Applications of nonstabilized enolates as nucleophiles in palladium catalyzed allylic alkylations t

Palladium catalyzed allylic alkylations are important reactions in organic synthesis. In general stabilized carbanions are used as soft C-nucleophiles. Nonstabilized enolates from ketones and esters often cause problems, and the developments and im provements made with these interesting nucleophiles are summarized in this review. With respect to ketone enolates, best results are obtained with the corresponding tin derivatives, but enoxyborates also can be used. With ketone enolates a-allylated products are obtained. In the presence of suitable chiral ligands, the optically active substitution products are obtained with high stereoselectivity. In contrast, sterically hindered ester enolates do not provide allylation products, but give rise to cyclopropane derivatives, while with silylketene acetals as nucleophiles, or vinylepoxidcs as allylic substrates, the a-allylated products arc obtained as well. This is also ture with more or less stabi lized enolates derived from pyrazinones or azlactones, giving rise to a-alkylated amino acid derivates. "Normal" unsatur ated amino acids can be obtained by using chelated amino acid ester enolates as nucleophiles. This enolates show a high reactivity and therefore these reactions can be carried out under \'ery mild conditions, conditions under which the n-u-n isomerization of the n-allyl intermediates can be suppressed. This opens up totally new synthetic possibilities. In the pres ence of chiral ligands optically active amino acids are obtained.


Introduction
Transition metal catalyzed reactions are becoming more and more popular in organic synthesis 1 . As many of these reactions tolerate a lot of functional groups, they are especially suited for the synthesis of complex molecules. Palladium, which can be used for cyclizations, cross-coupling reactions and allylic substitutions, holds a predominant position among the transition metals 2 . Especially the palladium-catalyzed allylic alkylation has developed into one of the standard reacuuns in organic synthesis, not the least because asymmetric variations of this reaction are pos-sible3. This reaction proceeds via an 1Z'-allylpalladium intermediate, which is formed from Pd(O) and an alkene with an leaving group in the allylic position.
The stereochemical course of the Pd-catalyzed allylic substitution has been studied extensively 4 . The first step, the attack of Pd(O) on a chiral allylic substrate such as I occurs from "the backside" under inversion of the configuration. The nucleophilic attack of 0-, N-, and soft C-nucleophiles on the 1Z'-ally I complex A again occurs from the ami face, and therefore overall retention of configuration is observed in the product 2 (Scheme 1). With unsymmetrical 1Z'-allyi-Pd complexes, attack of the nucleophile usually occurs at the less substituted position, but the tocdicated to Professor S. M. Mukherji.

JICS-2
regioselectivity is strongly dependent on the structural features of the substrate and the reaction conditions 2 . In most cases symmetrical nucleophiles such as malonates or disulfones are used to avoid a major problem of this reaction : the formation of a second stereogenic center in the "nucleophile moiety" of the product. Using unsymmetrical C-nucleophiles such as j3-ketoesters 5 or imines of amino acid esters 6 , mixtures of diastereomers are usually obtained. For example, reaction of the allylic acetate 3 with the unsymmetric nucleophile 4 gives the substitution products 5 in a I : 1 ratio 7 because of the configurational lability of the allylated nucleophile (Scheme 1). Thus, considerably better results are obtained with alkylated derivatives 8 . For example, allylation of "cyclic malonate" 6 with 7 provides 8 as a single stereoisomer 9 , while the attack of a chelated enolate obtained from 9 on allylic carbonate 10 gives rise to y,8-unsaturated amino acid derivative 11 in a highly stereoselecti ve fashion 10 . In contrast to the well investigated reactions of stabilized 'soft' carbanions, there are only a few reports concerning unstabilized enolates, for example those of ketones or esters, although the resulting products are often more interesting. The reactions of these enolates seems to be limited with respect to the substrates used, while one has to differentiate between ketone and ester enolates. This review will describe the developments and the improvement made during the last years. . Most attempts to improve the selectivity for monoallylation by variation of the catalyst, solvent and other reaction conditions failed. The problem arises from an equilibrium between the product and the starting enolate, which leads to loss of regioselectivity and to polyalkylation. Shortly after this first paper by Trost, Fiaud and Malleron reported on the use of sterically more demanding allylic substrates, such as cyclohexenylacetate (15), giving rise to the monoalkylated products 16 in acceptable yields 12 . The monoallylated product was also obtained exclusively in 59% by using the less reactive silyl enol ethers of 12, but herewith the reaction could not be extended to substituted allyl acetates.
A breakthrough brought a variation of the counterion of the enolate. Switching to tributyltin enolates such as 17 led to a remarkably rapid and clean monoalkylation with high regioselectivity (Scheme 3). Alkylation generally occurred at the sterically less hindered position of the n-allyl complex, and the £-configured product 19 was obtained preferentially independent of the olefin geometry of the starting allylic acetate (18) 13 . This can be explained by a fast n-a-nisomerization which will be discussed in detail later on. The palladium catalyzed allylic alkylation is compatible with a wide range of functional groups such as esters, ketones (20), alkyl bromides and enol ether, indicating a high chemo-selectivity.
The influence of the substituent in the allylic substrate on the regioselectivity (rs) of the reaction was investigated by Shi et at. 14 . They used several fluoro-and difluorosubstituted allyl acetates (20) and found a preferential attack of tin enol ate 17 at the terminus remote from the fluorine (21). The fluorinated acetates showed also a diminished reactivity in comparison to their nonfluorinated counterparts.
While the tin enolates can be generated in situ from the lithium enolates and tributyltin chloride, the yields in the alkylation step are not always satisfying. Better results are obtained if the tin enolates are formed from ketones and tin trifluoroacetates 15 . The reactivity of the tin enolates can be increased by switching from the tributyl-to the trimethylsubstituted derivatives. With these enolates acetophenone as well as several cyclic ketones can be allylated regioselectively, although the yields obtained are moderate in most cases 15 . The effect of the countercation on the outcome of the palladium catalyzed allylation was also investigated by Negishi and John in 1983 16 . They found, that allylation of potassium enoxyborates, obtained by reaction of ketones with either KH or KN(SiMe 3 ) 2 , followed by treatment with Kazmaier. : Applications of nonstabilized enolates as nucleophiles in palladium catalyzed ally tic alkylations BEt 3 , proceeded well in the presence of palladium phosphine complexes (Scheme 3). Various allylic halides and acetates can be used and the regioselectivity observed with both the "kinetic" and the "thermodynamic" enolates of 2methylcydohexanones (22) is in the range of90-95%. The allylation with trisubstituted allylic acetates such as geranyl and neryl23 derivatives proceeds with retention of the olefin configuration (24). Very similar results were obtained with the corresponding zinc enolates, while a wide range of other counterions (CIMg+, Me 2 At+, Me 3 AI!Li+, Cp 2 cm+, Me 3 si+) gave unsatisfying results or did not show any reaction at all.  If optically active ally tic substrates were used, the reactions proceeded with net retention. This can be explained by a double inversion, a mechanism which is typical for stabilized nucleophiles, and which is in sharp contrast to findings with aryl and vinyl metal compounds, which react under inversion (via transmetallation) 17 . Besides these substrate controlled allylic alkylations, asymmetric catalyzed ally lations became more and more important during the last years, and therefore it is not surprising, that this elegant technique found also some application to the allylation of ketones.
In 1999 Trost and Schroeder reported on the first asymmetric ally tic alkylation of tin enolates of2-substituted cyclohexanone derivatives 18 . In order for chiralligands to effect stereochemical control in this reaction, they must influence bond making and bond breaking events occurring outside the coordination sphere of the metal. Thus, they must transmit their stereochemical information through space. Discrimination of enantiotopic faces on the nucleophile is especially difficult as the nucleophile is segregated from the chiral environemnt by then-allyl moiety. For this purpose a chiral bidentate ligand 25 derived from trans-diaminocyclohexane was developed which gave good results with stabilized enolates and tin enolates of ketones as we11 19 . The O OAc 2 eq. LOA, OME ~ ) Scheme4 allylation of2-methyltetralone (26) (giving 27) with respect to the influence of the reaction parameters on the stereoselective outcome of the reaction was investigated (Scheme 4). While variations of the solvents had only moderate influence, the presence of additives and the choice of base had a significant effect. Best results are obtained by using 2 equiv. of LOA, and stannanes gave far superior results than did boranes and borates. Trimethylstannanes again proved to be the best choice. Other cyclohexanones gave comparable results.
Very recently Hou, Dai et at. reported on related investigations, also with 26, with a very similar chiral ligand 20 . They used phosphinocarboxylic acids derived from ferrocene instead of phosphinobenzoic acid and obtained similar results as reported by Trost.
Interesting investigations were carried out by Braun et at. not only on the enantiocontrol but also on the diastereoselective outcome of allylation reactions using 1.3diphenyl allyl acetate (28) in the presence of achiral and chiralligands 21 . Enolates of cyclohexanone (30) and propiophenones (such as 32) were used as nucleophiles. First of all they investigated the influence of the base and of achiral ligands on the diastereoselectivity of the allylation. Herein best results were obtained with lithium and magnesium enolates in the presence of ferrocenyl derived ligands 29, both with cyclic and aromatic ketones. While cyclohexanone (30) gave rise to the syn-product 31 preferentially, the antiproduct 33 was obtained from mesitylethyl ketone 32. Surprisingly, the enol ate geometry in the latter case had no significant influence on the stereoselective outcome of thereaction. The (E)-as well as the corresponding (Z)-enolate gave the same major diastereomer, albeit the (E)-isomer gave the better selectivity ((E) : 90% ds; (Z) : 67% ds). It was suggested that different transition states are involved in the allylation step, a theory that has to be veryfied.
With the chiral ligand (R)-Binap 34 excellent enantioas well as diastereoselectivities were obtained in the reaction of 30, if the magnesium enolate was used. Unfortunately, no yields were given for the allylation step 21 .

Allylic alkylations of ester enolates
A first report on the use of ester enolates as nucleophiles in allylic alkylations was described by Hegedus eta!. 1980 22 . They investigated the allylation of branched ester enolates in the presence of various ligands. The limitation of the palladium catalyzed allylic alkylation to relatively stabilized (pKa < -17) anions may be due to the propensity of nonstabilized carbanions to attack the metal in preference to the allyl group 2 3, resulting in reduction of the complex 960 rather than alkylation. However, addition of HMPA to this reaction completely suppresses reduction and permitts the alkylation to proceed with nonstabilized carbanions as well 24 . For example, treatment of n-allylpalladium chloride (36) with the enolate of methyl cyclohexanecarboxylate (37) 40  Surprisingly, cinnamyl acetate did not provide the phenyl substituted cyclopropane derivative but the linear cinnamylated product. Other substituted allylic acetates completely failed to undergo allylic alkylation. Branched ester enolates were the only carbanions studied that led to cyclopropanation. For example, anions of methyl hexanoate or acetophenone failed to give material containing an allyl fragment in any form under these reaction conditions, although the n-allylpalladium complex was consumed.
Unfortunately, attempts to trap some intermediates with carbon monoxide to obtain carbonylated products failed. Although carbon monoxide is not incorporated into the allylation product, Hoffmannn eta/. found an influence of CO on the yield of the reaction 25 . They observed, that the combination TMEDA/CO is superior to HMPA/NEt 3 under the same reaction cQnditions. Under these conditions not only sterically hindered ester enolates can be reacted with n-allylpalladium complexes such as 40, but also deprotonated amides, lactames, ketones, sulfones as well as Evans-enolates (41) 26 . Tertiary anions gave the best results.
Hoffmann et al. also undertook investigations concerning the mechanism of the reaction 27 . In the reaction of a diphenylallylpalladium complex 43 with deprotonated isobutyronitrile 44 they wre able to isolate the intermediate palladacyclobutane 45, which could be characterized by Xray structure analysis. It is very reasonable, that this palladacycle is a real intermediate in this reaction, because addition of CO resulted in a precipitation of Pd 0 and the formation of the cyclopropane derivative 46.
Obviously, the sterically hindered nucleophile attacks directly the central carbon atom of the allyl ligand and not the palladium. The reductive elimination to the cisdiphenylated cyclopropane is made more difficult because of sterical hindrance between the two phenyl rings. The CO can act as a n-acid coordinating to the 16-electron metalacycle. This results in a decrease of electron density on the palladium and an easier reductive elimination giving rise to the cyclopropane derivative.

75%E
The great importance of nonproteinogenic amino acids including a-substituted derivatives led to an investigation of modified amino acid ester enolates as nucleophiles in the palladium catalyzed allylic alkylation. Najera et a/. 30 reported on the use of a chiral pyrazinone derivative (54), obtained from (R)-valine and (S)-alanine, as a new chiral auxiliary for the synthesis of y,8-unsaturated amino acids via palladium catalyzed allylation. Pyrazinone 54 underwent highly regia-and diastereoselective ally lations (95-99% ds) under neutral conditions if allylic carbonates 55 are used as substrates. The alcoholate liberated is obviously basic enough to deprotonate the auxiliary, indicating that 54 forms a (partly) stabilized anion (Scheme 7). The free amino acid could be obtained under relatively drastic conditions (6 N HCI, 150"), a significant limitation of this protocol,.which is obviousiy only suitable for a-alkylated amino acids.
Trost et Ariza applied a-substituted azlactones 57 as nucleophiles for the allylation in the presence of allyl palladium chloride (36) and chiral ligands such as 253 1 . The requested products 59 were obtained from allylic acetate 58 in high yields and excellent stereoselectivities. Unfortunately, this protocol is also limited to a-alkylated amino acid derivatives (enolization of the azlactone) and the cleavage of theN-benzoyl group is not a trivial issue and might cause problems in the synthesis of amino acids with sensitive side chains. Besides allylic acetates and carbonates also gem diacetates 60 can be used for this purpose, as illustrated in a nice synthesis of Sphingofungin F based on this approach3 2 .
; coo:: Pd(0Ac) 2 Our group is also involved in the synthesis of unnatural and nonproteinogenic amino acids since several years, while y,8-unsaturated amino acids play a major role. Chelated amino acid ester enolates, such as 63 were found to give especially good results in various types of standard enol ate 962 reactions including alkylations, aldol reactions or Michael additions 33 . Chelation causes a marked enhancement of thermal stability without having any negative influence on the reactivity of these enolates. Due to the fixed enolate geometry, their conversions often proceed with a high degree of stereoselectivity. We are especially interested in reactions which can not be carried out with non chelated enolates. Thus, chelated enolates of amino acid esters 63 undergo Claisen rearrangement when being warmed up to room temperature providing y,8-unsaturated amino acids 65 (Scheme 8) 34 . Using substituted (E)-allylic esters 64 products with syn-configuration are obtained diastereoselectively via a chair-like transition state. The anti-products, however, cannot be obtained as easily from the corresponding (Z)-es-ters35. For that reason, it was interesting to see if the antiproducts might be accessible via palladium-catalyzed allylic alkylation. Stabilization of the enolate by chelation should also diminish the tendency of the enol ate to coordinate to the palladium-a solution of the 'enolate problem'? Therefore the reaction of different N-protected glycine esters with allyl carbonate 67 was examined. Instead of the carbonate 67, the corresponding acetate or benzoate can be used as well, but with the moe reactive carbonates the yields and selectivities are usually better. To avoid transesterifications by the liberated ethylate the tert-butyl esters 66 were chosen. These esters were deprolonated with excess LHMDS, and the resulting lithium enolates were transmetalated to the corresponding chelated zinc enolates by addition of zinc chloride. After a short time, a solution of the allyl carbonate 67 and the palladium source were added. In general, the best results were obtained with allylpalladium chloride 36 in the presence of triphenylphosphine. As aresult of the high reactivity of the chelated enolates, the allylation already takes place under very mild conditions at -78°, giving rise to the desired monoallylated amino acid derivative 68. Most common N-protecting groups can be used with comparable success, although the Z-derivative was superior in this special case 10 . To figure out if the ally lie alkylation with these nucleophi les can be carried out also diastereoselectively when substituted allyl substrates are used, the influence of theN-protecting group in the reaction of the 1 ,3-dimethylated allyl carbonate 69 was investigated. Compound 69 was chosen because this compound generates a symmetrical nallylpalladium complex, so one can ignore regioselectivity problems. By far the best results were obtained with the trifluoroacetyl-(Tfa)-protected glycine ester 70. The yield was very good and the diastereoselectivity excellent. The diastereomerically pure product 71 was accessible after a . . . . X 96%ds

TfaNH COOISu
C+l-71 single crystallization step. As determined by X-ray structure analysis, the anti-product was preferentially obtained from the palladium-catalyzed allylic alkylation. Thus, two different procedures for the synthesis of substituted y,8-unsaturated amino acids are avai I able, which complement each another. The syn-diastereomers can be obtained by ester enolate Claisen rearrangement, and the anti-diastereomers via palladium-catalyzed allylic substitution.
To extend the potential of this approach an asymmetric version of this protocol is desirable. Depending on the allylic substrate used, two different strategies can be applied. Substrates such as 70 with two identical substituents at both allyl termini form symmetrical, achiral n-allylpalladium complexes, and therefore the stereochemical outcome of the reaction can be controlled via chiral ligands on the palla-dium3. This straightforward approach is very elegant, but unfortunately more or less limited to such symmetrical substrates.
In order to assess the scope of ligand directed asymmetric ally lations with chelated enolates, a representative set of substrates, including a large and a small acyclic and two cyclic allylic derivatives, was investigated (Scheme 9) 36 . As chiral ligands phosphinooxazoline 72, which is particularly suited for acyclic substrates as well as the cymantrene derivative 73.and the phosphinomyrtanic acid 74 which are better suited for cyclic compounds, were used 37 .
High levels of selectivity were achieved with 1,3diphenylallyl acetate (75) as substrate, especially with the chiral ligands 72 a diastereoselcctivity of up to 95 : 5 and ee's up to 94% could be obtained. Recrystallization of the crude product gave diastereo-and enantiomerically pure  (77) led to cyclohexenyl-glycine derivative 78, respectively, which displays antibacterial activity and, therefore, is of particular interest 38 . The chiral ligand 73 gave a nearly 1 : I diastereomeric mixture, while with 74 the syn product was formed preferentially. ee-Val~ ues ofupto 93% could be obtained with these ligands, which is remarkable for such cyclic systems.
On the other hand, if allyl derivatives with different substituents are used, unsymmetrical n-allylpalladium complexes are formed. Attack of nucleophiles on these complexes in general provides mixtures of regioisomers, depending on the substitution pattern of the allyl moiety. Although this can be problematic, these substrates also have a big advantage : if optically active allyl substrates are used, the n-allylpalladium complexes formed are chiral, and nucleophilic attack on these complexes provides optically active substitution products.
Based on the reaction mechanism, the stereochemical control in the allyl fragment is not a problem (double inversion), in contrast to chiral centers formed in the 'nucleophile moiety'. The allylic substitutions with chiral allyl substrates (such as 79)proceeded cleanly and with good yields (Scheme 10). The only regioisomers obtained were those with the double bond in conjugation to the phenyl ring. The diastereoselectivities of the reaction were high, depending +Ph'Y Ph 2.5 eq. LHMDS -~ on the substitution pattern at the allyl moiety. The methyl derivative 79 was superior to the ethyl derivative, probably because of steric reasons. It was a striking feature that the diastereoselectivities obtained with the acetates were a little worse in comparison to those obtained with the carbonates. The chirality could be transfered almost completely from the allyl derivative 79 to the product 80.

82
The stereochemical course of the reaction can be explained by the following model. Then-allyl complex 79' is formed via attack of palladium(O) on the ally tic carbonate 79 with inversion of the configuration. Subsequently, this complex reacts with the chelate enolate 69'. The attack of the nucleophile occurs in such a way, that the resulting double bond is conjugated to the aromatic n-system. Because of the staggered arrangement of the substituents, steric interactions in the transition state between the n-allyl complex and the nucleophile are minimized. The bulky tert-butyl group is located at ttie less hindered methyl-substituted side of the allyl complex. No significant interactions should be expected here as well. Obviously these interactions are stronger in the case of the corresponding ethyl substituted derivatives, resulting in lower selectivities. This model is also in accordance with the observation, that no significant diastereoselectivity can be obtained if the corresponding methyl glycinate is used.
Since the palladium-catalyzed allylic substitution with chelated ester enolates already proceeds at -78°, one might see a good chance to circumvent a nearly unsolved problem in palladium-catalyzed allylic alkylations : the n-a-nisomerization.
With respect to the reaction mechanism the configuration of the substitution products depends on the configuration of these n-allyl intermediates 39 . Thus, the oxidative addition of palladium(O) to (£)-allyl-acetates 79 and -carbonates leads to allyl complexes with syn/syn-configuration (79'), which react with nucleophiles to provide the corresponding (E)-substitution products such as 80 (Scheme 10). The synlanti terminology is used to describe the orien- Exceptions can only be observed if steric interactions either between the substituents in the allyl substrate 41 or between the allyl moiety and the ligands 42 destabilize the synl.~yn-complex. As a result, the antilsyn-complex is enriched and reactions with nucleophiles give rise to an increased amount of (Z)-substitution product 84. However, selective palladium catalyzed 43 conversions of (Z)-allyl substrates with retention of the olefin geometry remain an unsolved problem. A transfer of the (Z)-configuration from the allyl substrate to the product would only be possible if one could run the reaction at temperatures where isomerization reactions do not yet take place. Since the palladium-Ph 2.5 eq. LHMDS ~ ~ catalyzed allylic substitution with ~helated ester enolates already proceeds at -78°, these nucleophiles have a good chance to fulfill these precondition.
Reactions of (Z)-substrates with the glycine ester 69 gave interesting results. The re.action with the (Z)-carbonate 85 almost exclusively yielded the desired (Z)-substitution product 86 CZIE: > 99/l) (Scheme 12). The outstanding selectivities (98% ds, 97% ee) observed even surpassed the very good results of the reaction with the (E)-carbonate 79. In

JICS-3
contrast, the reaction with 81 furnished a (EIZ)-mixture in a very low yield. The selectivities were markedly worse than those obtained with the carbonate. The major isomer (77%) was identical to the product 80 resulting from the reaction with the (E)-carbonate 79, which is in good agreement with the reaction mechanism discussed. This difference in product formation can be explained by the higher reactivity of the allyl carbonate 85. Because this substrate already reacts at -78°, n-a-n-isomerization obviously does not occur. In contrast, the reaction of the acetate 81 takes place at a higher temperature during the warm up. Simultaneously, the isomerization is setting in and a partial conversion of the primarily formed antilsyncomplex into the more stable syn/syn-complex can be observed. The same complex is also formed from the carbonate 79.
If the n-a-n-isomerization can be suppressed like in this case, further interesting questions .arise : What would happen to allylic substrates with ( (anti position) would provide a product 88 with (E)-configuration of the double bond, whereas the attack at b (syn position) would lead to a (Z)-double bond (89). Which position is the more reactive one, the syn or the ami position?
The clarification of this question is of major interest, because symmetrically substituted allyl derivatives are normally used in asymmetric catalyzed reactions. Irrespective of the configuration of the starting material, achiral syn/syn n-allylpalladium complexes are generally formed and the nucleophilic attack on complexes of this kind can be controlled, for example, by chiral ligands 3 . Therefore, if enantiomerically pure allyl substrates are used, the chiral information gets lost during the reaction. But if it is possible to suppress the n-a-n-isomerization during the reaction of {Z)-configured substrates, and if one of the two allylic positions is pronouncedly more reactive than the other one, it should be possible to generate optically active compounds with these substrates as well.
The allylic alkylation with racemic 90 gave a very pleasing result. The only product to be isolated was the (E)-configured product 76 with good yield and diastereoselectivity. Keeping in mind, that the n-a-n-isomerization does not play a significant role under these reaction conditions, one can conclude that the anti position is much more reactive than the syn position 44 . For that reason, the investigation of a feasible transfer of chirality from optically active, symmetrically substituted (Z)-allyl substrates was obvious. The reactions with chiral allyl substrate 91 also provided the chiral (E)-configured substitution products 71 exclusively in a very good yields. Also in these case the selectivities, which the anti-products were formed with, were excellent. Moreover, the almost complete transfer of chirality shows that the reaction proceeds via the antilsyn-complex and not via the synlsyn-complex, which would inevitably lead to racemization.

Acknowledgement
First of all I want to thank my highly engaged coworkers, especially Dr. Franz Zumpe and Matthias Pohlman for their tremendous work done during the last years. Financial support by the Graduiertenkolleg, the Deutsche Forschungsgemeinschaft (SFB 247) and the Fonds der Chemischen Industrie is gratefully acknowledged. I also want to thank Degussa-Htils AG for donations of transition metals.