Synthesis of 10-Aza-9-oxakalkitoxin by N-O Bond Formation

Abstract

We describe a formal synthesis of the hydroxalog 10-aza-9-oxakalkitoxin of the cytotoxic marine natural product kalkitoxin that features Mukaiyama Markovnikov silyl peroxidation of a terminal alkene and N-O bond formation as the central, enabling steps.

Graphical Abstract

We are interested in the development of N,N,O-trisubstituted hydroxylamines (-NR-O-) as novel bioisosteres, or hydroxalogs, of monosubstituted ethylene bridges (-CHR-CH₂-) or ether units (-CHR-O-) in bioorganic and medicinal chemistry. We designed 10-aza-9-oxakaliktoxin 1 as a hydroxalog of the cytotoxic marine natural product kalkitoxin^{1, 2} 2 (Figure 1) and described its synthesis by a multi-step procedure from L-arabinose (Scheme 1).³ 10-Aza-9-oxakalkitoxin 1 showed strong solid tumor selective activity in a disk diffusion assay⁴ and in particular nanomolar cytotoxicity for the solid hepato-carcinoma cell line HepG2 equal to that exhibited by the parent 2. Additionally, 10-aza-9-oxakalkitoxin 1 was found not be genotoxic in the Salmonella typhimurium test strain TA98, with and without metabolic activation by a PB/β-naphthoflavone-induced rat liver metabolic activation system,^{5, 6} thereby negating the common perception that hydroxylamines are inherently mutagenic.

Figure 1.

Structures of 10-Aza-9-oxakalkitoxin and Kalkitoxin

Scheme 1.

Mitsunobu and reductive amination sequence for hydroxylamine installation in the first synthesis of 10-aza-9-oxakalkitoxin

The synthesis of N,N,O-trisubstituted hydroxylamines by O-alkylation of readily available N,N-disubstituted hydroxylamines⁷ is inefficient and only practical for strongly electrophilic primary electrophiles while the Meisenheimer rearrangement of N-allyl amine oxides has limited scope.^{8, 9} This lack of broad effective methodology prompted us to develop alternative methods depending on the reduction of O-acyl-N,N,-disubstituted hydroxylamines,^{10, 11} or caused us to rely on reductive amination-based routes.¹² Indeed, our synthesis of 1 (Scheme 1) was designed with these constraints in mind and employed Mitsunobu inversion of alcohol 3 with N-hydroxyphthalimide^{13, 14} to give 4 in 89% yield. This was followed by hydrazinolysis to give an O-substituted hydroxylamine, which was condensed with formaldehyde to give the oxime 5 in 90% yield for the two steps. Reduction with sodium cyanoborohydride¹⁵ was followed by condensation with aldehyde 7, itself derived by ozonolysis of N-allyl 2(R)-methylbutyramide 6, and a second cyanoborohydride reduction to give 8 in 74% overall yield. The protected diol in 8 was then released and oxidatively cleaved to give the key acid 9, which was converted to the target 1 in four steps and 56% overall yield using standard procedures,^16-18 analogous to the final stages of the originals synthesis of kalkitoxin itself.² While successful, this synthesis suffers from the need to construct hydroxylamine precursor 3 with the 4,5-anti-configuration in order to establish the correct syn-configuration in 4, and the need to isolate oxime 5 so as to achieve the necessary level of control in the double reductive amination sequence.

With this in mind, subsequently we developed general methods for the convergent construction of N,N,O-trisubstituted hydroxylamines involving reaction of secondary amine-derived magnesium amides with i) methyltetrahydropyranyl (MTHP) derivatives of primary and secondary alkyl hydroperoxides, and ii) tert-butyl perbenzoates.^{19, 20} We now demonstrate that application of the first of these methods, coupled with the Mukaiyama Markovnikov hydroperoxidation of a terminal alkene²¹ prepared by chiral auxiliary-directed diastereoselective synthesis, affords a convenient entry into the key acid 9 and thereby a considerably shortened formal synthesis of the hydroxalog 10-aza-9-oxakalkitoxin 1.

The new synthesis (Scheme 2) of 9 began with conversion of a commercial L-phenylalaninol derived oxazolidinone to its N-crotyl derivative 10 in 88% yield as described by Evans and coworkers.²² Conjugate addition of vinylmagnesium bromide in the presence of copper(I) bromide•dimethyl sulfide and diethylaluminum chloride, as described by the Williams and Song laboratories,^{23, 24} gave adduct 11 in 77% yield with a diastereomeric ratio of 98:2. Lithium borohydride reduction then gave 3(R)-methyl-4-penten-1-ol 12 in 87% yield, which was protected in the form of the naphthylmethyl (Nap) ether 13 in 84% yield. The requisite peroxide was then installed by the Mukaiyama method²¹ on stirring 13 in the presence of triethylsilane and 5 mol% of Co(acac)₂ in a dichloromethane:2-propanol mixture under an oxygen atmosphere. Working with 0.5 g of 13, this protocol gave an overall 67% yield of a reproducible 1.1:1 mixture of the diastereomeric peroxides syn-14 and anti-14, which were separated by preparative HPLC over a C18 reverse phase column. Treatment of the individual silyl peroxides, syn-14 and anti-14, with trimethylphosphine in toluene, followed by TBAF in THF according to a literature protocol,²⁵ gave the corresponding alcohols, syn-15 and anti-15, whose configurations were determined by Mosher analysis²⁶ as detailed in the supporting information.²⁷

Scheme 2.

Formal synthesis of 10-aza-9-oxakalkitoxin with installation of the hydroxylamine moiety by N-O bond formation

The syn-isomer of 14, with the 3R,4R configuration, was stirred in dichloromethane over 4 Å acid-washed molecular sieves in the presence of 2-methyl-2-tetrahydropyranol¹⁹ and p-toluenesulfonic acid, when it afforded the MTHP derivative syn-16 as a mixture of diastereomers in 88% yield. Standard tritylation of N,N’-dimethylethylene diamine gave amine 17, which was deprotonated with n-butyllithium in THF at −40 °C and then treated with syn-16 in THF affording, after 0.5 h at −40 °C, the desired trisubstituted hydroxylamine syn-18 as a single diastereomer in 68% isolated yield. Detritylation to generate syn-19 was achieved in 91% yield, and was followed by coupling to commercial 2(R)-butyric acid with the PyBOP reagent,²⁸ to providee the tertiary amide syn-20 in 88% yield. Finally, removal of the naphthylmethyl ether with DDQ²⁹ produced syn-21 in 87% yield, and was followed by Parikh-Doering³⁰ and Pinnick³¹ oxidations to afford acid 9 in 85% yield, whose NMR spectra corresponded with those of the previous sample (Supporting Information), and thus completed the formal synthesis of 10-aza-9-oxakalkitoxin 1 (Scheme 2).³² As a final confirmation, the anti-peroxide anti-14, obtained from the Mukaiyama peroxidation sequence was taken through the same sequence of reactions via anti-16, anti-18, anti-19, anti-20, and anti-21, in comparable yields, to give the anti-isomer 22 of acid 9, which displayed markedly different NMR spectra to 9 (Supporting Information).

Several features of this new synthesis of 1 (Scheme 2) deserve comment. First, at 12 steps and 7.0% overall yield from the Evans auxiliary to the common acid 9 it is considerably shorter but not necessarily higher yielding than the original synthesis (Scheme 1), which required 21 steps from L-arabinose to reach 9, of which it gave a 5.7% overall yield. The significant reduction in step count represents a considerable increase in efficiency, even if the overall yield is only marginally increased, and is achieved by the use of a robust chiral-auxiliary-based entrance, rather than a chiral-pool-based one, and the novel strategic disconnection enabled by the novel N-O bond forming reaction. Second, reaction of the lithium amide of 17 with the silyl peroxides 14 to give the hydroxylamines 18 directly was inefficient underscoring the importance of the MTHP moiety in this and the related ether chemistry.³³ Third, the use of the lithium amide of 17 was superior to that of the corresponding magnesium amide, whereas in the general method described earlier the magnesium amides were preferred.¹⁹ We tentatively ascribe this phenomenon to differences in the aggregation states and so-reactivity of lithium and magnesium amides and the influence of chelation thereon. Finally, it is noteworthy that, the modest yield in the new synthesis stems mainly from the lack of selectivity in the installation of the silyl peroxide, which also necessitates HPLC separation of the isomers. In this regard, we note that more highly functionalized secondary MTHP peroxides can be readily formed by displacement of secondary carbohydrate-based triflates,^{19, 34} which are easy to prepare and handle,³⁵ but application of such methods in the absence of strongly electron-withdrawing vicinal C-O bonds is complex and inefficient. Other existing methods for the synthesis of simple secondary alkyl hydroperoxides are either limited in scope or do not lend themselves readily to the preparation of enantiopure substances and/or scale-up.^36-41 Clearly, there is a need for the development of an efficient, stereocontrolled synthesis of chiral secondary alkyl hydroperoxides to support the further development of our N-O bonding chemistry and the related ether-forming reactions of the Dussault and Herzon groups.^{33, 34}

In conclusion, we describe a much shortened and improved formal synthesis of 10-aza-9-oxakalkitoxin 1, that is enabled by installation of the central hydroxylamine moiety by N-O bond formation. In addition to highlighting the applicability of this N-O bond forming reaction in the construction of complex trisubstituted hydroxylamines, the synthesis draws attention to the need for the development of an effective asymmetric synthesis of secondary alkyl hydroperoxides and their derivatives.