Regulation of Antimycin Biosynthesis Is Controlled by the ClpXP Protease

Natural products produced by Streptomyces species underpin many industrially and medically important compounds. However, the majority of the ∼30 biosynthetic pathways harbored by an average species are not expressed in the laboratory. This unrevealed biochemical diversity is believed to comprise an untapped resource for natural product drug discovery. Major roadblocks preventing the exploitation of unexpressed biosynthetic pathways are a lack of insight into their regulation and limited technology for activating their expression. Our findings reveal that the abundance of σAntA, which is the cluster-situated regulator of antimycin biosynthesis, is controlled by the ClpXP protease. These data link proteolysis to the regulation of natural product biosynthesis for the first time to our knowledge, and we anticipate that this will emerge as a major strategy by which actinobacteria regulate production of their natural products. Further study of this process will advance understanding of how expression of secondary metabolism is controlled and will aid pursuit of activating unexpressed biosynthetic pathways.

and are not dedicated regulators of one biosynthetic pathway. Indeed, cluster-situated ECF factors have been observed previously only in the biosynthesis of lantibiotics produced by so-called rare actinomycetes. In Microbispora corallina, MibR and MibX regulate microbisporicin biosynthesis and MibX is controlled by the anti-factor MibW (18); in Planomonospora alba, PspR and PspX regulate planosporicin production and PspX is controlled by the anti-factor PspW (19). Interestingly, unlike the canonical ECF factors MibX and PspX , whose activities are controlled by cognate anti-factors, AntA lacks an identifiable anti-factor partner and as a consequence has created curiosity about how its activity is controlled.
The Clp-protease system is essential for normal bacterial proteostasis and is best characterized in Escherichia coli (20,21). The Clp protease is a multienzyme complex composed of a barrel-shaped peptidase, ClpP, and a regulatory enzyme, either ClpA or ClpX (or ClpC in some organisms). ClpA and ClpX (and ClpC) are AAAϩ-family protein unfoldases that recognize an N-terminal and/or C-terminal recognition signal (degron) and utilize ATP to unfold and translocate proteins to the peptidase chamber, where they are degraded into short peptides (22). In Streptomyces species, the peptidase is specified by two genes instead of one and is redundantly encoded (23). The primary peptidase is encoded by clpP1P2, whose corresponding proteins form a complex with ClpX or ClpA to facilitate normal proteostasis; the second peptidase is encoded by clpP3P4, but its expression occurs only when the primary system is compromised (24,25). The best-understood degron is the SsrA tag from E. coli (AANDENYALAA), which is added cotranslationally to polypeptides stalled on ribosomes (26,27). The E. coli SsrA tag has been comprehensively studied, and Ala-Ala-COO-, at the C-terminal region of this motif, is essential for proteolysis by ClpXP (28). Intriguingly, the C terminus of AntA harbors the sequence Ala-Ala-COO-, which previously led us to speculate that ClpXP may modulate its level/activity (9).
Here, we reconstituted ClpXP proteolysis of AntA in vitro and showed that it is dependent upon the C-terminal Ala-Ala. We also found that the abundance of AntA in vivo was higher when Ala-Ala was changed to Asp-Asp and that the abundance of AntA was elevated in the absence of genes encoding the primary peptidase, ClpP, and its unfoldase, ClpX. These data establish direct proteolysis as an alternative, and thus far unique, control strategy of ECF RNA polymerase factors, expanding the paradigmatic understanding of microbial signal transduction regulation.

RESULTS AND DISCUSSION
AntA orthologues are a new subfamily of ECF factors that regulate production of the antimycin biosynthetic starter unit. Since its initial discovery 6 years ago, more than 70 ant BGCs have been identified within the members of Actinobacteria, including Actinospica, Saccharopolyspora, Streptacidiphilus, and Streptomyces (5). Each of these BGCs harbors a single regulator, AntA (53% to 100% shared amino acid identity across all orthologues), which lacks a cognate anti-factor partner (5,9). Our previous work with S. albus S4 established that AntA orthologues comprise a new subfamily of ECF factors (9,29). We demonstrated that AntA is required for expression of antGF and antHIJKLMNO, which encode a standalone ketoreductase (AntM) and proteins required for the production/activation of the starter unit, 3-formamidosalicylate (3-FSA) (Fig. 1). We also mapped the transcriptional start sites and identified conserved promoter sequences for these operons in all ant BGCs known at the time (9). The conservation of AntA and target promoters within ant BGCs from taxonomically diverse species suggests that AntA -mediated regulation of these genes is direct. To verify this hypothesis, we performed chromatin immunoprecipitation sequencing (ChIP-seq) with a S. albus S4 ΔantA mutant complemented with a version of AntA with a 3xFLAG tag at its N terminus, which we demonstrated restored antimycin production (see Fig. S2 in the supplemental material). Analysis of the resulting data revealed only one ChIP-seq peak across the whole chromosome for which the number of mapped reads was enriched for both biological replicates of ΔantA/3xFLAG-antA compared to that of the wild-type mock-immunoprecipitated control. This region corresponded to the inter-genic space (297 bp) between antG and antH, which upon inspection revealed a prominent peak for the closely spaced and divergent AntA -target promoters antGp and antHp and a second, albeit smaller peak corresponding to the 5= end of the antH coding sequence (Fig. 2). Taken together, these data are consistent with the hypothesis that AntA is a cluster-situated regulator that directly activates expression of genes for the production of 3-FSA during antimycin biosynthesis.
AntA is degraded by the ClpXP protease in vitro. The activities of almost all characterized ECF factors are modulated by a cognate anti-factor, which is typically a protein coencoded within the same operon. Intriguingly, AntA lacks an anti-factor and is therefore an orphan, indicating that a unique mechanism is likely at work to control AntA activity. An inspection of AntA amino acid sequences revealed a C-terminal Ala-Ala in 67 of the 71 orthologues (Fig. S1). A C-terminal Ala-Ala is an important component of a common class of degrons for the ClpXP protease (28). This observation led us to hypothesize that the activity of AntA might be modulated by proteolysis instead of by an anti-factor. To test this hypothesis, we performed in vitro proteolysis. Previous work indicated that S. albus S4 AntA was insoluble when overproduced by E. coli, so we pursued the overproduction and purification of the orthologue from Streptomyces ambofaciens ATCC 23877, which has been experimentally demonstrated to be a producer of antimycins (30). S. ambofaciens AntA (75% amino acid identity with S. albus S4 AntA , including 13 of 15 amino acid residues at the C terminus) was purified as an N-terminal (His) 6 -SUMO-fusion protein. The (His) 6 -SUMO tag increases solubility and eases purification of putative substrates, without altering recognition of C-terminal degrons by ClpXP. ClpX orthologues from E. coli and S. ambofaciens possess 60% shared amino acid identity and therefore likely recognize similar substrates for degradation. Thus, ClpXP from E. coli was purified (Fig. S2) and its ability to degrade (His) 6 -SUMO-AntA was assessed. Degradation of (His) 6 -SUMO-AntA was apparent as early as 2.5 min after addition of ATP, and all of the sample was degraded by 15 min (Fig. 3). Substrates of ClpXP become resistant to proteolysis by specific alterations of the C-terminal Ala-Ala (28). Therefore, to investigate degradation specificity in the experiment described above, we constructed and tested a variant of S. ambofaciens AntA in which the C-terminal Ala-Ala was mutated to Asp-Asp [(His) 6 -SUMO-AntA-DD ]. Strikingly, the Asp-Asp variant was stable against ClpXP degradation over the lifetime of the assay (Fig. 3). Thus, the degradation of (His) 6 -SUMO-AntA and the characteristic resistance afforded by the Ala-Ala-to-Asp-Asp mutation demonstrated that AntA is a substrate of ClpXP in vitro.
AntA is degraded by the ClpXP protease in vivo. To investigate if the in vitro degradation of AntA demonstrated above is relevant to its regulation in vivo, we deleted the operon consisting of the clpX, clpP1, and clpP2 genes from S. albus S4. The resulting ΔclpXclpP1clpP2 mutant still harbored the second Clp peptidase encoded by clpP3clpP4 and underwent a normal developmental cycle, albeit sporulation was less robust, which is consistent with growth characteristics reported for mutation of equivalent genes in S. coelicolor ( Fig. S3) (31). Next, genes encoding the 3xFLAG-AntA or 3xFLAG-AntA-DD fusion proteins were generated and introduced into the parental strain and the ⌬clpXclpP1clpP2 mutant so the abundance of these proteins could be assessed over a developmental time course by Western blotting with anti-FLAG antisera. This experiment was initially performed with the AntA fusions integrated on the chromosome under the control of the native promoter. However, a reliable signal could not be detected for 3xFLAG-AntA and only a trace amount of the Asp-Asp variant was observed, presumably indicating that the cellular level of AntA is normally low because the native promoter is relatively weak. The experiment was therefore repeated with 3xFLAG-AntA and 3xFLAG-AntA-DD expression driven by a stronger, constitutive promoter, ermE* (32). Analysis of the resulting immunoblot revealed that 3xFLAG-AntA-DD was more abundant than 3xFLAG-AntA in vegetative mycelium of the parent and ΔclpXclpP1clpP2 strains ( Fig. 4; see also Fig. S5). Strikingly, in the later stages of development after aerial mycelium had formed (24 h and 30 h), 3xFLAG-AntA and 3xFLAG-AntA-DD were detected only in the ΔclpXclpP1clpP2 strain and not the parent; the Asp-Asp variant was also present in greater relative abundance (Fig. 4), which was consistent with our previous experiments that showed that the ant BGC is downregulated at the level of transcription upon the onset of aerial growth (9). Interestingly, the conspicuous absence of 3xFLAG-AntA and the presence of 3xFLAG-AntA-DD in protein samples prepared from the latest time point sampled suggest the potential involvement of an additional degradative factor(s). Taken together, these data support the hypothesis that the levels of AntA , and thus its ability to activate gene expression of antFGHIJKLMNO, are modulated by the ClpXP protease.
Antimycins are not overproduced in the absence of ClpXP. The results of the experiments described above indicate that the cellular level of AntA was more abundant in the absence of the ClpXP protease. In order to determine if an increased level of this transcription factor ultimately influenced the final production titer of antimycins, we used liquid chromatography-high-resolution mass spectrometry (LC-HRMS) to assess the abundance of antimycins in chemical extracts generated from the ΔclpXclpP1clpP2 and parental strains grown atop a cellophane disk on mannitol-soya flour (MS) agar in triplicate. The extracted ion chromatograms representing antimycin A 1 , A 2 , A 3 , and A 4 were used to determine the peak area for each compound, which was subsequently normalized based on the wet mycelium weight of the sample. Interestingly, the results indicated that the total levels of antimycin production by the ΔclpXclpP1clpP2 mutant (15.57 arbitrary units [AU] Ϯ 2.86) and the parental strain (16.59 AU Ϯ 1.12) were not statistically significantly different (P value, 0.59) (see Table S1 in the supplemental material). This result is consistent with a previous experiment where overexpression of antA did not increase the titer of antimycins, because that experiment showed that it resulted in overexpression of only antGF and antHI-JKLMNO (genes encoding the production of the AntG-S-3-formamidosalicylate starter unit) and not the remaining genes (antABCDE) in the BGC (9). This also presumably indicates that starter unit biosynthesis is not rate limiting for antimycin production.
Model for the regulation of antimycin biosynthesis. Our model for the regulation of antimycin biosynthesis is depicted in Fig. 5. Expression of the ant BGC is crossactivated by FscRI, a LuxR-family regulator, from the candicidin BGC, which activates expression of antBA and antCDE (10). This regulation in turn enables direct activation of the 3-FSA biosynthetic operons (antGF and antHIJKLMNO) by AntA . The expression of antBA and antCDE is downregulated following the onset of morphological differentiation, presumably because the ligand sensed by the FscRI PAS domain is no longer available (9, 10). The cellular level of AntA is antagonized by the ClpXP protease, for which it is a direct target, and is ultimately responsible for clearing residual AntA when FscRI is inactivated following the onset of morphological differentiation (10). While ClpXP proteolytic control of transcription factor activity, and in particular that of ECF factor/anti-factors, has been shown previously (33)(34)(35)(36)(37)(38)(39)(40)(41), it has thus far not been directly linked to the control of cluster-situated regulators of natural product biosynthesis. This finding provides a new lens through which to examine microbial signal transduction and the regulation of natural product biosynthesis in Streptomyces spe- Thirty-microgram volumes of total protein were analyzed by Western blotting with anti-FLAG antisera. The images shown are derived from uncropped original images shown in Fig. S4 in the supplemental material, and the corresponding densitometry analysis is shown in Fig. S6. cies. Understanding the diversity of regulatory strategies controlling the expression of these pathways is critical for the development of new tools for exploiting the "silent majority" of biosynthetic pathways harbored by these organisms.

MATERIALS AND METHODS
Growth media, strains, cosmids, plasmids, and other reagents. Escherichia coli strains were propagated on Lennox agar (LA) or broth (LB) (42,43), and Streptomyces albus S4 strains were cultivated using LA, LB, and mannitol-soya flour (MS) agar or broth (42). Development of clp mutants was assessed on MS and ISP2 medium (42). Culture medium was supplemented with antibiotics as required at the following concentrations: apramycin, 50 g ml Ϫ1 ; carbenicillin, 100 g ml Ϫ1 ; chloramphenicol, 25 g ml Ϫ1 ; hygromycin, 50 g ml Ϫ1 ; kanamycin, 50 g ml Ϫ1 ; nalidixic acid, 25 g ml Ϫ1 . Streptomyces strains were constructed by conjugal mating with E. coli ET12567 as previously described (42). Enzymes were purchased from New England Biolabs unless otherwise stated, and oligonucleotides were purchased from Integrated DNA Technologies, Inc. All of the strains, cosmids, and plasmids used in this study are described in Table S2 in the supplemental material, and all of the oligonucleotides used are provided in Table S3.
Construction of plasmids. The insertion for each plasmid generated in this study was prepared by PCR amplification with Q5 High-Fidelity DNA polymerase and oligonucleotides containing restriction sites. PCR-amplified insertions were restricted and cloned into the relevant plasmids cut with the same enzymes by standard molecular biology procedures. All clones were sequenced to verify the integrity of insertion DNA. The names of the restriction sites used for cloning are provided with the plasmid descriptions in Table S2.
ChIP-seq and bioinformatics analyses. The antA coding sequence was amplified with RFS629 and RFS630, which contain KpnI and EcoRI restriction sites, respectively. The restricted PCR product was cloned into pSETNFLAG (10) digested with the same enzymes. The resulting plasmid was then restricted with NotI and EcoRI to release ermE*p-3xFLAG-antA, which was subsequently cloned into pAU3-45 (44) digested with the same enzymes. pAU3-45-3xFLAG-antA was mobilized to an apramycin-marked ΔantA strain (9). Cultivations of the wild-type and ΔantA/pAUNFLAG-antA strains for ChIP-seq were performed exactly as described previously (10). Pure DNA resulting from immunoprecipitates from two biological The cellular level of AntA is antagonized by the ClpXP-protease system, for which it is a direct target and is ultimately responsible for clearing residual AntA when FscRI is inactivated following the onset of differentiation.
replicates of the wild-type and ΔantA/pAUNFLAG-antA strains and nonimmunoprecipitated chromosomal DNA were sequenced with a Illumina HiSeq 3000 platform with 150-nucleotide (nt) paired-end reads by the University of Leeds Next Generation Sequencing Facility at the St. James Teaching Hospital NHS Trust. The resulting reads were analyzed exactly as described previously (10). The graphic in Fig. 2 was generated using DeepTools computeMatrix in scale-regions mode with a bin size of 54 and plotProfile functions (45).
Construction of the S. albus S4 ⌬clpXclpP1clpP2 mutant strain. Deletion of clpXclpP1clpP2 was carried out using RecET recombineering in E. coli as follows. The clpXclpP1clpP2-containing cosmid, cos117, was obtained by screening a previously constructed S. albus S4 Supercos1 cosmid library (8) by PCR using oligonucleotides PBB001 and PBB002. Cos117 was mutagenized as required using E. coli recombineering with strain GB05-red (46) and a deletion cassette. The deletion cassette was generated by PCR from paac-apr-oriT (47) and consisted of the aac(3)IV apramycin resistance gene and a conjugal origin of transfer (oriT), which was flanked by ⌽C31-attL and ⌽C31-attR sites for excision of the cassette. The oligonucleotides used to generate deletion cassettes included 39 nt of homology upstream or downstream of the target open reading frame(s) and are listed in Table S3. The resulting PCR product was digested with DpnI, gel purified, and electroporated into arabinose-induced E. coli GB05-red harboring cos117. Transformants were screened for the presence of mutagenized cosmid by PCR using oligonucleotides listed in Table S3, and the integrity of the locus was verified by DNA sequencing. The mutagenized cosmid was electroporated into E. coli ET12567/pUZ8002 and mobilized to a strain of S. albus S4 harboring an entire antimycin BGC deletion (Δantall) (48) by conjugation as described previously (42). Transconjugants were screened for apramycin resistance and kanamycin sensitivity. The integrity of an apramycin-marked mutant was verified by PCR using the oligonucleotides listed in Table S3. The apramycin deletion cassette was subsequently excised from the chromosome by conjugal introduction of pUWLint31, which is a replicative plasmid with a temperature-sensitive origin of replication that expresses the ⌽C31 integrase required for removal of the cassette (47). Transconjugants were screened for loss of apramycin resistance, and excision of the cassette was verified by polymorphic shift PCR and DNA sequencing of the product.
Immunoblot analysis. Spores of the parental strain and of S. albus Δantall and the ΔclpXclpP1clpP2 mutant harboring pPDA or pPDD were grown on MS agar (buffered with 50 mM TES, pH 7.2) covered with cellophane discs. Protein was isolated from mycelium collected during growth at the following regular intervals: 14 h, 17 h, 24 h, and 30 h for the Δantall and ΔclpXclpP1clpP2 mutants harboring 3xFLAG-AntA constructs and 17 h, 20 h, 23 h, and 30 h for the Δantall and ΔclpXclpP1clpP2 mutants harboring the 3xFLAG-FscRI construct. Protein samples were generated as follows: 100-mg volumes of cells were resuspended in 200 l lysis buffer (50 mM sodium phosphate buffer [pH 7.0]; 150 mM sodium chloride; 10 mg ml Ϫ1 lysozyme; cOmplete, Mini, EDTA-free protease inhibitors [Roche]; 100 mg of 0.1-mm-diameter glass beads [PowerLyzer]) and lysed by vortex mixing for 30 min at 2,000 rpm and 37°C, with a subsequent incubation for another 30 min at 37°C. The obtained suspension was centrifuged for 20 min at 20,000 ϫ g at 18°C. Each clarified protein sample (30 g) was subjected to SDS-PAGE and then transferred to a nitrocellulose membrane (pore size, 0.2 m) for Western blot analysis. The membrane was probed with mouse monoclonal anti-FLAG M2-horseradish peroxidase (HRP) antibody (Sigma) (1:10,000, and the signals were detected using Pierce 1-Step Ultra TMB blotting solution (Thermo Scientific).
In vitro ClpXP proteolysis assays were performed at 30°C by preincubating 0.3 M ClpX 6 and 0.8 M ClpP 14 with an ATP regeneration system (4 mM ATP, 50 g ml Ϫ1 creatine kinase, 5 mM creatine phosphate) in 25 mM HEPES-KOH (pH 7.5)-20 mM KCl-5 mM MgCl 2 -10% glycerol-0.032% NP-40 -0.2 mM DTT and adding substrate to initiate the reactions. Samples of each reaction mixture were taken at specific time points, and the reactions were stopped by addition of SDS-PAGE loading dye and boiling at 100°C before loading on Tris-Glycine-SDS gels. Bands were visualized by staining with colloidal Coomassie G-250 and quantified by ImageQuant (GE Healthcare) after scanning by Typhoon FLA 9500 (GE Healthcare). The (His) 6 -SUMO-AntA fraction remaining was calculated by dividing the (His) 6 -SUMO-AntA density at a given time point by the density at time zero and normalizing by ClpX density. Chemical analysis. S. albus S4 strains were cultivated atop a cellophane disc on MS agar at 30°C for 7 days in triplicate. At the time of harvest, the cellophane disc containing mycelium was removed and the quantity of biomass was determined. Bacterial metabolites were extracted from both the mycelium and the "spent" agar for 1 h using 50 ml of ethyl acetate. Thirty-milliliter volumes of ethyl acetate were evaporated to dryness under conditions of reduced pressure, and the resulting residue was resuspended in 100% methanol (300 l). Immediately prior to LC-HRMS analysis, methanolic extracts were centrifuged at 16,000 ϫ g in a microcentrifuge tube for 5 min to remove insoluble material. Only the supernatant (3 l) was injected into a Bruker Maxis Impact time of flight (TOF) mass spectrometer equipped with a Dionext Ultimate 3000 HPLC system as previously described (52). The peak area associated with the extracted ion chromatograms for antimycin A 1 , A 2 , A 3 , and A 4 present in agar and mycelium extracts was determined and used to calculate the total level of antimycins produced for each replicate. These values were subsequently used to determine the arithmetic mean for total antimycin production for each strain. Statistical significance was assessed in Microsoft Excel by a homoscedastic Student's t test with a two-tailed distribution.
Data availability. The next-generation-sequencing data obtained in this study are available under ArrayExpress accession numbers E-MTAB-7700 and E-MTAB-5122.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.