High-yield synthesis of the enterobactin trilactone and evaluation of derivative siderophore analogs

Article abstract of DOI:10.1021/ja970718n

A novel one-step synthesis of the macrocyclic triserine trilactone scaffold of the siderophore enterobactin, which eliminates the β-lactonization step of N-tritylserine, is presented. The cyclization reaction is based on a stannoxane template and leads to an overall yield of ~ 50%. This enables the practical functionalization of the trilactone by attaching chelating groups other than catecholamides. The conformational stability of the trilactone ring has been examined by high-resolution X-ray diffraction studies of the N-trityl intermediate: crystals grown from methylene chloride:methanol are orthorhombic, space group P2₁2₁2₁ With unit cell dimensions a = 9.2495(5) A?, b = 11.3584(1) A?, c = 48.945(1) A?, V = 5142.1(2) A?³, and Z = 4. A hydroxypyridinonate analog of enterobactin, N,N',N''-tris[(3-hydroxy-1-methyl-2-oxo-(1H-pyridinyl)carbonyl]-4-cyclotriseryl trilactone (hopobactin), has been prepared by attachment of three 3-hydroxy-1-methyl-2(1H)-pyridinonate (3,2-HOPO) moieties to the triserine trilactone. This ligand represents the first enterobactin analog that retains the trilactone scaffold, but employs chelates other than catecholamides. Crystals of the chiral ferric complex grown from DMF:diethyl ether are monoclinic, space group P2₁, with unit cell dimensions a 13.0366(9) A?, b = 22.632(2) A?, c = 27.130(2) A?, b = 100.926(1)(o), V = 7860(1) A?³, and Z = 8. The Δ configuration of enterobactin metal complexes is also enforced in those of hopobactin and persists in aqueous or methanolic solution, as demonstrated by circular dichroism. The ferric hopobactin complex is the first reported chiral complex of hydroxypyridinonate ligands. The solution coordination chemistry of this new ligand and its iron(III) and iron(II) complexes have been studied by means of ¹H NMR, potentiometric, spectrophotometric, and voltammetric methods. The average protonation constant of the hopobactin free ligand (log K(av) = 6.1) is typical of other 3-hydroxy-1-methyl-2-oxo-1H-pyridin-4-carboxamide ligands. The stability constants of the iron(III) complex formed with hopobactin (log β₁₁₀ = 26.4) and with the tris(2-aminoethyl)amine-based analog, TRENHOPO, (log β₁₁₀ to = 26.7) are of the same order of magnitude, unlike the catecholamide-based species, where enterobactin (log β₁₁₀ = 49) is 6 orders of magnitude more stable than TRENCAM (log β₁₁₀ = 43.6). The stability enhancement reflects the specific predisposition by the triserine scaffold of the catecholamide binding units. In spite of a significantly lower affinity of 3,2-hydroxypyridinonates for iron(III) compared with the more basic catecholates, hopobactin is an extraordinarily powerful chelating agent under acidic conditions: No measurable dissociation is observed even in 1.0 M HCl. In contrast to enterobactin and its synthetic derivatives, the hopobactin ferric complex undergoes no sequential protonation above pH 1. The affinity of hopobactin and TRENHOPO for iron(III) relative to iron(II) results in strongly negative reduction potentials, -782 mV vs 0.01 M Ag⁺/Ag in CH₃CN or -342 mV vs NHE in water and -875 mV vs 0.01 M Ag⁺/Ag in CH₃CN or -435 mV vs NHE in water, respectively.

Full text of DOI:10.1021/ja970718n

J. Am. Chem. Soc. 1997, 119, 10093-10103
10093
High-Yield Synthesis of the Enterobactin Trilactone and
Evaluation of Derivative Siderophore Analogs1
Michel Meyer,† Jason R. Telford, Seth M. Cohen, David J. White, Jide Xu, and
Kenneth N. Raymond*
Contribution from the Department of Chemistry, UniVersity of California,
Berkeley, California 94720
ReceiVed March 6, 1997X
Abstract: A novel one-step synthesis of the macrocyclic triserine trilactone scaffold of the siderophore enterobactin,
which eliminates the â-lactonization step of N-tritylserine, is presented. The cyclization reaction is based on a
stannoxane template and leads to an overall yield of ∼50%. This enables the practical functionalization of the
trilactone by attaching chelating groups other than catecholamides. The conformational stability of the trilactone
ring has been examined by high-resolution X-ray diffraction studies of the N-trityl intermediate: crystals grown
from methylene chloride:methanol are orthorhombic, space group P212121 with unit cell dimensions a ) 9.2495(5)
Å, b ) 11.3584(1) Å, c ) 48.945(1) Å, V ) 5142.1(2) Å3, and Z ) 4. A hydroxypyridinonate analog of enterobactin,
N,N′,N′′-tris[(3-hydroxy-1-methyl-2-oxo-(1H-pyridinyl)carbonyl]-4-cyclotriseryl trilactone (hopobactin), has been
prepared by attachment of three 3-hydroxy-1-methyl-2(1H)-pyridinonate (3,2-HOPO) moieties to the triserine trilactone.
This ligand represents the first enterobactin analog that retains the trilactone scaffold, but employs chelates other
than catecholamides. Crystals of the chiral ferric complex grown from DMF:diethyl ether are monoclinic, space
group P21, with unit cell dimensions a ) 13.0366(9) Å, b ) 22.632(2) Å, c ) 27.130(2) Å, b ) 100.926(1)°, V )
7860(1) Å3, and Z ) 8. The ∆ configuration of enterobactin metal complexes is also enforced in those of hopobactin
and persists in aqueous or methanolic solution, as demonstrated by circular dichroism. The ferric hopobactin complex
is the first reported chiral complex of hydroxypyridinonate ligands. The solution coordination chemistry of this new
ligand and its iron(III) and iron(II) complexes have been studied by means of 1H NMR, potentiometric,
spectrophotometric, and voltammetric methods. The average protonation constant of the hopobactin free ligand (log
Kav ) 6.1) is typical of other 3-hydroxy-1-methyl-2-oxo-1H-pyridin-4-carboxamide ligands. The stability constants
of the iron(III) complex formed with hopobactin (log â110 ) 26.4) and with the tris(2-aminoethyl)amine-based analog,
TRENHOPO, (log â110 ) 26.7) are of the same order of magnitude, unlike the catecholamide-based species, where
enterobactin (log â110 ) 49) is 6 orders of magnitude more stable than TRENCAM (log â110 ) 43.6). The stability
enhancement reflects the specific predisposition by the triserine scaffold of the catecholamide binding units. In
spite of a significantly lower affinity of 3,2-hydroxypyridinonates for iron(III) compared with the more basic
catecholates, hopobactin is an extraordinarily powerful chelating agent under acidic conditions: No measurable
dissociation is observed even in 1.0 M HCl. In contrast to enterobactin and its synthetic derivatives, the hopobactin
ferric complex undergoes no sequential protonation above pH 1. The affinity of hopobactin and TRENHOPO for
iron(III) relative to iron(II) results in strongly negative reduction potentials, -782 mV vs 0.01 M Ag+/Ag in CH3CN
or -342 mV vs NHE in water and -875 mV vs 0.01 M Ag+/Ag in CH3CN or -435 mV vs NHE in water, respectively.
Introduction
negative bacteria) and recognition of the iron-siderophore
complex based on the chirality of the octahedral metal center.6,7
Because the growth of pathogenic bacteria is dependent on iron
availability, iron metabolism has increasingly been recognized
as significant to medicine as a determinant in bacterial disease.8-13
Isolated over 20 years ago,14,15 enterobactin is one of the best
characterized siderophores with respect to the mechanism and
genetics of its cellular transport and production.16-18 Entero-
Of the nutrients required by bacteria and fungi for growth,
one of the most difficult to obtain is iron. The hydrolysis of
Fe(III) limits its concentration at neutral pH to about 10-18 M,
requiring organisms to develop some form of complexation and
active transport. To this end microbes (and in some cases
plants) produce low-molecular weight chelating agents,
siderophores.2-5 Novel features of siderophore-mediated iron
transport by aerobic bacteria involve energy-dependent transport
across cell membranes (including the outer membrane of Gram-
(6) Mu¨ller, G.; Matzanke, B. F.; Raymond, K. N. J. Bacteriol. 1984,
160, 313.
(7) Matzanke, B. F.; Mu¨ller-Matzanke, G.; Raymond, K. N. Siderophore
Mediated Iron Transport. In Physical Bioinorganic Chemistry Series; Loehr,
T. M., Ed.; VCH: New York, 1989; p 1.
(8) Holbein, B. E.; Jericho, K. W. F.; Likes, G. C. Infect. Immun. 1979,
24, 545.
* Author to whom correspondence should be addressed.
†Present address: Laboratoire d’Inge´nierie Mole´culaire pour la Se´paration
et les Applications des Gaz (LIMSAG), UMR 5633, Universite´ de
Bourgogne, Dijon, France.
X Abstract published in AdVance ACS Abstracts, October 1, 1997.
(1) Coordination Chemistry of Microbial Iron Transport. 60. For the
previous paper in this series, see: Kersting, B.; Meyer, M.; Powers, R. E.;
Raymond, K. N. J. Am. Chem. Soc. 1996, 118, 7221.
(9) Bullen, J. J.; Leigh, L. C.; Rogers, H. J. Immunology 1968, 15, 581.
(10) Expert, D.; Gill, P. R. Iron: A Modulator in Bacterial Virulence
and Symbiotic Nitrogen-Fixation; CRC Press: Boca Raton, 1992.
(11) Mart´ınez, J. L.; Delgado-Iribarren, A.; Baquero, F. FEMS Microbiol.
ReV. 1990, 75, 45.
(2) Winkelmann, G. CRC Handbook of Microbial Iron Chelates; CRC
Press: Boca Raton, 1991.
(3) Winkelmann, G. Mycol. Res. 1992, 96, 529.
(12) Tai, S.-P. S.; Krafft, A. E.; Nootheti, P.; Holmes, R. K. Microb.
Pathogen. 1990, 9, 267.
(4) Neilands, J. B. J. Biol. Chem. 1995, 270, 26723.
(13) Griffiths, E. Biol. Met. 1991, 4, 7.
(5) Telford, J. R.; Raymond, K. N. Siderophores. In ComprehensiVe
Supramolecular Chemistry; Atwood, J. L., Davies, J. E. D., MacNicol, D.
D., Vo¨gtle, F., Eds.; Elsevier Science Ltd.: Oxford, 1996; Vol. 1; p 245.
(14) O’Brien, J. G.; Gibson, F. Biochim. Biophys. Acta 1970, 215, 393.
(15) Pollack, J. R.; Neilands, J. B. Biochem. Biophys. Res. Commun.
1970, 38, 989.
S0002-7863(97)00718-X CCC: $14.00 © 1997 American Chemical Society

10094 J. Am. Chem. Soc., Vol. 119, No. 42, 1997
Meyer et al.
bactin forms a remarkably stable complex with iron19 and is
the primary siderophore of enteric bacteria. (Although a
triphenolate derivative of 1,4,7 triazacyclononane has been
reported to have a higher stability constant,20a this ligand is
insoluble in water. The results in 75% ethanol/25% H2O
solution cannot be directly compared with aqueous solution.
However a related hydroxypyridyl derivative is reported to bind
Fe3+ more strongly than enterobactin at neutral pH20b.) These
features have prompted a range of studies of enterobactin’s
coordination chemistry and the structural origins of its unusual
stability.21-28 Synthetic enterobactin mimics have been hexa-
dentate triscatecholamide ligands coupled to a variety of triamine
scaffolds including carbocycles,29,30 tripodal31,32 or cyclic32,33
polyamines, mesityl,32-35 and scyllo-inositol36 derivatives. Most
of these analogs form Fe(III) complexes about 106 less stable
than that of enterobactin.
Recently, the origin of this enhanced stability has been
attributed to the positioning of the ligand groups by the trilactone
ring of enterobactin.35,37 This rigid scaffold imposes a ∆
stereochemistry on the high-spin Fe(III) metal center.23 How-
ever, the difficulty of synthesizing the 12-membered trilactone
ring has been a major barrier to the preparation of enterobactin
analogs based on the same trilactone scaffold. The first
syntheses gave overall yields from multistep reactions of only
about 1%.38-40 The organotin template strategy developed by
Shanzer et al. increased the yield to about 6%.41 Recently,
Gutierrez et al. improved the multistep Shanzer synthesis to give
practical yields.42
Figure 1. The natural siderophore enterobactin (R1 ) catechol) and
its synthetic 3,2-hydroxypyridinonate analog hopobactin (R2 ) 3,2-
HOPO).
enterobactin that retain the trilactone ring. For the first time,
the predisposition ability of the enterobactin scaffold has been
gauged with chelates other than catecholamides. To this end,
a novel ligand, N,N′,N′′-tris[(3-hydroxy-1-methyl-2-oxo-1H-
pyridin-4-yl)carbonyl]cyclotriseryl trilactone, named here hopo-
bactin (Figure 1), has been synthesized by attaching 3-hydroxy-
1-methyl-2(1H)-pyridinonate (3,2-HOPO) binding groups to the
trilactone triamine. The X-ray structure of ferric hopobactin
along with the spectroscopic and thermodynamic properties of
this novel ligand and its iron(III) complex have been fully
investigated.
Experimental Section
This paper presents a novel, one-step synthesis of the
enterobactin scaffold in yields as high as 50%. This high-yield
synthesis enables the preparation of new synthetic analogs of
General. Unless otherwise noted, starting materials were obtained
from commercial suppliers and used without further purification.
Solvents were distilled prior to use. HCl in dry ethanol was prepared
by passing dry HCl gas through absolute ethanol, previously distilled
over CaH2. The concentration was estimated by weight difference and
measured by titrating an aliquot (1.0-2.0 mL) of the solution with
standardized base. Flash silica gel chromatography was performed
using Merck 40-70 mesh silica gel. Radial chromatography was
performed on a Chromatotron (Harrison Research) with silica-coated
disks. Microanalyses were performed by the Microanalytical Services
Laboratory, College of Chemistry, University of California, Berkeley,
CA. Mass spectra were recorded at the Mass Spectrometry Laboratory,
College of Chemistry, University of California, Berkeley, CA. 1H NMR
spectra were recorded on either AMX 300, AMX 400, or AM 500
Bruker superconducting Fourier-transform spectrometers operating at
300, 400, and 500 MHz, respectively. Infrared spectra were measured
using a Nicolet Magna IR 550 Fourier-transform spectrometer. Melting
points were taken on a Bu¨chi melting apparatus and are uncorrected.
Circular dichroism spectra were recorded using a quartz cell of 1 cm
optical path length (Hellma, Suprasil) on a Jasco J500C spectrometer
which was equipped with an IF-500 II A/D converter and controlled
by a microcomputer.
(16) Ecker, D. J.; Matzanke, B. F.; Raymond, K. N. J. Bacteriol. 1986,
167, 666.
(17) Raymond, K. N.; Cass, M. E.; Evans, S. L. Pure Appl. Chem. 1987,
59, 771.
(18) Poole, K.; Young, L.; Neshat, S. J. Bacteriol. 1990, 172, 6991.
(19) Loomis, L. D.; Raymond, K. N. Inorg. Chem. 1991, 30, 906.
(20) (a) Clarke, E. T.; Martell, A. E. Inorg. Chim. Acta 1991, 186, 103.
(b) Motekaitis, R. J.; Sun, Y.; Martell, A. E. Inorg. Chim. Acta 1992, 198,
421.
(21) Isied, S. S.; Kuo, G.; Raymond, K. N. J. Am. Chem. Soc. 1976, 98,
1763.
(22) Salama, S.; Stong, J. D.; Neilands, J. B.; Spiro, T. G. Biochemistry
1978, 17, 3781.
(23) McArdle, J. V.; Sofen, S. R.; Cooper, S. R.; Raymond, K. N. Inorg.
Chem. 1978, 17, 3075.
(24) Harris, W. R.; Carrano, C. J.; Raymond, K. N. J. Am. Chem. Soc.
1979, 101, 2722.
(25) Lee, C.-W.; Ecker, D. J.; Raymond, K. N. J. Am. Chem. Soc. 1985,
107, 6920.
(26) Scarrow, R. C.; Ecker, D. J.; Ng, C.; Liu, S.; Raymond, K. N. Inorg.
Chem. 1991, 30, 900.
(27) Tor, Y.; Libman, J.; Shanzer, A.; Felder, C. E.; Lifson, S. J. Am.
Chem. Soc. 1992, 114, 6661.
Syntheses. 1,1,6,6-Tetra-n-butyl-1,6-distanna-2,5,7,10-tetraox-
acyclodecane, Stannoxane (1). Dibutyltin oxide (100 g, 0.4 mol),
ethylene glycol (25 g, 0.4 mol), and benzene (1 L) were added to a
flask fitted with a condenser and a Dean-Stark trap. The reaction
mixture was refluxed overnight, collecting about 5 mL of liquid in the
trap. The solution was cooled to 4 °C whereupon white crystals formed.
These were collected by filtration, washed with benzene, and dried.
Yield: 90%. Mp: 222-224 °C (lit. mp 223-227 °C).43 FT IR (thin
film cast from CHCl3): ν 2900, 1420, 1050, 910, 590 cm-1 (no evidence
for hydroxyl absorption). 1H NMR (500 MHz, CDCl3): δ 0.87 (t, J
) 2.92 Hz, 12H), 1.25 (br t, J ) 3.2 Hz, 8H), 1.34 (m, J ) 2.94 Hz,
J′ ) 2.92 Hz, 8H), 1.6 (m, J ) 3.2 Hz, J′ ) 2.94 Hz, 8H), 3.57 (s,
(28) Karpishin, T. B.; Raymond, K. N. Angew. Chem., Int. Ed. Engl.
1992, 31, 466.
(29) Corey, E. J.; Hurt, S. D. Tetrahedron Lett. 1977, 45, 3923.
(30) Tse, B.; Kishi, Y. J. Org. Chem. 1994, 59, 7807.
(31) Rodgers, S. J.; Lee, C.-W.; Ng, C. Y.; Raymond, K. N. Inorg. Chem.
1987, 26, 1622.
(32) Harris, W. R.; Raymond, K. N.; Weitl, F. L. J. Am. Chem. Soc.
1981, 103, 2667.
(33) Harris, W. R.; Raymond, K. N. J. Am. Chem. Soc. 1979, 101, 6534.
(34) Pecoraro, V.; Weitl, F. L.; Raymond, K. N. J. Am. Chem. Soc. 1981,
103, 5133.
(35) Stack, T. D. P.; Hou, Z.; Raymond, K. N. J. Am. Chem. Soc. 1993,
115, 6466.
(36) Tse, B.; Kishi, Y. J. Am. Chem. Soc. 1993, 115, 7892.
(37) Karpishin, T. K.; Dewey, T. M.; Raymond, K. N. J. Am. Chem.
Soc. 1993, 115, 1842.
(38) Corey, E. J.; Bhattacharyya, S. Tetrahedron Lett. 1977, 45, 3919.
(39) Rastetter, W. H.; Erickson, T. J.; Venuti, M. C. J. Org. Chem. 1980,
45, 5011.
(40) Rastetter, W. H.; Erickson, T. J.; Venuti, M. C. J. Org. Chem. 1981,
46, 3579.
(41) Shanzer, A.; Libman, J.; Lifson, S.; Felder, C. E. J. Am. Chem.
Soc. 1986, 108, 7609.
(42) Marinez, E. R.; Salmassian, E. K.; Lau, T. T.; Gutierrez, C. G. J.
Org. Chem. 1996, 61, 3548. Ramirez, R. J. A.; Karamanukyan, L.; Ortiz,
S.; Gutierrez, C. G. Tetrahedron Lett. 1997, 38, 749.
(43) Considine, W. J. J. Organomet. Chem. 1966, 5, 263.

High-Yield Synthesis of the Enterobactin Trilactactone
J. Am. Chem. Soc., Vol. 119, No. 42, 1997 10095
8H). (+)-FABMS: m/z 555 (MH+), isotope pattern consistent with
formula. Anal. Calcd (Found) for C20H44O4Sn2: C, 41.0 (41.3); H,
7.57 (7.48).
two drops of acetic acid. The reaction mixture was stirred under H2
gas for 10 min. The deprotected ligand precipitated as a yellow solid.
Much of the ligand adhered to the carbon catalyst as a gel, so the
catalyst was repeatedly washed with hot ethanol. Yield: 54%. 1H
NMR (300 MHz, DMSO-d6): δ 3.46 (s, 9H), 4.39 (dd, J ) 11 Hz,
J′) 4.4 Hz, 3H), 4.49 (dd, J ) 11 Hz, J′) 7.4 Hz, 3H), 4.90 (m, J )
7.4 Hz, J′) 4.4 Hz, 3H), 6.48 (d, J ) 7.2 Hz, 3H), 7.19 (d, J ) 7.2
Hz, 3H), 8.81 (d, J ) 7.02, 3H). 1H NMR (300 MHz, CDCl3): δ 3.46
(s, 9H), 4.44 (dd, J ) 11.6 Hz, J′) 1.8 Hz, 3H), 5.0 (dd, J ) 11.6Hz,
J′) 3.4 Hz, 3H), 5.25 (br d, J ) 7.9 Hz, 3H), 6.68 (s, 6H), 8.65, (d, J
) 7.9 Hz, 3H). 13C NMR (400 MHz, DMSO-d6): δ 37.2, 52.4, 65.7,
104.8, 116.8, 126.7, 146.8, 158.5, 166.9, 168.0. (+)-FABMS: m/z
714 (MH+). HRMS Calcd (Found) for C30H31N6O15: m/z 715.18680
(715.18474).
Methyl N-trityl-L-serinate (2). To an ice cold slurry of methyl
L-serinate hydrochloride in chloroform (25 g, 0.16 mol) was added 48.5
g (0.49 mol) triethylamine. Solid triphenylmethyl chloride (50 g, 0.18
mol) was added over 3 h. The solution was allowed to warm to room
temperature and stirred overnight. The organic solution was washed
with 5% citric acid (3 × 100 mL), water (3 × 100 mL), and brine (3
× 100 mL). The solution was dried over MgSO4, filtered, and
evaporated to dryness. Recrystallization from benzene:hexanes gave
a beige crystalline material. Yield: 77%. Mp: 146 °C (lit. mp 148
°C).44 FT IR (thin film CHCl3 cast): ν 1703 cm-1
. Anal. Calcd
(Found) for C23H23NO3: C, 80.90 (81.20); H, 6.79 (6.63); N, 12.31
(12.27).
Ferric Hopobactin (8). Ferric acetylacetonate (1 equiv) in methanol
was added to 1 equiv of 7 dissolved in methylene chloride. The solution
immediately turned dark red and was allowed to stir for 2 h. The ferric
complex was purified by flash chromatography on silica gel (methylene
chloride:methanol 96:4 v/v). Crystals of the neutral complex were
grown in a variety of solvents: diffusion of ether into acetone, THF,
or methylene chloride gave fused plates. (+)-FABMS: m/z 768 (MH+).
X-ray Crystal Structure of Tris(N-trityl-L-serine) Trilactone (3).
Crystals of the N-trityl trilactone scaffold were obtained by slow
evaporation of the compound from a solution of CH2Cl2 and CH3OH
over a period of 2 weeks. A crystal of approximate dimensions 0.3 ×
0.2 × 0.1 mm was mounted on a glass fiber in a droplet of Paratone
oil. All measurements were made on a Siemens SMART diffractometer
with graphite-monochromated Mo Ka radiation.45 The data were
collected at -138(1) °C using the ω scan technique with a total frame
collection time of 30 s. Data analysis was performed using Siemens
XPREP program.46 No decay or absorption corrections were applied.
The structure was initially solved using SIR92 using the program
teXsan.47 The absolute configuration was established by the method
of Flack.48 After all the atoms were located, the data set was refined
using the SHELXTL (Version 5) software package.46 The structure
was refined on F2 in the orthorhombic space group P212121 (no. 19)
using full-matrix least squares. All non-hydrogen atoms in the molecule
were refined anisotropically. Hydrogen atoms were calculated and
refined on their respective carbon atoms. The final cycle of refinement
converged to R1 ) 0.0549 and R2 ) 0.0926 for 676 parameters and
7394 reflections.
X-ray Crystal Structure of Ferric Hopobactin. Dark red plates
of ferric hopobactin were grown from diffusion of ether into wet DMF
(∼ 5% water). A crystal of approximate dimensions 0.2 × 0.1 × 0.01
mm was mounted on a glass fiber in a droplet of Paratone oil. All
measurements were made on a Siemens SMART diffractometer with
graphite monochromated Mo Ka radiation.45 The data were collected
at -101(1) °C using the ω scan technique with a frame width of 0.3°
and a total frame collection time of 30 s. The intensity data were
extracted from the frames using the program SAINT and integration
box parameters of 1.0 (XY) × 0.5° (Z) to a maximum 2q value of
46.5°.49 Data analysis was performed using the Siemens XPREP
program.46 No decay correction was applied. A semiempirical
absorption correction based on ψ scans was applied to the data with
minimum and maximum transmission factors of 1.00 and 0.827,
respectively.
Tris(N-trityl-L-serine) Trilactone (3). Methyl N-trityl-L-serinate
(2) (50 g, 0.2 mol) and stannoxane (1) (12 g, 17.8 mmol) were added
to a flask fitted with a condenser and a Dean-Stark trap filled with 4
Å molecular sieves. The reaction mixture was refluxed under a nitrogen
atmosphere in toluene for 3 days. After cooling, the solution was
evaporated to dryness to get a beige solid. The solid was dissolved in
methylene chloride and purified on a flash silica plug. Evaporation of
solvent gave the product as a white powder. Yield: 56%. FT IR (thin
film CH2Cl2 cast): ν 1735 cm-1
.
1H NMR (400 MHz, CDCl3): δ
2.65 (d, J ) 6 Hz, 3H), 3.45 (m, 6H), 4.05 (t, J ) 10.4 Hz, 3H), 7.2-
7.5 (m, 27H). 13C NMR (400 MHz, CDCl3): δ 54.7, 66.2, 71.1, 126.7,
128.0, 128.6, 145.5, 172.2. (+)-FABMS: m/z 988 (M+), 744 (M+
-
trityl), 243 (trityl group). Anal. Calcd (Found) for C66H57N3O6: C,
80.22 (80.51); H, 5.81 (5.65); N, 4.25 (4.03). The preparation was
repeated with 10 g of methyl N-trityl-D-serine to give 2.3 g of the
enantioenterobactin precursor D-serinate trilactone.
Triserine Trilactone Trihydrochloride (4). To a slurry of 1 g of
3 in 25 mL degassed ethanol was added 1.5 mL of HCl (0.404 M) in
dry, degassed ethanol, and the slurry was brought to reflux in a
preheated oil bath for 15 mins. The reaction flask was cooled in an
ice bath, and roughly half of the solvent was removed in Vacuo, keeping
the solvent cold. The solid material was quickly filtered, while kept
under a gently blowing dry nitrogen stream, and washed with ethanol,
chloroform, and ether. Yield: 98%. FT IR (KBr pellet): ν 1776 cm-1
.
1H NMR (500 MHz, DMSO-d6): δ 4.29 (dd, J ) 12.5 Hz, J′) 2.46
Hz, 3H), 4.60 (s, 3H), 5.09 (dd, J ) 12.5 Hz, J′) 1.69 Hz, 3H), 9.25
(br s, 9H). This material was not further characterized.
3-(Benzyloxy)-1-methyl-2(1H)-4-carboxoylpyridinone Chloride
(5). To a chilled slurry of 3-(benzyloxy)-1-methyl-2-oxo-1H-pyridine-
4-carboxylic acid (2.0 g, 0.8 mmol) in freshly distilled benzene was
added oxalyl chloride (1.0 g, 0.8 mmol). Addition of DMF (∼0.1 mL)
catalyzed the reaction. The solution was stirred for 1 h and evaporated
to dryness at ∼5 °C. The remaining yellow oil was dissolved in
methylene choride and used without further purification.
N,N′,N′′-Tris[(3-(benzyloxy)-1-methyl-2-oxo-1H-pyridin-4-yl)car-
bonyl]cyclotriseryl Trilactone, Benzyl-Protected Hopobactin (6). 5
(10 equiv) was added quickly to a chilled solution of 4 (0.3 g) containing
10 equiv of triethylamine in methylene chloride (20 mL). After 1 h
the solvent volume was reduced to ∼1 mL. The remaining solution
was applied to a silica gel column eluted with chloroform and 2%
ethanol. The first several fractions were combined and washed with
5% citric acid, water, and brine. The organic solution was dried with
MgSO4 and filtered, and the volume was reduced. The remaining
solution was applied to a rotary chromatograph and compound 6 was
eluted using solvent gradient (3-0% hexanes in methylene chloride
followed by 0-2% methanol). Yield: 3% (based on triamine). 1H
NMR (300 MHz, CDCl3): δ 3.56 (s, 9H), 4.07 (q, J ) 11 Hz, J′) 7.1
Hz, 3H), 4.18 (q, J ) 11 Hz, J′) 4.4 Hz, 3H), 4.8 (m, J ) 7.1 Hz, J′)
4.4 Hz, 3H), 5.44 (d, J ) 9.5 Hz, 3H), 5.51 (d, J ) 9.5 Hz, 3H), 6.65
(d, J ) 7.2 Hz, 3H), 7.1-7.4 (m, 18H), 8.55 (d, J ) 7.1 Hz, 3H).
(+)-FABMS: m/z 985 (MH+), 894 (MH+ - Bnz).
The structure was solved by direct methods (SHELXTL Version 5
software) and refined on F2 in the space group P21 (no. 4) using full-
matrix least squares.46 Although there was some indication of
C-centering from the data, this was determined to be pseudosymmetry
and not crystallographically imposed. The iron atoms and selected
oxygen atoms were refined anisotropically, while the remaining non-
hydrogen atoms were refined isotropically. Hydrogen atoms were
calculated and refined on their respective carbon atoms. The final cycle
(45) SMART, Area-Detector Software Package; Siemens Industrial
Automation, Inc.: Madison, WI, 1993.
(46) SHELXTL, Crystal Structure Determination Package; Siemens
Industrial Automation, Inc.: Madison, WI, 1994.
(47) teXsan, Crystal Structure Analysis Package; Molecular Structure
Corporation, 1992.
(48) Flack, H. D. Acta Crystallogr. 1983, A39, 876.
(49) SAINT, SAX Area-Detector Integration Program v. 4.024; Siemens
Industrial Automation, Inc.: Madison, WI, 1994.
N,N′,N′′-Tris[(3-hydroxy-1-methyl-2-oxo-1H-pyridin-4-yl)carbo-
nyl]cyclotriseryl Trilactone, Hopobactin (7). Compound 6 (20 mg)-
was dissolved in a 1:1:1 mixture of ethanol:methanol:ethyl acetate (50
mL total). To this was added 10% Pd on carbon catalyst (<2 mg) and
(44) Guttman, S. HelV. Chim. Acta 1962, 45, 2622.

10096 J. Am. Chem. Soc., Vol. 119, No. 42, 1997
Meyer et al.
of refinement was based on 10510 reflections, 87 restraints, and 1036
variable parameters and converged to R1 ) 0.1165 and R2 ) 0.2661.
Solution Thermodynamics. General Methods. All titrant solu-
tions were prepared using distilled water that was further purified by
passing through a Millipore Milli-Q reverse osmosis cartridge system
(resistivity 18 MW cm). Titrants were degassed by boiling for 1 h
while being purged by argon. Solutions were stored under an
atmosphere of purified argon (Ridox Oxygen Scavenger (Fisher) and
Ascarite II (A. H. Thomas) scrubbers) in order to prevent absorption
of oxygen and carbon dioxide. Carbonate-free 0.1 M KOH was
prepared from Baker Dilut-It concentrate and was standardized by
titrating against potassium hydrogen phthalate using phenolphthalein
as an indicator. Solutions of 0.1 M HCl were similarly prepared and
were standardized by titrating against the KOH solution to phenol-
phthalein endpoint. The combined pH glass electrode (Orion semim-
icro) filled with 3 M KCl (Orion filling solution) was calibrated in
hydrogen ion concentration units (p[H] ) -log [H+]) by titrating 2.000
mL standardized HCl diluted in 50 mL of 0.100 M KCl (Mallinckrodt,
ACS grade) with 4.200 mL of standardized KOH. The Nernst
parameters of the electrode and the pKw value (13.78 ( 0.05) were
refined using a nonlinear least-squares program.50 All titrations were
performed using a Dosimat 665 (Metrohm) piston buret and an Accumet
925 (Fisher) pH meter, both instruments being controlled by an IBM-
PC computer.32,51 Errors are reported in parentheses and correspond
to the standard deviations. In all titrations the p[H] was kept below 8
to prevent the hydrolysis of the trilactone ring.
1H NMR Titration. Due to the low solubility of hopobactin in
water, ∼6 mg of ligand was dissolved in 0.5 mL DMF-d7 (Aldrich, %
D > 99.5) and the resulting solution diluted to 1 mL with D2O (Aldrich,
% D > 99.9). The titration was performed directly in the NMR tube
by adding small amounts of ∼10% diluted NaOD (Aldrich, 40% wt,
% D > 99) or ∼10% diluted DCl (Aldrich, 20% wt, % D > 99.5)
solutions in D2O with a Hamilton microsyringe. The solution was
purged with nitrogen prior to the pH measurements. Apparent values
were measured with an 3 mm Ag/AgCl combined microelectrode
(Mettler-Toledo) filled with 3 M KCl and an Accumet 925 (Fisher)
pH meter. The electrode was calibrated with three aqueous buffer
solutions (Fisher) at pH ) 4.00, 6.98, and 10.00. The measured pH
values were corrected for the deuterium effect and the presence of
∼50% DMF using the equation pH* ) pHmes + 0.05.52 The NMR
spectra were recorded on a 500 MHz Bruker AM500 spectrometer and
referenced to dioxane (δ ) 3.75 ppm).
Spectrophotometric Titrations. The apparatus and methods for
spectrophotometric titrations have been described in detail elsewhere.53
The path length of the quartz cell (Hellma, Suprasil) was 10 cm. All
solutions were maintained at constant ionic strength (0.100 M KCl),
under argon atmosphere and at constant temperature (25.0 ( 0.2 °C)
by using a jacketed titration vessel fitted to a Lauda K-2/R water bath.
To 40 mL of 0.100 M KCl, 10 mL of a freshly prepared ligand solution
(3 mM in DMF:water) was added. The effect of DMF was neglected,
since the final concentration was less than 0.4%. After the p[H] was
decreased to about 3 with 0.1 M HCl, the ligand was titrated with
standardized base to about p[H] 8.0. The thermodynamic reversibility
was checked by cycling each of two independent titrations from low
to high p[H] and back to low p[H]. The data (including absorbances
at 150 different wavelengths ranging between 250 and 400 nm, p[H]
values, and respective volumes of about 45 solutions per titration) were
analyzed to determine the protonation constants using the factor analysis
program FINDCOMP and the nonlinear least-squares refinement
program REFSPEC.53
as above and added as a stock solution. Between 8 and 12 solutions
were prepared in 10 mL volumetric flasks, containing ferric hopobactin
(∼8 mM), CDTA (99.9 mM), and 0.05 M 3-(N-morpholino)propane-
sulfonic acid (MOPS, Sigma, Cell Culture) as a buffer. The p[H] of
the solutions was adjusted to 6.3-8.0 and the ionic strength to 0.1 by
addition of 1.0 M KOH and 1.0 M KCl respectively. The solutions
were allowed to stand in the dark at room temperature for 1 week to
ensure thermodynamic equilibrium prior to recording the p[H] values
and absorption spectra (HP 8452A diode array spectrophotometer, 1
cm Hellma quartz cell). Absorbances at 25 wavelengths (10 nm
intervals between 400 and 640 nm inclusive) were used by the program
REFSPEC to calculate the overall formation constant.53 Protonation
constants of CDTA and the stability constant of FeCDTA used in the
refinement were taken from the literature.54
Electrochemical Measurements. Cyclic voltammograms of a 0.35
mM ferric hopobactin solution in acetonitrile (Fisher) containing 0.1
M tetrabutylammonium perchlorate (Fluka, Electrochemical grade) as
supporting electrolyte were recorded on a BAS100A electrochemical
analyzer. Working and auxiliary platinum electrodes were used with
a freshly prepared Pleskow reference electrode (0.01 M AgNO3 in 0.1
M (C4H9)4NClO4 in CH3CN/Ag). The half-wave potentials were also
measured against the ferrocinium/ferrocene internal reference electrode.
Results and Discussion
Synthesis. Previous methods of synthesizing the enterobactin
scaffold were laborious and gave very low yields. Corey38 and
Rastetter39,40 each obtained the scaffold in roughly 1% yield.
Shanzer improved the synthesis with the use of a tin template,
but the overall yield was still low (∼6%) due to the low-yield
formation of a serine â-lactone precursor.41 Gutierrez et al.
further developed the methodology used by Shanzer and
obtained the trityl-protected trilactone in 54% yield.42 Following
these precedents and Seebach’s syntheses of cyclopolyhydrox-
ybutanoate derivatives,55,56 a one-step synthesis of the protected
enterobactin scaffold that eliminates the â-lactone intermediate
was devised, as shown in Scheme 1.
Removal of the N-trityl protecting groups under acidic
conditions gave the enterobactin scaffold triamine trihydrochlo-
ride (Scheme 2). Bidentate 3-hydroxy-1-methyl-2(1H)-pyridi-
nonate groups (3,2-HOPO) were coupled via their acid chloride
to the triamine scaffold. The 3,2-HOPO ligands were chosen
because they have been identified in natural siderophores,57 are
easily activated for amide formation, and can be deprotected
by reductive hydrogenation,58,59 conditions in which the en-
terobactin scaffold is stable (Scheme 2). Hopobactin, like the
other 3,2-HOPO derivatives, has low pKa’s and a high affinity
for iron(III) (Vide infra). The ferric complex was obtained by
combining at room temperature free ligand dissolved in meth-
ylene chloride with ferric acetylacetonate dissolved in methanol,
followed by column chromatography purification. The complex
could be further purified by recrystallization from DMF:ether.
The one metal/one ligand stoichiometry was confirmed by mass
spectroscopy.
X-ray Crystal Structure of Tris(N-trityl-L-serine) trilac-
tone (3). The N-trityl trilactone is soluble in methylene chloride,
chloroform, and hot toluene. Slow evaporation from a meth-
ylene chloride:methanol solution produced colorless crystals,
Spectrophotometric Competition Titrations. The formation con-
stant of the ferric hopobactin complex was determined by competition
against trans-1,2-diaminocyclohexane-N,N,N′,N′-tetraacetic acid mono-
hydrate (CDTA, Aldrich, ACS grade 99%). For each of four
independent titrations, fresh ferric hopobactin was prepared and purified
(54) Martell, A. E.; Smith, R. M. Critical Stability Constants; Plenum
Press: New York, 1974-1981; Vols. 1-5.
(55) Plattner, D. A.; Brunner, A.; Dobler, M.; Mu¨ller, H.-M.; Petter, W.;
Zbinden, P.; Seebach, D. HelV. Chim. Acta 1993, 76, 2004.
(56) Mu¨ller, H.-M.; Seebach, D. Angew. Chem., Int. Ed. Engl. 1993,
32, 477.
(57) Meyer, J.-M.; Hohnadel, D.; Halle´, F. J. Gen. Microbiol. 1989, 135,
1479.
(58) Xu, J.; Kullgren, B.; Durbin, P. W.; Raymond, K. N. J. Med. Chem.
1995, 38, 2606.
(50) Martell, A. E.; Motekaitis, R. M. The Determination and Use of
Stability Constants; VCH: New York, 1988.
(51) Kappel, M.; Raymond, K. N. Inorg. Chem. 1982, 21, 3437.
(52) Perrin, D. D.; Dempsey, B. Buffers for pH and Metal Ion Control;
Chapman and Hall: London, 1974.
(53) Turowski, P. N.; Rodgers, S. J.; Scarrow, R. C.; Raymond, K. N.
Inorg. Chem. 1988, 27, 474.
(59) Xu, J.; Franklin, S. J.; Whisenhunt, D. W.; Raymond, K. N. J. Am.
Chem. Soc. 1995, 117, 7245.

High-Yield Synthesis of the Enterobactin Trilactactone
J. Am. Chem. Soc., Vol. 119, No. 42, 1997 10097
Scheme 1
Scheme 2
with a needlelike morphology. An X-ray analysis conforms to
an earlier report41 that included a cogent discussion of the
stereochemistry, but no full report of the structure. To provide
atom coordinates and a comparison structure, a full analysis
was performed.
macrocycle and one other on the opposite face. A trans
conformation has been observed in other examples only when
the stereocenters are heterochiral.55 Of the three N-trityl amine
substituents, two occupy equatorial sites, while the third occupies
an axial site on the macrocycle.
The N-trityl trilactone displays an unusual geometry for cyclic
trilactone compounds (Figure 2). Of those enterobactin analogs
that have been structurally characterized, all that have homo-
chiral stereocenters, including ferric hopobactin (Vide infra),
adopt a geometry with the carbonyl moieties in a cis conforma-
tion.37,60,61 These groups in the N-trityl structure adopt a trans
conformation with two carbonyl groups on one face of the
Superpositions of the trilactone rings from the N-trityl scaffold
and [V(enterobactin)]2- display significantly dissimilar confor-
mations, as shown in Figure 2. A least-squares fitting of the
12 atoms that comprise the macrocycles give a root mean
squares (RMS) deviation of 0.6 Å. The general form of the
macrocyclic rings are the same.28,61 However, the amide
nitrogens and serine carbonyl oxygens adopt the all-cis con-
formation in [V(enterobactin)]2- requisite for metal coordination,
which overlap poorly with the trans conformation in the trityl-
protected macrocycle.28 A comparison of the N-trityl trilactone
and the ferric hopobactin macrocycle (Vide infra) leads to the
(60) Shanzer, A.; Libman, J.; Frolow, F. J. Am. Chem. Soc. 1981, 103,
7339.
(61) Seebach, D.; Muller, H.-M.; Burger, H. M.; Plattner, D. A. Angew.
Chem., Int. Ed. Engl. 1992, 31, 434.

10098 J. Am. Chem. Soc., Vol. 119, No. 42, 1997
Meyer et al.
Table 1. Crystal Data and Structure Refinement for 3 and Ferric
Hopobactin (8)
formula
formula weight
C66H57N3O6
988.15
FeC30H28N6O15
768.43
T, °C
crystal system
space group
-138(1) °C
orthorhombic
P212121 (no. 19)
9.2495(2)
11.3584(1)
4.945(1)
-101(1) °C
monoclinic
P21 (no. 4)
13.0366(9)
22.632(2)
27.130(2)
100.926°
7859(1)
8
1.311
0.455
1.53 to 23.27
31676
-14 e h e 13
-25 e k e 23
-30 e l e 23
21051 [R(int) )
0.0981]
a, Å
b, Å
c, Å
â
V, Å3
Z
5142.1(2)
4
1.276
0.082
0.83 to 23.28
22178
-10 e h e 6
-10 e k e 12
-47 e l e 54
7395 [R(int) )
0.0396]
D
calcd, g cm-3
µ (Mo KR), mm-1
θ range, deg
reflections collected
index ranges
independent reflections
data/restraints/parameter
7394/0/676
10510/87/1036
final R indices (I > 2σ(I))a R1 ) 0.0448, R2 ) R1 ) 0.1165, R2 )
0.0926
1.117
largest diff. peak and hole 0.331 and
-0.311 eÅ-3
0.2661
1.171
1.038 and
-0.428 eÅ-3
goodness of fitb
2
4
a R1 ) ∑||Fo| - |Fc||/∑|Fo|, R2 ) {∑[w(Fo - Fc2)2]/∑[wFo ]}1/2
.
b GOF ) [∑w(|Fo| - |Fc|)2/(No - Nv)]1/2, where w ) 1/(σ2|Fo|).
Figure 2. An ORTEP diagram (top) of the tris(N-trityl-L-serine)
trilactone (3) (hydrogen atoms are omitted for clarity). The numbering
scheme is indicated for one third of the molecule only. A CHEM3D
overlay (bottom) of the triserine trilactone from 3 and V(enterobactin)2-
viewed down (left) and perpendicular to (right) the approximate
molecular 3-fold axis. The RMS error in atom positions is 0.600 Å.
same conclusion: the trityl protecting groups are too bulky to
conform to the all-cis conformation. The structure of the N-trityl
trilactone essentially establishes a limit on the predisposition
for complex formation of derivatives of the macrocycle.
X-ray Crystal Structure of Ferric Hopobactin. The ferric
hopobactin complex crystallizes in the monoclinic space group
P21 with Z ) 8 (Table 1). The absolute configuration was
established by the method of Flack.48 The very small crystal
(0.01 mm in one dimension) gave a low intensity data set with
many weak reflections, especially at high scattering angles, and
resulted in a low-resolution structure. However, the quality of
the analysis is more than adequate for establishing the coordina-
tion and molecular geometry (Figure 3). The generally low
intensities of reflections where h + k values are odd suggested
that the cell might be C centered. However, a significant
number of apparent violations (I/σ(I) > 30) and the final
structure showed that this is pseudo C centering of the iron
atoms, with significant deviations in the remaining parts of the
apparently C-centered related molecules.
The complex shown in Figure 3 is a trischelate with
approximate C3 point symmetry. Since the trilactone scaffold
is based on three L-serine units, the introduction of chirality at
the metal center produces a preferred diastereoisomeric neutral
complex. The metal center has a right-handed configuration
(∆ chirality) as found for the vanadium(IV) structure.28 The
same configuration has also been determined by circular
dichroism (CD) spectroscopy for the iron(III), chromium(III),
and rhodium(III) enterobactin complexes.23,62
Figure 3. A view of one of two crystallographically independent ferric
hopobactin complexes viewed (top) perpendicular to and (bottom) down
the approximate molecular 3-fold axis (hydrogen atoms are omitted
for clarity).
hydroxypyridinonate complex.63 Average lengths over the four
molecules in the asymmetric unit cell are 1.98(2) and 2.06(3)
Å respectively. The 3,2-HOPO chelates twist from octahedral
geometry (a 60° angle is expected for a perfect octahedron)
around the coordination polyhedron of the central iron atom.
The average Fe-O distance for the 2-hydroxy oxygen is
shorter by 0.08 Å than for the 3-carbonyl oxygen, typical of a
(62) Stack, T. D. P.; Karpishin, T. B.; Raymond, K. N. J. Am. Chem.
Soc. 1992, 114, 1512.

High-Yield Synthesis of the Enterobactin Trilactactone
J. Am. Chem. Soc., Vol. 119, No. 42, 1997 10099
The average twist angle, defined as the angle between two
coordinating atoms projected onto the plane perpendicular to
the idealized 3-fold axis, is 35(2)°. By comparison, the twist
angle in vanadium(IV) enterobactin is only 28°.37 Considering
that the twist angles in iron(III) triscatecholate complexes
are 4-6° larger than in the corresponding vanadium(IV)
complexes,62,64-66 the twist angle of iron(III) enterobactin has
been estimated to be close to 33°,37 which is in excellent
agreement with the value found for ferric hopobactin. The
optimum twist angle in a trischelate metal complex has been
predicted from the normalized bite of the bidentate chelating
units (the normalized bite value is the distance between the two
coordinating atoms of a chelate, divided by the metal-
heteroatom bond length).67 Since hydroxypyridinonates have
two significantly different metal-oxygen distances, only an
approximate normalized bite can be calculated. For ferric
hopobactin the normalized bite value is 1.27 and the corre-
sponding optimum twist angle based on the lowest repulsive
energy between the ligating atoms is 45°.67 This value is
supported by the structure of the simple Fe(3,2-HOPO)3
complex, which has a twist angle of 42.1°.63 The substantially
smaller twist angle observed in ferric hopobactin seems to
be a result of steric constraints imposed by the rigid scaf-
Figure 4. A CHEM3D overlay of two conformationally different ferric
hopobactin molecules from the asymmetric unit (hydrogen atoms are
omitted for clarity). The RMS error in atom positions is 0.461 Å.
fold. For comparison, the complexes V(TRENCAM)3- 68
,
Fe(TRENCAM)3- 62 V(Et3MECAM)2- [Et3MECAM ) 1,3,5-
,
tris(dihydroxybenzamidomethyl)-2,4,6-triethylbenzene],69 and
V(enterobactin)2- 37 have twist angles 4.8°, 3.2°, 6.6°, and 8.3°
3-
smaller than their nontethered analogs V(cat)3 (cat ) cat-
3- 64
2-
echol),65 Fe(cat)3
,
and V(eba)3
(eba ) N-ethyl-2,3-
dihydroxybenzamide)37 respectively.
Superposition of two molecules from the asymmetric unit
shows some conformational differences, with a RMS deviation
of atoms of 0.461 Å. As seen in Figure 4, the predominant
deviations occur in the tilting of the scaffold carbonyl groups
relative to the plane of the macrocycle. The overlay of ferric
hopobactin and V(enterobactin)2- molecules produces a RMS
deviation of 0.351 Å, showing that the structures are essentially
identical. Thus, the rigid structure of vanadium(IV) enterobactin
is not markedly perturbed by substituting the catecholamide by
hydroxypyridinonate binding groups, nor by the somewhat
smaller vanadium(IV) cation (ri ) 0.58 Å for V(IV) and 0.65
Å for Fe(III), respectively).70
Spectroscopic Properties of Ferric Hopobactin. The
visible absorption spectrum of ferric hopobactin (Figure 6)
shows two LMCT transitions centered at 432 nm (ꢀ ) 6200
M-1 cm-1) and 534 nm (ꢀ ) 5200 M-1 cm-1). The energy
and intensity of the bands are typical for pseudooctahedral ferric
trishydroxypyridinonate complexes. The chirality of the ferric
hopobactin complex was probed by circular dichroism spec-
troscopy. Figure 7 shows the CD spectrum recorded in pure
methanol. The band at 340 nm is due to the chiral trilactone
scaffold. Two additional transitions are observed in the visible
region at 410 nm and 550 nm. These bands arise from ligand-
Figure 5. A CHEM3D overlay of ferric hopobactin and V-
(enterobactin)2- based on the atomic coordinates from the X-ray
structures (hydrogen atoms are omitted for clarity). The RMS error in
atom positions is 0.351 Å.
to-metal charge transfer (LMCT) transitions and are therefore
sensitive to the chirality at the metal center.71 The CD spectrum
of ferric enterobactin (Figure 7) also shows two LMCT bands
of the same sign and magnitude as the transitions of ferric
hopobactin. On the basis of these similarities, the ferric
hopobactin spectrum is fully consistent with the assigned D
absolute configuration and represents the first chiral hydroxy-
pyridinonate complex.
(63) Scarrow, R. C.; Riley, P. E.; Abu-Dari, K.; White, D. L.; Raymond,
K. N. Inorg. Chem. 1985, 24, 954.
(64) Raymond, K. N.; Isied, S. S.; Brown, L. D.; Fronczeck, F. R.; Nibert,
J. H. J. Am. Chem. Soc. 1976, 98, 1767.
(65) Cooper, S. R.; Koh, Y. B.; Raymond, K. N. J. Am. Chem. Soc.
1982, 104, 5092.
(66) Karpishin, T. B.; Stack, T. D. P.; Raymond, K. N. J. Am. Chem.
Soc. 1993, 115, 182.
(67) Kepert, D. L. Inorganic Stereochemistry; Springer-Verlag: Berlin,
1982.
(68) Bulls, A. R.; Pippin, C. G.; Hahn, F. E.; Raymond, K. N. J. Am.
Chem. Soc. 1990, 112, 2627.
(69) Hou, Z.; Stack, T. D. P.; Sunderland, C. J.; Raymond, K. N. Inorg.
Chim. Acta 1997, submitted for publication.
Ligand Protonation Constants. The hopobactin ligand is
a triprotic acid. Its low solubility in water prevents the
determination of the stepwise protonation constants by poten-
1
tiometry. Therefore, a H NMR potentiometric titration was
(71) Karpishin, T. B.; Gebhard, M. S.; Solomon, E. I.; Raymond, K. N.
(70) Shannon, R. D. Acta Crystallogr. 1976, A32, 751.
J. Am. Chem. Soc. 1991, 113, 2977.

10100 J. Am. Chem. Soc., Vol. 119, No. 42, 1997
Meyer et al.
Figure 8. Variations of the 1H NMR chemical shifts recorded at 500
MHz for the aryl protons of hopobactin (Ha (b) and Hb (2)) as a
function of pH* in DMF-d7:D2O 50:50 v/v. T ) 22 °C; and [hopobactin]
∼ 8 mM.
Figure 6. The UV-vis absorption spectrum of ferric hopobactin in
pure methanol. T ) 22 °C.
Figure 9. Spectrophotometric titration of hopobactin by KOH. I )
0.1 KCl; T ) 25.0(2) °C; l ) 10 cm; [hopobactin] ) 0.6 mM; and
[KOH] ) 0.1 M.
Figure 7. Circular dichroism spectra of (a) ferric hopobactin in pure
methanol and (b) ferric enterobactin in water. T ) 22 °C.
behavior is consistent with a deprotonation of the chelating
units.
carried out in a mixed DMF-d7:D2O (∼1:1 v/v) solvent system.
The two aryl protons of the three equivalent 3,2-HOPO groups,
which appear as an AB system, were suitable nuclei to observe
the protonation of the free ligand. The assignment and chemical
Factor analysis of the spectrophotometric data recorded
between 250 and 400 nm indicated the presence of only two
distinct absorbing species in solution. Therefore, analysis of
the data with a model involving four independent absorbing
species (deprotonated, monoprotonated, diprotonated, and tripro-
tonated) was not successful in giving values for the protonation
constants K011, K012, and K013 defined by
1
shift changes for the aryl region of the H NMR spectra as a
function of pH* are depicted in Figure 8. Over the pH* range
2-8, the higher field signal (assigned to proton Hb in the para
position with respect to the hydroxyl protonation site) showed
no significant shift. In contrast, the lower field signal (assigned
to proton Ha located in the meta position) shifts from δ ) 7.17
to 6.59 ppm, as pH* increases from 5 to 8, with only one
inflection point at pH* ) 6.0(2). Although full deprotonation
is not achieved at pH* ) 8, the basicity was not further increased
in order to prevent hydrolysis of the trilactone scaffold.
The protonation properties of the chelator were further
investigated by UV absorption spectroscopy up to p[H] 8 as
shown in Figure 9. Below p[H] 3 the fully protonated ligand
shows a broad absorption band at 330 nm assigned to a
hydroxypyridinone-centered π f π* transition. As the p[H] is
raised, this band undergoes a progressive bathochromic and
hyperchromic shift. Simultaneously, an additional broad,
unresolved band appears at approximately 340 nm. This
LHnn--14 + H+ {K01n} LHnn-3
[LHnn-3
[LHn-4][H+]
]
K01n
)
(1)
n-1
â013 ) K011K012K013
The data were refined successfully using a model including only
protonated (LH3) and deprotonated (L3-) ligand. This suggests
that the spectroscopic properties of the mono- and diprotonated
species are linear combinations of the spectra of the fully
protonated and deprotonated ligand. Refinement of the overall

High-Yield Synthesis of the Enterobactin Trilactactone
J. Am. Chem. Soc., Vol. 119, No. 42, 1997 10101
1 week. The rigidity of the competitor, the steric hindrance of
the ferric hopobactin complex, and the high p[H] values, account
for the slow kinetics encountered to reach equilibrium.74 The
existence of an isosbestic point at 388 nm suggests that no
ternary iron-CDTA-hopobactin species forms in a significant
concentration.
The actual competition equilibrium can be expressed by the
following equations:
FeL + CDTA {K } FeCDTA + L
(3)
(4)
comp
FeCDTA
[FeCDTA]tot[L]tot â′
R
RL
110
Kcomp
)
)
â110 RCDTA
[FeL][CDTA]tot
where L, CDTA, and FeCDTA refer to all forms of unbound
hopolactin, CDTA, and iron(III)-complexed CDTA respec-
tively.
Figure 10. Spectrophotometric competition titration of ferric hopo-
bactin by KOH in the presence of CDTA. [MOPS] ) 0.05 M; I ) 0.1
KCl; T ) 22(2) °C; l ) 10 cm; [Fehopobactin] ) 7.8 mM; and [CDTA]
) 99.9 mM. Spectrum 1 corresponds to the pure complex.
[L]tot ) [L3-] + [LH2-] + ... + [LH3]
[CDTA]tot ) [CDTA4-] + [CDTAH3-] + ... + [CDTAH4]
[FeCDTA]tot ) [FeCDTA-] + [FeCDTA(OH)2-]
protonation constant yielded a log â013 of 18.2(3), which gives
an average protonation constant log Kav of 6.1 to each proton.
This value agrees well with the value estimated from the NMR
titration, those of TRENHOPO (log Kav ) 5.9, disregarding the
central amine proton),72 and those of the monochelate model
compound 1-methyl-4-(n-propylamido)-3,2-hydroxypyridinone
(log K011 ) 6.12).58 This behavior can be predicted, because
the hydroxypyridinonate units are identical and well isolated.
If the microscopic protonation constants are identical, the
macroscopic constants for the three steps would be equal to
Kav/3, Kav, and 3Kav.73 However, the ratio of two successive
protonation constants for the hexadentate catecholamides and
TRENHOPO is at least twice the value calculated for a statistical
distribution.31,32,59 A comparable behavior is expected in the
case of hopobactin. Indeed, the large absorbance increase
between p[H] 3.3 and 4.5 indicates a greater than statistical
separation of the protonation constants around the average value.
Stability of Ferric Hopobactin. The metal vs proton affinity
of hopobactin is sufficently large that the stability constant
cannot be determined directly. In an attempt to investigate the
protonation behavior of the ferric hopobactin complex, spec-
trophotometric titrations were performed. The iron complex is
stable in acidic conditions and shows no evidence of protonation
at low p[H]: no changes in the absorption spectrum could be
observed down to p[H] 2, suggesting that the complex remains
intact. Even the addition of 1.0 M HCl to dilute solutions of
ferric hopobactin induced no spectroscopic change after 12 h.
Therefore, the formation constant â110 (eq 2) was determined
by spectrophotometric competition experiments.
The aL, aCDTA, and aFeCDTA coefficients are defined else-
where.75 With the known protonation constants of hopobactin
and CDTA (log K011 ) 12.4, log K012 ) 6.15, log K013 ) 3.53,
log K014 ) 2.42),54 the complexation (log â′110 ) 30)54 and
hydroxylation (log â1′11h ) 20.30)54 constants of ferric CDTA,
â
110 was refined using a nonlinear least-squares method.53 The
average formation constant of ferric hopobactin, from three
independent titrations, was determined to be log â110 ) 26.4-
(3).
Hopobactin provides for the first time the opportunity to
directly assess the contribution toward metal binding of the
predisposition induced by the macrocyclic triserine scaffold on
chelating groups other than catecholamides. The almost identi-
cal formation constant of ferric hopobactin and ferric TREN-
HOPO clearly indicates that the trilactone scaffold does not
induce additional stabilization for hydroxypyridinonate ligands.
In contrast, ferric enterobactin is more stable than model
compounds, such as TRENCAM or MECAM (1,3,5-tris-
(dihydroxybenzamidomethyl)benzene), by a factor of 106.19,31
Although the structures of the hopobactin and enterobactin free
ligands are not known, the structural comparison of the
vanadium(IV) enterobactin scaffold with related trilactones
synthesized by Shanzer et al.,60 or the cyclic trilactones of
polyhydroxybutyric acids prepared by Seebach et al.,61 suggests
that there is very little reorganization of the ligand upon
complexation. Indeed, a least-squares analysis of the overlay
of the vanadium(IV) enterobactin scaffold28 and Seebach’s
trilactone55 gives a root mean-squares deviation of atom
positions of only 0.069 Å, whereas a significantly higher RMS
deviation (0.309 Å) is found with the ferric hopobactin scaffold
(Figure 11). Empirical force field calculations41 and analysis
of the X-ray structures of the free Et3MECAM ligand and its
vanadium(IV) complex35 provide additional evidence that the
ground state of the enterobactin scaffold is the same in the free
and metal-complexed form. These structures show that the only
degree of freedom is in the rotation of each of the catechol
groups. The predisposition of the pendant catecholamide
â
mFe3+ + lLn- + hH+ { mlh} FemLlHh
3m-nl+h
3m-nl+h
(2)
[FemLlHh
]
âmlh
)
[Fe3+]m[Ln-]l[H+]h
The stability constant of the ferric hopobactin was measured
by competition against 10-fold excess of CDTA over the p[H]
range 6.3-8.0 (Figure 10). Under these conditions, almost no
exchange can be seen at p[H] 6.3, whereas complete exchange
is observed around p[H] 7.5. At this p[H] the hydrolysis of
the enterobactin trilactone scaffold is negligible. Due to slow
exchange rates, the competition titrations were performed over
(74) Albrecht-Gary, A. M.; Palanche-Passeron, T.; Rochel, N.; Hennard,
C.; Abdallah, M. A. New J. Chem. 1995, 19, 105.
(75) Ringbom, A. Complexation in Analytical Chemistry; Interscience:
New York, 1963.
(72) Xu, J.; Raymond, K. N. Unpublished results.
(73) Perlmutter-Hayman, B. Acc. Chem. Res. 1986, 19, 90.

10102 J. Am. Chem. Soc., Vol. 119, No. 42, 1997
Meyer et al.
the scaffold on metal binding. The p[H] effect on the pM values
for the four hexadentate ligands is shown in Figure 12; at low
p[H], hopobactin becomes a more effective ligand than entero-
bactin. The reduced affinity of enterobactin for iron in acidic
conditions is due to the basicity of the free ligand and three
successive protonation steps of the ferric complex (log K111
)
4.95, log K112 ) 3.52, and log K113 ) 2.5),19 the latter resulting
in a salicylate mode of binding.78 For hopobactin no protonation
is observed: the complex stays intact and the coordination mode
unchanged, even down to p[H] 1.
Figure 11. A Stereoview overlay of the crystallographic structures of
the ferric hopobactin scaffold with Seebach’s trilactone (ref 61). The
RMS error in atom positions is 0.309 Å.
Electrochemistry. Due to the low solubility of the neutral
ferric hopobactin complex in water (∼10 mM), the cyclic
voltammograms were mesured in acetonitrile. For comparison,
cyclic voltammograms of ferric TRENHOPO were also re-
corded. Quasireversible signals centered at -782 and -875
mV/0.01 M AgNO3 are seen for hopobactin and TRENHOPO,
respectively. In both cases, the peak separation between the
cathodic and anodic waves (∆E) is 80 mV, and is independent
of the sweep rate in the range of 50-200 mV/s. The peak
current ratio Iox/Ired is comprised between 0.47 and 0.9 for both
complexes. Iox and Ired are both linear functions of the square
root of the sweep rate, indicating diffusion limited oxidation
and reduction processes. Under the same conditions, ferrocene
exhibits similar behavior (∆E ) 75 mV, Iox/Ired ) 0.9) with a
reduction potential of 91 mV/0.01 M AgNO3/Ag.79
binding units which are oriented to one face of the scaffold in
Et3MECAM, or in the analog synthesized by Tse and Kishi
(based on three fused cyclohexyl rings in chair conformation),
reduces the conformational space the chelates can occupy.35,36
This induces an increase in the overall stability of the complex.
For example, Et3MECAM gains 103 in stability over MECAM,
and the saturated analog of Tse and Kishi has a stability constant
with ferric ion that is the same as enterobactin.36 The free
energy advantage26 of enterobactin over its analogs is derived
from the scaffold architecture.
The variation of the stability constants on going from
bidentate to hexadentate catecholamide, hydroxamate, and 3,2-
hydroxypyridinonate ligands based on the TREN and entero-
bactin scaffolds are given in Table 2. Although no data are
yet available for a hydroxamic acid analog of enterobactin, the
magnitude of the formation constant for an iron(III) hydroxamate
complex is the same whether the binding groups are part of a
hexadentate molecule or three independent bidentate ligands.
Initially pointed out by Schwarzenbach and Schwarzenbach,76
and later emphasized by Carrano et al.,77 the free energy per
hydroxamate group is constant at a standard state of 1 M.
Interestingly, this apparent lack of chelate effect seen for the
hydroxamate-based ligands is more general and can be extended
to the more basic catecholamide chelators such as TRENCAM
(Table 2) or MECAM. As discussed above, the large “positive”
chelate effect observed for Et3MECAM and enterobactin is
solely due to the predisposition of the three chelating arms to
bind a metal ion. For the more acidic 3,2-hydroxypyridinonate
ligands, both TRENHOPO and hopobactin form a 100-fold
weaker ferric complex than the simple bidentate 1-methyl-4-
n-propylamido-3,2-hydroxypyridinonate model compound (Table
2). By relying on the structural comparison of ferric hopobactin
and vanadium(IV) enterobactin, a similar contribution of the
triester scaffold toward predisposition of the binding groups is
expected for both ligands. Hence, the gain in stability of ferric
hopobactin over ferric TRENHOPO is canceled by an antago-
nistic tethering effect of same magnitude.
Since proton competition for weak acid ligands depends on
their relative acidity and pH, the pM value (pM ) -log [Fe3+])
is a better measure of the relative ligand complexing strength
under given conditions ([Fe]tot ) 10-6 M, [L]tot ) 10-5 M).
Calculated pM values at the physiological p[H] of 7.4 for
hopobactin, enterobactin, TRENHOPO, and TRENCAM are
presented in Table 2. An increase of about 7.2 in the pM value
is seen between TRENCAM and enterobactin, whereas the
binding affinity of hopobactin increases only by one logarithmic
unit when compared with TRENHOPO. At p[H] 7.4, all of
the hydroxypyridinonate ligands are nearly fully deprotonated
(>90%), so any extra stability would reflect the influence of
By relying on extrathermodynamic assumptions discussed by
Alexander et al.,80 the half-wave potential vs NHE in water can
be estimated by the formula:81
1/2
NHE
1/2
Ag /Ag
E
(H2O) ) E
(S) + E0NHE(Ag+,H2O) -
+
{∆Gtr + RT ln aAg+}/nF (5)
) EA1/g2 /Ag(S) + 440 mV
+
1/2
NHE
+
E
(H2O), E1A/g2+/Ag (S), ∆Gtr, and aAg stand for the estimated
reduction potential (vs NHE) in water, the measured reduction
potential vs the Pleskow electrode in the organic solvent, the
free energy change of transfer of Ag+ from water to the organic
solvent, and Ag+ ion activity in the reference compartment,
respectively. The ∆Gtr values (-23.2 kJ mol-1 for acetonitrile)
have been tabulated by IUPAC.82 On the basis of eq 5,
estimates of the reduction potentials for ferric hopobactin and
TRENHOPO in water are -342 and -435 mV/NHE, which
are well within the range of biological reductants and higher
than those typically observed for hydroxamate ligands. Fer-
richrome A and ferrioxamine B have reduction potentials of
E1/2 ) -446 and -454 mV/NHE, respectively.83 Noteworthy,
TRENHOPO has a higher specificity for iron(III) over iron(II)
than hopobactin. The same trend is seen in the high pH limiting
reduction potentials of enterobactin25 and TRENCAM31 (E1/2
) -990 and -1040 mV/NHE, respectively). On the basis of
these examples, the triserine ring stabilizes the ferrous over the
ferric state compared with the TREN scaffold.
(78) Cass, M. E.; Garrett, T. M.; Raymond, K. N. J. Am. Chem. Soc.
1989, 111, 1677. Cohen, S. M.; Meyer, M.; Raymond, K. N. Manuscript
in preparation.
(79) Diggle, J. W.; Parker, A. J. Electrochim. Acta 1973, 18, 975.
(80) Alexander, R.; Parker, A. J.; Sharp, J. H.; Waghorne, W. E. J. Am.
Chem. Soc. 1972, 94, 1148.
(81) Gagne, R. R.; Koval, C. A.; Lisensky, G. C. Inorg. Chem. 1980,
19, 2855.
(82) Marcus, Y. Pure Appl. Chem. 1983, 55, 977.
(83) Cooper, S. R.; McArdle, J. V.; Raymond, K. N. Proc. Natl. Acad.
Sci. U.S.A. 1978, 75, 3551.
(76) Schwarzenbach, G.; Schwarzenbach, K. HelV. Chim. Acta 1963, 46,
1390.
(84) Hou, Z. Ph.D. Disseration, University of California at Berkeley,
1995.
(77) Carrano, C. J.; Cooper, S. R.; Raymond, K. N. J. Am. Chem. Soc.
1979, 101, 599.
(85) Ng, C. Y.; Rodgers, S. J.; Raymond, K. N. Inorg. Chem. 1989, 28,
2062.

High-Yield Synthesis of the Enterobactin Trilactactone
J. Am. Chem. Soc., Vol. 119, No. 42, 1997 10103
Table 2. Formation Constants and pM Values for Bidentate and Hexadentate Ligands Based on Catecholamides, Hydroxamates and
3,2-Hydroxypyridinonates Binding Groupsa
a pM ) -log [Fe3+] at p[H] ) 7.4 with [Fe]tot ) 10-6 M, [L]tot ) 10-5 M. b R′ ) C2H5, ref 84. c Reference 31. d Reference 19. e Acetohydroxamic
acid, ref 76. f Reference 85. g R′ ) nC3H7, ref 58. h Reference 72. i This work.
enterobactin analogs that have appended chelating units other
than catecholamides. The 3,2-hydroxypyridinonate derivative,
hopobactin, has been synthesized and its iron(III) complex fully
characterized. It is the first chiral representative of an iron
hydroxypyridinonate complex and under mild acid conditions
is the most stable ferric complex known.
Hopobactin and ferric hopobactin provide a good model for
studying the structure/function relationships of enterobactin and
ferric enterobactin. These and other derivatives, employing
hydroxamates, terephthalamides, and salicylates will be exam-
ined with respect to their chemical and microbial iron transport
properties.
Acknowledgment. This paper is dedicated to Professor
Dieter Seebach on the occasion of his 60th birthday. This work
was supported by NIH Grants AI 11744 and DK32999. The
authors thank Dr. F. J. Hollander for assistance with the X-ray
diffraction studies and an anonymous reviewer for finding an
important numerical error in the stability constants.
Figure 12. Plot of pM vs p[H] for hopobactin (b), TRENHOPO (9),
enterobactin (() and TRENCAM (2). [Fe]tot ) 10-6 M; [L] tot ) 10-5
M.
Supporting Information Available: Table of atomic co-
ordinates and equivalent isotropic displacement parameters, bond
lengths and angles, anisotropic displacement parameters, hy-
drogen coordinates and isotropic displacement parameters for
3 and 8 (21 pages). See any current masthead page for ordering
and Internet access instructions.
Conclusion
A facile method of generating the enterobactin scaffold has
been developed. This one-step synthesis of the trityl-protected
enterobactin trilactone scaffold gives yields as high as 50%.
This synthetic approach enables the development of direct
JA970718N