Total synthesis and antimicrobial evaluation of natural albomycins against clinical pathogens

Development of effective antimicrobial agents continues to be a great challenge, particularly due to the increasing resistance of superbugs and frequent hospital breakouts. There is an urgent need for more potent and safer antibiotics with novel scaffolds. As historically many commercial drugs were derived from natural products, discovery of antimicrobial agents from complex natural product structures still holds a great promise. Herein, we report the total synthesis of natural albomycins δ1 (1a), δ2 (1b), and ε (1c), which validates the structures of these peptidylnucleoside compounds and allows for synthetic access to bioactive albomycin analogs. The efficient synthesis of albomycins enables extensive evaluations of these natural products against model bacteria and clinical pathogens. Albomycin δ2 has the potential to be developed into an antibacterial drug to treat Streptococcus pneumoniae and Staphylococcus aureus infections.

S ideromycins are a class of antibiotics covalently linked to siderophores 1 . They are actively transported into bacterial cells via siderophore uptake pathways commonly found in bacterial pathogens, by the so-called "Trojan horse" strategy, resulting in outstanding cell envelope permeability and very low minimum inhibitory concentrations (MICs). These pathogenspecific antibiotics are promising drug candidates for the treatment of various bacterial infections [2][3][4] . A few naturally occurring sideromycins have been discovered. Among these, albomycins, originally reported as grisein, were first isolated from soil microorganism Streptomyces griseus in 1947 [5][6][7][8][9] . Albomycins exhibited potent inhibitory activities against a number of Gramnegative, as well as Gram-positive bacteria, including multi-drug resistant strains 1,10,11 . For instance, albomycins exhibited an MIC value of 10 ng/mL against Streptococcus pneumoniae and 5 ng/mL against Escherichia coli, which is almost tenfold more potent than penicillin 12 . Moreover, no toxicity was observed during in vivo studies of albomycins, and it was well tolerated and safe up to a maximum dose evaluated in mice 10 . Albomycins have been successfully used to treat human bacterial infections in the Soviet Union 10 .
The structures of albomycins were fully elucidated by Benz and coworkers in 1982 13,14 , 35 years after their initial isolation. Albomycins δ 1 (1a), δ 2 (1b), and ε (1c) are all composed of a triδ-N-hydroxy-L-ornithine peptide siderophore and a thionucleoside warhead with six consecutive chiral centers, and differ only in the C4 substituent (R) of the pyrimidine nucleobase ( Fig. 1). As for 1b, the thionucleoside warhead is a potent seryl-tRNA synthetase inhibitor known as SB-217452 15 . The highly complex and densely functionalized structures, together with their important therapeutic potential, have made albomycins attractive targets for synthesis. Different synthetic strategies for the tri-δ-N-hydroxy-L-ornithine peptide siderophore have been described by the groups of Benz 16,17 and Miller [18][19][20] . The synthesis of the thionucleoside moiety of 1a was briefly described by Holzapfel et al. in 1991 21 , though with incomplete data. However, no total synthesis of albomycins has been reported. In one study 22 , an oxygen analog of 1a was synthesized. Surprisingly, a single replacement of the sulfur with oxygen resulted in the complete loss of antibacterial activity, suggesting a critical role of the sulfur atom in the activity of albomycins. A biosynthetic approach has also been attempted for the generation of albomycin analogs 23 , but albomycin production by S. griseus is difficult and has yet to be scaled up efficiently 24 . Herein, we describe the total synthesis of the three natural albomycins (1a-c), which features a Pummerer reaction for nucleobase introduction and an aldol reaction to expand the side chain of thionucleoside. Their biological evaluations demonstrate that albomycin δ 2 is a promising lead candidate for treating S. pneumoniae and S. aureus infections and warrants further development.

Results
Synthetic strategy. The retrosynthetic analysis of albomycins (1a-c) in a collective fashion is shown in Fig. 1. The amide bond linking to the thionucleoside core was first disconnected to generate tetrapeptide fragment 2 and thionucleosides 3a-c. Tetrapeptide 2 could be readily accessed via the condensations of tripeptide 4 with L-serine tert-butyl ester. Tripeptide 4 could be derived from amino acid 6, which could arise from the direct oxidation of protected L-ornithine 8. We anticipated that Amidation Fig. 1 Chemical structures of albomycins δ 1 , δ 2 , and ε, and their synthetic strategies. The key disconnections involve a Pummerer reaction for nucleobase introduction and an aldol reaction to expand the side chain of thionucleoside. tBu tert-butyl, Bz benzoyl, Fmoc fluorenylmethyloxycarbonyl, PMB p-methoxybenzyl thionucleosides 3a-c could be installed by substrate-directed asymmetric aldol condensation reaction from thionucleosides 5a-c, which could in turn be prepared through a Pummerer reaction from thiosugar 7 and uracil. Thiosugar 7 could be further traced back to the known L-(+)-lyxose derivative 9. Due to the structural feature and labile nature of albomycins, the protecting group strategy had to be delicately selected to accomplish the total synthesis. Because of the presence of reducible hydroxamic acids in the siderophore moiety and an imine group in the thionucleoside moiety of 1b and 1c, protecting groups such as Cbz and Bn which usually need to be deprotected under reducing conditions of H 2 and Pd/C in the final stage were avoided. As for the hydroxyl groups in the thionucleoside fragment, orthogonality of protecting groups and their influence on stereoselectivity of uracilation were also key considerations.
Total synthesis of albomycins δ 1 , δ 2 , and ε. Our synthetic efforts commenced with the preparation of tetrapeptide 2 (Fig. 2). The core component hydroxamic acid 6 of tetrapeptide 2 was first synthesized from N 2 -Boc-L-ornithine tert-butyl ester 8 25 . Oxidation of the free amine in N 2 -Boc-L-ornithine tert-butyl ester 8 with benzoyl peroxide [25][26][27] , followed by acylation under biphasic conditions afforded the fully protected hydroxamate 6 in 80% yield. Removal of tert-butyl-carbonate (Boc) and tert-butyl (t-Bu) groups with trifluoroacetic acid (TFA) provided the TFA salt 10. The resulting amine group was protected with Fmoc to deliver acid 11 in 91% yield over two steps. Then, we turned to coupling 11 and 10 under active ester-mediated coupling conditions (DCC, NHS) 20 , which proceeded smoothly to produce dipeptide 12. Unfortunately, application of the same conditions to synthesize tripeptide 4 led to significant epimerization. Failure to optimize the condition after numerous efforts prompted us to explore an alternative strategy. Fortunately, under active amide-mediated coupling conditions developed by  , tripeptide 4 could be obtained in an iterative fashion from 11 in 77% yield over two steps without any detectable epimerization. Tripeptide 4 was condensed with L-serine tert-butyl ester hydrochloride in the presence of HATU and DIPEA to generate tert-butyl ester 13 in 95% yield. Treatment of 13 with TFA in dichloromethane at 0°C proceeded smoothly to yield the corresponding tetrapeptide 2.
With tetrapeptide 2 in hand, we initiated the synthesis of thionucleosides 5a-c ( Fig. 3). At first, lactone 9 32 was transformed into lactone 14 by a two-step sequence involving Mitsunobu reaction and removal of the isopropylidene protecting group with TFA. Selective mono-tosylation of lactone 14 with tosyl chloride/DABCO generated tosylate 15 in 87% yield. In this case, use of pyridine led to low conversion because of its weaker basicity, whereas Et 3 N to a bis-tosylation byproduct which further underwent elimination to produce the corresponding α,β-unsaturated tosylate. The transformation of 15 to 16 with K 2 CO 3 in MeOH at room temperature gave the product with the undesired configuration at C4' exclusively. A variety of bases and temperatures were examined for the 5membered thio-ring closure reaction, and eventually the use of K 2 CO 3 as a base at −30°C in MeOH smoothly led to 16 in 81% yield without any detectable epimerization. It was surprising that the reaction temperature had such an impact on the stereoselectivity. As mentioned above, the proper choice of protecting group for the trans-1,2-diol is crucial, because the neighboring group effect could influence the stereoselectivity of the following uracilation reaction 33 . Initially, when the trans-1,2-diol was protected as a diester, the Pummerer reaction indeed provided the desired product as a single diastereomer with moderate yield. However, this protected diester was incompatible with the subsequent selective methyl ester reduction and later stage aldol reaction. Thus, trans-1,2-diol 15 was protected with 2,3-butanedione under acid catalysis, followed by reduction of the methyl ester with DIBAL-H to afford sulfide 7. Treatment of sulfide 7 with m-CPBA provided the corresponding sulfoxide 17 in 95% yield, and subsequent Pummerer reaction 34,35 with DIPEA as a base gave rise to 18 and epi-18 in 86% yield as a 1:1 diastereoisomeric mixture. ARTICLE Commonly used Et 3 N was not suitable for the Pummerer reaction because it could act as a nucleophile, leading to formation of a triethylammonium adduct, and thus reduced yield 36 . The N-methyl group was introduced by a conventional procedure to afford 5a, which is an advanced intermediate for the synthesis of albomycin δ 1 . The structure of 5a was verified by X-ray crystallographic analysis (Fig. 3). To achieve the synthesis of albomycin δ 2 and ε, a practical access to the imine 20 was required. First, 18 was protected as its TBS ether 19, which was treated with TPSCl in the presence of DMAP and Et 3 N, followed by NH 4 OH to produce a cytosine derivative in 90% yield over two steps in one pot, and then methylation of N3 delivered 20 in 96% yield. For the subsequent installation of the N4 carbamoyl group, aminolysis of N4-phenoxycarbonyl 37 did not take place due to the steric hindrance of the N3 methyl group. Inspired by the N-PMB carbamoylimidazole urea formation reported by Batey et al. 38 , we developed a scalable and efficient protocol to prepare 22b. In the presence of Et 3 N, 20 was allowed to react with 21 under reflux, giving rise to 22b in 96% yield. Additionally, Fmoc was chosen as the protecting group for 20, leading to 22c, which is a key intermediate for the synthesis of albomycin ε.
With fragments 5a-c in hand, the next challenge was to install the side chain with the desired stereochemistry (Fig. 4). Oxidation of alcohols 5a-c with IBX gave the corresponding aldehydes 23a-c in excellent yields. Sidechain extension via aldol reaction required a delicate choice of protecting groups for the glycine moiety, which is essential for good stereoselectivity. Inspired by Trost's work 39 , we found that the lithium salt of N-(diphenylmethylene) glycine methyl ester could react with aldehydes 23a-c at −78°C, and subsequent treatment of the condensation products with 2 M aqueous HCl yielded 3a (86%, 2 steps, d.r. 4:1), 3b (83%, 2 steps, d.r. 3:1), and   3c (75%, 2 steps, d.r. 3:1) respectively. The structures of 3a and 3b were unambiguously established by X-ray crystallographic analysis, and the molecular structure of 3c was substantiated by X-ray crystallography of its benzoyl derivative 3c' (Fig. 4).
With key fragments 2 and 3a-c in place, their assembly into albomycins (1a-c) was embarked (Fig. 5). Thionucleosides 3a-c were first condensed with tetrapeptide 2 to afford 25a-c in excellent yields. The final deprotection steps from 25a-c to the final albomycins (1a-c) required delicate adjustments of the reaction conditions in order to preserve the sensitive amide bond 40 on the cytosine of 1b and 1c. First, oxidative deprotection of the N-PMB group in 25b proceeded smoothly. Addition of CF 3 CO 2 H and H 2 O to the resulting crude product removed the bisacetal moiety within the molecules. The reaction time for these two deprotection steps needed to be carefully controlled, otherwise it would lead to hydrolysis of the imine. After surveying numerous reaction conditions, we found that aqueous K 2 CO 3 (150 mg/mL) could efficiently remove the remaining protecting benzoyl, methyl ester, and Fmoc groups in one pot to afford albomcyin δ 2 (1b). These three deprotection steps (25b to 1b) were carried out without purifying the intermediates and the final product albomycin δ 2 (1b) was purified with Sephadex TM G-15. We subsequently exploited the flexibility of this sequential deprotection. To our pleasure, following the same protocals, both albomycins δ 1 (1a) and ε (1c) were obtained in good yields. It's worth noting that albomcyin ε (1c) was not stable in D 2 O, and slowly converted to albomycin δ 1 (1a) over time. Albomycin δ 2 (1b) was much more stable than albomcyin ε (1c), and remained unchanged in D 2 O at 4°C for over one month.
Biological assessment. As the three naturally occurring albomycins became synthetically accessible, we evaluated their potential as therapeutic agents and determined their MIC values against three Gram-positive and three Gram-negative bacteria species following the protocol from the Clinical and Laboratory Standards Institute (CLSI). Commercial antibiotic, ciprofloxacin, was used as a positive control in the MIC determination experiment. As shown in Table 1, albomycin δ 1 (1a) and δ 2 (1b) exhibited 8fold more potency than ciprofloxacin against S. pneumoniae ATCC 49619. 1b inhibited S. aureus USA 300 strain NRS384 41 , a virulent methicillin resistant MRSA strain, with an MIC of 0.125 µg/mL, and was 16-fold more potent than ciprofloxacin. 1b exhibited an MIC value of 0.5 µg/mL against Bacillus subtilis ATCC 6633, which was less active compared to ciprofloxacin. As for Escherichia coli BJ 5183 42 , 1b showed about 8-fold higher potency than 1a. Particularly impressive is the potency of compound 1a towards the fastidious Neisseria gonorrhoeae ATCC 49226 with a 3.9 ng/mL MIC value while 1b was completely inactive. These results led us to conclude that the C4 substituent of nucleobase in albomycins played an important role rendering their antibacterial activity. Biochemical analysis of albomycin nucleobase analogs will test this hypothesis. All three albomycins displayed no activity towards Gram-negative bacteria Salmonella typhi, and albomycin ε (1c) was inactive to all these strains, which was previously unknown. World Health Organization (WHO) published a priority list of antibiotic-resistant pathogenic bacteria for developing new and effective antibiotic treatments. Both S. pneumoniae and S. aureus are on the list due to their increasing multidrug resistance. Ciprofloxacin, vancomycin and penicillin G are on the WHO Model List of Essential Medicines (EML). Penicillin G is among the first medications against bacterial infections and vancomycin has been hailed as the last line of defense. To explore the potential to develop albomycin δ 2 into an effective antibiotic, we screened albomycin δ 2 against a random collection of 27 clinical S. pneumoniae and S. aureus isolates (three of them are MRSA strains), and compared it with ciprofloxacin, vancomycin, and penicillin G (Table 2) in inhibiting the growth of these clinical pathogens. All these S. pneumoniae and S. aureus strains were freshly isolated from patients in clinic. As shown in Table 2, albomycin δ 2 exhibited excellent anti-S. pneumoniae and S. aureus activities better than the other three antibiotics in most cases, while many of these strains displayed severely natural resistance to penicillin G. The MIC values of albomycin δ 2 were well below those of ciprofloxacin, vancomycin, and penicillin G, and in a number of cases reaching 1000 times lower. The influence of iron concentration on the antibacterial activity was also studied by conducting assays in iron-rich and iron-deficient media ( Table 3). The antibiotic activity of albomycin δ 2 against S. pneumoniae strains was significantly increased in iron-deficient media, which most closely mimics the physiological situation in a human host, wherein iron is sequestered in macromolecules such as heme 43 . Two isolates, strains S15 and S29, which showed the most resistance to albomycin δ 2 in iron-rich media, became highly susceptible under iron-depleted conditions. The MIC values of albomycin δ 2 against S.  aureus and E. coli strains were not influenced by iron concentration. The three control antibiotics did not show any dependence on iron concentration in any of the strains tested. These results suggest that albomycin δ 2 is a promising antibiotic candidate for further clinical drug development.

Discussion
In summary, the successful execution of the convergent strategy has led to the total synthesis of albomycins δ 1 (1a), δ 2 (1b), and ε (1c). Antibacterial assessment of albomycins revealed that C4 substituent on the nucleobase in albomycin plays an essential role in their antibacterial activity. Albomycin δ 2 exhibited potent antimicrobial activities against clinical S. pneumoniae and S. aureus isolates including MRSA. Further studies to evaluate albomycin δ 2 as a potentially effective and safe antibiotic are ongoing.

Methods
General. All air-sensitive and water-sensitive reactions were carried out under a nitrogen atmosphere with dry solvents under anhydrous conditions, unless otherwise noted. Tetrahydrofuran (THF) was distilled over sodium and benzophenone, dichloromethane (CH 2 Cl 2 ), N, N-dimethylformamide (DMF), triethylamine (Et 3 N), and N,N-diisopropylamine (DIPEA) over calcium hydride. All other solvents, as well as starting materials and reagents were obtained from commercial sources and used without further purification. Reactions were monitored by analytical thin-layer chromatography (TLC) on Merck silica gel 60 F 254 plates (0.25 mm), visualized by ultraviolet light and/or by staining with phosphomolybdic acid in EtOH. Retention factor (R f ) values were measured using a 5 × 2 cm TLC plate in a developing chamber containing the solvent system described. Yields refer to the isolated yields after silica gel flash column chromatography, unless otherwise stated. 1 H NMR spectra were obtained on an Agilent 400MR or 600MR DD2 spectrometer at ambient temperature. Chemical shifts were reported in parts per million (ppm), relative to either a tetramethylsilane (TMS) internal standard or the signals due to the solvent. 13