Pd‐Catalyzed Allylation of Imines to Access α‐CF3‐Substituted α‐Amino Acid Derivatives

We herein report a high yielding protocol for the direct α‐allylation of easily accessible trifluoropyruvate‐derived imines using Pd‐catalysis. The reaction gives access to a variety of different α‐allylated‐α‐CF3‐amino acids in a straightforward manner, starting from commercially available trifluoropyruvate. We also provide a proof‐of‐concept for an enantioselective protocol (up to er = 75:25) by using chiral phosphane ligands.


Introduction
Syntheses and applications of fluorine-containing α-amino acids (α-AA) are heavily investigated research fields and the incorporation of such amino acids in peptides or proteins is an appealing strategy to alter their (bio)-chemical and (bio)-physical properties. [1][2][3] One particularly powerful approach to influence the nature of biologically active molecules is the introduction of a trifluoromethyl group. [4] Not surprisingly, α-CF 3 -containing α-amino acids (α-CF 3 -α-AA) have thus emerged as target molecules of high interest. [2] A broad variety of complementary strategies for the synthesis of (chiral) CF 3 -containing compounds are known, [5] either relying on the nucleophilic [6] or electrophilic [7] (late-stage) CF 3introduction on already appropriately substituted compounds, or making use of simple CF 3 -containing (commercial) building blocks to access further structural complexity. With respect to the synthesis of quaternary α-CF 3 -α-AA one common methodology is to make use of simple CF 3 -ketimines and install the α-AA motive by means of Strecker-type chemistry. [8] An alternative approach relies on the use of commercially available 3,3,3trifluoropyruvates 1, which are commonly used building blocks to access quaternary α-CF 3 -esters by means of nucleophilic additions either to the pyruvates themselves, [9] or by using the corresponding ketimines 2, which upon addition of different nucleophiles give diversely substituted quaternary α-CF 3 -αamino acid derivatives 3 (Scheme 1A). [10] In addition to using imines as acceptors for nucleophilic additions, the last years saw an increasing number of reports demonstrating that the inherent reactivity of imines can be inverted (making the imine carbon nucleophilic). Such an imine umpolung can be achieved by using a suited benzylamine protecting group that allows for the formation of ambident nucleophilic azaallyl anions under basic reaction conditions, which then preferably react with electrophiles in the α-position (compare with Scheme 1B). [11][12][13][14][15][16] Accordingly, these umpolung approaches allow for the synthesis of valuable acyclic [12,13] or cyclic [14,15] products by starting from easily accessible imines. Interestingly however, CF 3 -containing benzylic imines 2 have only sparingly been reported and utilized as building blocks so far. [15,17] This comes as a surprise, as the reaction between benzylamines and pyruvates 1 proceeds easily and already in the initial reports by Soloshonok and co-workers the very rapid tautomerization of the initially formed imines 2 to imines 4 was observed. [17] This fast isomerization allows for the direct formation of 4 by heating 1 with benzylamines and most likely proceeds via formation of the intermediate azaallyl anion I. Given this observation, it seemed very likely to us that reactions of both, imines 2 (provided that these can be isolated) and the thermodynamically more stable imines 4 with suited electrophiles under basic conditions would proceed via α-attack predominantly. Overall such this concept would therefore result in a formal reactivity umpolung of the commercially available simple starting materials 1 and will give access to a variety of (chiral) α-CF 3 -α-amino acid derivatives in a direct and, compared to established protocols, complementary manner (Scheme 1B).
We have recently shown that preformed imines 4 undergo highly diastereoselective (3+2)-type cyclizations with Michael acceptors to access CF 3 -proline derivatives 6, with the α-position acting as the donor site. [15] Based on these observations, we now became interested in developing this concept further towards a more general approach to access (novel) acyclic α-CF 3 -α-amino acid derivatives in an unprecedented fashion. We opted for Pd-catalyzed α-allylation approaches, [18] as these would give access to highly functionalized α-allylated trifluoroalanine derivatives 5 (and upon imine hydrolysis 7) straightforwardly (Scheme 1C). [19,20]

Results and Discussion
We started our investigations by carrying out the racemic reactions between the ethyl ester 4a and the simple allylic acetates 8a and 8b (Scheme 2). Literally the first attempt with 4a and 8a in the presence of bis(dibenzylideneacetone)palladium (Pd(dba) 2 ) as a simple and cheap Pd(0)-source and 1,4-bis(diphenylphosphanyl)butane (dppb) resulted in the almost quantitative formation of 5a (using aqueous KOH as a base in acetonitrile as the solvent). The reaction was then found to be rather tolerant to different solvents and CH 2 Cl 2 , toluene and THF were equally well suited (with NMR yields > 90 % in all cases after 1-2 h). Other aqueous alkali hydroxide bases were well tolerated too, and a catalyst loading of 3 mol-% Pd(dba) 2 and 3 mol-% Scheme 2. Optimized reaction conditions for the racemic α-allylation of 4a with allylic acetates 8a and 8b (dba = dibenzylideneacetone; dppb = 1,4bis(diphenylphosphanyl)butane: IST = internal standard).
Eur. J. Org. Chem. 2019, 7122-7127 www.eurjoc.org dppb was found to be the optimum (lower amounts unfortunately did not allow for full conversions even after longer reaction times). Unfortunately, however, product 5a hydrolyzed relatively quickly during normal phase silica gel column chromatography and could therefore only be isolated in around 35 % yield after column chromatography. Thus, crude 5a was directly hydrolyzed in a quantitative manner by treatment with either HCl or TFA. This procedure also allowed for a simple extractive purification, giving isolated 7a in around 85 % isolated yield (over both steps) on up to 1 mmol scale (Scheme 2A).
Having identified operationally simple and high yielding conditions for the formation of the α-allylated amino acid derivatives 5a and 7a, we next tested the reaction of 4a with the cinnamyl acetate 8b (Scheme 2B). Interestingly, this transformation was found to be a bit more sensitive to the used solvent and incomplete conversion was observed in acetonitrile, while CH 2 Cl 2 and toluene allowed for complete conversion to 5b within 2 h again (other Pd-sources did not perform better in CH 3 CN). Concerning the catalyst loading again 3 mol-% of (Pd(dba) 2 ) were found to be the optimum. Surprisingly in this case however the use of 6 mol-% of the ligand (dppb) were necessary and other (i.e. chiral) ligands like e.g. BINAP were giving lower yields (please see Scheme 3 below for our attempts to develop an asymmetric variant). In contrast to 5a, compound 5b was found to be more stable under column chromatography conditions and could be isolated in 82 % isolated yield. The reaction turned out to be robust and could easily be carried out using 1 g of 4a (3.85 mmol) giving 5b in 77 % isolated yield. Hydrolysis with HCl then gave the ammonium chloride 7b quantitatively. The later compound could also be successfully employed for standard amide bond forming reactions with benzoyl chloride or N-protected glycine under classical peptide coupling conditions, as outlined in the online supporting information. [21] With these robust racemic conditions in hand, we next evaluated the scope of the reaction between imines 4 and differently substituted γ-substituted acetates 8 (Table 1). We initially wondered if a change of the imine-protecting group may influence the outcome, but when replacing the phenyl imine group in 4a by a p-NO 2 -phenyl group the outcome was not much different (entry 1), showing that this group does not significantly influence the reactivity of the nucleophile 4. A broad variety of different aryl-based acetates 8b-l were well accepted, all resulting in isolated yields between 70-90 % (entries 1-11). Only the thienyl-based acetate 8m reacted somewhat slower and the corresponding product 5m could only be isolated in 54 % yield (entry 12). We also tested the crotyl-based acetate 8n (entry 13), but unfortunately conversion was much slower in that case and the product tends to decompose rather quickly. Finally, the branched acetate 8o was employed as well (entry 14), which resulted again in the formation of the linear allylation product 5b under the Pd-catalyzed conditions, but in notably lower yield (and much slower) compared to the analogous reaction with 8b (compare entries 1 and 14).
Having demonstrated the applicability of the racemic protocol, we then put our efforts on developing an asymmetric variant. Here we focused on the use of the four commercially Table 1. Application scope of the racemic α-allyation using different γ-substituted acetates 8. [a] [a] All reactions were run at room temperature using 0.1 mmol 4 and 0.2 mmol 8. available chiral diphosphane ligands L1-L4. [22] We started by investigating the reaction between imine 4a and cinnamyl acetate 8b (Scheme 3). Unfortunately, the chiral ligands slowed down the reaction significantly, compared to the use of the achiral dppb (compare with Scheme 2B), and even with 6 mol-% Pd(dba) 2 and 12 mol-% of the ligands the reaction did not proceed to completion (longer reaction times did not allow to overcome this limitation as the reactions stalled before completion). In addition, in most of these attempts pronounced amounts of unidentified side products were formed, thus rationalizing the considerable difference between conversion and isolated yields. Besides these limitations also the enantioselectivity was found to be modest only (highest er = 71:29 obtained with ligand L2). A screening of different solvents and dilutions did not improve this outcome. In addition, we also tried other imine protecting groups (like the above-mentioned p-NO 2phenyl group) but without any improvement. It has been well described that the synergistic combination of chiral transition metal catalysis and asymmetric ammonium salt ion pairing catalysis can be rather fruitful to facilitate asymmetric allylation reactions. [13c,23] We therefore also tested the reactions with chiral phosphane ligands and in the presence of different asymmetric ammonium salts (i.e. Cinchona alkaloid-based ones [24] ) but in neither case any positive effect on the enantioselectivity was observed. Scheme 3. Attempted asymmetric α-allylation of 4a with cinnamyl acetate 8b (er determined by HPLC analysis using a chiral stationary phase).
As we found the racemic allylation with allylic acetate 8a to be somewhat more robust and tolerant to different solvents and lower ligand loading than the one with 8b (Scheme 2), we also investigated the asymmetric allylation of 4a with 8a under a variety of different conditions (Scheme 4). Unfortunately, the quick hydrolysis of 5a under protic and/or slightly acidic conditions turned out to be challenging for determining the enantioselectivity. With any mobile phase and HPLC column the hydrolysis of 5a was found to be rather fast, making a reliable chiral HPLC analysis of this compound impossible. On the other hand, the quantitative hydrolysis to 7a was easily possible, but we were not able to identify any suited HPLC method to directly Scheme 4. Direct chiral 19 F NMR analysis of 7a using the chiral aluminium complex A1. determine the er of compound 7a. One option to overcome this limitation would be to carry out the derivatization of the free amine with FMOC, which was found to be possible in principle. However, this turned out to be rather time consuming, low yielding, and required a not that easy column chromatographic purification, which made this approach not practical for a fast screening. Alternatively, the direct use of chiral NMR shift reagents may allow for a simple determination of the enantiomeric composition by just mixing the (crude) reaction product with the shift reagent. [25] Some of us (H. Kim's group) have recently introduced a novel class of chiral aluminum complexes A (Scheme 4A) that allowed for the straightforward analysis of chiral amines using 1 H NMR spectroscopy. [26] Considering the simplicity of this approach, we reasoned that this would allow us to overcome the difficulties observed for the analysis of compound 7a with HPLC by using standard 19 F NMR to directly quantify the enantiomeric composition of the CF 3 -amino acid 7a. We were glad to see that mixing racemic 7a with 6 equiv. of the aluminum complex A1 leads to 1:1 splitting of the CF 3signal in the 19 F NMR spectrum (Scheme 4B vs. 4C). Applying the same method to enantioenriched 7a (see Table 2 for the reaction optimization) then allows for the rapid determination of the er as shown in Scheme 4D (the integrity of these results was confirmed by carrying out HPLC analysis of corresponding FMOC-derivatives as well [21] ).
With a reliable and fast chiral analysis method at hand, we then screened the asymmetric Pd-catalyzed allylation of imines 4 with allylic acceptors 8 ( Table 2 gives an overview of the most significant results obtained in a very detailed screening Eur. J. Org. Chem. 2019, 7122-7127 www.eurjoc.org of different conditions). First experiments between the phenylbased imine 4a and allylic acetate 8a using BINAP (L1) as a ligand showed that toluene may be slightly better suited than CH 2 Cl 2 or Et 2 O with respect to asymmetric induction (entries 1-3). However, the selectivities were quite low and thus other ligands were tested next (entries [4][5][6]. Here the DACH-based ligand L3 turned out to be the most selective, resulting in an er of 65:35 (entry 6). We then carried out a very detailed optimization with this ligand for the reaction of 4a with 8a (see entries 6-11 for the most interesting results). It turned out the combination of acetonitrile with aqueous KOH as a base under relatively dilute conditions gives the hydrolyzed product 7a in high yield (78 % yield over both steps) and with a modest enantioselectivity of 75:25. Unfortunately, this was the best result we could obtain under a variety of different conditions and also lowering the temperature or using other amounts of catalysts did not improve this outcome. Also, the analogous Me-ester of nucleophile 4 did give more or less the same result as well. We again tested the addition of chiral PTCs but those had no positive effect either. [21] We then changed the nature of the electrophile leaving group but neither the Boc-protected allylic alcohol (entry 12) nor allyl bromide (entry 13) did allow for higher selectivities. Finally, the imine protecting group was varied as well (entries [14][15][16], but none of these groups performed better than the simple phenylimine. Accordingly, despite this reaction performs well under a variety of conditions with respect to conversion and yield, the asymmetric protocol is currently limited to an er of 75:25 (see entry 9 for the most selective conditions).

Conclusions
In conclusion, we have developed a high yielding protocol for the direct allylation of easily accessible trifluoropyruvate 1based imines 4 under Pd-catalysis. This protocol gives access to a variety of different α-allylated-α-CF 3 -amino acids 5 and 7 straightforwardly and with modest enantioselectivities (up to er = 75:25) in the presence of chiral phosphane ligands. In addition, we also demonstrated the use of the chiral aluminum complex A1 as an operationally simple and reliable tool to determine the enantiomeric composition of the target molecules by 19 F NMR analysis.

Experimental Section
General details as well as the analytical details and characterization data of all the novel compounds can be found in the online supporting information. [21] General racemic allylation procedure: To a stirred solution of 3 mol-% Pd(dba) 2 and 3 or 6 mol-% dppb in CH 2 Cl 2 the corresponding acetate 8 (2 equiv.), the imine 4 (1 equiv.), and KOH (aq. 50 %, 1 equiv.) were added successively. The reaction mixture was stirred for 2 h at room temperature. After completion, the mixture was filtered through a pad of Na 2 SO 4 and washed with Et 2 O. After evaporation of the solvent the product was purified by column chromatography with CH 2 Cl 2 and heptanes (2:1) to yield products 5a-n in the reported yields (please note that compound 5a rapidly hydrolyses during column chromatography).
Hydrolysis: Compound 5a or 5b was dissolved in 1 mL of CH 2 Cl 2 and 2 mL of 3 N HCl were added. The mixture was stirred at room temperature for 45 min and the layers were separated. Afterwards the aqueous layer was evaporated to dryness to get the hydrolyzed products 7a and 7b in quantitative yield.