(+)-Zwittermicin A: assignment of its complete configuration by total synthesis of the enantiomer and implication of D-serine in its biosynthesis

(+)-Zwittermicin A (1), ^[1] a water-soluble natural antibiotic reported in 1994 from the fermentation of the soil-borne bacterium Bacillus cereus, shows significant activity against phytopathogenic fungi.^[2] Most importantly, 1 synergizes the bioactivity of the endotoxin produced by Bacillus thuringensis (BT), a ‘green’ insecticide used globally protection of vegetable crops and eradication of gypsy moth from forest trees.^[2,3] BT toxin and related biocontrol agents are important commodities used in the fight against declining agricultural production and rising third world food shortages.^[3b] The biosynthesis of the sugar-like 1 is very unusual; the molecule does not derive from carbohydrate metabolism, as the structure may suggest, but arises from a non-ribosomal peptidyl-polyketide synthase (NRPS-PKS) pathway, starting with an activated serine (Ser, C13–C15, zwittermicin A numbering). Zwittermicin A is the first described polyketide in which C₂ chain extensions occur by condensations of C₂ units (followed by loss of CO₂) derived from hydroxymalonate (HM, C7–C8) and aminomalonate (AM, C9–C10) in addition to the more common extender, malonate. ^[4]

Combinatorial biosynthetic engineering of AM PKS modules has great potential for production of exotic ‘non-natural’ amino-polyketides^[4c] and possible remodeling of PKS structures alkaloids by exploitation of the innate nucleophilicity of the NH₂ group. Despite high interest in 1, the structure of zwittermicin A has eluded attempts to define its configuration for 14 years until now.^[1c]

We report here the complete absolute stereostructure of (+)-1 by way of deductive reasoning and the first total synthesis of its enantiomer (−)-1. Our surprise finding – that C13-C15 formally derives from D-Ser, rather than L-Ser^[5] – has implications for structure-function of the loading domain in the NRPS-PKS complex that initiates biosynthesis of (+)-1.

Azidodiol 2, prepared from L-serine as described earlier, ^[5] was refunctionalized by TBDPS protection⁶ of the terminal alcohol, MOM protection^[7] of the secondary alcohol and removal of the TBDPS group^[6] to give 3 in high yield (85% three steps, Scheme 1).^[8] Transformation of the azido group in 3 to an N,N-dibenzyl group by hydrogenolysis (Lindlar’s catalyst^[9]) followed by N-benzylation^[6] gave a primary alcohol that was easily oxidized to the stable aldehyde 4 (84% three steps).

Scheme 1.

Synthesis of (−)-10. Reagents and conditions: (a) TBDPSCl, imidazole, DMF, 0 °C-rt, 4 h, 91%; (b) MeOCH₂Cl, Hünig’s base, CH₂Cl₂, 0 °C-rt, 56 h, 98%; (c) TBAF, THF, −10 °C, 4 h, 95%; (d) Lindlar cat., H₂, (1 atm), EtOH, 14 h, 98%; (e) BnBr, K₂CO₃, CH₃CN, 31 h, 91%; (f) (i) (COCl)₂, DMSO, CH₂Cl₂, −78 °C, (ii) Et₃N, 94%; (g) (i) 5, n-Bu₂BOTf, Et₃N, CH₂Cl₂, −78 to 0 °C, 3 h, (ii) 4, −78 to 0 °C, 2.5 h, 77%, dr 24:1; (h) H₂O₂, LiOH, 0 °C, 30 min, 96%; (i) TFA, 0 °C, 1 h, 98%; (j) (i) 6, EDCI, HOBt, DMF, 0 °C, 10 min, (ii) (−)-8, Et₃N, 0 °C-rt, 1 h, 81%; (k) (i) HCl, MeOH, H₂ (5 atm), Pd/C, 1 h, (ii) HCl, H₂O, H₂ (5 atm), Pd/C, 1 h, 76%. TBDPSCl= tert-butyldiphenylsilyl chloride, EDCI= 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloide, hOBt= 1-hydroxybenzotriazole.

Evan’s aldol addition of the chiral glycolate equivalent 5^[10] to 4 followed by removal of the chiral auxiliary under standard conditions afforded carboxylic acid 6 in 74% yield and 92% de (two steps)^[11] ready for coupling to the N-ureido-L-1,3-diaminopropionamide (−)-8 that was easily derived from the known amide 7.^[12]

Coupling of 6 and 8^[13] gave an amide 9 (81%) that was globally deprotected^[14] to afford compound (−)-10 with the configuration proposed for (+)-1.^[5] Although the ¹H and ¹³C NMR spectra (400 MHz, D₂O) of (+)-1^[15,16] and (−)-10 were almost identical at C10–C15 (see Supporting information), chemical shift differences at H8 [(−)-10, δ 4.53, d, J = 2.0 Hz: (+)-1, δ 4.56, d, J = 2.0 Hz] were readily revealed upon measurements of a mixture of the two compounds (Figure 2). Additionally, the specific rotation of (−)-10 ([α]_D −23.0°, H₂O) was opposite in sign and of larger magnitude than values measured for natural (+)-1 ([α]_D = +8.1°, H₂O; lit.^[1a] +8.9°) under the same conditions.

Figure 2.

¹H NMR spectra (400 MHz, D₂O): (a) 1:3 mole ratio of synthetic (−)-10 and natural (+)-1. (b) (−)-10 (c) (−)-1, and (d) 1:2 ratio of (−)-1 and natural (+)-1. Concentrations ~10 mM, no solvent suppression.

The relative configuration of the C8–C15 segment of (+)-1 was certain from analysis of ¹H NMR spin system topicities and ¹³C NMR chemical shift differences from a C₂ symmetric diamino tetraol derived from 2.^[5] Considering that the amino acid configuration in (+)-1 was unequivocally L-, ^[5] the ¹H and ¹³C NMR signals at C10–C15 showed negligible differences, and the largest ¹H NMR difference occurred at H8, we hypothesized that the mismatch was due to inversion of all configurations in the diaminopolyol-carboxylate moiety of (−)-10: C8–C11, C13, and C14. The latter would negate the original assumption of a formal biosynthesis of 1 derived from an L-Ser starter unit^[4a,5] in the NRPS loading domain and require involvement of D-Ser. In order to test this hypothesis, 12, a diastereomer of 9, was prepared (Scheme 2) by coupling carboxylic acid 6 with D-α-aminoamide (+)-8 (88%) (the latter compound was derived in two steps from the known acyl azide 11^[17] by Curtius rearrangement followed by ammoniolysis). Deprotection of 12 under the conditions used previously (Scheme 1)^[14] gave (−)-1 in 75% yield.

Scheme 2.

Synthesis of model (−)-1. Reagents and conditions: a) (i) μW, toluene, 110 °C, 15 min, (ii) THF, NH₃, 30 min, (iii) 2M NH₃, MeOH, 5 h, (iv) 1N NaOH, MeOH, 4.5 h, 62%; (b) TFA, 0 °C, 1 h, 99%; (c) (i) EDCI, HOBt, DMF, 0 °C, 10 min, (ii) (+)-8, Et₃N, 0 °C-rt, 1 h, 88%; (d) (i) HCl, MeOH, H₂ (5 atm), Pd/C, 1 h, (ii) HCl, H₂O, H₂ (5 atm), Pd/C, 1 h, 75%.

The NMR spectra of synthetic (−)-1 and natural (+)-1 were identical in all respects; co-addition of natural (+)-1 to (−)-1 and NMR measurements gave a single discrete set of ¹H (Figure 1) and ¹³C signals corresponding to those of natural (+)-zwittermicin A.^[1a]

Figure 1.

Natural (+)-zwittermicin A (1)

Finally, the specific rotation of synthetic (−)-1 ([α]_D −7.9°, H₂O) was opposite in sign and equal in magnitude of natural zwittermicin A. Therefore the configuration of zwittermicin A [(+)-1] is (4S,8R,9S,10S,11S,13S,14R) as depicted. The configurational assignment described here has implications for the biosynthesis of (+)-1. Three scenarios can be considered to explain the unexpected 14R configuration of zwittermicin A; direct incorporation of D-Ser at C13–C15, similar to that observed for the starter D-Ala residue of cyclosporine, ^[18a] α-epimerization of a carrier protein-bound L-Ser by an embedded epimerization domain, or the involvement of a dual function condensation/epimerization domain, such as those operating in the biosynthesis of arthrofactin^[18b] and enduracidin.^[18c] In the latter case, a single catalytic domain may be responsible for inversion of the α-configuration and coupling of the resultant thio-acyl D-Ser residue with a down-stream acceptor residue, however, this mechanism has yet to be associated with a mixed NRPS-PKS system. Although details have been reported for gene products ZmAG-ZmAI responsible for the AM extender unit, ^[4c] the identification of the genes and a mechanism responsible for the Ser loading domain and incorporation into C13–C15 of 1 are still unclear. Resolution of this mystery awaits more detailed annotation of the gene cluster for biosynthesis of (+)-1.

We briefly compared the biological activity of (+)-zwittermicin A with that of its synthetic enantiomer (−)-1 and by measuring the susceptibility to pathogenic fungi and Fluconazole-resistant pathogens of the genus Candida.

The minimum inhibitory activities of authentic natural (+)-1 against Candida albicans ATCC 14503 (MIC 55.7 μg/ml) and the Fluconazole-resistant strain C. albicans 96–489 (MIC 59.5 μg/ml) were found to comparable to the antifungal activities found by Handelsman et al. for (+)-1 against a range of plant pathogenic fungi of agricultural importance. ent-Zwittermicin A (−)-1, on the other hand was inactive (MIC > 128 μg/ml) under the same conditions. This interesting result implies that the activity of (+)-1 is due not to non-specific interactions with the diaminopolyol, but more closely allied to either transport across the cell wall or membrane, or a mechanism that implicates a more subtle chiral recognition motif at an as-yet unidentified intra-cellular target.

In summary, the absolute stereostructure of (+)-zwittermicin A (1) has been assigned unambiguously by total synthesis of (−)-1 in an overall yield of 1.9% (20 steps from N,N-dibenzyl-L-serine methyl ester). Interpretation of the configuration of (+)-1 implicates a ‘D-Ser’ motif in the biosynthesis of C13–C15 consistent with an antipodal configuration of the propagated Ser starter unit. Zwittermicin A and its enantiomer exhibit a differential activity against fungal pathogens that underscores the importance of chirality to the biological activity of these acyclic diaminopolyol natural products.

Table 1.

In vitro minimum inhibitory activities of natural (+)-1 and (−)-1, against pathogenic Candida species.

Fungal Strains	(+)-1 MIC (μg/ml)a	(−)-1 MIC (μg/ml)a
Candida albicans 96–489b	55.7	>128
C. glabrata	59.5	>128
C. albicans UCDFR1c	>128	>128
C. albicans ATCC 14503	>128	>128
C. krusei	>128	>128

^aCompounds tested as their free-bases. MIC defined as the lowest concentration eliciting 90% growth inhibition.

^bclinical isolate.

^cFluconazole-resistant Candida strain raised from C. albicans ATCC 14503 by passage through sub-inhibitory Fluconazole. See Supporting Information for details of culture conditions.