THEJOURNAL OF BIOLOGICAL

CHEMISTRY Val. 264. No . 6. Issue of February 25. pp. 3187-3193.1989

Q 1989 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in U.S.A.

Isolation, Characterization, andSynthesis of Chrysobactin, a

Compound with Siderophore Activity from Erwinia chrysanthemi*
(Received for publication, September 19, 1988)

Magnus Persmark, Dominique Expert$, and J. B. NeilandsP

From the Biochemistry Department, University of California, Berkeley, California 94720

A catechcol-type siderophore, assigned the trivial pathogenicity and theabsence of three outer membrane pro-
name chrysobactin, was isolated from the phytopath- teins. These proteinswere induced in thewild type underlow
ogenic bacterium Erwinia chrysanthemi and charac- iron conditions (Expert and Toussaint, 1985). In the same
terized by degradation and spectroscopic techniques as strain, Erwiniachrysanthemi 3937 whichcausessystemic
N-[N2-(2,3-dihydroxybenzoyl)-~-lysyl]-~-serine. disease in Saintpaulia plants, it was found recently that a
Chrysobactin, which was also obtained by chemical catechol-type, siderophore-dependent iron assimilation sys-
synthesis, was shown to be active in supplying iron to tem was requiredfor the expression of bacterial virulence
a group of mutants of E. chrysanthemi defective in throughout the plant (Enardet al., 1988).
biosynthesis of the siderophore. In the present communication we present the isolation,
characterization, synthesis, and certain biological properties
of the novel siderophore chrysobactin,which may be described
Most cells have an absolute requirement for iron. Due to as N-[N2-(2,3-dihydroxybenzoy1)-D-lysyl]-L-serine. Chryso-
the insolubility of ferric hydroxide, theelementisoften bactin has the unusual property for a catechol siderophore of
growth limitingunder aerobic conditions. Siderophores, possessing only three coordination sites. A minimum of two
highly efficient and virtually Fe3’-specific chelators, are syn- molecules is thus required to chelate hexacoordinated iron.
thesized by a majority of aerobic and facultative anaerobic Purifiedchrysobactin could supportthe growth of iron-
microorganisms under conditions of iron deprivation (Nei- starved mutantsdefective in the conversion of dihydroxyben-
lands, 1981). zoic acid (DHBA)’ to chrysobactin. The synthesis of chryso-
Ironcan also be growth limiting for microbes inhost- bactin was accomplished through traditional methods. Syn-
microbe interactions. In mammals, where most of the iron is thetic chrysobactin was chemically and biologically indistin-
eitherstoredintracellularly or bound to proteins such as guishable from the natural product.
transferrin or lactoferrin, bacterial virulence has been corre-
lated to siderophore production (Weinberg, 1984).A correla- EXPERIMENTALPROCEDURES
tion between siderophore production and virulence has also Strains and Growth Conditions-Chrysobactin was prepared from
been found in fish pathogens (Crosa, 1984). The situation in E. chrysanthemi PMV 4098, a transport-defective derivative of the
plant-microbe interactions is less clear (Neilands andLeong, wild type E. chrysantherni 3937 (Enard etai., 1988). The biosynthetic
1986). Siderophores from plant growth-promoting rhizobac- mutantsPMV 4096, PMV 4087, and PMV 4088, blocked in the
teria have been shown to increase crop yields by inhibiting biosynthesis of chrysobactin at three different stagesbetween DHBA
growth of fungal plant pathogens through chelating iron in and the final product, and the transport mutant PMV 4082 (Enard
the rhizosphere (Kloepper et al., 1980). Plant deleterious and et al., 1988) were used as indicator strains in a siderophore bioassay.
The strains were maintained in stabs at16 ”C (10 g of Difco nutrient
plant growth-promoting pseudomonads can beclassified ac- broth, 5 g of NaCI, 6 g of Difco agar/liter).
cording to their ability either to utilize or be inhibited by Materials-All glassware was washed in 6 N HCl and rinsed in
heterologous siderophores (Leong, 1986).In a survey of sev- double-distilled water prior to use. 5-Dimethylaminonaphtalene-1-
eral bacterial plant pathogens, the majoritywas found to sulfonyl chloride (dansyl chloride), N-ethylmorpholine, and polyam-
produce siderophores(LeongandNeilands, 1982). Sidero- ide sheets (Cheng Chin) were from Pierce Chemical Co. Benzalde-
hyde, D20,XAD-4, EDDA, carboxypeptidase Y, lysine decarboxylase,
phore production was in the case of Agrobacteriurn tumefa- pyridoxal phosphate, N-ethoxycarbonyl-2-ethoxy-l,2-dihydroquino-
ciens A217 shown not to be linked t o pathogenicity (Leong line (EEDQ), dansylamino acids, Ne-benzyloxycarbonyl-D-lysine (W-
and Neilands, 1981). CBZ-D-lysine), N‘CBZ-L-lysine, L-serine benzyl ester hydrochloride
Inagriculturallyimportant pectinolytic Erwinia, which and l-ethyl-3(3-dimethylaminopropyl)-carbodiimideHC1 were from
causes soft-rot diseases in a variety of plants (Collmer and Sigma. Hydrazine, DHBA, and 1,5-dlaminopentane were from Ald-
Keen, 1986),a correlation was found between the lack of rich. Sephadex was from Pharmacia LKB Biotechnology Inc., and
DEAE-trisacryl M was from LKB. Desferal was a gift from Ciba
~

* This work was supported in part by Grants A104156 from the Geigy.
National Institutes of Health, PCM78-12198 from the National Sci- Isolation of Chrysobactin-For chrysobactin production 10 ml of
ence Foundation, and CRCR-1-1633 from the United States Depart- an overnight culture of E. chrysantheni 4098 in LB was inoculated
ment of Agriculture. The costs of publication of this article were into each of 10 2800-ml Fernbach flasks containing 500 ml of MM9
defrayed in part by the payment of page charges. This article must (Schwyn and Neilands, 1987) buffered with 100 mM Tris/HCl, pH
therefore be hereby marked “aduertisement” in accordance with 18 7.4, and with 0.2% glucose as carbon source. Flasks were incubated
U.S.C. Section 1734 solely to indicate thisfact.
$ Charge de recherche from the Centre National de la Recherche The abbreviations used are: DHBA, dihydrobenzoic acid; EDDA,
Scientifique, supported in part by the Institute National de la Re- ethylenediaminedi-[(o-hydroxypheny1)aceticacid]; EEDQ, N-ethox-
cherche Agronomique. Permanent address: Laboratoire de Pathologie ycarbonyl-2-ethoxy-l,2-dihydroquinoline; CBZ, benzyloxycarbonyl;
Vbgetale, Institut National Agronomique Paris-Grignon, 75231 Paris FAB,fastatombombardment;Bis-Tris,2-[bis(2-hydroxyethyl)
Cedex 05, France. amino]-2-(hydroxymethyl)-propane-1,3-diol;MOPS,4-morpholine-
§ To whom correspondence should be addressed. propanesulfonic acid.

3187
3188 Siderophore
Chrysobactin,
a from E. chrysanthemi
onarotary shaker at 30 "C for 36 h, after which the cells were and 10 p1 was spotted on a Whatman No. 1 paper and chromato-
iemoved by centrifugation. graphed at pH 1.9 with serine and lysine as standards.
The cell-free medium was passed slowly througha5 X 60-cm Chirality-An attempt to determine the chirality of the serine
column of XAD-4, and after washingthe column with water adsorbed residue was performed by treating chrysobactin with carboxypepti-
chrysobactin was eluted with water:methanol (1:l). Fractions positive dase Y, as described previously (Hayashi, 1977). A 0.5-mg (1.4 pmol)
by the Arnow assay for catechols (Arnow, 1937) were pooled and quantity of chrysobactin was incubated with 1.5 units of carboxypep-
evaporated under reduced pressure at 30 'C. The dark-brown mate- tidase Y at pH 6.0 in 1 ml of 0.1 M pyridine/acetate buffer for 8 h at
rial, approximately 7.5 m!, was filtered through a 2.5 X 90-cm Seph- 30 "C. The lyophilized material was dissolved in 50 pl of water and
adex G-25 column equilibrated with 5 mM ammonium acetate, pH 10 yl was chromatographed at pH 1.9 as above. The remaining
5.5. Arnow-positive fractions were collected, evaporated, and rechro- material was subjected to amino acid analysis. The configuration of
matographed on G-25 as above. Fractions containing catechol mate- serine was also determined at the Amino Acid Dating Laboratory,
rial were lyophilized and dissolved in a minimal volume of water. Scripps Institution of Oceanography, University of California, San
This material was chromatographed on a Hewlett-Packard HP1090 Diego, as follows: Following hydrolysis in 6 N HCl for 24 h of 180 pg
liquid chromatograph, equipped with a UV/visible diode-array detec- (500 nmol) of chrysobactin, the sample was derivatized with o-
tor, using a 10 X 250-mm LiChrosorb RP-18 column (Merck). Sepa- PhthaldialdehydelN-acetyl-L-cysteineand analyzed by high perform-
ration was achieved at a flow rate of 5 ml/min by isocratic elution in ance liquid chromatography.
10 mM triethylammoniumacetate, pH 5.5, 3% acetonitrile, 2.5% Lysine decarboxylase was used to determine the chirality of lysine,
methanol for 5 min and by a linear gradient to 15% methanol for 15 following the method of Boeker and Fischer (1983). Two mg (5.4
min and to 85% methanol for 10 min in the same buffer. Fractions pmol) of chrysobactin was hydrolyzed in vacuo in 0.5 ml of 6 N HCI
with significant absorbance at 315 nm were collected. The chrysobac- for 24 h. The hydrolysate was extracted with 3 X 500 pl of ether to
tin-containing major peak which eluted after approximately 10 min remove DHBA, after which excess HC1 was removed under reduced
(data not shown), was concentrated and passed through a 1.5 x 60- pressure. The hydrolysate was dissolved in water and divided into
cm Sephadex G-10 column equilibrated in 5 mM ammonium acetate, two test tubes, to one of which was added 2.7 pmol of L-lysine. Both
pH 5.5, and Arnow-positive fractions were pooled and lyophilized. As samples were lyophilized and dissolved in 400 p1of0.2 M sodium
a final purification step, asolution of chrysobactin in water was acetate buffer, pH 5.7, containing 60 p~ pyridoxal phosphate, after
passed through a short column of DEAE-trisacryl M, chloride form, which 0.45 units of lysine decarboxylase was added. Samples were
equilibrated in water. Pure chrysobactin was lyophilized to a white incubated at 37 "C for 8 h and reaction products identified by paper
powder and stored over nitrogen at -20 "C. chromatography at pH 6.5 with 1,5-diaminopentane, lysine, and
Absorption Spectroscopy-Bausch and Lomb Spectronic 2000 or serine as references, and by amino acid analysis. A sample containing
Shimadzu UV 160 spectrophotometers were used for visible and 2.7 pmol each of L-serine and L-lysine was treated identically as a
ultraviolet spectrophotometry. reference.
Paper Electrophoresis-Paper electrophoresis was performed in a Chemical Synthesis-The two lysine isomers of chrysobactin were
flat bed apparatus at 1500 V. Chrysobactin and its hydrolysate were synthesized basically as described by Chimiak and Neilands (1984).
chromatographed a t p H1.9 in 8% acetic acid, 2% formic acid. Chry- One mmol (280 mg) of CBZ-L- or -D-lySine was dissolved in 20 ml of
sobactin and ferric chrysobactin were chromatographed a t pH 4.4 in tetrahydrofuran, 8 ml of HzO, and 0.6 ml of triethylamine. To this
0.4% pyridine, 0.8% acetic acid and a t p H6.5 in 10% pyridine, 0.3% was added 1 mmol (455 mg) of p-nitrophenyl esterof 2,3-dibenzylox-
acetic acid. Standards included pseudobactin (+1)and ferric pseu- ybenzoic acid (Chimiak and Neilands, 1984) in 2 ml of tetrahydrofu-
dobactin (0) (Teintze et al., 19811, ferrioxamine B (+1) (Prelog and ran. After brief heating, the solution was stirred at room temperature
Walser, 1962), ferrichrome A (-3) and its methyl esters (0 to -2), overnight and organic solvents removed by evaporation. The residue
(Emery and Neilands, 1961), and pseudobactin 589 (-1).2Spots were was acidified with 2 ml of 12 N HCl, washed with saturated NaCl
identified by UV illumination or spraying with 0.5% ninhydrin in solution, dried over MgSO,, and evaporated to give approximately 1
acetone. g of yellowish oil. The oil was dissolved in chloroform and chromat-
Amino Acid Analysis-Chrysobactin equivalent to approximately ographed on a 2 X 30-cm column with Sephadex LH-20 using chlo-
90 nmol of DHBA by the Arnow assay was hydrolyzed a t 100 'C in roform:ethyl acetate (1:l) as solvent. Fractions containing N2-(2,3-
U ~ C U Oin 6 N HC1 for 18 h. Following evaporation of acid at reduced dibenzoylbenzoic acid)-CBZ-lysine were evaporated and dissolved in
pressure, the hydrolysate was analyzed by paper chromatography or 10 ml of acetonitrile. Molar equivalentsof EEDQ, serine benzyl ester
with the amino acid analyzer operated by the Protein Chemistry and triethylamine were added and the solution stirred at room tem-
Laboratory, University of California, Davis. perature overnight. Yields were typically 80%. Coupling was checked
Oxazoline Test-Two methods were used to check the presence of by paper electrophoresis. The reaction mixture was evaporated and
an oxazoline ring. In the first (Ong et al., 1979), chrysobactin was dissolved in ethyl acetate. The organic phase was washed with 10%
dissolved in 0.1 N HCl to a concentration of 0.025 mM, and spectra citric acid, saturated ammonium carbonate and brine, after which it
from 200 to 350 nm were recorded a t regular intervals during 24 h. was dried with MgS04, filtered and evaporated. The residual, yellow-
In the second method, spectra were recorded of 0.025 mM chrysobac- ish oil was taken up in 20 ml ethanol and 1 M equivalent HCI to trap
tin in 0.1 M sodium citrate buffer at pH values of 1.2, 2.0, 2.5, 3.0, the hydrogenated product (Folsch, 1966). Hydrogenation was carried
4.0, and 4.6. out overnight with 5% palladium chloride on activated carbon. The
NMR Spectroscopy-Proton nuclear magnetic resonance (NMR) catalyst was hydrogenated immediately before use and washed in
spectra were recorded on a Bruker 500 AM and a Bruker 400 AM ethanol. After evaporation the residue was dissolved in ethyl acetate,
NMR spectrophotometer at the University of California, Berkeley, which gave a white precipitate of chrysobactin. The precipitate was
NMR Facility. The chrysobactin solution had apH of approximately taken up in 5 mM ammonium carbonate buffer, pH 5.5, and pure
3.5 before deuterium exchange. Chemical shifts were expressed rela- chrysobactin was obtained after chromatography on Sephadex G-10
tive to an internal standard of acetone, equal to 2.225 ppm. Peaks and DEAE-trisacryl M as above. The yields of pure products were
were assigned by double resonance irradiation. approximately 65%.
Mass Spectroscopy-The triethylammonium salt of chrysobactin The acid and ethyl ester of the glycylglycine analogs of chrysobac-
was subjected to fast atom bombardment mass spectroscopy (FAB) tin were synthesized according to the procedure of Ito and Neilands
at the University of California, Berkeley and at the University of (1958), with the following exceptions; Glycylglycine ethyl ester HCI
California, San Francisco, Mass Spectrometry Facilities. was neutralized with N-ethylmorpholine and dissolved in dimethyl
Determination of Amino Terminus-Approximately 10 ng of chry- formamide. As coupling agent l-ethyl-3(3-dimethyl-aminopro-
sobactin was dansylated following methods described previously py1)carbodiimide HCl was used. The free acid was purified bygel
(Gray, 1972; Weiner et al., 1972). Dansylated samples were chromat- filtration on a column with Sephadex G-10 as above.
ographed on polyamide sheets, with commercial dansyl amino acids Circular Dichroism-Circular dichroism (CD) spectra were taken
applied on the opposite side of the sheet. on the NH&l salt of chrysobactin and its L-lysine and D-lysine
Determination of Carboxyl Terminus-The carboxyl-terminal synthetic isomers at a concentration of about 0.3 mg/ml in HzO.
amino acid was determined by hydrazinolysis according to Ghuysen Analyses were performed with a JASCO 5600 Spectropolarimeter at
et al. (1966).A quantity of 1 mg of chrysobactin in 200 pl of anhydrous Triton Biosciences, Alameda, CA.
hydrazine was incubated in uacuo for 20 h at 60 "C. For identification Stoichiometry of Ferric Chrysobactin-Since ferric chrysobactin
of amino acids, the evaporated sample was dissolved in 50 p1 of H20, did not give any peak by FAB-mass spectroscopy, two other methods
were used to establish the stoichiometry of the Fe3+.chrysobactin
M. Persmark and B. Mattiason, unpublished data. complex. In the first method, 0.25 mM chrysobactin in 0.1 M Bis-
 Chrysobactin, a Siderophore
from E. chrysanthemi 3189
one, as is shown in Table I. Proton NMR of chrysobactin in
Tris, pH 6.0, was titrated with 5-rl portions of 1.0 mM FeC13 and
absorbance changes at 550 nm recorded. In addition we used Job’s90% H20, 10% D20revealed twoadditional peaks, which were
method of continuous variation to assess the formation of a single
assigned to the aNH-serine and the aNH-lysine protons,
metal chelate compound (Chaberek and Martell, 1959). The molar
respectively. The chemical shifts of most peaks corresponded
chrysobactin fraction in 10 samples of a 0.2 mM chrysobactin/Fe3+
well to literature values (Llinas et al., 1973;Bundi and Wuth-
mixture in 0.1 M MOPS buffer, pH 6.5, was varied from 0.33 to 0.88
and absorbances were measured. rich, 1976;Grog and Kalbitzer, 1988). However, it should be
Siderophore Bioassay-Biological activity of chrysobactin for one
noted that the aNH-lysine and aCH-lysine signals were sig-
transport mutant and three mutants of E. chvsanthemi 3937 defi- nificantly lower than expected. Nonequivalence of the lysine
cient in the synthesis of chrysobactin was determined in a bioassay
underlowiron conditions. Iron deficient L-broth containing 1.5% ,8 protons caused the signal to be split into two multiplets.
agar was preparedby adding 100 pg/ml filter-sterilized EDDA to the Mass spectral analysis of the triethylammonium salt of
liquid medium. The agarwas allowed to stand at 4 ”C for 48 h to chrysobactin gave an m/z peak at 370.1614 in the cationic
allow slow chelation of iron. The mediumwas then remelted and detection mode. This peak was assigned to [MH]’. In the
seeded with 10‘ or lo6 cfu/ml of indicator cells from exponential
anionic detection mode, a peak was observed a t m/z 368 and
cultures in L-broth. Sterile disks with 6-mm diameter were placed on
was assigned to [MI-. Chrysobactin therefore had a M , of
the agar surface and 10-pl portions of the following solutions were
added: chrysobactin, 0.01, 0.12, and 1.2 mM; supernatant from E.369.An m/z peak at 739,assigned to the adduct ion [MZH]’,
chvsanthemi PM 4098 grown in MM9,O.Ol and 0.1 mM; enterobactin, was also observed. Ferric chrysobactin failed to produce a
0.01 and 0.12 mM; DHBA, 0.1 and 1.0 mM; FeC13,1.0 and 10 mM; andmolecular ion (data not shown).
HzO. All concentrations refer to DHBA equivalents, as determined Chrysobactin thus consisted of 1DHBA, 1 Lys, and 1 Ser.
by the Arnow assay. Plates were incubated at 30 “C and diameters of
Furthermore, the downshifted a-lysine signals from proton
zones of growth of the indicator strains measured after24,36, and 48
h. NMR indicated that DHBA was attached to the &-nitrogen
of lysine. Evidence for the presence of a free e-amino group
RESULTS was obtained by dansylation, where the labeled product from
a chrysobactin hydrolysate comigrated with commercial N“
General Characteristics-Growth of E. chrysanthemi PM dansyllysine. A spot comigrating with serine on paper electro-
4098 in MM9 mediumwas accompanied by production of phoresis following hydrazinolysis, confirmed the presence of
Arnow-positive material, equivalent to at most 0.13 mM a free seryl carboxyl terminus and the peptide sequence as
DHBA after 24 h of incubation. Purified chrysobactin had Ne-DHBA-Lys-Ser.
spectral properties typical for catechol compounds (Ito and Chirality-Chrysobactin was not a substrate to carboxypep-
Neilands, 1958;Corbin and Bulen, 1969).At pH 6.5 in 0.1 M tidase Y. The configuration of the serine residue was therefore
MOPS buffer, absorption maxima were found at 227 and 318
determined by chemical modification, which revealed it to be
nm (emM = 3.0), with a shoulder at 247.The ferric complex, at the L-isomer.
the same pH, had absorption maxima at 228, 333 (cFe,mM =
The absolute structure of chrysobactin was established by
11.2),and 570 nm (cF~,mM = 3.8),with a shoulder at 255 nm.
the finding that lysine decarboxylase did not affect lysine in
The color of the ferric complexwashighly dependent on
a chrysobactin hydrolysate. Commercial L-lysine in a L-ly-
[H+], andalso on the ratio of iron to chelator. At high pH or
sine/L-serine mixture, as well as L-lysine added to the chry-
at a low iron to chelator ratio, the color was burgundy. With
decreasing pH or increasing iron ratio at higher pH, thecolor sobactin hydrolysate, was degraded and comigrated with 1,5-
shifted toward violet blue. diaminopentane upon paper electrophoresis. The ratioof Lys/
Chrysobactin gave a violet ninhydrin reaction, was neutral Ser, as determined by amino acid analysis, was 0.83 and 0.82
upon electrophoresis at pH 4.4 and 6.5, and showed a net for the chrysobactin hydrolysate and chrysobactin hydroly-
charge of +1, pH 1.9.Ferric chrysobactin had a net charge of sate spiked with an equimolar amount of L-lysine, respec-
-1 upon electrophoresis at pH 6.5. tively. The lysine residue therefore had D-chirality.
Amino acid analysis revealed lysine and serinewith a molar The absolute structure of chrysobactin was thus established
ratio of 1.00.94. This corresponded to roughly 1.4 DHBA as N-[N2-(2,3-dihydroxybenzoyl)-~-lysyl]-~-serine (Fig. 2).
equivalents with the Arnow test. A slightly elevated value for Synthetic chrysobactin was indistinguishable from the nat-
DHBA relative to threonine in agrobactin was found by Ong ural product by NMR (Fig. lb), UV/visible absorption spec-
et al. (1979),which was attributed to a combination of incom- troscopy, and FAB. The CD spectrum of the D-lysine isomer
plete scission and destruction duringhydrolysis. The reasons had a shape very similar to thatof natural chrysobactin, thus
may be similar in the case of chrysobactin. In order to deter- confirming the chirality (Fig. 3).
mine the number of DHBA molecules per chelate, the CAS Stoichiometry of Ferric Chrysobactin-Ferric chrysobactin
assay of Schwyn and Neilands (1987)was used. had a net charge of -1 upon electrophoresis at pH 6.5,as
One CAS equivalent was equal to one chelated Fe3+,whereas predicted fora 21 complex. Titration with Fe3+ gave an
one Arnow equivalent was defined as one DHBA. Enterobac- inflection point at 1.8DHBA residues/chelate. By varying the
tin which has 3 DHBA residues/molecule (Pollack and Nei- mole fraction of chrysobactin and Fe3+ in a solution with
lands, 1970),showed an CAS/Arnow ratio of 0.31,as expected. constant totalmolarity, absorbance maxima were obtained at
In thecase of chrysobactin 1 DHBA was found to correspond the preferred molar ratio. The obtained value of 0.69 corre-
to between only 0.05 and 0.1 CAS equivalents (datanot sponds well to a 2:l stoichiometry (Fig. 4).As mentioned, the
shown). ratio of CASto Arnow wassurprisingly low; about 0.1 instead
Since chrysobactin contained serine and DHBA the pres- of the expected 0.5.This discrepancy might be explained, as
ence of an oxazoline ring was thought possible, as inagrobac- pointed out by Schwyn and Neilands (1987),by a formation
tin (Ong et ai., 1979), parabactin (Peterson and Neilands, constant too low to allow complete exchange of iron from the
1979),and vibriobactin (Griffiths et al., 1984).Spectroscopy, CAS complex.
however, indicated an absence of an oxazoline ring (data not The glycylglycine and glycylglycine ethyl ester derivatives
shown). of DHBA were synthesized as a means to assess the involve-
StFUCtUFe of Chrysobactin-The ‘H NMR spectrum of chry- ment of the carboxyl-terminal carboxyl oxygen in the chela-
sobactin in D20is shown in Fig. la. The three aromatic tion of Fe3+.In these model compounds, the “backbone” is
protons and the lysine and serine aC protons had integrals of identical to that of chrysobactin. However, the carboxyl oxy-
3190 Chrysobactin, a Siderophore fromE. chrysanthemi

FIG. 1. Proton NMR spectrum of

chrysobactin (a)and synthetic chry-
sobactin ( b ) in DtO ( 8 ) . Chemical
shifts are referred to an internal stand-
ard of acetone ( r ) and arelisted in Table
I.

TABLEI
Proton NMR chemical shifts and coupling constants of chrysobactin in DzO
Resonance 6 Coupling constants IJ I Multiplicity" Integral
ppm Hz
Benzoyl
O-CH 7.30 J,, = 1.4, J,, = 8.1 1.0
m-CH 6.89 J,, = 8.0, J,, = 8.0 1.0
p-CH 7.10 Jp,,= 1.4, Jpm= 7.9 0.9
LYSY1
a-NH 8.75
a-CH 4.65 J , ~ I= 5.4, J o p z = 8.8 1.0
@-CHz 1.91 1.1
2.01 1.2
Y-CK 1.53 2.2
6-CHz 1.73 2.3
c-CH~ 3.00 J,a = 7.5 2.0
Seryl
a-NH 8.14 d
(u-CH 4.37 Jmo= 4.8 t 1.o
KHz 3.87 JBa = 4.8 d 2.1
Abbreviations: d, doublet; t, triplet; q, quartet; m, multiplet.

gen of the esterwould not likely function as an iron chelatingsixth ligands are supplied by the carboxyl oxygens or water.
ligand. Chrysobactin and the freeacid glycine analog had Biological Activity-Purified chrysobactin,togetherwith
identical spectral properties, as well as identical Job's plots. other iron sources, was investigated for its qualitative ability
However, the ethyl ester was more similar to chrysobactin to enhance growth of non-producing mutants. The resultsof
than to DHBA (data not shown). At high ligandto iron ratios, the bioassay, summarized in Table11, showed that both chry-
3:l or higher, or at high pH at ratios of ligand to iron at2:l sobactin and ferric chrysobactin could reverse iron starvation
or lower, the complex had a red color similar to that of of biosynthetic mutants, but not of the transport mutant
enterobactin. It is not clear if the red color signifies a 3:l lacking the putative chrysobactin receptor. This indicated
complex and, in the case of a 21 complex, if the fifth and that the purified compound is actually a biologically active
 an L-0
I
Siderophore
Chrysobactin,

OH
a from E. chrysanthemi
TABLEI1
Stimulation of growth of biosynthetic and transport mutants
by iron source
3191

Figures refer to the diameterof the growth zones surrounding the

-.
disks. no growth; k, 4 2 mm; +, 12-19 mm; ++, 20-30 mm; +++,
>30 mm; ND, not done.
Iron source‘ PMV 4086 PMV 4087 PMV 4088 PMV 4082
InM
Chrysohactin
0.01 t - - -
0.12 + +
- -4- -
1.2 + + ++ -

:
Ferric chrysobactin
60.000 1.0 ++ + f+ -
Supernatant
0.01 - - + -e
0.1 +. + ++ +
Enterobactin
0.01 ++ ++ ++ ++
0.12 +++ +++ +++ +++
DHBA
0.1 - - - ND
1.0 - t * ND
FeC13
1.o ++ ++ ++ ND
10 +++ +++ +++ ND
Water - - - -
220
180 200 240
280 260 300 For further details, see “Experimental Procedures.”
Wavelength (nm)
FIG. 3. CD spectrum of chrysobactin (a),synthetic chryso- growth also of the receptor mutant, it may be assumed that
bactin ( b ) and the synthetic L-Lys-t-Ser isomer (c). deg., de- iron is taken up through a second, not yet determined path-
grees. way.
Finally, the exogenous siderophore enterobactin was highly
efficient in promoting growth of all mutants. This suggested
that ferric enterobactin probably not was taken up through
the chrysobactin outer membranereceptor. It should be men-
tioned that in the concentration range used, the enterobactin
preparation stimulated the growth of an Escherichia coli ent
A mutant, but not a fep A mutant lacking the enterobactin
outer membrane receptor (data notshown).
In summary, the proposed structure of chrysobactin, which
had a molecular formula of C16H24N307and anaccompanying
calculated mass spectral molecular weight of370.1614, was
consistent with its ‘H NMR spectrum and its observed spec-
tral molecular weight of 370.1609. Biological activity of chry-
sobactin was shown by its ability to support growth of iron-
0 . 0 0 ~ ~ ” ’ ~ ‘ ~ ” ” ’
starved E. chrysanthemimutants. Finalproof for the proposed
0.40 0.50 0.60 0.70 0.00 0.90 structure was obtained by synthesis. Synthetic chrysobactin
Chrysobactin (Molefraction) was identical to the natural product, as determined by UV/
visible absorption spectroscopy, FAB, NMR, and CD. Growth
FIG. 4. Job’s plot of absorbance at 570 nm (0°C)) and 620 of iron-starved mutants was supported by chrysobactin and
nm (U ) continuous variation of the mole fraction
versus
chrysobactin in a solution of constant total [chrysobactin + its L-lysine analog.
FeS+].The highest absorption intensity corresponds
to the maximum
concentration of ferric chelate. DISCUSSION
In this study, we have purified, characterized, and synthe-
siderophore, which is specifically taken up by the cell when sized chrysobactin, anovel siderophore from E. chrysanthemi.
chelated to ferric iron. In these preliminary experiments ac- Chrysobactin, with the structure N-[N2-(2,3-dihydroxyben-
tivity was, unexpectedly, associated also with the synthetic L- zoy1)-D-lysyl]-L-serine,can supply at most three coordination
lysine analog. Growth stimulation, characterized by distinct sites for iron. The presence of D-lysine is notable, although
colonies surrounding the disks, was also observed with Fe3+ D-amino acids are commonly found with peptide siderophores
at levels high enough to saturate the EDDA. Material from from, for example, Pseudomonas (Teintze et al., 1981). The
two of the major Arnow-positive peaks separated from chry- only siderophores with structures resembling that of chryso-
sobactin by high performance liquid chromatography did not, bactin of which chirality has been determined is azotochelin
however, support growth of the mutants (data not shown). (Corbin and Bulen, 1969) and 2,3-dihydroxy-N-benzoylserine
These fractions might represent degradation products. (O’Brien et al., 1969). Both amino acids were determined to
In the same concentration range, the purified compound have the L-configuration.
appeared slightly less active than crude supernatant fluid of The ability of the L-Lys-L-Ser isomer to support growth of
a low iron-induced culture of E. chrysanthemi PMV 4098. the iron-starved mutants was unexpected and remains unex-
Since this culture supernatant could significantly enhance the plained. O’Brien et al. (1970) found that the isomers of 2.3-
3192 Chrysobactin, a Siderophore from E. chrysanthemi
TABLE 111
Linear amino acid- or peptide-substituted DHBA siderophores
Substituent Source Trivial name Reference
GlY subtilis Bacillus Ito and Neilands (1958)
Rhizobium RA1
Modi et al. (1985)
Ser E. coli Brot et al. (1966), O’Brien et al. (1969)
Klebsiella oxytoca Korth (1970)
Aerobacter aerogenes O’Brien et al. (1969)
(Klebsiella pneumo-
niae)
Thr K. oxytoca Korth (1970)
Rhizobium RA1
Modi et al. (1985)
Putrescine Azotobacter
vinelandii Aminochelin
andPage von Tigerstrom (1988)
Lys (Bis) A . vinelandii andAzotochelin
Corbin Bulen (1969)
Om, Ser Azospirillum brasilense Spirilobactin
Bachawat and Ghosh (1987)
a-Lys-Ser E. chrysanthemi Chrysobactin This study
Lys, Leu lipoferum
Azospirillum Saxena et al. (1986)
Lys, Gly, Phe Aeromonns
hydrophila Amonabactin P Byers (1987)
LYS,G ~ YTrp
, A . hydrophila Amonabactin T Byers (1987)

dihydroxy-N-benzoylserinewere equally active in supporting Pseudomonads show promiscuity in their ability to utilize
growth of E. coli. However, in this case DHBAwasmore structurally related pseudobactin-type siderophores from
active and ittherefore can not be ruled out that 2,3-dihydroxy- other Pseudomonas strains (Leong, 1986). Griffiths et al.
N-benzoylserine could have served as a source of DHBA for (1984)showed that both vibriobactin and agrobactin sup-
enterobactin synthesis, and not as asiderophore per se. ported growth of iron starved Vibrio cholerae Loul5. The
In hexacoordinated siderophores of both catechol and hy- similarity in general structural features, as well as aroundthe
droxamate type, it has been shown that the iron center is iron center in the case of both pseudobactins and vibriobactinl
specifically recognized by the outer membrane receptor. A agrobactin, makes the use of a common receptor conceivable.
change in amino acid chirality may lead to a change in iron However, chrysobactin and enterobactin are structurally, as
center configuration and a biologically inactive siderophore well as stoichiometrically, quite distinct and would therefore
(Neilands et al., 1981; Huschka et al., 1985). It is not known require separate receptors.
whether this is also the case for non-hexadentate sidero- Despite its unfavorable stoichiometry, chrysobactin was
phores, such as chrysobactin, or if other parts of the chelate biologically active, as shown by its ability to support growth
also play an important role in receptor recognition. The of iron-starved E. chrysanthemi mutants deficient in chryso-
question is presently under investigation. bactin synthesis. Likewise, biologicalactivity with linear, non-
E. chrysanthemi 3937 might produce additional sidero- hexacoordinated catechol-containing siderophores has been
phore(s), since thesupernatant of minimal mediumwas shown with a number of iron-induced monoamino acid cate-
slightly more effective than purified chrysobactin in support- chols or decarboxylation products, isolated from several mi-
ing growth of the three mutants defective in chrysobactin crobial sources (Table 111). It has been suggested (O’Brien et
synthesis. More importantly, in contrast to chrysobactin it al., 1970) that 2,3-dihydroxy-N-benzoylserinewas present in
also supported growth of a transport mutant. It appears that the supernatant only as a degradation product of biologically
this putative siderophore belongs neither to the catechol nor active enterobactin. The number of simple, biologically active
the hydroxamate class of siderophores. The two major frac- DHBA derivatives described (Table 111)appears, however, to
tions, separated from chrysobactin by high performance liquid indicate that these compounds are in fact functioning as
chromatography, that were both Arnow-positive and had ab- siderophores. Furthermore, no enterobactin-like trimer of the
sorption at 318 nm, failed to support growth of the biosyn- di- and tripeptide DHBA derivatives has been isolated. With
thetic mutants. Apart from those three fractions, virtually no the exception of azotochelin (Pageand Huyer, 1984), the
catechol material was present in the culture supernatant. structure of the monoamino acid compounds would make a
Furthermore, it was found (Enard et al., 1988) that no hy- 3:l siderophore-iron complex likely. No studies on the stoi-
droxamate positive material could be identified by the Csiky
chiometry have, however, been undertaken. Siderophore-iron
test (Csiky, 1948). complexes that do not show 1:l stoichiometry have been
E. chrysanthemi could utilize enterobactin, and it is con-
described previously. Rhodotorulic acid, a hydroxamate sid-
ceivable that one of the three membrane proteins induced
erophore from Rhodotorukz pilimanue, was shown by Carrano
under low iron conditions (Enard et al., 1988) might be a
and Raymond (1978) to form an Fe2RA3 complexat physio-
ferric-enterobactin receptor. This would thus be a second
logical pH. In chrysobactin the o-hydroxyl and carboxyl-
receptor beside that for ferric chrysobactin. E. chrysanthemi
terminal oxygens are separated by nine atoms, a distance
3937 has furthermore been shown to utilize ferrichrome? It
appears unlikely, however, that E. chrysanthemi 3937 pro- theoretically large enough to form a tridentate ligand and
duces enterobacitin, since no Arnow-positive material could thus to allow the chelation of one hexacoordinated ferric ion
be extracted into ethyl acetate from acidified minimal me- by two chrysobactin molecules. This assumption was experi-
dium, And as mentioned above, virtually all Arnow-positive mentally supported. However, the synthesis of 2,3-dihydrox-
material was present inthe form of either chrysobactin or the ybenzoyl glycylglycine and its ethyl ester did not unequivo-
two putative degradation products. The ability to utilize het- cally show whether or not the carboxyl oxygens are involved
erologous siderophores is a relatively common phenomenon in iron chelation.
among bacteria. Several enteric bacteria have receptors for Chrysobactin might be the first characterized siderophore
ferrichrome and other fungal siderophores (Neilands, 1982). from a plantpathogen, where iron has been directly correlated
to virulence. Its stoichiometry and structure areunusual, but
D. Expert, unpublished data. similar siderophores have been isolated. It thus appears that
 Chrysobactin, a Siderophore from E. chrysanthemi 3193
of a class of siderophores which
chrysobactin is representative Griffiths, G. L., Sigel, S. P., Payne, S. M., and Neilands, J. B. (1984)
are DHBA derivatives of amino acidsor linear peptides. J. Biol. Chem. 259,383-385
GroR, K.-H., and Kalbitzer, H. R. (1988)J. Magn. Reson. 76,87-99
Hayashi, R. (1977)Methods Enzymol. 25.84-93
Acknowledgments-We thank Drs. E. Alvarado, J. L. Bada, P. Huschka, H., Naegeli, H. U., Leuenberger-Ryf, H., Keller-Schierlein,
Hoeprich, and A. Smith for experimental assistance. We also ac- W., and Winkelmann, G. (1985)J. Bacteriol. 162,715-721
knowledge the staffs at the University of California, Berkeley, Mass Ito, T., and Neilands, J. B. (1958)J.Am. Chem. SOC. 80, 4645-4647
Spectrometry Facility and the University of California, San Fran- Kloepper, J. W., Leong, J., Teintze, M., and Schroth, M. N. (1980)
cisco, Bio-Organic Mass Spectrometry Resource (A. L. Burlingame, Nature 286,885-886
Director) for technical assistance. The latter was supported by Na- Korth, H. (1970)Arch. Mikrobiol. 70, 297-302
tional Institutes of Health Division of Research Resources Grant Leong, J. (1986)Annu. Reo. Plant Pathol.24,187-209
RR01614. Leong, S. A., and Neilands, J. B.’(1981)J. Bacteriol. 147, 482-491
Leong, S. A., and Neilands, J. B. (1982)Arch.Bwchem.Biophys.
REFERENCES 218,351-359
Llinis. M.. Wilson. D. M.. and Neilands. J. B. (1973)
. . Biochemistry
Arnow, L. E. (1937)J . B i d . Chem. 118,531-537 12,3836-3843 ‘
Bachawat, A. K., and Ghosh, S. (1987)J . Gen. Microbiol. 133,1759- Modi. M.. Shah. K. S.. and Modi. V. V. (1985)
1765 . , Arch. Mikrobiol. 141.
156-158
Boeker, E. A., and Fischer, E. H. (1983)Methods Enzymol. 94, 180- Neilands, J. B. (1981)Annu. Reo. Nutr. 1,27-46
184 Neilands, J. B. (1982)Annu. Reu. Microbiol. 36,285-309
Brot, N., Goodwin, J.., and Fales, H.(1966)Biochem. Biophys. Res. Neilands, J. B., and Leong, S. A. (1986)Ann. Reo. Plunt Physiol. 37,
Commun. 25,454-461 187-208
Bundi, A., and Wuthrich, K. (1979)Biopolymers 18,285-297 Neilands, J. B., Erickson, T. J., and Rastetter, W. H. (1981)J. Biol.
Byers, B. R. (1987)in Iron Transport in Microbes, Plants and Animals Chem. 256,3831-3832
(Winkelmann, G., van der Helm, D., and Neilands, J. B., eds) pp. O’Brien, I. G., Cox, G. B., and Gibson, F. (1969)Biochim. Biophys.
111-116,VCH Verlagsgesellschaft, Weinheim Acta 177,321-328
Carrano, C. J., and Raymond, K. N. (1978)J. Am. Chem. SOC.100, O’Brien, I. G., Cox, G. B., and Gibson, F. (1970)Biochim. Biophys.
5371-5374 Acta 201,453-460
Chaberek, S., and Martell, A. E. (1959)Organic SequesteringAgents, Ong, S. A., Peterson, T., and Neilands, J. B. (1979)J. Biol. Chem.
pp. 78-81,John Wiley & Sons, Inc., New York 264,1860-1865
Chimiak, A., and Neilands, J. B. (1984)Structure Bonding 58, 89- Page, W. J., and Huyer, M. (1984)J. Bacteriol. 158,496-502
96 Page, W. J., and von Tigerstrom, M. (1988)J. Gen. Microbiol. 134,
Collmer, A., and Keen, N. T. (1986)Annu. Reo. Phytopathol. 24, 453-460
383-409 Peterson, T., and Neilands, J. B. (1979)Tetrahedron Lett. 50, 4805-
Corbin, J. L., and Bulen, W. A. (1969)Biochemistry 8, 757-762 4808
Crosa, J. H. (1984)Annu. Reu. Microbiol. 38,68-89 Pollack, J. R., and Neilands, J. B. (1970)Biochem. Biophys.Res.
Csaky, T. (1948)Acta Chem. Scand.2,450-454 Comm. 38,989-992
Emery, T.,and Neilands, J. B. (1961)J. Am. Chem. SOC.83, 1626- Prelog, V., and Walser, A. (1962)Helo. Chim. Acta45, 631-637
1628 Saxena, B., Modi, M., and Modi, V.V. (1986)J. Gen. Microbiol. 132,
Enard, C., Diolez, A., and Expert, D. (1988)J. Bacteriol. 170,2419- 2219-2224
2426 Schwyn, B., and Neilands, J. B. (1987)Anal. Biochem. 160,47-56
Expert, D., and Toussaint, A. (1985)J. Bacterwl. 163, 221-227 Teintze, M., Hossain, M. B., Barnes, C.L., Leong, J., and van der
Folsch, G. (1966)Acta Chem. Scand.20,459-473 Helm, D. (1981)Biochemistry 20,6446-6457
Ghuysen, J.-M., Tipper, D. J., and Strominger, J. L. (1966)Methods Weinberg, E. D. (1984)Physiol. Reo. 64,65-102
Enzymol. 8,685-699 Weiner, A. M., Platt, T., and Weber, K. (1972)J. Bioi. Chem. 247,
Gray, W. (1972)Methods Enzymol. 25, 121-138 3242-3251