Stereodivergent entry to β-branched β-trifluoromethyl α-amino acid derivatives by sequential catalytic asymmetric reactions

Currently, conventional reductive catalytic methodologies do not guarantee general access to enantioenriched β-branched β-trifluoromethyl α-amino acid derivatives. Herein, a one-pot approach to these important α-amino acids, grounded on the reduction – ring opening of Erlenmeyer–Plöchl azlactones, is presented. The configurations of the two chirality centers of the products are established during each of the two catalytic steps, enabling a stereodivergent process.


Introduction
b-Branched a-amino acids (AAs) carrying different b-substituentsthus bearing two vicinal chirality centres, at the aand bcarbonsare important yet challenging synthetic targets. 1 The stereocontrolled preparation of these compounds has been tackled and realised with different (catalytic) methods. 2-7 A noncomprehensive list of examples includes enantioselective conjugate additions and alkylations of glycinate imines, 2 palladium catalysed b-C(sp 3 )-H alkylation and arylation of AAs, 3 aziridine ring-opening, 4 an engineered tryptophan synthase, 5 multi-enzymatic b-methylation of AAs, 6 and catalytic asymmetric hydrogenation of a,b-dehydro-amino acids (DHAAs), in its implementation with tetra-substituted substrates. 7 This latter approach (Scheme 1(a)) has disclosed diastereo-and enantioselective entries to various classes of b-branched AAs, including 7d less common yet relevant b-triuoromethyl AA derivatives. 8, 9 Given the common stereospecicity of the hydrogenation, the E/Z geometry of the substrate dictates the relative conguration of the product (Scheme 1(a)). Thus, a diastereoisomer is obtainable only if the parent olen isomer can be prepared. This constraint can have negative implications. For example, a straightforward synthesis of a preclinical drug candidate via asymmetric hydrogenation of the corresponding b-aryl-b-triuoromethyl DHAA was envisioned (Scheme 1 (b)). 9a However, the required E-olen isomer could not be accessed with sufficient selectivity. Furthermore, enantioselective hydrogenation of tetrasubstituted olens can occasionally be challenging. In fact, the b-triuoromethyl Z-DHAAs required for the target was found to be reluctant to asymmetric hydrogenation, 9a preventing application of Turner's formal stereodivergent reduction of DHAAs. 10 This ingenious chemoenzymatic approach, which unlocks access to the isomer not obtainable by hydrogenation via downstream inversion of the aamino chirality centre in the reduced DHAA, is in fact established only for b-aryl-b-methyl AAs. Eventually, to obtain the target syn-triuoromethylated amino alcohol, § Alimardanov et al. resorted to a longer, yet effective and scalable, route, encompassing a chiral auxiliary and introduction of the amino functionality at a late stage. 9a With this background in mind, we envisioned an original stereodivergent 11 entry to b-branched AAs grounded on the dynamic stereoselective ring-opening of enantioselectively reduced Erlenmeyer-Plöchl azlactones (Scheme 1 (c)). In contrast with catalytic hydrogenation, this formal hydrogenation of the azlactone olen (a DHAA derivative) xes the congurations of the two hydrogenated centres in different steps, lending itself to stereodivergency. Herein, we present the rst demonstration of this strategy by its application to the onepot preparation of b-triuoromethyl AA derivatives 4 (Scheme 1 (d)). In more detail, enantioselective transfer hydrogenation of readily available substrates 1 (ref. 9a) with Hantzsch esters 12 sets the absolute conguration of the triuoromethylated b-centre. 13 Subsequent dynamic alcoholytic ring-opening 14 of intermediates 3 xes the absolute conguration of the a-carbon. In the ring-opening reaction, the substrates 3 feature a considerable bias towards the formation of anti-isomers 4. Such bias was readily leveraged with conventional Cinchona catalysts to obtain a range of anti-4 products with very high stereoselectivities (including compounds not suited for hydrogenation). Conversely, the development of the syn-selective process was less straightforward, requiring a peculiar ammonia-derived squaramide catalyst to afford the syn-4 isomers with variable diastereoselectivities.
It is worth stressing that b-branched b-triuoromethyl AAs, and derivatives thereof, have found widespread interest in medicinal chemistry (Fig. 1). b-Triuoromethylated analogues of natural AAs have been incorporated into peptides, wherein the triuoromethyl group can give unique effects on stability, acidicy/basicity, folding behaviour, hydrophobicity, and ultimately biological activity. 8a The b-triuoromethyl AA framework can also be found in less canonical structures. Besides the drug candidate mentioned in Scheme 1 (b), which showcases inhibition of b-amyloid production, 15 another b-triuoromethyl AA structure of medicinal interest is an analogue of thalidomide, 2i which feature enhanced congurational stability compared to thalidomide. An additional example is represented by the heterocyclic compound derived from a 3-(triuoromethyl)azetidine carboxylic acid, shown in Fig. 1. This compound is a member of a library of related heterocycles, investigated for their activity as inhibitors of phosphoinositide 3-kinases. 16

Results and discussion
At the outset, with transfer hydrogenation to b,b-disubstituted nitroalkenes promoted by Jacobsen-type catalysts as a lead, 17 we explored different variables in the reduction of b-phenyl-b-tri-uoromethyl Erlenmeyer-Plöchl substrates 1 with HEs.{ Preliminary screening of different catalysts and HEs identied the conditions outlined in Table 1, entry 1, as promising starting point. In more detail, using 20 mol% of catalyst 2a in combination with the isobutyl HE, in dichloromethane (reagent grade) as solvent at low temperature (À30 C), the 2-phenyl azlactone 1a could be reduced with full conversion and promising enantioselectivity (70% ee). Products 3 of the transfer hydrogenation reaction are relatively unstable. Thus, for CSP HPLC analysis they are converted to the ultimate products 4 by ring-opening with allyl alcohol using achiral tertiary amine promoters, furnishing the two diastereoisomers 4. These two isomers feature comparable ees, validating this method for the evaluation of the enantioselectivity of the reduction step. Thus, a systematic variation of the modules of this type of Jacobsen catalyst 18 (AA portion, double H-bond donor, amide portion, terminal N-group), of C2 substituents of the azlactone, and of reaction conditions, was undertaken.{ Exploration of catalysts based on double hydrogen bond donors other than thiourea (ureas and squaramides), and another AA portion (L-valine), conrmed the L-tert-leucine derived thiourea as the most efficient catalyst core. Whereas variation of solvent and/or dilution  was not fruitful, investigation of different terminal N-groups in the thiourea (2a-f) indicated that the 4-(triuoromethyl)phenyl residue (2b) could provide a distinct improvement, compared to the common 3,5-bis(triuoromethyl)phenyl (2a) and other aryl/ alkyl groups (2c-f) (entries 1-6). With catalyst 2b, three differently C2 substituted Erlenmeyer-Plöchl azlactones 1b-d were screened (entries 7-9). The 4-methoxyphenyl (PMP) derivative 1d gave better results, and was thus selected for further catalyst development, which focussed on exploring various N-substituents on the amide (entries 10-14). Amongst catalysts 2g-k, the N-benzyl-N-methyl amide derivative 2k performed best, affording the transfer hydrogenation product 3d with a respectable 85% ee, measured on the corresponding ring-opened derivative 4d.
A closer inspection at the alcoholytic ring-opening step of compounds 3 promoted by achiral tertiary amines, indicated that the ring-opened products 4 were obtained with higher anti/ syn ratios (5.5 : 1 for 4d) than the parent azlactones 3 (ca. 1.5 : 1). Thus, the alcoholytic process was dynamic, and biased towards the anti-isomer. We initially surmised that such substrate-bias, hindering access to syn-4, 19 would be circumscribed to the tertiary amine promoted alcoholytic ringopening. Since a variety of DKRs of azlactones by ringopening reactions, using different nucleophiles and catalytic approaches (Lewis bases, Lewis acids, Brønsted acids, enzymes), are available in the literature, 14 we hoped that one of these could be subdued to our aims.
However, a preliminary screening of many of these methods suggested the squaramide Cinchona catalyzed alcoholytic ring opening 20 as most promising option, despite its resemblance with the biased achiral amine promoted reaction. State-of-theart Cinchona squaramide dimeric catalysts derived from quinidine (QD-1) and quinine (QN-1) (Scheme 2) were initially employed with enantioenriched azlactone 3d under standard conditions (dichloromethane, allyl alcohol, 0 C). While, unsurprisingly, "matched" QD-1 increased the anti/syn ratio to a high 10.0 : 1 value, compared to an achiral tertiary amine (ca. 5.5 : 1), we were pleased to observe that the corresponding "mismatched" QN-1 could reverse the selectivity of the process, forcing the ring-opening reaction towards a moderate preference (1 : 2.3) for syn-4d. The products 4d displayed a higher enantiomeric excess than 3d, in accordance with the Horeau effect. 21 Adjusting the reaction conditions and testing additional QD and dihydroquinidine (dhQD) derived structures led to a highly anti-selective protocol. Catalyst dhQD-2 improved in fact the diastereomeric ratio of the product 4d up to 14.2 : 1 in favour of the anti-isomer. However, application of its quasienantiomeric derivative dhQN-2 did not result in the expected improvement in the syn-selectivity, providing a result similar to QN-1 (1 : 2.4 vs. 1 : 2.3). This result emphasized that the transition states leading to the anti and syn-products are "intrinsically" diastereomeric, 19b due to the presence of the chiral (R)-congured b-branched chain of azlactone 3d. On these grounds, different (i.e. non quasi-enantiomeric) catalyst structures may be required for anti-and syn-selective processes. Thus, a range of (dh)QN derived squaramide catalysts and reaction conditions were examined (see also ESI †). The diastereomeric catalyst dhQN-3 (from (S)-a-methylbenzylamine instead of (R)-a-methylbenzylamine of dhQN-2) was tested rst, giving however a poor result. Subsequent catalyst screening, performed at RT, suggested that the main factor affecting the stereoselectivity is the bulkiness of the squaramide portion. While catalysts QN-4, 5, 6, wherein the squaramide bears a methylene group, gave slight improvements compared to dhQN-2 (1 : 2.7-2.8 vs. 1 : 2.4), the more bulky tert-butyl substituted QN-7 provided a lower d.r. (1 : 1.9). The similar performances of the prototypical 22 3,5-bis(triuoromethyl)benzyl catalyst QN-4 and the simple benzyl derivative QN-5 point to a negligible inuence of the electronics of this group on selectivity. Thus, aiming at reducing bulkiness, catalyst QN-8 derived from methylamine was applied, providing indeed a rewarding improvement (anti/ syn ¼ 1 : 3.1). A further reduction in bulkiness could be achieved only by entirely removing the N-substituent, which was nalized preparing and testing the ammonia derived catalyst QN-9. Pleasingly, this peculiar and unprecedented structure was able to afford syn-4d with a notable 1 : 4.6 selectivity. The very  poor solubility of QN-9 in CH 2 Cl 2 resulted however in sluggish reactivity, with only 50% conversion aer 42 h at RT. Such shortcoming was overcome by switching to the more soluble dihydroquinine derivative dhQN-10, which gave 4d with >90% conversion, even by performing the reaction at 0 C, and in a 1 : 5.9 diastereomeric ratio favouring the syn-isomer. 23 Enantiomeric excess was found to be good (92%), as expected. At this stage, additional experiments indicated the unique requirements of the syn-selective reaction: catalyst dhQD-10, quasienantiomeric of dhQN-10, gave an anti-selectivity in the process comparable to an achiral tertiary amine (ca. 5 : 1).
Aiming at streamlining the overall process (1d / 4d) by implementing a one pot procedure, thus circumventing the problematic purication of azlactone intermediate 3d, it was found that an excess of HE in the transfer hydrogenation reaction has to be avoided, since this species inhibits the basic squaramide catalyst used in the ring-opening step (see ESI †). In contrast, the other components of the transfer hydrogenation (thiourea catalyst 2k and pyridine co-product) do not interfere with the second step. Fortunately, it was possible to drive the transfer hydrogenation reaction to completion even by using just 1.1 equiv. of HE. Ultimately, this modication was sufficient to develop an efficient one-pot procedure.
Then, in line with the notion that enhancing the enantiopurity of 3d would result in additional improvement of synselectivity, 19 an additional round of optimization of the catalyst used in the hydrogen transfer step was undertaken (Scheme 3).
A tentative transition state picture can be built from a computationally validated model for the transfer hydrogenation of nitroalkenes with HEs catalysed by Jacobsen-type thiourea catalysts (2), 17c complemented with recognition studies of lactones by a thiourea. 25 Coordination of the acidic thiourea hydrogens to the lactone moiety, possibly assisted by its N-aryl group, and simultaneous stabilisation of the positive charge on the HE by the amide oxygen, are the key interactions between catalyst and substrates (Fig. 2). The tert-butyl group serves to "lock" the conformation of the catalyst as shown, thus leading to a match between the catalyst polar functionalities and a transition state leading to (3R)-3d. While this model does not help rationalizing the subtle effects of the amide and the thiourea aryl groups on the enantioselectivity of the reaction, it reconciles with the observed comparably high, but opposite, sense of enantioinduction exerted by catalyst 2p on the two isomeric olens (i.e. Z-1 / (3R)-3, and E-1 / (3S)-3, see ESI †). From the experimental results shown in Scheme 3 (compare 2p with 2q and 2r), the oen encountered positive relationship between the extension of the p-system of the 2-substituent of the pyrrolidine and the enantioselectivity 24b is not apparent. Stabilising cation-p interactions might not be helpful to selectivity in this case.
The stage was thus set for the full unravelment of the stereodivergent methodology (Scheme 4). It is clear from the results reported that the improvement in enantioselectivity provided by catalyst 2p in the rst step was indeed benecial to the diastereoselectivities of the whole processes. Its combination with catalyst dhQD-2 furnished anti-4d in good yield and in essentially diastereo-and enantiopure form, while use of "mismatched" dhQN-10 in the second step afforded syn-4d in 72% yield and >99% ee, with a notable 7.5 : 1 diastereomeric ratio. These results are to be compared with the 14.2 : 1 d.r. and 89% ee for anti-4d, and the 5.9 : 1 d.r. and 92% ee for syn-4d obtained when catalyst 2k was used in the rst step (Scheme 2). Scheme 4 shows also how the different combinations of catalysts (2p and ent-2p, dhQD-2 and dhQN-2, dhQD-10 and dhQN-10) could permit the obtainment of the full set of stereoisomeric products 4d with moderate to excellent results. Moreover, although the one pot protocol required longer reaction times (2-6 days instead of 24 h), for the ring opening step, the Cinchona loading could be lowered to 5 mol% without affecting the selectivity of the process.
The scope of the one-pot procedure was then studied (Table  2), by rst applying a range of b-triuoromethyl Erlenmeyer-Plöchl azlactones 1d-j, bearing electron-donating (1g, i) or electron-withdrawing (1e, f, h, j) groups at the b-aryl ring, and bheteroaromatic substituents (1k, l). Entries 1-9 show that these substrates behaved very well in the anti-selective reaction, providing the corresponding anti-4d-l with results comparable to the parent anti-4d, that is, in good yields and outstanding diastereo-and enantioselectivities. The syn-selective processes provided variable results in terms of diastereoselectivities, ranging from a fully satisfactory 8.5 : 1 value for product syn-4g to less pleasing ca. 2 : 1 results for the b-heteroaromatic derivatives syn-4k and syn-4l. The latter results can be ascribed to a very high substrate bias towards anti-4k, l in the ring-opening process (>10 : 1 employing Et 3 N), rather than to poor catalyst dhQN-10 efficiency. Syn-4k and syn-4l were also obtained in lower yields compared to the other compounds. Nevertheless, the enantiomeric excesses of the major syn-4 isomers were found to be excellent in all cases examined ($99% ee). Substrate 1m bearing a b-peruoro residue rendered results similar to the b-heteroaromatic derivatives 1k and 1l, that is, excellent selectivity in the anti-4m isomer, and moderate yield and diastereoselectivity, but with >99% ee, for the syn-4m diastereoisomer (entry 10). The application of b-aliphatic substrates 1n-p required an adjustment to the conditions used in the transfer hydrogenation step, which was performed with higher (20 mol%) catalyst loading and at higher temperatures (À20 C for the ethyl and methyl derivatives 1o and 1p, 0 C for the more hindered cyclohexyl counterpart 1n). With these adjustments, it was possible to obtain anti-4n-p with good selectivities, while the results for syn-4n-p vary from the satisfactory level of syn-4n , 2p (0.01 mmol, 20 mol%), CH 2 Cl 2 (0.300 mL), À20 or 0 C, 24-48 h, then CH 2 Cl 2 (0.60 mL), dhQD-2 for anti-4 or dhQN-10 for syn-4 (0.01 mmol, 10 mol%), allyl alcohol (0.1 mmol, 2 equiv.), 0 C, 3-6 d. to the less pleasing syn-selectivity for 4p (entries [11][12][13]. The latter result was ascribed to the deleterious combination of moderate enantioselectivity in the transfer hydrogenation step (ca. 70% ee) with high substrate bias in the ring-opening step (ca. 10 : 1). The last three entries 14-16 of Table 2 display the results obtained by applying alcohols other than allyl in the alcoholytic process with substrate 1d. The peculiarity of the present reaction system makes the tolerance to different primary alcohols, known for the DKR of simple azlactones, 20 less than obvious, especially in the case of the syn-selective protocol. However, it was pleasing to observe that results in line with the allyl derivatives 4d were obtained for the products of methyl, benzyl and isobutyl alcohols 4q-s, although lower yields were observed in the latter case.k Compounds 4j were separately subjected to a two-step reduction-hydrolysis sequence (Scheme 5), delivering the corresponding aminoalcohol hydrochlorides 6, via amides 5. It is worth stressing that neither syn-6intermediate en route to the drug candidate (see Scheme 1(b)) 9anor anti-6 can be easily accessed by conventional asymmetric hydrogenation. 7d

Conclusions
We have proved that the conceptually new combination of two catalytic processes (transfer hydrogenationdynamic ringopening) on Erlenmeyer-Plöchl azlactones can provide a new stereodivergent strategy to enantioenriched b-branched AAs. The realization of this tactic with triuoromethylated substrates has disclosed a one-pot entry to b-aryl-b-triuoromethyl AA derivatives. Using the appropriate catalyst combination, the anti-bias of the ring-opening reaction was leveraged, giving antiproducts with excellent stereoselectivities (d.r. up to >20 : 1, ee $ 89%). The scope of this reaction includes substrates reluctant to enantioselective hydrogenation. A newly designed ammonia derived squaramide catalyst afforded the syn-isomers, not obtainable by hydrogenation, with variable diastereoselectivities (d.r. up to 8.5 : 1) and high enantioselectivities (ee $ 99%).

Author contributions
Conceptualization and supervision: VC and LB. Investigation and methodology: all authors. Writing -original dra: LB. Writing -review and editing: VC, MF, LB. Funding acquisition, project administration and resources: MF and LB.

Conflicts of interest
There are no conicts to declare. Notes and references § All through the paper, to identify the diastereomers of the b-branched b-tri-uoromethyl AA derivatives, we have used Masamune's syn and anti descriptors, arbitrarily setting the b-aryl/alkyl group of these compounds in the main chain. Using CIP descriptors for the relative conguration (e.g. R*, S*), although more rigorous, would result in a less clear identication of the diastereoisomeric pairs. { For a more comprehensive list of screening results, see ESI. † k For limitations in terms of substrate variations, see ESI. †