Polyketide Starter and Extender Units Serve as Regulatory Ligands to Coordinate the Biosynthesis of Antibiotics in Actinomycetes

ABSTRACT Polyketides are one of the largest categories of secondary metabolites, and their biosynthesis is initiated by polyketide synthases (PKSs) using coenzyme A esters of short fatty acids (acyl-CoAs) as starter and extender units. In this study, we discover a universal regulatory mechanism in which the starter and extender units, beyond direct precursors of polyketides, function as ligands to coordinate the biosynthesis of antibiotics in actinomycetes. A novel acyl-CoA responsive TetR-like regulator (AcrT) is identified in an erythromycin-producing strain of Saccharopolyspora erythraea. AcrT shows the highest binding affinity to the promoter of the PKS-encoding gene eryAI in the DNA affinity capture assay (DACA) and directly represses the biosynthesis of erythromycin. Propionyl-CoA (P-CoA) and methylmalonyl-CoA (MM-CoA) as the starter and extender units for erythromycin biosynthesis can serve as the ligands to release AcrT from PeryAI, resulting in an improved erythromycin yield. Intriguingly, anabolic pathways of the two acyl-CoAs are also suppressed by AcrT through inhibition of the transcription of acetyl-CoA (A-CoA) and P-CoA carboxylase genes and stimulation of the transcription of citrate synthase genes, which is beneficial to bacterial growth. As P-CoA and MM-CoA accumulate, they act as ligands in turn to release AcrT from those targets, resulting in a redistribution of more A-CoA to P-CoA and MM-CoA against citrate. Furthermore, based on analyses of AcrT homologs in Streptomyces avermitilis and Streptomyces coelicolor, it is believed that polyketide starter and extender units have a prevalent, crucial role as ligands in modulating antibiotic biosynthesis in actinomycetes.

corresponding starter and extender units for avermectin or actinorhodin biosynthesis also serve as the ligands of AcrT homologs.

RESULTS
A novel regulator is discovered with high-affinity binding to the eryAI promoter. The DNA affinity capture assay (DACA) is an in vitro method to directly capture DNAbinding TFs (40), which has been efficiently utilized to discover underlying regulatory pathways for the biosynthesis of antibiotics in actinomycetes (41,42). Considering that the three eryA genes (eryAI, eryAII, and eryAIII) encoding PKSs for erythromycin biosynthesis are cotranscribed under the control of P eryAI (43), we used biotinylated P eryAI to isolate regulators interacting with the probe from total proteins of Sac. erythraea strain A226 (hereafter named A226). As determined by mass spectrometry (MS) analysis, 48 TFs mapped to .1 peptide fragments were identified as potential P eryAI -interactive regulators, in which AcrT (SACE_3980) possessed the highest number of detectable peptide fragments (Fig. 1A). Based on the genome annotation of Sac. erythraea (28), we found that AcrT is a TetR family transcriptional regulator and its homologs are widespread in polyketide-producing actinomycetes (see Fig. S1 in the supplemental material), suggesting that this type of TF has physiologically conserved regulatory roles.
Electrophoretic mobility shift assays (EMSAs) and the enhanced green fluorescent protein (EGFP) reporter system in Escherichia coli were used to investigate the regulatory pattern of AcrT acting on P eryAI (Fig. 1B to D). Results showed that AcrT specifically bound to P eryAI in vitro ( Fig. 1B; Fig. S2A), and bioluminescence was diminished with the expression of AcrT in vivo (Fig. 1D). Furthermore, a DNase I footprinting assay showed that a 20-nucleotide sequence, CTTGATCCTTCCTATATTGT (termed site A), was protected by AcrT (Fig. 1E). EMSAs with probes covering different sequences indicated that site A was indispensable for the binding of AcrT ( Fig. S2B and C). Biolayer interferometry (BLI) assays confirmed that the interaction between the 50-bp fragment containing site A and AcrT had an affinity comparable to an equilibrium dissociation constant (K D ) of 16.4 nM (Fig. 1F). Moreover, by monitoring acrT transcription and Er-A production in A226 during the fermentation successively with reverse transcriptionquantitative PCR (RT-qPCR) and high-performance liquid chromatography (HPLC) analyses, we found that Er-A production was low and transcription of acrT appeared in the early stage, and as the yield of Er-A increased, acrT transcription peaked at 48 h and subsequently decreased to a very low level (Fig. 1G). Taken together, these results indicate that AcrT is a direct repressor of eryAI in the early stage of Sac. erythraea fermentation.
AcrT directly represses the transcription of ery cluster genes and its own gene. To investigate the function of AcrT, acrT was first disrupted with thiostrepton resistance gene (tsr) replacement in A226, which was confirmed by PCR ( Fig. S3A and B). The difference in Er-A yield between A226 and A226DacrT appeared after 2 days of fermentation in R5 liquid medium, and A226DacrT showed a 28.3% increase in Er-A production compared to A226 on the 6th day ( Fig. 2A and B). However, A226DacrT showed sporulation and growth rates similar to those of A226 ( Fig. S3C and D). Complementation of acrT in A226DacrT pushed back Er-A production to its original level (Fig. 2B). Furthermore, when pIBacrT (see Table S1 in the supplemental material) was transformed into A226, the Er-A yield of A226/pIBacrT decreased by 20% compared to that of A226/pIB139 (Fig. 2B). Using the same method, we inactivated acrT in the industrial strain Sac. erythraea WB, and the Er-A yield of WBDacrT increased by 15.8% over that of WB ( Fig. S3E and F). These results further suggest that AcrT functions as a repressor to control erythromycin biosynthesis in Sac. erythraea.
Since AcrT had a high affinity for P eryAI , we wondered whether it interacted with other promoters within the ery cluster (Fig. 2C). EMSAs showed that AcrT also specifically binds to P eryK , P eryBVI , eryBI-BIII-int, and ermE-eryCI-int (Fig. 2D). Furthermore, we performed RT-qPCR experiments on eight genes within the ery cluster. The transcriptional levels of eryAI, ermE, eryBI, eryBIII, eryBVI, eryCI, eryBIV, and eryK in A226DacrT grown for 24 or 48 h exhibited overall increases compared with those in A226 ( Fig. 2E and F). We FIG 1 AcrT is a novel regulator interacting with P eryAI . (A) Screening of the potential P eryAI -interactive regulators by DACA. The PepCount and UniquePepCount values represent the total number of peptide fragments and the number of unique peptide fragments detected, also found that the acrT transcript markedly increased in A226DacrT cultured for both 24 and 48 h, respectively, in comparison to that in A226 ( Fig. 2E and F). Moreover, EMSAs showed that AcrT specifically interacts with its own promoter (Fig. 2D). Therefore, our results verify that AcrT directly represses all genes in the ery cluster and itself.
P-CoA and MM-CoA are ligands of AcrT for the regulation of eryAI. Our recent investigations showed that ligands play pivotal roles in mediating the regulation of TFs for antibiotic biosynthesis in actinomycetes (37,44,45). Others reported that antibiotics or their biosynthetic intermediates as ligands can control their own biosynthesis by modulating the DNA-binding activity of TFs (8)(9)(10)(11)(12)(13)(14). Therefore, we tested whether Er-A and its biosynthetic intermediates could influence AcrT interacting with P eryAI and found that Er-A, Er-B, Er-C, and Er-D had no effect on AcrT binding to P eryAI (data not shown).
Since AcrT transcriptionally repressed eryA genes encoding PKSs to catalyze the condensation of P-CoA and MM-CoA, we explored effects of the two substrates on the binding ability of AcrT to P eryAI . It was demonstrated that P-CoA and MM-CoA could cause the dissociation of AcrT from P eryAI , and 10 mM P-CoA or 15 mM MM-CoA was sufficient for this, whereas A-CoA had no effect on the DNA-binding activity of AcrT ( Fig. 3A and B). BLI assays also revealed that when 400 mM or 800 mM P-CoA was added, the affinity between AcrT and the 50-bp fragment containing site A was reduced from a K D of 16.4 nM without ligands to that of 81.1 nM or 194.0 nM, respectively ( Fig. 1F; Fig. S4A and B). When MM-CoA was added at the same concentrations, the affinity dropped to a K D of 50.8 or 154.8 nM ( Fig. S4C and D). Furthermore, the interactions between the two acyl-CoAs and AcrT were analyzed using circular-dichroism (CD) spectroscopy. Results showed that the a-helix content of AcrT markedly decreased after addition of P-CoA or MM-CoA, indicating that the two acyl-CoAs could interact with AcrT, while A-CoA did not obviously affect the signal intensity of AcrT ( Fig. 3C to E). Subsequently, the EGFP reporter system in E. coli was used to investigate whether P-CoA and MM-CoA relieved the repression of AcrT on eryAI (Fig. 1C). When 0.5 to 5 mM P-CoA or MM-CoA was added to the system, bioluminescence was stimulated in a dose-dependent manner ( Fig. 3F and G). The addition of A-CoA at the same concentrations made no differences in bioluminescence (Fig. 3H). Taken together, these findings corroborate that erythromycin biosynthetic starter and extender units, P-CoA and MM-CoA, play a novel role as effectors.
P-CoA and MM-CoA are increased by redistribution of A-CoA under AcrT inactivation. Since P-CoA and MM-CoA were identified as the ligands of AcrT, we wondered if their metabolism was in turn controlled by AcrT. To this end, we compared the intracellular amounts of several acyl-CoAs between A226 and A226DacrT. As shown in Fig. 4A, the levels of M-CoA, P-CoA, and MM-CoA in A226DacrT were higher than those in A226, whereas A-CoA and S-CoA levels were not affected by acrT deletion. Hence, AcrT might control the biosynthesis of M-CoA, P-CoA, and MM-CoA via the ACC and PCC pathways. To verify this hypothesis, an untargeted multiple MS analysis was applied to profile intracellular metabolites within A226 and A226DacrT (Fig. 4B). In total, 348 metabolites from the major metabolic pathways were identified (Table S2A), among which 39 with variable influence on projection (VIP) values greater than 1.0 and P values less than 0.05 were considered to be significantly different (highlighted with blue in Table S2A), and some were involved in carbohydrate, lipid, and amino acid metabolism (Fig. 4C). Although the majority of detectable metabolites in the glycolytic pathway, the pentose phosphate pathway, and the tricarboxylic acid (TCA) cycle respectively. Columns that are blue only indicate that the PepCount and UniquePepCount values were the same. Among those regulators, BldD and PhoP have been reported to bind to P eryAI in Sac. erythraea (33,39). (B) EMSA of AcrT with P eryAI . Competing assays were performed using a 50-fold excess of unlabeled P eryAI or a 50-fold excess of nonspecific probe poly(dI-dC). (C) Illustration of the EGFP reporter system in E. coli DH5a. The system contained two plasmids, pKC-EE, expressing egfp under P eryAI without acrT, and pKC-acrT-EE, expressing egfp under P eryAI with acrT driven by the promoter of apramycin-resistance gene aac (  EMSAs with AcrT binding to the probes of P eryK , P eryBVI , eryBI-BIII-int, ermE-eryCI-int, and acrT-3981-int. Competing assays were performed using a 50-fold excess of unlabeled probes or a 50-fold excess of nonspecific probe poly(dI-dC). EMSA with AcrT binding to the probe of eryAI-BIVint is shown in Fig. 1B with P eryAI instead. The two shifted bands (complex 1 and 2) were shown in EMSA with AcrT binding to acrT-3981-int, implying the existence of two AcrT-binding sites in acrT-3981-int. (E) RT-qPCR analyses of ery cluster genes and acrT in A226 and A226DacrT cultured for 24 h. (F) RT-qPCR analyses of ery cluster genes and acrT in A226 and A226DacrT cultured for 48 h. For these experiments, the mean values of 3 measurements are shown with SDs. *, P , 0.05; **, P , 0.01; ***, P , 0.001; ns, not significant. Regulatory Role of Polyketide Starter/Extender Units did not significantly change, the amount of citrate evidently decreased after the deletion of acrT ( Fig. 4C to E). This implied that the enhancement of intracellular M-CoA, P-CoA, and MM-CoA might be derived from A-CoA, which previously flowed to citrate. Simultaneously, we noticed that the intracellular level of citrate was much higher than those of isocitrate, succinate, and (S)-malate in both strains (Fig. 4D), suggesting that citrate was sufficient to maintain the normal metabolism of the TCA cycle, even if some A-CoA turned to M-CoA, P-CoA, and MM-CoA instead of citrate.
P-CoA and MM-CoA can coordinate their own supplies for erythromycin biosynthesis. To explore the regulatory pattern of AcrT with respect to these ACC, PCC, and CS genes ( Fig. 5A; Fig. S5C and D), EMSAs were carried out. Results showed that AcrT specifically bound to P 0018-0028 , P 3400 , P 7038-7039 , and P 0632-0633 ( Fig. S5E to H) but not to P 3398-3399 and P 4237 , as well as the negative controls P 3241-3242 and P 0649 (Fig. S5I to L). Using the motif-finding program MEME (http://meme-suite.org/) with the upstream sequences of those genes, a conserved AcrT-binding motif (nttGaTc, n: C, G, A; t: T, G/C; a: A, G; c: C, G) similar to site A was identified (Fig. 5B).
Since P-CoA and MM-CoA could dissociate AcrT from P eryAI , we wondered whether the two acyl-CoAs also affected the interaction between AcrT and P 0018-0028 , P 3400 , P 7038-7039 , or P 0632-0633 . EMSA results showed that P-CoA could efficiently pull AcrT down from the four probes, and MM-CoA had a bit weaker effect (Fig. 5C to F), while A-CoA had no any effect; however, P-CoA and MM-CoA did not influence AcrT binding to its own promoter (data not shown). Based on these findings, it is proved that P-CoA and MM-CoA, as ligands, can synthetically coordinate erythromycin biosynthesis by multiple approaches.
Polyketide starter and extender units prevalently act as regulatory ligands. Considering that the homologs of AcrT are widespread in the polyketide-producing actinomycetes (Fig. S1), we wanted to know whether the regulatory mechanism is universal.
Avermectin is a typical type I polyketide that is constructed using the starter unit methylbutyryl-CoA (MB-CoA) or isobutyryl-CoA (IB-CoA) and extender units M-CoA and MM-CoA in S. avermitilis (46). A TetR family regulator, SAV4017, here named AcrT Sa , shared 61% amino acid identity with AcrT (Fig. S1). Genetic experiments with disruption and complementation of acrT Sa demonstrated that AcrT Sa negatively affected the production of avermectin B1a (Fig. 6A). Using EMSA, RT-qPCR, and EGFP reporter system experiments, we found that AcrT Sa directly suppressed the transcription of aveA1, which encodes a PKS for avermectin biosynthesis (Fig. S6A to E), and that the four acyl-CoA precursors could mediate the dissociation of AcrT Sa from the promoter of aveA1 (P aveA1 ), whereas A-CoA had no effect on the DNA-binding activity of AcrT Sa (Fig. 6B). (F) EGFP reporter system to assay the interaction between P-CoA and AcrT. P-CoA was added to DH5a/pKC-acrT-EE. DH5a/pKC-EE was used as a control. (G) EGFP reporter system to assay the interaction between MM-CoA and AcrT. MM-CoA was added to DH5a/ pKC-acrT-EE. DH5a/pKC-EE was used as a control. (H) EGFP reporter system to assay the interaction between A-CoA and AcrT. A-CoA was added to DH5a/pKC-acrT-EE. DH5a/pKC-EE was used as a control. Mean values of 3 measurements are shown with SDs. *, P , 0.05; **, P , 0.01; ns, not significant.
With the individual addition of those four acyl-CoAs (0.5 to 5 mM), bioluminescence was stimulated for all in a dose-dependent manner, whereas the addition of A-CoA did not affect the intensity of bioluminescence ( Fig. 6C; Fig. S6F to J). Therefore, these results indicate that MB-CoA, IB-CoA, M-CoA, and MM-CoA serve as ligands to coordinate avermectin biosynthesis.
Moreover, we explored the ligand-mediated regulatory mode in the biosynthesis of actinorhodin, a type II polyketide in S. coelicolor. The TetR family regulator SCO4194, here named AcrT Sc , likewise had high amino acid identity (58%) with AcrT (Fig. S1). Correspondingly, AcrT Sc was found to specifically bind to the promoter of actI-ORF1 (P actI-ORF1 ), which encodes a PKS for actinorhodin biosynthesis (Fig. S7A and B). Distinct only from AcrT and AcrT Sa , AcrT Sc acted as an activator to stimulate the transcription of the actI-ORF1 gene (Fig. S7C to E) and displayed a positive correlation with actinorhodin production (Fig. 6D). As A-CoA and M-CoA are the starter and extender units in actinorhodin biosynthesis (47), we further explored whether they also influenced the binding activity of AcrT Sc to P actI-ORF1 . Results from EMSAs showed that A-CoA or M-CoA could promote AcrT Sc to interact with P actI-ORF1 (Fig. 6E). Bioluminescence was stimulated in a dose-dependent manner after the addition of 0.5 to 5 mM A-CoA or M-CoA to the EGFP reporter system ( Fig. 6F; Fig. S7F and G). These results indicate that A-CoA and M-CoA also act as ligands to modulate the biosynthesis of actinorhodin.
Therefore, our findings reveal that the starter and extender units function as ligands to allosterically modulate the DNA-binding activities of AcrT-like TFs, regardless of activators or repressors, ultimately promoting the biosynthesis of polyketides in actinomycetes.

DISCUSSION
Acyl-CoAs are involved in more than 100 cellular reactions in various biological processes of microorganisms, including glycolysis, TCA cycle, metabolism of amino acids and fatty acids, and biosynthesis of secondary metabolites (48). It is well documented that certain acyl-CoAs, as original building blocks, can be condensed to generate diverse polyketides in actinomycetes (21). A recent report has shown that some acyl-CoAs are also functional as major donors in the acylation of biosynthetic enzymes to modulate the synthesis of natural products (26). Our present work unprecedentedly found that the starter unit P-CoA and extender unit MM-CoA act as the ligands of AcrT to construct the P-/MM-CoA-AcrT-PKS circuit coordinating the synthesis of erythromycin in Sac. erythraea. AcrT could control the supply of the two acyl-CoAs via distribution of A-CoA metabolic flux, which in turn was modulated by the two acyl-CoAs as ligands. Based on these data, as well as those from S. avermitilis and S. coelicolor, we conclude that polyketide starter and extender units universally play an alternative role as ligands to coordinate antibiotic biosynthesis in actinomycetes.  In recent years, several types of TFs have been investigated in Sac. erythraea; however, elucidation of the regulatory network governing erythromycin biosynthesis remains limited. DACA is a very effective strategy to capture and identify potential TFs (41,42). Using this method, we identified at least 48 potential P eryAI -interactive TFs, among which BldD and PhoP have been reported to bind to P eryAI in Sac. erythraea (33, 39) (Fig. 1). Unfortunately, SACE_7301 and SACE_3446 (35,36), other previously published TFs that directly interact with P eryAI , were not detected among these TFs. Furthermore, we compared the affinities of AcrT and these four TFs toward P eryAI and found that AcrT exhibited an affinity with a K D of 11 nM under our experimental conditions, which was approximately 15-, 20-, 22-, and 33-fold higher than values for SACE_3446 (165 nM), PhoP (219 nM), SACE_7301 (245 nM), and BldD (364 nM), respectively. Therefore, it seems necessary to further optimize the conditions of this strategy for a promoter to capture its every potential TF.
Our recent investigations have shown that there exist complex regulatory mechanisms in the biosynthesis of erythromycin (34)(35)(36)38), and the ligands of TFs play nonnegligible roles in antibiotic biosynthesis (37,44,45). Antibiotics or their intermediates were previously found to function as ligands of TFs for feedback or feed-forward modulation in actinomycetes (3). Herein, for the first time, we verified that the direct precursors, P-CoA and MM-CoA, can coordinate the biosynthesis of erythromycin as ligands. An in vivo reporter system showed that 0.5 mM P-CoA or MM-CoA, which is approximately the physiological concentration (Fig. 4), can effectively relieve AcrT repressive effect on P eryAI (Fig. 3), suggesting that these two acyl-CoAs probably serve as signal molecules to promote erythromycin biosynthesis in Sac. erythraea. Very recently, methylcrotonyl-CoA, P-CoA, and A-CoA were found to be ligands of AccR in S. avermitilis, but they were not the starter and extender units for avermectin biosynthesis (49).
Moreover, our metabolomic analysis suggested that AcrT acts as a key coordinator to distribute the metabolic flux of A-CoA through the ACC and CS pathways. When acrT was deleted, A-CoA was more converted to M-CoA via the ACC path than via the CS path despite that it remained at relatively constant levels (Fig. 4). Although the level of citrate decreased in response to acrT deletion, it was still maintained at a high enough level to generate sufficient isocitrate, succinate, and (S)-malate as measured within the TCA cycle (Fig. 4). In Sac. erythraea, there are eight sets of putative ACC enzymes that may transform A-CoA to M-CoA and two CS enzymes that may transform A-CoA to citrate. RT-qPCR and EMSA analyses showed that AcrT directly represses three sets of ACC genes and stimulates one CS operon ( Fig. 4; Fig. S5), inferring that there may be additional pathways to control the other ACC or CS genes. Therefore, AcrT might be one center to regulate A-CoA metabolic flux by the opposite modulation of the ACC and CS pathways, reducing P-CoA and MM-CoA and increasing citrate for cell growth at the early stage. Meanwhile, with the accumulation of P-CoA and MM-CoA, they in turn served as the ligands of AcrT to modulate more A-CoA metabolic flux from the CS path to the ACC path (Fig. 5), indicating that the two acyl-CoAs also coordinate their own metabolism to supply more precursors for erythromycin biosynthesis.
Based on these findings, we propose a regulatory model for polyketide starter and extender units as ligands to coordinate erythromycin production (Fig. 7).
Furthermore, according to the precisely identified site of AcrT binding, we employed the PREDetector software to predict potential AcrT-binding target genes across the Sac. erythraea genome. A total of 359 putative target genes (cutoff score, $ 8) were identified ( Fig. S8 and Table S2B), among which 91 were functionally unassigned and the remaining 268 were divided into 19 categories involved in major metabolic pathways, such as transport and metabolism of carbohydrates and amino acids, as well as metabolism of lipids. This implies that AcrT plays a global regulatory role.
To explore the universality of polyketide starter and extender units acting as ligands in polyketide-producing actinomycetes, we first chose S. avermitilis, which produces the type I polyketide avermectin. As expected, the starter unit (MB-CoA or IB-CoA) and extender units (M-CoA and MM-CoA) were the ligands of AcrT Sa , a homolog of AcrT, and suppressed the biosynthesis of avermectin (Fig. 6). Subsequently, we assessed the type II polyketide producer S. coelicolor and confirmed that the starter unit A-CoA and extender unit M-CoA were also the ligands of AcrT Sc , which was also homologous to AcrT, except that AcrT Sc exhibited activation of actinorhodin biosynthesis (Fig. 6). Interestingly, in the biosynthesis of type I polyketides erythromycin and avermectin, AcrT and AcrT Sa showed the same regulatory pattern, whereas AcrT Sc exhibited an opposite effect on type II polyketide actinorhodin biosynthesis, indicating that AcrT homologs play complicated regulatory roles in transcriptional suppression or activation of antibiotic biosynthesis in actinomycetes. However, whether AcrT homologs suppress or activate the targets, the precursors as ligands are always beneficial to promote the biosynthesis of polyketides in actinomycetes.
All of our findings expand the knowledge that polyketide starter and extender units, beyond building blocks, play a vital role in coordinating the biosynthesis of antibiotics and enrich our understanding of the regulatory network in actinomycetes.

MATERIALS AND METHODS
Materials and culture conditions. The bacterial strains, plasmids, and primers used in this study are listed in Table S1 in the supplemental material. The sources of enzymes, chemicals, reagents, primers, and DNA sequencing services are shown in Table S2C. E. coli strains were grown in Luria-Bertani (LB) medium (in g/liter: yeast extract, 5; tryptone, 10; NaCl, 10) at 37°C. E. coli DH5a was used for DNA cloning, E. erythraea. When the intracellular P-CoA and MM-CoA pools are initially at low levels, AcrT interacts with the promoters of eryAI, ACC, PCC, and CS genes to inhibit the production of erythromycin and stimulate the biosynthesis of citrate beneficial to cell growth. With bacterial growth, the two acyl-CoAs accumulate enough to dissociate AcrT from those target promoters, increasing their own metabolic pools and turning to produce more erythromycin. ACC, A-CoA carboxylase; PCC, P-CoA carboxylase; CS, citrate synthase; 6-dEB, 6deoxyerythronolide B; Er-A, erythromycin A. Double arrows represent reactions of two steps or more. Whitefilled arrows represent translation. Black arrows represent transcriptional activation. Black blocked line represents transcriptional inhibition. Interruption is indicated by a red X. in a 50-ml total volume at 30°C for 30 min. DNase I (1 U/mg, Promega) digestion was carried out at 25°C for 60 s and stopped by adding DNase I Stop Solution (Promega) and heating at 65°C for 10 min. After purification, the samples were detected with a 3730XL DNA analyzer (Applied Biosystems), and data analyses were performed using the GeneMarker v2.2 software program.
BLI analyses. The binding affinity between the regulator (AcrT, PhoP, BldD, SACE_7301, or SACE_3446) and P eryAI were detected using the Octet K2 system with SA sensors (ForteBio) as previously described (55). The biotinylated P eryAI was obtained with the biotin-labeled primers BLI-eryAI-F1/R1 and immobilized on SA-coated biosensor tips. The reactions were conducted at 25°C in a buffer (1 mM Tris-HCl [pH 8.0], 5 mM MgCl 2 , 60 mM KCl, 10 mM DTT, 50 mM EDTA, and 10% glycerol), and the tips were immersed into wells containing purified proteins with appropriate concentration gradients. The data were set to an average model to determine the kinetic parameters K on and K off . The binding affinities (K D ) were then estimated as a ratio (K off /K on ) of the rate constants. Based on the DNase I footprinting assay, a 50-bp biotinylated probe containing site A (the precise binding site of AcrT) was obtained by directly annealing the biotin-labeled primers BLI-eryAI-F2/R2, and the K D value was calculated according to the same procedure.
Gene deletion, complementation, and overexpression. Gene deletion, complementation, and overexpression in Sac. erythraea were performed as previously described (34). Two 1.5-kb DNA fragments flanking acrT were successively obtained using the primer pairs acrT-F1/R1 and acrT-F2/R2 with the A226 genome as the template. The amplified fragments were digested, individually, with XbaI/HindIII and EcoRI/KpnI restriction enzymes and ligated into the corresponding sites of pUCTSR to obtain pUCDacrT. Through the homologous recombination of linear fragments, a 394-bp fragment within acrT was replaced by tsr in A226. Using the primers acrT-F3/R3, the desired thiostrepton-resistant mutant, named A226DacrT, was confirmed by PCR. A 639-bp acrT fragment was amplified with the primers acrT-F4/R4 and cloned into the NdeI/XbaI sites of pIB139 (54) to generate pIBacrT. Then, the complementation strain A226DacrT/pIBacrT and the overexpression strain A226/pIBacrT were obtained by apramycin resistance screening, and the strains A226DacrT/pIB139 and A226/pIB139 were used as the controls. Similarly, acrT was disrupted in the industrial Sac. erythraea strain WB, generating WBDacrT.
S. avermitilis mutant construction was done in light of the procedure described previously herein with minor revisions. A 3.6-kb fragment containing tsr and the two homologous arms was ligated into the HindIII/EcoRI sites of pKC1139 (53). The obtained pKCDacrT Sa was introduced into S. avermitilis NRRL8165 and integrated into the chromosome by single crossover recombination. The strain could lose the plasmid at 37°C, generating the DacrT Sa mutant. The DacrT Sa /pIBacrT Sa complementation strain and DacrT Sa /pIB139 control strain were likewise obtained.
The construction of the S. coelicolor mutant was done by the intergeneric conjugation method (44). Similarly, pKCDacrT Sc was obtained as described previously herein and introduced into E. coli ET12567 (pUZ8002). The strain was mixed with S. coelicolor M145 and cocultured on solid SFM medium, followed by coating with sodium naphthyridine and apramycin, generating a single crossover strain. The DacrT Sc mutant was subsequently obtained through temperature change as described for S. avermitilis. Based on intergeneric conjugation, the DacrT Sc /pIBacrT Sc complementation strain and DacrT Sc /pIB139 control strain were constructed.
Antibiotic fermentation and measurement. For flask fermentation of Sac. erythraea A226 and its derivatives, spores from R3M agar plates cultured for 3 days were inoculated into 50 ml TSB seed medium and shaken at 220 rpm at 30°C for 48 h. Five milliliters of culture was transferred into 50 ml of R5 liquid medium and grown for 6 days. Sac. erythraea strain WB and its derivative were cultivated in 50 ml of industrial fermentation medium [g/liter: cornstarch, 40; dextrin, 30; soybean flour, 30; soybean oil, 10; (NH 4 ) 2 SO 4 , 2; CaCO 3 , 6). After 24 h of fermentation, n-propanol (1.0 ml) was added to the broth, which was further shaken for 5 days at 30°C. Using a previously described method (35), Er-A was extracted from the fermentation culture and quantified by HPLC analysis.
Flask fermentation of S. coelicolor M145 and its derivatives was performed as previously described (44). Well-grown spores were first inoculated into 50 ml of TSBY medium at 30°C for 2 days. Five milliliters of culture was transferred into 50 ml of R5 liquid medium and further cultured for 7 days. To measure the actinorhodin yields of M145 and its derivatives, fermentation broths were treated with KOH solution at a final concentration of 1 M. After centrifugation (14,000 Â g, 5 min, 4°C), the supernatant was quantified at a 640-nm wavelength. The level of actinorhodin was normalized to the biomass of mycelia.
RNA preparation and RT-qPCR assays. Using a TransZol kit (Transgen), total RNA was prepared from bacteria in liquid fermentation medium at different time points. The quality and quantity of RNA were examined using a microplate reader (BioTek) and confirmed by electrophoresis. RNA samples were treated by reverse transcription using a HiScript II Q RT supermix (Vazyme) to obtain cDNAs for RT-qPCR. The assays were performed on the QuantStudio 6 Flex system (Applied Biosystems), using a Maxima SYBR green/ROX qPCR master mix (Vazyme). The experiments were carried out with three technical replicates and three independent biological replicates. Endogenous hrdB was used as a control. The transcript levels of various genes were determined according to the manufacturer's instructions.
CD spectroscopy. CD spectroscopic assays were recorded with a bandwidth of 2 nm at 25°C on a MOS-500 spectropolarimeter (Biologic) within the wavelength range of 200 to 250 nm. The protein was dissolved in 50 mM phosphate buffer solution (pH 8.0) at a final concentration of 0.1 mg/ml. The secondary structural characteristics of the proteins were estimated.
Untargeted metabolomic analyses. The mycelia of A226 and A226DacrT grown in R5 liquid medium for 72 h were harvested by centrifugation and washed at least three times with 50 mM phosphate buffer (pH 8.0). To remove proteins and obtain diverse metabolites, a mixture of methanol, acetonitrile, and water (2:2:1, vol/vol/vol) was added to the samples. After ultrasonic treatment, proteins were precipitated at 220°C for 1 h, followed by centrifugation (14,000 Â g, 20 min, 4°C). The supernatant was used to monitor and balance the system for quality control (QC) purposes. Metabolomic analysis was performed using an LC-MS/MS system consisting of a model 1290 Infinity UPLC (Agilent), a model 6550 mass spectrometer (Agilent), and a TripleTOF 6600 mass spectrometer (AB SCIEX) at Shanghai Applied Protein Technology Co., Ltd. Based on the data library from Shanghai Applied Protein Technology Co., Ltd., MetaboAnalyst 5.0 (https://www.metaboanalyst.ca) was used for the multivariate statistical analysis, including principal-component analysis (PCA), partial least-squares discrimination analysis (PLS-DA), and orthogonal partial least-squares discrimination analysis (OPLS-DA).
Detection of intracellular acyl-CoAs. Based on a previously described method (29), the extraction and detection of various intracellular acyl-CoAs were performed with minor revisions. The mycelia of Sac. erythraea were separated from fermentation broths grown in R5 liquid medium for 72 h by centrifugation, washed at least three times with 50 mM phosphate buffer (pH 8.0), and lysed in buffer (10 mM DTT and 10% trichloroacetic acid). The lysates were frozen and thawed three times with liquid nitrogen and ice water. After centrifugation at 4°C, the supernatants were transferred to Sep-Pak C 18 solid-phase extraction columns (Waters), washed with 0.1% trifluoroacetic acid (TFA), and eluted with 40% acetonitrile containing 0.1% TFA. The eluent was dried by a vacuum freeze dryer (Scientz). Acyl-CoAs were isolated and determined using an HPLC (Thermo Fisher Scientific), which was equipped with an InertSustain C 18 column (5 mm, 4.6 by 250 mm; Shimadzu) and equilibrated with a mixture of 98% solution A (50 mM KH 2 PO 4 , pH 5.5) and 2% solution B (acetonitrile). Samples were detected at 254 nm with a flow rate of 0.8 ml/min. The mobile-phase compositions were set to several gradients of 0 to 8 min (solution A from 98% to 95%), 8 to 12 min (solution A from 95% to 90%), 12 to 15 min (solution A from 90% to 85%), 15 to 19 min (solution A from 85% to 70%), and 19 to 22 min (solution A from 70% to 98%). The column was equilibrated with the aforementioned mixture for 10 min.
Statistical analysis. All data are presented as means 6 standard deviations (SDs) and were estimated by Student's two-tailed t test. P values of less than 0.05 were considered statistically significant. Significance is indicated as P , 0.05 (*), P , 0.01 (**), and P , 0.001 (***); ns indicates not significant. All error bars represent the SDs between independent experimental replicates.
Data availability. All data supporting the findings of this work are presented in the paper and the supplemental material.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.