Further Studies on the [1,2]-Wittig Rearrangement of 2-(2-Benzyloxy)aryloxazolines

The behaviour of 14 ortho-functionalised 2-aryloxazolines (11 of them prepared and characterised for the first time) with butyllithium has been examined. Significant limitations to the Wittig rearrangement of such systems are revealed. In terms of asymmetric Wittig rearrangement, good diastereoselectivity is obtained with a valine-derived 4-isopropyl oxazoline, but this is compromised by racemisation upon hydrolysis. More encouraging selectivity is achieved in the Wittig rearrangement of an acyclic phenylalanine-derived ortho-benzyloxy benzamide.


Introduction
Some time ago we described the reaction of 2-(2-benzyloxyphenyl)oxazoline 1 with strong base to give either the 3-aminobenzofuran product 2 resulting from intramolecular nucleophilic ring-opening of the oxazoline by the benzyl anion, or the oxazoline 3 in which the benzyloxy group has undergone a Wittig rearrangement (Scheme 1) [1]. While the aminobenzofuran formation could be optimised by using 3.3 equiv. of Schlosser's base (BuLi/KOBu t ) and applied to a number of substituted examples [1], the Wittig rearrangement process was not so favourable and, under optimal conditions of 2.2 equiv. butyllithium (BuLi) in THF, an isolated yield of just 29% was obtained. More recently, we have studied the competition between Wittig rearrangement and direct anion cyclisation in a series of three isomeric (benzyloxythienyl)oxazolines 4-6 and found the outcome to depend upon the distance between the two groups in the starting compound [2]. In the meantime, we have found the N-butylamide group, CONHBu, to be far superior in promoting the Wittig rearrangement [3], but several aspects of the oxazoline chemistry remain unexplored. Specifically, although thia-and aza-analogues of the direct cyclisation of 1 giving benzothiophene and indole products were described [1], Wittig rearrangement of these substrates has not been examined until now. In addition, cyclisation of the α-branched benzhydryl ether 7 occurred to give the 3-iminodihydrobenzofuran 8 (Scheme 2) and for the α-methylbenzyl ether 9 cyclisation gave the spiro oxazolidine-dihydrobenzofuran 10 with some stereochemical control [1]. Given the illustrious history of chiral oxazolines in controlling a wide range of asymmetric processes [4][5][6], we were interested in exploring their ability to direct the reactions of an α-branched ortho-benzyloxy group. In general terms, we were interested to examine the stereoselective cyclisation of the anion derived from 11 to give either 12 from direct cyclisation or 13 from cyclisation following Wittig rearrangement (Scheme 3), with these products leading, respectively, to 2-substituted dihydrobenzofuran-3-ones 14 or 3-substituted dihydroisobenzofuranones ('phthalides') 15 following hydrolysis. Introduction of the 2-benzyloxy group by a S N Ar reaction to give 21 was followed by carbonyldiimidazole-mediated condensation with 2-amino-2-methylpropanol to afford 22. Attempted cyclisation of this to oxazoline 23 using thionyl chloride resulted in loss of the O-benzyl group, but treatment with methanesulfonyl chloride in the presence of triethylamine yielded the required product. Similar condensation of 21 with butylamine gave the amide 24 in good yield. Unfortunately, treatment of both 23 and 24 with Schlosser's base, butyllithium or LDA under a wide variety of conditions did not yield any useful products and it appears that, far from promoting the Wittig rearrangement as in benzene-based systems, the oxazoline and amide groups actually prevent it in the pyridine systems.

Synthesis and Reactivity of 2-benzyloxy-3-pyridyloxazoline and N-butylamide
2-Benzyloxypyridines readily undergo Wittig rearrangement on treatment with base [12][13][14], and so we expected that they would do so even more readily in the presence of an activating oxazoline or amide group at the 3-position. We therefore targeted compounds 23 and 24, which were readily prepared starting from 2-chloronicotinic acid 20 (Scheme 5). Introduction of the 2-benzyloxy group by a SNAr reaction to give 21 was followed by carbonyldiimidazole-mediated condensation with 2-amino-2-methylpropanol to afford 22. Attempted cyclisation of this to oxazoline 23 using thionyl chloride resulted in loss of the O-benzyl group, but treatment with methanesulfonyl chloride in the presence of triethylamine yielded the required product. Similar condensation of 21 with butylamine gave the amide 24 in good yield. Unfortunately, treatment of both 23 and 24 with Schlosser's base, butyllithium or LDA under a wide variety of conditions did not yield any useful products and it appears that, far from promoting the Wittig rearrangement as in benzene-based systems, the oxazoline and amide groups actually prevent it in the pyridine systems.

Synthesis and Reactivity of Chiral 2-(2-benzyloxyphenyl)oxazolines
Since all our previous studies have involved 4,4-dimethyloxazolines, we first prepared the racemic 4-phenyl compound 28, starting from (±)-phenylglycinol and 2-benzyloxybenzoyl chloride 25 (Scheme 6). The resulting hydroxy amide 26 was expected to cyclise upon treatment with thionyl chloride, but instead the stable chloro amide 27 was formed. However, this could be cyclised to give the desired oxazoline 28 in essentially quantitative yield, using a literature method [15].

Synthesis and Reactivity of Chiral 2-(2-benzyloxyphenyl)oxazolines
Since all our previous studies have involved 4,4-dimethyloxazolines, we first prepared the racemic 4-phenyl compound 28, starting from (±)-phenylglycinol and 2-benzyloxybenzoyl chloride 25 (Scheme 6). The resulting hydroxy amide 26 was expected to cyclise upon treatment with thionyl chloride, but instead the stable chloro amide 27 was formed. However, this could be cyclised to give the desired oxazoline 28 in essentially quantitative yield, using a literature method [15].
For synthesis of the chiral oxazolines we made use of the fact that treatment of 2-(2allyloxyphenyl)oxazolines with two equivalents of Schlosser's base in toluene results in removal of the allyl group to give the corresponding phenol [1]. Thus, allyl could be used as a protecting group for the phenolic OH through the oxazoline synthesis (Scheme 7). Starting from 2-allyloxybenzoyl chloride 29, and treating with (S)-and (R)-phenylglycinol, (S)-phenylalaninol and (S)-valinol, respectively, gave the hydroxy amides 30-33 in good yield, which were converted into oxazolines 34-37 using methanesulfonyl chloride and triethylamine. Deprotection to give hydroxyphenyl oxazolines 38-41 was then followed by O-alkylation, using α-methylbenzyl bromide to give the target oxazolines 42-45 as mixtures of diastereomers. For synthesis of the chiral oxazolines we made use of the fact that treatment of 2-(2allyloxyphenyl)oxazolines with two equivalents of Schlosser's base in toluene results in removal of the allyl group to give the corresponding phenol [1]. Thus, allyl could be used as a protecting group for the phenolic OH through the oxazoline synthesis (Scheme 7). Starting from 2-allyloxybenzoyl chloride 29, and treating with (S)-and (R)-phenylglycinol, (S)-phenylalaninol and (S)-valinol, respectively, gave the hydroxy amides 30-33 in good yield, which were converted into oxazolines 34-37 using methanesulfonyl chloride and triethylamine. Deprotection to give hydroxyphenyl oxazolines 38-41 was then followed by O-alkylation, using α-methylbenzyl bromide to give the target oxazolines 42-45 as mixtures of diastereomers. Since some of the synthetic steps for the valinol-derived oxazoline 45, which turned out to be the most synthetically useful, proceeded in low yield, an alternative synthesis involving installation of the α-methylbenzyl ether before oxazoline formation was examined and this did indeed give a higher overall yield (Scheme 8). Hydrolysis of the ester 46 formed by O-alkylation of methyl salicylate gave the acid 47 which was converted into its acid chloride and treated with valinol to give hydroxy amide 48 and this could be cyclised in the normal way to afford 45. Attempted formation of the acid chloride from 47 in dichloromethane, as opposed to toluene, instead led to unexpected intramolecular transfer of the α-methylbenzyl group giving the known [16] salicylate 49. Scheme 6. Synthesis of racemic oxazoline 28.

Scheme 6. Synthesis of racemic oxazoline 28.
For synthesis of the chiral oxazolines we made use of the fact that treatment of 2-(2allyloxyphenyl)oxazolines with two equivalents of Schlosser's base in toluene results in removal of the allyl group to give the corresponding phenol [1]. Thus, allyl could be used as a protecting group for the phenolic OH through the oxazoline synthesis (Scheme 7). Starting from 2-allyloxybenzoyl chloride 29, and treating with (S)-and (R)-phenylglycinol, (S)-phenylalaninol and (S)-valinol, respectively, gave the hydroxy amides 30-33 in good yield, which were converted into oxazolines 34-37 using methanesulfonyl chloride and triethylamine. Deprotection to give hydroxyphenyl oxazolines 38-41 was then followed by O-alkylation, using α-methylbenzyl bromide to give the target oxazolines 42-45 as mixtures of diastereomers. Since some of the synthetic steps for the valinol-derived oxazoline 45, which turned out to be the most synthetically useful, proceeded in low yield, an alternative synthesis involving installation of the α-methylbenzyl ether before oxazoline formation was examined and this did indeed give a higher overall yield (Scheme 8). Hydrolysis of the ester 46 formed by O-alkylation of methyl salicylate gave the acid 47 which was converted into its acid chloride and treated with valinol to give hydroxy amide 48 and this could be cyclised in the normal way to afford 45. Attempted formation of the acid chloride from 47 in dichloromethane, as opposed to toluene, instead led to unexpected intramolecular transfer of the α-methylbenzyl group giving the known [16] salicylate 49. Since some of the synthetic steps for the valinol-derived oxazoline 45, which turned out to be the most synthetically useful, proceeded in low yield, an alternative synthesis involving installation of the α-methylbenzyl ether before oxazoline formation was examined and this did indeed give a higher overall yield (Scheme 8). Hydrolysis of the ester 46 formed by O-alkylation of methyl salicylate gave the acid 47 which was converted into its acid chloride and treated with valinol to give hydroxy amide 48 and this could be cyclised in the normal way to afford 45. Attempted formation of the acid chloride from 47 in dichloromethane, as opposed to toluene, instead led to unexpected intramolecular transfer of the α-methylbenzyl group giving the known [16] salicylate 49.
When the four chiral oxazolines 42-45 were treated with 3.3 equivalents of Schlosser's base, as used in the cyclisation of 9 to give 10, there was complete decomposition and no useful products could be identified. It is clear that moving to the 4-monosubstituted oxazolines has opened the way to unwanted reaction pathways, and since deprotonation at a benzylic position may contribute to this in 42-44, attention was focused on the valine-derived compound 45, where this is not possible, to optimise the conditions. When oxazoline 45 was treated with 2.2 equivalents of butyllithium in the absence of potassium tert-butoxide, a cyclic product was formed which proved to have the iminophthalide structure 50 resulting from Wittig rearrangement followed by cyclisation (Scheme 9). Although this was isolated in only 50% yield after preparative TLC, it appeared to be a single stereoisomer so it was of interest to determine the stereochemistry at the newly formed centre by formation of 50, followed by direct hydrolysis to the corresponding phthalide 51 for which the optical rotation is known. This proceeded in reasonable yield and showed that 51, and thus also 50, had the S configuration at the newly formed centre, but the rotation corresponded to only 5% e.e. It is interesting to compare the outcome here with the alternative oxazoline-based approach reported by Meyers [17], where reaction of the (2-acetylphenyl)oxazoline 52 with phenyl Grignard reagent gave the iminophthalide 53 which was then hydrolysed to afford 51 in 80% e.e. Application of the oxalic acid hydrolysis method in our case did not improve the e.e., and it seems that the iminophthalide 50 is much more susceptible to racemisation upon hydrolysis than 53. When the four chiral oxazolines 42-45 were treated with 3.3 equivalents of Schlosser's base, as used in the cyclisation of 9 to give 10, there was complete decomposition and no useful products could be identified. It is clear that moving to the 4-monosubstituted oxazolines has opened the way to unwanted reaction pathways, and since deprotonation at a benzylic position may contribute to this in 42-44, attention was focused on the valine-derived compound 45, where this is not possible, to optimise the conditions. When oxazoline 45 was treated with 2.2 equivalents of butyllithium in the absence of potassium tert-butoxide, a cyclic product was formed which proved to have the iminophthalide structure 50 resulting from Wittig rearrangement followed by cyclisation (Scheme 9). Although this was isolated in only 50% yield after preparative TLC, it appeared to be a single stereoisomer so it was of interest to determine the stereochemistry at the newly formed centre by formation of 50, followed by direct hydrolysis to the corresponding phthalide 51 for which the optical rotation is known. This proceeded in reasonable yield and showed that 51, and thus also 50, had the S configuration at the newly formed centre, but the rotation corresponded to only 5% e.e. It is interesting to compare the outcome here with the alternative oxazoline-based approach reported by Meyers [17], where reaction of the (2-acetylphenyl)oxazoline 52 with phenyl Grignard reagent gave the iminophthalide 53 which was then hydrolysed to afford 51 in 80% e.e. Application of the oxalic acid hydrolysis method in our case did not improve the e.e., and it seems that the iminophthalide 50 is much more susceptible to racemisation upon hydrolysis than 53.  When the four chiral oxazolines 42-45 were treated with 3.3 equivalents of Schlosser's base, as used in the cyclisation of 9 to give 10, there was complete decomposition and no useful products could be identified. It is clear that moving to the 4-monosubstituted oxazolines has opened the way to unwanted reaction pathways, and since deprotonation at a benzylic position may contribute to this in 42-44, attention was focused on the valine-derived compound 45, where this is not possible, to optimise the conditions. When oxazoline 45 was treated with 2.2 equivalents of butyllithium in the absence of potassium tert-butoxide, a cyclic product was formed which proved to have the iminophthalide structure 50 resulting from Wittig rearrangement followed by cyclisation (Scheme 9). Although this was isolated in only 50% yield after preparative TLC, it appeared to be a single stereoisomer so it was of interest to determine the stereochemistry at the newly formed centre by formation of 50, followed by direct hydrolysis to the corresponding phthalide 51 for which the optical rotation is known. This proceeded in reasonable yield and showed that 51, and thus also 50, had the S configuration at the newly formed centre, but the rotation corresponded to only 5% e.e. It is interesting to compare the outcome here with the alternative oxazoline-based approach reported by Meyers [17], where reaction of the (2-acetylphenyl)oxazoline 52 with phenyl Grignard reagent gave the iminophthalide 53 which was then hydrolysed to afford 51 in 80% e.e. Application of the oxalic acid hydrolysis method in our case did not improve the e.e., and it seems that the iminophthalide 50 is much more susceptible to racemisation upon hydrolysis than 53. Scheme 9. Products from Wittig rearrangement of 45 and an alternative approach to the same product [17]. Scheme 9. Products from Wittig rearrangement of 45 and an alternative approach to the same product [17].
It seemed likely that the change from direct cyclisation, as observed for 7 and 9, to Wittig rearrangement followed by cyclisation in the case of 45 is a result of the change from Schlosser's base to butyllithium, and this was confirmed by subjecting compound 9 [1] to the latter conditions which indeed produced the rearranged product 54 (Scheme 10). Interestingly, this product seems to exist in solution in equilibrium with a minor quantity of the spiro-oxazolidine form 54a, for which separate signals are observed by NMR. It seemed likely that the change from direct cyclisation, as observed for 7 and 9, to Wittig rearrangement followed by cyclisation in the case of 45 is a result of the change from Schlosser's base to butyllithium, and this was confirmed by subjecting compound 9 [1] to the latter conditions which indeed produced the rearranged product 54 (Scheme 10). Interestingly, this product seems to exist in solution in equilibrium with a minor quantity of the spiro-oxazolidine form 54a, for which separate signals are observed by NMR. In an attempt to achieve higher selectivity, we examined the use of the 4-isopropyl-5,5-dimethyloxazoline group [18,19]. Reaction of the amino alcohol with the acid chloride derived from 47 led to the hydroxy amide 55 (Scheme 11). Unfortunately, the high yielding cyclisation of this to form the oxazoline was accompanied by loss of the α-methylbenzyl group to give 56, which could be re-alkylated to afford the target oxazoline 57. In an attempt to achieve higher selectivity, we examined the use of the 4-isopropyl-5,5-dimethyloxazoline group [18,19]. Reaction of the amino alcohol with the acid chloride derived from 47 led to the hydroxy amide 55 (Scheme 11). Unfortunately, the high yielding cyclisation of this to form the oxazoline was accompanied by loss of the α-methylbenzyl group to give 56, which could be re-alkylated to afford the target oxazoline 57.

Scheme 10. Formation of Wittig rearrangement products from oxazoline 9.
In an attempt to achieve higher selectivity, we examined the use of the 4-isopropyl-5,5-dimethyloxazoline group [18,19]. Reaction of the amino alcohol with the acid chloride derived from 47 led to the hydroxy amide 55 (Scheme 11). Unfortunately, the high yielding cyclisation of this to form the oxazoline was accompanied by loss of the α-methylbenzyl group to give 56, which could be re-alkylated to afford the target oxazoline 57. Scheme 11. Synthesis of a 4-isopropyl-5,5-dimethyl oxazoline.
When this was treated with 2.2 equivalents of butyllithium there was extensive decomposition and no useful products could be separated. It therefore appeared that the simple valine-derived 4-isopropyloxazoline was the most promising auxiliary group and, with this in mind, the secondary alkoxy group was varied by alkylation of phenolic oxazoline 41 with the appropriate alkyl halides to give new chiral oxazolines 58-61 (Scheme 12). Although most of these were obtained in low yield they were fully characterised in each case. Scheme 11. Synthesis of a 4-isopropyl-5,5-dimethyl oxazoline.
When this was treated with 2.2 equivalents of butyllithium there was extensive decomposition and no useful products could be separated. It therefore appeared that the simple valine-derived 4-isopropyloxazoline was the most promising auxiliary group and, with this in mind, the secondary alkoxy group was varied by alkylation of phenolic oxazoline 41 with the appropriate alkyl halides to give new chiral oxazolines 58-61 (Scheme 12). Although most of these were obtained in low yield they were fully characterised in each case. Unfortunately, treatment of each of these compounds with 2.2 equivalents of butyllithium under the conditions optimised for 45 led only to decomposition and no useful products could be separated from the complex product mixtures.
In the light of the overall disappointing results obtained from Wittig rearrangement of the chiral oxazolines, and our discovery in the meantime [3] that a secondary benzamide could act as a much better promoter of the Wittig rearrangement, we decided to examine an asymmetric version of this process using the phenylalanine-derived alkoxy amine auxiliary 62 (Scheme 13). The amide 63 was readily prepared and treatment with 3.3. equivalents of an alkyllithium base did result in Wittig rearrangement. While use of n-butyllithium at either room temperature, 0 °C or −78 °C gave diastereomeric ratios of about 60:40, the best selectivity of 68:32 was obtained by using s-butyllithium at 0 °C. Furthermore, the resulting hydroxy amide underwent spontaneous cyclisation with regeneration of the chiral alkoxy amine auxiliary upon storage under normal laboratory conditions for a few weeks. Purification produced the phthalide 64 in good yield and with an e.e. of 37%. Unfortunately, treatment of each of these compounds with 2.2 equivalents of butyllithium under the conditions optimised for 45 led only to decomposition and no useful products could be separated from the complex product mixtures.
In the light of the overall disappointing results obtained from Wittig rearrangement of the chiral oxazolines, and our discovery in the meantime [3] that a secondary benzamide could act as a much better promoter of the Wittig rearrangement, we decided to examine an asymmetric version of this process using the phenylalanine-derived alkoxy amine auxiliary 62 (Scheme 13). The amide 63 was readily prepared and treatment with 3.3. equivalents of an alkyllithium base did result in Wittig rearrangement. While use of n-butyllithium at either room temperature, 0 • C or −78 • C gave diastereomeric ratios of about 60:40, the best selectivity of 68:32 was obtained by using s-butyllithium at 0 • C. Furthermore, the resulting hydroxy amide underwent spontaneous cyclisation with regeneration of the chiral alkoxy amine auxiliary upon storage under normal laboratory conditions for a few weeks. Purification produced the phthalide 64 in good yield and with an e.e. of 37%.
zamide could act as a much better promoter of the Wittig rearrangement, we decided to examine an asymmetric version of this process using the phenylalanine-derived alkoxy amine auxiliary 62 (Scheme 13). The amide 63 was readily prepared and treatment with 3.3. equivalents of an alkyllithium base did result in Wittig rearrangement. While use of n-butyllithium at either room temperature, 0 °C or −78 °C gave diastereomeric ratios of about 60:40, the best selectivity of 68:32 was obtained by using s-butyllithium at 0 °C. Furthermore, the resulting hydroxy amide underwent spontaneous cyclisation with regeneration of the chiral alkoxy amine auxiliary upon storage under normal laboratory conditions for a few weeks. Purification produced the phthalide 64 in good yield and with an e.e. of 37%. Scheme 13. Wittig rearrangement of a chiral benzyloxy amide and an alternative approach to the same product [20].
It is interesting to compare this result with the work of Matsui and coworkers [20] who used the same secondary amide auxiliary group to direct ortho-lithiation of 65 and reaction with benzaldehyde to give 66, the diastereomer of the Wittig rearrangement intermediate, which was then cyclised upon acid treatment to produce the enantiomer of 64 (Scheme 13). Given the much more encouraging result with amide 63, future work will focus on secondary amides rather than oxazolines to direct the asymmetric Wittig rearrangement of an adjacent benzylic ether group. Scheme 13. Wittig rearrangement of a chiral benzyloxy amide and an alternative approach to the same product [20].
It is interesting to compare this result with the work of Matsui and coworkers [20] who used the same secondary amide auxiliary group to direct ortho-lithiation of 65 and reaction with benzaldehyde to give 66, the diastereomer of the Wittig rearrangement intermediate, which was then cyclised upon acid treatment to produce the enantiomer of 64 (Scheme 13). Given the much more encouraging result with amide 63, future work will focus on secondary amides rather than oxazolines to direct the asymmetric Wittig rearrangement of an adjacent benzylic ether group.

General Experimental Details
NMR spectra were recorded on solutions in CDCl 3 , unless otherwise stated, using Bruker instruments and chemical shifts are given in ppm to high frequency from Me 4 Si with coupling constants J in Hz. IR spectra were recorded using the ATR technique on a Shimadzu IRAffinity 1S instrument. The ionisation method used for high-resolution mass spectra is noted in each case. Column chromatography was carried out using silica gel of 40-63 µm particle size and preparative TLC was carried out using 1.0 mm layers of Merck alumina 60 G containing 0.5% Woelm fluorescent green indicator on glass plates. Melting points were recorded on a Gallenkamp 50 W melting point apparatus or a Reichert hot-stage microscope.
To a stirred solution of the acid chloride prepared as above (12.65 g, 48.5 mmol) in CH 2 Cl 2 at 0 • C was added Et 3 N (6.76 cm 3 , 4.91 g, 48.5 mmol) and (S)-2-amino-3methylbutan-1-ol (5.00 g, 48.5 mmol) and the mixture stirred at rt for 18 h. The reaction mixture was poured into water (100 cm 3 ) and extracted with CH 2 Cl 2 (2 × 100 cm 3 ), the combined organic layers were dried over MgSO 4 and concentrated to give, after purification via flash column chromatography (Et 2 O/hexane 4:1) at R f 0.18, 48 (9.13 g, 58%) as a colourless oil as an inseparable 1:1 mixture of diastereomers; ν max /cm −  . The 1 H spectral data were in accordance with those previously reported [16], and the 13 C data are reported for the first time.  To a stirred solution of 2-(1-phenylethoxy)benzoic acid 47 (4.0 g, 16.5 mmol) in CH 2 Cl 2 (165 cm 3 ) was added (COCl) 2 (2.97 cm 3 , 4.40 g, 34.65 mmol) and two drops of DMF, and the mixture stirred at rt for 2 h, then cooled to rt and concentrated.