Stereocontrolled Synthesis of Syn-β-Hydroxy-α-Amino Acids by Direct Aldolization of Pseudoephenamine Glycinamide

Abstract

β-Hydroxy-α-amino acids figure prominently as chiral building blocks in chemical synthesis, serving as precursors to numerous important medicines. We have developed and here report a method for the synthesis of β-hydroxy-α-amino acid derivatives by aldolization of pseudoephenamine glycinamide, which can be prepared from pseudoephenamine in a one-flask protocol. Enolization of (R,R)- or (S,S)-pseudoephenamine glycinamide with lithium hexamethyldisilazide in the presence of lithium chloride followed by addition of an aldehyde or ketone substrate affords aldol addition products that are stereochemically homologous with L- or D-threonine, respectively. These products, which are typically solids, can be obtained in stereoisomerically pure form in yields of 55–98%, and are readily transformed into β-hydroxy-α-amino acids by mild hydrolysis or into 2-amino-1,3-diols by reduction with sodium borohydride. This new chemistry greatly facilitates the construction of novel antibiotics of several different classes.

As part of a program to develop practical synthetic chemistry for the discovery of new antibiotics we investigated and here report a two-step method for the constructive assembly of enantiomerically pure syn-β-hydroxy-α-amino acids from simple starting materials. These products figure prominently as chemical precursors to a number of important medicines, most notably antibiotics, as evidenced by the fact that five of the compounds prepared in this study have been transformed into antibiotics from four different structural classes: amphenicols, monobactams, vancomycins, and macrolides. The chemistry we describe offers a number of practical advantages relative to existing methodology, which we discuss after presentation of our results.

The basis of the new methodology stems from the discovery that pseudoephenamine glycinamide (1) undergoes efficient and diastereoselective syn-aldolization with both aldehyde and (remarkably) ketone substrates.^[1] The key precursor in this transformation, pseudoephenamine glycinamide (1), is readily available in both enantiomeric forms on multi-gram scale from the appropriate enantiomer of pseudoephenamine^[2] and N-Boc glycine using either one- or two-step protocols (the yields are effectively the same, Scheme 1). Compound 1 is conveniently recrystallized from absolute ethanol and forms a free flowing, white crystalline solid (mp 168–170 °C, 78% overall yield employing the one-flask protocol followed by recrystallization, 30-g scale). X-ray crystallographic analysis reveals that the crystalline lattice is free of any solvent or water molecules. Furthermore, unlike pseudoephedrine glycinamide,^[3] in crystalline form 1 shows little or no propensity to hydrate upon exposure to the air and thus is easily weighed and transferred in the laboratory.

Scheme 1.

Synthesis of pseudoephenamine glycinamide (1).

Enolization–syn-aldolization of 1 was readily achieved by the following general protocol. Freshly (flame) dried anhydrous lithium chloride (saturating, ~7.8 equiv)^[4] and 1 (1.3 equiv)^[5] were combined at 23 °C in anhydrous THF (~0.15 M in 1) and the resulting suspension was stirred at 23 °C until 1 dissolved; a portion of the excess LiCl did not dissolve. The resulting suspension was cooled to −78 °C whereupon a freshly prepared solution of lithium hexamethyldisilazide in THF (1 M, 2.5 equiv) was added by syringe. After stirring at −78 °C for 5 min, the reaction flask was transferred to an ice–water bath for 25 min, then was re-cooled to −78 °C where a solution of an aldehyde or ketone substrate in THF (1 M, 1 equiv) was added. The progress of the aldol addition was conveniently monitored by TLC analysis; aldehyde reactants were typically completely consumed within 30 min at −78 °C, whereas reactions with ketone substrates proceeded more slowly and in certain cases required warming to 0 °C to achieve complete conversion (see Table 1 and Supporting Information). In all cases only one of the four possible diastereomeric aldol addition products predominated (Table 1), and this product was typically readily isolated in diastereomerically pure form by flash column chromatography (55–98% yield of purified product). The minor diastereomeric aldol addition product(s) typically constituted <15% of the product mixture.^[6],[7]

Table 1.

Aldolization of pseudoephenamine glycinamide (1) with aldehyde and ketone substrates.

^[a]X₂₊ = (R,R)-pseudoephenamine. Reactions were run at a final concentration of 0.1 M (aldolate) and were performed on at least a 1-mmol scale. Isolated yields of major diastereomers are reported; diastereomeric ratios can be found in the Supporting Information.

^[b]Reaction was run on a 20-g scale. Enolization was conducted at 0 °C, and the final concentration of aldolate was 0.05 M.

As shown in Table 1, many different aldehydes and ketones were found to be effective substrates. We observed that the majority of the purified primary aldol products were solids; in the case of product 4 (from isobutyraldehyde), crystals suitable for X-ray analysis were obtained. The solid state structure of 4 derived from (R,R)-1 revealed it to be the syn-aldol product stereochemically homologous with L-threonine. In addition, the absolute and relative stereochemistries of syn aldol adducts 8 and 9 (from para-nitrobenzaldehyde and para-methanesulfonylbenzaldehyde, respectively) were rigorously established to form a homochiral series with 4 on the basis of their successful conversion to active antibiotics identical with chloramphenicol and thiamphenicol, respectively (vide infra). Stereochemical assignments of the remaining aldehyde addition products from Table 1 were made by analogy. The stereochemistry of these products conforms with the diastereofacial preferences for alkylation reactions of pseudoephenamine amide enolates, provided that a (Z)-enolate (with the α-amino group and enolate oxygen cis) is invoked, which seems to us quite reasonable.^[2b] Syn stereochemistry presumably arises from conventional Zimmerman–Traxler-type arguments.^[8]

In addition to its general, efficient, and stereoselective reactions with aldehyde substrates (linear, branched, and α-tetrasubstituted aliphatic, aromatic, α-oxygenated, and α,β-unsaturated), pseudoephenamine glycinamide (1) also serves as an exceptional substrate for aldolization with ketone substrates, providing aldol adducts with fully substituted β-centres, as illustrated by the seven examples 13-19 in Table 1. The stereochemistry of aldol adduct 16 (from methyl isopropyl ketone) was established unambiguously by X-ray analysis of its crystalline hydrate; not surprisingly, it was found to be fully consistent with the stereochemistry of the aldehyde aldol adducts (the methyl group acts as the “small” group). We also rigorously established the stereochemistry of the aldol adduct 18 by X-ray analysis of a crystalline derivative (vide infra), and this also conformed to that of the other aldol products. This product appears to represent a case of stereochemical matching, where the diastereofacial preferences of the enolate and the chiral ketone substrate (the latter consistent with a Felkin-Ahn trajectory)^[9] are reinforcing, accounting for the extraordinarily high stereoselectivity and yield of this particular transformation. Product 19 (55% isolated yield), from methyl styryl ketone, was formed least efficiently, we believe as a consequence of competitive conjugate addition (est. ~15%).

As a seemingly minor point, we note that careful analysis of the ¹H NMR spectra of the majority of the purified aldol adducts from Table 1 reveals that in addition to the two rotameric forms of the expected syn-aldol diastereomers, trace (≤5%) amounts of an “impurity” corresponding to the N→ O-acyl transfer product, a βamino ester, are present.^[10] This reveals that the latter constitutional isomer is only slightly higher in energy than the tertiary amide form, providing a rationale for the remarkable facility of the subsequent transformations of the direct aldol products discussed below, namely their hydrolysis and reduction.

In contrast to conditions typical for hydrolysis of tertiary amides, hydrolysis of the aldol adducts of Table 1 proceeds under remarkably mild conditions, more consistent with saponification of an ester than hydrolysis of a tertiary amide (Table 2). For example, hydrolysis of aldol adduct 4 was complete within 4 h at 23 °C in the presence of 1 equiv of sodium hydroxide in 1:1 THF:methanol. Once hydrolysis was complete, pseudoephenamine was recovered by extraction with dichloromethane in quantitative yield (≥95% purity), and the alkaline aqueous solution was lyophilized to provide the β-hydroxy-α-amino sodium carboxylate 22 in 92% yield and ≥98% ee (Table 2). The inclusion of methanol was critical to avoid retroaldol fragmentation during the hydrolysis, which was otherwise facile, especially with aromatic aldol addition products. In a noteworthy example, use of the THF-methanol-sodium hydroxide protocol with substrate 10 afforded the aromatic aldolate 25 in 94% yield and ≥98% ee (auxiliary recovery: 97% yield). A protected form of the latter α-amino acid served as a key starting material in the synthesis of vancomycin reported by the Nicolaou group.^[11]

Table 2.

Mild alkaline hydrolysis of aldol adducts.^[a]

^[a](R,R)-Pseudoephenamine was recovered in ≥90% yield in each case in high purity. Enantiomeric purity was determined by ¹H-NMR analysis of the (R)- and (S)-MTPA amides. For experimental details, see Supporting Information.

Interestingly, the present hydrolysis conditions are much milder than those required for hydrolysis of pseudoephedrine^[10] and pseudoephenamine^[2b] amide alkylation products, suggesting that the β-hydroxy group of the aldol adducts may facilitate N→ O-acyl transfer. In this regard, it is notable (though not surprising) that X-ray crystallographic analysis (structures 4 and 16) reveals an internal hydrogen bond between the amide carbonyl groups and their β-hydroxy functions. We believe that facile hydrolysis (and reduction, vide infra) of pseudoephenamine amide aldol products occurs by rapid N→ O-acyl transfer followed by saponification (reduction) of the resulting β-amino ester, as we have previously proposed for alkaline hydrolyses of pseudoephedrine amides.^[10]

The α-amino sodium carboxylates obtained upon alkaline hydrolysis can be converted to α-amino acid methyl esters upon exposure to acidic methanol (e.g., 20 → 26, Scheme 2). Alternatively, treatment of the same substrates with di-tert-butyldicarbonate affords N-Boc-protected amino acids in high yield (e.g., 23 → 27, Scheme 2). The N-Boc α-amino acid 27 is noteworthy for it serves as precursor to the fully synthetic monobactam antibiotic BAL30072, which is currently in phase I clinical trials as an anticipated treatment for infections caused by Gram-negative bacteria.^[12]

Scheme 2.

Esterification and N-Boc protection of amino carboxylates.

Alkaline hydrolysis conditions were not uniformly successful with every substrate; in certain cases retroaldol fragmentation was faster than hydrolysis, even when employing our optimal protocol. For example, treatment of the ketone aldol adduct 17 with 1 equiv of sodium hydroxide in 1:1 methanol:water at 23 °C provided primarily three products: acetophenone, pseudoephenamine, and sodium glycinate (the latter two products presumably result from hydrolytic cleavage of 1); none of the desired β-hydroxy-α-amino sodium carboxylate was observed.^[13] We envisioned that retroaldol fragmentation would be avoided if the β-hydroxy substituent were shielded, and for this purpose we chose a cyclic carbamate, which can easily be introduced and removed^[14] under very mild conditions and has the added benefit of protecting the α-amino function. Treatment of aldol adduct 17 with phosgene (1.1 equiv) and diisopropylethylamine (3 equiv) at −78 °C in dichloromethane formed within 30 min the cyclic carbamate 28, isolated in pure form by simple aqueous extraction. Although carbamate 28 was resistant to alkaline hydrolysis (presumably due to the acidity of the carbamate function) we found that heating a solution of 28 in a 1:1 mixture of dioxane and pure water at reflux for 24 h effected clean hydrolysis of the auxiliary. Straightforward acid-base extraction then provided acid 29 in 85% yield (and, separately, pseudoephenamine in 97% yield). By an analogous sequence, treatment of aldol adduct 18 with phosgene provided carbamate 30, (the stereochemistry of which was rigorously established by X-ray crystallography). This intermediate has been transformed into >100 novel macrolide antibiotics in ongoing research in our laboratory.^[15] Hydrolysis of 30 provided acid 31 in 94% yield (90% recovered pseudoephenamine).

To apply our new aldol methodology to synthesize chloramphenicol and thiamphenicol, antibiotics which are on the essential medicine list published by the World Health Organization^[16] and play critical roles in the treatment of infectious disease, especially in developing countries,^[17] we investigated reductive cleavage of the auxiliary to produce 2-amino-1,3-diols. Remarkably, treatment of aldol adduct 8 with the mild reducing agent sodium borohydride (5.0 equiv) in ethanol at 40 °C provided the 2-amino-1,3-diol 32 in 80% yield (Scheme 4); the auxiliary was recovered quantitatively in pure form. We are aware of only one previous report of the reduction of tertiary amides (α-hydroxy morpholinamides) to the corresponding alcohols with sodium borohydride.^[18] Reduction of pseudoephedrine and pseudoephenamine amides to the corresponding primary alcohols has historically been achieved using lithium amidotrihydroborate (LAB),^{[2b, 3b, 10]} a much more reactive hydride donor that we introduced in 1996.^[19] Again, we believe that the facile reduction with sodium borohydride we observe is due to intramolecular N→ O-acyl transfer followed by reduction of the resulting α-amino ester.^[20] The synthesis of chloramphenicol was completed by acylation of 32 with methyl dichloroacetate (Scheme 4), providing the antibiotic in excellent yield in just three steps from (R,R)-pseudoephenamine glycinamide (1) and para-nitrobenzaldehyde. Thiamphenicol was synthesized by an identical 2-step sequence from the aldol adduct 9. In contrast to the 3-step routes to chloramphenicol and thiamphenicol reported here, the commercial routes to these substances require ~6 linear steps, including a resolution.^[21]

Scheme 4.

Mild reductive cleavage of aldol adducts applied to the syntheses of chloramphenicol and thiamphenicol.

Commensurate with their importance in medicine, chemists have developed an extraordinarily diverse array of methods to synthesize enantiomerically enriched β-hydroxy-α-amino acids. These may be divided into two broad categories: constructive syntheses (as in the present work) and nonconstructive syntheses. The latter include the Sharpless asymmetric aminohydroxylation of certain alkenyl esters,^[22] multi-step transformations of Garner aldehyde-type intermediates,^[23] asymmetric hydrogenation of 2-amino-β-ketoesters,^[24] as well as other strategies.^{[14f, 25]}

Constructive syntheses are generally more powerfully simplifying, for they enable retrosynthetic targeting of the C–C bond linking the stereogenic, heteroatom-bearing centres. The pioneering advances of the Schöllkopf group employing bis-lactim ethers^[26] and the Seebach group employing masked glycine-derived heterocycles^[27] as substrates in diastereoselective aldol additions remain important enabling methodologies. To reveal the parent β-hydroxy-α-amino acids or esters, however, strongly acidic conditions are required and auxiliary-derived by-products can complicate isolation of the products.^{[26e, 26f]} Evans and Weber developed α-isothiocyanato acyl oxazolidinones as substrates in their diastereoselective tin-mediated aldol chemistry,^[28] and notable advances have been recorded by the Willis,^[29] Feng,^[30] and Seidel^[31] groups to transform this method into processes mediated by chiral catalysts. These α-isothiocyanate methodologies afford thiocarbamate heterocycles as products, which conveniently serve to protect the amine and alcohol functionalities of the aldol adducts, but require a 3-step procedure to reveal the embedded α-amino acids. Methods employing chiral glycine enolate equivalents have also been reported by the Bold,^[32] Iwanowicz,^[33] Caddick,^[34] and Franck^[35] groups. Hydroxymethylations of alanine equivalents to form α-alkyl serine derivatives have also been reported.^[36]

Another notable approach employs Schiff bases of glycine tert-butyl esters in aldol reactions with aldehyde substrates to provide aldol addition products that are then treated with acid to reveal the embedded β-hydroxy-α-amino esters. Advances in this area were reported by the Mukaiyama,^[37] Belokon,^[38] Miller,^[39] and Corey^[40] groups, and subsequently several modifications have emerged that provide both syn^[41] and anti^[42] products. While these methods are convenient due to the facile enolization of glycine Schiff bases and the direct conversion of the aldol products into β-hydroxy-α-amino esters, they often suffer from poor diastereoselectivities, narrow substrate scope, and frequently require further functionalization to permit separation of syn and anti aldol addition products.

Ito, Hayashi, and coworkers employed α-isocyano esters and amides in aldol reactions catalyzed by chiral gold(I) complexes, providing oxazoline-4-carboxylate products that can be converted to β-hydroxy-α-amino acids upon treatment with strong acid.^[43] Oxazoline-4-carboxylates have also been constructed by the addition of 5-alkoxyoxazoles to aldehydes catalyzed by chiral aluminum catalysts, as demonstrated by Suga and Ibata^[44] and the Evans group.^[45] These systems were found to be highly effective only with aromatic aldehyde substrates, and conversion of the oxazoline products to β-hydroxy-α-amino acids requires three steps and harshly acidic conditions. Barbas, Tanaka, and coworkers reported a method for the aldolization of phthalimidoacetaldehyde catalyzed by proline that achieved high enantio- and diastereoselectivities, but only with α-branched aldehyde substrates.^[46] The Wong group has developed methodology for chemoenzymatic aldolization of glycine catalyzed by threonine aldolases that, while highly stereoselective for certain aldehyde substrates, is limited in scope.^[47]

We believe aldolization of pseudoephenamine glycinamide offers a number of advantages. Enolization of 1 proceeds under very mild conditions (LiHMDS, LiCl) without metal additives, and the syn aldol products are readily obtained in stereoisomerically pure form by column chromatography. A broad selection of electrophiles, including alkyl and aryl aldehydes and ketones, undergo efficient aldolization with 1, whereas many other glycine equivalents react efficiently only with aryl or alkyl aldehydes, and very few are reported to react efficiently with ketones.^[48] With the exception of chemoenzymatic approaches,^[47] the aforementioned glycine equivalents all require shielding of the α-amino group, but this is not necessary with our method. Hydrolysis of the aldol adducts of 1 proceeds under unusually mild conditions compared to other glycine equivalents, and both the product and the auxiliary can be isolated by straightforward biphasic extraction. Additionally, reduction of pseudoephenamine glycinamide aldol adducts to the corresponding primary alcohols can be accomplished with the mild reducing agent sodium borohydride. We believe pseudoephenamine glycinamide (1) is an exceedingly practical reagent for the synthesis of β-hydroxy-α-amino acids and chiral 2-amino-1,3-diols, and anticipate the methods reported herein will have broad applicability in chemical synthesis.

Scheme 3.

Cyclic carbamate formation followed by hydrolysis under neutral conditions affords protected α-amino acid derivatives. X₃₊ = (R,R)-pseudoephenamine. [a] Product contained ≤8% TBDPS-OH after aqueous extraction.