The parent strain used in this study, S. hygroscopicus var. ascomyceticus FS35, was selected from S. hygroscopicus var. ascomyceticus ATCC 14891 after femtosecond laser irradiation (Qi et al., 2014a). All the strains and plasmids used in this study are listed in Supplementary Table 1. Strain FS35 and its derivatives were cultured and passaged on MS solid medium (20 g/L soybean cake meal, 20 g/L mannitol, and 20 g/L agar) at 28∘C. When the colonies on MS medium produced black spores, the spores were used to inoculate liquid seed medium and growth for 60 h at 28∘C and 220 rpm. The composition of the seed medium was same as described previously (Qi et al., 2014b). Then, the seed cultures were transferred into fermentation medium at an inoculation ratio of 10%. The fermentation broth was incubated for 192 h at 28∘C and 220 rpm. The fermentation medium used for the measurement of various fermentation parameters contained 20 g/L soluble starch, 40 g/L dextrin, 5 g/L yeast powder, 5 g/L peptone, 5 g/L corn steep liquor, 1 g/L K₂HPO₄, 1.5 g/L (NH₄)₂SO₄, 0.5 g/L MnSO₄, 1 g/L MgSO₄⋅7H₂O, 1 g/L CaCO₃, and 2.5 mL/L soybean oil. E. coli DH5α was used for plasmid construction. E. coli ET12567 (pUZ8002) was used as the donor strain for intergeneric conjugation with S. hygroscopicus var. ascomyceticus. E. coli BL21(DE3) was used for heterologous protein expression. All E. coli strains were cultured in Luria-Bertani (LB) medium at 37∘C. The plasmid pET28a(+) was used to construct the protein expression vector. The plasmid pKC1139 was used to construct the gene deletion vector. The plasmid pSET152 was used for the construction of a gene complementation vector. The plasmid pIB139, which contains the strong promoter ermE^∗p, was used for the construction of a gene overexpression vector.

Transformation of Streptomyces was done using intergeneric conjugation with an E. coli donor strain according to standard methods (Kieser et al., 2000; Ma et al., 2018). The raw genome sequences were uploaded to the Gene Expression Omnibus (GEO) database at the National Center for Biotechnology Information (NCBI) (accession number: GSE143832). All primers used for gene manipulation are listed in Supplementary Table 2. To construct the glnB deletion strain ΔglnB, the 1,009-bp upstream flanking region and 1,003-bp downstream flanking region of glnB were, respectively, amplified from the genome of FS35 using the primer pairs DglnB-UF/DglnB-UR and DglnB-DF/DglnB-DR. The obtained fragments were fused by overlap-extension PCR using the primer pair DglnB-UF/DglnB-DR. The fusion fragment was inserted into the plasmid pKC1139 between the HindIII and XbaI sites to obtain the glnB deletion vector pKCglnB. The recombinant plasmid pKCglnB was introduced into FS35 by traditional intergeneric conjugation. The single-crossover transformants were selected on MS solid medium containing 50 μg/mL apramycin. After two rounds of sporulation on MS solid medium without apramycin, the apramycin-sensitive colony was selected as the double crossover strain. The deletion strain was verified by PCR and sequencing using the primer pair VDglnB-F/VDglnB-R. The glnK deletion strain ΔglnK and the glnB-glnK double deletion strain ΔglnBΔglnK were constructed using the same method, and were verified by PCR and sequencing using the primer pairs VDglnB-F/VDglnB-R and VDglnK-F/VDglnK-R, respectively.

For the construction of the glnK complementation strain ΔglnBΔglnK/pSETglnK based on the double deletion strain ΔglnBΔglnK, the native promoter sequence of the amtB-glnK-glnD operon was amplified from the genome of FS35 using the primer pair CglnK-PF/CglnK-PR. The DNA sequence of glnK was amplified from the genome of FS35 using the primer pair CglnK-KF/CglnK-KR. The obtained fragments were fused by overlap-extension PCR using the primer pair CglnK-PF/CglnK-KR. The fusion fragment was inserted into the vector pSET152 between the BamHI and XbaI sites to construct the complementation vector pSETglnK. The recombinant vector pSETglnK was introduced into the ΔglnBΔglnK strain by traditional intergeneric conjugation to obtain the complementation strain ΔglnBΔglnK/pSETglnK. which was verified by PCR and sequencing using the primer pair pSET152-F/pSET152-R.

For the construction of the pccB overexpression strain ΔglnB/pIBOpccB based on the ΔglnB strain, the DNA sequences of pccB was amplified from the genome of FS35 using the primer pair OpccB-F/OpccB-R. The obtained fragment was inserted into the vector pIB139 between the NdeI and XbaI sites to construct the overexpression vector pIBOpccB, in which the pccB coding sequence was placed under the control of the strong constitutive promoter ermE^∗p. The recombinant vector pIBOpccB was introduced into the ΔglnB strain by traditional intergeneric conjugation to obtain the overexpression strain ΔglnB/pIBOpccB, which was verified by PCR and sequencing using the primer pair pIB139-F/pIB139-R. The pccE overexpression strain ΔglnB/pIBOpccE and the pccB-pccE co-overexpression strain ΔglnB/pIBOpccBE were constructed based on the ΔglnB strain using the same method. They were verified by PCR and sequencing using the primer pair pIB139-F/pIB139-R.

To measure the production of FK520 and FK523, 2 ml of fermentation broth was mixed with 3 ml of ethanol. After 30 min of ultrasonic extraction, and 10 min of centrifugation at 8,000 × g, the supernatant was filtered through a 0.2 μm syringe-driven filter (Spartan10463100, Whatman, England), suitable for organic solvents. The concentrations of FK520 and FK523 were quantified by liquid chromatography on a 1100 series instrument (Agilent, United States), equipped with a C-18 column (150 mm × 4.6 mm, 3.5 μm; Agilent). The mobile phase and gradient elution program were same as reported previously (Yu et al., 2019). The flow rate was 2 mL/min and the detection wavelength was 205 nm. The injection volume was 20 μL and the column temperature was 60∘C.

To measure the biomass concentration, mycelia from 5 ml of fermentation broth were washed once with 0.1 M-HCl solution and twice with Milli-Q water. After centrifugation for 10 min at 8,000 × g, the wet cell pellet was dried in an oven at 80∘C until a constant weight to measure the dry cell weight (DCW). The biomass concentration was defined as the ratio of DCW to the volume of the fermentation broth.

In order to obtain the native or His₆-tagged proteins, the target proteins were overexpressed in E. coli BL21 (DE3). All primers used for heterologous protein expression are listed in Supplementary Table 3. The nucleotide sequences of target proteins were amplified from the genome of FS35 using the corresponding primers. These resulting DNA fragments were digested with the listed restriction enzymes (Supplementary Table 3), and inserted into the corresponding sites of the pET28a (+) vector through enzymatic ligation. The constructed recombinant plasmids were introduced into E. coli BL21 (DE3) for the heterologous expression of target proteins. All recombinant strains were cultivated in Luria-Bertani (LB) medium with 50 μg/mL of kanamycin at 37∘C until the optical density of the fermentation broth at 600 nm (OD₆₀₀) reached 0.6–0.7, after which 0.5 mM of Isopropyl β-d-thiogalactoside (IPTG) was added to induce the expression of target proteins. After 3 h of expression at 37∘C, the cells were collected by centrifugation at 12,000 × g for 10 min and resuspended in lysis buffer (50 mM Tris-HCl, 0.1 M KCl, and 20% glycerol, pH 7.5). The cells were then disrupted by sonication, the lysate was centrifuged at 12,000 × g for 10 min to remove cell debris, and the resulting cleared supernatant was collected for protein purification. All α subunits used in this study were completely biotinylated according to a published method before purification (Rodrigues et al., 2014).

The His₆-tagged proteins were purified using a Ni²⁺-nitrilotriacetic acid (Ni²⁺-NTA) agarose affinity chromatography column (Qiagen, Germany), according to the manufacturer’s protocol. The native proteins were purified using a Hi-Trap Heparin column (GE Healthcare), which was pre-equilibrated with a previously published buffer (Moure et al., 2012). The proteins were eluted with a buffer gradient from 0.1 to 1 M KCl. After further dialyzed in a previously published buffer (Rodrigues et al., 2014), the eluted proteins were verified by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) (Supplementary Figure 1). The concentration of the purified proteins was measured using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, United States). The purified proteins were stored at −80∘C for subsequent protein co-precipitation and electrophoretic mobility shift assays (EMSA).

The formation of complexes between proteins was assessed using in vitro protein co-precipitation as described previously (Huergo et al., 2007). Briefly, 5 μL of His-magnetic beads were washes twice with buffer containing 50 mM Tris-HCl pH 8, 0.1 M NaCl, 0.1% (w/v) lauryl dimethyl amine oxide (LDAO), 20 mM imidazole, 10% glycerol and 5 mM MgCl₂, together with the indicated effectors (2-OG and ATP) to achieve the pre-equilibrium state. The purified proteins (20 μg GlnB, 20 μg His-GlnB, 20 μg GlnK, 20 μg α subunit, 20 μg His-α subunit) required for the different experiments were added into 500 μL of buffer to induce complex formation. After binding for 20 min, the magnetic beads were washed three times with the binding buffer. The washed beads were then incubated in 20 μL of buffer containing 0.5 M imidazole for 5 min, and the resulting eluent was analyzed by 15% acrylamide SDS-PAGE followed by staining Coomassie brilliant blue R-250 to determine the components of the protein complex.

The in vitro activity of ACC/PCC was measured using previously described coupled enzyme activity assay (Beez et al., 2009; Gerhardt et al., 2015), with some modifications. The hydrolysis of ATP catalyzed by ACC was coupled to the formation of pyruvate catalyzed by pyruvate kinase (PK) and then coupled to the oxidation of NADH catalyzed by the lactate dehydrogenase (LDH). The coupled enzymatic reactions were carried out in buffer containing 10 mM NaHCO₃, 10 mM ATP, 1 mM phosphoenolpyruvate, 0.2 mM NADH, 20 mM MgCl₂, 50 mM KCl, 0.5 mM dithiothreitol (DTT), 50 mM imidazole, 4.4 units of LDH, and 6 units of PK. The reaction systems also included 10 nM AccB or 10 nM PccB, 20 nM AccA, along with the indicated concentrations of PII proteins and 2-OG, in a final volume of 400 μL with a final pH of 7.5. After the reaction systems were pre-incubated at 25∘C for 15 min, 400 μM acetyl-CoA or 400 μM propionyl-CoA was added to start the enzymatic reactions. The oxidation of NADH over a period of 20 min at 25∘C was measured by recording the decrease at 340 nm using a SPECORD 200 photometer (Analytik, Jena).

In order to measure the content of intracellular coenzyme A esters, mycelia obtained form 5 mL of fermentation broth were washed twice with deionized water and centrifuged at 8,000 × g for 10 min. The wet hyphae were suspended in 300 μL of 15% trichloroacetic acid and lysed by vortexing for 3 min at 4∘C with 150 μL of glass beads. After centrifugation for 10 min at 8,000 × g and 4°C, the supernatant was passed through an OASIS HLB SPE cartridge (waters, United States) under vacuum to extract coenzyme A esters, as reported previously (Mo et al., 2012a). The analysis of coenzyme A esters was performed using Ultra-high performance liquid chromatography (UPLC, Waters, United States)-electrospray ionization (ESI)-tandem mass spectrometry (MS; Xveo TQ-XS, Waters, United States), equipped with an ACQUITY UPLC HSS T3 column (100 mm × 2.1 mm, 1.7 μm; Waters, United States), as previously reported, with some modifications (Park et al., 2007). The mobile phase was composed of 5 mM ammonium acetate and 0.05% acetic acid in water (A), and 80% acetonitrile with the same additive concentration (B). The gradient elution program included: 0–0.8 min, 90% A/10% B; 0.8–2 min, a linear gradient from 90% A/10%B to 60% A/40% B; 2–6 min, a linear gradient from 60% A/40% B to 20% A/80% B; 6–8 min, 20% A/80% B; 8–12 min, and a linear gradient from 20% A/80% B to 90% A/10% B. The quantification was done in multiple reaction monitoring (MRM) mode with two mass ions: m/z parent > m/z daughter (acetyl-CoA, 810 > 303; malonyl-CoA, 854 > 347; propionyl-CoA, 824 > 317; methylmalonyl-CoA, 868 > 361).

For the EMSA, the promoter sequence of the amtB-glnK-glnD operon was amplified from the genome of FS35 using the primer pair pBKD-F/pBKD-R (Supplementary Table 3). The Cy5-labeled primer 5′-AGCCAGTGGCGATAAG-3′ was used for the secondary amplification of the obtained promoter sequence to generate the Cy5-labeled DNA probes. Then 5 ng of the labeled probe was incubated for 30 min at 28∘C with different concentration of His₆-GlnR protein in a previously described reaction buffer (Zhang et al., 2019). After mixing with the loading buffer (0.25 × Tris borate EDTA (TBE) buffer, 60% glycerol), the reaction mixture was loaded on to 6% native polyacrylamide gels and separated for 40 min in an ice bath containing 0.5 × TBE at 90 V. Then, the DNA bands with bound or unbound proteins were visualized by fluorescence imaging using a Typhoon Trio variable mode imager (GE Healthcare, United States).

The transcriptional levels of gene glnK in the parent strain FS35 and the deletion strain ΔglnB were measured by qRT-PCR at 60 h. The primers used for qRT-PCR were listed in Supplementary Table 3. Firstly, the fermentation broth of FS35 and ΔglnB was collected at 60 h. After centrifugation at 8,000 × g for 10 min, the wet hyphae were rapidly frozen in liquid nitrogen. Then the total RNA was extracted from the frozen hyphae using RNAprep Pure Cell/Bacteria Kit (Tiangen, Beijing, China). The concentration and integrity of the RNA sample was detected by 1% agarose gel electrophoresis. Then the RNA sample was reversely transcribed into cDNA by using PrimeScript^TM RT reagent Kit with gDNA Eraser (takara, Japan). To exclude DNA contamination, the RNA sample which treated by gDNA Eraser but not reverse transcription was used as a template for negative control. With the cDNA as template, qRT-PCR was carried out on a LightCycler^® 480 using SYBR Green Master Mix (Roche, Switzerland). The 16S rRNA was used as the internal reference gene, and the change folds of the transcriptional levels were quantified relatively with the comparative C_T method (Wang et al., 2018).

In this study, the samples used for the measurement of biomass and FK520 yield were taken from five independent technical replicates. The samples used for the analysis of coenzyme A esters’ content, enzyme activity and quantitative real-time PCR were taken from three independent technical replicates. All data were presented as the mean values of respective independent technical replicates and the error bars indicate the standard deviations (SD).

The signal transduction proteins belonging to the PII family are classified into three subgroups, GlnK, GlnB, and NifI (Huergo et al., 2013). The genome-wide sequencing results revealed that there are two PII signal transduction proteins (1_2537 and 1_5736) in S. hygroscopicus var. ascomyceticus. The amino acid sequence alignment showed that these two PII signal transduction proteins shared 66.96% amino acid sequence identity (Figure 1B). In order to determine which subgroups these two PII signal transduction proteins belong to, the amino acid sequences encoded by the 1_2537 and 1_5736 genes of S. hygroscopicus var. ascomyceticus were aligned with the GlnK proteins from three other actinobacteria (S. coelicolor, Mycobacterium tuberculosis, and Corynebacterium glutamicum) and the GlnB protein from E. coli (Figures 1C,D). The sequence alignment indicated that the protein encoded by 1_5736 shared approximately 81% amino acid sequence identity with the GlnK proteins of other actinobacteria (Figure 1C), while the protein encoded by 1_2537 shared approximately 61% amino acid identity with the GlnB protein of E. coli (Figure 1D). This implied that the protein encoded by 1_2537 belonged to the GlnB subgroup of PII signal transduction proteins, while the protein encoded by 1_5736 belonged to the GlnK subgroup.

The gene encoding GlnK protein in actinobacteria is located adjacent to the amtB gene (responsible for ammonia transfer) and the glnD gene (responsible for uridylyl transfer) in an amtB-glnK-glnD operon (Arcondeguy et al., 2001; Huergo et al., 2013). The genome-wide sequencing results and sequence homology analysis revealed that 1_5735 and 1_5737, the genes adjacent to 1_5736 in S. hygroscopicus var. ascomyceticus, respectively, encoded the proteins AmtB and GlnD (Supplementary Figure 2). They also formed a complete operon with gene 1_5736, as in other actinobacteria (Figure 1E). This further confirmed that the 1_5736 gene of S. hygroscopicus var. ascomyceticus belongs to the GlnK subgroup of PII signal transduction proteins.

In E. coli, the BC and BCCP functional subunits of ACC are, respectively, composed of two different polypeptides (Figure 2A). However, the BC and BCCP functional units of actinobacteria are fused into a single protein, which is called the α subunit of ACC (Tong, 2013). The amino acid sequence alignment between the α subunit of ACC annotated in the genome-wide sequencing and the α subunit of ACC with known biological function from other actinobacteria showed that the 1_3403 gene (accA) encoded the α subunit of ACC in S. hygroscopicus var. ascomyceticus (Supplementary Figure 3). This also confirmed that the BC and BCCP functional units were also fused together to form the α subunit of ACC in S. hygroscopicus var. ascomyceticus, as in other actinobacteria (Figure 2A).

To assess whether there was a similar interaction between the α subunit of ACC and two PII signal transduction proteins in S. hygroscopicus var. ascomyceticus as is found in E. coli, where the GlnB protein inhibits the activity of ACC by binding with the BC-BCCP complex (Gerhardt et al., 2015), the His-tagged α subunit and two native PII signal transduction proteins were purified and mixed with ATP or 2-OG for protein co-precipitation experiments (Figure 2B). The results of SDS-PAGE analysis showed that the His-tagged α subunit and GlnB protein could be simultaneously eluted by imidazole in the presence of ATP (Figure 2B, lane 3), and this complex disassociated in the presence of a sufficient 2-OG concentration (Figure 2B, lane 4). By contrast, GlnK protein was not found in the eluent irrespective of the presence of ATP or 2-OG (Figure 2B, lane 6 and 7). This suggested that GlnK protein could not bind to the α subunit, so it might not affect the activity of ACC in S. hygroscopicus var. ascomyceticus.

To further investigate the effects of 2-OG at different concentrations on the interaction between the GlnB protein and the α subunit, protein co-precipitation experiments were carried out with His-tagged GlnB as the bait at different concentrations of 2-OG. The corresponding SDS-PAGE gels showed that when His-tagged GlnB was immobilized on the Ni²⁺ beads as bait, both His-tagged GlnB protein and the α subunit could be eluted in the presence of ATP (Figure 2C, lane 1). This in turn confirmed the binding of GlnB protein to the α subunit. In addition, when the concentration of 2-OG was controlled within a certain range, GlnB protein could still bind to the α subunit to different degrees, but this interaction gradually weakened with the increase of 2-OG concentration (Figure 2C, lanes 2–4). When the concentration of 2-OG increased to 500 μM, the protein complex completely disassociated (Figure 2C, lanes 5–7). This indicated that the interaction between the GlnB protein and the α subunit was regulated by 2-OG, which is consistent with the regulation of its homologs in E. coli and Synechocystis sp. PCC 6803 (Gerhardt et al., 2015; Hauf et al., 2016).

Considering that PCC had been reported to share the same α subunit with ACC in actinobacteria (Diacovich et al., 2002; Gago et al., 2006; Oh et al., 2006), the effects of the interaction between GlnB protein and the α subunit on the in vitro activity of ACC and PCC was assessed using coupled enzyme activity assays. Compared to the activity of ACC and PCC in the reaction system without GlnB protein, which was defined as 100%, the activity of ACC and PCC, respectively, decreased to 54.23 and 76.51% in the presence of GlnB protein (Figure 3A). This result confirmed that the interaction between GlnB protein and the α subunit could inhibit the activity of ACC and PCC. However, GlnK protein had almost no influence on the activity of these two enzymes, which was consistent with the co-precipitation results.

Further assays with different concentrations of PII protein showed that when the concentration of GlnB protein increased from 0 to 1 μM, the inhibition of ACC and PCC intensified sharply, and gradually stabilized at a GlnB protein concentration above 1.5 μM (Figure 3B). The maximal calculated inhibition of ACC by GlnB protein reached 50.04%, while the maximal calculated inhibition of PCC by GlnB protein reached only 24.69% (Figure 3B). This indicated that GlnB protein had a stronger inhibitory effect on ACC than on PCC. In order to analyze whether 2-OG could eliminate the inhibitory effects of GlnB protein on ACC and PCC, the coupled enzyme activity assays were carried out in the presence of different concentrations of 2-OG. The results showed that when the concentration of 2-OG increased from 0 to 20 μM, the inhibition of ACC and PCC was greatly alleviated, and when the concentration of 2-OG reached 500 μM, the inhibition of ACC and PCC was almost completely eliminated (Figure 3C). This demonstrated that the inhibition of ACC and PCC activity by GlnB protein could be alleviated by 2-OG in a concentration-dependent manner.

The effect of GlnB protein on the activity of ACC and PCC in vivo was assessed by measuring the changes in the intracellular concentrations of substrates (acetyl-CoA and propionyl-CoA) and products (malonyl-CoA and methylmalonyl-CoA) caused by the knockout of the glnB gene. As shown in Figure 4, during the stationary phase of fermentation (96–168 h), the amount of acetyl-CoA in the knockout strain ΔglnB was always lower than in the parental strain FS35, while the content of malonyl-CoA was always higher (Figures 4A,B). In detail, the content of acetyl-CoA in strain ΔglnB increased at a lower rate than in FS35 during 96–120 h, and then decreased at a faster rate in the subsequent fermentation period. However, the amount of malonyl-CoA in strain ΔglnB increased at a higher rate during 96–144 h (Figures 4A,B). These results illustrated that the removal of GlnB protein enhanced the conversion of acetyl-CoA into malonyl-CoA, indicating the release of ACC inhibition in vivo. However, the increase of malonyl-CoA concentration in strain ΔglnB during 120–144 h was lower than during 96–120 h, and the amount of malonyl-CoA decreased at a faster rate than in FS35 during 144–168 h (Figure 4B). These changes might be caused by the increased consumption of malonyl-CoA as a substrate for downstream biochemical reactions or as a precursor for the biosynthesis of FK520 in strain ΔglnB (Figure 1A).

Due to the increased supply of malonyl-CoA, the amount of propionyl-CoA in strain ΔglnB was always higher than in the parental strain FS35. In detail, the increase rate of propionyl-CoA in strain ΔglnB was lower than in FS35 during 96–120 h, and the decrease rate was higher after 120 h (Figure 4C). In addition, the content of methylmalonyl-CoA in strain ΔglnB was higher than in FS35 during the early stationary phase (Figure 4D). These results demonstrated that the removal of GlnB protein also increased the conversion of propionyl-CoA into methylmalonyl-CoA, indicating the release of PCC inhibition in vivo. The dramatic reduction of the methylmalonyl-CoA concentration in strain ΔglnB during the stationary phase, which was even faster than in the parental strain FS35, might be caused by the increased consumption of methylmalonyl-CoA as a precursor for the biosynthesis of FK520. In summary, the deletion of the glnB gene alleviated the inhibition of ACC and PCC activity by its encoded GlnB protein, and increased the supply of the precursors malonyl-CoA and methylmalonyl-CoA for the biosynthesis of FK520.

The fermentation results showed that the yield of FK520 in the knockout strain ΔglnB was significantly improved (to 390 ± 10 mg/L) due to the increased supply of the precursors malonyl-CoA and methylmalonyl-CoA. However, the yield of FK520 in the knockout strain ΔglnK was not significantly different from the parent strain FS35 (Figure 5A). The biomass measurement results showed that the individual deletion of glnB or glnK did not affect the biomass accumulation, while the double deletion of glnB and glnK resulted in a significant decrease of the biomass yield (Figure 5B). Furthermore, colonies of the glnB-glnK double mutant on agar plates also showed poorer mycelial growth (Supplementary Figure 4). At the same time, the yield of FK520 was also decreased sharply in the glnB-glnK double knockout strain (Figure 5A). These results indicated that the double deletion of glnB and glnK might cause a disturbance of intracellular metabolism. However, when GlnK expression was complemented in the glnB-glnK double deletion strain, the biomass accumulation and the FK520 production both recovered to the same levels as in the parental strain FS35 (Figures 5A,B). This suggested that the dysregulation caused by the removal of GlnB protein might be compensated by GlnK protein, which appears to have a partially redundant role.

However, unlike the constitutive expression of GlnB protein, the expression of GlnK protein is controlled by nitrogen regulator I (NRI) in E. coli, or by the transcriptional regulator GlnR in actinobacteria (vanHeeswijk et al., 1996; Fink et al., 2002; Amon et al., 2008). Therefore, to assess whether GlnK protein is also regulated by GlnR in S. hygroscopicus var. ascomyceticus, the known DNA binding sites of the GlnR regulator in S. coelicolor were collected and loaded into WebLogo¹ to create a sequence logo (Figure 5C), and the binding site of the GlnR regulator in the promoter region of the amtB-glnK-glnD operon from S. hygroscopicus var. ascomyceticus was identified using the Motif Alignment and Search Tool (MAST) (Figure 5C). An EMSA with a His-labeled promoter probe and different concentrations of GlnR protein demonstrated that the GlnR regulator did bind to the promoter region of amtB-glnK-glnD operon (Figure 5D). The specificity of this binding was further confirmed by EMSA, in which specific and non-specific unlabeled proteins were added for the competitive binding (Figure 5E). These results indicated that when GlnB protein was absent, GlnK protein regulated the related metabolism under the control of the global regulator GlnR to maintain the normal growth of the strain. This was also demonstrated by the up-regulated expression of glnK in the glnB deletion strain ΔglnB compared with the parent strain FS35 (Supplementary Figure 5).

Although the removal of GlnB protein relieved its inhibition of ACC and PCC, increasing the yield of FK520 to a certain extent, the content of methylmalonyl-CoA in the late period of fermentation was still low (Figure 4D). Previous research suggested that the specific β subunit (responsible for substrate recognition) and ε subunit (responsible for protein- protein interactions) are crucial for the carboxylation activity of PCC (Diacovich et al., 2002; Demirev et al., 2011; Tong, 2013). The results of amino acid sequence alignment showed that the 1_3398 gene (pccB) encoded the β subunit, while the 1_3397 gene (pccE) encoded the ε subunit of PCC in S. hygroscopicus var. ascomyceticus (Supplementary Figure 6). Therefore, in order to further improve the production of FK520, the pccB and pccE genes were overexpressed alone or in combination in the glnB knockout strain to further increase the supply of methylmalonyl-CoA. The fermentation results showed that the single overexpression of pccB or pccE improved the supply of methylmalonyl-CoA during the early stationary phase (96–144 h), and increased the yield of FK520 to a certain extent (Figures 6A,B). Moreover, the co-overexpression of pccB and pccE in strain ΔglnB ensured a sufficient supply of methylmalonyl-CoA throughout the stationary phase (96–168 h), thus increasing the production of FK520 to 550 ± 20 mg/L, which was 1.9-fold higher than the yield of the parent strain FS35 (287 ± 9 mg/L) (Figures 6A,B).

FK523 is the main impurity in the production of FK520, which is formed by the assembly of methylmalonyl-CoA at the C21 position of the macrolide skeleton instead of the specific precursor ethylmalonyl-CoA (Yu et al., 2019). The quantification of the FK523 yield and the FK523/FK520 ratio in the parental strain FS35 and the co-overexpression strain ΔglnB/pIBOpccBE showed that although the FK523 yield of strain ΔglnB/pIBOpccBE was also increased, the FK523/FK520 ratio did not increased significantly (Supplementary Figure 7). This confirmed that the genetic engineering methods for increasing the yield of FK520 described in this study are reasonable and effective. Nonetheless, the reduction of the FK523 ratio requires further efforts in future studies.