The strains and plasmids used or constructed in this study are listed in Table 1, and the primers in Supplementary Table S1. Culture conditions for Escherichia coli and S. avermitilis strains were as described previously (Liu W. et al., 2015). MM, R2YE (Kieser et al., 2000) and YMS (Ikeda et al., 1988) plates were used for phenotypic observation of S. avermitilis strains. Insoluble fermentation medium FM-I (Chen et al., 2007) was used for routine avermectin production. Soluble fermentation medium FM-II (Guo et al., 2010) was used to grow mycelia for biomass analysis, and for RNA isolation following stress treatment.

A sig8 in-frame gene deletion mutant was constructed using a homologous recombination strategy. Two homologous fragments flanking sig8 were amplified by PCR from wild-type (WT) genomic DNA. A 529-bp 5′-flanking region (positions -337 to +192 relative to the sig8 start codon) was amplified with primers SD2A/SD2B, and a 541-bp 3′-flanking region (positions +835 to +1375) was amplified with primers SD3A/SD3B. The two fragments were cut with HindIII/BamHI and BamHI/EcoRI, respectively, and then simultaneously cloned into HindIII/EcoRI-digested pKC1139 to generate sig8 deletion vector pKCDsig8. Non-methylated pKCDsig8 was transformed into WT protoplasts, and double-crossover recombinant strains were isolated as reported previously (Zhao et al., 2007). Resulting sig8 deletion mutants were verified by colony PCR using primers SD23A/SD23B (flanking the exchange regions) and SD39A/SD39B (located within the deletion region of sig8) (Supplementary Figure S1), followed by DNA sequencing. When primers SD23A/SD23B were used, a 1.2-kb band appeared, whereas a 1.9-kb band was detected in WT genomic DNA. When primers SD39A/SD39B were used, only WT DNA generated a 390-bp band. We thus obtained sig8 gene deletion mutant Dsig8, in which 840-bp sig8 ORF was mostly deleted (from positions +193 to +834 relative to the start codon) (Supplementary Figure S1). The deletion part of sig8 covered coding region for all three functional domains (σ₂, σ₃, and σ₄); Thus, the remaining fragment was unlikely to be functional.

To construct a sig8sav_742 double deletion mutant, the pKCDsig8 vector was transformed into D742 protoplasts (Sun et al., 2016). The expected mutant, Dsig8–742, was isolated by selection of the Dsig8 mutant.

For complementation of Dsig8, a 1618-bp PCR fragment containing the sig8 promoter and open reading frame (ORF) was amplified with primers SD1B/SD1D, and inserted into the integrative vector pSET152 to give sig8-complemented vector pSET152-sig8, which was then transformed into Dsig8 to obtain complemented strain Csig8.

For overexpression of sig8, a 987-bp fragment carrying the sig8 ORF was amplified using primers SD1B/SD1C, and inserted into pJL117 to generate pJL117-sig8, in which sig8 was controlled by ermE^∗p (Streptomyces strong constitutive promoter). The 1.3-kb EcoRI/HindIII fragment containing ermE^∗p and sig8 ORF from pJL117-sig8 was ligated into pKC1139 to generate sig8 overexpression vector pKC1139-erm-sig8, which was transformed into WT protoplast to construct sig8 overexpression strain Osig8.

Fermentation of S. avermitilis strains and HPLC (high-pressure liquid chromatography) analysis of avermectin yield in fermentation broth were performed as described previously (Chen et al., 2007).

S. avermitilis mycelia grown in FM-I or FM-II were collected at various time points for RNA isolation. Triturated samples were suspended in TRIzol reagent (Tiangen, China) for RNA isolation. Genomic DNA contamination was removed by treatment of RNA samples with RNase-free DNase I (TaKaRa, China). 4 μg total RNA was used for cDNA synthesis. qRT-PCR was performed using primers listed in Supplementary Table S1 to analyze transcription levels of the tested genes as described previously (Luo et al., 2014), with expression level of housekeeping gene 16S rRNA as internal control. Each experiment was repeated three times.

The 953-bp sig8 coding region (279 amino acids) was amplified from WT genomic DNA using primers SD1A/SD1B. The obtained PCR product was inserted into pET-28a (+) to generate pET-sig8. After confirmation by DNA sequencing, pET-sig8 was transformed into E. coli BL21 (DE3) for His₆-σ⁸ (N-terminal His₆-tagged σ⁸ recombinant protein) overexpression. Following IPTG induction, cells containing His₆-σ⁸ were harvested, washed, resuspended in a lysis buffer (Luo et al., 2014), and disrupted by sonication on ice. His₆-σ⁸ present as inclusion body in E. coli was purified, solubilized, and renatured as described previously (Luo et al., 2014). Purified renatured soluble His₆-σ⁸ was quantified and stored at -80°C for use in vitro transcription experiments and electrophoretic mobility shift assays (EMSAs).

EMSA probes carrying respective tested promoter regions were obtained by PCR using primers listed in Supplementary Table S1, and labeled at their 3′ ends with digoxigenin (DIG). Conditions for binding reaction and detection were as described previously (Sun et al., 2016). To confirm binding specificity between His₆-σ⁸ and DNA probes, a ∼300-fold excess of each specific or non-specific (hrdB) unlabeled probe was added to the reaction mixture before incubation.

DNA templates containing respective promoter and partial coding region were amplified by PCR using primers listed in Supplementary Table S1. Conditions for transcription assays and detection of transcripts were as described previously (Luo et al., 2014). Reaction mixtures contained 0.4 pmol DNA template and various amounts of renatured His₆-σ⁸. Transcription was initiated by addition of 3.7 pmol E. coli core RNA polymerase (core-RNAP) (New England Biolabs, USA).

The transcriptional start site (TSS) of selected genes was identified using a 5′/3′ RACE kit (Roche, 2nd Generation). S. avermitilis strains were cultured in FM-II at 28°C for 2 days, and then subjected to various stress conditions. Mycelia were harvested after various durations of stress treatment and used for RNA isolation. 4 μg total RNA was reverse transcribed with 40 pmol gene-specific primer SP1. Purified cDNAs were added to oligo-dA tails at the 3′ end by TdT (terminal deoxynucleotidyl transferase, TaKaRa) treatment at 37°C for 30 min. The tailed cDNA was used as template for first-round PCR using second inner gene-specific primer SP2 and oligo-dT anchor primer. The resulting PCR product was diluted to appropriate concentration, and used as template for second-round PCR with nested primer SP3 and an anchor primer. The purified final PCR product was sequenced (Invitrogen Biotechnology Corporation, China). TSS was determined as the first nucleotide following the oligo-dT sequence.

The sig8 (sav_741) gene in the S. avermitilis chromosome contains 840 nucleotides (nt) and encodes a 279-amino-acid σ⁷⁰-family alternative σ factor, σ⁸. The downstream convergently transcribed gene sav_742, located 377 nt from sig8 (Figure 1A), encodes an AraC-family transcriptional regulator that was recently characterized as a global regulator of avermectin biosynthesis, cell growth, and morphological development (Sun et al., 2016). The upstream convergently transcribed gene sav_740, located 215 nt from sig8, encodes a hypothetical protein. BLAST analysis revealed that σ⁸ displays high amino acid sequence identities with S. coelicolor σ^B (77.22%) and S. hygroscopicus σ^B (78.29%).

To investigate the role of σ⁸ in avermectin biosynthesis, we constructed sig8 deletion mutant Dsig8, complemented strain Csig8 and overexpression strain Osig8, and compared their avermectin yields with that of WT ATCC31267 cultured in FM-I for 10 days. Relative to WT yield, that of Dsig8 was ∼96% higher, that of Osig8 was ∼50% lower, and that of Csig8 was similar (Figure 1B). Yields of two vector control strains, WT/pSET152 and WT/pKC1139, were almost the same as that of WT (Figure 1B). These findings indicate that σ⁸ has a negative effect on avermectin production.

To assess the effect of σ⁸ on S. avermitilis growth, we measured biomasses of WT, Dsig8 and Csig8 cultured in soluble FM-II. Biomass of Dsig8 and Csig8 was similar to that of WT (Figure 1C), indicating that σ⁸ has no effect on cell growth, and that avermectin overproduction in Dsig8 does not result from alteration of growth. Dsig8 and Osig8 grew normally on YMS, R2YE, and MM plates (Supplementary Figure S2), indicating that σ⁸ also has no effect on morphological differentiation.

To investigate the relationship between σ⁸ and avermectin production, we examined the sig8 transcription profile by qRT-PCR using RNAs isolated from WT in FM-I. sig8 transcript was detected throughout the fermentation process. Its level increased gradually from day 1, reached a maximum on day 6, and remained high thereafter, suggesting that σ⁸ plays its regulatory role mainly during the late stage of fermentation (Figure 1D).

The effect of σ⁸ on expression of ave genes was assessed by qRT-PCR analysis of transcription levels of aveR (CSR gene) and aveA1 (structural gene encoding AVES1 polyketide synthase) in WT and Dsig8 grown in FM-I for 2 days (early exponential phase) or 6 days (stationary phase). Transcription levels of aveR and aveA1 in Dsig8 did not differ from those in WT on day 2, but were notably higher on day 6 (Figure 1E), consistent with the observed avermectin overproduction in Dsig8. These findings indicate that σ⁸ controls avermectin production at the transcription level by repressing ave genes, primarily in the late fermentation stage.

σ factors function as initiators of transcription. The negative regulatory effects of σ⁸ on aveR and aveA1 are therefore likely to be indirect. To test this idea, we performed EMSAs using refolded soluble recombinant His₆-σ⁸ and probes aveRp and aveA1p prepared by labeling promoter regions of aveR and aveA1. His₆-σ⁸ did not bind to the probes even at high concentration (1 μM) (Figure 2A), confirming that σ⁸ indirectly represses ave genes.

σ^B homologes typically initiate their own transcription (Hecker et al., 2007). The finding that σ⁸ indirectly represses ave genes suggested that σ⁸ initiates transcription of direct repressor(s) of aveR or aveA1. We recently characterized PhoP (Yang et al., 2015) and AvaR2 (Zhu et al., 2016) as direct repressors of aveR. We also found that SAV742 directly represses transcription of several ave structural genes including aveA1, but not aveR (Sun et al., 2016). To test the hypothesis that σ⁸ directly controls sig8 and regulatory genes phoP, avaR2 and sav_742, we performed EMSAs, qRT-PCR analysis and in vitro transcription assays.

EMSA results revealed that His₆-σ⁸ formed complexes with promoter regions of sig8 (probe sig8p) and sav_742 (probe sav_742p), but not with those of avaR2 (probe avaR2p) or phoR-phoP operon (probe phoRp) (Figure 2A). No shifted band was observed for negative probe control hrdB or protein control BSA. Binding specificity was confirmed by competitive assays using excess unlabeled specific probe (lanes S) and non-specific probe hrdB (lanes N). qRT-PCR analysis showed that transcription levels of sig8 and sav_742 on days 2 and 6 were lower in Dsig8 than in WT (Figure 2B), indicating that σ⁸ positively regulates expression of its own gene and adjacent gene sav_742. The lower sav_742 transcription level and higher avermectin yield in Dsig8 are consistent with SAV742’s function as a repressor of avermectin production (Sun et al., 2016).

In vitro transcription experiments were performed to determine whether σ⁸ initiates transcription of sig8 and sav_742. Linear DNA fragments harboring sig8, sav_742, and aveR promoter regions were used as templates, and various amounts of renatured His₆-σ⁸ were mixed with sufficient E. coli core-RNAP to reconstitute RNAP holoenzyme (Eσ⁸). Core-RNAP alone did not initiate transcription at the sig8 or sav_742 promoter, whereas the Eσ⁸ holoenzyme did (Figure 2C). When core-RNAP concentration was kept constant and σ⁸ concentration was increased, transcript levels of sig8 and sav_742 increased, but no aveR transcript was detected. These findings indicate that σ⁸ initiates transcription of its own gene and sav_742, but not that of aveR. The presence of sig8 and sav_742 transcripts in Dsig8 suggests that additional σ factors may recognize sig8 and sav_742 promoters and initiate their transcription, but with lower transcription efficiency.

Avermectin production in Dsig8 was very close to that in sig8sav_742 double deletion mutant Dsig8–742 (Supplementary Figure S3), but was higher than that in sav_742 deletion mutant D742 (Sun et al., 2016), suggesting that other σ⁸ targets may also affect avermectin production in Dsig8.

Alternative σ factors are usually involved in modulation of stress responses (Osterberg et al., 2011). To determine which type(s) of stress σ⁸ responds to, we performed stress tests on YMS plates. Relative to growth of WT, that of sig8 deletion mutant Dsig8 was more sensitive to NaCl, KCl, H₂O₂, diamide, and heat (42°C) stresses, but similarly sensitive to tert-butyl hydroperoxide (TBHP) and sucrose stresses (Figure 3A). These findings indicate that σ⁸ is required for WT levels of resistance to multiple stresses, e.g., heat stress, salt stress, and oxidative stress (from H₂O₂ or diamide).

The observations that sig8 mutant Dsig8 was more sensitive to various environmental stresses, and that σ⁸ was autoregulated, suggested that sig8 itself may be induced by these environmental stresses in a σ⁸-dependent manner. We tested this possibility by qRT-PCR comparison of sig8 transcription levels in WT and Dsig8 under stress conditions. Cells were cultured in soluble FM-II for 2 days, and then treated with heat (42°C), H₂O₂, diamide, or NaCl. RNA samples were isolated from cells before (0 min) and after treatment (10, 20, 30, 40, 60, 80, and 100 min). For WT, sig8 transcription level increased to a maximal value within 10–60 min for each stress type (∼13-fold for heat; ∼20-fold for H₂O₂, ∼17-fold for diamide; ∼54-fold for NaCl). For Dsig8, maximal induction was sharply decreased under each stress type (∼6-fold for heat; ∼2-fold for H₂O₂, ∼8-fold for diamide; ∼17-fold for NaCl) and delayed for heat (from 10 to 30 min) and salt (from 60 to 80 min) (Figure 3B). These findings indicate that σ⁸ itself responds to various environmental stresses at the transcription level in either a σ⁸-dependent or σ⁸-independent manner. The rapid and robust induction of sig8 in WT may be due to σ⁸-dependent control, whereas the decreased and/or delayed induction in Dsig8 may be due to σ⁸-independent control.

σ⁸ was strongly induced in response to a variety of stresses, suggesting that it mediates defensive systems against stresses. To identify σ⁸ target genes that respond to heat stress, we initially performed a series of EMSAs using His₆-σ⁸ and potential promoter probes of heat shock genes, including dnaK1p for dnaK1-grpE1-dnaJ1-hspR operon, dnaK2p for dnaK2-grpE2-dnaJ2 operon, groES1p for groES1-groEL1 operon, groEL2p, htpGp, hsp18_1p and hsp18_2p. His₆-σ⁸ bound specifically to dnaK1p, but not to other probes tested (Figure 4A). In vitro transcription analysis confirmed that dnaK1 transcription was initiated by σ⁸ (Figure 4B). dnaK1 transcription level in Dsig8 was downregulated at two time points (Figure 4C). These findings demonstrate that σ⁸ is a direct activator of dnaK1.

grpE1, dnaJ1, and hspR are cotranscribed with dnaK1, and therefore are also σ⁸ targets. dnaK1, grpE1, and dnaJ1 encode molecular chaperones for interaction with denatured proteins and facilitate refolding to the native state following heat stress, and hspR encodes a putative transcriptional repressor of its operon. Transcription levels of dnaK1 in WT and Dsig8 under heat stress were recorded to determine whether dnaK1p for its operon was induced in a σ⁸-dependent manner. In WT, dnaK1 transcription level increased rapidly to ∼45-fold after 10 min exposure to 42°C, and then gradually declined. In Dsig8, maximal induction was reduced to ∼35-fold and delayed after 30 min (Figure 4D). These findings indicate that σ⁸ facilitates rapid adaption to heat stress mainly by direct regulation of the dnaK1-grpE1-dnaJ1-hspR operon. The slower, lower-degree induction of dnaK1p in Dsig8 presumably depends on other factors not addressed here.

σ⁸ responds to H₂O₂ and diamide, which respectively cause peroxidative and thiol-oxidative stress. Bacteria often respond to H₂O₂ stress by producing catalases and peroxidases that degrade H₂O₂. We showed that responses to H₂O₂ stress in S. avermitilis involve three catalase genes (katA1, katA2, katA3), the ahpCD operon (encoding alkyl hydroperoxide reductase and alkylhydroperoxidase), and two peroxide-sensing transcriptional factor genes (oxyR, catR) (Liu et al., 2016). In S. coelicolor, ECF-σ^R plays a key role in response to thiol-oxidative stress, and major σ^R targets include trx genes (encoding the thioredoxin system) and msh genes (for biosynthesis of mycothiol, the major thiol buffer) (Kallifidas et al., 2010). S. avermitilis contains one sigR homologous gene (sig22), six thioredoxin genes (trxA1, trxA2, trxA3, trxA4, trxA5, trxA6), two thioredoxin reductase genes (trxB1, trxB2), and four msh genes (mshA, mshB, mshC, mshD). Possible interactions of σ⁸ with these oxidative stress-related genes were investigated by EMSAs. His₆-σ⁸ bound specifically to the bidirectional promoter probes catR_katA1_int and oxyR_ahpCD_int, and to promoter probes trxA3p, trxB2p, mshAp, mshCp, mshDp and sig22p, but not to probes katA2p, katA3p, trxA1p, trxA2p, trxA4p, sav_2810p for sav_2810-2809-trxA5-2807-cyp12 operon, trxA6p, trxB1p, or mshBp (Figure 5A). In vitro transcription analysis showed that σ⁸ initiated transcription of catR, oxyR, trxA3, trxB2, mshA, mshC, mshD and sig22, but not that of katA1 or ahpC (Figure 5B). Transcription levels of catR, oxyR, trxA3, trxB2, mshA, mshC, mshD, and sig22 were all reduced in Dsig8 relative to WT on day 6, or days 2 and 6 (Figure 5C). These findings indicate that catR, oxyR, trxA3, trxB2, mshA, mshC, mshD, and sig22 are directly activated by σ⁸.

For analysis of oxidative stress responses, WT and Dsig8 were treated with 1 mM H₂O₂ or diamide for various durations. In WT, catR and oxyR were induced to maximal transcription level (∼20- and ∼23-fold, respectively) by H₂O₂ within 40 min (Figure 5D). In Dsig8, H₂O₂ treatment caused only a slight increase of transcription level (∼4-fold induction) of these two genes. These findings indicate that σ⁸ mediates H₂O₂ induction of the peroxide-sensing regulators CatR and OxyR. Diamide treatment of WT caused notable induction of trxA3 (∼43-fold), mshA (∼17-fold) and sig22 (∼17-fold) within 20 min, and slight induction of mshC (∼2.2-fold), mshD (∼3-fold) and trxB2 (∼1.8-fold) within 40 min, whereas diamide treatment of Dsig8 had much lower (for trxA3, mshA, mshC, mshD, and sig22) or no effect (for trxB2) on expression of these genes (Figure 5E). These findings indicate that σ⁸ helps S. avermitilis to cope with thiol-oxidative stress by activating transcription of its target genes trxA3, trxB2, mshA, mshC, mshD, and sig22.

The σ⁸ homolog, S. coelicolor σ^B, responds to osmotic stress caused by NaCl and KCl. To investigate σ⁸ target genes involved in osmoprotective functions, a series of candidates were selected for EMSA analysis: ect genes (encoding enzymes for biosynthesis of ectoine) (Bursy et al., 2008), opu genes (encoding ABC transporters for uptake of osmoprotectants) (Horn et al., 2005), osmC (encoding a putative osmotically inducible protein) (Gutierrez and Devedjian, 1991), osaAB (encoding a two-component system) and katB (encoding a catalase). osaAB and catB (katB homolog) were reported to be essential for osmoadaptation in S. coelicolor (Cho et al., 2000; Bishop et al., 2004). EMSAs revealed specific binding of His₆-σ⁸ to the promoter probes ectAp for ectA-ectB-ectC-ectD operon, opuBA1p, opuBC1p, sav_5148p for sav_5148-opuBB2-opuBA2-opuBB1-opuBC2-sav_5143 operon, osaAp, osaBp, and katBp, but not to those of opuAAp for opuAA-opuAB operon or osmCp (Figure 6A). In vitro transcription assay confirmed initiation of transcription of ectA, opuBA1, opuBC1, sav_5148, osaA, osaB, and katB by σ⁸ (Figure 6B). Transcription levels of seven newly identified σ⁸ target genes were all lower in Dsig8 than in WT (Figure 6C), demonstrating the role of σ⁸ as an activator of these genes. ectB, ectC, and ectD are cotranscribed with ectA, and opuBB2, opuBA2, opuBB1, and opuBC2 are cotranscribed with sav_5148; therefore, these are also target genes of σ⁸.

σ⁸ may also affect induction of its target osmoprotection-related genes by salt stress in S. avermitilis. ectA transcription in WT was sharply increased to ∼490-fold by 80 min of NaCl treatment (Figure 6D), suggesting that ectoine plays as a key osmoprotective role. Maximal induction by NaCl treatment in WT was ∼3.5-fold to 8.7-fold for opuBA1, opuBC1, opuBB2 (selected for qRT-PCR analysis for its operon), osaA, osaB, and katB (Figure 6D). In Dsig8, NaCl treatment had no effect on transcription of opuBC1, osaA, osaB, or katB, indicating that salt stress induces these genes in a σ⁸-dependent manner. Maximal induction by NaCl treatment on ectA, opuBA1, and opuBB2 transcription was lower in Dsig8 than in WT, suggesting that these three genes are induced by salt stress in both σ⁸-dependent and σ⁸-independent manners.

σ⁸ in S. avermitilis evidently plays a pleiotropic role in control of avermectin production and in protection against a variety of stresses. To clarify broader roles of σ⁸ in this species, more σ⁸ target genes need to be identified. The recognition and binding sites of σ⁷⁰-family factors are the -35 and -10 hexamers of its target promoters. We conducted 5′-RACE assays to determine promoter structures of several σ⁸ targets and identified a consensus σ⁸-binding sequence. TSSs of sig8, dnaK1, oxyR, trxA3, sig22, and opuBC1 were determined by 5′-RACE analysis of WT gene transcripts under stress conditions (Supplementary Figure S4), leading to the putative -35 and -10 promoter sequences shown in Figure 7. Analysis of these promoter sequences using the PREDetector web-based program¹ revealed a consensus sequence BGVNVH-N₁₅-GSNNHH (Figure 7), which resembles that of σ^B-specific promoters of S. coelicolor (GNNTN-N_14-16-GGGTAY) (Y: C or T) (Lee et al., 2004b).

RNA polymerase is not sensitive to variations in spacer length between the -35 and -10 regions, and the consensus σ^B-binding sequence is GNNTN-N_14-16-GGGTAY. We therefore employed the sequence pattern BGVNVH-N_14-16-GSNNHH to scan the S. avermitilis genome using the PREDetector program to predict candidate members of the σ⁸ regulon, and selected putative promoters that were located within 300 nt upstream of the translational start codons. Using a cut-off score of ≥6.5, we identified 940 putative σ⁸ target genes, of which 453 have unknown function or are unclassified (Supplementary Table S2). The remaining 487 genes were assigned to 17 functional groups. Among these 487 genes, 179 are associated with regulatory functions, according to the KEGG pathways database for S. avermitilis². These findings suggest the extent biological significance of σ⁸ in S. avermitilis.