Identification of the grincamycin gene cluster unveils divergent roles for GcnQ in different hosts, tailoring the L-rhodinose moiety

Article abstract of DOI:10.1021/ol401253p

The gene cluster responsible for grincamycin (GCN, 1) biosynthesis in Streptomyces lusitanus SCSIO LR32 was identified; heterologous expression of the GCN cluster in S. coelicolor M512 yielded P-1894B (1b) as a predominant product. The ΔgcnQ mutant accumulates intermediate 1a and two shunt products 2a and 3a bearing l-rhodinose for l-cinerulose A substitutions. In vitro data demonstrated that GcnQ is capable of iteratively tailoring the two l-rhodinose moieties into l-aculose moieties, supporting divergent roles of GcnQ in different hosts.

Full text of DOI:10.1021/ol401253p

ORGANIC
LETTERS
XXXX
Vol. XX, No. XX
000–000
Identification of the Grincamycin Gene
Cluster Unveils Divergent Roles for
GcnQ in Different Hosts, Tailoring
the L‑Rhodinose Moiety
Yun Zhang,† Hongbo Huang,† Qi Chen, Minghe Luo, Aijun Sun, Yongxiang Song,
Junying Ma, and Jianhua Ju*
CAS Key Laboratory of Marine Bio-resources Sustainable Utilization, Guangdong Key
Laboratory of Marine Materia Medica, RNAM Center for Marine Microbiology, South
China Sea Institute of Oceanology, Chinese Academy of Sciences, 164 West Xingang
Road, Guangzhou 510301, China
jju@scsio.ac.cn
Received May 5, 2013
ABSTRACT
The gene cluster responsible for grincamycin (GCN, 1) biosynthesis in Streptomyces lusitanus SCSIO LR32 was identified; heterologous
expression of the GCN cluster in S. coelicolor M512 yielded P-1894B (1b) as a predominant product. The ΔgcnQ mutant accumulates intermediate
1a and two shunt products 2a and 3a bearing L-rhodinose for L-cinerulose A substitutions. In vitro data demonstrated that GcnQ is capable of
iteratively tailoring the two L-rhodinose moieties into L-aculose moieties, supporting divergent roles of GcnQ in different hosts.
The angucycline antibiotic grincamycin (GCN, 1) was
first discovered in 1987 from Streptomyces griseoincarnatus1
and has been recently rediscovered by our group along
with GCN B (2), GCN E (3), and GCNs C, D, and F as
minor metabolites from the deep sea derived Streptomyces
lusitanus SCSIO LR32.2 GCN has a tetrangomycin skeleton
in which the 3-O- and 9-positions are substituted by sacchar-
ides, containing R-L-rhodinose and R-L-cinerulose A for the
disaccharide and consisting of β-D-olivose, R-L-rhodinose, and
R-L-cinerulose A for the trisaccharide (Figure 1B). Closely
related to 1 is vineomycin A1 also known as P-1894B (1b),
isolated from Streptomyces matensis subsp. vineus and
S. albogriseolus subsp. No. 19843ꢀ5 featuring two L-aculose
moieties in place of the two L-cinerulose A moieties of 1
(Figure 1B). GCN was reported to inhibit the growth of P388
murine leukemia cells and displayed cytotoxicities toward a
panel of tumor cell lines,1,2 whereas P-1894B was reported
to be active against Sarcoma 180 solid tumors in mice.3
Although numerous gene clusters responsible for angucy-
cline antibiotics have been cloned,6 the biosynthetic gene
clusters for 1 and 1b have not been identified. Consequently,
the direct biosynthetic link between 1and 1b has been unclear.
Herein we report (i) the cloning, sequencing, and anno-
tation of the gene cluster governing GCN biosynthesis in
S. lusitanus SCSIO LR32; (ii) heterologous expression of
the entire GCN biosynthetic gene cluster in S. coelicolor
M512 yielding P-1894B (1b) as a predominant product;
(iii) inactivation of gcnQ leading to accumulation of
† These authors contributed equally.
(1) Hayakawa, Y.; Iwakiri, T.; Imamura, K.; Seto, H.; Otake, N. J.
Antibiot. 1987, 40, 1785–1787.
(2) Huang, H.; Yang, T.; Ren, X.; Liu, J.; Song, Y.; Sun, A.; Ma, J.;
Wang, B.; Zhang, Y.; Huang, C.; Zhang, C.; Ju, J. J. Nat. Prod. 2012, 75,
202–208.
(3) Omura, S.; Tanaka, H.; Oiwa, R.; Awaya, J.; Masuma, R.;
Tanaka, K. J. Antibiot. 1977, 30, 908–916.
(4) Ohta, K.; Mizuta, E.; Okazaki, H.; Kishi, T. Chem. Pharm. Bull.
(5) Imamura, N.; Kakinuma, K.; Ikekawa, N.; Tanaka, H.; Omura,
S. Chem. Pharm. Bull. 1981, 29, 1788–1790.
(6) Kharel, M. K.; Pahari, P.; Shepherd, M. D.; Tibrewal, N.; Nybo,
1984, 32, 4350–4359.
S. E.; Shaaban, K. A.; Rohr, J. Nat. Prod. Rep. 2012, 29, 264–325.
r
10.1021/ol401253p
XXXX American Chemical Society

Figure 1. (A) The color-coded biosynthetic gene cluster of GCN (1); overlapping cosmids are indicated by solid lines. (B) GcnQ in vitro
catalyzes transformation of 1a and 2a to 1b and 2b via intermediates 1c and 2c, respectively. In vitro, GcnQ converted 3a to 3c, failing to
produce anticipated final product 3b. GcnQ in vivo tailors L-rhodinose units to L-cinerulose A units in S. lusitanus SCSIO LR32 and
tailors L-rhodinose units to L-aculose units in S. coelicolor M512.
biosynthetic intermediate 1a and two shunt products 2a and
3a; and (iv) in vitro GcnQ-catalyzed transformations of
1aꢀ3a into P-1894B (1b), vineomycin B2 (2b), and 3c; 1b
and 2b are generated through the intermediacy of 1c and 2c,
respectively.
polyketide synthases and cyclases associated with angucy-
cline core construction (GcnHIJKL). Also found in the
cluster of 1 were (i) seven oxidoreductases for postmodi-
fications (GcnACFMQTU), (ii) a set of eight proteins for
deoxysugar biosynthesis (GcnS1-S8), (iii) three glycosyl-
transferases for sugar attachment (GcnG1-G3), (iv) two
regulatory proteins (GcnDR), (v) two proteins for transpor-
tation (GcnBN), (vi) a decarboxylase (GcnP), and (vii) two
proteins with unknown functions (GcnEO) (Table S1). The
nucleotide sequences have been deposited in the GenBank
with accession number KC962511. The genetic organization
of the biosynthetic gene cluster is shown in Figure 1A.
The sugars contained in GCN are all 2,3-deoxy sugars.
Accordingly, we set out to clone the gene cluster of 1 using
degenerate primers targeting the sugar 2,3-dehydratase.7,8
A clear PCR band was amplified using the genomic DNA
as a template, then ligated into pCR 2.1 vector, and
transformed into E. coli DH5R cells. Sequence analysis
of five clones revealed that the cloned fragments possessed
identical sequences, which, upon Blast analysis, showed
high homology to the 2, 3-dehydratase (UrdS) in the
urdamycin biosynthetic pathway.9 This specific 2,3-dehy-
dratase probe was utilized to screen ∼2300 clones of a
SuperCos1-based genomic library of S. lusitanus SCSIO
LR32. The target 2, 3-dehydratase allele was then inacti-
vated by replacement with an aac(3)IV/oriT cassette using
PCR-targeting mutagenesis methods.10 The resulting mu-
tant (ΔgcnS7) lost the ability to produce 1 but generated
three analogues (4ꢀ6) (Figure 2, trace II), suggesting
involvement of the cloned gene in GCN biosynthesis.
Subsequent restriction enzyme mapping and end-sequen-
cing of the 10 screened cosmids revealed that cosmid
179F may contain the whole biosynthetic pathway for 1.
Sequencing of cosmid 179F led to a contiguous DNA
sequence of 50 Kb, from which 30-open reading frames
spanning 37 Kb were identified probably accounting for
biosynthesis of 1. The gene cluster contains five type II
Figure 2. HPLC profiles of S. lusitanus wild-type (I), ΔgcnS7
(II), ΔgcnH (III), S. coelicolor/179F (IV), S. coelicolor M512
(V), and ΔgcnQ (VI). See Figure 1B for structures.
The 2,3-dehydratase mutant (ΔgcnS7) was fermented on
a large scale (8-L), enabling purification of analytically
1
pure 4ꢀ6. HRMS, H, and 13C NMR data analyses and
(7) Zhang, X.; Alemany, L. B.; Fiedler, H. P.; Goodfellow, M.; Parry,
R. J. Antimicrob. Agents Chemother. 2008, 52, 574–585.
(8) Sasaki, E.; Ogasawara, Y.; Liu, H.-w. J. Am. Chem. Soc. 2010,
132, 7405–7417.
(9) Hoffmeister, D.; Ichinose, K.; Domann, S.; Faust, B.; Trefzer, A.;
Drager, G.; Kirschning, A.; Fischer, C.; Kunzel, E.; Bearden, D. W.;
Rohr, J.; Bechthold, A. Chem. Biol. 2000, 7, 821–831.
(10) Gust, B.; Challis, G. L.; Fowler, K.; Kieser, T.; Chater, K. F.
Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 1541–1546.
comparisons with previously reported spectroscopic data
revealed the structures of 4ꢀ6 to be tetrangomycin (4),11
fridamycin (5),12 and fridamycin methyl ester (6).12
(11) Kaliappan, K. P.; Ravikumar, V. J. Org. Chem. 2007, 72, 6116–
6126.
B
Org. Lett., Vol. XX, No. XX, XXXX

To further confirm that the candidate gene cluster is, in
fact, responsiblefor biosynthesis of1, the minimal ketoacyl
synthase GcnH was inactivated. The resultant ΔgcnH
mutant proved incapable of 1 production as well as produc-
tion of its analogues (Figure 2, trace III), validating that the
cloned gene cluster is indeed responsible for 1 biosynthesis.
To confirm the sufficiency of the gene cluster for GCN
biosynthesis, heterologous expression of cosmid 179F,
likely harboring the entire GCN biosynthetic cluster, was
carried out. Cosmid 179F was generated by replacing the
kanamycin resistance gene (from the Supercos1 vector)
with a fragment excised from pSET152AB containing the
apramycin resistance gene and elements necessary for
conjugation and site specific recombination (oriT, inte-
grase gene and jC-31 site), using λ-RED-recombination
technology. The resulting cosmid, termed 179F-pSET152AB
(Supporting Information), was transferred into S. coelicolor
M512 by conjugation to generate the S. coelicolor/179F
strain. The engineered strain was fermented using the
same medium previously used for fermentation of wild-type
S. lusitanus SCSIO LR32. Fermentation broth was then
extracted with butanone, and the resulting metabolite con-
tents were analyzed by HPLC. Unexpectedly, we observed
generation of a predominant peak with a slightly longer
retention time than that for 1 and having UV absorption
signals characteristic of GCN at 425, 319, and 230 nm
(Figure 2, trace IV). Further HRMS analysis of this peak
revealed that it represents a compound having molecular
formula C49H58O18, four mass units smaller than that of 1,
and indicative of a structure known as P-1894B. Large-scale
fermentation (3-L) of the recombinant strain enabled isola-
tion of sufficient quantities of this material for NMR data
To explore the role of GcnQ in the GCN biosynthetic
pathway, we inactivated gcnQ. Fermentation of the resul-
tant ΔgcnQ mutant and HPLC metabolite analysis re-
vealed that inactivation of gcnQ abolished production of 1
and its corresponding analogues. Instead, this mutant
produced three major products 1a, 2a, and 3a (Figure 2,
trace VI). Large scale fermentation (8 L) of the ΔgcnQ
strain enabled isolation of 1aꢀ3a in quantities sufficient
for thorough structure elucidation. Molecular formulas of
C49H66O18 (1a), C49H66O18 (2a), and C49H64O17 (3a) were
determined on the basis of HRMS data. 1H and 13C NMR
spectroscopic data of 1a, 2a, and 3awere similar to those of
GCN (1), GCN B (2), and GCN E (3) (Tables S7ꢀS9),
respectively.2 However, the 13C NMR signals for the two
carbonyl carbons in the two L-cinerulose A units were
missing in each compound. Moreover, two additional
oxygen-bearing methine signals at about δC 67 ppm were
observed in all three compounds. Furthermore, the 13C
NMR resonances of C-3 in the L-cinerulose A units were
shifted upfield from δC ∼33 to δC ∼24 ppm. These
observations suggested that the two L-cinerulose A units
at the end of the sugar chains in 1, 2, and 3 were replaced by
two L-rhodinoses in 1a, 2a, and 3a, respectively. Detailed
analysis of their respective 2D (COSY, HMQC, HMBC,
and NOESY) NMR data confirmed the elucidated struc-
tures. We envision that 2a and 3a are shunt products
derived from 1a during fermentation. Hydrolysis of the
C-1/C-12b bond in 1a likely affords 2a whereas UV or
heat-mediated rearrangement of 1a may generate 3a.2,16 To
test this hypothesis, 1a was dissolved in 0.1% HOAcꢀH2O
and allowed to stand for 2 h; conversion to 2a was readily
apparent upon HPLC analysis and comparisons with a
standard sample (Figure S5). Conversely, when dissolved in
MeOH 1a was converted to 3a following UV irradiation for
12hatrt(FigureS6). Theisolationof1aꢀ3a from the ΔgcnQ
mutant implies that 1a is a direct precursor to 1 and that
GcnQ is responsible for tailoring of the two L-rhodinose units
of 1a into the respective L-cinerulose A units of 1. Impor-
tantly, we witnessed no signs of olefination of either ciner-
ulose A units with the wild-type producer S. lusitanus SCSIO
LR32 (Figure 1B). We also investigated the prospect that the
L-cineruloses of 1ꢀ3 might be generated by 2,3-reduction of
appropriate L-aculose-containing substrates. We, thus, per-
formed feeding experiments in which the ΔgcnH mutant
strain provided L-aculose substrate 1b for possible conversion
into 1. Under no circumstances could 1 be generated from 1b
thereby further supporting the divergent roles of GcnQ in
different hosts (Supporting Information, Figure S4).
1
acquisition. Both H and 13C NMR data confirmed the
identity of the heterologously expressed compound as
P-1894B (1b).4 That the GCN gene cluster when expressed
in S. lusitanus SCSIO LR32 (Figure 2, trace I) provides 1, yet
when expressed in S. coelicolor M512 affords 1b as the
predominant product, is truly remarkable.
Both 1 and 1b possess keto and R,β-unsaturated keto
sugars, respectively. How these highly unusual keto sugars
are formed and then sequentially attached to each mole-
cule are intriguing questions. Within the GCN cluster,
gcnQ encodes a 529 aa oxidoreductase showing homology
with AknOx (56% identity), TrdL (49% identity), and GilR
(38% identity) in the aclacinomycin,13 tirandamycin,14 and
gilvocarcin V15 biosynthetic pathways, respectively. All
these proteins are characterized by bicovalent attachment
of the FAD cofactor to conserved histidine and cysteine
residues (Supporting Information, Figure S2).
To evaluate the activity of GcnQ in vitro, the coding
gene was cloned into the NdeI and HindIII sites of the
pET28a(þ) vector and the resulting vector was trans-
formed into E. coli BL21(DE3). GcnQ was overexpressed
as an N-terminal His6-tagged soluble protein and purified
to homogeneity by Ni affinity chromatography. The
UVꢀvis spectrum of GcnQ exhibited two absorption
(12) Ueberbacher, B. J.; Osprian, I.; Mayer, S. F.; Faber, K. Eur. J.
Org. Chem. 2005, 7, 1266–1270.
€
€ €
(13) Alexeev, I.; Sultana, A.; Mantsala, P.; Niemi, J.; Schneider, G.
Proc. Natl. Acad. Sci. U.S.A. 2007, 104, 6170–6175.
(14) (a) Mo, X.; Wang, Z.; Wang, B.; Ma, J.; Huang, H.; Tian, X.;
Zhang, S.; Zhang, C.; Ju, J. Biochem. Biophy. Res. Commun. 2011, 406,
341–347. (b) Mo, X.; Huang, H.; Ma, J.; Wang, Z.; Wang, B.; Zhang, S.;
Zhang, C.; Ju, J. Org. Lett. 2011, 13, 2212–2215.
(15) Noinaj, N.; Bosserman, M. A.; Schickli, M. A.; Piszczek, G.;
Kharel, M. K.; Pahari, P.; Buchanan, S. K.; Rohr, J. J. Biol. Chem. 2011,
286, 23533–23543.
(16) Sezaki, M.; Kondo, S.; Maeda, K.; Umezawa, H.; Ohno, M.
Tetrahedron 1970, 26, 5171–5190.
(17) Imamura, N.; Kakinuma, K.; Ikekawa, N.; Tanaka, H.; Omura,
S. J. Antibiot. 1982, 35, 602–608.
Org. Lett., Vol. XX, No. XX, XXXX
C

bands at 350ꢀ375 and 450 nm, very similar to the bicovalent
flavinylated proteins BBE and TrdL.14b Heat denaturation
of GcnQ resulted in a colorless supernatant and yellow
precipitate; no FAD was detected in the supernatant by
UVꢀvis spectroscopy or HPLC analysis demonstrating
that FAD is, in fact, covalently attached to the protein.
Enzymatic reactions with GcnQ were conducted at 30 °C
in a volume of 50 μL consisting of 50 mM Tris-HCl (pH 7.5),
20 μM substrate, and 0.5 μM GcnQ. Reactions with boiled
GcnQ served as controls; all reactions were carried out in
parallel to ensure reaction consistency. Reactions were
terminated by 2-fold extraction with 100 μL EtOAc; EtOAc
layers were evaporated to dryness, the resulting contents
were then dissolved in 10 μL of MeOH, and the samples were
subjected to HPLC analyses. HPLC time-course analyses
revealed the GcnQ-catalyzed transformation of 1a into 1b
through intermediate 1c (Figure 3A). Similarly, 2a was
converted to 2b through the agency of 2c (Figure 3B), and
3a was converted to 3c (Figure 3C). Notably, GcnQ failed to
generate 3b from 3a; it appears the reaction sequence stalls
following generation of 3c (Figure 3C). Large scale (100 mL)
GcnQ-catalyzed reactions of 1a allowed the isolation of 1b
and 1c in quantities sufficient for structural elucidation.
Compounds 2b and 2c were obtained in similar yields from
large scale (100 mL) reaction of 2a with GcnQ; small scale
reaction (50 mL) of 3a afforded 3c. In all cases inter-
mediates and products were isolated in quantities enabling
rigorous structure elucidation. Significantly we found that,
in vitro, GcnQ failed to generate either set of products 1bꢀ3b
or 1cꢀ3c from the respective starting materials 1ꢀ3
(Figure 3DꢀF).
6.95 (dd, J = 9.8, 3.4 Hz), δC 144.1, δH 6.10 (d, J = 9.8 Hz),
δC 127.5 were noted. Moreover, a new 13C NMR signal at
δC 197.9 was apparent, indicating the presence of a (cis)-R,
β-unsaturated carboxy moiety in 1c. Consequently, it was
apparent that an L-rhodinose unit in 1b was replaced by an
L-aculose in 1c. Further HMBC correlations of 100000-H/C-40000
and 40000-H/C-100000 revealed this unanticipated L-aculose
moiety to be positioned at the end of the disaccharide chain
(Figure S12). Similarly, structures 2c and 3c were determined
by NMR data analysis.
The in vitro enzyme assays reveal a number of new and
interesting features about GcnQ. First, GcnQ demonstrates
broad substrate specificity with regards to the aglycon; all
angucyclic, tricyclic, and tetracyclic rings served as effective
substrates. Second, oxidation of the terminal sugar (either
di- or trisaccharide) at C-4 to its keto form and subsequent
desaturation were found to occur in tandem, indicating that
the L-rhodinose moiety is readily converted to the L-aculose
moiety and does not involve the intermediacy of L-ciner-
ulose A. This path deviates from the related aclacinomycin
case.13 Third, for both the native substrate 1a and non-
native substrates 2a and 3a, the L-rhodinose components of
both di- and trisaccharide chains were found to be very
efficiently desaturated to their L-aculose moieties though
desaturation of the disaccharide L-rhodinose preceeds that
of the trisaccharide. In the case of 3a, only the disaccharide
element serves as a substrate for desaturation.
In summary, we have conclusively identified the gene
cluster responsible for GCN biosynthesis. Inactivation of
GcnQ, a bicovalent FAD-bonded enzyme, affords a strain
that accumulates biosynthetic intermediate 1a, a hydro-
lysis dependent shunt product 2a, and a UV-dependent
rearranged product3a; both 2aand 3acontainL-rhodinose
moieties at both termini of their tri- and disaccharide chains.
In vitro experiments demonstrate that GcnQ catalyzes the
conversion of L-rhodinose residues into L-aculose in both
the tri- and disaccharide elements of 1aꢀ3a, to afford 1b, 2b,
and 3c. Importantly, 1b and 2b are generated through the
intermediacy of 1c and 2c, respectively. However, GcnQ
fails to analogously convert 3c to product 3b. Nonetheless,
these studies demonstrate that GcnQ is compatible with a
broad range of substrates. The GCN gene cluster, when
heterologously expressed in S. coelicolor M512, yielded
P-1894B (1b) as the sole product, supporting divergent roles
of GcnQ in different hosts. These findings support the
notion that GcnQ may be a prime candidate for combina-
torial biosynthetic generation of new non-natural products
based on the grincamycin scaffold.
Figure 3. GcnQ time-course experiments involving HPLC ana-
lyses of GcnQ-containing reactions in which 1a (A), 2a (B), 3a
(C), and 1 (D), 2 (E), 3 (F) were incubated with GcnQ for (II)
1.5 min, (III) 5 min, (IV) 30 min, and (V) 60 min. Trace I is a
negative control (1 h) with boiled GcnQ.
Acknowledgment. This work was supported by MOST
(2012AA092104, 2010CB833805), NSFC (31290233,
41106138), and STPPGP (2011B031200004).
Compounds 1b and 2b were confirmed to be P-1894B
and vineomycin B2 by comparisons and analyses of the 1H,
13C, and 2D NMR data with previously reported data.4,17
Intermediates 1cꢀ3c were identified on the basis of MS,
1H and 13C NMR data analyses (Tables S7ꢀS9). HRMS
spectra revealed that 1c had the molecular formula C49H62O18.
1H and 13C NMR spectroscopic data of 1c resembled those
of 1b. Close comparisons of these data sets revealed the
absence of two methylenes in 1b.Instead,asetofsignalsatδH
Supporting Information Available. Detailed experimen-
tal procedures, NMR data and spectra for compounds 1a,
2a, 3a, 1b, 2b, 1c, 2c, and 3c. This material is available free
of charge via the Internet at http://pubs.acs.org.
The authors declare no competing financial interest.
D
Org. Lett., Vol. XX, No. XX, XXXX