Strains used in this study are listed in Supplementary Table S1.

Escherichia coli DH5α and HB101 were used for routine transformation and subcloning. The E. coli strains were grown routinely in Luria Broth (LB) medium (10 g/L peptone, 5 g/L yeast extract, and 5 g/L NaCl, pH 7.2). M. xanthus strains were grown in CYE medium [10 g/L casitone, 5 g/L yeast extract, 10 mM 3-(N-morpholino) propanesulfonic acid (MOPS; pH 7.6), and 4 mM MgSO₄] for routine growth and CMO medium (CYE medium plus 7 mL/L methyl oleate) for the production of epothilones. When appropriate, kanamycin (40 μg/mL), tetracycline (10 μg/mL), and apramycin (25 μg/mL) were added. The growth temperatures were 37°C for E. coli and 30°C for M. xanthus.

Total RNA was extracted from cell cultures using BIOZOL kits (Total RNA Extraction Reagent, BioFast, China) and then transcribed reversely into cDNA with PrimeScript™ reagent kit with DNAase (Takara, Japan). Primers were designed at the junction between two genes (Supplementary Table S2). Total RNA without inverse transcription and the cells containing the epothilone gene cluster were used as templates in the negative and positive control groups, respectively. In experimental groups, the cDNAs were used as templates.

We used the promoter prediction software “Neural Network Promoter Prediction” (https://www.fruitfly.org/seq_tools/promoter.html) to predict internal promoters in the epothilone gene cluster derived from Sorangium cellulosum strain So0157-2. The threshold was set as 0.8. At the same time, another online promoter prediction software “BPROM” (http://www.softberry.com/berry.phtml) was also employed in the prediction. The −35 and −10 binding regions were predicted in comparison to the σ70 consensus -35 (TTGACA) and -10 (TATAAT) promoter regions of E. coli. The threshold was set greater than 0.8.

The primers and plasmids used in this study are provided in Supplementary Tables S2 and S3.

The plasmids pkk-232-P_epoP, pkk-232-P_epoB ∼ pkk-232-P_epoF were used as a promoter activity reporter vector in E. coli (Supplementary Figure S1). The promoters P _epoP , P _epoB ∼ P _epoF were obtained by PCR with primers P_epoP-F/R, P_epoB-F/R ∼ P_epoF-F/R from the epothilone gene cluster in So0157-2. The plasmids pkk-232-P_epoP, pkk-232-P_epoB ∼ pkk-232-P_epoF were constructed by inserting P _epoP , P _epoB ∼ P _epoF into the HindⅢ/BamHⅠ sites of plasmid pkk-232-8. The above plasmids were used to detect the activity of internal promoters from the epothilone operon in E. coli. The plasmid pkk-232-aphⅡ was constructed by inserting the promoter aphⅡ into the HindⅢ/BamHⅠ sites of pkk-232-8 and employed as the positive control. The promoter activity was characterized by detecting the activity of the report gene chloramphenicol acetyltransferase (CAT). The activity of the reporter gene CAT was detected by using the CAT ELISA kit. The CAT ELISA kit was purchased from Roche and operated according to the instructions provided.

In our previous work (Peng et al., 2018), we constructed the plasmid pSWcuomxdCas9-ω (Supplementary Figure S2). It contains codon-optimized mxdcas9 [D10A (GAC→GGC) and H840A (CAC→GCC)], linked to the EcoRI/HindⅢ site of the pSWU30 plasmid. We fused the Omega (ω) protein gene from M. xanthus DK1622 to the 3′-end of mxdcas9 with a 21-nt linker (AAG​CTT​TCT​GGA​TCA​AGT​TCT). In front of mxdcas9 is the promoter P _cuoA from M. xanthus DK1622, which is inducible by the addition of cupric ions (200 μM).

The plasmid pZJY41-sgRNA was used to express sgRNA (Supplementary Figure S3). To construct the pZJY41-sgRNA series plasmids, we designed spacers targeting different internal promoters with the online software “CasOT” (Xie et al., 2015) and employed the spacer with the lowest off-target efficiency for each promoter. The spacer sequences were added to the forward primer of each sgRNA using the p41sg (Zhao et al., 2008; Peng et al., 2018) plasmid as template; the linear plasmid containing different sgRNA fragments was amplified with primers sgRNA-P-F, sgRNA-B-F ∼ sgRNA-F-F, and sgRNA-R and then ligated with T4 DNA ligase with themselves, resulting pZJY41-sgRNA-P, pZJY41-sgRNA-B ∼ pZJY41-sgRNA-F plasmids. We obtained the sequence PpilA-spacerA-sgRNA by PCR from the plasmid pZJY41-sgRNA-A with the primer NdeI-F/R and then digested and connected it to the NdeI site of plasmid pZJY41-sgRNA-P to obtain two simultaneous expressions of sgRNA plasmid pZJY41-sgRNA-AP. Plasmids pZJY41-sgRNA-AB, pZJY41-sgRNA-APB, and pZJY41-sgRNA-DEF were constructed in the same way with different restriction sites.

We transformed plasmid pSWcuomxdCas9-ω into E. coli competent HB101 cells and obtained the pkk-CuOm strain. Then we transformed two plasmids into E. coli pkk-CuOm at the same time: pZJY41-sgRNA series plasmid carrying sgRNA with different spacer sequences and pKK232-Pepo series plasmid carrying an internal promoter and reporter gene CAT. These three plasmids all replicate autonomously in E. coli. Similarly, we constructed CRISPRa-dCas9 in M. xanthus by introducing plasmids pSWcuomxdCas9-ω and pZJY41-sgRNA. Notably, pSWcuomxdCas9-ω in M. xanthus was inserted at the attB site of the genome, while the plasmid pZJY41-sgRNA replicated autonomously.

ZE9 and mutants were grown overnight in 50 ml of CYE medium supplemented with Apra (25 μg/mL). The cultures were inoculated at a ratio of 2:100 into 50 mL of CMO medium containing 2% of XAD-16 resin. The resin was harvested with a strainer after 6 days and extracted with 3 mL of methanol by shaking at room temperature overnight. The supernatant was centrifuged for 10 min at 12,000 rpm and filtered with a 0.22-μm filter to remove the impurities. Twenty microliter of the sample was injected into a high-performance liquid chromatography (HPLC, SHIMADZU, Japan) system, analyzed on a Shim-pack MRC-ODS RP C18 column (4.6 mm × 250 mm, 4.60 μm; SHIMADZU, Japan), and monitored at 250 nm, with a mobile phase of 60% of methanol (HPLC grade) and 40% of H₂O at a flow rate of 1.0 mL/min.

The major epothilone products are epothilones A and B. The structural difference between the two compounds is that there is an extra methyl on 12C in epothilone B (Supplementary Figure S4). Epothilones A and B are suggested to be produced by the nonspecific receiving of malonyl-CoA (producing epothilone A) or methyl-malonyl-CoA (producing epothilone B) by the second acyltransferase domain in EPOSC (Gerth et al., 2001). The yields of epothilones were quantified from the peak area in the UV chromatogram, by reference against a calibration standard. According to the elution time of standard samples, the peak of epothilone A appears at 12.5 min and that of epothilone B at 15 min.

We collected samples continuously from the fermentation culture after 48 h of incubation. Then, total RNA of the samples was extracted using BIOZOL kits (Total RNA Extraction Reagent, BioFast, China) and then transcribed reversely into cDNA with the PrimeScript™ reagent kit with DNAase (Takara, Japan). The gapA gene (glyceraldehyde-3-phosphate dehydrogenase gene, MXAN_2815) was chosen as the reference gene for normalization. The transcriptional level of the epothilone gene cluster was analyzed by RT-qPCR on the LightCycler®480 system (Switzerland) with SYBR® Premix Ex Taq™ GC dye (Takara, Japan). All the primers used in RT-qPCR are listed in Supplementary Table S2.

The epothilone biosynthetic gene cluster from S. cellulosum So0157-2 consists of seven ORFs transcribed in the same direction. These ORFs are shortly overlapped or separated with a short distance (15–147 bp), and no terminator structure was found in the intergenic regions (Figure 1A). The RT-PCR results showed that all the adjacent ORFs were co-transcribed (Supplementary Figure S5). Thus, the whole gene cluster is a huge operon. However, the transcriptional levels of the operon ORFs varied significantly. For example, M. xanthus ZE9 is an epothilone producer constructed with the whole epothilone gene cluster from S. cellulosum So0157-2 (Zhu et al., 2015). In ZE9, epoP, epoB, or epoC was transcribed about a quarter of the level of the first gene epoA, while the epoD transcription was 2-fold higher than that of the upstream epoC, and the last epoF had the highest expression level among all the operon genes, approximately 2.5 times that of epoA (Figure 1B). The differential expressions of operon genes suggested that additional promoters were probably present in the operon to coordinate the transcription of individual genes.

We searched for possible promoters in the epothilone operon using the “Neural Network Promoter Prediction” (https://www.fruitfly.org/seq_tools/promoter.html) and “BPROM” (http://www.softberry.com/berry.phtml) programs, both of which have been successfully employed for promoter prediction (Yang et al., 2013; Cebrián et al., 2014; Westholm et al., 2014; Umarov and Solovyev, 2017). The “Neural Network Promoter Prediction” program found that internal promoters occurred upstream of each of the operon genes, except for epoD (shown as red arrows in Figure 1A). Prediction with “BPROM” (http://www.softberry.com/berry.phtml) revealed similar results but excavated more potential promoters, including a promoter upstream of epoD (shown as green arrows in Figure 1A). These promoters were mostly located upstream of the initiation codons of downstream ORFs, and detailed information is provided in Supplementary Table S4.

To verify the transcriptional activities of internal promoters, we amplified 1000-bp junction regions between the adjacent operon ORFs as the promoter fragments, named PepoP, PepoB ~ PepoF. These regions included approximately 800 bp of the upstream gene and 200 bp of the downstream gene, which were amplified from M. xanthus ZE9 using the corresponding primer pairs (listed in Supplementary Table S2). These fragments, each containing at least one predicted internal promoter, were cloned into the pKK-232-8 plasmid, respectively, to control the expression of chloramphenicol acetyltransferase (CAT) (Supplementary Figure S1). The aphII promoter (Zhu et al., 2013) was constructed upstream of the CAT reporter gene as a positive control, and the pKK-232-8 plasmid containing the CAT gene with no promoter was used as a negative control. The activity of each internal promoter fragment was characterized by the CAT activity assay.

As shown in Figure 1C, the PepoP promoter had the highest activity, which was close to that of the aphII promoter, while the PepoB and PepoE promoters showed low activities, slightly higher than that of the negative control. The PepoC, PepoD, and PepoF promoters also exhibited remarkable transcriptional activities, which were lower than that of PepoP. Thus, multiple internal promoters are present in the epothilone operon. However, the detected promoter activities in E. coli were significantly inconsistent with the transcriptional levels of the genes in the epothilone operon. For instance, we did not find a strong promoter upstream of the epoF, which, however, displayed the highest transcriptional level among the ORFs in the epothilone operon. The results suggested that the functions of internal promoters might be complexly interfered in the operon.

The CRISPR/dCas9 activation system (CRISPRa) is derived from the RNA-mediated CRISPR Cas system by fusing the nuclease-deficient Cas9 (dCas9) with a transcription activator, which combines with the sgRNA to guide the fused dCas9 protein to the target promoter to recruit more RNA polymerases and promote transcription (Bikard et al., 2013). CRISPRa has been used to activate gene expression in bacteria and fungi (Konermann et al., 2015; Peng et al., 2018; Cámara et al., 2020; Kiattisewee et al., 2021). To assay the effects of internal promoters on the transcription of operon genes, we performed in situ activation by introducing plasmids pSWcuomxdCas9-ω and pZJY41-sgRNA into the epothilone-producing M. xanthus ZE9. The pSWcuomxdCas9-ω plasmid carried the dCas9-ω gene encoding the dCas9 protein fused with the omega subunit of RNA polymerase (Peng et al., 2018) under the control of a cupric ion-induced promoter (Supplementary Figure S2), while the pZJY41-sgRNA plasmid carried an sgRNA targeting a specific promoter (Supplementary Figure S3). The principle of CIRSPR-dCas9 activation is shown in Figure 2A, and the sgRNA sequences targeting internal promoters are shown in Table 1. After incubation with cupric ions (200 μM), the transcriptional levels of the activated genes were significantly increased (t-test, p < 0.05), except for epoD (Figure 2B). According to the growth curves, the addition of cupric ions had no obvious influence on the growth of ZE9 or the activated mutants (Supplementary Figure S6).

For comparison, we also constructed a CRISPRa system by transferring the pSWcuomxdCas9-ω and pZJY41-sgRNA plasmids into the above constructed E. coli strains to activate separate promoters of the epothilone operon. According to the CAT activity assay, the transcription activities of all these promoters were significantly improved (t-test, p < 0.05; Figure 2C). Consistent with the previous report (Bikard et al., 2013), weak promoters were activated more efficiently, and strong promoters were also activated but to low extents. For example, the transcription activity of the weakest promoter PepoB was increased by nearly 32-fold, while that of the strongest promoter PepoP was activated by approximately 60%. However, PepoD was activated by 260% in E. coli, but not significantly in situ activated in the operon in M. xanthus. The differences between the in situ activation in the operon and ectopic activation in a separate form suggested that the transcription activities of internal promoters were interfered in the operon.

We further assayed the yields of epothilones in these activation mutant strains. As shown in Figure 3A, activation of internal promoters also significantly increased the epothilone yields, except for that in ZE9-D, in which the expression of epoD failed to be activated. Comparably, the activation of PepoC, PepoE, and PepoF produced similar yields of epothilones, which, however, were lower than that from the activation of PepoP or PepoB. The result was inconsistent with the activation results of transcriptional levels, where the best activation was achieved on epoE, increasing by ∼ 4-fold compared to that in ZE9 (Figure 2B).

Furthermore, RT-qPCR analysis of all the operon genes in different activation mutants revealed that activating a specific internal promoter promoted the transcription of not only the target gene but also the nearby genes (Figure 3B). For example, when PepoP was activated, the transcriptional levels of epoP, the downstream epoB, and the upstream epoA were increased by 2.3-fold, 2.3-fold, and 43%, respectively. The activation of PepoB significantly promoted the transcription of epoP, epoB, and epoC. The downstream or upstream distant genes were also often affected by the activation of the promoters. For example, activation of PepoC resulted in a 1.9- and 1.2-fold transcriptional increase of epoC and epoB but decreased the transcription of epoD, epoE, and epoF by 76, 24, and 31%, respectively. Similarly, activating the posterior promoter (PepoE or PepoF) not only promoted the transcription of epoE or epoF genes but also affected the transcription of anterior genes of the operon, and even inhibited the first epoA gene. The above results suggested that in situ activation of internal promoters might change the operon transcriptional profile, thus leading to changes in epothilone yields.

Obviously, high yields of epothilones required an overall transcriptional increase of all the operon genes. In our previous work, activation of the starting promoter PepoA (the promoter of epoA) significantly increased the transcription of the operon genes and improved the yield of epothilones by 1.5-fold (Peng, et al., 2018). To further improve the transcriptional efficiency, we activated the PepoA and internal promoters together. Since independent activation of PepoP or PepoB resulted in relatively more epothilones (Figure 3A), we combined the activation of PepoA with PepoP or PepoB by introducing the plasmids pZJY41-sgRNA-AB and pZJY41-sgRNA-AP into ZE9-CuOm, a M. xanthus strain containing the plasmid pSWcuomxdCas9-ω. According to the RT-qPCR analysis, activating two promoters at the same time might lead to higher transcription levels of genes. For example, compared with that of ZE9, activation of PepoP alone increased the transcription levels of epoA and epoP by 1.4- and 2.3-fold, respectively, in ZE9-P (Figure 3B), but the combined activation of PepoA and PepoP improved the transcription levels of epoA and epoP by 1.8-fold and 6.5-fold, respectively, in ZE9-AP (Figure 4A). In ZE9-AB, compared with that of ZE9-B, combined activation of PepoA and PepoB produced higher transcription of epoA, epoP, and epoC but lower transcription of epoB. Similar to that in ZE9-P or ZE9-B, combined activation of PepoA with PepoP or PepoB did not increase the transcription of the posterior genes in the operon.

Corresponding to the transcription of the epothilone operon, the production abilities of epothilones were improved similarly in ZE9-AP and ZE9-AB: The epothilone yields were increased from 4.95 mg/L in ZE9 to 11.17 mg/L in ZE9-AP and 9.59 mg/L in ZE9-AB (Figure 4B). No significant difference was found between ZE9-AP and ZE9-AB. However, compared with ZE9-P, in which the PepoP was activated alone resulting in a 1.17-fold improvement of epothilone yield, further increase in transcription of the epothilone operon did not lead to a higher yield in ZE9-AP. The potential reason is that transcription of the posterior genes was not efficiently activated by the activation combination. Notably, further combined activation of PepoA, PepoP, and PepoB did not lead to higher transcription of the anterior genes in the operon and neither increased the transcription of the posterior genes, thus leading to similar epothilone production in the ZE9-AP, ZE9-AB, and ZE9-APB strains (Figure 4).

We previously constructed dozens of epothilone-producing M. xanthus strains, in which the epothilone biosynthetic gene cluster was inserted into different sites of the DZ2 genome, producing mutants with varied epothilone production abilities (Zhu et al., 2015). We also combined promoter activation in three M. xanthus strains with different transcription levels of the epothilone genes, that is, ZE9, ZE5, and ZE10. Among these three strains, ZE9 had the highest epothilone production ability, followed by ZE10 and ZE5 exhibiting the lowest epothilone yields (Figure 5A). Consistently, the transcriptional levels of the seven ORFs in ZE5 were mostly lower than those of the other two strains (Figure 5B). Specifically in ZE10, epoP, epoC, and epoE were transcribed at higher levels, but epoA and epoF were significantly lower than that in ZE9. In ZE5, the operon genes were transcribed at much lower levels than that in ZE9, especially the posterior genes in the operon. The transcription levels of epoD, epoE, and epoF in ZE5 were only 1–2% of those in ZE9, which were evidently the limitation for the expression of the whole gene cluster.

We combined the activation on the posterior promoters (PepoD, PepoE, and PepoF) in these epothilone-producing strains. As expected, the epothilone yields were all increased in these strains, and the highest 15-fold increase of epothilone yield was obtained in ZE5 with the DEF promoter activation (Figure 5A). Consistent with the yields of epothilones, the highest activation efficiency also occurred in the ZE5-DEF strain (the relative transcriptional increases in the three strains are shown in Figures 5C–E). In ZE5-DEF, the transcriptional levels of the three activated genes epoD, epoE, and epoF were increased by 9.6, 3.1, and 51.7 times, respectively, and epoP and epoB were also increased slightly. However, the transcriptional changes in operon genes suggested that the interferences between operon promoters were complex. For example, significantly increased transcription occurred with epoC and epoE in the ZE9-DEF strain, epoE and epoF in ZE10-DEF, and epoD, epoE, and epoF in ZE5-DEF. The transcription of the anterior genes was mostly unchanged significantly. Comparably, while the transcription of operon genes and yield of epothilones in ZE9 were both higher than those of ZE5 and ZE10, the activation efficiency in ZE9-DEF was much lower. Seemingly consistent with the above activation results of separate promoters, activation of weak promoters in lower epothilone-producing strains produces better activation effects.