PhzA/B catalyzes the formation of the tricycle in phenazine biosynthesis

Article abstract of DOI:10.1021/ja806325k

Phenazines are redox-active bacterial secondary metabolites that participate in important biological processes such as the generation of toxic reactive oxygen species and the reduction of environmental iron. Their biosynthesis from chorismic acid depends

Full text of DOI:10.1021/ja806325k

Published on Web 11/17/2008
PhzA/B Catalyzes the Formation of the Tricycle in Phenazine
Biosynthesis
Ekta G. Ahuja,† Petra Janning,† Matthias Mentel,‡,§ Almut Graebsch,‡
,#
Rolf Breinbauer,†,‡,§,| Wolf Hiller,‡ Burkhard Costisella,‡ Linda S. Thomashow,
Dmitri V. Mavrodi, and Wulf Blankenfeldt*,†
Max-Planck-Institute of Molecular Physiology, Otto-Hahn-Strasse 11,
44227 Dortmund, Germany, Technical UniVersity of Dortmund, Faculty of Chemistry,
Otto-Hahn-Strasse 6, 44221 Dortmund, Germany, UniVersity of Leipzig, Institute of Organic
Chemistry, Johannisallee 29, 04103 Leipzig, Germany, Graz UniVersity of Technology, Institute
of Organic Chemistry, Stremayrgasse 16, 8010 Graz, Austria, Washington State UniVersity,
Pullman, Washington 99164-6430, and USDA, Agricultural Research SerVice, Root Disease and
Biological Control Research Unit, Pullman, Washington 99164-6430
Received August 10, 2008; E-mail: wulf.blankenfeldt@mpi-dortmund.mpg.de
Abstract: Phenazines are redox-active bacterial secondary metabolites that participate in important biological
processes such as the generation of toxic reactive oxygen species and the reduction of environmental
iron. Their biosynthesis from chorismic acid depends on enzymes encoded by the phz operon, but many
details of the pathway remain unclear. It previously was shown that phenazine biosynthesis involves the
symmetrical head-to-tail double condensation of two identical amino-cyclohexenone molecules to a tricyclic
phenazine precursor. While this key step can proceed spontaneously in vitro, we show here that it is
catalyzed by PhzA/B, a small dimeric protein of the ∆5-3-ketosteroid isomerase/nuclear transport factor 2
family, and we reason that this catalysis is required in vivo. Crystal structures in complex with analogues
of the substrate and product suggest that PhzA/B accelerates double imine formation by orienting two
substrate molecules and by neutralizing the negative charge of tetrahedral intermediates through protonation.
HPLC-coupled NMR reveals that the condensation product rearranges further, which is probably important
to prevent back-hydrolysis, and may also be catalyzed within the active site of PhzA/B. The rearranged
tricyclic product subsequently undergoes oxidative decarboxylation in a metal-independent reaction involving
molecular oxygen. This conversion does not seem to require enzymatic catalysis, explaining why phenazine-
1-carboxylic acid is a major product even in strains that use phenazine-1,6-dicarboxylic acid as a precursor
of strain-specific phenazine derivatives.
model of lung infections.2,3 Since the lungs of nearly all victims
of cystic fibrosis are chronically colonized by P. aeruginosa
and this infection contributes significantly to the low life
expectancy of these patients,4 phenazine production may present
an attractive target for pharmaceutical intervention.
Phenazines play multiple roles in the bacteria that generate
them, and only recently has an understanding of these roles
begun to emerge.5 Phenazine-producing strains residing on roots
can protect cultivated plants against susceptible bacterial and
fungal pathogens.6,7 In soil, phenazines can participate in iron
acquisition by reducing environmental Fe3+ to the more soluble
Fe2+, and they also function in primary energy metabolism by
1. Introduction
The production of phenazine pigments is among the most
readily observable phenomena in bacteria, due to their intense
color. This pigmentation triggered the early discovery of the
blue phenazine derivative pyocyanin (1-hydroxy-5N-meth-
ylphenazinium betaine), which was isolated in the 1850s from
wound dressings of soldiers infected with Pseudomonas aerugi-
nosa.1 Phenazines are redox-active and can reduce molecular
oxygen, leading to the generation of toxic reactive oxygen
species and explaining their broad-spectrum antibiotic activity.
They also act as virulence factors in human infectious disease.
For example, pyocyanin induces apoptosis in neutrophils, and
strains of Pseudomonas aeruginosa defective in pyocyanin
synthesis are more susceptible to immune responses in a mouse
(2) Allen, L.; Dockrell, D. H.; Pattery, T.; Lee, D. G.; Cornelis, P.;
Hellewell, P. G.; Whyte, M. K. J. Immunol. 2005, 174, 3643–3649.
(3) Lau, G. W.; Hassett, D. J.; Ran, H.; Kong, F. Trends Mol. Med. 2004,
10, 599–606.
(4) Heijerman, H. J. Cyst. Fibros. 2005, 4 Suppl 2, 3–5.
(5) Price-Whelan, A.; Dietrich, L. E.; Newman, D. K. Nat. Chem. Biol.
2006, 2, 71–78.
† Max-Planck-Institute of Molecular Physiology.
‡ Technical University of Dortmund.
§ University of Leipzig.
(6) Giddens, S. R.; Feng, Y.; Mahanty, H. K. Mol. Microbiol. 2002, 45,
769–783.
| Graz University of Technology.
Washington State University.
(7) Cook, R. J.; Thomashow, L. S.; Weller, D. M.; Fujimoto, D.; Mazzola,
M.; Bangera, G.; Kim, D. S. Proc. Natl. Acad. Sci. U.S.A. 1995, 92,
4197–4201.
# USDA.
(1) Fordos, J. Recl. TraV. Soc. d’Emulation Sci. Pharm. 1859, 3, 30.
9
10.1021/ja806325k CCC: $40.75
2008 American Chemical Society
J. AM. CHEM. SOC. 2008, 130, 17053–17061 17053

A R T I C L E S
Ahuja et al.
Scheme 1. Biosynthesis of Phenazine-1-carboxylic acid (7)a
a Steps upstream of PhzA/B catalysis are shown in green, and intermediates discussed here are shown in red. Orange indicates byproducts. Reactions that
lead to the formation of PhzA/B’s product are boxed.
mediating the reoxidation of NADH and utilization of pyruvate
under oxygen-limiting conditions.8,9 In P. aeruginosa, pyocyanin
is a signaling molecule that controls gene expression by directly
activating the iron-containing transcription factor SoxR.10 Col-
lectively, these observations reveal some of the ways in which
phenazines are exploited by their producers to interact with the
environment, showing that they are more than simple secondary
metabolites.
Dozens of naturally occurring phenazine derivatives have
been isolated, and existing knowledge indicates that all of them
originate from either phenazine-1-carboxylic acid (7) or phena-
zine-1,6-dicarboxylic acid (9) (Scheme 1).11-13 These “core”
phenazines are synthesized from two molecules of chorismic
acid (1) in a pathway catalyzed by enzymes encoded in the
conserved “phz operon”. A compilation of available sequence
data indicates that five Phz-enzymes are indispensable for
production of 7 and/or 9 (Scheme 1 and Supporting Information
Figure S1). Of these, PhzE converts 1 to 2-amino-4-desoxy-
isochorismate (2).14 Second, isochorismatase PhzD cleaves 2
to generate trans-2,3-dihydro-3-hydroxyanthranilate (3),15 the
last intermediate that can be isolated. In the next step, PhzF
isomerizes 316,17 to 6-amino-5-oxocyclohex-2-ene-1-carboxylic
acid (4) by initiating a [1,5]-prototropic shift.16 This ketone can
undergo spontaneous dimerization by symmetrical head-to-tail
double condensation in Vitro, yielding a tricyclic phenazine
precursor 5/5′/5a that must be further oxidized to arrive at the
aromatic phenazine moiety. In Pseudomonas, the steps past
ketone condensation include oxidative decarboxylation to 6, and
core phenazine biosynthesis ends at 7, whereas other bacteria,
e.g. Pantoea agglomerans, generate 9 as their core phenazine.6
Intriguingly, in Vitro studies have revealed that none of the steps
following ketone formation require enzyme catalysis, and PhzF
alone can trigger conversion of 3 to 7, albeit with low yield.16,17
This raises questions about the function of the remaining
conserved enzymes PhzA/B and PhzG. While it is relatively
(14) McDonald, M.; Mavrodi, D. V.; Thomashow, L. S.; Floss, H. G. J. Am.
Chem. Soc. 2001, 123, 9459–9460.
(15) Parsons, J. F.; Calabrese, K.; Eisenstein, E.; Ladner, J. E. Biochemistry
2003, 42, 5684–5693.
(16) Blankenfeldt, W.; Kuzin, A. P.; Skarina, T.; Korniyenko, Y.; Tong,
L.; Bayer, P.; Janning, P.; Thomashow, L. S.; Mavrodi, D. V. Proc.
Natl. Acad. Sci. U.S.A. 2004, 101, 16431–16436.
(8) Hernandez, M. E.; Kappler, A.; Newman, D. K. Appl. EnViron.
Microbiol. 2004, 70, 921–928.
(17) Parsons, J. F.; Song, F.; Parsons, L.; Calabrese, K.; Eisenstein, E.;
Ladner, J. E. Biochemistry 2004, 43, 12427–12435.
(9) Price-Whelan, A.; Dietrich, L. E.; Newman, D. K. J. Bacteriol. 2007,
189, 6372–6381.
(18) Parsons, J. F.; Calabrese, K.; Eisenstein, E.; Ladner, J. E. Acta
Crystallogr., D: Biol. Crystallogr. 2004, 60, 2110–2113.
(19) Mavrodi, D. V.; Ksenzenko, V. N.; Bonsall, R. F.; Cook, R. J.;
Boronin, A. M.; Thomashow, L. S. J. Bacteriol. 1998, 180, 2541–
2548.
(10) Dietrich, L. E.; Price-Whelan, A.; Petersen, A.; Whiteley, M.; Newman,
D. K. Mol. Microbiol. 2006, 61, 1308–1321.
(11) Turner, J. M.; Messenger, A. J. AdV. Microb. Physiol. 1986, 27, 211–
275.
(20) Mavrodi, D. V.; Bonsall, R. F.; Delaney, S. M.; Soule, M. J.; Phillips,
G.; Thomashow, L. S. J. Bacteriol. 2001, 183, 6454–6465.
(21) Haagen, Y.; Gluck, K.; Fay, K.; Kammerer, B.; Gust, B.; Heide, L.
ChemBioChem 2006, 7, 2016–2027.
(12) Laursen, J. B.; Nielsen, J. Chem. ReV. 2004, 104, 1663–1686.
(13) Mavrodi, D. V.; Blankenfeldt, W.; Thomashow, L. S. Annu. ReV.
Phytopathol. 2006, 44, 417–445.
9
17054 J. AM. CHEM. SOC. VOL. 130, NO. 50, 2008

The Phenazine Biosynthesis Protein PhzA/B
A R T I C L E S
clear that the FMN-dependent PhzG is involved in one of the
terminal oxidation reactions,18 the function of PhzA/B is not
obvious. Working with E. coli expressing fragments of the phz
operon, it was shown that this enzyme, while not required for
phenazine biosynthesis, increases the efficiency of the pathway
severalfold.14,19 Deletion of the gene in P. agglomerans led to
complete abrogation of phenazine biosynthesis,6 arguing for an
absolute in ViVo requirement, and Parsons et al. hypothesized
that it may be involved in the dimerization step that generates
the phenazine tricycle.17 This could explain why phenazine
production is so prominent in the pseudomonads, which, in
contrast to other phenazine producing bacteria, possess two
adjacent, highly homologous copies of this gene, phzA and phzB
(approximately 80% amino acid identity, Figures S1 and
S2).19,20 Two copies of the gene are also found in Streptomyces
cinnamomensis, but they are not neighbored.21 Because of the
high similarity of PhzA, PhzB, and the corresponding proteins
in other strains, they are termed “PhzA/B” here when not
referring to one of the two specific Pseudomonas genes.
Fold prediction indicated that PhzA/B belongs to the ∆5-3-
ketosteroid isomerase (KSI)/nuclear transport factor 2 (NTF2)
family.22 Because members of this family participate in a
plethora of reactions, serve in nuclear transport processes, and
act as assembly platforms for larger protein complexes, it is
not possible to deduce the function of PhzA/B from this fold
relationship alone. We have therefore developed biochemical
assays and performed crystal structure analysis of PhzA/B from
Burkholderia cepacia R18194 to learn more about the role of
this protein in phenazine biosynthesis.
that this product is the nonsymmetrical molecule 5a, containing
four conjugated double bonds (Figure 1c, Figures S10 and S12).
This shows that condensation is accompanied by rearrangement
reactions that may be required to stabilize double imine 5, which
probably is water-sensitive, against back-hydrolysis to 4.
PhzA/B could fulfill two different roles in the reaction cascade
leading from 4 to 5a. First, PhzA/B could utilize ketone 4 as
its substrate and catalyze condensation to 5, followed by
spontaneous rearrangement to 5a. Second, if condensation
already takes place in PhzF as we previously speculated,16
PhzA/B could isomerize 5 to 5a by acid/base catalysis. In this
case, 4 would not be released from PhzF and the ketones 4 and
4a detected by HPLC-MS would stem exclusively from back-
hydrolysis of 5. We therefore assessed the incorporation of 18
O
into 4 and 4a in reactions carried out in H218O. Only minor
incorporation of the label was found shortly after initiation of
the reaction, when ketone concentrations are highest (Figure
S6), thus demonstrating that PhzF indeed releases ketone 4 as
its product. This suggests that PhzA/B catalyzes the subsequent
condensation reaction, which is also supported by an experiment
in which PhzA/B was added after PhzF had already fully
depleted its substrate 3 and the concentration of the condensation
product 5a was again declining (Figure S7). Here, a new
maximum of 5a developed, indicating that PhzA/B utilizes free
ketone 4 as its substrate. It also shows that formation of 5a
does not involve the substrate of PhzF 3 in, for example, a
reaction condensing one molecule of 3 and one molecule of
ketone 4.
2.2. The Product of PhzA/B Is Oxidation Sensitive. Mass
spectrometry reveals that 5a undergoes oxidative decarboxyla-
tion to give a species that most likely is the partially aromatized
intermediate 6. To identify the electron acceptor of this and the
following reactions, we monitored dissolved O2 with a Clark
electrode. This shows that oxygen is depleted once PhzF begins
turnover of 3 and that O2-depletion is accelerated with increasing
concentrations of PhzA/B (Figure 1d). With a carbon dioxide
selective electrode, it was also possible to confirm the release
of CO2 (Figure S13). Following the reaction by HPLC-MS
shows that intermediate 6 forms earlier in the presence of
PhzA/B but that its overall quantity does not change significantly
(Figure 1e), which may be a consequence of limited oxygen
availability in the reaction mixture. Because efficient oxygen
consumption is already triggered by PhzF alone and because
the higher rate of depletion likely results from PhzA/B’s
acceleration of the step preceding oxidative decarboxylation,
we hypothesize that oxidation of 5a to 6 proceeds without
enzymatic catalysis. This would explain why 7 is found as a
major byproduct even in strains that utilize 9 rather than 7 as a
precursor of their strain-specific phenazines.6
Experiments with chelating reagents show that oxygen
consumption does not depend on metal ions (Figure S14),
leaving the question about oxygen activation unanswered. In
reactions with low initial amounts of 3, approximately half of
the depleted oxygen was recovered by addition of catalase,
demonstrating the generation of H2O2, which was also detectable
with Amplex Red (Figure S15).
While details of the spontaneous conversion of 5a to 6 remain
elusive, it is apparently propagated by the irreversibility of
decarboxylation and by the stability gained in generating the
aromatic π-system of 6. In this regard, phenazine biosynthesis
is reminiscent of pathways involving oxidative cascades, in
which cyclization is followed by oxidation that is often
uncatalyzed and restores aromaticity on the substrate.23 More
2. Results
2.1. PhzA/B Catalyzes the Condensation Step in Phen-
azine Biosynthesis. In order to identify if and which of the steps
in phenazine biosynthesis PhzA/B catalyzes, we applied different
ratios of PhzF and PhzA/B from B. cepacia to 3 and employed
HPLC-coupled mass spectrometry to follow the turnover of 3,
ketone 4, tricyclic intermediate 5/5′/5a, the product of its
oxidative decarboxylation 6, and the end product 7 in addition
to some minor species including probably the less reactive vinyl
ketone 4a and the byproducts 9 and phenazine (8) (Figure 1a,
b, and e and Figures S3-S9). The same intermediates were also
observed with Phz-enzymes from Pseudomonas, showing that
phenazine biosynthesis follows the same pathway in both
species.
Increasing concentrations of PhzA/B led to greater amounts
of condensation product 5/5′/5a, which also peaked at earlier
times (Figure 1a). Simultaneously, ketone 4 and its isomer 4a
were drastically decreased (Figure 1b and Figure S5a). To
determine the structure of the unstable condensation product,
we conducted HPLC-coupled NMR experiments, which reveal
(22) Ahuja, E. G.; Mavrodi, D. V.; Thomashow, L. S.; Blankenfeldt, W.
Acta Crystallogr., D: Biol. Crystallogr. 2004, 60, 1129–1131.
(23) Dorrestein, P.; Begley, T. P. Bioorg. Chem. 2005, 33, 136–148.
(24) Kis, K.; Kugelbrey, K.; Bacher, A. J. Org. Chem. 2001, 66, 2555–
2559.
(25) Burgi, H. B.; Dunitz, J. D.; Lehn, J. M.; Wipff, G. Tetrahedron 1974,
30, 1563–1572.
(26) Kim, S. W.; Cha, S. S.; Cho, H. S.; Kim, J. S.; Ha, N. C.; Cho, M. J.;
Joo, S.; Kim, K. K.; Choi, K. Y.; Oh, B. H. Biochemistry 1997, 36,
14030–14036.
(27) Doublie, S. Methods Enzymol. 1997, 276, 523–530.
(28) Wolf, C.; Liu, S. L.; Mei, X. F.; August, A. T.; Casimir, M. D. J.
Org. Chem. 2006, 71, 3270–3273.
(29) Smallcombe, S. H.; Patt, S. L.; Keifer, P. A. J. Magn. Reson. A 1995,
117, 295–303.
9
J. AM. CHEM. SOC. VOL. 130, NO. 50, 2008 17055

A R T I C L E S
Ahuja et al.
Figure 1. Functional characterization of PhzA/B. (a) Influence of increasing concentrations of PhzA/B on ketone condensation product 5a, observed by
HPLC-MS without column separation. The initial concentration of 3 was 0.1 mM, and PhzF was applied at 1 µM. (b) Depletion of ketone 4 by PhzA/B,
observed by HPLC-MS with column separation. The initial concentration of 3 was 1 mM, and turnover was initiated with 1 µM PhzF. Note the different
time scales of (a) and (b). (c) 1H NMR and assignment of ketone condensation product 5a recorded by HPLC-coupled NMR spectroscopy. (d) Depletion of
oxygen during turnover of 1 mM 3 in the presence of increasing amounts of PhzA/B. Reactions were initiated with 1 µM PhzF. (e) Influence of PhzA/B on
6 observed by HPLC-MS with column separation.
research will be needed to understand this step in phenazine
biosynthesis, but it seems that 5a is primed to lose CO2 due to
the conjugated π-system in the ꢀ-position to the carboxylate
group.
The ability to accelerate oxygen depletion was used as an
activity test for assessing the effect of mutations in PhzA/B
(Table S4).
2.3. PhzA/B Is a Dimeric KSI/NTF2 Protein. To gain further
insight into the role of PhzA/B, we have determined the crystal
structure of the B. cepacia enzyme in two different crystal forms
with 1.65 and 2.2 Å resolution, respectively (Figure 2 and Table
S2). PhzA/B indeed possesses a KSI/NTF2-fold, but the dimer
features an arm-exchange not observed in related protein
structures (listed in Table S3). Consequently, the C-terminus
of each chain contributes to the formation of the active center
of the neighboring monomer (Figure 2). Residues of the
C-terminus refine to higher B-factors than the rest of the
structure and could not be traced in some of the data sets
collected for this study (from R160 or D161 to S165 in the
higher or lower resolution crystal form, respectively; Figure
S18). Since the active site is fully shielded from the solvent
when all residues of the C-terminus are present, we hypothesize
that the C-terminus acts as a flexible lid that controls access to
the active site. It is also critical for activity and stability (Table
S4): while introduction of a stop codon at position 164 yields
a fully active enzyme, truncation at residue 162 impairs activity
slightly and shortening the protein to 159 residues renders
PhzA/B nonfunctional, which is expected, since R160 partici-
pates in ligand binding (see below). Truncation at P156 results
in a protein that is no longer stable. PhzA/B also possesses
flexible R-helical N-termini that point out of the fold core and
are too mobile to be fully traced. Their function is not obvious.
The two active sites of the dimer reside in large cavities lined
mainly by aromatic and hydrophobic residues together with
several polar side chains (Figure 2b-d), whose importance is
highlighted by their high degree of conservation (Figure S2).
Two acetate ions from the crystallization buffer bind to R41
and R160* in the active site of the apo structure (* indicates
the second monomer). Their methyl groups point in opposite
parallel directions, hinting at the position and approximate
9
17056 J. AM. CHEM. SOC. VOL. 130, NO. 50, 2008

The Phenazine Biosynthesis Protein PhzA/B
A R T I C L E S
Figure 2. Crystal structure of Burkholderia cepacia PhzA/B. (a) Overall structure of the dimer. N- and C-termini are indicated. Compound 11 is shown in
one of the active centers. The active site cavity is shown as a gray mesh in one of the monomers. (b) Binding of two acetate ions (ACT) in the apo structure
with emphasis on nonpolar residues lining the active center. (c) Binding of 10 with emphasis on conserved polar residues of the active center. (d) Binding
of 11 and azide. Thin gray lines indicate the shifted position of acetate and R41 in the active center of the second monomer.
orientation with which substrate molecules bind (Figure 2b).
Because this information is not sufficient to derive the details
of the interaction between the protein and its natural ligands
and because these ligands are too unstable to be isolated for
crystallographic experiments, we used isothermal titration
calorimetry (ITC) to test substrate analogues and synthetic
mimics of the product of PhzA/B (Figures S16, S17, S19, and
Table S1). While none of the analogues of ketone 4 showed
binding in ITC, it was still possible to obtain complexes with 3
and with 3-oxocyclohexanecarboxylic acid (10) by soaking
(Figure 2c and Figure S19). The carboxylate group of these
ligands interacts either only with R41 in the form of a salt bridge
(complex with 3, Figure S19a) or with R41 and S77 simulta-
neously (complex with 10, Figure 2c and Figure S19b). It is
possible that this ambivalence is required in the catalytic cycle
of PhzA/B (Figure 3). The 2-amino and 3-oxo/hydroxy groups
form hydrogen bonds to E140. E140, S77, and R41 adopt
multiple conformations in the various structures presented here,
and this mobility seems important for catalysis (Figure 3a,
Scheme 2). The 3-oxo/hydroxy groups of 3 and 10 also bind to
Y120 via a water molecule. Interestingly, even if a racemic
mixture of 10 was used, only the (R)-form binds to the active
center, and C1 of 3 is distorted toward (R)-configured sp3-
hybridization (Figure S19a), in both cases exhibiting the
expected stereochemistry at C1 of ketone 4. While only one
molecule of 3 binds to the active center, a second, weakly bound
molecule of 10, also in the (R)-configuration, replaces acetate
at R160*, where it also coordinates Q147 and, through the same
water molecule as the first molecule of 10, Y120. Its 3-oxo group
forms a hydrogen bond to H73.
Of the product analogues, the best compounds have
affinities between 2 and 25 µM (Figure S17 and Table S1).
Since all of the analogues investigated by crystallography
displayed a similar binding mode (Figure S20), only 2-(cy-
clohexylamino)benzoic acid (11) is discussed here (Fig-
ure 2d and Figure S19c). Like the second molecule of 10,
the carboxylate group of 11 replaces acetate at R160*, while
the other acetate is pushed to the side together with R41 in
one of the active centers or replaced by azide from the
crystallization buffer in the other (Figures 2d and 3a). Further,
the amino bridge in 11 forms a hydrogen bond to the side
chain of E140, indicating how an imine intermediate could
interact with PhzA/B. The planes of the two rings are tilted
such that substitution with carboxylate at C3′ of the cyclo-
hexyl moiety for interaction with R41 requires a chiral center
with the same configuration as C1 in ketone 4. This explains
why none of the synthetic ligands with two carboxylate
groups investigated here show binding to PhzA/B (Figure
S16).
9
J. AM. CHEM. SOC. VOL. 130, NO. 50, 2008 17057

A R T I C L E S
Ahuja et al.
Figure 3. Catalytic proposal for PhzA/B. The initial position of the two ketone substrate molecules 4 in (a) is based on the complexes with 3 and 10. Gray
sticks show the conformation of R41, S77, and E140 found in other data sets (Table S2). Orange dashes mark nucleophilic attacks; circles indicate catalysis
by proton transfer. THI1/2, first and second tetrahedral intermediate; I1, first imine.
3. Discussion
In this respect, the function of PhzA/B is reminiscent of the
otherwise unrelated lumazine synthase in riboflavin biosynthesis,
which catalyzes a similar condensation reaction that can also
proceed spontaneously.24
Our data show that PhzA/B catalyzes the formation of the
phenazine tricycle from two identical precursors, a reaction that
also proceeds without catalysis in Vitro. Why does the phz
operon contain the PhzA/B enzyme if it is apparently not
essential for phenazine biosynthesis? The answer may be that
the uncatalyzed condensation requires a bimolecular step, for
which the reaction rate declines exponentially upon dilution.
Accordingly, the uncatalyzed reaction is very slow at low
concentrations of 3 and catalysis by PhzA/B leads to great
acceleration as well as, within the time frame of the experiments
summarized here, overall yield, since side reactions are pre-
vented. On the other hand, only little acceleration results when
the initial amount of substrate is higher than 1 mM, indicating
that significant product forms uncatalyzed under these conditions
(Figure S8). In the bacterial cell, phenazine biosynthesis is
coupled to the shikimate pathway, which implies that the
intracellular concentration of phenazine precursors never reaches
millimolar levels. Under these circumstances, breakdown of the
reactive ketone 4 by, for example, oxidation or reaction with
free amino groups of proteins must be avoided, explaining why
PhzA/B catalysis is required in ViVo. This is, for example,
demonstrated by the complete abrogation of phenazine produc-
tion after deletion of the gene in Pantoea agglomerans Eh1087.6
The crystal structures presented here, together with mutagen-
esis data (Table S4), allow us to propose a catalytic mechanism
in which PhzA/B accelerates multiple reactions, leading from
ketone 4 to the rearranged condensation product 5a (Scheme 2
and Figure 3). The complex with 3 shows that the catalytic cycle
is initiated by binding of the first ketone molecule to R41 and
possibly also to S77 by its carboxylate group. Interaction of
the 2-amino and 3-oxo groups with E140 positions the ring.
Next, the second substrate molecule binds to R160* and H73,
as seen in the complex with 10. This is also expected to anchor
the mobile C-terminus, which would then shield the active site
from the solvent. The rings of the two ketones are positioned
at a nearly perpendicular angle, and the molecules approach
each other such that a nucleophilic attack by the 2-amino group
of the second ketone at C3 of the first can start the condensation
reaction (Scheme 2a and Figure 3a). The small structural
alterations required for this attack to follow a Bu¨rgi-Dunitz
trajectory25 seem easily achievable due to the mobility of R41,
S77, and E140 mentioned above. Catalysis is brought about by
the indispensible E140, which protonates the negatively charged
9
17058 J. AM. CHEM. SOC. VOL. 130, NO. 50, 2008

The Phenazine Biosynthesis Protein PhzA/B
A R T I C L E S
Scheme 2. Catalytic Proposal for PhzA/B Showing the Most Important Proton Transfer Reactionsa
a THI1/2, first and second tetrahedral intermediate; I1, first imine. The stages shown here correspond to those shown in Figure 3.
oxygen atom of the resulting tetrahedral intermediate (Scheme
2b and Figure 3b). The environment of E140 is hydrophobic,
which will increase its pKa, as in other members of the KSI/
NTF2 family, who often also possess a catalytically important
residue at this position or in the immediate vicinity.26 Loss of
water concludes formation of the first imine bond, which is in
the less favored cis conformation because of the relative initial
orientation of the substrate molecules. As a consequence of this
first condensation reaction, the 2-amino group of the first ketone
comes closer to the carbonyl group of the second, triggering
formation of the tricycle (Scheme 2c and Figure 3c). This
second, intramolecular condensation will be strongly facilitated
and may not require catalysis, but the model suggests that the
negative charge of the associated tetrahedral intermediate is
neutralized by H73 and S77 (Figure 3d). To participate in
catalysis, S77 must adopt the second conformation observed in
the complex with 11, and the positive charges of the nearby
R38 and R41 probably decrease its pKa for this task. While
mutation of H73 had only a minor impact on activity, S77 was
absolutely required, but it is unclear at present whether this is
due to reduced affinity for the first substrate molecule or because
of an incapacity to act in catalysis. Interestingly, H73 and S77
are both substituted with leucine in the PhzA proteins of
Pseudomonas, and PhzA was indeed found inactive in our assays
(Figure S9). We hypothesize that PhzA and PhzB from
Pseudomonas, due to their high similarity, can form het-
erodimers with only one catalytically competent active center.
This could increase the efficiency of phenazine biosynthesis in
ViVo because of the peculiarity that the condensation step
requires two identical substrate molecules. The homodimeric
B. cepacia PhzA/B protein possesses four binding sites for
ketone 4, two of which, according to the crystal structures
presented here, have high affinity but are located in the separate
active centers. As a consequence, the high-affinity binding site
of the second active center reduces the effective concentration
of ketone at the second, lower affinity substrate binding site of
the first active center, leading to a decreased condensation rate
at the substrate concentrations anticipated in the cell.
While the present data do not allow us to unequivocally rule
out that the two imine bonds form simultaneously, the model
indicates that PhzA/B could also participate in the final
“maturation” of double imine 5 by catalyzing the shift of one
of the protons in the R-position to the imine bonds to give 5′
(Schemes 1 and 2e, f and Figure 3e, f). The proximity and
mobility of E140 make this residue ideally suited for this
function. Rearrangement within PhzA/B is advantageous, since
contact with the solvent is avoided. However, an intermediate
5′ cannot be observed, probably because the final proton shifts
that convert 5′ to 5a are facile due to the stability gained by
conjugating the four double bonds.
The observation that oxygen participates in the spontaneous
oxidative decarboxylation of the product of PhzA/B is a new
aspect of phenazine biosynthesis. It suggests that phenazine
production is limited to aerobic lifestyle, which is in contradic-
tion to a recent study demonstrating that P. aeruginosa still
produced pyocyanin when oxygen was depleted.9 However, the
9
J. AM. CHEM. SOC. VOL. 130, NO. 50, 2008 17059

A R T I C L E S
Ahuja et al.
nonenzymatic nature of the oxidative decarboxylation makes it
possible that other chemical oxidants participate in this process.
The detection of phenazine-1,6-dicarboxylic acid (9) and
unsubstituted phenazine (8) as byproducts demonstrates that 5a
can follow several routes to gain aromaticity. These involve
either spontaneous oxidative decarboxylation, leading to 8 when
occurring twice, or oxidation without loss of CO2, leading to 9.
The later implies participation of a cofactor-dependent oxidase,
probably the FMN-containing PhzG of the phz operon, and we
suggest that bacteria that utilize 9 for the production of strain-
specific phenazine derivatives show different relative activities
of the enzymes PhzA/B and PhzG.
Because of the requirement of PhzA/B in ViVo, the synthetic
probes used for functional analysis in this study could act as
lead compounds against phenazine biosynthesis in infectious
disease. Using the complex with 11 as a guide, we are currently
synthesizing improved inhibitors of PhzA/B.
(Qiagen), using a protein concentration of 15 mg/mL. The
optimized setup consisted of a 1 µL + 1 µL hanging drop at 12
°C equilibrated against a 500 µL reservoir of 16-20% (w/v)
PEG 3350, 0.2 M NH4-acetate, and 0.1 M Bis-Tris pH 6.1-6.7.
Cryoprotection was achieved by briefly washing in mother liquor
with PEG 3350 increased to 25% (w/v). Both crystal forms
appeared under similar conditions and sometimes in the same
drop, with the C2221 crystals having a tendency to form at lower
pH and lower precipitant concentration. Complexes were
obtained by overnight soaking of the P3221 crystals in mother
liquor supplemented with 10 mM 2-(cyclohexylamino)benzoic
acid (11) or 50 mM of the other ligands. trans-2,3-Dihydro-3-
hydroxyanthranilate (3) was a kind gift from DSM Anti-
Infectives, and racemic 3-oxocyclohexanecarboxylic acid (10)
was from Sigma.
All data were collected on beamline X10SA of the Swiss
Light Source (Villigen, Switzerland). The structure was solved
from selenium-edge SAD data of a P3221 crystal. The C2221
structure was solved by molecular replacement. Full data
collection and refinement statistics are shown in Table S2.
5.5. HPLC-Coupled Mass Spectrometry. Turnover of 3
(0.1-5 mM) in 50 mM Tris-HCl pH 7.5, supplemented with
10 µM caffeine as internal reference, was initiated with 1 µM
PhzF and varying concentrations of PhzA/B at 10 °C. Reaction
mixtures were analyzed by an Agilent 1100 HPLC system
coupled to a diode array detector and an LTQ mass spectrometer
(Thermo Electron Corporation) operated with atmospheric
pressure chemical ionization in the positive mode.
4. Conclusion
The formation of the tricycle in phenazine biosynthesis
involves the condensation of two identical amino-cyclohexenone
precursors 4 to a double imine 5. While this bimolecular reaction
does not require catalysis in Vitro, it is catalyzed by the small
dimeric KSI/NTF-2 protein PhzA/B in ViVo, which acts as an
acid/base catalyst and significantly increases the reaction rate
at the low substrate concentration expected to exist in the
bacterial cell. Condensation is accompanied by rearrangement
reactions to yield an asymmetric molecule with four conjugated
double bonds 5a, thereby preventing back-hydrolysis to 4. The
rearranged condensation product is oxidation-sensitive and
undergoes oxidative decarboxylation in a nonenzymatic reaction
involving dissolved oxygen, explaining why phenazine-1-
carboxylic acid (7) and not phenazine-1,6-dicarboxylic acid (9)
is the major product of phenazine biosynthesis. However, the
detection of unsubstituted phenazine (8) and of phenazine-1,6-
dicarboxylic acid (9) in our in Vitro experiments indicates that
condensation product 5a can follow different oxidative pathways
to gain aromaticity, making phenazine biosynthesis promiscuous
to a certain extent.
1-µL samples were injected at either 11-min intervals and
separated on a Waters Atlantis dC18 column (3 µm, 4.6 × 100
mm) or at 2-min intervals without column separation. The
column was developed with a H2O/acetonitrile gradient supple-
mented with 0.1% formic acid. 2.5 µL injections were applied
in H218O labeling experiments. H218O was from Isotec.
Mass spectra were analyzed with Xcalibur (Thermo Electron
Corporation) by referencing the peak area of each species to
that of caffeine except in the case of experiments without column
separation, where caffeine could not be quantified.
5.6. Online HPLC-Coupled 1H NMR Spectroscopy. Experi-
ments were carried out on a Varian Inova 600 MHz NMR
spectrometer equipped with a pulse field gradient triple reso-
nance LC probe head coupled to a Varian ProStar HPLC system.
A ProntoSIL 120-5-C18-Aq 5 µm column (Bischoff Analysen-
technik) was used to separate 100 µL of a reaction mixture (100
mM of 3, 10 µM PhzF and 20 µM PhzA/B in 250 mM Tris-
HCl pH 7.5) after 40 min of incubation. The column was
developed in a gradient of degassed H2O and acetonitrile
(Riedel-deHae¨n Chromasolv, HPLC-NMR grade) supplemented
with 0.1% TFA. On-flow and stop-flow experiments were
carried out with WET suppression of the water and acetonitrile
resonances.29
5. Materials and Methods
For full experimental details, refer to the online Supporting
Information.
5.1. Cloning, Protein Expression and Purification, and
Mutagenesis. Proteins were expressed from pET15b in E. coli
Rosetta2 pLysS (Merck Biosciences). Seleno-methionine label-
ing was achieved by suppression of methionine biosynthesis in
M9 media.27 Mutations were introduced with the QuikChange
II XL system (Stratagene). Purification was achieved by Ni2+
affinity chromatography followed by a size-exclusion step.
-
5.2. Synthesis of 2-(Cyclohexylamino)benzoic acid (11). Com-
pound 11 was synthesized by copper-catalyzed amination of
2-bromobenzoic acid with cyclohexylamine following a pub-
lished protocol28 and purified by flash chromatography.
5.3. Isothermal Titration Calorimetry. Calorimetric experi-
ments were carried out with a VP-ITC system (MicroCal LLC)
by titrating 1 mM ligand in 20 mM Tris-HCl pH 8, 150 mM
NaCl to 0.1 mM PhzA/B in the same buffer at 25 °C. Data
were analyzed with Origin (OriginLab Corporation).
5.7. Oxygen Measurements and H2O2 and CO2 Detection.
Oxygen consumption was followed with a Clark electrode in a
sealed chamber at 25 °C. Turnover of 1 mM of 3 in 50 mM
Tris-HCl pH 7.5 was initiated with 1 µM PhzF and varying
amounts of PhzA/B (1-10 µM). Mutants of PhzA/B were
assessed at 10 µM. The influence of metal ions was tested with
Chelex beads or EDTA. H2O2 production was probed with
catalase or Amplex Red (Invitrogen).
5.4. Crystallization, Data Collection, Structure Solution,
and Refinement. Initial crystallization conditions were sought
at 20 °C with screens “The Classics” and “The PEG Suite”
The generation of CO2 was tested with a selective electrode
(Microelectrodes Inc.). One millimolar 3 was reacted with 3
µM PhzF and varying amounts of PhzA/B.
9
17060 J. AM. CHEM. SOC. VOL. 130, NO. 50, 2008

The Phenazine Biosynthesis Protein PhzA/B
A R T I C L E S
Acknowledgment. We thank Roger Goody for his generous
support of this project. Petra Geue and Christiane Pfaff are
acknowledged for excellent technical assistance, Georg Holtermann
for reviving a homemade Clark electrode from the 1970s, Aybike
Yektaoglu for synthesizing 2-(cyclohexylamino)benzoic acid, and
Bernhard Griewel for measuring NMR spectra. DSM Anti-
Infectives B.V. (Delft, The Netherlands) made a generous contribu-
tion of pure trans-2,3-dihydro-3-hydroxyanthranilate. We thank the
X-ray communities of the Max-Planck-Institutes of Molecular
Physiology (Dortmund, Germany) and for Medical Research
(Heidelberg, Germany) for collecting diffraction data at the Swiss
Light Source of the Paul Scherrer Institute (Villigen, Switzerland),
which gave us generous access and support for beamline X10SA.
This project was sponsored in part by grants of the Bioband of the
University of Dortmund to R.B. and W.B. and of the Deutsche
Forschungsgemeinschaft to W.B. (BL587).
Supporting Information Available: . Full experimental details,
protein sequence data, additional HPLC-coupled mass spec-
trometry data, 2D-NMR spectra of ketone condensation product
5a, CO2-detection, oxygen depletion experiments, H2O2 detec-
tion, results of ITC experiments, 2FOFC electron density of the
C-termini in P3221 and C2221 crystals, |FOFC| electron density
of ligands, additional structural presentations, crystallographic
data collection and refinement statistics, list of related protein
structures, and activity of mutants. This material is available
free of charge via the Internet at http://pubs.acs.org.
JA806325K
9
J. AM. CHEM. SOC. VOL. 130, NO. 50, 2008 17061