Catalytic Enantioselective Synthesis of Heterocyclic Vicinal Fluoroamines by Using Asymmetric Protonation: Method Development and Mechanistic Study

Abstract A catalytic enantioselective synthesis of heterocyclic vicinal fluoroamines is reported. A chiral Brønsted acid promotes aza‐Michael addition to fluoroalkenyl heterocycles to give a prochiral enamine intermediate that undergoes asymmetric protonation upon rearomatization. The reaction accommodates a range of azaheterocycles and nucleophiles, generating the C−F stereocentre in high enantioselectivity, and is also amenable to stereogenic C−CF3 bonds. Extensive DFT calculations provided evidence for stereocontrolled proton transfer from catalyst to substrate as the rate‐determining step, and showed the importance of steric interactions from the catalyst's alkyl groups in enforcing the high enantioselectivity. Crystal structure data show the dominance of noncovalent interactions in the core structure conformation, enabling modulation of the conformational landscape. Ramachandran‐type analysis of conformer distribution and Protein Data Bank mining indicated that benzylic fluorination by this approach has the potential to improve the potency of several marketed drugs.


Introduction
The incorporationo ff luorine into organic compounds is prominent in the pharmaceutical, agrochemical,and materials industries. [1] The unique characteristicso ft he CÀFb onde nable modulation of physicochemicalp roperties while mitigating steric contributions. [2] Ak ey attribute is the intrinsic polarityo f the CÀFb ond, whichc an induce conformational changes through electrostatic and dipolei nteractions with neighbouring functional groups. Installationof ac hiral CÀFb ond with a vicinal relationship to ah eteroatom or electron-withdrawing group is particularly valuable, as exploitation of the gauche effect allows predictable conformational control. [3] For exam-ple, fluorinated phenethylamines are especially valuableg iven the demonstrable utility of this compound class within bioactive molecules, [4] and the topological control afforded by the gauche effectc an enableb espoke biological targete ngagement. [5] Similarly,f luorine is often considered as am inimal change to block metabolism, for example of labile benzylp ositions. [2,6] Given the stereoselective nature of most metabolic processes,s ubstitution of as pecific CÀHf or CÀFi namethylene represents av ery efficient meanso fb enefiting from this effect, thus the introductiono ff luorine in as tereoselective fashionh olds significant appeal.
Deoxyfluorination of alcohols is the most commonm ethod to install C(sp 3 )ÀFb onds, and numerous reagents have been developed to facilitate this transformation (e.g.,S cheme 1a). [7,8] An oted problem with the introduction of ab enzyl fluoride by deoxyfluorination is the propensity for stereochemical erosion due to variable contribution of S N 1p athways (e.g.,S cheme1 b). [8a, 9] This can be ameliorated in some cases by altering the deoxyfluorination reagent in situ by using specific additives; however,t his generalp roblem arisess pecifically from the nucleophilic fluorinating reagents commonly used. An alternative strategy for C(sp 3 )ÀFb ond formation that avoids S N 2i sb y asymmetricprotonation of prochiral C(sp 2 )ÀFcentres; however, there are limited examples of this and none for benzylic C(sp 3 )ÀF. [10] Herein, we presentamethodf or the enantioselective synthesiso fb enzylic C(sp 3 )ÀFv icinal fluoroamines by asymmetric protonation of in situ-generated prochiral fluoroenamines (Scheme 1c). [10][11][12] This method allowst he formation of an ew CÀNb ond and ab enzylic stereogenic CÀFb ond in as ingle catalytic reaction, thereby providing direct modular access to chiral heterocyclic vicinal fluoroamines from readily accessible vinyl fluoride precursors. [13] As the process does not rely upon the use of af luorinating reagent, the issue of stereochemical erosioni sa voided. The mechanism is fully investigated,w ith refinement of previous proposals,a nd we also show how the relationship of the azaheterocycle, amine,a nd CÀFb onds provides unique conformational control,w hich could offer benefits in drugd esign by better alignment with bound ligand conformation.

Method development
We have previously shown that Brønsted acid catalysis enables conjugate addition and highly selectivea symmetricp rotonation of prochiral enamines (Scheme 2a). [14] However,t his was only amenable with the steric control of enamine geometry afforded by aryl substituents at the a-carbon: alkyl substituents led to poor geometry control of the intermediate enamine 3, resultingi nl ower enantioinduction in product 4.A lthough fluorineh as as mall steric footprint, we postulated that dipole minimization might provide an alternative selectivity determinant (Scheme 2b). Indeed, preliminaryD FT studies highlighted the preferred s-trans geometry for benchmark starting material 5a,w hichw as anticipated to assist geometricalc ontrol of the developing fluoroenamine 6 and enhancing enantioinduction in 7.
Based on this hypothesis, ab enchmark processw as established whereby 5a was subjected to aniline (8), and Brønsted acid catalyst 9a (Table 1; ar ange of catalysts were evaluated, vide infra and see the Supporting Information). Optimization of reaction parameters delivered as ystem that afforded the desired product 7a in high conversion and enantioselectivity (82 %a nd 96:4 er;e ntry 1). Severalo bservations relating to optimization were noted (see the Supporting Information for full details and additional experiments). Ethereals olvents (THF, CPME) were particularly effective (entries 1a nd 2), with other solvents affording good to excellent conversion but with notably poorer enantioselectivity (entries 3a nd 4). Lowering the reaction temperature from À10 to À20 8Ch ad little effect on enantioselectivity but impacted reactione fficiency (entry 5). A similar effect was observed by lowering catalystloading, where 10 mol %w as less efficient but maintained selectivity (entry 6). Controle xperimentss upported ac atalyst-promoted reaction that lacked background reactivity (entry 7).
The scope of the reactionw as investigated (Scheme 3). A range of aryl amine nucleophiles was accommodated with variation in functional group (e.g.,h alides, alkyl groups, BPin, heterocycles) and regiochemical substitution (ortho, meta, para) was typically accommodated while maintaining selectivity (Scheme 3a). Additionally,s ubstitution on the aniline nitrogen was tolerated (7m). It should be noted that the choice of solvent was important for conversion,d ue to solubility:p roduct precipitation as the reaction progressed becamep roblematic for certain substrates;h owever,c hanging ethereal solvent based on substrate (THF or CPME) resolved this issue. Scheme2.a) Our previous work and simplified mechanism indicating asymmetric induction is af unction of enamine geometry control. b) Preliminary DFT conformational analysissuggesting high enaminediastereoselectivity. Ar ange of vinylheterocycles wasa lso generally well accommodated( Scheme 3b). Variationi ns ubstitution of the benchmark 2-vinylquinoline was straightforward (7a, 7n-q)a nd the reactiont olerated quinoxaline (7r), benzothiazole (7s), and pyridine (7t), with the latter as ignificantlym ore challenging substrate due to its higher dearomatization barrier, hence requiring ah igh temperature for the reactiont op roceed, which negatively affects enantioinduction. Significantly, the reaction also allowse nantioselective formation of stereogenic CÀCF 3 bonds (11); however,c atalystl oadingh ad to be increased and reactiontemperature decreased to overcomeasignificant nonselectiveb ackground reactiono bserved for this substrate (see the SupportingI nformation).

Mechanistic analysis
Twom ain mechanistic pathways are possible for the key asymmetric protonation event (Scheme 4). The initial events commont ob oth pathways involve reversible protonation of the substrate (5a)b yt he catalyst to provide LUMO-lowered intermediate complex 12 ande nabling reversible aza-Michael using PhNH 2 (8)t od eliver key intermediate 13.T wo mechanistic pathways are then possible from this intermediate:p athway 1p roceeds via direct stereocontrolled protont ransfer from the aniliniumv ia TS1 and delivers the product-catalyst complex 15,w hich subsequently liberates the product (7a). Alternatively,i np athway 2p roton transfer from the aniliniumo f 13 to the phosphate (via TS2)d elivers 14,w hich undergoes stereocontrolled proton transfer via TS3 to deliver 15.I no ur previous report, [14] computational analysiss upported pathway 1, with selectivity arising from good shape and electrostatic complementarity between the catalyst and TS1 leading to the observed enantiomer.T hese purely quantum mechanical studies did not yield transition states that would have supported pathway 2(or other alternative mechanisms). Aseries of kinetic isotope effect experiments were conducted via the use of 15 Naniline and PhND 2 .H owever,t hese proved inconclusive, with independentr ate experiments (see the Supporting Information) resulting in observed 14/15 NK IE of approximately 0.8 and H/D KIE of approximately 1.8, which might be affected by the pre-RDS equilibrium associated with this reaction.
Goodman and others have shown that catalysis by BINOLderived catalysts,s uch as 9a,c an be studied effectively and efficientlyb yQ M/MM ONIOM calculations where the quantum mechanical aspects are described by B3LYP/6-31G** and the molecular mechanics by UFF. [15] Accordingly,amore exhaustive theoretical exploration was undertaken using this approach (Figure 1a nd Supporting Information).
Experimentally,n ob ackground reaction was observed, which was consistentw ith DFT calculations that indicated a Scheme3.Substrate scope and isolatedy ields. Enantiomericratios determined by HPLC analysis on ac hiral stationary phase. See the Supporting Information for details. [a] CPME, À20 8C, 5d.
prohibitively high barrierf or direct reaction of 5a with 8 (see the SupportingI nformation).
Complexation of 5a with 9a to give 12 is moderately favourable, with the preferred dipole-induced s-trans conformation of 5a also retainedi n12.T his initial complex is held together by aH -bond (OH···N = 1.63 )a nd aw eak peri CH···O=P interaction (2.40 ). Complex 12 then undergoes dearomatizing aza-Michael addition to deliver 13,w here the loss of aromaticity is compensated for by the formation of at ightly bound ionic interaction betweent he anilinium NH and phosphate (P=O···HNHPH = 1.38 and P=O···HN quin = 1.71 ). Rearrangementw ithin this complex by proton transfer from the aniliniumt ot he phosphate involves al ow barrier and yields a complexo ft he enamine (14)t hat is higheri ne nergy than reactants. All of these steps are therefore strongly reversible and no significant concentration of any of the intermediates subsequent to quinoline complexation would be expected-this was confirmed by parallel NMR experiments (see the Supporting Information).
Transition states leading to each low-energy conformation of complex 15-(R)a nd 15-(S)w ere optimized leading to an array of conformations for each of TS1 and TS3.C onsistent with our previous report, [14] pathway 1, direct proton transfer to the prochiral centre was identified (TS1). This process has a significant barrier (+ 35 kcal mol À1 )b ut is predicted to be highly stereoselective (DDG°= 4.4 kcal mol À1 , > 99:1 er)i n favouro ft he experimentally observede nantiomer.A lthough this rationalizes the stereoselectivity of the process, it is not consistentwith the experimental rate of reaction.
However,p athway 2w as more consistent with the experimentally observed rate. The key step, in which the stereochemistry is generated, involves protonation of the enamine by the POH in complex 14 via TS3 and exhibits ac lear preference for the experimentally observed enantiomer (DDG°= + 3.5 kcal mol À1 ), which arises from geometricalr estrictions between the catalyst and enaminei nt he developing transition states TS3-(R)a nd TS3-(S)( vide infra). The iPr substituents of catalyst 9a are also particularly important for imposing this geometrical restriction (vide infra:T able 2a nd Figure 3, below). This mechanistic overview reveals that the catalyst provides its effect by acting as both acid and base at each stage as required andd oes so in aw ay that imposes specific shape requirements on the substrate that interplay with the polar interactions that hold the complex together. .T wo pathwaysare compared: Pathway 1: directp rotont ransfer and pathway 2: proton transfer to and from the phosphate. Free energiesink cal mol À1 are reported relativetos eparated substrate and catalystbound to THF.All calculations were performedinG aussian09, and free energiesatÀ20 8Cand 1 m concentration were obtained by using goodvibes. [16] Based on these results, ac omplete reinvestigation of the computational analysis of our previousp rocess using aryl substituents (Scheme 2a) [14] using the approachd elineated above suggestst hat Pathway 2i samore likely reactionm echanism in this process. The full profile for this reaction is provided in the SupportingI nformation.
Despitem oderate yields of product for each, enantioinduction wasp oor,w hich arises from features that are not well-tolerated in the lowest energy transition state. Substrate 7u places the fused phenyl ring in ap osition that clashes with the iPr groups of 9a in TS3.I nc ontrast, 7v prevents the required simultaneous interactions of 5a and 8 with 9a and also lacks the dipole-induced geometry control. Lastly,t he essential enamine NH-OP H-bond in TS3 is impaired by the adjacent chlorine in substrate 7w,w eakening the association between the substrate and catalyst.
With regards to the optimal catalyst, ac atalyst survey demonstrated the superior level of asymmetric induction using 9a (Table 2; see the Supporting Information for full details). To determinet he origin of this enhanced selectivity,w ea nalysed 9a in comparison to the related 3,3'-mesityl (9b)a nd -phenyl (9c) analogues ( Figure 3). As the stereodirecting group on the catalyst is reduced in size from 9a to 9b and 9c,t here is ag eneral tendency for the barrier for the catalysed reactiont oi ncrease (from 16.2 to 16.6 and 18.5 kcal mol À1 ,r espectively), resulting in the observed diminished conversion. This is accompanied by as harp erosioni nDDG°between TS3-(R)a nd TS3-(S)-the energy associated with TS3-(R)r emains similarf or all three, and this erosion is principally driven by ac hange in energy of TS3-(S). This is highlightedi nt he preferred conformation of each of the three structures equivalentt oTS3-(S) ( Figure 3).
The red arrows (Figure 3a)i ndicate where the bulk of the iPr groups of 9a press against both ends of the bounds ubstrate. This causes the break-up of an intramolecular H-bond between the aniline nitrogen and the NH of the nitrogen arising from the quinoline (2.48 for 9a,b ut 2.02 and 2.03 for 9b and 9c,r espectively);t his interchanges with an interaction between the aniline NH and ap hosphate oxygen (2.01 for 9a, but 2.69 and 2.67 for 9b and 9c,r espectively). The combination of the steric clashing and this change in hydrogen bonding pattern clearly disfavours TS3-(S)c ompared to TS3-(R)f or 9a;this differenceiss ignificantly reduced for 9b and 9c.
In line with experimental observations, the computational model also confirms al ower rate of catalysed reactionf or 5t, associated with the larger dearomatization barrier (see the Supporting Information). The observed significant background reactionf or 10 was also investigated computationally,c onfirming the acceleratingr ole of the LUMO-lowering CF 3 unit as previously observed for other Michael acceptors. [17] Implications for conformational control The value of the substructures accessible using the developed protocolw as explored by investigatingt heir conformational properties. The crystal structureo f7a shows an anti relationship between CÀFa nd the aniline nitrogen (dihedral angle = 1798), which is likely preferred in comparison to the gauche due to af avourable N pz ···s* C-F interaction ( Figure 4). The CÀF bond is almostp erpendicular to the carbonf ramework of the quinoline (dihedrala ngle = 1078), which we believe arises due to af avourable s* C-F ···p Ar interaction competing with CÀF/N quin dipole minimization. [18] Hydrogen bonding of the quinoline nitrogen with ah ydrogen bond donor would reducet his dipole and is ak ey feature of this system:t he crystal structure has an intermolecular hydrogen bond between the quinoline nitrogen and the HNPh in an adjacent molecule.
To reduce the impact of hydrogen bonding and any N pz interactions, 7a was acetylated to give 16 (Figure 4a). The preference towards the anti configurationi sd iminished, with the gauche conformation noted in the crystal structure (FCCN Aniline dihedrala ngle = 588). The dihedral angle between the CÀF bond and quinoline nitrogen is 1748,e xplicitly affectedb yt he CÀF/N Ar dipole minimization and no longer modulated by the other effects described.
Ramachandran plots for dihedrala ngles 1a nd 2o f7a, 16, and the parent 2-pyridylethylamine (not shown) were computed and revealt hat the introduction of the benzylic fluorideh as  ap rofound effect on theo verall conformational landscape (Figure 4a nd Supporting Information). Compounds 7a and 16 display specific low-energy conformationsb iased by the presence of the fluorine and that are likely to be populated in solution. This presents opportunities for application in drug discovery by improving bindinga ffinity and selectivity by decreasing the population of alternative, less favourable conformations. The protein databank was searched for ligands that contain the 2-pyridylethylamine substructure and the conformations that are populated in thesec rystal structures are mapped onto the Ramachandranp lot for 7a (Figure 4b,b lack dots). Five compounds in particulara dopt ac onformation that would be enhanced by the introductiono ff luorine at the benzylic position (Figure 4b,c ircled), including inhibitors of HIV reverse transcriptase, [19] cathepsin L, [20] and purine nucleoside phosphorylases (Figure4c). [21] This highlights the value of this structural change for enhancing potency and selectivity of potential drug molecules.

Conclusions
In summary,aBrønsted acid-catalysed aza-Michael/asymmetric protonation method for the synthesis of heterocyclic vicinal fluoroamines has been developed. The methoda llows access to stereogenic CÀFb onds in high selectivity and on as election of different heterocyclic templates. The method also translates to establishing stereogenic CF 3 analogues. The origin of the reactivity and stereoinduction has been investigatedb ye xtended DFT calculations that have established ap hosphate proton transfer as being more consistent with experimental observations than ad irect proton-transferp rocess. This has led to ar evision of our interpretation of the mechanism associated with our previous report. Conformational control of the vicinal phenethylamine system has been interrogated by DFT and crystallography, identifying the likelyp referred topologies in the solid and solution state. This might have strategic applications in drug discoveryb yi ntroducing conformational bias to access conformations more like the bound state, illustrated with examplesextracted from the PDB. [22]