A biomimetic domino reaction for the concise synthesis of capreomycidine and epicapreomycidine

The non-proteinogenic amino acids capreomycidine and epicapreomycidine are constituents of antibiotically active natural products, but the synthesis of these unusual cyclic guanidine derivatives is challenging. The biosynthesis of capreomycidine has therefore been employed as a guideline to develop a concise biomimetic synthesis of both epimeric amino acids. The resulting domino-guanidinylation-aza-Michael-addition reaction provides the most convenient access to these amino acids in racemic form. Attempts to dissect the domino reaction into two separate transformations for a stereocontrolled version of this synthetic approach have also been made. The synthesized didehydro-arginine derivatives with urethane-protected guanidine moieties did not undergo the aza-Michael-addition anymore. These results may have wider implications for the 1,4-addition of guanidines to α,β-unsaturated carbonyl compounds, particularly to didehydro amino acids.


Introduction
Bacterial infections continue to be a major threat to human health, particularly with respect to emerging strains with resistances to established antibiotics (Taubes 2008;Cooper and Shlaes 2011). It is therefore highly desirable to develop novel antibacterial agents which should ideally display new or yet unexploited modes of action (Walsh 2003). Natural products have been used as lead structures for many antibacterial drugs (Mahady et al. 2008). However, their often complex structures represent important challenges to organic synthesis as a prerequisite for detailed structureactivity relationship (SAR) studies. Among the many bacterial secondary metabolites with antibiotic activity, a significant number of compounds are of peptidic nature and contain non-proteinogenic amino acid structures.
Due to their relevance in natural product chemistry, several approaches for the synthesis of amino acids 1 and 2 have been developed. A first non-stereoselective synthesis of 1 and 2 involved the preparation of an aromatic 2-aminopyrimidine precursor, which was then hydrogenated to provide a diastereomeric mixture of 1 and 2, each in racemic form. The separation of the diastereomers was carried out by crystallization of the respective picrates (Bycroft et al. 1971b). A slightly improved version of this synthesis has also been described (Yamashita et al. 2004). The first syntheses of stereochemically pure 1 and 2 employed an aldol reaction for the synthesis of a protected b-hydroxy-ornithine precursor, which then underwent acylase-mediated enzymatic resolution. Further conversion via an aziridine intermediate finally provided either 1 or 2, but the synthesis was lengthy, and the overall yields were rather low Wakamiya et al. 1978;Teshima et al. 1980). The first stereoselective synthesis of capreomycidine 1 involved a Mannich-type reaction with a chiral glycine derivative as the key step. The subsequent guanidinylation reaction only proceeded with one of the two diastereomers obtained from the Mannich transformation, thus finally giving pure 1 after further conversions Williams 2001, 2003). However, this also implies that an according synthesis of 2 using the same strategy is not feasible. An alternative ex-chiral pool synthesis of 13 C-labelled 1 starting from Garner's aldehyde has also been established (Jackson et al. 2002), but investigations in our own laboratory revealed that an application of this approach to the synthesis of 2 is not possible (unpublished data). The first practical synthesis of stereochemically pure epicapreomycidine has been established in recent years. The initial version of this synthetic route involved an ornithine-derived sulfamate which was then employed in a rhodium-catalyzed C-H insertion reaction to construct the stereocenter in the C-3 position. However, this elegant approach suffered from low yields and diastereoselectivities of the C-H activation key step (Tanino et al. 2008(Tanino et al. , 2010. A superior version of this C-H activation strategy starting from D-tyrosine and displaying perfect stereocontrol has recently been described though (Tanino et al. 2011).
One of the highly useful strategies for the development of efficient synthetic routes is to imitate the biosynthetic pathways found in nature. Such biomimetic syntheses have found widespread applications (Razzak and De Brabander 2011). Our goal was to investigate the feasibility of a biomimetic synthesis of both capreomycidines 1 and 2. The biosynthetic assembly of the cyclic guanidine structure of capreomycidine 1 as part of viomycin biosynthesis in Streptomyces vinaceus has been elucidated (Fig. 2, Yin and Zabriskie 2004;Ju et al. 2004). Stereoselective 3-hydroxylation of L-arginine 5, catalyzed by the non-heme 2-oxoglutarate (2-OG) dependent Fe(II)-oxygenase VioC, provides (3S)-3-hydroxy-L-arginine 6. The subsequent ring closure reaction is mediated by the pyridoxalphosphate-(PLP)-dependent enzyme VioD and most likely proceeds via the a,b-unsaturated didehydro-arginine intermediate 7, as demonstrated by labelling studies. Michael-type conjugate 1,4-addition of the guanidine moiety then stereoselectively provides capreomycidine 1  Fig. 1 Structures of non-proteinogenic amino acids capreomycidine 1 and epicapreomycidine 2 as well as structures of natural products viomycin 3 and muraymycin A1 4 containing 1 and 2, respectively with overall formal inversion of the stereocenter at C-3. The biosynthesis of 2 has not been elucidated yet, but the recent analysis of the biosynthetic gene cluster of muraymycin nucleoside antibiotics suggests that it probably occurs in a similar manner, just with a different stereochemical course (Cheng et al. 2011; also see Lemke et al. 2010). Hence, the goal of this study was to synthesize didehydro-arginine derivatives 8a-h as potential precursors for biomimetic Aza-Michael-additions in order to obtain amino acids 1 and 2 or protected derivatives thereof (Table 1). There is very limited precendent for 1,4-additions of protected guanidines, with a striking example provided by Baran, Seiple and coworkers as part of their synthetic studies on palau'amine and related compounds (Seiple et al. 2011). It was thus envisaged to investigate different protecting group patterns both at the guanidine group (to influence its nucleophilicity) and the didehydro amino acid moiety (to tune its electrophilicity).

Materials and methods
Chemicals were purchased from Sigma-Aldrich, Alfa Aesar, ABCR and VWR. Amino acid phosphonates 9a-c and 3-azido-propionaldehyde 10 were synthesized as previously reported (Ducho et al. 2009;Schmidt et al. 1984;Davies et al. 1967). Reactions involving oxygen and/or moisture-sensitive reagents were carried out under an atmosphere of argon using anhydrous solvents. Anhydrous solvents were obtained in the following manner: THF was dried over sodium/benzophenone and distilled, MeCN and CH 2 Cl 2 were dried over P 2 O 5 and distilled and DMF was dried over molecular sieves 4 Å . All other solvents were of technical quality and distilled prior to their use, and distilled water was used throughout. Column chromatography was carried out on silica gel 60 (0.040-0.063 mm, 230-400 mesh ASTM, VWR) except where indicated under flash conditions. TLC was performed on aluminium plates precoated with silica gel 60 F 254 (VWR). Visualisation of the spots was carried out using UV light (254 nm) where appropriate and/or KMnO 4 staining under heating (staining solution: 1 g KMnO 4 , 6 g K 2 CO 3 and 1.5 mL 5 % NaOH (aq)  General procedure A (synthesis of protected didehydro amino acids by Wittig-Horner reactions) To a solution of KOt-Bu in THF, a solution of the amino acid phosphonate in THF was added at -78°C. After 15 min, a solution of 3-azido-propionaldehyde 10 in THF was added dropwise at -78°C. The reaction mixture was stirred overnight and slowly warmed to rt during this period. The reaction was quenched by addition of MeOH. After the addition of EtOAc, the organic layer was washed with water, dried over Na 2 SO 4 and evaporated under reduced pressure. The resultant crude product was purified by column chromatography.
General procedure B (sequence of Staudinger reduction and guanidinylation reactions with reagent 13) To a solution of the protected d-azido-didehydro a-amino acid in THF and water, PPh 3 was added at rt. The reaction mixture was stirred overnight at rt. After the addition of EtOAc, the organic layer was washed with aqueous AcOH (10 %, 3 9). The combined aqueous layers were adjusted to pH 12 by the addition of diluted NaOH solution and extracted with CH 2 Cl 2 (3 9) and CH 2 Cl 2 /i-PrOH (3:1, 1 9). The combined organics were dried over Na 2 SO 4 and evaporated under reduced pressure. The resultant crude products were sufficiently pure ([95 % as judged by 1 H NMR) without further purification. With respect to the limited stability of the didehydro-ornithines, they were prepared freshly and directly used for the subsequent guanidinylation reaction. Thus, the crude didehydro-ornithines were dissolved in DMF, and NEt 3 and guanidinylation reagent 13 were added at rt. The reaction mixture was stirred at rt or elevated temperature (to drive the reaction to completion) for the indicated time. The solvent was evaporated under reduced pressure and the resultant crude product was purified by column chromatography.
General procedure C (guanidinylation reactions with reagent 31) To a solution of the crude ornithine derivative in DMF, guanidinylation reagent 31, NEt 3 and AgOTf were added. The reaction mixture was stirred at rt for 3 h, then filtered through Celite and the Celite were washed with EtOAc (3 9). The filtrate was washed with brine (3 9), dried over Na 2 SO 4 and evaporated under reduced pressure. The resultant crude product was purified by column chromatography.
The data for (Z)-11a were collected from shock-cooled crystals at 100(2) K (Kottke and Stalke 1993;Kottke et al. 1996;Stalke 1998). The data of (Z)-11a were collected on a INCOATEC Mo Microsource (Schulz et al. 2009) with mirror optics and APEX II detector with a D8 goniometer. The diffractometer was equipped with a low-temperature device and used MoK a radiation, k = 71.073 pm. The crystal of (Z)-11a used for data collection was a nonmerohedral twin with two twin domains. The two domains were separated using RLATT. Two orientation matrices were used for the integration with SAINT (Bruker AXS Inst. Inc. 2011) and a semi-empirical absorption correction with TWINABS (Sheldrick 2012) was applied. The structure was solved by direct methods (SHELXS-97) (Sheldrick 1990) using untwinned data (HKLF 4) and refined by full-matrix least-squares methods against F 2 (SHELXL-97) (Sheldrick 2008;Müller et al. 2006) using the twinned data (HKLF 5) within the SHELXLE GUI (Hübschle et al. 2011). The fractional distribution of the second domain refines to 0.4669 (9). All non-hydrogen-atoms were refined with anisotropic displacement parameters. The hydrogen atoms were refined isotropically on calculated positions using a riding model with their U iso values constrained to equal to 1.5 times the U eq of their pivot atoms for terminal sp 3 carbon atoms and 1.2 times for all other carbon atoms. Only the hydrogen atoms of the nitrogen atoms (H1 and H8) were found from difference density map and restrained to same distance within the esd (0.02 Å ). Disordered moieties were refined using bond lengths and angles restraints and anisotropic displacement parameter restraints.
Crystallographic data (excluding structure factors) for the structures reported in this paper have been deposited with the Cambridge Crystallographic Data Centre. The CCDC number, crystal data and further details for the X-ray measurement are listed in the Supplementary Material. Copies of the data can be obtained free of charge from The Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/data_request/cif or from the corresponding author.

Results and discussion
For the synthesis of potential cyclization precursors 8a-h (Table 1), a general synthetic approach was developed. In order to achieve diversity with respect to the protecting group pattern of the guanidine moiety, it was envisaged to introduce this functionality at a late stage. Guanidinylation of primary or secondary amines is a common way to prepare guanidines (for the terminology of guanidine formation, see Jones 2002). However, since damino-didehydro a-amino acid esters (i.e., didehydroornithine derivatives) with an unprotected d-amino group display limited stability, the corresponding d-azido analogues were synthesized first. Reduction of the azide moiety would then provide the amines suitable for guanidinylation reactions. This approach has previously been used in the synthesis of a protected didehydro-arginine derivative (Yonezawa et al. 2000). However, to the best of our knowledge, the literature precedent for the synthesis and further transformation of a d-azido-didehydro a-amino acid ester is limited to this one example.
Thus, rapid access to protected d-azido-didehydro a-amino acids was an essential requirement for the described synthetic strategy. All d-azido-didehydro a-amino acids used in this study were prepared by Wittig-Horner reactions of amino acid phosphonates 9a-c (Ducho et al. 2009;Schmidt et al. 1984) with 3-azido-propionaldehyde 10 ( Davies et al. 1967) (Fig. 3). This approach proved to be significantly more efficient than the previously used lengthy transformation of protected didehydro-glutamate into the respective d-azido compound (Yonezawa et al. 2000). Wittig-Horner reactions of amino acid phosphonates with aldehydes are known to furnish the according (Z)-didehydro amino acids with excellent diastereoselectivities (Schmidt et al. 1992; for further examples, see Ducho et al. 2009;Spork and Ducho 2010;Spork et al. 2011). Accordingly, the (Z)-configured products (Z)-11ac were obtained in isolated yields of 52-77 %. The synthesis of cyclization precursors 8b and 8g,h required the presence of a second Boc group at the a-amino moiety. Thus, (Z)-11a was treated with di-tert-butyl dicarbonate ((Boc) 2 O) to afford (Z)-11d in 92 % yield. For the stereochemical assignment of (Z)-11a, the according (E)-isomer was needed as a reference compound. Following careful column chromatography, the (E)-configured congener (E)-11a was isolated as a byproduct from the Wittig-Horner transformation of 9a with 10 in 4 % yield. Application of the established 1 H NMR criteria for the distinction of (Z)and (E)-didehydro amino acids (Mazurkiewicz et al. 2005) on both isomers of 11a clearly revealed the major product to display (Z)-configuration (for details see Supplementary Material). For a completely unambiguous assignment of the double-bond geometry, single crystals of (Z)-11a suitable for X-ray crystallography could be obtained by slow evaporation of a solution of the compound in diethyl ether. The elucidated structure confirmed the (Z)-configuration of the double bond (Fig. 4). The crystals of (Z)-11a grew as non-merohedral twins in the space group P2 1 /c with two hydrogen-bonded (N-HÁÁÁO d = 206(2) pm) molecules O 10 9a: R 1 = Boc, R 2 = t-Bu, R 3 = Et 9b: R 1 = Ac, R 2 = t-Bu, R 3 = Me 9c: R 1 = Ac, R 2 = R 3 = Me The azide groups of both molecules are disordered on two positions with an occupancy of 50 %. The first attempt to achieve the desired biomimetic cyclization was to synthesize precursor 8a without protecting groups at the guanidine moiety (Fig. 5). Azido derivative (Z)-11a was therefore reduced under Staudinger conditions to give amino intermediate 12, which was not purified with respect to anticipated stability issues (see Yonezawa et al. 2000). Guanidinylation of 12 was carried out using commercially available reagent 13 in the presence of triethylamine to liberate the reactive agent from the employed hydrochloride. However, instead of the expected product (Z)-8a, a mixture of protected capreomycidine (rac)-14 and epicapreomycidine (rac)-15 (diastereomeric ratio (d.r.) ca. 1 : 1) was isolated in 54 % yield over two steps from (Z)-11a. Thus, after formation of intermediate (Z)-8a, an immediate aza-Michael-addition occurred in a domino fashion (for the concept of domino reactions, see Tietze and Beifuss 1993;Tietze 1996;Tietze et al. 2006). Similar results were obtained with precursor (Z)-11d, which was reduced to ornithine derivative 16. Intermediate 16 was then used in the domino reaction to give a mixture of protected capreomycidine (rac)-17 and epicapreomycidine (rac)-18 (d.r.*1.8 : 1.0) in 49 % yield over two steps from (Z)-11d. Hence, the higher electrophilicity of the precursor 16 bearing two Boc groups did not result in higher isolated yields of the domino product (Fig. 5). Interestingly, when isomer (E)-11a was used in the sequence of Staudinger reduction and subsequent domino-guanidinylation-aza-Michael-addition, no product could be isolated. It was therefore concluded that either the (E)-configured ornithine derivative was more instable than the (Z)-isomer or that the (E)-didehydro-arginine underwent decomposition under the reaction conditions.
The thus synthesized mixtures of protected capreomycidine (rac)-14 and epicapreomycidine (rac)-15 as well as (rac)-17 and (rac)-18, respectively, were deprotected under acidic conditions to furnish a mixture of target compounds capreomycidine (rac)-1 and epicapreomycidine (rac)-2 as their dihydrochlorides in yields of 99 and 87 %, respectively (Fig. 5). A procedure for the separation of 1 and 2 by crystallization of the picrates is established (Bycroft et al. 1971b). Our novel biomimetic synthesis can therefore be considered a very convenient and rapid access to pure 1 and/or 2 in racemic form.
It was envisaged that a stereocontrolled version of the domino reaction would be difficult to achieve. First attempts to employ chiral bases such as (-)-quinine or (-)-sparteine instead of triethylamine resulted in low conversions and side reactions. Hence, it was decided to prepare N 2 -acetyl derivatives of 1 and 2 using the biomimetic route to enable acylase-mediated resolution of the product mixtures for separation of the L-from the D-isomers. Didehydro amino acids (Z)-11b and (Z)-11c were therefore used in the newly established protocol of Staudinger reduction (products 19a and 19b, not purified) and subsequent biomimetic domino transformation (Fig. 6). Surprisingly, the anticipated domino process stopped at the didehydro-arginine stage after the reduction and guanidinylation steps for both precursors (Z)-11b and (Z)-11c. Thus, compounds 8c and 8d were obtained in impure form and did not undergo cyclization towards the capreomycidine scaffold even at elevated temperatures up to 100°C. This indicates that the didehydro amino acid moiety is required to display a certain electrophilicity to enable the aza-Michael addition and that even slight changes, such as  the shift from carbamate to amide protection, can lead to a complete loss of reactivity. N-Acetylated capreomycidin derivatives for enzymatic resolution therefore had to be prepared via a different route. An attempted direct acetylation of target structures (rac)-1 and (rac)-2 did not provide the desired N a -acetylated products. It was therefore envisaged to synthesize the according N-Cbz-protected congeners first and then to exchange the N-Cbz for an N-acetyl moiety after the biomimetic cyclization process, followed by acidic cleavage of the tert-butyl ester (Fig. 7). When Cbz-protected phosphonate 20 (Schmidt et al. 1982(Schmidt et al. , 1984Schmidt and Wild 1985;Hamzavi et al. 2003) was employed in the Wittig-Horner reaction with aldehyde 10, Cbz-protected d-azidodidehydro a-amino acid ester (Z)-21 was obtained in 81 % yield (calculated as traces of aldehyde 10 could not completely be removed). Stereoisomer (E)-21 could also be isolated in 3 % yield, and again, application of the established 1 H NMR criteria for the distinction of (Z)-and (E)-didehydro amino acids (Mazurkiewicz et al. 2005) confirmed the stereochemical assignment. Subsequent Staudinger reduction of (Z)-21 furnished ornithine derivative 22, which was directly treated with guanidinylation reagent 13 to initiate the domino reaction. Thus, formation of the respective arginine derivative and domino-type ring closure provided a mixture of protected capreomycidine (rac)-23 and epicapreomycidine (rac)-24 (d.r.*1 : 1) in 62 % yield over two steps. The mixture of (rac)-23 and (rac)-24 was then subjected to a sequence of hydrogenolysis with in situ-acetylation (products (rac)-25 and (rac)-26) and acidic cleavage of the tert-butyl ester to afford a mixture of N a -acetylated products (rac)-27 and (rac)-28 in 77 % yield over two steps from the mixture of (rac)-23 and (rac)-24 (Fig. 7). However, attempts to  Fig. 7 Synthesis of N a -acetylated capreomycidines for enzymatic resolution: successful route via N a -Cbz-protected derivatives subject this mixture of N a -acetylated capreomycidines to acylase-mediated deacetylation to separate the L-from the D-isomers failed as no conversion could be detected. Apparently, N a -acetylated capreomycidines are no substrates for standard acylases. This might result either from steric hindrance due to the adjacent six-membered ring or from the positively charged guanidine group being fixed in a position close to the reaction site. The synthesis of guanidine-protected cyclization precursors 8e-h (see Table 1) was then attempted. It was anticipated that these compounds should not spontaneously undergo the aza-Michael addition as a result from the reduced nucleophilicity of the guanidine group. In contrast, one should be able to isolate fully protected didehydroarginine derivatives 8e-h and to subject them to aza-Michael cyclization in a separate step. This second reaction would then require activation either with a Lewis acid (in order to enhance the electrophilicity of the didehydro amino acid moiety) or a base (to deprotonate the urethaneprotected guanidine and make it nucleophilic). Thus, the domino reaction would be dissected into two separate transformations, and a chiral activating agent for the cyclization step might potentially provide stereoinduction and enable a stereoselective biomimetic synthesis of capreomycidines. In principle, the acidic properties of urethane-protected guanidines are established (Feichtinger et al. 1998). However, it was unclear if the guanidine moiety had to be fully protected in order to allow for sufficient deprotonation of the urethane moiety, and therefore, the additional N-benzyl group was introduced into potential cyclization precursors 8f and 8h.
For an efficient synthesis of 8e-h, d-azido-didehydro a-amino acids (Z)-11a and (Z)-11d were used and again reduced to moderately stable ornithine derivatives 12 and 16 (vide supra, Fig. 8). For the introduction of the benzyl group, 12 and 16 were subjected to reductive amination reactions with benzaldehyde to furnish N d -benzylated ornithines 29 and 30, which could not be obtained in pure form. All didehydro-ornithines 12, 16, 29 and 30 were transformed into the respective arginine derivatives by guanidinylation with reagent 31 (Tian et al. 1992) in the presence of silver(I) triflate. Thus, target structures 8eh were obtained in yields of 42-71 % (over two or three steps from the respective d-azido-didehydro a-amino acid).
Guanidine-protected cyclization precursors 8e-h were then subjected to numerous attempts to activate them for the aza-Michael addition. Initially, 8e was treated with a series of bases (K 2 CO 3 in MeCN, NaH, KHMDS or DBU in THF) for urethane deprotonation or with iron(III) chloride as a Lewis acid in dichloromethane for activation of the Michael acceptor system (reactions not displayed). No conversion towards the cyclic capreomycidine system was detected. As the bis-Boc-protected congener 8g was anticipated to be more reactive, its possible cyclization in the presence of DBU in THF was investigated next, but no conversion was observed even at elevated temperatures. It was therefore decided to skip cyclization studies on 8f and to conduct more detailed investigations regarding the aza-Michael cyclization of what should be the most reactive precursor both in terms of guanidine-urethane acidity and Michael acceptor activity, i.e. didehydro-arginine 8h. Hence, compound 8h was treated with a series of bases (K 2 CO 3 in DMSO, basic Al 2 O 3 in MeCN, KOt-Bu or NaH in THF) as well as a series of Lewis acids (FeCl 3 , SnCl 4 , Me 2 AlCl, EtAlCl 2 or Yb(OTf) 3 in dichloromethane, reactions not displayed). Furthermore, Takemoto's bifunctional chiral organocatalyst (Inokuma et al. 2006) was added to a solution of 8h in toluene to study if the combination of thiourea activation and basic moiety might be able to trigger the cyclization reaction. However, no conversion towards the cyclic capreomycidine system was detected in any of the listed cases. It was therefore concluded that the reactivity of the didehydro amino acid moiety was just not sufficient for aza-Michael additions if the nucleophilicity of the guanidine group is too low resulting from its derivatization. This again demonstrated that the biomimetic cyclization reaction could only work within a certain frame of reactivity of both moieties involved (vide supra). Thus, a dissection of the domino reaction into two separate transformations was impossible, preventing further

Conclusion
In summary, we have developed an unprecedented biomimetic pathway for the concise synthesis of the non-proteinogenic amino acids capreomycidine 1 and epicapreomycidine 2 in racemic form. The two key steps of this efficient synthetic route were (1) the preparation of protected d-azido-didehydro a-amino acid precursors using Wittig-Horner reactions and (2) a sequence of Staudinger reduction and a novel domino reaction for the construction of the capreomycidine scaffold. This domino-guanidinylation-aza-Michael-addition sequence mimics the biosynthetic ring closure of didehydro-arginine towards capreomycidine. Attempts to subject the furnished racemic mixtures of L-and D-amino acids to acylase-mediated enzymatic resolution were unsuccessful as N a -acetylated capreomycidines were apparently no substrates for acylases. Investigations on didehydro-arginine cyclization precursors with different protecting group patterns revealed that the domino reaction would only work within a certain frame of reactivity, i.e. both sufficient electrophilicity of the Michael acceptor and nucleophilicity of the guanidine moiety were essential. Fully protected didehydro-arginine derivatives 8e-h therefore did not undergo the aza-Michael addition anymore, even in the presence of highly reactive activating agents. However, both compounds 8e-h and their respective azide precursors might be useful synthetic intermediates for the preparation of isotope-labelled ornithine or arginine derivatives (also see Baldwin et al. 1993a, b). Overall, the obtained results demonstrate that a biomimetic 1,4-addition of a guanidine moiety to a didehydro amino acid unit is synthetically feasible, but that a fine-tuning of the reactivity is essential.