Biomolecular The α -hydroxyphosphonate-phosphate rearrangement of a noncyclic substrate – some new observations †

Racemic ethyl hydrogen (1-hydroxy-2-methylsulfanyl-1-phenylethyl)phosphonate was resolved with ( R )-1-phenylethylamine. The ( R )-con ﬁ guration of the ( − )-enantiomer was determined by chemical corre-lation. Esteri ﬁ cation of the ( − )-enantiomer with a substituted diazomethane derived from 3-hydroxy-1,3,5 (10)-estratrien-17-one delivered two epimeric phosphonates separated by HPLC. Methylation with methyl ﬂ uorosulfate at the sulfur atom and treatment with a strong base induced an α -hydroxyphosphonate-phosphate rearrangement with formation of dimethyl sulphide and two enantiomerically pure enol phosphates. Their oily nature interfered with a single crystal X-ray structure analysis to determine the stereochemistry at the phosphorus atom. 53.6, 118.0, 118.8 CF 321.0 Hz, CF 121.1, 139.6, 141.3, 147.4. Anal. calcd


Introduction
When phosphoric acid derivatives (±)-1 are treated with strong bases in stoichiometric amounts such as alkyl lithiums or lithium amides at low temperatures, they are deprotonated to give short-lived organolithiums (±)-2 containing dipole-stabilised 1 carbanions (Scheme 1). These undergo rearrangements via (±)-3 to lithiated α-substituted phosphonates (±)-4 and on work up to α-hydroxy-, α-sulfanyland α-aminophosphonates (±)-5. This isomerisation discovered for X = O by Sturtz and Corbel 2 is called phosphate-phosphonate or more specifically phosphate-α-hydroxyphosphonate rearrangement. This 3-5 and the versions for X = S 6,7 and N 8 have extensively been studied by Hammerschmidt's group. The reverse process with many examples 9-17 for X = O, the α-hydroxyphosphonate-phosphate rearrangement, also termed [1,2]-phospha-Brook rearrangement, has been found by Pudovik and Konovalova 17 before the phosphate-phosphonate rearrangement. This isomerisation is normally catalysed by a variety of catalytic bases such as e.g. NaOH, NaOEt and DBU. While the transformation of (±)-1 into (±)-5 for X = O is feasible even for R 2 = alkyl and R 3 = H, the reverse process not. At least one of the substituents, R 2 or R 3 , should stabilise the developing negative charge on the carbon atom in (±)-3 upon cleavage of the C-P bond. An aromatic substituent suffices to stabilise the intermediate carbanion. The driving force for the phosphate-phosphonate rearrangements is the stronger Li-O than Li-C bond. The reverse process (O-H + P-C → C-H + P-O) is dominated by the much higher P-O than P-C bond energy. These isomerisations are related to the Brook and retro-Brook rearrangements in silicon chemistry. 18 The phosphate-phosphonate rearrangement for X = O, 3,5 S 7 and N 8 and the reverse process for X = O 11,15,16 follow a retentive course at the respective carbon atoms. The stereochemistry at the phosphorus atom upon the a-hydroxyphosphonatephosphate rearrangement follows a retentive course too, proven only for α-hydroxyphosphonates with the phosphorus atom as part of a six-membered ring. 11,16 It was found that diastereomeric α-hydroxyphosphonates (R,S P )-and (R,R P )-7 obtained by esterification of enantiomer (R)-6 and fractional crystallisation rearrange stereospecifically (Scheme 2). 10 Here the phosphorus atom was not part of a ring system and the developing negative charge on the α-carbon atom upon Scheme 1 Phosphate-phosphonate rearrangements and reverse processes.
cleavage of the P-C bond eliminated a β-chloride, resulting in enantiomerically pure enol phosphates (+)-and (−)-8. As they were oils, their absolute configuration could not be determined by X-ray structure analysis and the stereochemistry at the phosphorus atom had to remain unanswered.

Results and discussion
The highly enantioselective synthesis of acyclic phosphate triesters is difficult and challenging. 19,20 While Hall and Inch 19 built their syntheses on 5-and 6-membered cyclic phosphorus compounds derived from (−)-ephedrine and D-glucose, Nakayama and Thompson 20 applied (S)-proline derivatives. We reasoned that the α-hydroxyphosphonate-phosphate rearrangement of acyclic substrates with a stereogenic P-atom of known configuration would give chiral, nonracemic phosphate triester and alkenyl dialkyl ester. In order to assign the configuration to the P-chiral product, the stereochemistry of the rearrangement at the phosphorus atom has to be known. Here we start another approach to unravel it.
Ethyl bis(trimethylsilyl) phosphite 3 was added to ketone 9 21 to give a protected α-hydroxyphosphonate as intermediate, that was hydrolysed to phosphonic acid monoethyl ester (±)-10 upon aqueous workup and isolated in 59% yield. The phenyl ketone was selected, because the phenyl substituent with its anion-stabilising effect will ascertain that the α-hydroxyphosphonate-phosphate rearrangement at the end of the sequence will be feasible. The methylsulfanyl substituent can be methylated and give the good leaving group dimethyl sulfide. (R)-(+)-1-Phenylethylamine [(R)-11] was found to be a better resolving agent for phosphonic acid (±)-10 than brucine. The crystals obtained from Et 2 O/CH 2 Cl 2 contained Et 2 O (by 1 H NMR, salt/Et 2 O, 2.6 : 1.0) and had a de of already 86% (by 1 H NMR, the two methylsulfanyl groups of the two diastereomers resonate as two singlets at δ 1.77 and 1.79). Two crystallisations from CHCl 3 delivered crystals of hemihydrate (R)-11 × (−)-10 × 0.5H 2 O of de > 98% in 56% yield. When this salt was dissolved in aqueous ammonia (25%) and extracted with CH 2 Cl 2 , the (R)-1-phenylethylamine was recovered. The free acid (−)-10 was isolated from the aqueous phase by passage through Dowex 50 W, H + and removal of water under reduced pressure.
To determine the absolute configuration of (−)-10, it was transformed into the known α-hydroxyphosphonate (S)-(−)-12 (Scheme 4). This was achieved by desulfurisation of the respective potassium salt with aged RANEY-nickel, 22 followed by passage through Dowex 50 W, H + to get the free acid. Esterification with diazoethane 23 furnished phosphonic acid diethyl ester (−)-12 (in 22% overall yield), which has (S)-configuration based on the comparison of the specific optical rotation with the literature value. 3 When freshly prepared RANEY-nickel was used, the CH 3 S and OH groups were both reductively removed. This experiment proved that phosphonic acid (−)-10 3 has (R)-configuration. The change of the descriptor is caused by the change in the priority for the substituents according to the CIP rules [for (S)-(−)-12: P > Ph > CH 3 ; for (R)-(−)-10: P > CH 2 SCH 3 > Ph.
Separation by preparative HPLC delivered the less polar 20 of 88% de and the more polar 23 of 96% de. Crystallisation of epimer 20 from hexanes or cyclohexane furnished crystals of >98% de, which contained solvent (20/hexanes, 3.13 : 1; 20/ cyclohexane, 2 : 1, by 1 H NMR). The more polar epimer 23 was crystallised from hexanes/i-PrOH to give crystals of also de >98%, which were suitable for an X-ray crystal structure. It allowed to assign (R,R P )-configuration ( Fig. 1) to the phosphonic acid part of 23 and consequently (R,S P )-configuration to that of 20 (the PvO bond is considered a single bond when the sequence rule is used!). Both epimers were methylated at the sulfur atom with methyl fluorosulfate at −35°C. The respective sulfonium salts 21 and 24 were deprotonated at the hydroxyl groups with phosphazene base P 1 -t-Bu, 31 a stronger base than DBU, to induce α-hydroxyphosphonate-phosphate rearrangements as detailed for 24 in Scheme 7.
The alkoxide 26 has two options. Firstly ( pathway A), it can disintegrate (retro-Abramov reaction 32 ) into phosphite anion 29 and sulfonium salt 28, which in turn react with each other to sulfonium ylide 30 and H-phosphonate 31. The carbonyl group of the ylide is not electrophilic enough to allow addition     33 We assume that 27 has a trigonal bipyramidal structure formed by an apical attack of the alkoxide anion on the electrophilc phosphorus atom from the less hindered side opposite to the EstCH 2 O substituent. The P-C bond will be equatorially orientated. The negative charge building up on the α-carbon atom in 25 upon cleavage of the P-C bond eliminates dimethyl sulphide. The two enol phosphates 22 and 25 were obtained in yields of 46% and 48%, respectively. Their specific optical rotations were [α] 20 D + 46.8 and + 41.9, respectively. These compounds contain beside the stereogenic phosphorus atom some stereogenic carbon ones in the steroidal substituent. Therefore the specific optical rotations cannot have the same absolute values with opposite signs. NMR spectroscopically, they are virtually identical ( 1 H, 13 C, 31 P) except for the resonances of EstCH 2 OP group (AB parts of ABP systems) in the 1 H NMR spectrum (Fig. 2). Inspection of the three segments of the relevant 1 H NMR spectra reveal that the two enol phosphates 21 and 25 are enantiomerically pure. Unfortunately, none of the two oils could be induced to crystallise and the absolute configuration of the stereogenic phosphorus atom could not be determined by single X-ray structure analysis. The stereochemical course of the α-hydroxyphosphonate-phosphate rearrangement of a noncyclic α-hydroxyphosphonates remains to be determined. However, it must be a stereospecific reaction yielding enantiomerically pure enol phosphates.

Conclusions
In summary, we prepared a racemic ethyl hydrogen α-hydroxyphosphonate, resolved it with (R)-1-phenylethylamine and esterified it with a diazomethane derived from 3-hydroxy-1,3,5(10)-estratrien-17-one. Each epimer obtained by HPLC separation was methylated at the methylsulfanyl substitutent and treated with base to induce α-hydroxyphosphonate-phos-phate rearrangements. We found that the rearrangement is stereospecific. However, the stereochemistry could not be determined as the obtained phosphate was not crystalline to perform a single crystal X-ray structure analysis. The sequence allows to prepare enantiomerically pure enol phosphates.
Diazoethane: 23 To a solution of KOH (15 g) in water (45 mL) and Et 2 O (30 mL) cooled at −35°C (bath temperature) N-nitroso-N-ethylurea 35 (4.0 g) was added in portions within 5 min. The mixture was stirred until the urea had dissolved (20 min). The yellow ethereal solution of diazoethane was used directly for esterification.

Conflicts of interest
There are no conflicts to declare.