Stereoselective Allylic Alkylations of Amino Ketones and Their Application in the Synthesis of Highly Functionalized Piperidines

Chelated ketone enolates are excellent nucleophiles for allylic alkylations. Electron‐withdrawing groups on the allyl moiety allow subsequent intramolecular Michael additions giving rise to piperidines with up to five stereogenic centers.


Introduction
Piperidines are widespread found in nature, for example, as alkaloids, and are privileged structures in medicinal chemistry. [1] Amino acids derived from piperidines,s uch as pipecolica cid and the elongated homopipecolica cid are also common, for example, in peptidic natural products and peptidomimetics. [2] Calvine, ap iperidinic lactone isolated from al adybird beetle of genus Calvia [3] can be seen as ab icyclic homopipecolic acid derivative ( Figure 1). Highly substituted homopipecolicacid derivatives, for example, A have been used as synthetic intermediates in the synthesis of decahydroquinolinea lkaloids such as the lepardins, [4] while polyhydroxylated homopipecolic acids such as B act as glycosidase inhibitors. [5] Therefore, it is not surprisingt hat aw ide range of protocols have been developed for the synthesis of piperidines in general, and also for homopipecolic acids in particular. Whereas these approaches are generally applicable for the synthesis of 2,6-disubstituted [6] or 3-hydroxylatedp iperidines, [7] methodsf or the synthesis of higher substituted piperidines are significantly less developed. [8] Many procedures advanced for b-amino acid syntheses [9] can also be applied for homopipecolic acids. Besides homologation of the corresponding pipecolica cids using the Arndt-Eistert protocol, [10] 1,4-additions of amines are also very common.W hile an intermolecular amine addition requires as ubsequentc yclization step, [11] the intramolecular version (Scheme 1) is more straightforward, but the stereochemical outcomeo ft he cyclizationstep is difficulttoc ontrol. [12] Our group is involved in the synthesis of amino acids using chelated enolates as thermally stable and highly selective nucleophiles. [13] Theseenolates can be used in awide range of reactions,s uch as transition-metal-catalyzed allylic alkylations [14] or Michael additions, for example, towards nitroalkenes [15] or a,b-unsaturated esters (Scheme 2a). [16] Allylic alkylations cannoto nly be used for the synthesis of g,d-unsaturated amino acids, but also for the modifications of peptides in a highly stereoselective fashion. [17] Recently,w ec ould showt hat Pd-catalyzed allylic alkylations can also be performed with chiral chelated a-aminoketonee nolates with excellent diastereoselectivity (Scheme 2b). [18,19] The stereochemical outcome of the reaction is mainly controlled by the enolate geometry Scheme1.Homopipecolic acids via intramolecular aza-Michael addition. and the stericalb ulk of the side chain, which shields onef ace of the enolatei nt he allylation step. Especially good results are obtained with arylated enolates. Here the (Z)-enolate A is formed almost exclusively,a voiding 1,3-allyl strain [20] between the aromatic ring and the side chain, which would be significant in the (E)-enolate.
Our goal was now to use this approach for the synthesis of highly substituted homopipecolica cids and related structures via subsequent intramolecular aza-Michael additions. Therefore, an electron-withdrawing functionality at the double bond is required.

Results and Discussion
We started our investigationsw ith the previously synthesized unsaturated a-amino ketone 1a, [18] which could be reducedt o the correspondinga lcohol with excellents electivity (Scheme 3). The configuration of the new formed stereogenic center was determined by NMR of the corresponding oxazolidinone, whichw as obtained by treatment of 2a with base. [18] For the introduction of the required electron-withdrawing group we decided to use across metathesis with Grubbs II catalyst. [21] 2.5 Equivalents of methyla crylate and 10 mol %o fc atalyst were necessary for complete conversion. The excess of acrylate was required, because the styrene formed during metathesis can also react with the acrylatei nc ross metathesis. Acetylation of the secondary alcohol 3a and subsequent deprotection of the nitrogen gave rise to the free amine salt, which was subjected to base treatment to undergo the desired aza-Michael addition in good yield. The two diastereomeric homopipecolic acids 5a were formed as almost 2:1m ixture and their configurationw as determined by NMR andX -ray structure analysis. Obviously,t his protocol is well suited for the synthesis of tetra-substituted piperidine rings.
To get access to even highers ubstituted piperidines we decided to use chiral allylic substrates such as 6 to introduce an additional substituent at the 3-position (Scheme 4). With (R)-6 am atched situation was observed and the allylation product 1b waso btained as as ingle diastereomer.A lso the next step, the stereoselective reduction workedp erfectly.U nfortunately,w ewere unable to convert 2b into the corresponding a,b-unsaturated ester 3b,a lthough aw ide range of metathesis protocolsh as been investigated.
Therefore, we decided to changeo ur strategy and to introduce the electron-withdrawingg roup (EWG) directly via allylic alkylation using functionalized allylic substrates C (Scheme 5). This is am ore straightforwardp rotocol, but it was unclear which substituents and EWG's are favoring the allylic alkylation and which the Michael addition, because in general,b oth reactions can proceed already at À78 8C( depending on the substitution pattern).
In addition, these allylic substrates might undergo deprotonation of the p-allyl Pd intermediate under the basic reaction conditions, providing electron-poord ienes, whichm ight also cause side reactions. This might be the reason why these types of allylic substrates have been used only very sporadically in palladium-catalyzed allylic alkylations. [22] The corresponding methyl-substituted allylic substrates can easily be obtained from O-protected lactic acid ester via Dibal reduction/H orner-Wadsworth-Emmons( HWE) olefination, which providest he corresponding a,b-unsaturated ester 7a and ketone 8a as single (E)-isomer,w hile in case of nitrile 9a a 3:7( E/Z)-mixture was obtained, whichc ould easily be separated by flash chromatography.I faBoc-protecting group was used on the lactate, the leaving group for the allylic alkylation could directly be introduced. In case of other carbamates it is recommended to introducet his leaving group after the HWE reaction, because Dibal reduction in this case resulted in the formation of side products. To investigate also the influence of the substitution pattern at the doubleb ond we synthesized the isopropyl-and phenyl-substituted esters 7b and 7c from the corresponding a-hydroxye sters (Scheme 6).
With thesea llylic substrates in hand, we next investigated the allylic alkylation of several amino ketones with the lactic acid-derived allylic substrate 7a.U nder standard conditions we used 2.5 equiv LHMDS for the deprotonation (conditions A) and 1.3 equiv ZnCl 2 fort he formation of the chelated enolate complex. Alternatively,i ns ome cases also 2.05 equiv LDA were used as base (conditions B). In this case, less base was used to avoid epimerizationo ft he allylation product after the reaction with this stronger base. In general,b etter yields were obtained if the amino ketone was used in excess (1.5 equiv) compared to the allyl carbonate because of complete conversion of the allylic substrate (according to crude NMR and TLC). The results obtaineda re summarized in Ta ble 1.
To determine the diastereomeric ratio, the crude product was analyzed by NMR and also by HPLC, especially in cases where more than two diastereomers were formed( mismatched situations). In general, the diastereomers could easily be separated by flash chromatography,p roviding enantio-and diastereomerically pure products.

Ratio Entry
Ketone yield (entry 1). Only the formation of the (E)-configured double bond was observed. The stereoselectivity could be improved by quenching the reactionm ixture at À20 8C( entry 2). The reaction starts at around À65 8Ca nd is finished at À20 8C. Interestingly, the opposite enantiomer of 7a,( R)-7a gave even a better yield, although this seemed to be the mismatched case (entry 3). Here, also traces of the (Z)-isomer could be determined.T his effect becamem ore significant in reactions of the valine-derived benzylk etone 11 a.T he yields and selectivities were comparable to the resultso btained with 10 a,w hile no big difference was observed if the differentb ases are used (entries 4,5). In case of the (R)-7a ah igher ratio for the (Z)-isomer was formed. Here LDA not only gave ah ighery ield, but also a better (E/Z)-ratio. As imilars ituation wasf ound with the isoleucine-derived benzylketone 12 a (entries 8-10). The high diastereoselectivity probably results from ac lear preference of the conjugated (Z)-enolate C over the (E)-isomer. One might expect that the stereoselective outcome of alkylsubstituted enolates should significantly depend on the steric size of the substituent. Therefore, we alsos ubjected the ethyl and isobutyl ketones 10 b and 11 b to our reactionc onditions. However,w ith the methyl-substitutede nolater esulting from 10 b,n os ignificant differencei nd iastereoselectivity (3:97) was observed, almosti ndependent on the base used (entries 11-13). With these n-alkyl substituted ketones the yields were significantly lower,c ompared to the aryl derivatives. Under our standard conditions using 2.5 equiv LHMDS ay ield of only 59 %w as obtained. It could be improved to 80 %b yu sing 4equiv of base or by switching to the strongerb ase LDA. With (R)-7a the yield dropped again,and in this case also the diastereoselectivity (entry 14). Here 5% of the (Z)-isomer were formed. Increasing the steric bulk of the enolate substituent also increased the diastereoselectivity of the reaction, although here with asmalldecrease in yield (entry 15).
With these resultsi nh and, we next investigated the influence of the substitution pattern on the allyl substrate. Replacing the methyl substituent of (S)-7a by as terically more demandingi sopropyl group resulted in the exclusive formation of as ingled iastereomer, independent which ketonew as used (entries16a nd 17). But also here, the alkyl ketone 10 b gave significantly lower yield (entry 16).
Obviously,a llylic carbonates bearing an ester functionality as an electron-withdrawing group are excellent substrates for allylic alkylation. In noneo ft he reactions 1,4-addition of the enolatet owards the double bond was observed. Therefore, we also investigated the lactic acid-derived ketones 8a and nitriles 9a.T he resultso bserved with ketone 8a were comparable to those obtainedw ith the corresponding ester (Table 2), the diastereoselectivities were only slightly worse. But in this case a stronger dependence on the workupt emperature was observed.H ere, quenching the reaction at À25 8Cr esulted in a significant increasei nt he diastereoselectivity (entries 2/3, 4/5). As expected, the yields and selectivities were lower with the alkyl ketone 10 b (entry 6). Prolonging the reactiont ime had no significant influence on the yield of 16 b,o nly more side products are formed. In this case, 20-25 %o ft he corresponding Michael adduct could be identifieda so ne of the major side products. In all examples investigated, the (E)-configured allylation product was formed exclusively.
As mentioned earlier, the HWE-olefination provided am ixture of (E)-and (Z)-a,b-unsaturatedn itrile 9a.I ng eneral, such unselective reactions are undesired, but in this case, it was an advantage, because it gave us the opportunity to investigate the two different allylic substrates separately (Table 3). One might expect the formation of four differentproducts, depending on the allylic substrate used. The configuration at C-4 results mainly from stereoretention in the allylic fragment, while the configurationa tC -5 is controlled by the adjacent side chain at C-7. In all previousc ases the (E)-configured allylation product was formed almost exclusively.
The situation seemed quite different in case of the nitriles 9. Here, the (S)-configuredc ompounds obviously represent the mis-matched case (Table 3). Althought he allylation products were formed with good diastereoselectivities, as ignificant isomerization of the double bond was observed. With LHMDS as ab ase, the (Z)-isomer was formed preferentially from the (E)-allylic substrate (ratio E/Z 1:2) (entry 1), whilew ith LDA almost a 1:1m ixture was obtained (entry 2). If the (S,E)-9a is the mismatched case, the (R,E)-isomer should represent the matched one. And indeed, with this (E)-isomer,t he (E)-allylation product was formed with good E/Z-selectivity and perfect diastereoselectivity (entry 3).
On the other hand, if the p-allyl-Pd complexi ntermediate undergoes fast equilibration, the same product as from (R,E)-9a shoulda lso be obtained from the (S,Z)-isomer.V ery similar resultsw ere obtained using LHDMS as ab ase, although the diastereoselectivity for the minor (Z)-isomer was lower in this case (entry 4). Also in this example with LDA ah igher amount of the (E)-isomer was obtained (entry 5). To prove the generality of this observation, we also subjected some of our other aminoketones to the same reactionc onditions. Comparable results were obtainedw ith the valine-derived ketone 11 a and ethyl ketone 10 b,a lthough the E/Z-selectivity was worse (entries 6-10). With these allylationp roducts in hand, we came back to our originali dea to synthesize highly substituted homopipecolic acid derivatives via aza-Michael addition. In principle, all positions of the piperidine ring can be modified by using suitable aminoketone and allylic substrates. Also differentf unctionalities (ester,k etone, nitrile)c an be introduced at C-1 allowing furthers ynthetic transformations. Exemplarily,w es ubjected ester (4S,5R)-13 a to the previously described (Scheme 1) reaction conditions. Reduction of the keto functionality delivered also here as ingled iastereomer( Scheme 7). Cleavage of the Boc-group and cyclization under basic conditions provided the desired homopipecolic ester 21 in good overall yield as a $2:1 diastereomeric mixture, as determined by NMR, while the (2R)isomer was formed preferentially.U nfortunately,t he two diastereomers could not be separated at this stage, that's why we tried to acylate the secondary piperidine-N to get separable amides. Surprisingly,n oN-acylation was observed neither with Boc 2 On or Cbz-Cl, only uncharacterized by-products were obtained. With Ac 2 Oa na cetylation was observed, but not as expectedo nt he N-b ut on the OH-functionality.B ut in this Oacetylated form as eparationo ft he two diastereomers (2R)and (2S)-22 could be accomplished. To prove, if the acetate substituent has an influence on the stereoselectivity,w e changed also the order of steps. But if the OH-functionality was first acetylated andt hen the amine deprotected/cyclized, the diastereoselectivity was comparable(70:30).
To investigate the influence of the 4-methyl group and the double bond geometry on the stereochemical outcome of the reaction,w es ubjected all four nitrile allylation products 19 a to deprotection/c yclization. In case of the (4R)i somers, the (Z)isomer gave definitely am uch higher selectivity in the cyclization step (Scheme8). The acetate was also here found to be superior to the alcohols. In addition, we subjected the alcohols and acetates of the (4S)-series to cyclization. In case of the (E)isomer,t he configurationa tC -4 obviously has no tremendous influence on the selectivity,w hile the effect was significant in case of the (Z)-isomer.H ere, the reactionm ixture had to be warmed to 30 8Cfor one hour forcomplete conversion.

Conclusions
In conclusion we could show,t hat allylic alkylations of amino ketones are versatile tools in organic synthesis, not only for the synthesis of highly functionalized ketones, but also the generation of highlys ubstituted piperidines and homopipecolic acid derivatives. Up to five stereogenic centers can be incorporated,w hile at least three of them are formed in ah ighly stereoselective fashion. The substituent at C-6 originates from an a-amino acid and controlsm ost of the others. The configuration at C-4 is the result of ah ighlys tereoselective allylation of ac helated amino ketone enolate and the stereoselective re-ductiono ft he ketone functionality (C-5) is directed by the two adjacent stereogenic centers. The configuration at C-3 is transferred from the allylic substrate. Dependingo nt he configuration of the allyl carbonate and its double bond geometry,b oth stereoisomers can be obtained in ah ighly stereoselective fashion. Severale lectron-withdrawing groups can also be part of the allylic substrate, whicha llows the direct incorporation of esters, ketones or nitriles onto the piperidine ring. Obviously, the allylic alkylation is faster than ac ompetitive Michael addition, because only with the highly reactive a,b-unsaturated ketones as ignificant amount of Michael adduct was obtaineda s side product. a,b-Unsaturated ketones and esters give (E)-configured allylation products exclusively,w hile nitriles provide( E/ Z)-mixtures. Nevertheless, in the cyclization step all stereoisomers deliver the (2R)-configured piperidines preferentially,a lthought he selectivity depends on the substitution pattern and the olefin geometry.

Experimental Section
General remarks:A ll air-or moisture-sensitive reactions were carried out in dried glassware (> 100 8C) under an atmosphere of nitrogen. Dried solvents were distilled before use. The products were purified by flash chromatography on silica gel (0.063-0.2 mm). Mixtures of EtOAc and petroleum ether were generally used as eluents. Analytical TLC was performed on pre-coated silica gel plates (Macherey-Nagel, Polygram SIL G/UV254). Visualization was accomplished with UV-light and KMnO 4 or Ninhydrin solution. Melting points were determined with aL aboratory Devices MEL-TEMP II melting point apparatus and are uncorrected. 1 Ha nd 13 CNMR spectra were recorded with Bruker AV II 400 [400 MHz ( 1 H) and 100 MHz ( 13 C)] spectrometer in CDCl 3 ,u nless otherwise specified. Chemical shifts are reported in ppm relative to TMS, and CHCl 3 was used as the internal standard. Mass spectra were recorded with aF innigan MAT9 5s pectrometer (quadrupole) using the CI technique.