10.1002/anie.202005770
Angewandte Chemie International Edition
RESEARCH ARTICLE
ΔqulA or ΔqulB mutant, and there might be an endogenous
amidase outside of the qul gene cluster that catalyzes the amide
hydrolysis and accelerates the hydrolysis process.
Next, we sought to find the actual precursor substrates for both
QulA and QulB. Compound 7 was accumulated in ΔqulA and
ΔqulB knockout strain metabolites, and also observed in the QulF
and amidase in vitro assays; however, 7 does not possess a
carboxylic group to be utilized by the NRPS A domain. Thus, we
proposed that the unstable compound 9 might be the actual
precursor substrate of QulA or QulB for the biosynthesis of 1.
Although multiple strategies were employed to synthesize 9, the
methylation step of the highly reactive primary amine group was
Amide hydrolysis of 8 by amidase. To verify our hypothesis, we
then aimed to identify the amidase that might locate outside of the
qul biosynthetic gene cluster and catalyze the amide hydrolysis of
8 to 9. BLAST analysis against the P. citrinum genome using an
acetamidase from Aspergillus oryzae[13] showed eleven hits
containing amidase signature sequence (Table S8, Figure
S15).[14] Next, we cloned and overexpressed six amidases (#2, #4,
#7, #9, #10, and #11) from E. coli for further characterization
(Figure S16A). Five of these amidases were successfully
expressed, except for amidase #9 (Figure S16B). Interestingly,
incubation of QulF and compound 4 with amidase #7 (9 hours) led
to complete conversion of 8 to 7, while the other amidases failed
to further facilitate the reaction (Figure 4A). In addition, a new
peak with m/z=194.1 [M+H]+ was also detected in the reaction
with amidase #7 (Figure 4A, trace iii). According to the UV and
mass data (Figure S6.2E), we proposed this new intermediate
could be the unstable precursor 9, which might be spontaneously
cyclized to 7 (Figure 3C). When a set of time-course assays using
amidase #7 were performed, the complete conversion of 8 to 7
was observed within 20 min (Figure 4B). These results indicate
that the intracellular amidase (amidase #7) could hydrolyze 8 to
9, followed by spontaneous cyclization to afford 7.
not successful. Considering that ΔqulM produced
3
(desmethylated 1, Figure 3A, trace iv), we speculated that QulA
and QulB could also react with 2-aminobenzoylacetate (10, the
desmethylated 9) and L-Ile to produce 3. To verify our hypothesis,
we chemically synthesized compound 10 (Figure S6.2D) and
verified its structure by NMR (Figure S10.12). We also tried to
methylate 10 to produce 9 through chemical synthesis, but doing
so proved challenging because 10 was also unstable in the
reaction solution and quickly converted to compound 11.[17] We
purified and characterized the structure of 11 via 1D and 2D NMR
1
(Figure S6.4A, Table S5, and Figure S10.13-S10.15). H NMR
and 13C NMR spectra matched the reported data of 4-hydroxy-2-
quinolone,[18] which is a desmethylated version of 7. The observed
spontaneous dehydration of 10 to 11 supported our hypothesis
that compound 7 should be the spontaneously dehydrated
product of compound 9. Notably, when 10 was incubated with
QulM and SAM, 9 was not produced, indicating that 10 was not a
substrate of QulM (Figure S8C) and thus could not be synthesized
enzymatically.
We then heterologously expressed and purified QulA and QulB
(Figure S7) from E. coli BAP1.[19] To measure the adenylation
activity of the A domains, the colorimetric detection of free
pyrophosphate assay was performed.[20] The high level of
pyrophosphate production was only observed when L-Ile and 10
were used as substrates for QulA and QulB, respectively (Figure
5B). These results suggested that L-Ile is the preferred substrate
for QulA, and that β-keto acid 10 is the preferred substrate for
QulB. It is worth noting that, although numerous nonproteinogenic
amino acids participated in NRPS biosynthesis, an NRPS A
domain recognizing a β-keto acid has been rarely reported. In
earlier reports, ten amino acid residues (A1-A10) have been
proposed to be involved in the substrate recognition, based on
information from the adenylation domain PheA structure in
GrsA.[21] Alignment of selective NRPS A domains showed the
highly conserved acidic amino acid Asp (A1-position, interacts
with the α-amino group of the substrate amino acids) in QulB is
changed into the basic amino acid His423 (Figure S18A). Docking
analysis of the QulB-A domain with 9 showed a favorable
interaction between the His423 (A1-position) and the β-ketone
group of 9 (Figure S18B). We thus proposed the His423 in A1-
position is crucial to accommodate the β-keto group, which
enables the NRPS QulB to recognize and incorporate β-keto acid.
Figure 4. Amide hydrolysis of 8 by amidase #7. (A) Amidase assays for 8
hydrolysis to 9. (i)-(vi) enzymatic assays of QulF, 4, and amidase #2, #4,
#7, #9, #10, and #11, respectively. (vii) standards of 4 and 7. (B) Time-
course assays of amidase #7. (i)-(iii) amidase #7 reaction for 0 min, 5 min,
and 20 min, respectively. (iv) control, incubation of 4 and amidase #7 for 20
min. (v) standards of 4 and 7.
QulA and QulB are involved in the formation of quinolone-γ-
lactam hybrid. NRPS module analysis shows that QulB consists
of condensation (C), adenylation (A), and thiolation (T) domains
(Figure 5A). A close examination showed that the conserved
HHxxxDG motif is completely missing in the C domain of QulB.[15]
This type of inactivated C domain was named the C* domain and
proposed to play a structural role for protein stability.[15] QulA
consists of C-, A-, T-, and reductase (R) domains (Figure 5A). The
phylogenetic analysis showed that the C-terminal domain of QulA
falls into the same group as the previously reported NAD(P)H-
independent R* domains found in the fungal PKS-NRPS hybrids,
e.g., CpaA, ApdA, and TenS (Figure S17), which usually catalyze
the Dieckmann condensation between a β-keto intermediate and
an amino acid.[16] Furthermore, like other R* domains, the C-
terminal domain of QulA lacks the critical Ser-Tyr-Lys catalytic
triad and the conserved NAD(P)H binding sites in typical
reductase domain protein (Figure S17B).
When QulA and QulB were incubated with 10 and L-Ile in the
presence of ATP, the formation of 3 was observed (Figure 5C,
trace ii, time-course analysis shown in Figure S19A), which
confirmed that QulA and QulB could catalyze the formation of the
tricyclic quinolone-γ-lactam hybrid. In the absence of QulA and
QulB, compound 10 spontaneously converted to compound 11
(desmethylated version of 7) (Figure 5C, trace i). The in vitro
enzymatic synthesis of 3 in the absence of NAD(P)H strongly
supported the supposition that the C-terminal domain of QulA is
4
This article is protected by copyright. All rights reserved.