High pressure chiral recognition experiments – a route to homochiral 4-fluorocyclohexenones

4-Fluorocyclohexadienones, prepared by an improved Jacquesy-process, were shown to undergo highly face-selective Diels-Alder reactions with the homochiral cyclopentadiene 1 . Some of the diastereoselective transformations, quite unique rearrangements of α -hydroxy compounds resulting from nucleophilic attack to the carbonyl group as well as selected Retro-Diels-Alder reactions, are reported.


Introduction
In recent papers, we communicated the potential of the homochiral cyclopentadiene 1 1 for sequential diastereoselective cycloadditions, 2, 3 kinetic resolution 4 and differentiation of enantiotopic double bonds. 4,5,6,7 Inthe latter case, cycloadditions with cyclohexadienones of type 2 were shown to follow the very obvious pathway which places the small residue "S" into the α-endo-position (Scheme 1).Since we deal with simple starting materials the optimization process was stopped at this stage.It should be noted that with these modifications even the bicyclic fluoride 7 and the tertbutyl derivative 5f are accessible, however, in the case of 5f only in 7 % yield.These data clearly demonstrate that the presence of an acceptor group (see 5d) as well as space demanding substituents (5e, 5f) lower the yield of the transformation.
As expected, all these cyclohexadienones (5a-5f) suffer from air oxidation and photodegradation.However, they can be kept for months if they are stored in the refrigerator at 0 °C under argon.

Cycloaddition experiments
With a good selection of dienophiles available, we focused on a detailed investigation of their cycloaddition behaviour.In order to demonstrate the rate enhancing effects of the fluoro atom 7 we started with a high pressure competition experiment employing methyl derivative 5a, which had already been shown to undergo high yielding, diastereoselective cycloadditions at 6,5 kbar, 5 and the corresponding quinol 8 (Scheme 3).Although this competition experiment was run at 14 kbar for six days, only the fluoro derivative 9a was formed and proved to be identical to the product obtained earlier at 6,5 kbar.Therefore the cycloadditions with 5b-5f could be run at both pressures and both cases smoothly provided the corresponding adducts 9b-9f in good to excellent yields.
A special observation that was particularly made starting from 5b and 5c deserves further comments.In some cases without any detectable reason or predictability, no cycloaddition took place at all and the whole material was completely destroyed leading to a black suspension.
Assuming the formation of radicals to be responsible for these undesired processes, we tried a few radical scavengers and were finally successful with the addition of 4-N-acetylamino tetramethylpiperidin-N-oxyl (TEMPO), 15 which leads to perfectly reproduceable yields for 9a, 9b and 9c as given in Scheme 3.
HPLC-as well as NMR-data proved all the cycloadducts to be single stereo-and regioisomers.Similarities to the well established data of 9a 5 together with NOE measurements (see experimental) indicated the configuration as portrayed in Scheme 3.
Finally, we also studied the Diels-Alder chemistry of the two bicyclic dienophiles 7 and 11.While 7 resulted directly from the fluorination process (see Scheme 2), 11 was prepared from 5e via an intramolecular Michael addition (Scheme 4).In both cases the yields at 6,5 kbar are low (starting from 11 only 4 % cycloadduct).Therefore, we repeated this cycloaddition at 14 kbar for a fortnight and obtained 36 % of cycloadduct 12 instead.Taking into consideration that a kinetic resolution is undergone starting from the racemic mixture 11 the yield was clearly improved.The retro process with the material thus obtained provides excellent yields of enantiomerically pure 11, affording a simple route to homochiral fluorinated alkaloid building blocks.

Selective transformations of cycloadducts
It had been observed in numerous cases before that the special endo-configuration of adducts 9, 10 and 12 together with their conformational rigidity result in completely diastereoselective βapproach -in electrophilic as well as nucleophilic attack to the cyclohexenone moiety. 4,5 r instance this could be shown for the "flash-dihydroxylation" 16 of 9a which can be followed by ketal formation and also for the chemo-and diastereoselective nucleophilic epoxidation 17 of the allyl derivative 9b (Scheme 5).Interestingly, the ester derivative 9c gave different results when we tried to apply the flashhydroxylation -ketal formation sequence.Although hydroxylation worked as expected, ketalisation, which leads to nearly quantitative yields with methyl derivative 14a, never exceeded 73 % in this case.As prolonged treatment in the presence of acid led to the regeneration of diene 1, we assume hydrogen fluoride elimination, followed by a retro-Diels-Alder process and aromatization.The reasonable and quite useful formation of lactone 17 was observed on treatment with HCl in tetrahydrofuran, albeit in low yield.
Things became more complicated when we turned to nucleophilic attack to the unsaturated carbonyl group of 9a.
As expected, treatment with just one equivalent of L-Selectride in toluene at -78 °C led to the saturated ketone 18 in high yield (Scheme 6).Subsequent reduction with K-Selectride gave rise to the α-carbinol 20 exclusively.However, the reduction of 9a with 2,2 equivalents of L-Selectride did not even provide a trace of the desired product.To our surprise, the high yield reaction product turned out to be polycyclic ether 19, which obviously had been formed under extrusion of hydrogen fluoride.i ii iii iv Scheme 6. i: L-Selectride 1 eq., toluene, -78 °C, 15 min, 86 %; ii: K-Selectride, toluene, -78 °C, 15 min, 64 %; iii: L-Selectride 2.2 eq., toluene, -78 °C, 15 min, 82 %; iv: LiPF 6 , CH 2 Cl 2 , RT, 5 min, 49 %.This unusual and on the first glance certainly unexpected structure was confirmed by the facts that no fluoro atom was present anymore, that a hydroxyl group was missing, and that the NMR-signal of the methyl group in the six-membered ring showed up as a doublet, giving rise to the assumption that the fluoro atom had been replaced by hydrogen.Additionally, the double bond in the five-membered ring had disappeared.An X-ray analysis finally proved structure 19 (Figure 1).The importance of the lithium counterion for this strange behaviour is clearly demonstrated by the fact that ether formation was also noticed during L-Selectride reduction of the saturated ketone 18, proving this compound to be an intermediate en route to 19.On the other hand, K-Selectride reduction of this ketone uneventfully led to the secondary alcohol 20, which on treatment with lithium hexafluorophosphate in dichloromethane was finally converted to ether 19 again.
Obviously, the lithium cation in inert solvents like toluene or dichloromethane operates as a Lewis acid coordinating with the fluorine, which triggers the rearrangement (see 21, Scheme 7).Assuming that formation of ether 19 is due to the combined action of a Lewis acid and a hydride donor and that the Lewis acid behaviour of the counterion determines the fate of the fluoro atom, one had of course to check on Grignard reagents.Those reagents represent a combination of Lewis acid and nucleophile which on predictable 1,2-addition to the unsaturated ketone 9a should result in a highly reactive tertiary, allylic fluoride (see 22).Accordingly, we were not surprised to again isolate 68 % of a non-fluorinated compound after performing the Grignard reaction at -78 °C.As NMR data of this reaction product clearly point to the cyclic ether 23, we assume in this case SN´-substitution of the fluoride anion (see

22).
A very similar cyclisation took place in the lithium aluminium hydride reduction of the tertbutyl derivative 9f leading to the cyclic ether 24 (Scheme 8).Although run with a huge excess of lithium aluminium hydride, ether formation in the case of 24 was a much slower process than the Grignard reaction described above.Additionally, it proved to be a specialty of the sterically hindered tert-butyl compound 9f, which obviously only leads to intramolecular SN´-substitution of the fluoro atom.In contrast 9a under similar conditions gave rise to a mixture of ketone 18 (66 %) and carbinol 20 (34 %) in quantitative yield via an intermolecular 1,4 -1,2 hydride addition sequence.These results obtained with a Lewis base reducing agent were compared with the outcome in the case of the Lewis acid reagent DIBAL.Here again SN´-substitution dominates the process leading to the homoallylic alcohol 26.It can be assumed that Lewis acid assistance from the aluminium alcoholate formed in the reduction step (see 25) is responsible for the outcome of this reaction.Scheme 8. i: LAH 20 eq., THF, reflux, 1 d, 38 %; ii: LAH 1,2 eq., THF, RT, 5 min, 66 % 18 and 34 % 20; iii: DIBAL 2.2 eq., toluene, -78 °C, 15 min, 76 %.
Although we are presenting only a comparatively small selection of results obtained on nucleophilic attack to the carbonyl system it demonstrates the enormous impact of the fluorine on the outcome of this process.Minor changes in the starting material as well as in the reagent may lead to a complete change in the reaction pathway.It would certainly need a profound and detailed investigation of all options to arrive at safe rules and predictions.

Retro Diels-Alder chemistry
Since all the cyloadducts described so far can be obtained with perfect face selectivity and undergo any subsequent transformation with very high diastereoselectivity, one could expect homochiral fluorine substituted cyclohexenones from their thermal retro Diels-Alder processes.
While compounds with free hydroxyl groups like 13 and 16 failed completely during Retro-Diels-Alder processes, good to excellent yields could be obtained starting from protected derivatives 14a, 14b and 15 (Scheme 9).It is worth to mention that the indolinone adduct 12, although suffering from poor yields in its preparation, surprisingly undergoes the retro process in quantitative yield.

Conclusions
All cycloadditions employing diene 1 and the fluorocyclohexadienones 5a-5f, 7, and 11 have been shown to proceed with excellent regio-and face-selectivity to provide in all cases a single homochiral Diels-Alder adduct.
Subsequent oxidation reactions at the remaining double bond were characterised by exclusive β-attack leading to reaction products that could be transformed easily into homochiral cyclohexenones in a thermal Retro Diels-Alder reaction.
While these oxidations parallel those with non-fluorinated compounds, diastereoselective nucleophilic attack to the unsaturated carbonyl group led to hydroxyl intermediates, which rearranged under formation of cyclic ethers and under extrusion of the fluoride anion.

Experimental Section
General Procedures.All solvents were destilled before use.Column chromatography was performed on silica gel (Baker, 30-60 µm) with the indicated eluent.High pressure of 14 kbar was achieved with a high pressure apparatus from Hofer.Reactions with 6,5 kbar were performed with a high pressure apparatus from Nova Swiss using the pneumatic pump 550.0441-A.FVP (flash vacuum pyrolysis) was carried out in an apparatus presented by Seybold und Jersak 19 .Melting points were determined with a melting point apparatus from Gallenkamp and are not corrected.IR-spectra were recorded with a Perkin-Elmer FT 1710 spectrometer.The wave numbers ν of the characteristic bands are given in cm -1 .The intensities are abbreviated as follows: s (strong), m (middle) und w (weak).NMR spectra were recorded on a Bruker AM 400 (400 MHz for H NMR and 13 CDCl 3 (77.0)for 13 C NMR. Coupling constants J are presented in Hz.
Multiplicities for 1 H NMR are abbreviated as follows: s (singlet), d (doublet), tr (triplet), q (quartet), m (multiplet).Broad signals are marked with a br.The number of protons is electronically integrated.Multiplicities for 13 C NMR have been obtained by using the DEPT (Distortionless Enhancement by Polarisations Transfer) technique.The signals are named as follows: p = primary, s = secondary, t = tertiary und q = quarternary carbon atoms.Mass spectra (MS) were determined on a Finnigan MAT 312 at an ionisation potential of 70eV and the indicated temperature.Intensities are in relative relation to the base peak.Mass spectra (HRMS) were performed on a VG AUTOSPEC spectrometer using the peak-matching method.FAB (fast atomic bombardment)-spectra were determined with a VG AUTOSPEC in a m-nitrobenzylic alcohol matrix at low resolution.Intensities are given relative to the base peak.

Acetonide adduct (14a).
To a solution of diol adduct 13 (102 mg, 0.255 mmol) and a catalytic amount of p-toluenesulfonic acid in dry DMF (5 ml) 2-methoxypropene (120 µl, 1.275 mmol) was added at 0 °C.The reaction was stirred over night at room temperature and then diluted with Et 2 O.The organic layer was washed two times with brine, dried over MgSO 4 , and filtered.Concentration in vacuo and subsequent flash chromatography on silica gel with petroleum ether/Et 2 O (3:1) provided 14a (111 mg, 99%) as a white solid, mp 163.9 °C.IR (CHCl
Teflon tube and exposed to a pressure of 14 kbar for 14 days.After evaporating the solvent under reduced pressure and flash chromatography on silica gel with petroleum ether/Et 2 O (5:1)