with 1.20 g of 23 and yielded 632 mg of α/β−29.^f

^fα/β−29 was contaminated with an unidentified TBS containing impurity

Unsatisfied with the yield of α/β−29, we reexamined the reaction from a mechanistic point of view. In doing so, we formed two hypotheses: (1) protonation of the C4 amine likely occurs faster than the desired protonation of the glycal and (2) hydration of the resulting ammonium species 30 likely requires the intermediacy of high energy, doubly charged species such as intermediates 31 and 32 (Scheme 3).¹⁶ As such, we reasoned that switching the mode of activation from a Brønsted to a Lewis acid would ameliorate this problem. Subjection of glycal 23 to hydroxymercuration-demercuration validated this conjecture as we increased our yield of α/β−29 to 58–60% (Table 1, Entries 4–5). Furthermore, these conditions allowed for increased material throughput as the reaction efficiency changed little when starting with 1.20 g of 23 (Table 1, Entry 7).

Scheme 3.

Proposed Mechanism for the Brønsted-Acid Mediated Hydration of 23.

Having solved the issue of glycal hydration, we pushed forward toward targeted thioglycoside 22. Acetylation of reducing sugar α/β−29 under standard conditions gave acetate α/β−33 in 85% yield as an inconsequential mixture of anomers (Scheme 4a). Finally, BF₃•OEt₂ activation of the anomeric ester in the presence of PhSH gave a co-eluting mixture of desired thioglycoside 22 in 54% yield (10:1 α:β) and elimination product 23 (7%). While we envision that the presence of glycal 23 will not prove deleterious in future glycosylations, we nevertheless wanted to improve the yield and selectivity of the reaction to increase material throughput. In this vein, a small-scale experiment (62 µmol α/β−33) showed that increasing the amount of BF₃•OEt₂ and adding powdered 4 Å molecular sieves allowed to us increase the yield of 22 to 78% and improve the ratio of 22:23 from 5:1 to 18:1.

Scheme 4.

Synthesis and Conformational Analysis of l-Lemonose Thioglycoside a/β−22. ^aData collected in CDCl₃ (¹H, 600 MHz). ^bData collected by Giuliano and co-workers⁷ on an inseparable mixture of α/β−17 in CDCl₃ (¹H, 300 MHz).

Structurally, major anomer α−22 exhibited larger anomeric coupling constants (³J_H1-H2 = 6.2, 4.8 Hz) than anticipated (Scheme 4bi).¹⁷ NOESY analysis showed correlations corresponding both to the expected ¹C₄ conformation (C3Me ↔ H5) and the chair flipped ⁴C₁ conformation (H1 ↔ TBS_t-Bu, H1 ↔ H5, TBS_t-Bu ↔ H5, and H2_pro-S ↔ H4). As such, we concluded that α−22 exists either in conformational equilibrium or in an intermediate conformation. In contrast, anomeric coupling constant (³J_H1-H2 = 12.4, 2.3 Hz) and NOESY analyses indicate that β−22 occupies the canonical ¹C₄ conformation (Scheme 4bii).

Interestingly, Giuliano and coworkers observed a related phenomenon in their synthesis of l-lemonose thioglycoside α/β−17.⁷ They found that α−17 exhibited even larger coupling constants (³J_H1-H2 = 10.4, 6.4 Hz) than we observed for α−22 (Scheme 4biii). An H2_pro-S ↔ H5 NOESY correlation indicated that α−17 assumed a boat conformation (^2,5B). A lack of NOESY correlations for the anomeric proton (only the H1 ↔ H2_pro-R correlation was observed) as well as an X-ray crystal structure support this conformational assignment.

Conclusion

In conclusion, we have synthesized l-lemonose thioglycoside 22 in 10 steps from d-threonine. This work featured the gram scale synthesis of vinylogous ester 24 and extensive NMR analysis of the conformationally deviant α−22. Further optimization of the thioglycosylation as well as the synthesis of an AMA disaccharide via donor α/β−22 will be reported in due course.