Complexity generation in fungal peptidyl alkaloid biosynthesis: Oxidation of fumiquinazoline A to the heptacyclic hemiaminal fumiquinazoline C by the flavoenzyme Af12070 from aspergillus fumigatus

Article

pubs.acs.org/biochemistry

Complexity Generation in Fungal Peptidyl Alkaloid Biosynthesis:

Oxidation of Fumiquinazoline A to the Heptacyclic Hemiaminal

Fumiquinazoline C by the Flavoenzyme Af12070 from

Aspergillus fumigatus

Brian D. Ames,^†,⊥Stuart W. Haynes,^†,⊥Xue Gao,^‡Bradley S. Evans,^§Neil L. Kelleher,^∥Yi Tang,^‡

and Christopher T. Walsh*^,†

^†Department of Biological Chemistry & Molecular Pharmacology, Harvard Medical School, Boston, Massachusetts 02115,

United States

^‡Department of Chemical & Biomolecular Engineering, Department of Chemistry & Biochemistry, University of California,

Los Angeles, Los Angeles, California 90095, United States

^§Institute for Genomic Biology, University of Illinois, Urbana, Illinois 61801, United States

^∥Department of Chemistry, Northwestern University, Evanston, Illinois 60208, United States

S

* Supporting Information

ABSTRACT: The human pathogen Aspergillus fumigatus

makes a series of fumiquinazoline (FQ) peptidyl alkaloids of

increasing scaffold complexity using ^L-Trp, 2 equiv of L-Ala,

and the non-proteinogenic amino acid anthranilate as building

blocks. The FQ gene cluster encodes two non-ribosomal

peptide synthetases (NRPS) and two flavoproteins. The

trimodular NRPS Af12080 assembles FQF (the first level of

complexity) while the next two enzymes, Af12060 and

Af12050, act in tandem in an oxidative annulation sequence

to couple alanine to the indole side chain of FQF to yield the imidazolindolone-containing FQA. In this study we show that the

fourth enzyme, the monocovalent flavoprotein Af12070, introduces a third layer of scaffold complexity by converting FQA to the

spirohemiaminal FQC, presumably by catalyzing the formation of a transient imine within the pyrazinone ring (and therefore

acting in an unprecedented manner as an FAD-dependent amide oxidase). FQC subsequently converts nonenzymatically to the

known cyclic aminal FQD. We also investigated the effect of substrate structure on Af12070 activity and subsequent cyclization

with a variety of FQA analogues, including an FQA diastereomer (2′-epi-FQA), which is an intermediate in the fungal

biosynthesis of the tremorgenic tryptoquialanine. 2′-epi-FQA is processed by Af12070 to epi-FQD, not epi-FQC, illustrating that

the delicate balance in product cyclization regiochemistry can be perturbed by a remote stereochemical center.

he secondary metabolome of the human fungal pathogen

T_{Aspergillus fumigatus is extensive, including several}

polyketides, but notably a range of peptidyl alkaloids that

include gliotoxin, fumitremorgins, fumigaclavines, fumagillins,

and fumiquinazolines.^1,2The fumiquinazolines (FQs) are

signature peptidyl alkaloids produced by all 40 A. fumigatus

strains that have been examined for natural products, although

their physiologic function remains to be elucidated.¹The FQs

are a group of compounds that display varying degrees of

structural complexity^3,4and include the tricylic FQF, which is

modified to the more architecturally elaborate framework seen

in FQA and then finally to a third level of complexity in FQC,

which contains a heptacyclic scaffold formed via a bridging

hemiaminal linkage (Figure 1).

the assembly of the fumiquinazolines (Figure 1A).^6,7Our initial

interest in these NRPS systems came as a result of the

identification of fungal adenylation domains that activate

anthranilate (a core building block in the fumiquinazoline

framework), and we have shown that this activity resides in

module one of the trimodular NRPS Af12080.⁶Two adjacent

genes in the cluster, af12060 and af12050, have been expressed

in E. coli and shown to (1) oxygenate the indole side chain of

FQF (Af12060) and then (2) activate ^L-Ala for annulation to

the oxidized indole scaffold (Af12050), thereby constructing

two new C−N bonds.⁷This set of oxidative annulation

operations stereospecifically transforms the bicyclic indole in

FQF to the tricyclic imidazoindolone in the product FQA

(Figure 1B).

The 29 Mb A. fumigatus genome encodes 14 predicted

polyketide and 14 non-ribosomal peptide gene clusters.⁵We

have recently assigned two NRPS genes (af12080 and af12050)

on chromosome 6 of A. fumigatus Af293 as being involved in

Received: August 15, 2011

Revised: September 6, 2011

Published: September 7, 2011

8756

dx.doi.org/10.1021/bi201302w|Biochemistry 2011, 50, 8756−8769

Biochemistry

Article

Figure 1. Gene cluster and proposed biosynthetic route for the production of fumiquinazoline alkaloids from Aspergillus fumigatus Af293.

Figure 2. Examples of FQA-like scaffold modification by intramolecular nucleophilic reaction/rearrangement.

In this work we have characterized Af12070, the fourth

enzyme encoded in the FQ gene cluster. Af12070 takes a

molecule with a fair degree of rotational freedom, which

consists of two distinct moieties (a pyrazinoquinazolinedione

and an imidazoindolone) linked by a single methylene “bridge”

and transforms it into the fixed and highly constrained

heptacyclic scaffold FQC (Figure 1B). This fascinating piece

of biosynthetic origami, though unusual, is not unique to

Af12070; instead, it appears to form part of a small group of

fungal biosynthetic enzymes that build up complex multicyclic

scaffolds. TqaG (which bears 72% amino acid sequence

similarity to Af12070) from the recently characterized

tryptoquialanine pathway in Pencillium aethiopicium,⁸and as

yet unassigned enzymes in Aspergillus versicolor^9,10and an

Acremonium sp.,¹¹appear to catalyze potentially analogous and

equally intriguing transformations en route to tryptoquialanine,

cottoquinazolines, and fumiquinazoline H, respectively (Figure

2). Intriguingly, whole-genome transcriptional profiling has

shown that af12070 and af12050 gene expression is down-

regulated in A. fumigatus Af293 ΔbrlA or ΔstuA mutants,

suggesting that the production of the fumiquinazoline

metabolites is developmentally regulated and tied to con-

idiation (asexual sporulation).¹²

This work forms the final piece of our in vitro studies of

fumiquinazoline biosynthesis, where together the four enzyme

pathway of Af12050 through Af12070 recapitulates all three

levels of FQ metabolite architectural complexity generation. We

show that Af12070 is a flavoprotein amide oxidase (acting on

R-[R′-]-CH-NH-CO-R″) of notable similarity to the more

prevalent amine oxidases (which act on R-[R′-]-CH-NH-R″¹³).

The regiospecific amide oxidation of FQA by Af12070 yields

the architecturally complex FQC as the kinetic product via the

formation of an unusual spirohemiaminal linkage. FQC is also

shown to slowly convert to the known cyclic aminal metabolite

FQD nonenzymatically. We also examine the Af12070

processing of 2′-epi-FQA and 11′-dimethyl-2′-epi-FQA, which

are known intermediates in the tryptoquialanine pathway of

P. aethiopicum that share a large degree of structural similarity

to FQA⁸(Figure S1). These studies allowed us to observe the

effects of these remote site substitutions on the cyclic aminal/

hemiaminal product ratio and investigate the affect of these

remote structural changes on product outcome.

8757

dx.doi.org/10.1021/bi201302w|Biochemistry 2011, 50, 8756−8769

Biochemistry

EXPERIMENTAL PROCEDURES

Article

appropriate antibiotics to an OD₆₀₀between 0.4 and 0.8, and

the temperature was lowered to 16 °C prior to induction with

0.1 mM IPTG. Cells were harvested 18−24 h postinduction by

centrifugation, suspended in lysis buffer (25 mM Tris-HCl [pH

7.5], 300 mM NaCl, 20% glycerol, 0.1% Tween 20), and lysed

using an EmulsiFlex-C5 homogenizer (Avestin). Insoluble

material was removed by centrifugation (35000 g), and soluble

protein was applied to 1−2 mL of Ni-NTA agarose (Qiagen)

equilibrated in lysis buffer. Ni-affinity purification was

performed by batch binding protein for 30 min at 4 °C,

washing the Ni-resin twice with 20 mL of buffer A (25 mM

Tris-HCl [pH 7.5], 300 mM NaCl, 10% glycerol, 0.5 mM

EDTA, 25 mM imidazole), and then eluting protein using three

5 mL volumes of buffer A including 250 mM imidazole. The

elutions were pooled and concentrated using a centrifugal

filtration device (30K MWCO, Amicon). Additional purifica-

tion of Af12070 Δ1−24 was performed using Strep-Tactin

Superflow Plus resin (Qiagen). Concentrated protein post Ni-

NTA (≈2 mL) was added to 1 mL of resin for batch binding

for 1−2 h at 4 °C, the resin was then washed with three 1 mL

volumes of buffer A (pH 8.0/NaOH), and bound protein was

eluted with five 0.5 mL volumes of buffer A including 2.5 mM

desthiobiotin. Elutions containing target protein were con-

centrated by centrifugal filtration. For both full-length and

truncated versions of Af12070, the concentrated protein was

either stored at 4 °C for immediate use or flash-frozen in liquid

N₂and stored at −80 °C.

■

General Materials and Methods. Oligonucleotide pri-

mers were purchased from Integrated DNA Technologies

(Coralville, IA). PCR reactions were carried out using Phusion

High-Fidelity PCR MasterMix (New England Biolabs). Plasmid

DNA was propagated in E. coli XL1 Blue (Stratagene) and

prepared using the QIAprep Spin Mini Kit (Qiagen). DNA

sequencing to confirm the correct construction of expression

vectors was performed by Genewiz (South Plainfield, NJ).

Protein concentration was determined spectrophotometrically

using theoretical extinction coefficients obtained from the

online ProtParam tool.¹⁴An Agilent Technologies 6520

Accurate-Mass QTOF instrument was used for high-resolution

LC-MS analysis, and a Beckman Coulter System Gold

instrument equipped with diode-array detection was used for

reverse-phase HPLC. NMR data were collected on a Varian

600 MHz spectrometer using the residual solvent peak from

incomplete deuteration as internal standard (CDCl₃, δ 7.26).

Additional Details for Bioinformatic Analysis. Se-

quence alignments were performed with ClustalW¹⁵and

colored using GeneDoc.¹⁶Homology modeling was performed

with HHpred.¹⁷Superimposition of Af12070 with VAO-family

oxidases of known structure allowed for manual docking of

FAD into the active-site of Af12070). NetNGlyc 1.0 was used

to predict Asn glycosylation sites, and the three sites identified

were checked by homology modeling to ensure the residues

were surface exposed. Of note is that while these servers are

dedicated to analysis of mammalian proteins, the N-X-S/T

sequence motif and the transfer of the GlcNAc substrate to the

polypeptide are conserved between mammalian and fungal

systems.¹⁸

For all results sections describing the in vitro characterization

of Af12070, the version of the protein referred to or utilized

in the assays is the N-terminal deletion construct Af12070

Δ1−24.

Identification of the Covalent Attachment Site of FAD

to Af12070 by Mass Spectrometry. Identification of the

flavin cofactor as FAD rather than FMN and localization of that

modification to His105 was performed using trypsin digestion

followed by capillary LC-MSⁿ. All data were recorded on a

custom 11T LTQ-FT Ultra (Thermo-Fisher Scientific)

equipped with an Eksigent 2D-LC nanocapillary LC system

and a Picoview 550 nanospray source (New Objective). Data

were analyzed manually using the Qualbrowser application of

Xcalibur (Thermo-Fisher Scientific). 20 μL of 40 μM Af12070

Δ1−24 was denatured using 5 μL of protein extraction reagent

4 (Sigma) for 10 min and then diluted 4-fold with 10 mM

NH₄HCO₃, pH 8.0, prior to application to an immobilized

trypsin column prepared according to the manufacturer’s

instructions (Sigma). Digestion was allowed to proceed for

1 h before the digested peptides were eluted. The digested

peptides were desalted using diafiltration over an Amicon

Ultracel 10K MWCO filter (Millipore). The retentate was then

concentrated to dryness in a speedvac. Concentrated peptides

were reconstituted in 50 μL of 0.1% formic acid in water, and

20 μL of this sample was injected onto a 75 μm × 80 mm

Biobasic C₈Picofrit column (New Objective). Peptides were

eluted using a linear gradient of 5% B to 75% B over 60 min at

a flow rate of 300 nL/min (A = 0.1% formic acid in 95% H₂O/

5% MeCN, B = 0.1% formic acid in MeCN). MS data were

acquired in profile positive ion mode at resolution 50 000 for

FT scans and positive centroid mode for ion trap scans. Data

dependent or targeted MSⁿscans were used as appropriate,

where n was 2, 3, or 4 and indicates the number of stages of

fragmentation used in tandem mass spectrometry.

Cloning, Expression, and Purification of Af12070. A.

fumigatus Af293 cDNA was used for all cloning. A full-length

version of Af12070 was cloned into pET-30 Ek/LIC to encode

a 536 residue (59 kDa) protein with an N-terminal 6x-His tag

using the PCR primers: 5′-GACGACGACAAGatgcagta-

catccccttcctgatctc-3′ (forward), 5′-GAGGAGAAGCCCGGT-

TAcagactcaacggaatcggagcc-3′ (reverse, bases complementary

to gene are in lowercase type, underline indicates a stop

codon). DNA sequencing revealed that the terminal (3′)

boundary for the first intron of the cloned gene was different

than the 3′ boundary identified for AFUA_6g12070 in the

NCBI database (GenBank accession XM_745993). The first

intron of the database entry ends at base 3,020,413 of the

genomic DNA for Af293 chromosome 6 (accession

NC_007199), whereas the termination site for the sequenced

clone is 12 nucleotides downstream at base 3,020,425.

Consequently, the cloned transcript encodes a protein that is

four amino acids shorter than the database entry

AFUA_6g12070, and the amino acid numbering subsequent

to residue 87 is different.

A truncated version of Af12070 which removes the first 24

amino acids of the full-length construct (Af12070 Δ1−24), was

cloned into pET-52b 3C/LIC to encode a 511 residue (56

kDa) protein with an N-terminal Strep-tag II and C-terminal

10x-His tag using the PCR primers: 5′-CAGGGACCCGGTag-

caacaccaggattgatgggaatg-3′ (forward), 5′-GGCACCA-

GAGCGTTcagactcaacggaatcggagcc-3′ (reverse).

Protein overproduction was performed in E. coli Origami-

(DE3) cells in order to promote disulfide bond formation. Both

constructs were expressed and protein purified in a similar

manner: 4 L of cells were grown at 37 °C in LB including the

Preparation and Purification of FQF. In order to

generate various substrates for the in vitro testing of Af12070,

8758

dx.doi.org/10.1021/bi201302w|Biochemistry 2011, 50, 8756−8769

Biochemistry

Article

the precursor FQF was synthesized (as previously described⁶)

or obtained by growth and extraction of P. aethiopicum ΔgsfA/

ΔtqaH (double knockout reported by Gao et al.⁸). Spores of P.

aethiopicum ΔgsfA/ΔtqaH were harvested from a single plate

(100 × 20 mm) following 4-day growth at 30 °C on glucose

minimal medium (GMM) agar and used to inoculate 3 L of

YMEG liquid medium (4 g/L yeast extract, 10 g/L malt extract,

5 g/L glucose) for stationary culture at 25 °C for 7 days (50 μL

of spore stock to ≈100 mL media placed in 150 × 25 mm

plates). The resulting fungal mat was separated from the media

and extracted three times with 1 L of ethyl acetate. The extract

was concentrated under vacuum and washed twice with ethyl

acetate:water (2:1), and the combined ethyl acetate layers were

dried under vacuum. The residue was purified by silica gel

chromatography using the solvent system ethyl acetate:hexane

(3:1).

Preparation and Purification of FQA, 2′-epi-FQA, and

11′-Dimethyl-FQA. The enzymatic synthesis of FQA was

performed in a 30 mL reaction by combining 5 μM holo-

Af12050, 2.5 μM holo-Af12060, 1 mM ATP, 2 mM MgCl₂,

1 mM ^L-Ala, 2 mM NADH, and 250 μM FQF in NaPi buffer

(50 mM sodium phosphate [pH 7.5], 100 mM NaCl and 5%

glycerol). 2′-epi-FQA was generated using a similar enzymatic

reconstitution procedure, but with 5 μM holo-TqaB (the

trimodular NRPS of the tryptoquialanine-pathway¹⁹) in place

of Af12050. 11′-Dimethyl-FQA was also prepared similarly, but

with 5 μM holo-Chimera2 (which contains the A-T domains of

TqaB for 2-AIB activation and the C-domain of Af12050¹⁹) in

place of Af12050 and 1 mM 2-AIB in place of ^L-Ala. Reactions

were incubated overnight (≈16 h) at room temperature,

quenched by adding an equal volume of MeCN, and

centrifuged to remove precipitate. The supernatant following

centrifugation was concentrated under vacuum to remove

MeCN, flash frozen, and then lyophilized. The resulting residue

was taken up in 5 mL of 25% MeCN, filtered (0.2 μm PTFE

membrane), and purified by reverse-phase HPLC (Phenomen-

ex Luna C18, 250 × 21.2 mm, 10 μm) with detection at 254

nm. Solvent systems A (water plus 0.1% TFA) and B (MeCN

plus 0.1% TFA) were held at 25% B for 1 min and then run

over a linear gradient of 25−53% over 30 min, followed by a

gradient of 53−95% B over 1 min before a holding at 95% B for

5 min; the column was then equilibrated back to initial

conditions by returning to 25% B and holding for 8 min.

FQA HRMS: m/z calculated for C₂₄H₂₃N₅O₄: 446.1823

[M + H]⁺. Found: 446.1826.

initial rate data were used for calculation of apparent velocity

and turnover number (taking into account 46% holoprotein).

HPLC and LC-MS Assays for Af12070. For time course

analysis of substrate utilization by Af12070, reactions (typically

400 μL) were setup containing 2 μM Af12070 and 200 μM

substrate (FQA, 2′-epi-FQA, 11′-dimethyl-FQA, or 11′-

dimethyl-2′-epi-FQA) in NaPi buffer. Reactions were initiated

with enzyme and time points taken between 0.1 and 20 h by

quenching 50 μL aliquots with an equal volume of MeCN. The

precipitant was removed from quenched reactions by

centrifugation, and 20 μL samples of the supernatant were

injected for HPLC using an Alltima C18 column (150 × 4.6

mm, 5 μm), with detection at 254 nm. Solvent system A (water

plus 0.1% TFA) and B (MeCN plus 0.1% TFA) was held at

25% B for 1 min and then run over a linear gradient of 25−55%

B over 20 min, followed by a gradient of 55−95% B over 1 min

before a holding at 95% B for 2.5 min; the column was then

equilibrated back to initial conditions by returning to 25% B

and holding for 5 min. To determine the rate of FQA

utilization, for each time point the peak corresponding to

substrate was integrated and converted to a concentration using

a standard curve made by injecting 20 μL samples of known

concentration. Plotting the concentration of FQA vs time

allowed for the calculation of initial rate data. Turnover was

calculated based on 46% holo-Af12070.

Genetic Manipulation and Transformation of P.

aethiopicum and Characterization of Metabolites. tqaB

was deleted in the wild-type P. aethiopicum strain using bar as a

selection marker as previously described.⁸pBARGPE1

(obtained from the Fungal Genetics Stock Center) is a fungal

expression vector and contains the A. nidulans gpdA promoter

and trpC terminator.²⁰The whole trpC promoter and bar gene

fragment were replaced by the trpC and Zeocin genes using SpeI

and PmlI restriction sites to yield the pZeoGPE1. af12050 was

amplified and digested with BamHI and AfeI and inserted into

pBARGPE1 (digested with BamHI and EcoRV) to yield

af12050-pZeoGPE1 plasmid. The gene for chimera2 was

amplified and digested with AfeI and inserted into pBARGPE1

(digested with EcoRV) to yield chimera2-pZeoGPE1 plasmid.

∼20 μg of af12050-pZeoGPE1 or chimera2-pZeoGPE1 plasmid

was transformed into the P. aethiopicum ΔtqaB strain using

Zeocin as a second selection. Poly(ethylene glycol)-mediated

transformation of P. aethiopicum was performed essentially as

previously described.^8,21Genomic DNA from P. aethiopicum

transformants was used for PCR screening. Integration of the

af12050/chimera2-overexpression cassette was confirmed by

PCR. For small-scale analysis, the P. aethiopicum wild-type and

transformants were grown in stationary YMEG liquid culture for

4 days at 25 °C. The cultures were extracted with equal volumes

of ethyl acetate and evaporated to dryness. The dried extracts

were dissolved in methanol for LC/MS analysis. LC/MS spectra

were obtained on a Shimadzu 2010 EV liquid chromatography

mass spectrometer using positive and negative electrospray

ionization and a Phenomenex Luna 5 μm, 2.0 mm ×100 mm

C18 reverse-phase column. Samples were separated on a linear

gradient of 5−95% CH₃CN in water (0.1% formic acid) for 30

min at a flow rate of 0.1 mL/min followed by isocratic 95%

CH₃CN in water (0.1% formic acid) for another 15 min.

Isolation and Characterization of FQC/FQD and

Analogues. (i) FQC and FQD. A 20 mL reaction containing

5 μM Af12070 and 200 μM FQA was incubated overnight at

room temperature, and the sample was quenched, processed,

and purified as described above for the scaled production of

2′-epi-FQA HRMS: m/z calculated for C₂₄H₂₃N₅O₄:

446.1823 [M + H]⁺. Found: 446.1824.

11′-Dimethyl-FQA HRMS: m/z calculated for C₂₅H₂₅N₅O₄:

460.1979 [M + H]⁺. Found: 460.1983.

For spectroscopic characterization and assignments of

fumiquinazoline A analogues and comparison to fumiquinazo-

line A see Table S4.

O₂Consumption by Af12070. A Hansatech oxygen

electrode (S1 Clark type) with an Oxygraph control unit was

used to monitor O₂levels in sample reactions reconstituting

Af12070 oxidation of FQA. A reaction volume of 300 μL was

used and contained 200 μM FQA and 2 μM Af12070 in NaPi

buffer. Substrate, buffer, and water were combined in the

reaction vessel (with stirring set to 100 rpm and a temperature

of 25 °C), and the baseline was allowed to stabilize before

initiating reactions by addition of enzyme. Controls were run

that omitted either enzyme (with buffer injected in its place) or

substrate. Reactions and controls were run in triplicate, and

8759

dx.doi.org/10.1021/bi201302w|Biochemistry 2011, 50, 8756−8769

Biochemistry

Article

FQA-(like) substrates. For FQC and FQD spectroscopic

characterization and assignments see Table S1.

HRMS: m/z calculated for C₂₅H₂₃N₅O₄: 458.1823 [M + H]⁺.

Found: 458.1822.

FQC HRMS: m/z calculated for C₂₄H₂₁N₅O₄: 444.1666

[M + H]⁺. Found: 444.1667.

RESULTS

■

FQD HRMS: m/z calculated for C₂₄H₂₁N₅O₄: 444.1666

[M + H]⁺. Found: 444.1662.

Bioinformatic Analysis of Af12070 and TqaG. To gain

insight into the possible function of the remaining ORF in the

FQ gene cluster, Af12070, we first turned to bioinformatic

analysis. Af12070 and TqaG (46% identity/72% similarity)

belong to the vanillyl alcohol oxidase (VAO) family of FAD-

dependent oxidoreductases.²²A characteristic feature of the

VAO family is their propensity to form mono- and bicovalent

linkages to the isoalloxazine of FAD, a self-catalytic process

(termed autoflavinylation) that increases the redox potential of

the enzyme, enhances structural integrity and protein stability,

and protects the coenzyme from modification and enzyme

inactivation.²³Sequence alignment and homology modeling

predict that Af12070 and TqaG are monocovalent flavopro-

teins, which form a histidyl linkage to the 8α methyl of FAD

(Figure 3). Homologues of Af12070, identified by a BLASTP

search, are predicted to possess both mono- and bicovalent

(ii) 2′-epi-FQD. A 25 mL reaction containing 2 μM Af12070

and 250 μM 2′-epi-FQA was incubated overnight at room

temperature, and the sample was quenched, processed, and

purified as described above for the scaled production of FQA-

(like) substrates. For 2′-epi-FQD spectroscopic characterization

and assignments see Table S2.

HRMS: m/z calculated for C₂₄H₂₁N₅O₄: 444.1666 [M + H]⁺.

Found: 444.166.

(iii) 1,1′-Dimethyl-FQC. A 25 mL reaction containing 2 μM

Af12070 and 300 μM 1,1′-dimethyl-FQA was incubated

overnight at room temperature, and the sample was quenched,

processed, and purified as described above for the scaled

production of FQA-(like) substrates. For 1,1′-dimethyl-FQD

spectroscopic characterization and assignments see Table S3.

Figure 3. Bioinformatic analysis of FAD binding by Af12070 and homologous proteins. (A) Homology model of Af12070 illustrating the predicted

covalent attachment of FAD to His105 via an 8α-N1-histidyl linkage (black dashed line), the position of residue Glu164 (which is replaced by a

cysteine in the bicovalent FAD proteins in order to generate an additional 6-S-cysteinyl attachment), and two cysteines (residues 156 and 338) which

may form a disulfide linkage near the dimethylbenzene of FAD. (B) 2D representation of the predicted 8α-N1-histidyl FAD linkage of Af12070. (C)

Sequence alignment of select vanillyl alcohol oxidase (VAO) family proteins homologous to Af12070. The top 17 sequences (labeled as groups

[Grp] 1, 2, and 3) were obtained from a BLASTP search using Af12070 as the query. The bottom two sequences are structural homologues

identified by HHpred and represent proteins with bicovalent FAD (3FW9) or monocovalent FAD (2BVF).

8760

dx.doi.org/10.1021/bi201302w|Biochemistry 2011, 50, 8756−8769

Biochemistry

Article

links to FAD (Figure 3C, Grp1 and Grp2). For the putative

bicovalent flavoprotein homologues (which constitute the

majority of the top 100 BLASTP results), the residue

a

Glu164 of Af12070 is replaced by a Cys residue, this Cys

forms the 6-S-cysteinyl FAD linkage in the known bicovalent

flavoprotein (S)-reticuline oxidase (PDB id: 3F9W) (Figure

3C). The BLAST search also shows that two Cys residues (166

and 338 of Af12070), located adjacent to the dimethylbenzene

portion of FAD and above the entrance to the active site, are

highly conserved within groups 1 and 2 (Figure 3C). Though

their inter-residue distance is large in the homology model

(7.5 Å), in actuality it is plausible to envision that these Cys

residues may form a disulfide bond, thus conferring even

greater structural stability on this set of proteins. A small

fraction of the top 100 results from BLAST (less than 10%)

have a Ser or Thr at the position equivalent to Cys166 of

Af12070 (Figure 3C, Grp3), consequently eliminating the

possibility of a disulfide at this position and setting this group

apart from other Af12070 homologues.

Bioinformatics analyses also indicate that Af12070 may

contain an N-terminal signal sequence and multiple N-

glycosylation sites. Sequence alignment and modeling identify

an unstructured region of Af12070 and TqaG which extends

≈35 amino acids beyond the FAD-binding domain (Figure S2).

PSORT and SignalP analysis predict that the first 17 residues of

Af12070 constitute a cleavable hydrophobic signal sequence

that directs protein export. A signal peptide is also predicted for

TqaG, though the location of the cleavage site is more

ambiguous. N-terminal signal sequences have been reported

for several VAO-family enzymes, including THCA synthase,²⁴

(S)-reticuline oxidase (berberine bridge enzyme),²⁵cytokinin

dehyodrogenase (maize isoform ZmCKX1),²⁶and aryl alcohol

oxidases.²⁷Of these, signal peptide directed secretory

trafficking has previously been described for THCA synthase

and cytokinin dehydrogenase.^24,26The possibility that the

N-terminal signal peptide of Af12070 directs secretion is

supported by the prediction that the enzyme is N-glycosylated

at up to 3 sites (Asn95, 257, and 284) each of which contains

the consensus sequence Asn-X-Ser/Thr and are predicted to be

surface exposed based on 3D modeling.

Taken together, the predicted signal peptide, disulfide bond,

and Asn glycosylation sites strongly suggest that Af12070 is

secreted. Future studies investigating the cellular trafficking of

Af12070 will be of interest in determining if, as predicted, the

enzyme is exported in order to function biosynthetically, for

example by deposition into developing conidia during

sporulation. Notably, as we describe below, the prediction of

this signal sequence was also experimentally consequential for

the heterologous expression of Af12070 as an active, soluble

enzyme.

Recombinant Production and Purification of Af12070

from E. coli. Initial efforts to overproduce the full-length

protein encoded by the gene AFUA_6g12070 (493 amino

acids, abbreviated as Af12070) in E. coli gave some soluble,

active protein, but yields were low and the protein was impure

and possessed low activity. Cloning of a construct that deleted

the first 24 residues of Af12070 (Af12070 Δ1−24) and

expression in E. coli Origami(DE3) led to improved solubility

and yield, increased purity and stability, and higher enzymatic

activity. Af12070 Δ1−24 was purified as a yellow protein of

>90% purity and with a yield of ≈1 mg/L following dual affinity

chromatography (utilizing Ni-NTA and Strep-Tactin resins)

(Figure 4). As purified from E. coli Origami(DE3) cells,

Figure 4. SDS-PAGE analysis of purified Af12070 Δ1−24 (N-terminal

truncation of the first 24 amino acids) with staining by Coomassie

(left) or visualization by fluorescence imaging (right, 365 nm

excitation).

Af12070 Δ1−24 migrates as two closely spaced bands in

nonreducing SDS-PAGE. Upon treatment with reducing agents

the top band disappears and the intensity of the bottom band

increases, suggesting that the doublet is composed of oxidized

(S−S, top band) and reduced (−SH HS−, bottom band) forms

of the protein. As there are only three Cys residues present in

Af12070, the top band likely represents the Cys166-Cys338

disulfide form of the protein as postulated in the previous

section. Though the oxidized species of the protein could be

eliminated by adding chemical reducing agents, converting the

protein from the reduced to oxidized form proved more

difficult; neither prolonged incubation and exposure to air nor

chemical treatment (with H₂O₂or oxidized glutathione)

afforded a significantly increased proportion of the oxidized

species. Attempts at overproducing the full-length or an N-

terminal deletion construct of the homologous TqaG utilizing

either bacterial or yeast expression systems have been

unsuccessful; therefore, we have focused on Af12070 for our

in vitro studies during this work.

Af12070 Is a Covalent Flavoprotein and an O₂-

Dependent FQA Oxidase. Denaturing SDS-PAGE with

fluorescence visualization (prior to Coomassie staining)

confirmed that Af12070 contained covalently bound flavin

(Figure 4). The flavin content of the purified enzyme was

calculated to be 46% by UV−vis analysis (Figure S3). Trypsin

digestion followed by capillary LC-MSⁿlocalized the flavin

attachment site to His105 and confirmed the identity of the

covalent flavin as FAD (Figure 5).

Given that Af12080, 12060, and 12050 act together to

generate FQA (Figure 1) and that FQC and FQD are more

complex scaffolds predicted to arise from enzymatic mod-

ification of FQA, we evaluated if Af12070 would, as predicted,

utilize FQA as a substrate. We generated substrate quantities of

FQA via incubations of synthetic FQF with Af12060/12050⁷

(described in Experimental Procedures) and then turned to

evaluate the ability of O₂to act as a reducible cosubstrate of

Af12070. O₂consumption by Af12070 was determined using

an oxygen electrode; combining 2 μM Af12070 Δ1−24 with

200 μM FQA (in buffer at pH 7.4) led to a 5-fold increase in

8761

dx.doi.org/10.1021/bi201302w|Biochemistry 2011, 50, 8756−8769

Biochemistry

Article

Figure 6. Af12070 Δ1−24 utilizes O₂as an electron acceptor in the

oxidation of FQA. (A) Oxygen electrode trace data representing (i) no

substrate control (buffer plus the addition of 2 μM Af12070 as

indicated), (ii) no enzyme control (buffer including substrate with the

addition of an extra volume of buffer as indicated), and (iii)

reconstitution of FQA oxidation by Af12070 (buffer including FQA

plus the addition of 2 μM Af12070 as indicated). (B) Initial rate data

obtained from triplicate measurements of the experiments in panel A.

FQA synthesis catalyzed in tandem by the monooxygenase

Af12060 and NRPS Af12050.⁷

HPLC Analysis of Af12070 Activity toward Its Natural

Substrate FQA. In order to identify the products resulting

from FQA oxidation in vitro, we followed the enzymatic activity

of Af12070 over time by HPLC. Incubating 2 μM Af12070

Δ1−24 with 200 μM FQA in buffer (pH 7.4) resulted in the

disappearance of FQA (retention time [t_R] = 14.2 min) with

the initial formation of a new peak at t_R= 15.9 min and then the

slow appearance of a second peak at t_R= 14.6 min (Figure

7A,B). Determining the rate of FQA disappearance through

peak integration provided an apparent turnover number for

FQA oxidation of 5.2 min⁻¹, which is consistent with the value

obtained by monitoring oxygen consumption with the oxygen

electrode. The addition of reducing agents to the in vitro

reaction had a variable effect on enzymatic activity, 5 mM

TCEP enhanced activity ≈2-fold, whereas 5 mM DTT

decreased activity ≈3-fold, and 5 mM β-ME essentially

inactivated the enzyme (Figure S4).

Figure 5. LC-MSⁿanalysis of Af12070 Δ1−24 covalent flavin

modification. (A) A [M+4H]⁴⁺ion corresponding to the tryptic

peptide Pro74-Lys104 with attached FAD (m/z = 967.6758) was

found in the LC-MS trace and subsequently targeted for MS². (B) The

MS²spectrum of the peptide from panel A yielded only a single

intense [M+3H]³⁺peak corresponding to the Pro74-Lys104 peptide

with the neutral loss of AMP from the FAD modification (m/z =

1174.2106), and this peak was subsequently targeted for MS³. (C) MS³

analysis localized the FAD modification to the C-terminal region of the

peptide. (D) MS⁴analysis of the y₇⁺ion at m/z 1047.6 yielded several

fragment ions including the y₄⁺at m/z 876.7 and the internal ion [y₇−

b₂₈]⁺at m/z 745.5 that localized the FAD modification precisely to

His105.

O₂-consumption compared to controls lacking enzyme or

substrate (Figure 6). This result indicates that O₂is the electron

acceptor for regeneration of oxidized flavin during the oxidative

half-reaction of the catalytic cycle.²⁸Under the conditions assayed,

and taking into account the observed ratio of holo-Af12070

(46%), the apparent turnover number for FQA oxidation was

4.6 min⁻¹. This rate is of comparable magnitude to the rate of

MS analysis indicated that both of the new peaks that appear

when Af12070 is incubated with FQA had a mass change of −2

Da compared to FQA, a mass that corresponds to two

compounds previously identified from A. fumigatus culture

extracts, FQC and FQD (FQC, [M + H]⁺observed 444.1667,

8762

dx.doi.org/10.1021/bi201302w|Biochemistry 2011, 50, 8756−8769

Biochemistry

Article

Figure 7. HPLC-based timecourse data (with detection at 254 nm) for 2 μM Af12070 Δ1−24 with 200 μM FQA as substrate. (A) HPLC

chromatogram overlay of reaction aliquots quenched at the indicated time by addition of MeCN prior to injection for analysis. (B) Graphical

representation of the time course data obtained by integrating the peak corresponding to FQA, FQC, and FQD. (C) Schematic illustrating the

observed pattern of product formation when Af12070 was incubated in the presence of substrate FQA.

expected 444.1666; FQD, [M + H]⁺observed 444.1666,

expected 444.1666).⁴¹H NMR characterization and compar-

ison with the literature confirmed the identity of these new

peaks as FQC and FQD (Table S1) and led to the peak

assignments as provided in Figure 7A. These products arise

from the intramolecular cyclization of FQA as a result of the

nucleophilic attack of either the free hydroxyl or the secondary

amine of the imidazoindolone moiety on C3 of the

quinazolinone (Figure 7C). The additional bond formed in

FQC generates a bridging 7-membered ring, while in FQD a

bridging 8-membered ring is formed. As illustrated in Figure 7,

FQA is almost entirely converted to FQC before the

appearance of FQD becomes obvious (by ≈0.7 h), and then

FQD slowly grows in (from 0.7 to 20 h) with the concomitant

decrease of FQC.

To evaluate if the slow FQC to FQD transformation was

catalyzed by Af12070 or was nonenzymatic, a sample of purified

FQC was incubated in buffer (pH 7.0) without enzyme.

Nonenzymatic, spontaneous rearrangement to FQD occurred,

along with the appearance of a substantial amount of a new

peak with m/z = 462 which eluted ≈1 min prior to FQD

(Figure S5A). The new peak has a mass ([M + H]⁺observed

462.1772, expected 462.1771) and UV−vis spectral properties

consistent with the previously reported FQB, a C3-hydroxyl

derivative of FQA (3-OH-FQA) that could arise from

hydration of FQD/FQC or more probably hydration of a

transient N2−C3 imine.⁴The accumulation of the hydrate

indicates that the intermolecular addition of water effectively

competes with the intramolecular capture of the imine by the

hydroxyl or secondary amine groups of the imidazoindolone.

Incubation of purified FQD without enzyme also resulted in

interconversion to FQC (to low level) and in the accumulation

of the putative hydration product 3-OH-FQA (Figure S5A).

Comparing the relative amounts of each compound formed (or

remaining) following the week-long incubation revealed that

spontaneous equilibration favors FQD over FQC, and

hydration is favored over either intramolecular product. We

hypothesize that FQC and FQD interconvert through the

formation of a transient N2−C3 imine and that this imine

intermediate is the nascent product of Af12070 enzymatic

action (Figure S5B).

With this knowledge we propose that Af12070 is an FAD-

dependent amide oxidase (acting on R-[R′-]-CH-NH-CO-R″)

compared to the more prevalent amine oxidases (which act on

R-[R′-]-CH-NH-R″).¹³The nascent imine generated by

Af12070 undergoes efficient intramolecular capture by the

pendant alcohol in FQA to form the 7-member spirohemiami-

nal FQC as the kinetic product. Slow equilibration ultimately

yields the 8-membered aminal containing FQD as the thermo-

dynamic adduct (Figure 8). It is interesting to speculate that

the intramolecular attack of the pendant −OH to form the

spirohemiaminal linkage in FQC may be favored within the

active site of Af12070 by the orientation of the −OH

nucleophile and the nascent imine.

Af12070 Activity toward Analogues of FQA Reveals

That Subtle Structural Differences in Substrate Can

Have a Dramatic Impact on End Products. From our

previously reported characterization of the enzymatic elabo-

ration of FQF to FQA by Af12060 and 12050,⁷we had noted

that the homologous enzyme pair TqaH and TqaB from

P. aethiopicum produced 2′-epi-FQA from FQF and ^L-Ala and

11′-dimethyl-2′-epi-FQA when 2′-aminoisobutyric acid (AIB) is

incorporated in place of ^L-Ala.⁸An HPLC-based assay was

performed to test the ability of Af12070 to utilize these two

FQA analogues as substrates and to investigate the effects of

structural changes on product outcome. 2′-epi-FQA was

processed smoothly by Af12070 Δ1−24 to give a single new

peak (Figure 9A). This compound was isolated and

characterized by NMR spectroscopy to be the amine-linked

2′-epi-FQD ([M + H]⁺observed 444.1662, expected 444.1666)

(Table S2), a result that is in contrast to the preference for

intramolecular capture by the pendant 3′-OH of FQA to form

FQC.

8763

dx.doi.org/10.1021/bi201302w|Biochemistry 2011, 50, 8756−8769

Biochemistry

Article

Figure 8. Proposed mechanism for Af12070.

When 11′-dimethyl-2′-epi-FQA (m/z = 460) was evaluated

with Af12070 Δ1−24, a series of five products were detected

(Figure 9B and Figures S6 and S7). Two peaks (3 and 4)

overlapped during HPLC analysis but separated when analyzed

by LC-MS (presumably due to differences in instrument

configuration). Peak 1 ([M + H]⁺observed 476.1929, expected

476.1928) and peak 2 ([M + H]⁺observed 476.1930, expected

476.1928) had a mass consistent with imine hydration and are

thought likely to be 3-OH diastereomers of the corresponding

dimethyl analogues of the previously characterized FQB (3-

OH-FQA).⁴Peak 3 ([M + H]⁺observed 458.1826, expected

458.1823) and peak 5 ([M + H]⁺observed 458.1828, expected

458.1823) appear to be the 11′-dimethyl-2′-epi analogues of

FQD and FQC, respectively (peak assignments are based on

relative retention times compared to the results in Figure 7A).

The mass data obtained for peak 4 indicates the presence of

three different compounds with m/z of 458, 459, and 477

(Figure S8). Interestingly, the retention time of peak 4 matches

that of the major peak present in the extract of P. aethiopicum

ΔtqaI, which has been postulated to be a keto acid resulting

from hydrolysis of the N2−C3 imine intermediate (TqaI

encodes a putative trypsin-like serine protease and is proposed

to act immediately downstream of the Af12070 homologue

TqaG) (Figure S8).⁸The previous characterization of the keto

acid (m/z = 477) by Gao et al. was accompanied by

spontaneous lactonization to yield deoxytryptoquialanone

(m/z = 459).⁸Comparison of our in vitro reaction with an

authentic standard of deoxytryptoquialanone reveals a small

peak in the standard that has a mass corresponding to the keto

acid and the same retention time as peak 4 (Figure S8). This

small peak presumably results from a degree of deoxytrypto-

quialanone lactone hydrolysis to form the keto acid during LC-

MS analysis. Importantly, the mass spectrum of this standard

does not contain a molecular ion with m/z = 458. It is also

important to note that the retention time of (unhydrolyzed)

deoxytryptoquialanone does not match that of peak 4 but that

we do observe a degree of MS-mediated dehydration of the

keto acid to give a species with the same mass and formula as

deoxytryptoquialanone; i.e., the presence of the keto acid can

explain the species with m/z of both 459 and 477 in peak 4, but

not that of 458. This indicates that the compound from our in

vitro reaction with m/z = 458 is structurally distinct from both

the keto acid and deoxytryptoquialanone. The identification of

a peak with m/z = 458, which appears intractable to isolation

and characterization, is consistent with the presence of a

nascent imine intermediate that hydrolyzes to the observed

keto acid under LC-MS analysis and is further masked by a

subsequent lactonization.

A third substrate analogue, 11′-dimethyl-FQA, was used to

further our analysis of the effects of molecular structure on the

fate of the putative enzymatic product of Af12070 (the N2−C3

imine). The substrate 11′-dimethyl-FQA (which is not a known

natural product) was obtained using a chimeric monomodular

NRPS which combines the A- and T-domains of TqaB (for

aminoisobutyric acid selection and incorporation) and the C-

domain of Af12050 (to retain the “natural” stereochemistry at

C2′).¹⁹When 11′-dimethyl-FQA was incubated with Af12070

Δ1−24, we observed the complete consumption of substrate

and the formation of a single new peak (Figure 9C);

characterization by MS ([M + H]⁺observed 458.1822,

expected 458.1823) and NMR spectroscopy revealed that this

peak is the hydroxyl-linked 11′-dimethyl-FQC (Table S3).

When compared to the results obtained using FQA as substrate

(Figure 7A), the absence of other product peaks appearing over

time indicates that the additional methyl substituent at C11′

discourages the spontaneous interconversion of 11′-dimethyl-

FQC to an amine-linked FQD analogue.

Genetic Knock-in Studies for the in Situ Production of

FQA and Dimethyl-FQA in P. aethiopicum. To provide

insight into the substrate specificity of the Af12070 homologue

TqaG, and because we could not produce it heterologously in a

soluble, active form in vitro, we utilized an in vivo platform

previously reported for our characterization of the tryptoquia-

lanine pathway.⁸We developed a P. aethiopicum strain that can

synthesize FQA derivatives that have the same trans stereo-

chemistry across C2′-C3′ (C2′-S) as in the A. fumigatus

fumiquinazoline pathway. To do so, we first deleted the tqaB

gene in the wild-type P. aethiopicum using bar as the selection

marker. We then knocked-in the functionally homologous

af12050 using the recombination plasmid pZeoGPE1, which

8764

dx.doi.org/10.1021/bi201302w|Biochemistry 2011, 50, 8756−8769

Biochemistry

Article

Figure 9. HPLC time course data (detection at 254 nm) for 2 μM Af12070 with 200 μM 2′-epi-FQA (A), 11′-dimethyl-2′-epi-FQA (B), or 11′-

dimethyl-FQA (C) as substrates. The structures of the products in panels A and C were determined by NMR spectroscopy, while the structures of

products 1−5 in panel B are proposed based on high-resolution MS, UV−vis, and published precedent.

includes the Zeocin resistance gene controlled by the trpC

promoter and the af12050 gene controlled by the gpdA

promoter.²⁰The replacement of TqaB with Af12050 is

expected to yield the intermediate FQA in vivo and thus

expose FQA to potential oxidative catalysis by TqaG. However,

experiments utilizing the ΔtqaB::af12050 mutant show efficient

in vivo production of FQA, but there were no compounds

detected corresponding to TqaG-mediated oxidation of FQA

(Figure 10A). We also used the P. aethiopicum ΔtqaB strain to

knock-in a hybrid monomodular NRPS (Chimera2¹⁹) for the in

vivo production of 11′-dimethyl-FQA. Similar to the results

obtained with the ΔtqaB::af12050 mutant, 11′-dimethyl-FQA is

produced by the ΔtqaB::chimera2 mutant, but oxidation of this

compound by TqaG is not observed (Figure 10B). For both

knock-in mutants, PCR verification was performed to ensure

that the chromosomal copy of tqaG gene was not interrupted

8765

dx.doi.org/10.1021/bi201302w|Biochemistry 2011, 50, 8756−8769

Biochemistry

Article

Figure 10. HPLC and LC-MS traces of metabolite extracts obtained from gene deletion/knock-in strains of P. aethiopicum used for the in situ

generation of FQA or 11′-dimethyl-FQA. (A) HPLC data (254 nm detection) of (i) extract from ΔtqaB, (ii) extract from ΔtqaB::af12050, and (iii)

an FQA standard. Lowercase letters labeling peaks: a, dechlorogriseofulvin; b, griseofulvin; c, oxy-FQF-dimer; d, viridicatumtoxin; e, FQA. (B) LC-

MS data mass filtered for m/z: 460 (equal to the [M + H]⁺species for dimethyl-FQA): trace i, extract from ΔtqaB; trace ii, extract from

ΔtqaB::chimera2; and trace iii, an 11′-dimethyl-FQA standard.

by plasmid insertion through random integration. These results

may indicate that Af12070 and TqaG have differing substrate

specificities, namely, that TqaG has a strict requirement for a

substrate with (R)-configuration at C2′ (e.g., the naturally

occurring 11′-dimethyl-2′-epi-FQA or 2′-epi-FQA). Alterna-

tively, these negative results could arise from differential

subcellular localization of substrate (FQA/11′-dimethyl-FQA)

and enzyme (TqaG) and simply reflect the inability of the

TqaG to act on a substrate that it does not come into contact

with. Obtaining and testing purified TqaG in vitro is one way to

distinguish between these possibilities, though our best efforts

to overproduce TqaG have been unsuccessful.

line.⁶By those efforts we identified an FQ biosynthetic gene

cluster in the A. fumigatus Af293 genome which utilizes

Af12080 as the synthase for the tripeptidyl-S-enzyme

intermediate that cyclizes to give FQF. A dedicated anthranilate

synthetase (Af12110) is observed four ORFs downstream of

the trimodular NRPS and presumably plays a role in generating

enough of this planar aromatic β-amino acid to get the pathway

started. Of the remaining predicted three enzymes in the

cluster, Af12050, Af12060, and Af12070, we have already

shown that the FAD-enzyme Af12060 and Ala-activating

monomodular NRPS Af12050 act in tandem to achieve the

oxidative annulation involved in the formation of the

imidazoindolone tricycle as FQF is processed to FQA.⁷

The remaining ORF Af12070, a predicted FAD-enzyme, was

then the likely candidate for introduction of the third stage of

scaffold complexity as FQA is converted into the heptacyclic

frameworks of FQC and FQD. After trimming off an N-

terminal region of the protein, likely to be involved in

localization/export in A. fumigatus, a soluble, active form of

Af12070 was obtained from E. coli expression. The enzyme

contains FAD covalently attached to His105 and functions as

an FQA oxidase. Kinetic analysis and product structure

determination reveal that FQA is converted to FQC and that

subsequent accumulation of FQD occurs via a nonenzymatic

equilibration.

Inspection of the FQC and FQD structures shows that in

FQC a new spirohemiaminal linkage results in the formation of

a seven-membered ring between the 3′-OH in FQA and the

carbon that was originally Cα of the ^L-Ala moiety incorporated

into the FQF quinazolinone scaffold (C3 in FQA). In

comparison, the aminal linkage observed in FQD forms an

eight-membered ring between the nitrogen of the second ^L-Ala

(incorporated by Af12050 in the observed indole annulation

reaction) and the same C3 carbon on the quinazolinone

scaffold as seen in FQC. Both the hemiaminal and aminal

linkages are formed by analogous addition reactions: the first

between an alcohol and an imine and the second between an

amine and the same imine. It is therefore not surprising that the

formation of the hemiaminal linkage in FQC could be

reversible, and the resultant imine would be available for

capture by the competing intramolecular amine to yield FQD.

DISCUSSION

■

The fumiquinazolines, as the prefix “fumi” implies, are a

signature class of tri- to tetrapeptidyl alkaloid natural products

from the human pathogen Aspergillus fumigatus. Three levels of

increasing complexity of the morphed peptide scaffold can be

discerned (Figure 1). The simplest is FQF (and the epimeric

FQG), where the tricyclic quinazoline scaffold is constructed

from three amino acid building blocks: the proteinogenic ^L-Trp

and ^L-Ala and the non-proteinogenic anthranilate (formally a

β-amino acid). The next level of complexity is exemplified by

conversion of FQF to FQA (and the congener FQI isolated

from Acremonium sp.¹¹) where a fourth amino acid, either L-Ala

(FQA) or ^L-Leu (FQI), is added in a net annulation to the

indole side chain of FQF. The third level of architectural

morphing, the focus of this study, is the conversion of FQA to

FQC (the corollary process in Acremonium is the conversion of

FQI to FQH¹¹) (Figure 2). The architectural complexity

increases dramatically in these three stages of maturation;

together they morph a tetrapeptide into a highly constrained

small molecule scaffold.

Our entry into this ubiquitous class of A. fumigatus

metabolites began with a search for the fungal enzymatic

machinery that could draw off a portion of the anthranilate pool

from tryptophan biosynthesis (and other primary metabolic

functions) for use in the production of natural products. We

anticipated an NRPS module with an adenylation domain that

could recognize and activate anthranilate for activation and

incorporation into a trimodular NRPS (Af12080) assembly

8766

dx.doi.org/10.1021/bi201302w|Biochemistry 2011, 50, 8756−8769

Biochemistry

Article

The time course of product appearance clearly indicates that

the initial accumulation of FQC is kinetically favored while

FQD is the thermodynamically favored product (Figure 7).

The formation of FQC and FQD, their spontaneous

interconversion, and the identification of a hydroxylated FQA

analogue (3-OH-FQA) are all consistent with the nascent

product of Af12070 catalysis being a N2−C3 imine species.

This, together with the fact that several FAD-containing

enzymes are known to oxidize carbon−nitrogen bonds,²⁹

suggests that it is highly likely that Af12070 is a regiospecific

amide oxidase that acts to dehydrogenate across the N2−C3

bond of the quinazolinone moiety of FQA. Despite the relative

abundance of FAD-dependent amine oxidases in biological

systems,¹³to the best of our knowledge this is the first

characterized example of an amide oxidase. It seems reasonable

to envision that the chemistry of hydride abstraction from a

carbon alpha to an amide, and thus oxidation, is significantly

hindered by the absence of lone-pair stabilization of the

resulting cation (as would be seen with an amine). It is

therefore intriguing to speculate that in the case of FQA

oxidation the extended delocalized system observed in the FQ

scaffold (i.e., the quinazolinone which neighbors the oxidized

pyrazinone) acts to stabilize the cation resulting from hydride

abstraction, such that the FQ system is unusually amenable to

the chemistry required for amide oxidation. Recent work

describing evidence in favor of amine oxidases mediating

transformation via single electron transfer (rather than direct

hydride transfer) would also be consistent with the FQ scaffold

being more susceptible to oxidation than a typical amide.³⁰The

presumed imine product resulting from amide oxidation does

not accumulate (when FQA is the substrate), and instead the

spirohemiaminal FQC is formed due to the intramolecular

capture of the imine by the pendant −OH at C3′ of the

imidiazoindolone.

The presumption is that Af12070 acts similar to a canonical

FAD-dependent amine oxidase to generate an imine product

and reduced flavin (via hydride transfer from substrate to N5 of

FAD) and bound imine product in the active site (Figure 8).¹³

However, an important distinction in the case of Af12070 is

that the −NH group that is enzymatically oxizidized is adjacent

to a carbonyl within the pyrazinone ring of FQA and thus

represents an unusual oxidation of an amide −NH. We have

shown that the resulting dihydroflavin is then reoxidized by

molecular oxygen in the second (oxidative) half reaction, thus

regenerating active enzyme without the requirement for further

cosubstrates/enzymes. The regeneration of FAD by O₂is

consistent with an oxidase mechanism yielding H₂O₂as a

coproduct (as opposed to an N-hydroxylation/dehydration

route to the imine which would yield two molecules of water

and also would require an external reductant such as NADH).

The biological role and the potential relevance of the

interconversion of the kinetically favored FQC and the

thermodynamically favored FQD within A. fumigatus is yet to

be determined. However, we speculate that the vast

architectural changes observed in going from FQA to either

FQC or FQD would have a significant impact on the

interaction of these molecules within biological systems. The

retention of Af12070 in the FQ biosynthetic gene cluster and

biosynthetic energy exerted in the observed regioselective

amide oxidation chemistry suggests that these complex

molecules are more than simply architecturally intriguing

natural products and must in fact have a tangible value to the

producing fungi within the complex ecological environment of

nature.

Four analogous substrates were tested with Af12070 in order

to assess substrate tolerance and gain insight into the effects of

molecular structure on product formation. Three naturally

occurring substrates were used: FQA (the biological substrate

for Af12070) and two compounds identified during genetic

characterization of the tryptoquialanine pathway from P.

aethiopicum, 11′-dimethyl-2′-epi-FQA and 2′-epi-FQA. In

addition, a fourth “unnatural” substrate, 11′-dimethyl-FQA,

was generated using a genetically engineered monomodular

NRPS. Taken together, these four substrates probe the effect of

C2′ stereochemistry and/or the effect of methyl vs dimethyl

substitution at C11′. These experiments enable us to investigate

how sites remote to the proposed site of Af12070 enzymatic

action influence molecular events, subsequent to catalysis, that

dictate the kinetics and thermodynamics of the apparently

spontaneous cyclizations that mediate the end-product out-

come.

In contrast to FQA, when the 2′-(R)-epi-FQA molecule is

evaluated as an Af12070 substrate, it is the 2′-epi-FQD scaffold

that forms with none of the 2′-epi-FQC hemiaminal product

being detected. Thus, the stereochemistry of the annulated

indole alters the kinetics of intramolecular capture of the

nascent imine product such that, in the 2′-epi-scaffold, 2′-epi-

FQD is both kinetically and thermodynamically favored. The

activity of TqaG toward 2′-epi-FQA has been tested in vivo by

deleting the genes in P. aethiopicum which encode TqaM or

TqaL,⁸the enzymes responsible for AIB biosynthesis. With AIB

synthesis shut down, the monomodular NRPS TqaB selects

and couples ^L-Ala to the FQF framework to give 2′-epi-FQA.^8,19

Culture extracts prepared from P. aethiopicum ΔtqaM or ΔtqaL

show that 2′-epi-FQA is a substrate for TqaG and that 2′-epi-

FQD is detected along with nortryptoquialanine, the des-

methyl analogue of the final natural product (Figure S10).⁸

While the appearance of 2′-epi-FQD as a shunt metabolite

clearly illustrates that the kinetics of the processing downstream

of TqaG (to give tryptoquialanine) is significantly perturbed by

the lack of the 11′-dimethyl group, it also provides clear

evidence that Af12070 and TqaG are functionally equivalent

with 2′-epi-FQA as substrate.

When 11′-dimethyl-FQA was incubated in vitro with

Af12070, only 11′-dimethyl-FQC is observed in the reaction

mixture even after extended periods of time. Therefore, it

appears that the presence of an 11′-dimethyl substitution on

FQA produces the opposite result to that observed with 2′-epi-

FQA. In this case 11′-dimethyl-FQC is both the kinetic and the

thermodynamic adducts resulting from intramolecular capture

of the nascent imine.

11′-Dimethyl-2′-epi-FQA is the proposed natural substrate for

TqaG, with its production in nature being more abundant than

the analogue 2′-epi-FQA due to the strong preference for

activation of AIB over ^L-Ala by TqaB in WT P. aethiopicum

(Figure S1). When 11′-dimethyl-2′-epi-FQA is incubated with

Af12070 in vitro, both the hydroxyl- and amine-linked products

are detected (11′-dimethyl-2′-epi-FQC and 11′-dimethyl-2′-epi-

FQD), as is the hydrate (consistent with the competitive

capture of the nascent imine product by intermolecular

addition of water) (Figures S6 and S7). Intriguingly, we can

also detect products consistent with the presence of the

proposed imine intermediate and that of the keto acid which

would be predicted to result from the hydrolysis of this

imine (Figures S6, S8, and S9). Overall, we have observed that

8767

dx.doi.org/10.1021/bi201302w|Biochemistry 2011, 50, 8756−8769

Biochemistry

Article

Funding

2′-epi-(R) stereochemistry appears to favor FQD formation and

that 11′-dimethyl substitution appears to favor FQC formation.

It is therefore intriguing to speculate that it is the combination

of these competing effects that prolongs the lifetime of the

nascent imine product of Af12070 (and TqaG) such that it can

be observed when 11′-dimethyl-2′-epi-FQA is used as substrate

for Af12070 catalysis. The hydrolysis of the imine to yield the

keto acid could be direct (i.e., the stepwise hydrolysis of the

carbonyl−nitrogen bond of the cyclic imine followed by

hydrolysis of the resulting acyclic imine⁸) or indirect following

the intramolecular capture of the imine as a hemiaminal/aminal

intermediate (Figure S9). If hydrolysis follows intramolecular

capture, ring strain could promote spontaneous hydrolysis of

the pyrazinone amide, a step which would be accelerated by the

trypsin-like serine protease TqaI in nature. In either case, the

resultant keto acid could undergo lactonization to give

deoxytryptoquialanone en route to the formation of trypto-

quialanine in P. aethiopicium (Figure S9). The mechanism of

this hydrolytic removal of the nitrogen from the tricyclic

quinazoline ring is not yet understood. However, comparative

analysis of the products resulting from Af12070 and TqaG

catalysis clearly suggest that molecular architecture is a critical

determinant in the reactivity of the nascent imine and

consequently fumiquinazoline vs tryptoquialanine pathway

differentiation (see Supporting Information Discussion for

details).

The studies described here on the initial characterization of

Af12070 raise the question of what are the biological functions

of each/any of the fumiquinazolines that make them

constitutive metabolites for all the A. fumigatus strains so far

examined. The complete biochemical characterization of the

fumiquinazoline biosynthetic enzymes now potentially opens

the door to genetic knockout experiments in A. fumigatus in

order to search for a phenotypic response to FQF, FQA, or

FQC/FQD production (such as reduced virulence or altered

patterns of development and/or sporulation) by knocking out

af12080, af12060 or af12050, and af12070, respectively.

From the point of view of biosynthesis of highly morphed

and constrained peptide-based architectural scaffolds, the

fumiquinazoline biosynthetic pathway comprises an efficient

and remarkably impressive set of catalytic machinery. In a short

metabolic pathway, with only four enzymes, two NRPS to

incorporate four amino acid building blocks and two FAD

enzymes to carry out key oxidations, a remarkable pair of

heptacyclic frameworks with five nitrogen atoms are con-

structed.

This work is supported in part by National Institutes of Health

Grant GM20011 (to C.T.W.), F32GM090475 (to B.D.A.),

1R01GM092217 (to Y.T.), and R01GM067725 (to N.L.K).

ACKNOWLEDGMENTS

■

We thank Steven Malcolmson for helpful discussions and

Thomas Gerken for providing A. fumigatus Af293 cDNA.

ABBREVIATIONS

■

2-AIB, 2-aminoisobutyric acid; AMP, adenosine-5′-mono-

phosphate; Ant, anthranilate (2-aminobenzoate); ATP, adeno-

sine-5′-triphospate; EDTA, ethylenediaminetetraacetic acid; epi,

epimer; FAD, flavin adenine dinucleotide; FMN, flavin

mononucleotide; FT, Fourier transform; FQ, fumiquinazoline;

HPLC, high-performance liquid chromatography; HRLC-MS,

high-resolution liquid chromatography−mass spectrometry;

IPTG, isopropyl-β-^D-galactopyranoside; LB, Luria−Bertani

medium; LIC, ligation-independent cloning; MeCN, acetoni-

trile; MWCO, molecular weight cutoff; NADH, nicotinamide

adenine dinucleotide; Ni-NTA, nickel nitrilotriacetic acid−

agarose; NMR, nuclear magnetic resonance; NRPS, non-

ribosomal peptide synthetase; PCR, polymerase chain reaction;

PDB ID, Protein Data Bank identifier; PTFE, polytetrafluoro-

ethylene; QTOF, quadrupole time-of-flight; SDS-PAGE,

sodium dodecyl sulfate−polyacrylamide gel electrophoresis;

TFA, trifluoroacetic acid; Tris, tris(hydroxymethyl)-

aminomethane.

ADDITIONAL NOTE

■

a

A discrepancy exists between the database entry for

AFUA_6g12070 and the coding sequence obtained from

cDNA cloning due to the predicted/annotated vs experimen-

tally determined boundaries for the first intron. Residue

numbering reflects the experimentally determined sequence

(see Experimental Procedures for details).

REFERENCES

■

(1) Frisvad, J. C., Rank, C., Nielsen, K. F., and Larsen, T. O. (2009)

Metabolomics of Aspergillus fumigatus. Med. Mycol. 47, 53−71.

(2) Rementeria, A., et al. (2005) Genes and molecules involved in

Aspergillus fumigatus virulence. Rev. Iberoam. Micol. 22, 1−23.

(3) Numata, A., et al. (1992) Fumiquinazolines, novel metabolites of

a fungus isolated from a saltfish. Tetrahedron Lett. 33, 1621−1624.

(4) Takahashi, C., et al. (1995) Fumiquinazolines A−G, novel

metabolites of a fungus separated from a Pseudolabrus marine fish.

J. Chem. Soc., Perkin Trans. 1, 2345−2353.

(5) Nierman, W.C., et al. (2005) Genomic sequence of the

pathogenic and allergenic filamentous fungus Aspergillus fumigatus.

Nature 438, 1151−1156.

(6) Ames, B. D., and Walsh, C. T. (2010) Anthranilate-activating

modules from fungal nonribosomal peptide assembly lines. Bio-

chemistry 49, 3351−3365.

(7) Ames, B. D., Liu, X., and Walsh, C. T. (2010) Enzymatic

processing of fumiquinazoline F: A tandem oxidative-acylation strategy

for the generation of multicyclic scaffolds in fungal indole alkaloid

biosynthesis. Biochemistry 49, 8564−8576.

(8) Gao, X., et al. (2011) Fungal indole alkaloid biosynthesis:

Genetic and biochemical investigation of the tryptoquialanine pathway

in Penicillium aethiopicum. J. Am. Chem. Soc. 133, 2729−2741.

(9) Fremlin, L. J., Piggott, A. M., Lacey, E., and Capon, R. J. (2009)

Cottoquinazoline A and cotteslosins A and B, metabolites from an

Australian marine-derived strain of Aspergillus versicolor. J. Nat. Prod.

72, 666−670.

ASSOCIATED CONTENT

■

S

* Supporting Information

Supplementary discussion of fumiquinazoline vs tryptoquiala-

nine pathway differentiation, figures, and tables of NMR data.

This material is available free of charge via the Internet at

http://pubs.acs.org.

AUTHOR INFORMATION

■

Corresponding Author

*E-mail: christopher_walsh@hms.harvard.edu. Phone: 617-

432-1715. Fax: 617-432-0483.

Author Contributions

^⊥Equally contributing authors.

8768

dx.doi.org/10.1021/bi201302w|Biochemistry 2011, 50, 8756−8769

Biochemistry

Article

(10) Zhuang, Y., et al. (2011) New quinazolinone alkaloids within

rare amino acid residue from coral-associated fungus, Aspergillus

versicolor LCJ-5−4. Org. Lett. 13, 1130−1133.

(11) Gilbert, N. B., Montserrat, A., Paul, R. J., William, F., and

Matthias, K. (2000) Oxepinamides A-C and fumiquinazolines H-I:

bioactive metabolites from a marine iIsolate of a fungus of the genus

Acremonium. Chem.Eur. J. 6, 1355−1360.

(12) Twumasi-Boateng, K., et al. (2009) Transcriptional profiling

identifies a role for BrlA in the response to nitrogen depletion and for

StuA in the regulation of secondary metabolite clusters in Aspergillus

fumigatus. Eukaryot. Cell 8, 104−115.

(13) Fitzpatrick, P. F. (2009) Oxidation of amines by flavoproteins.

Arch. Biochem. Biophys. 493, 13−25.

(14) Wilkins, M. R., et al. (1999) Protein identification and analysis

tools in the ExPASy server. Methods Mol. Biol. 112, 531−552.

(15) Larkin, M. A.et al. (2007) Clustal W and Clustal X version 2.0.

Bioinformatics 23, 2947−2948.

(16) Nicholas, K. B., Nicholas, H. B., and Deerfield, D. W. (1997)

GeneDoc: analysis and visualization of genetic variation. EMBNEW

News 4, 14.

(17) Soding, J., Biegert, A., and Lupas, A. N. (2005) The HHpred

interactive server for protein homology detection and structure

prediction. Nucleic Acids Res. 33, W244−248.

(18) Deshpande, N., Wilkins, M. R., Packer, N., and Nevalainen, H.

(2008) Protein glycosylation pathways in filamentous fungi.

Glycobiology 18, 626−637.

(19) Haynes, S. W., Ames, B. D., Gao, X., Tang, Y., and Walsh, C. T.

(2011) Unraveling terminal C-domain-mediated condensation in

fungal biosynthesis of imidazoindolone metabolites. Biochemistry 50,

5668−5679.

(20) Pall, M., and Brunelli, J. (1993) A series of six compact fungal

transformation vectors containing polylinkers with multiple unique

restriction sites. Fungal Genet. Newsl. 40, 59.

(21) Szewczyk, E., et al. (2007) Fusion PCR and gene targeting in

Aspergillus nidulans. Nature Protoc. 1, 3111−3120.

(22) Leferink, N. G. H., Heuts, D. P. H. M., Fraaije, M. W., and van

Berkel, W. J. H. (2008) The growing VAO flavoprotein family. Arch.

Biochem. Biophys. 474, 292−301.

(23) Heuts, D. P.H. M., Scrutton, N. S., McIntire, W. S., and Fraaije,

M. W. (2009) What’s in a covalent bond? On the role and formation

of covalently bound flavin cofactors. FEBS J. 276, 3405−3427.

(24) Sirikantaramas, S., et al. (2005) Tetrahydrocannabinolic acid

synthase, the enzyme controlling marijuana psychoactivity, is secreted

into the storage cavity of the glandular trichomes. Plant Cell Physiol.

46, 1578−1582.

(25) Bird, D., and Facchini, P. (2001) Berberine bridge enzyme, a

key branch-point enzyme in benzylisoquinoline alkaloid biosynthesis,

contains a vacuolar sorting determinant. Planta 213, 888−897.

̌

́

(26) Smehilova, M., et al. (2009) Subcellular localization and

biochemical comparison of cytosolic and secreted cytokinin

dehydrogenase enzymes from maize. J. Exp. Bot. 60, 2701−2712.

́

(27) Varela, E., Jesus Martinez, M., and Martinez, A. T. (2000) Aryl-

́

alcohol oxidase protein sequence: a comparison with glucose oxidase

and other FAD oxidoreductases. Biochim. Biophys. Acta 1481, 202−

208.

(28) Mattevi, A. (2006) To be or not to be an oxidase: challenging

the oxygen reactivity of flavoenzymes. Trends Biochem. Sci. 31, 276−

283.

(29) Fagan, R. L., Palfey, B. A., Lew, M., and Hung-Wen, L. Flavin-

Dependent Enzymes, in Comprehensive Natural Products II, pp 37−

113, Elsevier, Oxford.

(30) MacMillar, S., Edmondson, D. E., Matsson, O. Nitrogen kinetic

isotope effects for the monoamine oxidase B-catalyzed oxidation of

benzylamine and (1,1-²H₂)benzylamine: nitrogen rehybridization and

CH bond cleavage are not concerted, J. Am. Chem. Soc. 133, 12319−

12321.

8769

dx.doi.org/10.1021/bi201302w|Biochemistry 2011, 50, 8756−8769

Article Doi

Complexity generation in fungal peptidyl alkaloid biosynthesis: Oxidation of fumiquinazoline A to the heptacyclic hemiaminal fumiquinazoline C by the flavoenzyme Af12070 from aspergillus fumigatus

DOI: 10.1021/bi201302w

Source and publish data:

Authors:

Article abstract of DOI:10.1021/bi201302w

Full text of DOI:10.1021/bi201302w

Products guided by the article

R&D Labs maybe for 140852-71-7

Relevant to this article

Hot Product