The Stereochemical Course of the α-Hydroxyphosphonate–Phosphate Rearrangement

The phosphonate–phosphate rearrangement is an isomerisation of α-hydroxyphosphonates bearing electron-withdrawing substituents at the α-carbon atom. We studied the stereochemical course of this rearrangement with respect to phosphorus. A set of four diastereomeric α-hydroxyphosphonates was prepared by a Pudovik reaction from two diastereomeric cyclic phosphites. The hydroxyphosphonates were separated and rearranged with Et3N as base. In analogy to trichlorphon, which was the first reported compound undergoing this rearrangement. All four hydroxyphosphonates could be rearranged to 2,2-dichlorovinyl phosphates. Single-crystal X-ray structure analyses of the α-hydroxyphosphonates and the corresponding phosphates allowed us to show that the rearrangement proceeds with retention of configuration on the phosphorus atom.


Introduction
Racemic dimethyl 2,2,2-trichloro-1-hydroxyethylphosphonate (1), also known as trichlorphon, is an organophosphonate that was introduced as an insecticide in the early 1950s. [1] Later it was used in medicine under the trademark" metrifonate" to treat schistosomiasis, ap arasitic disease. This compound attracted the interesto ft he scientific community,w hen au nique kind of transformation was observed. Upon treatment with base, HCl was eliminated from the molecule (Scheme 1). [2] The observed elimination product was O,O-dimethyl O-2,2dichlorovinyl phosphate (2), also known as DDVP.T his was the first reported a-hydroxyphosphonate-phosphate rearrangement.
Subsequently,o ther a-hydroxyphosphonates,n ot having al eaving group like chloride at the b-carbon atom, but an electron-withdrawing group on the a-carbona tomw ere shown to undergo the same transformation. [3,4] Therefore, ag eneral reactionm echanism was proposed, which was later on proven by isotope labellinge xperiments (Scheme 2). [5] After deprotonation of the a-hydroxyl group by the base, anucleophilic attack of the oxyanion on the electrophilicphosphorusc entre takes place. This induces PÀCb ond cleavage and either elimination of the leaving group R 3 (if there is one at the C-2 atom, such as ah alogen) or formation of as hortlived carbanion at the C-1 atom (if R 2 is carbanion stabilising, such as aphenyl group), which is finally protonated.
The a-hydroxyphosphonate-phosphate rearrangement, as it is generally referred to, can be observed for a-hydroxyphosphonates bearing at least one electron-withdrawing substitu-ent on the a-carbona tomn ext to phosphorus, such as aryl-, cyano-or halide residues. [6] Similarly,c ertain a-mercapto-and a-aminophosphonates undergo this isomerisation to give thiophosphates and phosphoramidates,respectively.
The corresponding retro-reaction, the phosphate-phosphonate rearrangement is also known. When ap hosphate is treated with as toichiometric amount of BuLi at lowt emperature, aproton is abstracted in a positiontoanoxygen atom. The resulting carbanion is configurationally stable and the ensuing PÀCb ond formation proceeds with retention of configuration at the involved carbon atom. [7] This rearrangementi so fi ndustrial importance in the synthesis of DDVP (2). It was demonstrated that trichlorphon (1)h as no physiological activity andt hat its effects are exclusively due to the non-enzymatic transformation to DDVP,w hich is the active agent. Therefore, trichlorphon was replaced by DDVP in some cases. However, trichlorphon can be used as as low release formulation of DDVP, [1,8,9] which is an acetylcholinesterase inhibitor interfering with neuronal signal transmission. [3] If either the phosphorus centre or the neighbouring carbon atom involved in this reaction is bearing chiral information,t he stereochemistry of the transformation can be studied. The ahydroxyphosphonate-phosphate rearrangement was shown to proceedw ith retention of configurationw ith respect to the involved carbon centre by our group. [6] The reaction was carried out in am ixture of DMSO and water (95:5) by using 1,5-diazabicyclo [5.4.0]undec-5-ene (DBU) as abase (Scheme 3C).
Because decomposition of the a-hydroxyphosphonates to the dialkyl phosphite anion 16 and ketone 15 is af ast-and thus troublesome-side reactionu nder those conditions, water is essential to quickly protonate the intermediately formed anion. The recombination of those two fragments to racemic a-hydroxyphosphonates is also base catalysed. This leads to the predominant formationo ft he racemic phosphate 14.S till, the enantiomeric excess (ee)o ft he re-isolated starting materialw as high enough to deduce the stereochemical course of the reaction.
In af irst attempt, Brienne and co-workers studied the stereochemistry at phosphorus by using derivatives of trichlorphon. They performed ac hemical resolution of the racemic monomethyl ester of trichlorphona nd esterified the enantiomericallyp ure isomersb yu sing orthoformates to give the diastereomeric a-hydroxyphosphonates 9a and 9b.S eparation by fractional crystallisation and subsequent single-crystal X-ray structure analysis allowed them to assign the absolute configuration of each diastereomer.U pon rearrangement by using NaOH as base, they obtained ap air of enantiomeric phosphates,t hat is, compounds 10 a and 10 b (each phosphonate giving only one phosphate),b eing chiral only due to phosphorus. Their experiments showedt hat the obtained phosphates were of opposite configurationa sc ould be judged from the opticalr otation. However,t hey could not crystallise them and thus could only show that the reactioni ss tereospecific, but whether it proceeds with retention or inversion of the configuration remained elusive at the time (Scheme 3A). [10] Recently,J ankowski and co-workersf ound that the rearrangementp roceeds with retention of configuration on the involved phosphorus atom (Scheme 3B). [11] However,t he reaction was only performed with two diastereomeric a-hydroxyphosphonates 11 a and 11 b,b earing axial PÀCb onds, derived from as ix-membered cyclic phosphite. The two other a-hydroxyphosphonates,h aving equatorial PÀCb onds, were not isomerised. Therefore, they could not give ageneral conclusion on the stereochemistry of the rearrangement.

General thoughts
We investigated whether the assumption made by Jankowski et al. that the a-hydroxyphosphonate-phosphate rearrangement proceeds with retention of the configuration at the phosphorus atom, is ag eneral rule for all a-hydroxyphosphonates.
When the stereochemistry of the a-hydroxyphosphonatephosphate rearrangementi ss tudied, two main problems need to be overcome. Single-crystal X-ray structure analyses are necessary to assess the absolute configuration of the involved phosphorus atom before and after the rearrangement. The major difficulty in all prior attempts has been the poor tendency of phosphates to crystallise. Thus,first of all, it is essential to find as ystem of a-hydroxyphosphonates and phosphates giving crystalline compounds. Secondly, the decomposition of the a-hydroxyphosphonates to phosphites and aldehydes and the reverse process, both being base catalysed, have to be suppressed to prevent the formation of stereoisomers from homogeneous startingc ompounds.
In our current approachw ee nvisaged the synthesis of four diastereomeric, cyclic a-hydroxyphosphonate analogues of trichlorphon. We wanted to make use of the HCl elimination during the rearrangement as ad riving force for the reaction to preventt he decomposition of the a-hydroxyphosphonate to phosphite and aldehyde. Also, cyclic phosphates are known to crystallise more easily than acyclic ones. Using cyclic phophonates meanst hat interconversion between the possible orientations of the PÀCb onds in the molecules is impossible, as the rings fix their position. Thereby we wanted to overcome all prior problems.
Of our four diastereomerich ydroxyphosphonates two had axial and two had equatorial PÀCb onds, in order to compare their behaviour during the rearrangement. Another key point of this current approachi st he fact that upon rearrangement our substrates will give diastereomericr ather than enantiomeric phosphates. This will make their comparison and characterisation much easier compared to all former attempts.

Synthetic strategy
We planned to synthesise all four stereoisomeric a-hydroxyphosphonates through aP udovik reaction from an aldehyde and cyclic phosphites with an axial and an equatorial PÀH bond, respectively.W ea lready knew from other projects in our group that it is very easy to obtain six-membered cyclic phosphites by transesterification of the commercially available bis(2,2,2-trifluoroethyl) phosphite (21) [12] with 1,3-diols. [13] In case of a C 2 symmetric, enantiomerically pure diol this leads to the clean formation of one cyclic phosphite. The phosphorus, in this case, is then ac hirotopic, non-stereogenic centre. [14] Startingf rom an asymmetric1 ,3-diol, ap air of diastereomeric phosphites is obtained. If those phosphites are reactedw ith as uitable aldehyde in Pudovik reactions, diastereomeric a-hydroxyphosphonates are obtained, which can be used to study the phosphonate-phosphate rearrangement.
The reaction is very reliable and furnished the diol in 75 % overall yield for both steps withoutp urificationo ft he intermediate ketone. The resulting diol 20 was used to transesterify bis(2,2,2-trifluoroethyl) phosphite (21)a tr oom temperature in dry pyridine, which was the necessary base and the solvent at at ime (Scheme 5).
This reaction proceeded quantitatively as judged by NMR spectroscopy and produced both diastereomericc yclic phosphites 22 a and 22 b in a1:1 ratio.
Initially,w ew anted to continueo ur synthesis with this mixture of phosphites in order to keep the number of necessary purification steps as low as possible. Unfortunately,i tw as extremely difficult to separatea ll four diastereomeric a-hydroxyphosphonates 24 a-24 d in the next step. Therefore, we separated the two phosphites andr eacted them separately with anhydrous chloral( 23)t og ive two a-hydroxyphosphonates each. As the two phosphites partially decomposedo ns ilica gel, we lost some material and could only isolate ac ombined yield of 54 %o fb oth diastereomers. Both were solids, but only the more polar diastereomer 22 b [16] could be crystallised from hexanes/CH 2 Cl 2 to give colourless needles, suitable fors inglecrystal X-ray structure analysis. It revealed an equatorial PÀH bond for this diastereomer.H owever,w ep oint out that the phosphorus centrek eeps its tetrahedral configurationa nd distorts the six-memberedr ing. The chair conformation is rather flattened towards phosphorus. This makes it difficult to clearly define axial and equatorial positions. Therefore, we decided to have al ook at the position of the phenyls ubstituent in the ring and use it as an internal reference point of equatorial position. The configurationo ft he other diastereomer,t hat is, compound 22 a,w hich could not be determined by X-ray structure analysis, was underpinned by 1 HNOESY NMR spectroscopy. The homogenous phosphites were then reacted separately with chloral at À30 8Cc atalysed by Et 3 N( Scheme 6).
Depending on the diastereomer,t he starting phosphite was consumed to about 90 %a fter 5-7 h. In the case of the faster movingp hosphite 22 a the two a-hydroxyphosphonates were formed in a1 :1 ratio as judged by 31 PNMR spectroscopy.T hey could be separated by flash chromatography.I ti se ssential to remove Et 3 Nq uickly from the reactionm ixture in order to avoid rearrangement of the product. Flash chromatography Scheme4.Synthesis of the racemic diol through ac rossed aldolreaction followed by reduction. www.chemeurj.org with as hort column was thus performedf irst, followed by as econd one with al onger column to separatet he diastereomeric products.B oth homogenous a-hydroxyphosphonates 24 a and 24 b are colourless solids. The faster moving diastereomer 24 a could be crystallised from toluene/acetone at room temperature and furnished crystalss uitable for X-ray structure analysis. It showed the PÀCb ond to be axially orientated, meaning that the PÀHb ond of the starting phosphite was replaced with retention of configuration. The other diastereomerd erived from the same phosphite, that is, compound 24 b,m ust also have an axial PÀCb ond, but be of opposite configuration at the newly formed stereogenic centre.
In the case of the more polar phosphite 22 b the two a-hydroxyphosphonates were also formed in a1 :1 ratio (as could be judged from the final product yield). However,t his ratio could not be determined from the crude reaction mixture in CH 2 Cl 2 .O ne of the two diastereomers, that is, compound 24 d, has an extremely low solubility in all organic solvents except DMSO and DMF. Under the reactionc onditions it crystallises. Therefore, NMR spectroscopy of the resulting turbid mixture is not representative for the actual diastereomeric ratio. However, this property made the separation of the two compounds very easy.T hey could be separated by digestion with ethyl acetate or dichloromethane. Again, both a-hydroxyphosphonates are crystalline solids. The less polar compound 24 c yieldedc rystals suitable for X-ray structure determination,r evealing that the equatorial PÀHb ond of the startingp hosphite wasr eplaced by an equatorial PÀCb ond. As each pair of hydroxyphosphonates was obtained from one phosphite they had to share the configuration of the phosphorus atom, as well as of the stereocentre in the six-membered ring. Consequently,e ach pair differs only in its configurations at the carbon atom bearing the hydroxyl group.

The rearrangement
Having all four desired a-hydroxyphosphonates in hand, as well as the corresponding crystal structures of two of them, namely compounds 24 a and 24 c,w er earranged them (Scheme 7).
We rearranged all four diastereomeric a-hydroxyphosphonates separately in CHCl 3 at 0 8Cb yu sing Et 3 Na sb ase in stoichiometric amounts. The reactionp rogress wasm onitored by 31 PNMR spectroscopy.These experiments clearly demonstrated that the four phosphonates produce two different dichlorovinyl phosphates, compounds 25 a and 25 b.F urthermore, each pair of hydroxyphosphonates with the same configuration at the phosphorus atom gave exclusively one phosphate. This finding is an experimental prooff or the stereospecific character of the reaction.
As anticipated, decomposition of the hydroxyphosphonates to chlorala nd the respective phosphite interfered with the rearrangement. Even thoughd ecomposition was observed to ah igh extent,t he detected phosphite was always exclusively the one from which the studied hydroxyphosphonate was derived. Thus, no interconversion between the cyclic phosphites takes place by inversion of the configuration at the phosphorus atom. The amount of phosphite formed was dependento n the used base concentration. Surprisingly,n oc hloral could be detected by NMR spectroscopy.F urther,n or etro-reaction of phosphite andc hloralc ould be observed at anyt ime of the reaction, for which we did not find an explanation. We also tried to identify other possible side products. However,a ll our attempts to detect chloroform and carbon monoxide failed. [17] NMR spectroscopic kinetics for the rearrangemento fc ompound 24 a can be found in the Supporting Information and are representative for all hydroxyphosphonates. Luckily,b oth vinylphosphates 25 turned out to be crystalline solids.A single-crystal X-ray structure analysis was performed of the less polar compound 25 a having an axial dichlorovinyloxy substituent at the phosphorus atom. The isomeric cyclic phosphate 25 b has therefore an equatorial dichlorovinyloxy substituent at the phosphorus atom.
Comparison of the crystal structures of the a-hydroxyphosphonates and the corresponding dichlorovinyl phosphates showed that the phosphorus atom does not change its configuration during the rearrangement. However,t he sixmembered ring is not ap erfect chair due to the phosphorus atom as in the case of the phosphites. This leads to adistortion of the ring conformation of the a-hydroxyphosphonates, which is less dominant in case of the final phosphates. Still, the two molecules can be easily compared by studying the relative orientation of the phenyl and the 1-hydroxy-3,3,3trichloroethyl or dichlorovinyloxy substituent on the phosphorus atom as already discussed. Thereby we were ablet of ind that the a-hydroxyphosphonate-phosphate rearrangement proceeds with retention of configurationr egarding the phosphorusa tom in all four studied diastereomeric a-hydroxyphosphonates.
We assume that the reactionp roceeds through ap entacoordinated phosphorus spirocentre, with both the PÀCa nd the PÀOb ond still present in accordance to the mechanism described in Scheme 8. Upon treatment with base the a-hydroxyl group is deprotonated anda ttacks the phosphorus centre from the side opposing the P=Ob ond (intermediate 26). The resultings piro intermediate 27 cannotu ndergo aB erryp seudorotation as this is not possible for cyclic pentacoordinated phosphorus centres. [18] However,t his intermediate can undergo at urnstile rotationl eading to ac onfiguration where the oxyanion and the PÀCb ond are opposing each other (intermediate 28). [19] When the PÀCb ond breaks,t he phosphorus atom gets againt etracoordinated and adoptsi ts preferred tetrahedral structure. Thereby,t he emergingp hosphate 25 a re-Scheme7.The rearrangement was performedi nCHCl 3 with Et 3 Na t08 C. Still, it has to be mentioned that our data, as wella st hose already published by Jankowski et al. [11] were obtained by using cyclic a-hydroxyphosphonates. Data for at least one acyclic a-hydroxyphosphonate are still missing.

Furtherconsiderations
As there are some naturally occurring a-hydroxyphosphonic acids known, the retro-a-hydroxyphosphonate-phosphate rearrangement was once postulated as an alternative pathway to the rearrangement of phosphoenolpyruvate for the biosynthesis of phosphonates. It would have accounted fort he PÀC bond formation and the a-hydroxylg roup introduction at at ime. [20] This mechanism was ruled out by feeding experiments with Tetrahymena thermophila by using C-6 doubly deuterated glucose. However,r ecently an ew pathway for the biodegradation of 2-aminoethylphosphonic acid (2-AEP) wasd iscoveredi nm arine microorganisms. Twoe nzymesa re involved. The first one, PhnY, a-hydroxylates 2-AEP and the second one, PhnZ, effects the subsequento xidative PÀCb ond cleavage. [21,22] It was suggested that an a-hydroxyphosphonatephosphate-type rearrangement could form am echanistic basis for the PÀCbondcleavage.

Conclusion
Diastereomeric, cyclic phosphites 22 a and 22 b were prepared, which furthergave the four diastereomeric a-hydroxyphosphonate 24 a-24 d analogues to trichlorphon. These could be easily rearranged to give the corresponding dichlorovinyl phosphates 25 a and 25 b by using Et 3 Na sb ase. Four singlecrystal X-ray structure analyses allowed the assignment of the configuration of the phosphorus atom in the diastereomeric phosphites, a-hydroxyphosphonates andp hosphates. The combination of all theses tereochemical datasets provest hat the a-hydroxyphosphonate-phosphate rearrangement proceeds with retention of configurationr egarding the involved phosphorus atom. The same is true for the a-hydroxyphosphonate formation from aldehyde and phosphite in aP udovik reaction.