翻页电子书制作,电子书制作,电子杂志制作

云展网——上百万用户在此分享了PDF文档。上传您的PDF转换为3D翻页电子书,自动生成链接和二维码(独立电子书),支持分享到微信及网站!

Harnessing the intracellular triacylglycerols for titer improvement of polyketides in Streptomyces—翻页版预览

上传者:小笙 上传时间:2020-01-08 15:18:56 分享阅读:
小笙 上传于 2020-01-08 15:18:56

Harnessing the intracellular triacylglycerols for titer improvement of polyketides in Streptomyces

Articles

https://doi.org/10.1038/s41587-019-0335-4

Harnessing the intracellular triacylglycerols for
titer improvement of polyketides in Streptomyces

Weishan Wang   1,2,10*, Shanshan Li   3,10, Zilong Li1,10, Jingyu Zhang2,10, Keqiang Fan1, Gaoyi Tan   2,

Guomin Ai1, Sin Man Lam4, Guanghou Shui4, Zhiheng Yang2, Hongzhong Lu2, Pinjiao Jin3, Yihong Li1,

Xiangyin Chen2, Xuekui Xia5, Xueting Liu2,6, H. Kathleen Dannelly7, Chen Yang8, Yi Yang   2,

Siliang Zhang2, Gil Alterovitz9, Wensheng Xiang   3* and Lixin Zhang   2*

Pharmaceutically important polyketides such as avermectin are mainly produced as secondary metabolites during the sta-
tionary phase of growth of Streptomyces species in fermenters. The source of intracellular metabolites that are funneled into
polyketide biosynthesis has proven elusive. We applied multi-omics to reveal that intracellular triacylglycerols (TAGs), which
accumulates in primary metabolism, are degraded during stationary phase. This process could channel carbon flux from both
intracellular TAGs and extracellular substrates into polyketide biosynthesis. We devised a strategy named ‘dynamic degrada-
tion of TAG’ (ddTAG) to mobilize the TAG pool and increase polyketide biosynthesis. Using ddTAG we increased the titers of
actinorhodin, jadomycin B, oxytetracycline and avermectin B1a in Streptomyces coelicolor, Streptomyces venezuelae, Streptomyces
rimosus and Streptomyces avermitilis. Application of ddTAG increased the titer of avermectin B1a by 50% to 9.31 g l−1 in a 180-m3
industrial-scale fermentation, which is the highest titer ever reported. Our strategy could improve polyketide titers for
pharmaceutical production.

Streptomyces produce many biologically active polyketides, Previous studies have shown that the shift from growth to second-
including antibiotics in active clinic use, immunosuppres- ary metabolite production is accompanied by a series of transcrip-
sants, antiparasitics and antitumor agents1–3. The biology tomic and proteomic switches8,15,16. Both the levels of central carbon
of Streptomyces and the biochemistry of polyketide production metabolic intermediates and the transcripts of the corresponding
are well understood4–7. Streptomyces undergo a metabolic switch pathways decline after transition phase during fermentation8,15–18.
from primary to secondary metabolism during the course of Since the building blocks and energy for secondary metabolite bio-
fermentation4,8. During primary metabolism, cells consume exter- synthesis come from the primary metabolic pathways either directly
nal nutrients and grow rapidly. When external nutrients become or indirectly, we hypothesized that some intracellular metabolites
limited, cells stop growing and begin to produce secondary metab- might contribute to polyketide production in stationary phase
olites such as polyketides9,10. How do Streptomyces specifically (Fig. 1a). We reasoned that manipulating the pathways relating to
shift metabolic flux to extensive biosynthesis of secondary metabo- these intracellular metabolites might provide a new strategy for
lites at stationary phase? Although substantial progress has been polyketide titer improvement.
made in understanding transcriptional regulation during the
Streptomyces lifecycle5,11, the relationship between intracellular We report integrated metabolomic and transcriptomic data for
metabolites and polyketide production in stationary phase remains S. coelicolor in supplemented minimal medium (SMM) medium19
poorly understood. Understanding how metabolism alters during (phosphate limited medium for antibiotic production). By com-
growth could provide new ways to optimize metabolic engineering paring wild-type and high-yielding strains, we identified the
in Streptomyces species. accumulated cellular TAG pool as an intracellular carbon source
for polyketide biosynthesis during stationary phase. Moreover,
When Streptomyces are cultivated on solid media, substrate we demonstrated that mobilization of cellular TAG enables car-
mycelia are autolyzed in response to nutrient depletion during late bon flux to be redirected to polyketide biosynthesis. We devised
stationary phase, which releases nutrients for secondary metabo- a new ddTAG strategy that increases polyketide titers and applied
lite biosynthesis and morphological differentiation12. However, the ddTAG in four Streptomyces species to increase yields of actinorho-
same response to nutrient depletion is not observed in industrial wfdeiernmi(neAcnrcteta)at,sijeoadndotfhmreoymtcitien6r.Bo2,f0oaxtvoyetre9mt.r3ea1cc tgyi ncl−li1Bn, 1eahiaingnhdali1ag8vh0etr-inmmge3 citnthidneuBust1srae.iaFoli-fnsacoalullyer,
fermentations for polyketide production, where the carbon source ddTAG strategy.
is present at a constantly low concentration and other essential
nutrients (such as phosphate) are limited at stationary phase13,14.

1State Key Laboratory of Microbial Resources, Institute of Microbiology, Chinese Academy of Sciences, Beijing, China. 2State Key Laboratory of Bioreactor
Engineering, East China University of Science and Technology, Shanghai, China. 3State Key Laboratory for Biology of Plant Diseases and Insect Pests,
Institute of Plant Protection, Chinese Academy of Agricultural Sciences, Beijing, China. 4State Key Laboratory of Molecular Developmental Biology,
Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, Beijing, China. 5Key Biosensor Laboratory of Shandong Province, Biology
Institute, Qilu University of Technology (Shandong Academy of Sciences), Jinan, China. 6Laboratory for Marine Biology and Biotechnology, Qingdao
National Laboratory for Marine Science and Technology, Qingdao, China. 7Indiana State University, Terre Haute, IN, USA. 8CAS-Key Laboratory of Synthetic
Biology, CAS Center for Excellence in Molecular Plant Sciences, Shanghai Institute of Plant Physiology and Ecology, Chinese Academy of Sciences,
Shanghai, China. 9Boston Children’s Hospital, Boston, MA, USA. 10These authors contributed equally: Weishan Wang, Shanshan Li, Zilong Li, Jingyu Zhang.
*e-mail: wangws@im.ac.cn; xiangwensheng@neau.edu.cn; lxzhang@ecust.edu.cn

Nature Biotechnology | www.nature.com/naturebiotechnology

Articles Nature Biotechnology

a Primary metabolism Polyketide biosynthesis b 12 M145 1 150 c Identified metabolites Pathway distribution
HY01
Polyketides Glucose concentration (g l–1) 10
120
Cell growth 8 Act titer (mg l–1) 30 8 8 7
Cell growth (A595) 90 143
6 0.1 3
60 12
Essential nutrients 4 633 unknown 35
9
Time 2 30 8 13 10

Metabolic 0 0.01 0 TCA NUM FMM
Switch 0 12 24 36 48 60 72 84 96
Time (h) EMP AAM GM GDM
PPP LPM PGI Other

d 2.5 20 h 36 h 48 h e 3.0 TAGs f 0.33%
60 h 72 h 96 h PLs 0.45%
2.0 2.5 4.14% 1.08%
Relative levels of metabolites 1.5 Relative levels of different lipids 0.27% 11.23% TAG-iC13:0
1.0 ** 2.0 4.53% 0.87% TAG-aiC13:0
1.31% TAG-iC14:0
1.5 4.45% 13.93% TAG-C14:0
0.29% TAG-iC15:0
NS NS NS NS TAG-aiC15:0
20.89% TAG-C15:0
1.0 TAG-iC16:0
TAG-C16:1-Δ9
0.5 0.5 32.03% TAG-C16:0
TAG-iC17:0
4.18% TAG-aiC17:0
TAG-C17:0
0 0 TTAAGG--CC1188::10-Δ9
EMP 12 24 36 48 60 72 84 96
PPP TCA AAM LPM
Time (h)
Metabolic pathway

Fig. 1 | Identification of intracellular metabolites with relationship to polyketides production. a, Schematic of the switch between primary metabolism
and polyketide biosynthesis. b, Glucose consumption, cell growth and Act production of the fermentation run in SMM medium. Purple dot indicates
sampling time point for metabolome analysis. Data shown are the average and standard deviation (s.d.) of three independent experiments. c, Classification
of identified metabolites gathered by GC–MS. Left, number of identified (orange) and undetermined (gray) metabolites; Right, distribution of the identified
metabolites. NUM, nucleotide metabolism; FMM, fructose and mannose metabolism; GDM, glyoxylate and dicarboxylate metabolism; LPM, lipid
metabolism; and PGI, pentose and glucuronate interconversions. Metabolite details are listed in Supplementary Table 2. d, Trends of different metabolic
pathways in M145. Data represent the relative average level and s.d. of all identified metabolites in the corresponding pathway (details are shown in
Supplementary Fig. 1a–f). Differences were analyzed by Student’s t-test, and P < 0.05 was considered statistically significant. ***P < 0.001, **P < 0.01,
*P < 0.05, NS, not significant. e, Temporal profile of PLs and TAGs by TLC. For d and e, data from 20 h were set as one; data shown are the average
and s.d. of five independent experiments. f, Proportion of different fatty acid moieties of TAGs in M145 sampled at 48 h. TAG-iC13:0, isotridecanoic
acid; TAG-aiC13:0, anteisotridecanoic acid; TAG-iC14:0, isomyristic acid; TAG-C14:0, myristic acid; TAG-iC15:0, isopentadecanoic acid; TAG-aiC15:0,
anteisopentadecanoic acid; TAG-C15:0, pentadecanoic acid; TAG-iC16:0, isopalmitic acid; TAG-C16:0, palmitic acid; TAG-iC17:0, isoheptadecanoic
acid; TAG-aiC17:0, anteisoheptadecanoic acid; TAG-C17:0, heptadecanoic acid; TAG-C18:0, stearic acid; TAG-C16:1-Δ9, 9-hexadecenoic acid and
TAG-C18:1-Δ9, 9-octadecenoic acid.

Results (Fig. 1d and Supplementary Fig. 1f). Similar trends were observed
Cellular TAG is used as a carbon source in stationary phase. First, in the high-yielding strain HY01 (Supplementary Fig. 1g–l). We
we carried out comparative metabolomic analyses of S. coelicolor hypothesized that FFAs and MAGs were more likely related to our
strain M145, a model strain that produces the well-known polyketide seeking candidates contributing to polyketide biosynthesis during
Act, and HY01, a mutant strain derived from M145 that overpro- stationary phase.
duces Act, during a 96 h time course (Fig. 1b). Metabolomes were
obtained by gas chromatography–mass spectrometry (GC–MS) FFAs and MAGs are intermediates of lipid metabolism. In
(Methods) (Supplementary Note 1 and Supplementary Table 1). Streptomyces, phospholipids (PLs) and TAGs are the main com-
A total of 143 known metabolites from nearly all the main meta- ponents of cellular lipids20,21. As PLs and TAGs cannot be directly
bolic pathways were identified (Fig. 1c and Supplementary Table 2). detected using general GC–MS analytical methods for metabolomic
In M145 we found that the metabolites related to primary meta- data collection, we used thin-layer chromatography (TLC) to char-
bolic pathways, including glycolysis (Embden–Meyerhof–Parnas acterize the PL and TAG profiles of the M145 and HY01 strains dur-
pathway (EMP)), the pentose phosphate pathway (PPP), the tricar- ing the 96 h time course. TAGs and PLs accumulated gradually up to
boxylic acid cycle (TCA) and amino acid metabolism (AAM), accu- 48 h, but thereafter they had distinct profiles (Fig. 1e). The amount
mulated during exponential phase (12–36 h), then decreased to low of PLs remained relatively constant, whereas TAGs decreased dur-
levels (36–72 h) and remained relatively constant throughout sta- ing stationary phase. To examine TAG degradation in more details,
tionary phase (72–96 h) (Fig. 1d and Supplementary Fig. 1a–e). We we hydrolyzed the TAGs from our time-course samples, and quan-
noted a difference between patterns of metabolites in the lipid meta- tified fatty acid moieties using GC–MS (Supplementary Fig. 2a).
bolic pathway with that of the EMP, PPP, TCA and AAM pathways. All 15 detected fatty acid moieties of cellular TAGs in M145 (Fig. 1f)
Lipid metabolic pathway metabolites, including free fatty acids had dynamic profiles that corresponded to our TLC data
(FFAs) and monoacylglycerols (MAGs), reached maximum accu- (Supplementary Fig. 2b). We inferred that the TAG pool, which is
mulation around 48 h and then declined continuously thereafter catabolized into FFAs and MAGs, might represent one group of our
proposed key intracellular metabolites for polyketide biosynthesis.

Nature Biotechnology | www.nature.com/naturebiotechnology

Nature Biotechnology Articles

a 25 M145 b TAGTA-CG1-6iC:11-5Δ:90 c

HY01Consumed fatty acid moieties TAG-aiC17:0 Glucose 12 100 Relative level of TAGs2.0
of TAGs (mg l–1) AcCoA concentration (g l–1) 80 Act titer (mg l–1) 1.5
20 Metabolite 9 60 1.0
TAG-iC16:0 40 0.5
15 TAG-C15:0 72–96 h 6 20 0.0
TAG-iC14:0 0
10 TAG-aiC15:0 234 3 96
TAG-C14:0 VIP score
5 0
MalCoA 0
0 Unknown 173

CC1186a::aaiiiiiiii11--CCCCCCCCCCCCC1111111111111ΔΔ777555443366899 TAG-C16:0 24 48 72
TAG-iC17:0 Time (h)
Unknown 88

DHAP
TAG-iC13:0
TAG-C17:0
MG(18:0/0:0/0:0)
Unknown 592
TAG-C18:1-Δ9

1

d 1.00 C17:0 e Fatty acid biosynthesis f Fatty acid degradation (β-oxidation)
iC14:0
Correlation coefficient 0.96 aiC15:0 –3.0 0 3.0 4 –3.0 0 3.0 4
0.92 aiC17:0 3 sco1069
–0.96 C14:0 Relative transcript level 2 sco2774 3
–1.00 C16:1-Δ9 1 sco3079 2
aiC13:0 0 sco6026
1.0 C17:0 sco6027 1
0.9 iC15:0 24 36 48 60 sco6731
0.8 C15:0 –1 sco6732 0
–0.9 C18:0 –2 Time (h) sco6787 24 36 48 60
–1.0 sco6788
18 h sco678918 h –1
24 h 24 h –2 Time (h)
30 h 30 h
36 h 36 h
42 h 42 h
48 h 48 h
60 h 60 h
Relative transcript level
Correlated to glucose (20–48 h) sco0548
Correlated to Act (72–96 h) sco1271
aiC13 sco1814
iC15 sco1815
C18 sco2387
aiC15 sco2388
iC14 sco2390
C16:1-Δ9 sco3246
aiC17 sco3248
C14 sco6564
iC17
C17
C15
iC16
iC13
C16

Fig. 2 | Cellular TAG pool contributes to Act yield during stationary phase. a, Comparison of the amount of consumed TAG pool between M145 and HY01.
Columns show the consumed amount of fatty acid moieties from cellular TAG pool between 48 h and 96 h. Data shown are the average and s.d. of three
independent experiments. b, Top 20 metabolites with significant change between M145 and HY01 during late stationary phase. VIP score indicates the
variation level of a metabolite related to Act production. Orange bars highlight the fatty acid moieties of cellular TAG pool. c, Profiles of glucose consumption
(black), TAG metabolism (orange) and Act production (blue). Light and thick orange curves indicate the profiles of fatty acid moieties of cellular TAG pool
and their overall trend, respectively. Data of fatty acid moieties obtained at 20 h were set as one. d, Hunting of metabolites with negative correlation to
glucose consumption (upper) and Act biosynthesis (lower). Fatty acid moieties of cellular TAG pool were indicated by orange bar. e,f, Transcriptional profiles
of genes involved in fatty acid biosynthesis (e) and β-oxidation pathways (f). Blue curves indicate the overall transcriptional trend of the selected genes.

Cellular TAG pool contributes to high yield. The high-yielding production in M145. The TAG pool accumulated during primary
strain HY01 produces more Act but consumes less glucose than M145 metabolism and degraded during Act biosynthesis (Fig. 2c). To
(Fig. 1b). We found that more intracellular TAGs were degraded in explore the role of TAG in more depth, we evaluated the relation-
HY01 during late stationary phase compared with M145 (Fig. 2a), ships between all 794 intracellular metabolites and the uptake of
although the amount of cellular TAGs in each strain were almost glucose, as well as the production of desired polyketide Act through-
identical in transition phase (Supplementary Fig. 3a). Moreover, out the time course. We found that the accumulation patterns of
we observed that TAG-degradation products, including glycerol 14 TAG fatty acid moieties were negatively correlated with glucose
3-phosphate, 3-phosphoglycerate and dihydroxyacetone phosphate, use in primary metabolism (20–48 h; r < −0.8, P < 0.001) (Fig. 2d
accumulated to higher levels in HY01 than in M145 during station- and Supplementary Table 3), and that the degradation patterns for
ary phase (Supplementary Fig. 3b–e). Therefore, the degradation of almost all the fatty acid moieties from the TAG pool were consis-
cellular TAG pool may be the cause of the high-yielding strain HY01 tently among the most negatively correlated with the accumulation
producing more Act while consuming less glucose than M145. pattern of Act in late stationary phase (72–96 h; r < −0.8, P < 0.001)
in M145. Notably, the first nine of the top ranked metabolites iden-
To evaluate the contribution of the cellular TAG pool to Act tified at the late stationary phase were all TAG fatty acid moieties
production, we carried out a comparative analysis of metabolome (Fig. 2d and Supplementary Table 3). Similar correlations were
data (includes 796 putative metabolites) obtained from different also observed in the high-yielding strain HY01 (Supplementary
analytical approaches in both M145 and the high-yielding strain Table 4). Next, we examined the expression of genes related to
HY01 (Supplementary Note 2) over a time course of 96 h. As time TAG metabolism by profiling the time-course transcriptome of
progressed, the TAG pool increasingly contributed to Act produc- M145 in the same culture condition (Gene Expression Omnibus
tion as shown by their individual variable importance in projection (GEO) no. GSE53562), and found results consistent with cellular
(VIP) score (Supplementary Fig. 4a). The different levels of contri- TAG profiles. For example, transcripts of genes for fatty acid bio-
bution of the TAG pool to Act production among different strains synthesis were upregulated in primary metabolism and downregu-
became most pronounced in late stationary phase (Fig. 2b), which lated in Act production, whereas transcription of genes encoding
coincides with the biggest Act titer difference between M145 and β-oxidation enzymes was upregulated in Act production (Fig. 2e,f
HY01. By constrast, the contribution to Act production by other and Supplementary Table 5). We also reanalyzed previously pub-
metabolites (such as amino acids) decreased (Supplementary lished time-course transcriptome datasets for S. coelicolor grown
Fig. 4b). These results indicate that the cellular TAG pool contrib- in diverse media and culture conditions8,17,22,23, and found similar
utes to the higher Act yield in HY01. trends (Supplementary Fig. 5a–d).
TAG pool in polyketide-producing Streptomyces. We profiled
the concentration of cellular TAG, glucose consumption and Act Finally, we analyzed the cellular TAG pool in three industrial
species, avermectin producer S. avermitilis A56 (ref. 24), milbemycin

Nature Biotechnology | www.nature.com/naturebiotechnology

Articles Nature Biotechnology

a 1.8 × 106 Oleate feeding b Glucose feeding

1.5 × 106 Unlabeled 1.2 × 107 Unlabeled
1.2 × 106 Labeled 0.9 × 107 Labeled
M145 M145
HY01 HY01

Peak area
Peak area
0.9 × 106 0.6 × 107

0.6 × 106

0.3 × 107

0.3 × 106

0 0

γ-Act γ-Act
γ′-Act γ′-Act

Act Act
Unknown-Act Unknown-Act

ε-Act ε-Act

c 0.5 M145 HY01 d8 M145 HY01 e8 M145 HY01 i Glucose feeding (–) M145 HY01
TAG pool
0.4 *** *** *** *** *** *** Glc PYR

6 6
NADH/NAD+ ratio
ATP/ADP ratio0.3 10.88 9.81 16.62 35.59

NADH/NADP+ ratio4 4

0.2 AcCoA 2.04 1.93
Biomass
2 2
0.1
7.18 2.19 18.28 41.28

0 96 h 0 96 h 0 96 h TCA cycle Act
72 h 72 h 72 h

f 100 KGDHRelative activity (%) Relative activity (%)g 100 IDH Relative activity (%)h 100 CS Glucose feeding (+) TAG pool
Glc PYR
75 75 75
50 50 50 105.10 101.47 11.73 26.93
25 25 25 76.22 55.08
AcCoA 17.26 17.12
0 Biomass

23.35 56.20

0 0

NADH ATP NADH ATP NADH ATP TCA cycle Act
ATP ATP ATP

+ + +

NADH NADH NADH

Fig. 3 | High-yielding mechanism of mobilizing cellular TAG pool. a, Tracing the carbon flux from TAG degradation to Act biosynthesis using [U-13C]
oleate. b, Tracing the carbon flux from glucose to Act biosynthesis by feeding [U-13C] glucose. For a and b, Act here includes five detected congeners
produced by the S. coelicolor (Supplementary Note 3). c–e, Comparison of the NADH/NAD+(c) ATP/ADP (d) and NADPH/NADP+ (e) in M145 and
HY01 during stationary phase. f–h, Influence of NADH and/or ATP on the activity of KGDH (f), IDH (g) and CS (h). The enzyme activity of the M145 cell
extracts without adding NADH and ATP was set as 100%. i, Mass balance of the acetyl-CoA (AcCoA) node. Fluxes were expressed in unit of the acetyl
unit. Data for MFA were collected during stationary phase (72–120 h). Data shown were the average of three independent replicates. The actual fluxes
of the main metabolic network are shown in Supplementary Fig. 10. Reactions, stoichiometric equations and experimentally determined flux are listed in
Supplementary Tables 6–8, respectively. For a–h, the data shown were the average and s.d. of three independent experiments. Differences were analyzed
by one-way ANOVA and P < 0.05 was considered statistically significant. ***P < 0.001, **P < 0.01, *P < 0.05.

producer Streptomyces bingchenggensis BC04 (ref. 25) and oxytet- After 24 h cultivation, we observed more labeled and unlabeled Act
racycline producer S. rimosus M4018::2SFotcR (M2R)26 in a time- in HY01 than in M145 (Fig. 3a and Supplementary Note 3). Then,
course experiment. We observed the same dynamic profiles of we tested acetyl-CoA, the main precursor for Act biosynthesis and
cellular TAG pool in these strains (Supplementary Fig. 5e–g). Based found that the amount of labeled acetyl-CoA in HY01 was 1.63-fold
on these data we hypothesize that the cellular TAG pool is accumu- higher than in M145 (Supplementary Fig. 7a), indicating a faster
lated during growth and then degraded in stationary phase, when degradation rate of exogenous oleate and intracellar TAGs in HY01
polyketides are produced. compared with M145. The labeling ratio of acetyl-CoA in HY01 was
Metabolic flux during TAG breakdown. To better understand the lower (P < 0.05) than in M145 (Supplementary Fig. 8a), indicating
breakdown of the TAG pool in HY01, we first fed uniformly labeled more intracellular TAGs were degraded in HY01 and thus dilute the
[U-13C] oleate to trace carbon flux from intracellar TAGs at station- labeled acetyl-CoA.
ary phase (72 h) when Act is produced (Supplementary Fig. 6a).
We also tested whether the carbon flux from glucose contrib-
utes to Act biosynthesis in stationary phase. After feeding [U-13C]

Nature Biotechnology | www.nature.com/naturebiotechnology

Nature Biotechnology Articles

a b c Group d M145
HY01
TAGs 100 8 6
80 5 Relative transcription level
FAs Sequence identity (%) 60 Relative transcription level 5
40 1 2345
Acyl-CoA 20
synthetase 0 6

3 4
44
Groups 3

22 2

1

10 0
24 36 48 60 72 84 96
Acyl-CoA ssssscccccooooo72664241934468944863
(n + 2) Time (h)

β-oxidation e *** f g
*** 8
160 TAG
120 6 10
Act titer (mg l–1) FA 5 9 1112 13 14 15
MAG 12 34
DAG 7
PL M145
80
AcCoA 40

6196DM

0 6196OE
20 21 22 23 24 25 26 27 28 29 30 31 32
Polyketides M145 661199M661DO4EM5 Time (min)
6196DM
6196OE

Fig. 4 | Identification of SCO6196 responsible for cellular TAG mobilization. a, A brief illustration of metabolic pathway from cellular TAG pool

to polyketides. b, Pairwise sequence identity analysis among the 888 ACSs in 125 Streptomyces genomes. Protein sequences of ACSs are listed in

Supplementary Table 10. c, Relative transcript levels of the five conserved ACSs in S. coelicolor M145 at 72 h. d, Temporal profiles of sco6196 transcripts in

M145 and HY01. e, Effect of SCO6196 on Act production. f, TLC assay the remaining cellular TAG pools in M145, 6196DM and 6196OE at 96 h. g, Levels

of fatty acid moieties from remaining cellular TAG pools of M145, 6196DM and 6196OE. The peaks 1 to 15 are iC13:0, aiC13:0, iC14:0, C14:0, iC15:0,

aiC15:0, C15:0, iC16:0, C16:1-Δ9, C16:0, iC17:0, aiC17:0, C17:0, C18:1-Δ9 and C18:0, respectively. Data shown in c, d and e are the average and s.d. of three
independent experiments. Significant differences were analyzed by one-way ANOVA, and P < 0.05 was considered statistically significant. ***P < 0.001,
**P < 0.01, *P < 0.05.

glucose at 72 h (Supplementary Fig. 6b), we observed that the activities of KGDH, IDH and CS in S. coelicolor cell extracts with
amount of labeled acetyl-CoA in HY01 and M145 was almost iden- or without adding exogenous NADH and/or ATP. We found that all
tical (Supplementary Fig. 7b), indicating there a lack of notable the three enzymes were inhibited by the addition of NADH or ATP
differences in glycolysis pathway between M145 and HY01. The alone, and the largest extent of inhibition was observed when both
amount of unlabeled acetyl-CoA in HY01 was 2.05-fold higher NADH and ATP were added (Fig. 3f–h). This finding supports the
than in M145 (Supplementary Fig. 7b), again demonstrating that conclusion that the mobilization of cellular TAG pool weakened the
more cellular TAG pool was degraded in HY01. Although the carbon flux toward TCA cycle.
intake of labeled glucose and the cellular levels of labeled acetyl-
CoA was almost identical (Supplementary Fig. 7b), we observed To quantify the flux distribution of the proposed mechanism, we
more labeled Act in HY01 (Fig. 3b), suggesting that the TAG deg- constructed a simplified model to perform a straightforward meta-
radation in HY01 enables more carbon flux out of glucose to be bolic flux analysis (MFA) during stationary phase (Supplementary
channeled to Act production. Carbon flux from glucose and TAG Fig. 10 and Supplementary Tables 6–8). Metabolic flux data of fer-
is routed to the common intermediate acetyl-CoA, and then flows mentations with and without feeding showed that cellular TAG
into Act biosynthetic pathway and TCA cycle at stationary phase. pool degradation in HY01 can provide more acetyl-CoA build-
We next examined the carbon flux from acetyl-CoA to TCA cycle. ing blocks and redirects more carbon flux out of this node toward
We observed that, in both labeled oleate and glucose feeding experi- polyketide production by depressing the TCA cycle (Fig. 3i). To fur-
ments, the labeling ratios of all detected TCA intermediates were ther verify this flux redistribution, we also implemented flux sam-
lower in HY01 than those in M145 (Supplementary Fig. 8), indicat- pling analysis using the genome-scale metabolic model (GEM) of
ing the carbon flux from acetyl-CoA to TCA cycle was weakened S. coelicolor (iKS1317)28 and observed consistent results with MFA
in HY01. These results support the hypothesis that in HY01 more (Supplementary Fig. 11). These data suggested that rational mobili-
acetyl-CoA is diverted into Act production. zation of TAG might improve polyketide titers.
Genes involved in cellular TAG mobilization. Next, we focused
In both stable isotope feeding experiments, the cellular level of on the identification of key gene(s) involved in degradation of TAG.
labeled and unlabeled TCA intermediates in HY01 was lower than Fatty acyl-CoA synthetase (ACS) is an essential enzyme for lipid
in M145, except for α-ketoglutarate (Supplementary Fig. 7c–j). We catabolism, activating fatty acids by thioesterification with coen-
hypothesized that α-ketoglutarate dehydrogenase (KGDH), which zyme A (CoA) to enter β-oxidation cycle21 (Fig. 4a). Consistently,
consumes α-ketoglutarate, might be inhibited in HY01. It is reported we observed that the transcripts of genes responsible for β-oxidation
that the activities of KGDH, isocitrate dehydrogenase (IDH) and were all upregulated at stationary phase (Fig. 2f), whereas the tran-
citrate synthase (CS) in the TCA cycle are repressed by high levels of script profiles of ACS homologs were irregular (Supplementary Table
reducing equivalents and ATP in many species27. Our data showed 9). This indicates that fatty acyl-CoA synthesis is more likely the con-
that, in addition to acetyl-CoA, TAG degradation via multi-cycle trol point of TAG degradation in Streptomyces and that the appropri-
β-oxidation generates more reducing equivalents and ATP in HY01 ate ACS(s) will need to be identified to enable further engineering.
(Fig. 3c–e and Supplementary Fig. 9). We therefore measured the

Nature Biotechnology | www.nature.com/naturebiotechnology

Articles Nature Biotechnology

a sco6196 b 220 Glucose concentration (g l–1)c 12 M145 Feeding (+) 800
sco6196 200 M145-DT Feeding (–) 600
Cumate 30 180 10
OFF Cumate dosage (µM) 10 160 Act titer (mg l–1)8
ON 140 Act production (mg l–1)
5 120 6 400
3 100
1 80 4
0.5 200
0.1
0 2

30 36 42 48 54 60 0 0
Induction time point (h) 0 24 48 72 96 120 144 168 192

Time (h)

d 3.0 Feeding (–) e *** f ** g A56
A56-DT
Feeding (+) 150 12 10
9
Specific productivity of Act 2.4 ** JdB titer (mg l–1) 6 Avermectin B1a (g l–1) 8
(mg g–1 glucose per h) Otc titer (g l–1)
1.8 * 100 ***
NS
1.2 3 6 Induction
50 4
0.6
0 0 2
0 Strain ISP5230 Sv-DT Strain
M145 M145-DT Sv-DT M4018 M-DT M2R M2R-DT 0
10 µM 2 4 6 8 10 12 14
Cumate – – 16 h Cumate – 10 µM – 5 µM
72 h 60 h Time (day)

Fig. 5 | The ddTAG strategy for polyketide titer improvement. a, Schematic of temporal control of TAG mobilization. b, Determination of the optimal

conditions for Act production in M145-DT. Color bar indicates the titer of Act. c, Performance of ddTAG strategy under the feeding condition. d, Effect

of ddTAG strategy on specific productivity in fermentations with (+) or without (−) glucose feeding. The specific productivity was calculated from 72
to 192 h. e, Jadomycin B (JdB) titer improvement using the ddTAG strategy. S. venezuelae ISP5230 and Sv-DT are the parent and the engineered strain,

respectively. f, Oxytetracycline (Otc) titer improvement using the ddTAG strategy. M-DT and M2R-DT are engineered strains derived from S. rimosus

M4018 and M2R, respectively. g, Avermectin B1a titer improvement using the ddTAG in a 180-m3 fermenter. A56 and A56-DT are the start strain and the
engineered S. avermitilis strain, respectively. Representative data is shown. The arrow indicates the time point of adding inducer (7.5 μM of cumate). Data
shown in c–f are the average and s.d. of three independent experiments. Significant differences were analyzed by one-way ANOVA, and P < 0.05 was
considered statistically significant. ***P < 0.001, **P < 0.01, *P < 0.05.

Given that accumulation of a TAG pool is widespread in acti- into M145 to yield strain M145-DT and found that modulation of
nomycetes21, we implemented a pairwise sequence identity assay TAG degradation using the inducer can increase Act titers (Fig. 5b).
among the 888 ACSs present in 125 Streptomyces genomes in NCBI In optimal conditions (10 μM of cumate at 48 h), the Act titer of
(Supplementary Table 10), and found five groups of ACSs that the M145-DT reached 216.1 ± 15.1 mg l−1, which is 190% and 58%
are conserved in S. coelicolor and other Streptomyces (Fig. 4b). We higher than that of either M145 or the high-yielding strain HY01,
analyzed the transcription of ACS genes in S. coelicolor belonging respectively (Supplementary Fig. 14c). We analyzed the TAG pool
to these five groups using real-time quantitative PCR at 72 h. The (from 48 to 96 h) (Supplementary Fig. 14) and checked the meta-
expression of sco6196 was significantly higher in HY01 compared bolic flux in M145-DT (Supplementary Fig. 10b), and our data sup-
with M145 (Fig. 4c). Further, we found that the amount of sco6196 ported the notion that rational mobilization of cellular TAG pool
transcript was increased in HY01 in stationary phase (Fig. 4d). can provide and redirect more carbon flux toward Act synthesis. We
Therefore, we hypothesized that SCO6196 might be responsible for also observed an increased cellular level of reducing equivalents in
degradation the TAG pool. the M145-DT strain (Supplementary Fig. 14e).

To analyze the role of SCO6196 further, we created a sco6196 In industrial fermentation, carbon-source feeding is necessary to
deletion strain (6196DM) and a sco6196 overexpression strain maintain productivity. To test whether our ddTAG strategy enables
(6196OE) in M145 (Supplementary Fig. 12a,b). Compared with an engineered strain to maintain higher productivity with a feed-
parental strain M145, 6196DM produced substantially less Act ing carbon source at stationary phase, we fed glucose at 72 h in
(Fig. 4e and Supplementary Fig. 12c,d), and harbored more cellu- both M145 and M145-DT at the determined induction condition
lar TAG pool in late stationary phase (Fig. 4f and Supplementary (Fig. 5c). We found that M145-DT in the feeding condition gives
Fig. 12e). 6196OE had the opposite phenotype (Fig. 4e,f). We fur- the highest specific productivity, increasing by 50.1% compared to
ther analyzed the fatty acid moieties of the cellular TAG pool, and that of M145 (Fig. 5d). This result shows that selective control of
showed that deletion of sco6196 blocked the degradation of cellular TAG degradation using our ddTAG strategy is effective to improve
TAGs with a broad range of fatty acid moieties (Fig. 4g). The sub- polyketide titers in fed-batch fermentations.
strate promiscuity of SCO6196 was further confirmed by in  vitro Broad application of the ddTAG strategy. To investigate the
biochemical data (Supplementary Fig. 13). applicability of our ddTAG strategy in other polyketide-producing
ddTAG strategy for polyketide titer improvement. Finally, we Streptomyces species, we applied it in S. venezuelae ISP5230 that
placed the sco6196 gene under the control of a cumate-inducible synthesizes jadomycin B30. Under optimized induction conditions,
promoter29 to generate a ddTAG module, to enable selective con- engineered strain Sv-DT produced 133.0 ± 9.4 mg l−1 of jadomycin
trol of the timing and strength of TAG degradation (Fig. 5a and B in 48 h, which is 1.7-fold higher than the parent strain ISP5230
Supplementary Fig. 14a,b). We transformed the ddTAG plasmid (Fig. 5e). We also used ddTAG in the parental oxytetracycline

Nature Biotechnology | www.nature.com/naturebiotechnology

Nature Biotechnology Articles

producer S. rimosus M4018 and the engineered strain M2R (ref. 26) References
with the optimized expression of OtcR activator. ddTAG strategy
enabled higher yields from strains M-DT and M2R-DT, with titers 1. Chater, K. F. Streptomyces inside-out: a new perspective on the bacteria
of oxytetracycline increasing from 0.96 to 4.54 g l−1 and from 6.24 to that provide us with antibiotics. Phil. Trans. R. Soc. Lond. B. 361,
9.17 g l−1, respectively (Fig. 5f and Supplementary Fig. 14f). 761–768 (2006).
Finally, we used our ddTAG strategy in the ionbdtuaisntreidalaahveigrhm-eycietilndiBn1ga
producer strain S. avermitilis A56 (ref. 24). We 2. Zhang, L. & Demain, A. L. Natural Products: Drug Discovery and Therapeutic
strain A56-DT with 95.26% titer improvement in fed-batch fermen- Medicine 1st edn (Humana Press, 2005).
tation in a shake flask (Supplementary Fig. 14g), and we detected
mFigo.re15caarnbdonSuflpupxletomweanrtdarayveTrambleecsti1n1–B11a3p).roTdouacstcioernta(iSnupwphleemtheenr toauryr 3. Ashforth, E. J. et al. Bioprospecting for antituberculosis leads from microbial
ddTAG strategy could be scaled up, we also carried out experiments in metabolites. Nat. Prod. Rep. 27, 1709–1719 (2010).
a stirred-tank bioreactor. In a 180-m3 fermenter, the engineered strain
tAio5n6-iDncTreparsoeddufcreodm560.%20htiogh9.e3r1t git le−r1s(oFfiga.v5egr)m. TechtiisnisBa1a,vweriythhpigrhodtiutecr- 4. Hwang, K. S., Kim, H. U., Charusanti, P., Palsson, B. O. & Lee, S. Y. Systems
of avermectin B1a in an industrial fermentation. biology and biotechnology of Streptomyces species for the production of
secondary metabolites. Biotechnol. Adv. 32, 255–268 (2014).
Discussion
Understanding the shift from primary metabolism to secondary 5. Liu, G., Chater, K. F., Chandra, G., Niu, G. & Tan, H. Molecular regulation
metabolism could enable the development of strategies to improve of antibiotic biosynthesis in Streptomyces. Microbiol. Mol. Biol. Rev. 77,
the production of commercially relevant secondary metabolites 112–143 (2013).
such as antibiotics11,31. In the polyketide-producing Streptomyces, it
has remained unclear which intracellular metabolites contribute to 6. Kang, Q., Bai, L. & Deng, Z. Toward steadfast growth of antibiotic research in
polyketide biosynthesis in stationary phase and how metabolic flux China: from natural products to engineered biosynthesis. Biotechnol. Adv. 30,
is mobilized to produce polyketides in fermentations. In this report, 1228–1241 (2012).
we addressed this problem and in so doing identified a role for the
TAG pool in generation of precursors for polyketide biosynthesis, 7. Wang, Q. et al. Abyssomicins from the South China Sea deep-sea
and in redirection of the carbon flux into polyketide production sediment Verrucosispora sp.: natural thioether Michael addition adducts
during stationary phase. as antitubercular prodrugs. Angew. Chem. Int. Ed. Engl. 52,
Streptomyces species are capable of accumulating TAGs during 1231–1234 (2013).
growth21. The rational control of TAG mobilization during sta-
tionary phase increases the cellular levels of acetyl-CoA, reduc- 8. Nieselt, K. et al. The dynamic architecture of the metabolic switch in
ing equivalents and ATP. The high level of reducing equivalents Streptomyces coelicolor. BMC Genomics 11, 10 (2010).
and ATP further inhibit the activities of the key enzymes in TCA
cycle, thus the carbon flux from the key metabolic branchpoint 9. Bibb, M. J. Regulation of secondary metabolism in streptomycetes.
acetyl-CoA to TCA cycle was weakened and that toward polyketide Curr. Opin. Microbiol. 8, 208–215 (2005).
production was enhanced. This redirection of carbon flux may be
the main reason why HY01 and other engineered strains produce 10. Alam, M. T. et al. Metabolic modeling and analysis of the metabolic switch in
more polyketides but consume less glucose compared with parental Streptomyces coelicolor. BMC Genomics 11, 202 (2010).
strains. We maintain that the TAG pool has an important role in
polyketide production in stationary phase and is not simply a com- 11. Gao, Q., Tan, G. Y., Xia, X. & Zhang, L. Learn from microbial intelligence for
peting pathway for polyketide biosynthesis. avermectins overproduction. Curr. Opin. Biotechnol. 48, 251–257 (2017).
Inspired by our data, we devised a ddTAG strategy to fine-
tune the timing and amount of TAG mobilization and in so doing, 12. Rigali, S. et al. Feast or famine: the global regulator DasR links nutrient
increase polyketide titers. Using this strategy, the titers of Act, jado- stress to antibiotic production by Streptomyces. EMBO Rep. 9,
mycin B, oxytetracycline avnednezauveelrame,eSc.tirnimBo1sauswaenrde substantially 670–675 (2008).
improved in S. coelicolor, S. S. avermitilis,
respectively; therefore, our strategy is broadly applicable. Consistent 13. Masuma, R., Tanaka, Y., Tanaka, H. & Omura, S. Production of nanaomycin
results in shake flasks and in an industrial bioreactor demonstrate and other antibiotics by phosphate-depressed fermentation using phosphate-
the vast potential of the ddTAG approach in commercial applica- trapping agents. J. Antibiot. 39, 1557–1564 (1986).
tions. Also, ddTAG can be combined with existing engineering in
industrial strains to further increase titers (Fig. 5f,g). 14. Mendes, M. V. et al. The two-component phoR-phoP system of Streptomyces
In summary, our results have enabled a better understanding natalensis: Inactivation or deletion of phoP reduces the negative phosphate
of cellular TAG metabolism and provided a simple tool for titer regulation of pimaricin biosynthesis. Metab. Eng. 9, 217–227 (2007).
improvement of the polyketides in Streptomyces.
15. Wentzel, A., Sletta, H., Stream, C., Ellingsen, T. E. & Bruheim, P.
Online content Intracellular metabolite pool changes in response to nutrient depletion
Any methods, additional references, Nature Research reporting induced metabolic switching in Streptomyces coelicolor. Metabolites 2,
summaries, source data, extended data, supplementary informa- 178–194 (2012).
tion, acknowledgements, peer review information; details of author
contributions and competing interests; and statements of data and 16. Jankevics, A. et al. Metabolomic analysis of a synthetic metabolic switch in
code availability are available at https://doi.org/10.1038/s41587- Streptomyces coelicolor A3(2). Proteomics 11, 4622–4631 (2011).
019-0335-4.
17. Huang, J., Lih, C. J., Pan, K. H. & Cohen, S. N. Global analysis of growth
Received: 25 April 2018; Accepted: 30 October 2019; phase responsive gene expression and regulation of antibiotic biosynthetic
Published: xx xx xxxx pathways in Streptomyces coelicolor using DNA microarrays. Genes Dev. 15,
3183–3192 (2001).

18. D’Huys, P. J. et al. Genome-scale metabolic flux analysis of Streptomyces
lividans growing on a complex medium. J. Biotechnol. 161, 1–13 (2012).

19. Kieser, T., Bibb, M. J., Buttner, M. J., Chater, K. F. & Hopwood, D. A. Practical
Streptomyces Genetics (The John Innes Foundation, 2000).

20. Olukoshi, E. R. & Packter, N. M. Importance of stored triacylglycerols in
Streptomyces: possible carbon source for antibiotics. Microbiology 140,
931–943 (1994).

21. Alvarez, H. M. & Steinbuchel, A. Triacylglycerols in prokaryotic
microorganisms. Appl. Microbiol. Biotechnol. 60, 367–376 (2002).

22. Li, S., Wang, W., Li, X., Fan, K. & Yang, K. Genome-wide identification and
characterization of reference genes with different transcript abundances for
Streptomyces coelicolor. Sci. Rep. 5, 15840 (2015).

23. Waldvogel, E. et al. The PII protein GlnK is a pleiotropic regulator for
morphological differentiation and secondary metabolism in Streptomyces
coelicolor. Appl. Microbiol. Biotechnol. 92, 1219–1236 (2011).

24. Zhuo, Y. et al. Reverse biological engineering of hrdB to enhance the
production of avermectins in an industrial strain of Streptomyces avermitilis.
Proc. Natl Acad. Sci. USA 107, 11250–11254 (2010).

25. Zhang, Y. et al. Characterization of a pathway-specific activator of
milbemycin biosynthesis and improved milbemycin production by its
overexpression in Streptomyces bingchenggensis. Microb. Cell Fact. 15,
152 (2016).

26. Yin, S. et al. Identification of a cluster-situated activator of oxytetracycline
biosynthesis and manipulation of its expression for improved oxytetracycline
production in Streptomyces rimosus. Microb. Cell Fact. 14, 46 (2015).

27. Vemuri, G. N., Eiteman, M. A., McEwen, J. E., Olsson, L. & Nielsen, J.
Increasing NADH oxidation reduces overflow metabolism in Saccharomyces
cerevisiae. Proc. Natl Acad. Sci. USA 104, 2402–2407 (2007).

28. Kumelj, T., Sulheim, S., Wentzel, A. & Almaas, E. Predicting strain
engineering strategies using iKS1317: a genome-scale metabolic model of
Streptomyces coelicolor. Biotechnol. J. 14, e1800180 (2019).

Nature Biotechnology | www.nature.com/naturebiotechnology

Articles Nature Biotechnology

29. Horbal, L., Fedorenko, V. & Luzhetskyy, A. Novel and tightly 31. Chae, T. U., Choi, S. Y., Kim, J. W., Ko, Y. S. & Lee, S. Y. Recent advances in
regulated resorcinol and cumate-inducible expression systems for systems metabolic engineering tools and strategies. Curr. Opin. Biotechnol. 47,
Streptomyces and other actinobacteria. Appl. Microbiol. Biotechnol. 98, 67–82 (2017).
8641–8655 (2014).
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
30. Wang, W. et al. Development of a synthetic oxytetracycline-inducible published maps and institutional affiliations.
expression system for streptomycetes using de novo characterized genetic © The Author(s), under exclusive licence to Springer Nature America, Inc. 2019
parts. ACS Synth. Biol. 5, 765–773 (2016).

Nature Biotechnology | www.nature.com/naturebiotechnology

Nature Biotechnology Articles

Methods were extracted using the trichloroacetic acid (TCAA)-ester method35 with some
modifications. Ice-cold TCAA (1.3 ml, 10% (m/v)) was added to 200 mg of
Reagents. Derivatizing reagents and chemical standards were purchased from powdered cells. Following homogenization and vigorous vortex of the extract at
Sigma-Aldrich. Stable isotopic labeled oleate and glucose were purchased from 0 °C for 3 min, the TCAA suspension was centrifuged at 12,000g for 10 min at 0 °C.
Cambridge Isotope Laboratories, Inc. The mixture standards of fatty acid methyl The supernatant was collected and extracted with successive partitioning with
esters (FAMEs) were purchased from ZZBIO Co. All organic reagents used for 2 ml prechilled (−20 °C) diethyl ether to remove TCAA. The aqueous extract was
GC–MS or liquid chromatography–tandem mass spectrometry (LC–MS/MS) were recovered and lyophilized, then dissolved in 300 μl ice-cold 25 mM ammonium
high-performance liquid chromatography (HPLC) grade solvent. formate (pH 4.6, 2% methanol) and filtered with a 0.2 μm cellulose acetate
membrane filter (Pall) before analysis by LC–MS/MS.
Strains and growth conditions. All plasmids were propagated in Escherichia
coli JM109 (Novagen) cultured in Luria-Bertani (LB) medium with apramycin
(50 μg ml−1) or kanamycin (50 μg ml−1) at 37 °C. E. coli ET12567/pUZ8002 was Analysis of intracellular triacylglycerols. TAGs were purified following the
method described by Rodriguez et al.36 with some modifications. Cells were
used for conjugation and grown in LB at 37 °C with appropriate antibiotics collected by centrifugation at −9 °C, 13,000g for 1 min. Cells were immediately
(chloramphenicol, 25 μg ml−1 and kanamycin, 25 μg ml−1). To prepare spores,
S. coelicolor, S. rimosus and S. bingchenggensis were cultivated on mannitol-soya submerged into liquid nitrogen (2 min) and then lyophilized with a vacuum
concentrator. Total lipids were extracted from lyophilized cells (10 mg) by
flour agar plates19, S. venezuelae were grown on maltose-yeast extract-malt extract32 chloroform/methanol (2:1, v/v) in a water bath at 40 °C for 3 h. The mixture
agar plates and S. avermitilis were cultivated on yeast extract-malt extract-soluble
starch19 agar plates at 28 °C. was vortexed for 1 min every 30 min during the extraction period. TAGs and
PLs were analyzed by TLC on silica gel 60 F254 plates (Merck) as previously
All fermentations of S. coelicolor and S. venezuelae strains were conducted described37. Lipid fractions were visualized by Cu-phosphoric acid staining and
in 250 ml shake flasks with a work volume of 100 ml, while S. rimosus and
S. avermitilis strains were conducted in 250 ml shake flasks with a work volume bands of TAGs and PLs were quantified using a Tanon 1600 Gel Imaging System.
To get the purified TAGs, the TAG bands were dissolved in chloroform and after
of 40 ml, all of them were shaken at 250 r.p.m. at 28 °C. Spores of S. coelicolor centrifugation and evaporation to sequentially remove silica gel particles and
strains were inoculated in SMM19. For S. venezuelae strains, seed cultures were
first prepared by inoculating spores into 50 ml maltose-yeast extract-malt extract chloroform. For preparing FAMEs of fatty acid moieties, TAGs samples were
dissolved with 1.5 ml of 0.5 M sodium methoxide in methanol at 100 °C for 10 min.
and incubating at 220 r.p.m. for 20 h. Then, seed cultures (4%) were inoculated Then, 20 μl of nonadecanoic acid (4 mg ml−1), as internal standard, and 2 ml of 14%
into galactose metabolism (GM) medium with 4% (v/v) ethanol and cultured for
48 h. For S. rimosus, the seed culture and fermentation are carried out as previous foofrB1F m3 iinn.mAeftthearncoolowlinergetsoerqouoemntitaelmlypaedrdateudrea,n1d mcol notfinhuexoaunselywinacsuabdadteedd, at 100 °C
report26. For S. avermitilis strains, seed cultures were prepared by inoculating of followed
by 4 ml of a saturated solution of sodium chloride. The mixture was vigorously
spores into seed medium II and cultivating for 48 h. Then, the 1803-fermenter shaken for 1 min and centrifuged for 3 min at 12,000g. The upper phase contained
was carried out as described previously24. To perform MFA, S. avermitilis were
cultivated in FH medium (corn starch 30 g l−1, corn steep liquor 10 g l−1, yeast FAMEs and before GC–MS or GC–quadrupole time-of-flight–MS (GC–QTOF–
MS) analysis, trace moisture was removed using anhydrous sodium sulfate and the
extract 1a0t  1g2 l−01 h, C. oCl2 0.03 g l−1 and ɑ-amylase 0.04 g l−1) with glucose feeding sample was filtered with 0.2 μm polytetrafluoroethylene filter (Supelco).
(8 g l−1)

Construction of plasmids and strains. All the primers used in this study are Analytical conditions of GC–MS, GC–QTOF–MS and LC–MS/MS. The GC–MS
listed in Supplementary Table 14. The cumate induced promoter fragment was platform (Agilent) consisted of a 7890 GC, a 5975 MS system and a 7683 auto-
amplified from the plasmid pGCymRP21 (ref. 29) with a pair of primers CuF/CuR. sampler equipped with a HP-5 MS capillary column (30 m × 250 μm × 0.25 μm,
The sco6196 was amplified from the genome of S. coelicolor M145 with a pair of J&W Scientific). For metabolome analysis, 1 μl sample was injected with a split
primers 6196F/6196R. The two purified fragments were assembled with the ratio of 1:10. The chromatogram conditions were: 70 °C for 1 min, then increased
XbaI/EcoRV digested linear fragments of pSET152 (ref. 19) using the Gibson 5 °C/min to 230 °C, followed by increasing 10 °C min−1 to 290 °C and held for 6 min.
assembly method to generate pCu-SCO6196. The plasmid pCu-SCO6196 was Ions were generated by a 70 eV electron beam at an ionization current of 40 μA.
integrated into the genome of S. coelicolor M145, S. venezuelae ISP5230, S. rimosus Masses were acquired in the range of 50–650 m/z. To analyze fatty acid moieties
M4018 (ref. 26) and S. avermitilis A56 (ref. 24), generating M145-DT, Sv-DT, of TAGs, 1 μl sample was injected with a split ratio of 1:30. The chromatogram
M-DT and A56-DT, respectively. To construct pCuR-SCO6196, a 3,541-base condition was: 70 °C for 2 min, increased to 290 °C at a rate of 5 °C min−1 and held
pair fragment was cut from pCu-SCO6196 by XbaI/EcoRV and ligated with the for 3 min. Helium was the carrier gas at a flow rate of 1.0 ml min−1. The temperature
plasmid backbone from pLC803 (ref. 33) (amplified via primer LCF and LCR). of both injector and detector were set at 290 °C. To analyze the fatty acid moieties
Then, pCuR-SCO6196 was integrated into the genome of S. rimosus M2R (ref. 26) of TAGs in 6196DM and 6196OE, the 7200 GC–QTOF equipped with a DB-5MS
to generate M2R-DT. The construction of sco6196 deletion strain (6196DM) was capillary column (30 m × 250 μm × 0.25 μm, J&W Scientific) was used. Running
performed via homologous recombination. The plasmid backbone of pKC1132 condition was identified with that used in GC–MS.
was amplified using primer pair 1132F/1132R. The two fragments corresponding
to the upstream and downstream regions of sco6196 were amplified using the Acetyl-CoA and malonyl-CoA were analyzed using a 1260/6460 Triple
primer pairs 96LF/96LR and 96RF/96RR. Then, the two fragments and pKC1132 Quadrupole LC–MS/MS (Agilent) with a C18 analytical column (Gemini
backbone were ligated to generate pKC1132-96DM. This plasmid was transformed 150 × 2.0 mm, particle size 3 μm, Phenomenex). Quantification was carried out
into S. coelicolor M145 to by conjugating E. coli. The double-crossover of sco6196 with the multiple reaction monitoring mode (MRM) in the MS/MS using a
deletion strain was verified by PCR (using primer pair 96vF/96vR) and subsequent positive mode to detect the transition of the parent ion to the daughter ion (m/z
sequencing. To construct sco6196 overexpression strain (6196OE), the plasmid parent > m/z daughter) for acetyl-CoA (810 > 303) and malonyl-CoA (854 > 347).
pIMEP33 was treated with EcoRV and EcoRI to generate the backbone fragment.
The sequence of sco6196 was amplified using primer pair 96F1/96R1. Then, the Data processing and statistical analyses of metabolome data. The collected
two fragments were ligated using the Gibson assembly method to generate pIMEP- GC–MS data were automatically de-convoluted by AMDIS software38. Raw MS
96OE. The resulting plasmid was integrated into the genome of S. coelicolor M145 spectra data (.netCDF) of all samples were subjected to baseline correction, using
to generate 6196OE. To construct SCO6196 recombinant expression plasmid pET- MetAlign 3.0 software with the recommended parameters to smooth and remove
SCO6196, sco6196 was amplified using primer pair 96F2/96R2, and then ligated noise39. The resulting spectral peaks with signal-to-noise values less than four were
with pET-30a trimmed by HindIII and NdeI. The plasmid was transformed to eliminated. Minor variation data were corrected according to the intensities of the
E. coli BL21(DE3) for heterologous expression of the recombinant SCO6196. internal standards. Mass peaks with the same retention time belong to one putative
compound, and the peak with the highest mass intensity was selected for further
Sampling, quenching and preparation of samples for metabolome analysis. analysis. After integrity check, missing value estimation, data filtering (interquartile
Cells of M145 and HY01 were collected from different growth stages for range), normalization, centering (log-transformation) and scaling (Pareto scaling),
metabolome analysis: early and late exponential phases (20 and 36 h), transitional the data matrices were subjected to comparative metabolomic analysis by PLS-
phase (48 h), and early, mid and late stationary phases (60, 72 and 96 h). Five DA using the software MetaboAnalyst 3.0 (ref. 40). Metabolites contributing to
independent replicates were performed for each sample. Act production were identified via their VIP score. A VIP score higher than 1.5
indicates the significant contribution of the metabolites to Act production. Pearson
Details of sampling, quenching and metabolite extraction were performed correlation analysis was used to search for metabolites with close relationships
in a rapid and reproducible manner following the protocols illustrated in to the targets. Further information on research design is available in the Nature
Supplementary Note 1. The extractions were lyophilized and subject to two-stage Research Reporting Summary linked to this article.
chemical derivatization34 before analysis by GC–MS. All tubes and tips were
prechilled (−20 °C) before the start of experiments and kept below 0 °C during Isotope labeling experiments. Feeding experiments were performed with stable
experiments. uniformly 13C-labeled glucose (U-13C6, 99%) and 13C-labeled oleate, sodium salt
(U-13C18, 98%) as tracers, respectively. The [U-13C] glucose or [U-13C] oleate
Preparation of samples of acetyl-CoA and malonyl-CoA. Both unlabeled and were added at 72 h in SMM cultures of Streptomyces strains, and samples were
labeled acetyl-CoA and malonyl-CoA were extracted by the same method. After harvested after 24 and 48 h for the oleate and glucose feeding experiments,
the cells were collected, quenched and ground, acetyl-CoA and malonyl-CoA respectively. The intracellular metabolites were extracted and derivatized with

Nature Biotechnology | www.nature.com/naturebiotechnology

Articles Nature Biotechnology

3-nitrophenylhdyrazones. To prepare Act samples, pH of culture was adjusted to MFA. The fermentation stage from 72 to 120 h, where Act is extensively produced
3.0 with HCl, then extracted with ethyl acetate with the same volume, and finally at a constant specific production rate, was used to analyze the metabolic flux
Act were resolved in 1 ml of methanol after evaporation. For TCA intermediates distribution. We assumed cells under a steady-state condition during this period,
tracing, a ThermoFisher DGLC-3000 coupled to Sciex QTRAP 6500 Plus and the system dynamics could be described as Sv = 0, where S is a sparse matrix
system was used for data collection. Individual metabolites were separated on and v represents the flux through all reaction in the metabolic network. The
a Phenomenex Kinetex C18 column (100 mm × 2.1 mm, 2.6 µm) using 0.1% elemental composition (C, H, O, N) and precursors for biomass biosynthesis of
formic acid in water as mobile phase A and 0.1% formic acid in acetonitrile as M145 have been described previously47. The biomass composition of HY01 was
mobile phase B. Act congeners were analyzed on Thermo DGLC-3000 coupled to hypothesized as the same as that of M145. For each experiment, we measured the
Sciex QTRAP 6500 Plus in the enhanced MS mode. Individual polyketides were following parameters: dry cell weight, specific growth rate, specific Act production
separated on an Agilent Zorbax SB-Aq column (3.5 µm, 2.1 mm × 100 mm) rate, specific glucose uptake rate, specific TAG-degradation rate and specific CO2
using a gradient comprising 0.1% water as mobile phase A and 0.1% acetonitrile production rate. In addition to these parameters, metabolic flux out of different
as mobile phase B. The gradient started with 10% mobile phase A, which was kept precursors for biomass synthesis were calculated according to the stoichiometry of
constant for 2 min, and was then linearly increased to 100% mobile phase B over these precursors in the biomass biosynthesis reaction. On basis of these data, fluxes
6 min. The gradient was maintained for 4 min at 100% B before it was reduced of the listed reactions could be calculated by MFA. MFA for S. avermitilis were
again to 10% A over 1 min. The gradient was equilibrated for another 3 min before implemented in the same way by using the data collected from 120 to 192 h.
the injection of the next sample. Flux sampling analysis. The flux sampling analysis based on the GEM of
S. coelicolor (iKS1317)25 was conducted to compare the flux changes between
Natural abundance of 13C of raw data were corrected as described by S. coelicolor M145 and HY01. In the flux sampling analysis using cobrapy48, the
Yuan et al.41. To calculate the amount of labeled Act, the detectable Act with even GEM was constrained by the experimentally measured rate (Supplementary
number of labeled carbon atoms was used. For the other metabolites, the labeled Table 8) with the minimization of glucose uptake rate as the objective function
amount was calculated by adding all detectable labeled molecules. (we assume that the cell could use the carbon source in the most economical
Pairwise sequence identities assay of ACS. The ACS homologs were identified way at stationary phase49). During the simulation, the sampling algorithm
by BLASTP42 with an e-value threshold of 0.001 from 125 annotated Streptomyces optGpSampler50 was adopted. For each condition, 1,000 samples were used for
genomes published in NCBI (up to 29 September 2018), and the long-chain-fatty- valid sampling distributions. Python scripts for flux sampling analysis are shown in
acid-CoA ligases from E. coli K12 and Bacillus subtilis 168 (FadD, LcfA and LcfB) Supplementary Note 4.
were used as bait proteins. The resulting 888 ACSs were pairwise compared using Other analytical methods. Cell growth was measured by the diphenylamine
Needleman–Wunsch global alignment algorithm43, and clustered with a threshold colorimetric method51. Residual glucose concentration was determined
of sequence identity greater than 50%. enzymatically using a bioanalyzer (SBA-40C). Intracellular level of ATP, NADH,
NADPH, NADH/NAD+, NADPH/NADP+ and ATP/ADP were analyzed by
Influence of NADH and ATP on enzymatic activity of CS, IDH and KGDH. kits (BioVision) following their instructions. Production of Act, jadomycin B,
Relative enzymatic activity of CS, IDH and KGDH were evaluated by measuring oxytetracycline and avermectin B1a were assayed following previous protocols21,27,52.
the levels of generated products citrate, α-oxoglutarate, and succinyl-CoA for CS, Statistical analyses. All error bars shown in the present work are s.d. values
IDH and KGDH in crude enzyme extract, respectively. Cells of M145 cultivated of replicates. The number of replicates is provided in the corresponding figure
in liquid SMM were collected at 36 h. After ultrasonication, supernatant was caption. Where relevant, two-tailed Student’s t-tests or one-way analysis of variance
preserved for enzymatic activity assays. Generation of citrate and α-oxoglutarate (ANOVA) are used to demonstrate differences in values.
were carried out in a 2 ml reaction system as described by Xia et al.44. Generation Reporting Summary. Further information on research design is available in the
of succinyl-CoA was implemented in a 2 ml reaction system as described by Nature Research Reporting Summary linked to this article.
Weitzman45. When necessary, 0.1 mM NADH or/and ATP was added. After
incubating the mixture at 30 °C for 15 min, citric acid was measured by HPLC Data availability
(Agilent 1200) with an Agilent Zorbax SB C8 column (250 mm × 4.6 mm, 5 μm)
coupled to a photodiode array detector at 210 nm, the column was eluted The genome sequence of S. coelicolor strain is available at NCBI (accession number
wwaitsha2n0a mlyzMedobf yHH2SPOL4Catw3i0th °Ca with a flow rate of 0.6 ml min−1; α-oxoglutarate NC_003888). Time-series transcriptome data of S. coelicolor stains are available
Bio-Rad Aminex HPX-87 column (300 mm at NCBI GEO database (accession numbers GSE2983, GSE18489, GSE30569,
× 7.8 mm, 9 μm) coupled to a photodiode array detector at 215 nm, the column GSE30570 and GSE53562). Other data supporting the findings of this study are
qwuaasnetliufitceadtiwonithof5s mucMcinoyf lH-C2SoOA4 at 30 °C with a flow rate of 0.6 ml min−1 and included in the published article and supplementary information. Requests for any
was carried out with the MRM mode in the additional information can be made to the corresponding authors.
MS/MS using a positive mode to detect the transition of the parent ion to the
daughter ion (868 > 361) using the same condition as that for acetyl-CoA Code availability
detection mentioned above.
The custom Python scripts used for flux sampling analysis are provided in
RNA isolation and quantitative PCR with reverse transcription (RT–qPCR) Supplementary Note 4.
assay. RNA from S. coelicolor were isolated using RNeasy Midi Kit (Qiagen).
Isolated RNA was treated with DNase (Qiagen) before being reverse transcribed References
with random hexamers using SuperScript III (Invitrogen). RT–qPCR was
performed on a Roche LightCycler 480 using SYBR FAST qPCR master mix 32. Bai, C. et al. Exploiting a precise design of universal synthetic modular
(KAPA). The gene sco0710 of S. coelicolor was used as constitutive reference to regulatory elements to unlock the microbial natural products in Streptomyces.
normalize gene expression19. Technical triplicates of three biological repeats were Proc. Natl Acad. Sci. USA 112, 12181–12186 (2015).
performed per condition.
LC–MS/MS analysis of SCO6196 specificity in vitro. The specificity of the 33. Wang, W. et al. Angucyclines as signals modulate the behaviors of
ACS SCO6196 was determined based on each acyl-CoA formation from iC14:0, Streptomyces coelicolor. Proc. Natl Acad. Sci. USA 111, 5688–5693 (2014).
iC15:0, aiC15:0, iC16:0, aiC17:0 and C18:1-Δ9. The reaction mixture consisted
of each fatty acid (2 mmol), ATP (10 mmol), CoA (1.5 mmol), MgCl2 (4 mmol), 34. Winder, C. L. et al. Global metabolic profiling of Escherichia coli cultures: an
KCl (5 mmol) and SCO6196 (0.7 mg protein) in a final volume of 1 ml of 0.2 M evaluation of methods for quenching and extraction of intracellular
Tris–HCl buffer (pH 8.5) and incubated at 37 °C for 5 min (ref. 46). The reaction metabolites. Anal. Chem. 80, 2939–2948 (2008).
was terminated by the addition of 1 ml of cold acetonitrile. The solution was
evaporated in a centrifugal evaporator and the residue was dissolved with 200 μl of 35. Perera, M. A., Choi, S. Y., Wurtele, E. S. & Nikolau, B. J. Quantitative analysis
absolute methanol. After brief centrifugation, the supernatant was filtered and then of short-chain acyl-coenzymeAs in plant tissues by LC–MS-MS electrospray
subjected to LC–electrospray ionization–MS/MS analysis. ionization method. J. Chromatogr. B. 877, 482–488 (2009).

CoA thioesters analysis was performed on an Agilent 1260/6460 HPLC-triple 36. Rodriguez, E., Navone, L., Casati, P. & Gramajo, H. Impact of malic enzymes
quadrupole mass spectrometer (Agilent) with a ZORBAX SB-Aq C18 column on antibiotic and triacylglycerol production in Streptomyces coelicolor. Appl.
(3.5 μm, 2.1 mm × 100 mm). The mobile phases were A, 10 mM ammonium acetate Environ. Microbiol. 78, 4571–4579 (2012).
(pH 6.8) and B, acetonitrile. The eluting gradient was as follows: the column was
equilibrated with 5% B, 5% B for 3 min, 5% B to 20% B in 15 min, 20% B to 95% B 37. Arabolaza, A., Rodriguez, E., Altabe, S., Alvarez, H. & Gramajo, H. Multiple
in 5 min, 95% B for 6 min, 95% B to 5% A in 2 min and 5% B for 13 min. The flow pathways for triacylglycerol biosynthesis in Streptomyces coelicolor. Appl.
rate was 0.25 ml min−1. The mass spectrometer was operated in the negative-ion Environ. Microbiol. 74, 2573–2582 (2008).
mode with the source temperature set to 350 °C. Each acyl-CoA was monitored
by the daughter ions, which were selected based on m/z 425.9 and 408 in common 38. Stein, S. E. An integrated method for spectrum extraction and compound
and specific ions of [M-H-80]− and [M-H-347]−. identification from gas chromatography/mass spectrometry data. J. Am. Soc.
Mass Spectrom. 10, 770–781 (1999).

39. Lommen, A. & Kools, H. MetAlign 3.0: performance enhancement by efficient
use of advances in computer hardware. Metabolomics 8, 719–726 (2012).

Nature Biotechnology | www.nature.com/naturebiotechnology

Nature Biotechnology Articles

40. Xia, J., Sinelnikov, I. V., Han, B. & Wishart, D. S. MetaboAnalyst 3.0–making Acknowledgements
metabolomics more meaningful. Nucleic Acids Res. 43, W251–W257 (2015).
We dedicate this paper to the memory of our beloved mentor, friend and colleague,
41. Yuan, J., Bennett, B. D. & Rabinowitz, J. D. Kinetic flux profiling for K. Yang. We thank H. Tan, Y. Tao, Y. Chen, M. Li, S. Chen, S. Gao and Q. Gao for
quantitation of cellular metabolic fluxes. Nat. Protoc. 3, 1328–1340 (2008). helpful discussions. We acknowledge the financial support from the National Natural
Science Foundation of China (grants 31430002 and 31720103901 to L.Z.; 31922002
42. Camacho, C. et al. BLAST+: architecture and applications. BMC Bioinforma. and 31570031 to W.W.; 31772242 to S.L. and 31672092 to W.X.). This work was also
10, 421 (2009). supported by the Open Project Funding of the State Key Laboratory of Bioreactor
Engineering, the 111 Project (B18022), the Major Basic Program of the Natural
43. Rice, P., Longden, I. & Bleasby, A. EMBOSS: the European molecular biology Science Foundation of Shandong Province (ZR2017ZB0206), the Shanghai Science and
open software suite. Trends Genet. 16, 276–277 (2000). Technology Commission (18JC1411900) and the Shandong Taishan Scholar Program
of China to L.Z. and the International Partnership Program of Chinese Academy of
44. Xia, J., Xu, Z., Xu, H., Feng, X. & Bo, F. The regulatory effect of citric acid on Sciences (153211KYSB20170014) and the Young Scientists Innovation Promotion
the co-production of poly(epsilon-lysine) and poly(L-diaminopropionic acid) Association of CAS (2016087) to W.W.
in Streptomyces albulus PD-1. Bioprocess Biosyst. Eng. 37, 2095–2103 (2014).
Author contributions
45. Weitzman, P. D. J. Regulation of α-ketoglutarate dehydrogenase activity in
Acinetobacter. FEBS Lett. 22, 323–326 (1972). L.Z., W.W. and S.L. conceived and supervised the project. W.W., S.L., Z.L. and W.X.
designed and performed the main experiments. J.Z. performed the large-scale
46. Kasuya, F. et al. Determination of acyl-CoA esters and acyl-CoA synthetase fermentations. G.A. performed the GC–MS analyses. G.S. and S.M.L. collected LC–MS/MS
activity in mouse brain areas by liquid chromatography-electrospray data for isotope tracing experiments. K.F., G.T., Z.Y., H.L., P. J., Y.L., X.C., X.X., X.L.,
ionization-tandem mass spectrometry. J. Chromatogr. B. 929, 45–50 (2013). H.K.D., C.Y., Y.Y., G.A. and S.Z. participated in the experiments. W.W., S.L. and L.Z.
wrote the manuscript.
47. Coze, F., Gilard, F., Tcherkez, G., Virolle, M. J. & Guyonvarch, A. Carbon-flux
distribution within Streptomyces coelicolor metabolism: a comparison between Competing interests
the actinorhodin-producing strain M145 and its non-producing derivative
M1146. PLoS ONE 8, e84151 (2013). The authors have filed a provisional patent for this work to the China National Intellectual
Property Administration (CNIPA). L.Z., W.W., S.L., Z.L., G.T., K,F. and J.Z. are inventors
48. Ebrahim, A., Lerman, J. A., Palsson, B. O. & Hyduke, D. R. COBRApy: on the provisional patent application (CN201910411123.7, filed 17 May 2019).
constraints-based reconstruction and analysis for Python. BMC Syst. Biol. 7,
74 (2013). Additional information

49. Schuetz, R., Kuepfer, L. & Sauer, U. Systematic evaluation of objective Supplementary information is available for this paper at https://doi.org/10.1038/
functions for predicting intracellular fluxes in Escherichia coli. Mol. Syst. Biol. s41587-019-0335-4.
3, 119 (2007). Correspondence and requests for materials should be addressed to W.W., W.X. or L.Z.
Reprints and permissions information is available at www.nature.com/reprints.
50. Megchelenbrink, W., Huynen, M. & Marchiori, E. optGpSampler: an
improved tool for uniformly sampling the solution-space of genome-scale
metabolic networks. PLoS ONE 9, e86587 (2014).

51. Zhao, Y., Xiang, S., Dai, X. & Yang, K. A simplified diphenylamine
colorimetric method for growth quantification. Appl. Microbiol. Biotechnol.
97, 5069–5077 (2013).

52. Wang, W. et al. An engineered strong promoter for streptomycetes. Appl.
Environ. Microbiol. 79, 4484–4492 (2013).

Nature Biotechnology | www.nature.com/naturebiotechnology



点击阅读翻页书版本