Protecting‐Group‐Free Amidation of Amino Acids using Lewis Acid Catalysts

Abstract Amidation of unprotected amino acids has been investigated using a variety of ‘classical“ coupling reagents, stoichiometric or catalytic group(IV) metal salts, and boron Lewis acids. The scope of the reaction was explored through the attempted synthesis of amides derived from twenty natural, and several unnatural, amino acids, as well as a wide selection of primary and secondary amines. The study also examines the synthesis of medicinally relevant compounds, and the scalability of this direct amidation approach. Finally, we provide insight into the chemoselectivity observed in these reactions.


Introduction
Chemical reactions for the formation of amide bonds are among the most commonly employed transformationsi no rganic chemistry;a mides featuring in 25 %o fi ndustrially important pharmaceuticals as well as aw ide selectiono fb ioactive natural products and polymericm aterials. [1] The synthesis of amides derived from amino acids constitutes am ajor application of amide bond formation chemistry.W hile the synthesis of amino amides by the assembly of N-protected a-amino acids and amines is aw ell-known and accepted methodology, [2] the synthesis of amides directly derivedf rom unprotected amino acids (Scheme 1B)i sr are, despites hortening the synthetic sequenceb yt wo steps. This class of amidation reactionp resents issues relatingt oc ontrolo fr eactivity,b ut the lack of research effort in this area is probably due to the preconceptiont hat such ar eaction simply would not work (Scheme 1C). Nevertheless, as mall number of methods for direct amidation of unprotected amino acids have been developed.
The first known method for the chemoselective amidation of unprotected amino acids was developed by Leuchs, and employed phosgene or derivatives thereof to generate reactive N-carboxyanhydrides (NCAs)f or subsequentr eaction with amines (Scheme 2). [3] However the activationo ccurs to such an extentt hat it leads to polymerization [4] and the aminoa cid still requiresprotection for af ully selectivemonoacylation. [5] Since then,t here have been severalr eports of methods for the synthesis of amides derived froma mino acids employing a transienta ctivating group for the carboxylic acid that simultaneouslyp rotects the amine (Scheme 2A). Burger et al. reported the use of gaseous hexafluoroacetone( HFA) for the formation of activated oxazolidinones, [6,7] which are reactivet owards amines.T he applicabilityo ft his method was demonstrated through the shortests ynthesis of aspartame to date (Scheme 2B). [6] Later,s imilar approaches were exploited by Liskamp et al. using dichloroalkyl silanes [8] or boron trifluoride [9] to produce activated Lewisa cid conjugates in which the aminoa cid was proposed to coordinate in ab identate fashion. Remarkably,B ristol-Meyers-Squibbh ave employed the silane-based methodology for the synthesis of an amino amide drug candidate, [10] demonstrating benefits with regard to both costeffectiveness and atom economy( Scheme2B). [11] More recently,S harma et al. [12] reported the amidation of someu nprotected amino acids employing a" classical" coupling reagent, 1'carbonyldiimidazole (CDI), in water.N evertheless,t he method displayed serious drawbacks with regard to scope, only succeeding in the amidation of four amino acids (Scheme 2C).
We have recently reported that the borate ester B(OCH 2 CF 3 ) 3 is an effective reagent for the direct synthesis of a-amino amides from unprotected amino acids and amines. [13,14] In this study,w eo utline the full scopeo fthed irect amidation of unprotected amino acidse mploying both stoichiometric and catalytic quantities of av ariety of boron Lewis acids as well as group(IV)m etal salts and "classical" coupling reagents.

Classical coupling reagents
The acylation of amines with activated carboxylic acids is the most common way to make an amide, as ac onsequence of the widespread availability and high stability of both of these buildingb locks. In fact, based on literature surveys, the most commonly used methodsf or amidation involvet he formation of intermediary acid chlorides, O-acyl ureas or (mixed) anhydrides. [15] We therefore began our investigation by looking at the efficiency of classical coupling reagents for the chemoselective amidation of unprotected phenylalanine with benzylamine. Using polar solvents with the potentialt op artially solubilize unprotecteda mino acids (H 2 O, EtOH), biphasic systems (CH 2 Cl 2 /H 2 O) as well as non-protic solvents (DMF,C H 2 Cl 2 )w e looked at the competency of 'classical" coupling reagents in amidation. In most cases, little or no amino amide was pro-duced (Table 1, entries1-6). In the case of amidation employing CDI in water (entry 7), the formationo ft he desired amino amide was observed in small quantities, accompanied by the formation of three other amide species, highlighting issues relating to control of reactivity.O verall, we believe that the lack of reactivity,a ss uggested previously, [12] is mainly due to the poor solubility of zwitterionic a-amino acids both in organic solvents and in aqueous solution.
Following this study,w ew ent on to look at the reaction conditions in more detail in order to minimize the levelsofr acemization seen for some examples. The reactiono ft ryptophan with propylamine was examined with various conditions, and it was found that racemization could be minimizedb ys hortening the reactiont ime and/orb ya dding the borate reagent dropwise (Table 3).
We have also attempted the reaction to synthesise 1a under similar conditions with av ariety of group (IV) metal salts (Table 4), as these types of species have been reported to be active catalysts for amide bond formation in recent years. [16,17] While reactions employing zirconium-based reagents solidified and produced only minor quantities of amide 1a,T i(OiPr) 4 was identified as as uitable alternative amidation reagent. The reactions were also attempted in the presence of molecular sieves but the resulting yields were slightly lower.As electiono f amino amides were synthesised using 1equiv of Ti(OiPr) 4 ( Figure 1). Amide 2n was synthesised in high yield but with significant racemisation.A lthough lower yields weres een for the synthesis of 2r and 2t,t he use of Ti(OiPr) 4 furnished products with highere nantiopurity (! 95:5 e.r.) than the reactions mediated by B(OCH 2 CF 3 ) 3 .H ence, Ti(OiPr) 4 represents an easily accessible andl ower cost alternative to our borate ester,e specially for more reactive aminoacids. Table 2. B(OCH 2 CF 3 ) 3 -mediated amidation of free amino acids with propylamine.

Catalytic amidation employing Lewis acids
Having recently developed am ethod for general amidation employing catalytic B(OCH 2 CF 3 ) 3 , [14] we wishedt oe xplore the applicability of this approach in the amidation of unprotected amino acids. In the optimisation of the general catalytic amidation reaction, we screened aw ide selection of organic solvents under Dean-Stark conditions and identified tert-amyl methyl ether (TAME) as as uitable alternative to CPME and PhMe, which crucially allows for azeotropic removal of water to be conducted at lower temperatures (86 8C). [14] Design of Experiments (DoE) reactiono ptimisation [18] was then conducted to understand the factorsp laying ar ole in the amidation of an unprotected amino acid,a nd to maximise the yield of product 3a and minimise the formation of diamide byproduct 4a,t he product of amidation of the amino acid with the desired amino amide product. The catalyst loading, amine loading and volumeo fs olvent used were investigateda sf actors (Scheme3).
Good quality modelsf or predictingt he yield of both products were obtained from the DoE study.T he results indicated that only the amine loading and catalystl oading had as ignificant effect on the yield of the desired amino amide. Unsurprisingly,w ef ound that excess amine was beneficial for minimising byproduct formation.I ncreasing the catalyst loading led to increased formation of both products. As ar esult of this study, we were able to identify conditions to obtain high yields of the desired amino amide by running the reactions with only 1.5 equivalents of amine and 20 mol %B (OCH 2 CF 3 ) 3 catalyst. Lowering the amountso fa mine further led to increased formation of diamino amide 4a,a lthough less than statistically expected (usually ranging between 1-8 %, separabled uring purification).
With effective conditions in hand for the use of catalytic B(OCH 2 CF 3 ) 3 for the direct amidation of an unprotecteda mino acid, we then went on to explore the use of alternative Lewis acid catalysts under these conditions. The use of 2-Cl and 3,4,5-trifluorophenylboronic acids ( Table 5, entries 2-3, 9) has been previously reported to be effective for amidation under dehydrative conditions. [19,20] Such catalysts are generally unsuccessfule ven for amidation of protecteda mino acids, and, to the best of our knowledge, they have never been explored with unprotected aminoa cids. [19][20][21] However, to our surprise, these boronic acidc atalysts were reasonably effective fort he amidation of Phe with benzylamine under Dean-Stark conditions. Group (IV) metals (entries [4][5] were, as with the stoichio-metric conditions, suitable catalysts for this transformation, with Ti(OiPr) 4 ap articularly cost-effectivea lternative to our boratee ster providing amide 1a in excellent yield. When employing ZrCl 4 ,i solation of the product was problematic (product lost in workup), most likely due to the formation of complexesb etween amino amides andZ ra nd the formation of hydrochloride salts of the amines present in the reactionm ixture.
The high levelso fc onversion to 1a in these reactions prevents accurate differentiation of the reactivity of the various catalysts, so we went on to screen selected catalysts fort he direct amidation of valine, am uch less reactive substrate. Again, the amidation could be achieved chemoselectively with both group(IV) catalysts and boron-based catalysts. Although amidation employing ZrCl 4 gave the highest level of conversion, attempts at isolation failed due to issues relatingt o adduct/salt formation (vide supra, for use of stoichiometric ZrCl 4 ). B(OCH 2 CF 3 ) 3 was therefore identified as the catalyst of choice, but Ti(OiPr) 4 is as uitable alternative for more reactive amino acids.

Substrate scope of borate-catalysed amidation
With B(OCH 2 CF 3 ) 3 as the catalyst, we explored the scope of the amino acidc omponent with benzylamine as the amine (Scheme 4). All 20 common proteinogenic, as wella sn ine nonproteinogenic aminoa cids were tested. In general, the less polar amino acids wereg ood substrates for the reaction, with two aromatic (1a, 1d)a nd most aliphatic (1f-i)a mino acids yieldingt he corresponding amide in good to excellent yields (Scheme 4A,B). The amide derived from alanine and benzylamine could not be readily separated from unreacted benzylamine during purification, so ar eactionw as, therefore, per- formed with phenethylamine to give amide 1f in excellent yield. The reactionw ith glycine led to the formation of am ixture of the desired amide 1e and its diamido counterpart 5e under the standardc onditions, so al arger excesso fa mine had to be employed (3 equiv) and only am oderate yield of 1e was obtained.W ew ere also able to use catalytic quantities of Ti(OiPr) 4 for the preparation of amino amides derived from more reactive amino acids (e.g. 1a, 1g). This catalystw as much less effective for more hindered amino acids, however (e.g. 1b). Pleasingly,b yu sing B(OCH 2 CF 3 ) 3 we were also able to access amides of more polar amino acids with hydroxyl (1l-m) and sulfur (1j,1k)m oietiesp resent,a nd good conversions were observed( Scheme 4C). Amide 1k (from cysteine) was partially oxidized to the corresponding disulfide over the course of the reactionand upon exposure to air. Acidic amino acids underwent amidation effectively (Scheme4D), with glutamic acid cyclizing intramolecularly to give pyroglutamide 1n and aspartic acid undergoing double amidation to give ad iamide (1o). The increased degree of racemization for 1o is probablyd ue to ac ompeting dehydrative mechanism involvinga5 -membered acid anhydride,k nown to have ap ropensity for racemization. [24] The basic (1q-s)a nd amidic (1t-u)a mino acids were somewhat less reactive (Sche-me 4E,F), with only histidine giving an amide derived from benzylamine (1r). As expected, lysine spontaneously cyclized to form a-aminocaprolactam 1s. b-Amino acids generally workedlesswell than their a-amino counterparts (Scheme4G). As with glycine, b-alanine underwent extensive over-reaction (1v), and even with three equivalents of amine did not give a clean amino amide product. However, b-amino acids could be coupled successfully to give amides 1w-x,a lbeit in lower yields relative to their a-amino acid counterparts.
Amino acids with protected side chains were also tested (Scheme4H). Ester-protected glutamica cida nd the methylether derivative of serine gave amides 1y and 1z,r espectively, in good yields, despite showing signs of significant racemization. Protected cysteine formed desired amide 1aa along with minor amounts of arearranged product (see ESI); S-methyl cysteine yieldeda mide 1bb in excellent yield, but with significant racemization. Sarcosine, l-a-aminobutyric acid and phenylglycine all underwent successful amidation (Scheme 4I), although the latter (1ee)u nderwent completeracemization, as expected based on its known propensity to racemise even under mild conditions (e.g. Cbz-PhGly + NH 4 Cl yields amide in 43 %y ield, 46 % ee using ethyl chloroformate, 58 C, 1h). [25] Overall,t he methods howed similar trends in reactivity to the stoichiomet-  Table 2). The relevance of this chemistry can easily be exemplified by the fact that most of the amides synthesized in Scheme 4( or their enantiomers) are members of aw ell-documented class of potent anticonvulsants and agentsf or neuropathic pain treatment, commonly referred to as primary aminoacid derivatives(PAADs). [22,23] The scope of the reaction with regard to the amine component was investigated next through the preparation of amides derived from as election of different amines employing both stoichiometric B(OCH 2 CF 3 ) 3

(denoted as [A]) and the catalytic reactionconditions (denoted as [B],S cheme 5).
Simple aliphatic amines workedw ell under both sets of conditions (6a-g,S cheme5A). However,t he reactiono fi sobutylamine with tryptophan underc atalytic conditions resulted in a low yield of amide most likely due to the volatility of the amine, which is probablyr emoved into the side arm of the Dean-Stark apparatus. Benzylic and allylic amines were also successful (Scheme 5B), with substituted benzylamines bearing am ethoxy group (6i, 6l), fluoride (6j)o rp olyfluoroalkane (6n) giving good yields of the corresponding a-aminoa mides under both sets of conditions. Heterocyclic tryptamine (6m) and 2-picolylamine (6o)a lso underwent amidation successfully, with significant amountso fr acemization in the latter case when stoichiometric B(OCH 2 CF 3 ) 3 wasu sed. Pleasingly,t he degree of racemization wassignificantly reduced under catalytic conditions.
Amino amides could also be prepared from reactive secondary amines in good yield (6q-s, 6u-w,S cheme 5C), buto nly with the stoichiometricm ethod. The reactions with leucine and phenylalanine with pyrrolidine required forcingr eaction conditions which led to significant levels of racemization (6s, 6u). When employing catalytic B(OCH 2 CF 3 ) 3 fort he synthesis of at ertiarya mide, even with al arge excesso fa mine (3 equiv) the reactionw as not selectivea nd led to further reaction of 6t to give diamide 7t.U nder the catalytic set of conditions, glutamic acid was the only amino acid to show selectivityf or mono-amidation with as econdary amine, due to intramolecular cyclisation. Indeed, it was possible to synthesise 6w and the pharmaceutical Fasoracetam 6z (Scheme 5E)i no ne step from glutamic acid.
It was also possible to prepared ipeptide derivatives using glycine and alanine tert-butyl esters as the nucleophiles (Scheme 5D,E). Reaction of both Phe (6y)a nd Pro (6aa)w as attempted with OtBu Gly,a lthougho nly the latter gave ag ood yield and enantiopurity,w ith phenylalanine undergoing significant racemization (6y). Conveniently, 6aa is ap recursor to a marketed nootropic, Noopept, which can be accessed through af urther condensation employing our catalytic amidation conditions. [14] Glutamica cid also underwent successful cyclisation/ amidation with Ala-OtBu to give 6x in good yield. We were also ablet os ynthesize ab enzodiazepine derivative 8,w hich belongs to ac lass of anti-anxiolytic drugs,f rom 2-aminobenzophenonea nd l-Phei n5 2% yield, albeit under forcing conditions, which again led to an earf ull racemisation of the final product.
As the synthesis of tertiarya mides was problematic in most cases, we set out to explore an alternative protocol using aminoboranes (Scheme 6), which have been previously demonstrated to promotea midation of carboxylic acids [26,27] and esters. [28] Commercially available tris(amino) boranes were found to be effective ford irect amidation of unprotected amino acids, affording amides in moderate yield and enantiopurity.D ue to the volatility of amines, such as dimethylamine, these amides would be difficult to prepare using the boratemediated reactions outlined above. Using equimolar amounts of aminoborane under conditions analogoust ot he stoichio-metric borate reaction, amino amides 6u, 6s and 6bb-gg were synthesized in low to good yields andw ith little sign of racemization. In acetonitrile, the remarkable reactivity of these aminoboranes enabled the synthesis of amides 6ff-gg and 6ee under very mild conditions at room temperature. This methodi s, however,l imited by the requirement to employ an excess of amine in the synthesis of the tris(amino)b orane, something whichi so nly likely to be economically viable with low-cost readilya vailableamines.

Sequential amidation reactions
With amethod to provide direct access to amino amide derivatives, we reasonedt hat we could use the free amine group in furthert ransformations to access useful compounds. We envisaged that the direct synthesis of free amino amides could be combined with our previously reported amidation processes [13] to provide access to relativelyc omplex a-amido amides in a simple operation.
We started by exploring sequential amidation reactions with our stoichiometric reaction conditions. To this end, direct amidation of af ree amino acid with propylamine wasf ollowed by af iltrationw orkup to remove unreacteda mino acid and boron residues and give the crude a-amino amide. This was then subjected to direct borate-mediated amidation with ac arboxylic acid to give an a-amido amide, which was purified by a second filtration workup( Ta ble 6). The diamides 10 a-e were obtained efficiently over the two-step sequence in all cases, with no requirement for chromatographic purification.
We then went on to explore an analogous transformation employing the catalytic conditions in ao ne-pot procedure ( Table 7). Following the standard protocol for direct amidation of af ree amino acid, as olution of carboxylic acidi nT AME was Scheme6.Tris(amino)boranes in amidation. Enantiomeric/diastereomeric ratios > 95:5, unless otherwise stated. Table 6. Sequential double amidation reactions.

Amide
Step 1 yield [%] Step added to enableasecond amidationr eaction to take place. In general these double amidation reactions gave high conversions, although in the case of chiral substrates, significant epimerization took place most likely due to the extendedr eaction times. Purification for diamides 11 employing columnc hromatography was particularly difficult, due to the formation of amide side products from the reaction of the reactant amine with the second carboxylic acid coupling partner.N onetheless, the one-pot synthesis of Lacosamide was successful with this method, yieldingdesired amide 11 d,albeit with alow enantiopurity.I ti sw orth noting, however,t hat this substrate is particularly pronet or acemization, [29,30] and that the first step of the amidation reaction (1z,S cheme 4) also showed erosion of enantiopurity (vide supra).

Sequential condensation reactions
Given that B(OCH 2 CF 3 ) 3 has previously been shown to promote imine formation when used stoichiometrically, [31] we also explored aone-pot unprotectedamino acid amidation/condensation reactiont op rovide access to imidazolidinones in as ingle step (Scheme 7). In this case the cyclisation reactionw as much quicker (1-2 h), and yieldedp roducts with very little or no signs of racemization.I midazolidinones derived from as election of amino acids (Phe, Ala and Sar) were cyclized with aliphatic (12 c,d, 12 g-i, 12 k), benzylic (12 a), heterocyclic (12 b, 12 j)a nd hydroxyl containing (12 a)a ldehydes or ketones in good to excellenty ields. Only the reactiono fc amphor (12 e) failed to yield the desired heterocycle. We were also able to synthesise precursors to two naturalp roducts( AE)-Tricladin A and Bf rom alanine in one step. [32] Scalability and green metrics As our goal was to develop ah ighly efficient and scalablea midation protocol, we soughtt ot est the synthesis of as et of substrates on al arger scale. Both the chemoselective amidation protocol (25-50 mmol) and the sequential amidation/condensation procedures (250 mmol) were amenablet os cale up to access multigram quantitieso fm aterial, although in slightly lower yields than the smaller scale reactions in the case of 1b and 1d (Figure 2). This is likely due to the heterogeneity of the system,w hich makesa dequate mixingm ore difficult on a larger scale.
Next we set out to demonstrate the efficiency and cost-effectiveness of our method by benchmarking it against "classical" amide coupling approaches (Scheme 8). We examined the synthesis of amide 1a by using our method,a nd using literature approaches startingf rom either the free amino acid or the tert-butyloxycarbonyl (Boc)-protected derivative, which is also commercially available and commonly used as as tartingm ate- rial. However,d espite Boc-Phe-OH being availablea tw hat is often consideredanominal cost, it is > 20 times more expensive than the free amino acid (198 vs. 10 E mol À1 )! [33] It is clear on the basis of the total material input required in terms of reagents and solvents, that the directc hemoselecitve amidation route offers significant benefits. The literature route from the free amino acid requires at otal materiali nput of 308 Kg per Kg of amide product, and the approach from the Boc-protected amino acid requires an even larger material input (412 Kg per Kg of amide product). In comparison, our chemoselective amidationm ethod requires am ateriali nput of only 5Kgp er Kg of amide product obtained, % 60-80 times more efficient. As these methods for the syntheses of amino amides employ column chromatographic purification, we are unable to calculate the process mass intensities (PMIs). [34] However,t he synthesis of amide 1a was carriedo ut on a2 5mmol scale and crystallization was used to obtain the product with high purity,t hough in al ower yield (57 %) than from chromatographic purification. Nevertheless, this process proceeds with an impressive PMI of only 13, which compares very favourably to established amidationm ethods (Typical PMI values in the range of 150-300). [34] Originsofc hemoselectivity The interactions between amino acids and boron Lewis acids were investigated by NMRs tudies. The NMR spectra suggest that the Lewis acid coordinates to the amino acid to form a cyclic intermediate such as 13 or 14 (Scheme 9). As hift in the 11 BNMR spectrum from the trigonal to the tetrahedral region suggestst he formationo fastructure such as 13 (Figure 3). The interaction between the borate speciesa nd the amino acid effectively solubilizes the amino acid in organic solvent, allowingi tt oreact in an amidationr eaction.
Interestingly,a mino acids do not react with an externalc arboxylic acid such as phenylacetic acid under our standard catalytic amidation conditions (Scheme 10). This suggests that complexes such as 13 are not reactive at the amine. Similarly, in the absence of an amine reaction partners elf condensation Scheme8. Efficiency and cost-effectiveness of chemoselectivea midation in comparison to regular amide bond-formationprocesses.
Scheme9.Interaction of boron Lewis acids with amino acids. Scheme10. Reactivity of amino acids with carboxylic acids (top) and themselves( bottom). of the amino acid to give diketopiperazine did not take place to any significant extent (Scheme 10). Of all the aminoa cids screened,o nly prolinef ormed trace amountso fd iketopiperazine when subjected to the standardc atalytic amidationc onditions (< 5% yield over 24 h).
From theseo bservations, it can be deduced that, only small quantities of amino acidw ill be solubilized when employing catalytic amounts of Lewisa cid, creatingasystem in which the amine is in large excess to the solubilised amino acid, whichi s not reactive as an ucleophile (Scheme 11). This explainsh ow self-condensation of the amino acid is prevented, and also explains the lack of reactivity of the amino acid with ac arboxylic acid. The solubilized amino acid complex 13 is then able to undergo boron-catalysed amidation with the amine to generate the amino amide product coordinated to the catalyst, which then undergoes exchange with another molecule of free amino acid to continue the cycle. On the basis of our recent studies, [35] the amidation of the amino acid complex is likely to be mediated by interaction with as econd catalystm olecule to form as pecies with ab ridging acylboron unit.

Conclusions
In summary,w eh ave identified effectivem ethods for the direct amidation of unprotecteda mino acids with amines by using catalytic or stoichiometricq uantities of boron or titanium Lewis acids. In this study,adetailed exploration of the scope of these reactions has been carriedo ut, enabling the advantagesa nd limitations of each approacht ob ei dentified. In Scheme 12, we provide af lowchart to enable the best method for ap articulara midation reaction to be identified. Weh ope that this guide will prove useful in promoting the direct amidation of amino acids as au seful transformation for the chemistry community.W ith burgeoning interesti nt he development of novel catalytic methodsfor amide bond formation, [36] we anticipate that other amidation catalysts may well be applicable to this reaction.W ea lso anticipate that implementationo ft his synthetic strategy in the pharmaceutical sector can lead to improvedc ost-effectiveness and reducedl evels of waste in the synthesis of complex medicinally relevant compounds.

Experimental Section
Experimental procedures, 1 Ha nd 13 CNMR spectra, and characterization data for all compounds are available in the Supporting Information.