Structural and Functional Analysis of Pantocin A: An Antibiotic from <i>Pantoea agglomerans</i> Discovered by Heterologous Expression of Cloned Genes

Abstract

Small-molecule chemistry and biochemical techniques have been used to determine the structure and mode of function of pantocin A (see structure), a naturally occurring antibiotic. Pantocin A functions as an inhibitor of histidine biosynthesis, as shown by its ability to inhibit the action of L-histidinol phosphate aminotransferase, which converts imidazole acetol phosphate to L-histidinol phosphate in Escherichia coli. The epiphytic bacterium Erwinia amylovora causes fire blight, a devastating disease of rosaceous plants such as apple and pear.1 The closely related but nonpathogenic bacterium Pantoea agglomerans (formerly E. herbicola), which usually accompanies E. amylovora in the wild,2 produces a family of antibiotics that inhibit E. amylovora in a poorly understood strain-dependent fashion.3, 4 The numerous members of this antibiotic family constitute a source of biocontrol agents for fire blight, a pool of new and potentially useful antibiotics against human bacterial pathogens, and a model of nature's ability to produce a “library” of biologically active molecules. The complexity of antibiotic production, especially the multiple antibiotics produced by a given strain and low production in liquid culture, has hampered chemical investigations. We have taken a genomic approach to deconvolute the multiple antibiotics and to address the production problems by a particular strain. Heterologous expression of a genomic DNA library from P. agglomerans in Escherichia coli provides access to the small-molecule antibiotics. Since both P. agglomerans and E. coli are members of the widely distributed Enterobacteriacea, E. coli should be a suitable host for heterologous expression of P. agglomerans genes that encode antibiotics. Herein we report the isolation and structure determination of pantocin A, a new peptide-derived antibiotic, and its molecular target. A genomic library of P. agglomerans strain Eh318 was constructed in E. coli strain DH5α.5 Two distinct antibiotic-producing cosmid clones, pCPP702 and pCPP719, were identified, and the two antibiotics they produced were trivially named pantocin A and B, respectively. One indication that the two antibiotics were different came from studies which showed that arginine suppresses the antibiotic activity of pantocin B while histidine suppresses the antibiotic activity of pantocin A. In previous work, pantocin B (PB, 1)1 was structurally characterized, and its ability to potently inhibit N-acetylornithine transaminase, which catalyzes a committed step in arginine biosynthesis, was established.6–8 The characterization of pantocin A (PA, 2) was plagued by inconsistent and low levels of production in liquid culture, together with its acid, base, and thermal lability. Significant reproducible production was eventually achieved by culturing E. coli XL2Blue harboring a 3.5 kb subclone of pCPP702 in the high copy number plasmid pUC19 (pUC449). Isolation of 2 was guided by antibiotic activity against the E. amylovora strain Ea273 (IC50=200 nM). Pantocin A (2) was purified from large-scale fermentation of E. coli XL2Blue-pUC449 by four chromatographic separations using anion-exchange, reverse-phase, and hydrophilic interactions. This compound is an optically active small molecule, and its structure was established by spectroscopic methods and synthesis of a hydrogenated derivative. PA (2) has a molecular formula of C13H16N4O5 as determined from HRFAB-MS studies. The 13C NMR (Table 1) and DEPT spectra indicate four carbonyl, four olefinic, three methine, and two methylene carbon atoms. Three vinylic, a highly deshielded methine, two Cα amino acid, four Cβ amino acid, and three exchangeable protons could be characterized from 1H and 1H-13C HMQC NMR spectra recorded in D2O and [D6]DMSO. A loss of m/z=132 in the tandem ESI-MS study suggested the presence of an asparagine residue, which was confirmed by 2D NMR studies. Two spin systems (H6/H7a,b/H8 and H1/H2/H3) identified by COSY experiments coupled with HMBC correlations (H6/C5, H8/C9, H1/C9, and H3/C1′) established a continuous nine-carbon chain (C5 to C1′) containing the non-asparagine carbon atoms in 2. Two additional HMBC correlations (2′NH/C1′ and H3′/C1′) showed that the C9 and asparagine fragments must be connected by an amide bond. Atom No. 13C[a] 1H[b] 15N[c] 1 126.3 6.35 (1 H, dd, 6.0, 2.0) 2 131.5 6.15 (1 H, dd, 6.0, 1.7) 3 66.4 5.21 (1 H, br s) 4-N 145.1 (s) 5 167.9 6 50.2 3.58 (1 H, dd, 13.6, 7.2) 6-NH2 36.5 (s) 7 28.6 2.57 (1 H, m) 2.33 (1 H, m) 8 95.5 5.12 (1 H, dd, 6.0, 2.4) 9 143.3 1′ 166.2 2′-NH 8.27 (1 H, d, 7.5) 126.0 (d, 93.0) 3′ 49.8 4.29 (1 H, m) 4′ 37.9 2.50 (1 H, dd, 15.2, 7.0) 2.39 (1 H, dd, 15.2, 5.7) 5′ 171.6 5′-NH2 6.84, 7.72 (2 H, br s) 113.6 (t, 90.0) 6′ 172.6 15N NMR experiments on 15N-labeled 2, which was made using 15N ammonium sulfate in minimal culture media, revealed the nitrogen connectivity. A highly downfield shifted 15N resonance (δ=145.1 ppm) with correlations to H1, H2, H3, H6, and H8 required N4 to be connected to C3, C5, and C9. A 15N resonance at δ=36.5 ppm, typical of an amine, and its HMBC correlations with H6 and H7 required an amino group at C6, thus completing the planar structure of 2. The stereocenters at C3′, C3, and C6 were defined by the hydrogenation of 2, NOESY experiments, and synthesis. Hydrogenation of 2 gave 3 as a single diastereomer, and NOESY correlations between H3/H9 and H6/H9 require H3 and H6 to be on the same face (Scheme 1). Hydrolysis of 3 gave 4 and L-aspartic acid (5), which set the absolute configuration at C3′ as S. Both 3 and its 3(R),6(R),9(R) diastereomer were prepared from 6, which was synthesized from L- or D-Glu(OMe)-OMe following a literature procedure9 (Scheme 1). HPLC analysis of synthetic 3 and 3 derived from 2 indicated that the absolute configuration of 3 was 3(S),6(S). Thus, C3, C6, and C3′ all have the absolute configuration corresponding to an L-amino acid, and the absolute stereostructure of 2 is as shown. a) NaOH/dioxane; CH3I/K2CO3/DMSO; b) H2/Pd-C; c) (Boc)2O/TEA/DMF; NaOH/CH3OH-H2O; d) H-Asn-OtBu⋅HCl, EDC/NMM/DMF; e) TFA/CH2Cl2. Boc=tert-butoxycarbonyl, TEA=triethylamine, EDC=3-(3-dimethylaminopropyl)-1-ethylcarbodiimide, NMM=N-methylmorpholine, TFA=trifluoroacetic acid. Perhaps the most striking feature of 2, which to the best of our knowledge has no precedents in reported natural or synthetic compounds, is its very existence given the apparent ease with which it could aromatize to a pyrrole. Removal of a proton from C3, which is α to the carbonyl group at C1′, followed by reprotonation at C8 would generate a pyrrole, and this process likely underlies the lability of 2. A minor inactive compound was also recovered on isolating 2, and this was characterized as 8.0 Compound 8 arises from the aromatization and hydrolysis of the bicyclic amide bond. No material representing aromatization unaccompanied by hydrolysis was isolated—an observation consistent with the known rapid hydrolysis of N-acylpyrroles.10 While 2 is sensitive to acid, base, and elevated temperatures, it survives for days at room temperature in aqueous solution under neutral pH conditions. Two structural features of 2 undoubtedly contribute to its modest persistence: the C3H bond is shown in energy-minimized structures of 2 not to be aligned with the π bond of the C1′ carbonyl group, and this lack of conjugation creates a kinetic barrier to proton removal (see structure 9). The out of conjugation conformation is at least partially enforced by hydrogen bonds from asparagine-derived NH groups to the C5 carbonyl oxygen atom. The electron-withdrawing effect of the C5 carbonyl group also diminishes the thermodynamic benefit of aromatization, and this electron-withdrawing effect is enhanced by the same hydrogen bonds. With the structure of 2 established, we addressed how such a polar molecule gains entry into E. amylovora cells and the nature of its molecular target.11 Addition of the synthetic tripeptide Ala-Gly-Gly to the test medium, in amounts exceeding three times that of 2, suppresses the antibiotic activity of 2. Thus, 2 likely gains entry through the tripeptide transporter, and the added tripeptide monopolizes the transporter thereby limiting cellular entry of 2.6, 11 L-histidinol phosphate aminotransferase is a pyridoxal-5′-phosphate (PLP) dependent enzyme that couples the reductive amination of imidazole acetol phosphate with the conversion of L-glutamate into α-ketoglutarate. Pantocin A and B share a close functional relationship that is not apparent from their structures. Both gain access to their cells via the tripeptide transporter and inhibit transaminase-catalyzed amino acid biosynthetic steps that introduce nitrogen functionality.6, 11 Whereas 1 inhibits arginine biosynthesis, 2 inhibits histidine biosynthesis. This functional similarity is unsurprising since they have the common goal of suppressing the growth of bacterial competitors in a nitrogen-poor environment—an environment where import by tripeptide transporters and amino acid biosynthesis are essential. The structural and functional analysis of 2 (PA) reported here used techniques of small-molecule chemistry and biochemical analysis. The cosmid library/heterologous expression approach provides ready access to the DNA conferring production of 2 on E. coli XL2Blue-pUC449. Analysis of this DNA provides insights into the biosynthesis of 2 and identifies strains producing antibiotics related to 2 as discussed in the following communication. Cultivation and isolation: E. coli XL2Blue-pUC449 was grown in a shaker maintained at 30 °C for 24 h in minimal media supplemented with 100 μg mL−1 ampicillin.5 The culture supernatant was applied to a Dowex 1X8-200 (HCO3−) anion-exchange column and eluted with CO2-saturated H2O. The active fractions were lyophilized, then applied to a C18 solid-phase extraction column and eluted with H2O/CH3CN (100/0 to 90/10). Further fractionation by HPLC, first with a reverse-phase column (Vydac, 218TP510) using H2O/0.1 M NH4OAc/CH3CN (94/5/1), then with a polyhydroxyethyl A column (Poly LC, Inc.) using H2O/0.1 M NH4OAc/CH3CN (20/5/75) gave compounds 2 and 8 in yields of approximately 1 mg L−1 and 0.2 mg L−1, respectively. Cloning of E. coli-pUC449: Pantocin A producing cosmid clone pCPP702 was subcloned to generate pCPP1051, from which a XbaI-HindIII fragment was further subcloned into pBluescript to yield plasmid pCPP717.5 The cloned insert was sequenced (GenBank accession No. U81376). A 3.5 kb DNA fragment was then amplified by a polymerase chain reaction from pCPP717 using primer (5′-CCGCATCTAGAGTAGGTATGAC-3′) and primer (5′-ATACTCTGCAGAGTTGGTGCTCCA-3′), and ligated into pUC19 at the XbaI and PstI sites. Recombinants were selected from transformed E. coli XL2 Blue cells on Luria–Bertani plates supplemented with ampicillin at 100 μg mL−1, to yield clone E. coli XL2 Blue-pUC449. 2: [α]=−291.1° (c=1.8, H2O); IR (KBr): max=3700–2400 (br), 1676, 1578, 1402 cm−1; UV (H2O): λmax (ε)=200 (12 089), 272 nm (5447); HRFAB-MS: m/z: 331.1018 [M+Na]+, calcd m/z: 331.1018 for C13H16N4O5Na; ESI-MS/MS: m/z: 309 [M+H]+, 177, 159, 151, 149. 8: 1H NMR (500 MHz, D2O, 25 °C): δ=6.81 (1 H,d, J=3.7 Hz, H3), 6.12 (1 H, d, J=3.7 Hz, H4), 4.66 (1 H, dd, J=9.0, 4.5 Hz, H3′(Asn αH)), 3.75 (1 H, t, J=6.5 Hz, H3′′), 2.85 (1 H, dd, J=15.0, 4.5 Hz, H4′a), 2.80–2.69 (3 H, m, H1′′ and H4′b), 2.25–2.10 ppm (2 H, m, H2′′); 13C NMR (100 MHz, 5 %CD3OD in D2O, 25 °C): δ=178.6 (C6′, COOH), 177.0 (C5′, CONH2), 175.4 (C4′′) 163.4 (C1′), 137.4 (C5), 125.0 (C2), 113.5 (C3), 108.6 (C4), 55.3 (C3′′), 52.9 (C3′), 39.1 (C4′), 31.5 (C2′′), 23.8 ppm (C1′′); ESI-MS/MS/MS: m/z: 327 [M+H]+, 310, 195, 177, 149, 132, 106; UV (H2O): λmax (ε)=200 (6438), 220 (2994, sh), 275 nm (10 887); IR (thin film): max=3700–2400 (br), 1668, 1604, 1533, 1404 cm−1. Preparation of 3 from 2: A solution of 2 (2.4 mg) in deionized H2O (300 μL) was treated with Pd/C (10 wt %, 0.66 mg) and stirred under hydrogen at room temperature for 8 h. The reaction mixture was filtered over celite, washed with deionized H2O (2 mL), and the combined aqueous fractions were dried to give 1.7 mg of 3. 1H NMR (500 MHz, D2O, 25 °C): δ=4.46 (1 H, d, J=9.0 Hz, H3), 4.40 (1 H, dd, J=8.0, 5.0 Hz, H3′), 3.93 (1 H, t, J=8.5 Hz, H6), 3.74 (1 H, m, H9), 2.76 (1 H, dd, J=15.0, 5.0 Hz, H4′a), 2.65 (1 H, dd, J=15.0, 8.0 Hz, H4′b), 2.35 (1 H, m, H7a), 2.25–2.10 (4 H, m, H2a, 1a, 8a, 2b), 1.91 (1 H, m, H7b), 1.67 (1 H, m, H8b), 1.64 ppm (1 H, m, H1b); 13C NMR (100 MHz, D2O + 5 %CD3OD, 25 °C): δ=177.8 (C6′), 176.6 (C5′), 173.6 (C1′), 169.5 (C5), 61.0 (C3), 59.2 (C9), 53.2 (C3′), 49.6 (C6), 38.6 (C4′), 32.3 (C1), 30.2 (C2), 26.2 (C8), 25.6 (C7); [α]=−57.8° (c=0.16, H2O); UV (H2O): λmax (ε)=200 nm (2200); IR (thin film): max=3700–2400 (br), 2954 (sh), 1668, 1529 (sh), 1448, 1392, 1342 cm−1; ESI-MS/MS: m/z: 313 [M+H]+, 181, 153, 125. Chiral TLC: Chiral TLC analyses were performed on CHIRALPLATE (Macherey-Nagel) using solvent A (CH3OH/H2O/CH3CN, 5/5/20) or B (acetone/H2O/CH3OH, 10/2/2). 3 (0.15 mg, 0.5 μmol) was heated in 6 M HCl (300 μL) at 85 °C for 19 h to give 4 and 5. Authentic D- and L- Asn were also treated in the same manner. Compound 5 migrated the same distance as treated L-Asn (Rf=0.47 in A, 0.66 in B), which was easily distinguishable from treated D-Asn (Rf=0.40 in A, 0.58 in B) 3: As shown in Scheme 1, N-(PhF) L-or D-dimethyl glutamate was used in a Claisen condensation, reductive amination, and lactam-cyclization sequence to give enantiopure (3S,6S,9S)-7 or (3R,6R,9R)-7 following the procedure of Lombart and Lubell.9 Coupling with tert-butyl-L-asparagine ester followed by removal of protecting groups then gave (3S,6S,9S)-3 or (3R,6R,9R)-3 in 7 % overall yield. (3S,6S,9S)-3: [α]=−50.7° (c=1.01, H2O); 1H NMR (500 MHz, D2O, 25 °C): δ=4.61 (1 H,dd, J=6.0, 7.0 Hz), 4.51 (1 H, d, J=8.8 Hz), 4.06 (1 H, t, J=8.5 Hz), 3.78 (1 H, m), 2.87–2.77 (2 H, m), 2.42 (1 H, m), 2.31–2.18 (3 H, m), 2.13 (1 H, m), 1.97 (1 H, m), 1.73–1.63 ppm (2 H, m). (The small differences in δ values between (3S,6S,9S)-3 and 3 derived from 2 are the result of differences in pH values and concentration. Mixing two samples resulted in overlapping peaks). 13C NMR (100 MHz, D2O + 10 %CD3OD, 25 °C): δ=175.9, 175.6, 174.0, 168.0, 60.8, 59.1, 51.1, 49.4, 37.5, 32.3, 30.2, 26.3, 25.0 ppm; IR (thin film): max=3700–2400 (br), 2956 (sh), 1670, 1530 (sh), 1448, 1390, 1340 cm −1; ESI-MS/MS: m/z: 313 [M+H]+, 181, 153, 125. (3R,6R,9R)-3: [α]=+45.7° (c=0.98, H2O); 1H NMR (500 MHz, D2O, 25 °C): δ=4.62 (1 H, dd, J=8.0, 5.0 Hz), 4.53 (1 H, d, J=8.8 Hz), 4.06 (1 H, t, J=8.5 Hz), 3.78 (1 H, m), 2.84 (1 H, dd, J=15.5, 5.0 Hz), 2.71 (1 H, dd, J=15.5, 8.0 Hz), 2.42 (1 H, m), 2.30–2.17 (3 H, m), 2.05 (1 H, m), 1.96 (1 H, m), 1.75–1.64 ppm (2 H, m); 13C NMR (100 MHz, D2O+10 % CD3OD, 25 °C): δ=175.9, 175.3, 173.9, 168.1, 60.8, 59.1, 50.9, 49.5, 37.6, 32.3, 30.6, 26.3, 25.0 ppm; IR (KBr): max=3700–2400 (br), 2955 (sh), 1671, 1528, 1447 cm−1; ESI-MS/MS: m/z: 313 [M+H]+, 181, 153, 125. HPLC analysis was performed on an Aqua C18 column (250×4.6 mm, 5 μ, 125 Å, Phenomenex) with a mobile phase of H2O(0.1 % TFA)/CH3CN(0.1 % TFA) with a gradient of 100/0 (0–5 min); 100/0 to 85/15 (5–20 min) at 1 mL min−1. (3R,6R,9R)-3 and (3S,6S,9S)-3 showed base-line separation with retention times of 15.6 and 15.9 min, respectively. 3 derived from 2 co-migrated with (3S,6S,9S)-3 but not (3R,6R,9R)-3. Supporting information for this article is available on the WWW under http://www.wiley-vch.de/contents/jc_2002/2003/z51053_s.pdf or from the author. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.