Introduction

Organic fluorides are a class of unique fluorine compounds, which feature with strong electronegativity, similar size to hydrogen, great influence on pKa and lipophilicity, and behavior of hydrogen-bond receptors1,2,3,4,5,6,7. Although fluorine-containing compounds have special physicochemical and biochemical properties, naturally occurring organofluorine compounds are extremely rare8. Fluorine earns other extensive applications in modern chemistry. For example, at least 25% of drugs and 50% agrochemicals contain fluorine9,10,11. Chiral-fluorinated amino acids are a class of fluorine-containing building blocks, which are wildely utilized in biorthogonal chemistry, drug modification, and asymmetric catalysis12,13,14,15,16. However, the development of chiral fluorine-containing amino acids still has numerous limitations17 as below. (i) The preparation of fluorine-containing peptides from natural amino acids is extremely rare, and the application of natural chemical connection is still in its infancy;18,19,20,21 (ii) compared with aromatic fluorination, the structural diversity of alkyl-fluorinated amino acids needs further study;22,23 (iii) the self-disproportionation of enantiomers is also an important factor to prevent the formation of optically pure fluorinated amino acids24,25.

Difluoromethylene has the highest dipole moment in the fluoromethane series and also exerts as bioisostere of ketone and ether, which may change the protein structure26,27,28,29. Chiral β,β-difluoro-α-amino acids (dFAAs) are vital building blocks for assembling bioactive molecules30,31. As shown in Fig. 1, the β,β-difluoro compound III acts as a better substrate for human CCRF-CEM folypoly-γ-glutamate synthetase;32 3,3-difluoro-3,4-dideoxy-KRN7000 analog II is identified as potent immunostimulator;33 the β,β-difluorophenylalanyl puromycin I reacts faster than puromycin at neutral or acidic condition34. However, the synthetic methods of chiral dFAAs are not mature, especially in catalytic asymmetric induction. There are only three main routes to prepare chiral dFAAs (Fig. 2). The group of Ayi reported the enzymatic hydrolysis of β,β-difluoro-α-amino esters via kinetic resolution35. Liu and his colleagues disclosed an effective method for controlling enantioselectivity by introducing chiral auxiliaries36,37,38,39,40. Uneyama uncovered an amazing asymmetric hydrogenation of α-fluoroiminoesters via catalytic amount of chiral BINAP ligand and palladium41,42,43. However, in the current situation, it is urgent to develop a general, efficient, and highly enantioselective dFAAS method under mild conditions.

Fig. 1: Representative β,β-difluoro-α-amino acid derivatives.
figure 1

I: β,β-difluorophenylalanyl puromycin; II: 3,3-difluoro-3,4-dideoxy-KRN7000 analog; III:CCRF-CEM folypoly-γ-glutamate synthetase.

Fig. 2: Strategies for asymmetric synthesis of β,β-difluoro-α-amino acid derivatives.
figure 2

Strategies, a enzymatic hydrolysis; b asymmetric hydrogenation; c chiral auxiliaries. Valuable molecules: difluoro peptides, difluoro amino alcohols, and difluoro alkaloid.

In the 1990s, Soloshonok et al. first studied the biomimetic [1,3]-proton shift, which opened up a new door for the synthesis of chiral β-fluoroalkyl β-amino acids44,45,46. Then, Jørgensen47, Deng48,49,50, Shi51,52, and others53,54,55 further extended this strategy to construct chiral amino acids and chiral trifluoromethylamines. Building upon our continued interests in organic fluorides56,57,58,59,60,61 and the limitations for the rapid synthesis of chiral fluoro amino acids, we report here an approach to achieve dFAAS via biomimetic enantioselective [1,3]-proton-shift reactions. Notably, the yielded chiral dFAAs are useful synthons for producing various valuable molecules, such as difluoro peptides, difluoro amino alcohols, and difluoro alkaloid62.

Results and discussion

Reaction optimization

We first studied the reaction of N-benzyl-2-((2-chlorobenzyl)imino)-3,3-difluoro-5-phenylpentanamide 1a as model substrate and toluene as solvent. The results are shown in Table 1. When cinchonan-6′, diol A was exploited, the expected 2a was acquired in 34% yield and 81% ee (entry 1). Almost racemic 2a was obtained by replacing the catalyst with quinine B (entry 2), while the yield and enantioselectivity were improved by using catalysts C and D (entries 3, 4). Further tests on catalysts E and F showed no significant change in yield or ee (entries 5, 6). To our delight, when G was tested, 2a was successfully formed with 87% yield and 95% ee and pseudoenantiomer H was also tested for delivering the enantiomer of 2a with 80% yield and 95% ee (entries 7 and 8). Solvent evaluation experiments reveal that toluene is the best choice (entries 9–12). Furthermore, reducing the loading of catalyst G to 5 mol% (entry 13) without decreasing the yield or ee implies that this highly enantioselective biomimetic [1,3]-proton shift can be achieved under suitable conditions. Even if the time is extended to 24 h and the loading is further reduced to 1 mol% (entry 14), the requirement of useless yield can not be met.

Table 1 Optimization of asymmetric isomerization of 1aa.

Substrate scope

Under the optimized catalytic conditions, we turned our attention to explore the generality of biomimetic [1,3]-proton-shift reaction. As illustrated in Fig. 3, R1 group was examined first. In the process of generating the [1,3]-proton-shift products 2a and 2b, it was found that the yield and enantioselectivity are independent of the adjacent CH2CH2 groups of the substituents (1a and 1b). Alkene (1c) is also compatible under this condition. Furthermore, the CH2 adjoint groups containing cyclohexyl, 1,3-dioxolan-2-yl, isopropyl, phenyl, methyl and alkene were well tolerated and provided good yields and high enantioselectivities (2d–2i). β,β-difluoro-2-aminobutyric amide (2j) was obtained with good yield and excellent ee. In the case of 2l that is a hydrolyzed derivative and bearing ester group, 72% ee and 62% yield were observed. Then, we turned to amide group R2, where isobutyl can be installed smoothly (2m). It is gratifying that dipeptides (2n and 2o) bearing leukine have been successfully provided for peptide synthesis and the diastereoisomers indicated that the source of chirality is induced by the catalyst rather than substrate itself. Pleasingly, the imines with or without substituted phenyl group are well tolerated under these conditions (2p–2r). β,β-difluoro natural amino acid derivatives (e.g., β,β-difluoro glutamine, β,β-difluoro leucine, and β,β-difluoro phenylalanine) were all obtained in good yields and high ee’s.

Fig. 3: Scope of β,β-difluorinated imines[a,b,c].
figure 3

a Conditions: 1a (0.1 mmol), cat. G (5 mol%), toluene (1.0 mL), room temperature, 4–16 h. b Isolated yield after flash-column chromatography. c Enantiomeric excess (ee) determined via chiral-phase HPLC analysis. d Recrystallization from hexane/ethyl acetate (20/1). e After hydrolysis. f dr was determined via crude 1H NMR. g Cat. H (5 mol%) was used.

Mechanistic studies and postulated mechanism

Next, we started to investigate the reaction mechanism. We want to understand whether the [1,3]-proton-shift process is intermolecular or intramolecular. Benzyl-deuterated 1a-d2 was prepared and the reaction carried out under standard conditions. The results show that 2k has no erosion on deuterated ratio (Fig. 4a). As a contrast, when deuterium oxide was added to the reaction system, no deuterium product was found (Fig. 4b).

Fig. 4: Mechanistic studies.
figure 4

Deuterization and cross deuterization experiment.

When 1a-d2 cross-reacted with 1t, no deuterium was found in product 2t (Fig. 4c). These experiments reveal that the [1,3]-proton-shift process is intramolecular. The parallel kinetic isotope effects (KIE) were measured with kH/kD of 4.0, indicating that the rate-determining step (RDS) is the process of hydrogen leaving from benzyl group (Fig. 5).

Fig. 5: KIE from parallel experiments.
figure 5

The parallel experiments of 1a and 1a-d2.

Based on the condition optimization, substrate scope, and the initial mechanistic experiments, a plausible mechanism for the biomimetic enantioselective [1,3]-proton-shift reaction was proposed (Fig. 6). First, the possible intramolecular hydrogen-bond interaction (blue) of the amide NH to imine of the substrate is essential, and without it, the ee decreased dramatically (e.g., 2l bearing an ester group resulted in 72% ee). Then, the catalyst with free OH of phenol is needful (e.g., cat. B with OMe only induced 2% ee) for assembling another hydrogen bond between it and the N of the amide (green). The bulk isopentyloxy of the catalyst enforced deprotonation of the inner H of benzyl (red) and this process was a RDS (TS-I). Finally, the asymmetric protonation of 2-azaallyl anion from Si face (TS-II) to deliver the target product with R-configuration.

Fig. 6: Proposed transition model.
figure 6

The possible mechanistic of the enantioselective [1,3]-proton-shift reaction.

Synthetic transformations and applications

In addition, our protocol is potentially suitable for large-scale preparation. As illustrated in Fig. 7, the use of 2 mol% catalyst G is sufficient to produce 2i (0.91 g) in 80% yield and 95% ee. Compound 3 can be achieved via hydrolysis and N-Phth protection. Next, the conversion of amide to ester group can generate the key intermediate 463. Finally, 3,3-difluoro-3,4-dideoxy-KRN7000 analog II can be made smoothly from 4a based on formal synthesis33. Meanwhile, the corresponding Fmoc-protected amino acid 5 can be prepared from 4b for further medicinal study. To further prove the practicability of dFAAs, difluoro leukine was assembled into linear difluoro-oxytocin (28% yield for overall steps), which could be folded smoothly by using glutathione oxidation (63% yield) (Fig. 8). Additionally, we demonstrated that the stability of folded difluoro-oxytocin (3 mM GSH, 100 mM PBS, pH 7.0, rt) is higher than WT oxytocin64,65,66,67(Fig. 9).

Fig. 7: Gram-scale synthesis and transformations.
figure 7

The gram-scale synthesis of 1i. The transformation of 2i to II and 2t to 5.

Fig. 8: Solid-phase peptide synthesis of difluorinated oxytocin.
figure 8

a About 20% piperidine/DMF, 10 min, double; b 1.5 equiv. fluorinated leucine, 1.5 equiv. HOBt, 1.5 equiv. DIC, minimum DMF, rt, 16 h; c 10% piperidine/DMF, 3 mins, double; d 3 equiv. amino acid, 2.9 equiv. HBTU, 6 equiv. DIPEA, minimum DMF, rt, 1.5 h; e 10% piperidine/DMF, 3 mins, double; f TFA cocktail, global deprotection, rt, 3 h; g GSH/GSSG folding, rt, over-night. Fmoc = fluorenylmethoxycarbonyl, HOBt = hydroxybenzotriazole, DIC = N,N’-diisopropylcarbodiimide, HBTU = N,N,N’,N’-tetramethyl-O-(1H-benzotriazol-1-yl)-uronium hexafluorophosphate, DIPEA = N,N-diisopropylethylamine, GSSG = glutathione disulfide.

Fig. 9: Folded difluoro-oxytocin stable than folded WT oxytocin.
figure 9

s4: fold WT-oxytocin, s5: linear WT oxytocin; s4′: fold difluoro-oxytocin, s5′: linear difluoro-oxytocin, s1–s3/s6–s9 scrambling, s1′–s3′/s6′–s9′ scrambling.

In summary, a biomimetic enantioselective [1,3]-proton shift of difluoroimines has been developed. This new protocol allows the rapid assembly of enantiomerically enriched dFAAs from readily available starting materials under mild conditions. Preliminary mechanism studies show that the proton transfer is intramolecular and the deprotonation step is RDS. More details on enantiocontrol are being conducted in our laboratory and will be reported in due course.

Methods

Synthesis of 2

To a flame-dried Schlenk reaction tube equipped with a magnetic stir bar, was added the 1 (0.1 mmol) and cat. G (1.9 mg, 0.005 mmol). The Schlenk tube was closed with a septum, and toluene (1.0 mL) was added. The mixture was then stirred at 25 °C and monitored by TLC until 1 was consumed. The mixture was concentrated under reduced pressure and purified by column chromatography on silica gel (hexane/EtOAc = 20:1–10:1) to afford the desired product 2. Full experimental details can be found in the Supplementary Methods.