# REGULATION OF ANTIBIOTIC PRODUCTION IN ACTINOMYCETES

EDITED BY : Yvonne Mast, Evi Stegmann and Yinhua Lu PUBLISHED IN : Frontiers in Microbiology

#### Frontiers eBook Copyright Statement

The copyright in the text of individual articles in this eBook is the property of their respective authors or their respective institutions or funders. The copyright in graphics and images within each article may be subject to copyright of other parties. In both cases this is subject to a license granted to Frontiers. The compilation of articles constituting this eBook is the property of Frontiers.

Each article within this eBook, and the eBook itself, are published under the most recent version of the Creative Commons CC-BY licence. The version current at the date of publication of this eBook is CC-BY 4.0. If the CC-BY licence is updated, the licence granted by Frontiers is automatically updated to the new version.

When exercising any right under the CC-BY licence, Frontiers must be attributed as the original publisher of the article or eBook, as applicable.

Authors have the responsibility of ensuring that any graphics or other materials which are the property of others may be included in the CC-BY licence, but this should be checked before relying on the CC-BY licence to reproduce those materials. Any copyright notices relating to those materials must be complied with.

Copyright and source acknowledgement notices may not be removed and must be displayed in any copy, derivative work or partial copy which includes the elements in question.

All copyright, and all rights therein, are protected by national and international copyright laws. The above represents a summary only. For further information please read Frontiers' Conditions for Website Use and Copyright Statement, and the applicable CC-BY licence.

ISSN 1664-8714 ISBN 978-2-88963-995-3 DOI 10.3389/978-2-88963-995-3

#### About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

#### Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

#### Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

#### What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

# REGULATION OF ANTIBIOTIC PRODUCTION IN ACTINOMYCETES

Topic Editors: Yvonne Mast, German Collection of Microorganisms and Cell Cultures GmbH (DSMZ), Germany Evi Stegmann, University of Tübingen, Germany Yinhua Lu, Shanghai Normal University, China

Citation: Mast, Y., Stegmann, E., Lu, Y., eds. (2020). Regulation of Antibiotic Production in Actinomycetes. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-995-3

# Table of Contents


Dekun Kong, Xia Wang, Ju Nie and Guoqing Niu

*42 New Molecular Tools for Regulation and Improvement of A40926 Glycopeptide Antibiotic Production in* Nonomuraea gerenzanensis *ATCC 39727*

Oleksandr Yushchuk, Andres Andreo-Vidal, Giorgia Letizia Marcone, Mervyn Bibb, Flavia Marinelli and Elisa Binda

*56 The Balance Metabolism Safety Net: Integration of Stress Signals by Interacting Transcriptional Factors in* Streptomyces *and Related Actinobacteria*

Juan F. Martín and Paloma Liras

*75 sRNA scr5239 Involved in Feedback Loop Regulation of* Streptomyces coelicolor *Central Metabolism*

Franziska Engel, Elena Ossipova, Per-Johan Jakobsson, Michael-Paul Vockenhuber and Beatrix Suess


Alberto Vassallo, Emilia Palazzotto, Giovanni Renzone, Luigi Botta, Teresa Faddetta, Andrea Scaloni, Anna Maria Puglia and Giuseppe Gallo


Natalia M. Vior, Eva Cea-Torrescassana, Tom H. Eyles, Govind Chandra and Andrew W. Truman


Cláudia M. Vicente, Jean-Michel Girardet, Laurence Hôtel and Bertrand Aigle

# Editorial: Regulation of Antibiotic Production in Actinomycetes

Yvonne Mast 1,2,3 \*, Evi Stegmann<sup>4</sup> and Yinhua Lu<sup>5</sup>

<sup>1</sup> Department Bioresources for Bioeconomy and Health Research, Leibniz Institute DSMZ -German Collection of Microorganisms and Cell Cultures, Brunswick, Germany, <sup>2</sup> Department of Microbiology, Technical University Braunschweig, Brunswick, Germany, <sup>3</sup> German Center for Infection Research (DZIF), Partner Site Tübingen, Tübingen, Germany, <sup>4</sup> Department of Microbial Bioactive Compounds, Interfaculty Institute of Microbiology and Infection Medicine, Faculty of Science, University of Tübingen, Tübingen, Germany, <sup>5</sup> College of Life Sciences, Shanghai Normal University, Shanghai, China

Keywords: actinomycetes, antibiotic, secondary metabolism, regulation, signaling

**Editorial on the Research Topic**

#### **Regulation of Antibiotic Production in Actinomycetes**

#### Edited by:

Rustam Aminov, University of Aberdeen, United Kingdom

#### Reviewed by:

Polpass Arul Jose, The Hebrew University of Jerusalem, Israel Ryan Seipke, University of Leeds, United Kingdom

> \*Correspondence: Yvonne Mast yvonne.mast@dsmz.de

#### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

> Received: 02 June 2020 Accepted: 16 June 2020 Published: 22 July 2020

#### Citation:

Mast Y, Stegmann E and Lu Y (2020) Editorial: Regulation of Antibiotic Production in Actinomycetes. Front. Microbiol. 11:1566. doi: 10.3389/fmicb.2020.01566 Actinomycetes are a well-known resource for the discovery of novel bioactive compounds. Up to two-thirds of all antibiotics in use today are natural products from actinomycetes or semisynthetic derivatives thereof (Barka et al., 2016). Streptomyces species have thus far proven to be a particularly useful resource. It is now 80 years ago (1940) when Selman Waksman isolated the first antibiotic from an actinomycete's species, which was actinomycin A from Actinomyces (Streptomyces) antibioticus (Kresge et al., 2004). Shortly after, he coined the term "antibiotic," which is just one reason why he is considered widely to be the true "Father of Antibiotics" (Kresge et al., 2004). In the following years Waksman's group successfully applied his screening technique for the identification of natural substances and isolated more than 20 additional new bioactive compounds, thereby initiating the so-called "golden era of antibiotic discovery." One of the most important Waksman compounds was streptomycin, which Albert Schatz and he identified in 1943 as a product from Streptomyces griseus. As discoverer of the "first antibiotic effective against tuberculosis" he was rewarded with the Nobel Prize in 1952 (Kresge et al., 2004). Since that time, actinomycetes, especially streptomycetes, have been studied intensively, not only because of their gifted metabolic potential but also due to their interesting biological features: e.g., streptomycetes have extraordinarily large genomes with up to ∼13 Mbp, which are of linear style and with a high GC content (>70%). They are characterized by an unique life-cycle, involving the formation of mats of fungus-like mycelium from which aerial branches are formed that carry chains of spores. The spores are then readily released for dissemination (Chater, 2016). Nowadays, actinomycetes are still, or maybe more than ever, a valuable resource for novel drugs. New antibiotics are urgently needed due to the constant rise of infections caused by drug-resistant pathogens (O'Neill, 2016). In recent years, a wealth of expertise has been gained regarding especially knowledge on antibiotic biosynthesis in these organisms. Particularly since the advent of next-generation sequencing technologies, the field of antibiotic research has experienced a remarkable revival, which is due to the upcoming of a large amount of genome sequence information that gave rise to novel approaches in order to exploit actinomycetes for drug discovery. Comparable few studies focus on a better understanding of regulatory mechanism of antibiotic biosynthesis. Understanding regulatory principles in antibiotic producing organisms is, however, fundamental for targeted optimization of antibiotic production yields, activation of silent gene clusters to find novel antibiotics, or facilitating genetic engineering approaches, e.g., to allow heterologous expression of engineered gene clusters. Besides that, regulation is the interface connecting aspects of chemical differentiation with complex

morphological differentiation events in actinomycetes. Thus, understanding regulation means also learning about actinomycetal biology.

Regulation of secondary metabolism in actinomycetes is a complex process depending on various factors, such as nutrient limitation, oxygen supply or pH conditions (Liu et al., 2013; Wohlleben et al., 2017). Regulation can occur at the level of transcription and/or translation. Concerning the former, regulatory signaling cascades can include global, as well as pathway-specific (cluster-situated) transcriptional regulators. Thereby, various chemical substances may serve as effectors for transcriptional regulators, such as quorum-sensing-like γbutyrolactone (GBL) molecules, antibiotics, or intermediates thereof. Groundbreaking scientific achievement in this field has been made already in the 1980s by Sueharu Horinouchi using the above mentioned famous S. griseus strain (Ohnishi, 2010). Horinouchi et al., showed that streptomycin production and morphological differentiation of S. griseus is under control of a complex regulatory signaling cascade triggered by the microbial hormone-like substance A-factor (Horinouchi et al., 1983; Horinouchi, 2002). The A-factor is a GBL molecule and was initially discovered in the 1960s by Khokhlov et al., as a substance controlling streptomycin production in the originally named strain Actinomyces streptomycini (S. griseus) (Khokhlov et al., 1967). Horinouchi thereafter worked out in detail that the Afactor is produced in a growth-dependent manner at nanomolar concentrations. When A-factor concentrations reach a certain threshold the cognate repressor-type receptor ArpA dissociates from its target gene adpA. adpA itself encodes the transcriptional regulator AdpA, which acts as a pleiotropic regulator controlling several hundred genes in S. griseus, including a number of genes involved in morphological differentiation and secondary metabolite biosynthesis (Horinouchi, 2002). The A-factor signaling cascade became a model system for understanding regulatory principles in antibiotic-producing actinomycetes. Indeed, GBLs are produced by many, if not all streptomycetes (Bibb, 2005), whereby the signaling cascades that are governed by GBLs and which control transcriptional activation of antibiotic gene clusters can have different compositions in various antibiotic producers (Bibb, 2005; Wohlleben et al., 2017). Furthermore, not only GBLs may function as signaling molecules but also antibiotics or their intermediates, which altogether result in complex regulatory networks including elaborate feedback and feedforward loops (Xu and Yang, 2019). Besides that, such networks themselves are embedded in larger regulatory circuits involving transcriptional as well as translational regulation, which may respond to diverse environmental signals as mentioned above (Bibb, 2005; Liu et al., 2013). Most of these various regulatory interactions have not been explored yet. Thus, there is still a lot to investigate for actinomycetes.

With this Frontiers Research Topic we aimed to create space for current advances and knowledge on regulation of antibiotic biosynthesis in actinomycetes and highlight the importance of this topic for antibiotic research. We collected 13 articles, including nine original articles and four reviews covering a variety of different regulatory aspects of antibiotic production. The articles address the topics of regulatory networks, signaling molecules for antibiotic production, studies on regulator-driven activation of silent gene clusters, optimization of antibiotic production, as well as regulatory effects from small open reading frames (ORFs) and small non-coding RNAs in actinomycetes.

The overarching review article from Juan Martin and Paloma Liras reports on the interplay between various environmental stress signals (e.g., variations in carbon and nitrogen sources, phosphate, oxygen, iron, and other nutrients) and major transcriptional factors in actinomycetes. It is described how these different signals are integrated at the molecular level (Martín and Liras). The authors propose a cross-talk of the transcriptional regulators, which bind to specific DNA regions, so-called "integrator sites," that allow to compensate for unbalances produced by metabolic stresses—a phenomenon designated as the "balance metabolism safety net" (Martín and Liras). The review article from Kong et al., concentrates on autoregulatordriven regulation of secondary metabolite biosynthesis. Here, the authors highlight the recent findings on regulation of antibiotic biosynthesis by small molecules including GBLs, antibiotics and their intermediates. This review is accompanied by four original research articles, which address specific signaling cascades in streptomycetes: Vicente et al., report on the regulatory role of the GBL-receptor-type regulator AlpZ in the kinamycin producer Streptomyces ambofaciens. Here, the authors present a molecular dynamics approach, which was applied to investigate the DNAbinding properties of AlpZ to its target DNA sequence and the impact of the potential signaling molecule on receptor-DNA interaction. Misaki et al., explore the repressive function of the pseudo-GBL receptor SrrB on lankacidin and lankamycin production in Streptomyces rochei and its hierarchical positioning within the regulatory signaling cascade. The study of Kang et al.., shows the involvement of the pleiotropic AdpA-like regulator AdpAlin from Streptomyces lincolnensis in lincomycin production and morphological differentiation; whereas Vior et al., report on a so far less studied cluster-situated gene btmL, which encodes a protein that acts as a modulator of the biosynthesis of the ribosomally synthesized and post-translationally modified peptide (RiPP) antibiotic bottromycin from Streptomyces scabies. Bottromycin shows an activity against clinical relevant pathogens and exhibits a unique chemical structure and mode of action, which makes it a promising drug lead compound. However, production yields in the known producers are very low, which hampered further drug development so far (Vior et al.). Indeed, this is a well-known issue for several drug development approaches, where further advancement is hindered by low production yields. Understanding regulatory principles is a key aspect, which allows the targeted improvement of antibiotic production yields. This aspect is largely covered by two reviews from Xia et al. and Li et al.: The article from Xia et al., summarizes the latest knowledge on regulatory cascades in streptomycetes and their importance for yield improvement applications and metabolite mining purposes, whereas the article from Li et al. concentrates on recent advances of synthetic biology approaches involving also regulatory components for yield improvement strategies. The importance of regulation for antibiotic yield improvement is also highlighted by two original research articles (Park et al.; Yushchuk et al.): Park et al. report on the overexpression of a set of pathway-specific regulatory genes (nppRI–nppRVI) in Pseudonocardia autotrophica, which significantly improved production of the polyene-like substance NPP B2, whereas the study of Yushchuk et al. extended the genetic toolkit for members of the Nonomurea genus, involving the usage of suitable promoters and cluster-situated regulators to improve production yields of the glycopeptide antibiotic A40926 in Nonomurea gerenzanensis. As outlined in the review of Xia et al. understanding regulatory signaling can also help to activate silent gene clusters for novel natural compounds discovery. This is underpinned by an original research article from Krause et al. who report on the activating function of the SARP-type regulator PapR2 from Streptomyces pristinaespiralis and describe how these type of regulators can be used as general devices to activate antibiotic gene clusters in actinomycetes. Besides several studies that deal with transcriptional regulators, also posttranscriptional regulation is the topic in two research articles: The study of Engel et al. disclosed the post-transcriptional regulatory effect of a small RNA scr5239 on the primary metabolic enzyme phosphoenolpyruvate carboxykinase PEPK and report on the interconnection between central carbon metabolism and secondary metabolism via this small RNA. Furthermore, the study of Vassallo et al. shows that the conserved small ORF trpM, of which the product modulates L-tryptophan biosynthesis, has a regulatory effect on morphological development and antibiotic production in Streptomyces coelicolor. TrpM is suggested to interact with the putative cytosol aminopeptidase PepA (SCO2179) via a post-transcriptional and/or post-translational regulatory mechanism, whereby PepA plays a key role in antibiotic production and sporulation. Overall, this collection

#### REFERENCES


of articles gives a profound overview on the importance of knowledge on regulation of antibiotic biosynthesis for antibiotic research. We are delighted to present this Research Topic in Frontiers in Microbiology and hope that it will promote further research in this area.

# AUTHOR CONTRIBUTIONS

YM, YL, and ES edited the Frontiers Research Topic on Regulation of Antibiotic Production in Actinomycetes and wrote the manuscript. All authors contributed to the article and approved the submitted version.

# FUNDING

This work was supported by funding received from the Baden-Württemberg Stiftung (BWST\_WSF-035) and the German Center for Infection research (DZIF) (TTU 09.819) given to YM and ES, and the National Natural Science Foundation of China (31770088) given to YL.

# ACKNOWLEDGMENTS

The Editors would like to thank all the authors who contributed their valuable work to this Research Topic. We also acknowledge the important comments and suggestions of all the designated peer reviewers, who ensured robustness of research data and helped to improve the articles for the Research Topic. Finally, the authors thank the Chef Editor and the Frontiers Team for their support during handling of the Research Topic.


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2020 Mast, Stegmann and Lu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# AdpAlin, a Pleiotropic Transcriptional Regulator, Is Involved in the Cascade Regulation of Lincomycin Biosynthesis in Streptomyces lincolnensis

Yajing Kang<sup>1</sup> , Yingying Wang<sup>1</sup> , Bingbing Hou1,2, Ruida Wang<sup>1</sup> , Jiang Ye<sup>2</sup> \*, Xiaoyu Zhu<sup>1</sup> , Haizhen Wu1,2 and Huizhan Zhang1,2 \*

<sup>1</sup> State Key Laboratory of Bioreactor Engineering, East China University of Science and Technology, Shanghai, China, <sup>2</sup> Department of Applied Biology, East China University of Science and Technology, Shanghai, China

#### Edited by:

Yinhua Lu, Shanghai Normal University, China

#### Reviewed by:

Xu-Ming Mao, Zhejiang University, China Ying Wen, China Agricultural University (CAU), China

#### \*Correspondence:

Jiang Ye yej@ecust.edu.cn Huizhan Zhang huizhzh@ecust.edu.cn

#### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 14 August 2019 Accepted: 08 October 2019 Published: 23 October 2019

#### Citation:

Kang Y, Wang Y, Hou B, Wang R, Ye J, Zhu X, Wu H and Zhang H (2019) AdpAlin, a Pleiotropic Transcriptional Regulator, Is Involved in the Cascade Regulation of Lincomycin Biosynthesis in Streptomyces lincolnensis. Front. Microbiol. 10:2428. doi: 10.3389/fmicb.2019.02428 Lincomycin is one of the most important antibiotics in clinical practice. To further understand the regulatory mechanism on lincomycin biosynthesis, we investigated a pleiotropic transcriptional regulator AdpAlin in the lincomycin producer Streptomyces lincolnensis NRRL 2936. Deletion of adpAlin (which generated 1adpAlin) interrupted lincomycin biosynthesis and impaired the morphological differentiation. We also found that putative AdpA binding sites were unusually scattered in the promoters of all the 8 putative operons in the lincomycin biosynthetic gene cluster (BGC). In 1adpAlin, transcript levels of structural genes in 8 putative operons were decreased with varying degrees, and electrophoretic mobility shift assays (EMSAs) confirmed that AdpAlin activated the overall putative operons via directly binding to their promoter regions. Thus, we speculated that the entire lincomycin biosynthesis is under the control of AdpAlin. Besides, AdpAlin participated in lincomycin biosynthesis by binding to the promoter of lmbU which encoded a cluster sited regulator (CSR) LmbU of lincomycin biosynthesis. Results of qRT-PCR and catechol dioxygenase activity assay showed that AdpAlin activated the transcription of lmbU. In addition, AdpAlin activated the transcription of the bldA by binding to its promoter, suggesting that AdpAlin indirectly participated in lincomycin biosynthesis and morphological differentiation. Uncommon but understandable, AdpAlin auto-activated its own transcription via binding to its own promoter region. In conclusion, we provided a molecular mechanism around the effect of AdpAlin on lincomycin biosynthesis in S. lincolnensis, and revealed a cascade regulation of lincomycin biosynthesis by AdpAlin, LmbU, and BldA.

Keywords: AdpA, lincomycin, Streptomyces lincolnensis, transcriptional regulator, cascade regulation

# INTRODUCTION

Lincomycin is a naturally occurring antibiotic isolated from soil sample, and it was first introduced into clinical practice in 1963 (Macleod et al., 1964). Lincomycin and its derivatives belong to lincosamide antibiotics and exhibit biological activities against anaerobic and some protozoal infections by inhibiting protein synthesis in sensitive targets (Spizek and Rezanka, 2017).

**8**

Clindamycin is a semi-synthetic chlorinated derivative of lincomycin and it is marked by being one of the 20 most important antibiotics (Spizek and Rezanka, 2004a). Given the extensive clinical application of lincomycin, multiple attempts have been taken into industrial practice to increase the production yields of lincomycin or to optimize the products (Spizek and Rezanka, 2004b; Li et al., 2007). Genetic manipulations are also adopted as complement to enhance the production of lincomycin (Pang et al., 2015; Xu et al., 2018). Though the pathway of lincomycin biosynthesis was assembled mainly within the recent 10 years (Neusser et al., 1998; Novotna et al., 2004; Sasaki et al., 2012; Lin et al., 2014; Pang et al., 2015; Zhao et al., 2015; Jiraskova et al., 2016), studies on the regulation mechanism of lincomycin biosynthesis are prompted quite slowly. Thus, various methods have limited effect on yield improvement of lincomycin.

Biosynthesis of antibiotics is controlled by elaborate regulatory mechanisms. Hormone-like signaling molecules, for example γ-butyrolactone (Takano et al., 2000; Hsiao et al., 2009; Du et al., 2011), serve as stimuli that interact with their receptor proteins to prelude the secondary metabolism (Niu et al., 2016). Global regulators and/or pleiotropic regulators then deliver these signals to CSRs which directly control the onset of antibiotic biosynthesis. In addition, researches on the secondary metabolism in Streptomyces is expanded to the posttranscriptional regulation. For example, BldA, a rare tRNA in Streptomyces, has significant importance on morphological differentiation and antibiotic biosynthesis (Hackl and Bechthold, 2015). As for lincomycin biosynthesis, only limited researches contribute to decipher the regulatory network. Lu et al. (2008) found that LmbU contributes to lincomycin biosynthesis. Hou et al. (2018a, 2019) demonstrated that LmbU, as a CSR, positively regulates lincomycin biosynthesis by controlling the transcription of lmbA, lmbC, lmbJ, lmbK, lmbW, and lmbU itself, and subsequently solved the subtle mechanism of LmbU regulon. Besides, Hou et al. (2018b) also found that BldA functions as a global regulator on both morphological differentiation and lincomycin biosynthesis at the level of translation, and genes lmbB2, lmbY, and lmbU, which all contain TTA rare codon, get involved in the regulon. Meng et al. (2017) revealed the regulatory network between nitrate metabolism and lincomycin biosynthesis where GlnR activates the transcription of lmrA, the lincomycin exporter gene. Lately, a TetR-type regulator SLCG\_2919 has been identified as a repressor of lincomycin biosynthesis that controls the transcription of lmbA, lmbC, lmbE, lmbG, lmbK, lmbR, lmbV, and lmbW (Xu et al., 2019). However, to complete the regulatory network of lincomycin biosynthesis, there are lots of gaps remained.

AdpA was previously found to be an A-factor dependent regulator and repressed by ArpA (Kato et al., 2007). AdpA amplifies the A-factor signal and thus participates in morphological differentiation and secondary metabolism. By means of chromatin affinity precipitation (ChAP) and chromatin immunoprecipitation (ChIP), Higo et al. (2012) found that AdpA controls more than 500 genes in Streptomyces griseus. Afterward AdpA is considered to be a regulator of great importance in Streptomyces. In S. chattanoogensis, AdpAch controls the expression of wblAch, and thus participates in the regulation of natamycin biosynthesis (Yu et al., 2014). In S. roseosporus, AdpA controls the expression of atrA and indirectly control daptomycin biosynthesis (Mao et al., 2015). In a recent research, AdpA interacts with the twocomponent system PhoRP and both of them contribute to the transcription of atrA (Zheng et al., 2019). Effects of AdpA on antibiotic biosynthesis is always a popular topic. AdpA always serves as an activator of antibiotic biosynthesis, a classic case is that AdpA activates the transcription of strR, which encodes the CSR of streptomycin biosynthesis. Therefore, AdpA regulates streptomycin biosynthesis positively and indirectly (Tomono et al., 2005). Similarly, for the biosynthesis of grixazone (Higashi et al., 2007), nikkomycin (Pan et al., 2009), and natamycin (Yu et al., 2018), AdpA activates the transcription of CSRs in their BGCs and indirectly regulates antibiotic biosynthesis. On the other hand, AdpA has a negative impact on oviedomycin biosynthesis in S. ansochromogenes by repressing the transcription of CSR (Xu et al., 2017). Very recently, it was reported that AdpA from S. xiamenensis 318 negatively regulates morphological differentiation as well as polycyclic tetramate macrolactams (PTMs) production, while positively regulates xiamenmycin production by activating the transcription of two of the structural genes ximA and ximB (Bu et al., 2019). As we can speculate from previous studies, AdpA typically controls the CSR and/or some structural genes in an antibiotic BGC. So, we scanned the lincomycin BGC and found that putative AdpA binding sites were extraordinarily scattered upstream all of the 8 putative operons with different amounts and locations. Thus, we focused on the pleiotropic regulator AdpA (GenBank accession no. ANS65440.1) and attempted to investigate its regulatory mechanism of lincomycin biosynthesis in Streptomyces lincolnensis in this study.

#### MATERIALS AND METHODS

# Bacterial Strains, Plasmids, and Culture Conditions

Bacterial strains and plasmids used in this study are listed in **Table 1**. S. lincolnensis NRRL 2936 which served as wild type (WT) and its mutants were incubated at 28◦C on mannitol soya flour (MS) medium (Kieser et al., 2000) for 3–5 days for routine cultivation, phenotype observation, and strain preservation, and then cultivated at 28◦C in YEME liquid medium [10 g/L yeast extract (OXOID, United States), 5 g/L polypeptone (Nihon Pharmaceutical, Japan), 10 g/L glucose (Lingfeng, China), 3 g/L maltose (Generay, China), 5 mM MgCl2·2H2O (Lingfeng, China), 340 g/L sucrose (Titan, China), dissolved in dH2O] with shaking (210 rpm) for 3–5 days for routine cultivation, total DNA extraction, and sporeless strain preservation. Fermentation medium FM1 [20 g/L lactose (SCRC, China), 20 g/L glucose, 10 g/L polypeptone, 10 g/L corn steep liquor (Aladdin, China), dissolved in dH2O] is used for primary cultivation, and FM2 [20 g/L lactose, 20 g/L glucose, 10 g/L polypeptone, 10 g/L corn steep liquor, 4 g/L


(Continued)

#### TABLE 1 | Continued


CaCO<sup>3</sup> (Lingfeng, China), dissolved in dH2O] is used for dry cell weight determination and lincomycin production assays. ISP4 medium [10 g/L soluble starch (Lingfeng, China), 1 g/L K2HPO<sup>4</sup> (Lingfeng, China), 5 g/L MgSO4·7H2O (Lingfeng, China), 1 g/L NaCl (Titan, China), 2 g/L (NH4)2SO<sup>4</sup> (Lingfeng, China), 2 g/L CaCO3, 15 g/L Agar (Shize, China), 0.001 g/L FeSO4·7H2O, 0.001 g/L MnCl2·4H2O (Lingfeng, China), 0.001 g/L ZnSO4·7H2O (Lingfeng, China), 0.02 mol/L MgCl2, dissolved in dH2O] was used for conjugation of Escherichia coli and S. lincolnensis. Antibiotics were supplemented on demand with the following final concentration: 20 µg/mL apramycin (Sangon Biotech, China), 20 µg/mL kanamycin (Kinglyuan, China), 12 µg/mL chloramphenicol (Sigma-Aldrich, United States), and/or 20 µg/mL nalidixic acid (Aladdin, China).

Escherichia coli strains were cultivated at 37◦C in Luria-Bertani (LB) liquid medium with shaking (180 rpm) or on LB solid media. Antibiotics were supplemented on demand with the following final concentration: 50 µg/mL apramycin, 50 µg/mL kanamycin, and/or 30 µg/mL chloramphenicol.

Micrococcus luteus 28001 were cultivated at 37◦C in medium III (5 g/L polypeptone, 1.5 g/L beef extract (SCRC, China), 3 g/L yeast extract, 3.5 g/L NaCl, 3.68 g/L K2HPO4, 1.32 g/L KH2PO<sup>4</sup> (Lingfeng, China), 1 g/L glucose, 18 g/L agar) for 16 to 18 h.

#### Deletion and Complementation of AdpAlin

To construct AdpAlin disruption strain 1adpA in S. lincolnensis, DNA fragments of upstream and downstream region of AdpAlin were amplified separately using primers ad-F1/R1 and ad-F2/R2. Then digested, respectively, with restriction enzyme Hind III/Xba I and BamHI/EcoR I (Thermo Fisher Scientific, United States) and ligated into the E. coli-Streptomyces shuttle plasmid pMJ1 to generate plasmid pADNU. Then E. coli ET12567/pUZ8002 was used to introduce pADNU into S. lincolnensis NRRL 2936 by conjugal transfer (Hou et al., 2018b). As a result of homologous recombination 1adpA was constructed. DNA sequencing with

primers IDad-F1/IDneo-R1 and IDneo-F2/IDad-R2 was adopted for further identification.

To construct adpA complementation strain 1adpA:adpA, a DNA fragment covering AdpAlin was amplified by PCR with primers ad-C-F/R and then digested with Nde I/EcoR I (Thermo Fisher Scientific, United States). Then ligated into the corresponding sites of the integrative vector pSET152. The resulting plasmid pADC was introduced into 1adpA by conjugal transfer and integrated into the chromosome to generate 1adpA:adpA where the complemented AdpAlin was under the control of the promoter ermE∗p. DNA sequencing with primers 152-F/R was adopted for further identification.

All primers used in this study are listed in **Supplementary Table S1**, and synthesized by Genewiz (China).

#### Scanning Electron Microscope (SEM)

Scanning electron microscope assay referred to a previously established method (Hou et al., 2018b) with some optimizations. S. lincolnensis NRRL 2936, 1adpA, and 1adpA:adpA were cultured on SMA medium at 28◦C for about 5 days. Equivalent areas of the lawn were harvested and placed in 2.5% glutaraldehyde solution overnight. Dehydrated by vacuum freezing drying and sprayed with platinum by Gatan ALTO 1000E (Gatan, United States). Then observed with Hitachi S-3400N scanning electron microscopy (Hitachi, Japan).

# Dry Cell Weight Determination and Lincomycin Bioassay Analysis

Streptomyces lincolnensis NRRL 2936, 1adpA, and 1adpA:adpA were inoculated from SMA medium into FM1 at 28◦C with shaking (210 rpm) for 3 days and then inoculated into FM2 and cultured for 6 days at 28◦C with shaking (210 rpm). Precipitate of each sample was harvested every 24 h and dried at 55◦C for 24 h. Then the weights of the dried precipitates represent the dry cell weights. Meanwhile, to analyze the bioassay of lincomycin, supernatant of each sample was harvested at the same time, and previously mentioned method (Pharmacopoeia of the People's Republic of China [PPRC], 1990; Hou et al., 2018a) with some modifications was adopted. M. luteus 28001, used as indicator, was cultured on medium III at 37◦C for 16–18 h and the lawn was washed off with 0.9% NaCl and suspended readily to use. Lincomycin standard solutions (4, 6, 8, 10, 12, 14, and 16 µg/mL) were used for the standard curve and internal control. Diameters of inhibition zone were linearized with the logarithmic values of the concentrations of the lincomycin standard solutions. Concentration of each sample was calculated on the basis of the standard curves. All assays in this section were performed in duplication and standard errors of the mean were calculated. The software GraphPad Prism 7.00 was used to draw the line graph of dry cell weight and histogram of lincomycin bioassay.

# RNA Extraction and Quantitative Real-Time PCR (qRT-PCR)

Streptomyces lincolnensis NRRL 2936, and 1adpA cultured on the second day in FM2 medium were used to extract total RNA. Precipitate of samples were ground in liquid nitrogen (Liu et al., 2013) and followed by the method using TRIzol (Thermo Fisher Scientific, United States) (Setinova et al., 2017). After reacting with Recombinant DNase I (Takara, Japan) to remove the trace amount of DNA, 800 ng of RNA samples (analyzed by NanoDrop 2000, Thermo Fisher Scientific, United States) were reverse transcribed to cDNA using Reverse Transcriptase M-MLV (Takara, Japan). SYBR green PCR master mix (ToYoBo, Japan) was used and qRT-PCR was performed in triplication for each transcript. qRT-PCR conditions were mentioned previously (Hou et al., 2018a). To detect the transcript level of AdpAlin targets, primers qbl-F/R, qU-F/R, qA-F/R, qC-F/R, qD-F/R, qJ-F/R, qK-F/R, qV-F/R, and qW-F/R in **Supplementary Table S1** were used. And primers qhrdB-F/R were used to detect the transcript level of hrdB which served as an internal control. qRT-PCR was performed with samples in triplication and data were treated with the threshold cycle (2−11<sup>C</sup> <sup>T</sup>) method (Livak and Schmittgen, 2001) and standard errors of the mean were calculated. GraphPad Prism 7.00 was used to draw the histogram of relative expression level of each AdpAlin target.

# Catechol Dioxygenase Activity Analysis

DNA fragment covering reporter gene xylTE was amplified by PCR with primers xyl-F/R. Promoters of different AdpAlin targets were amplified separately by PCR with primers adp-xyl-F/R (for adpAp, from −610 to +4), blp-xyl-F/R (for bldAp, from −799 to +52), Up-xyl-F/R (for lmbUp, from −730 to +17), Ap-xyl-F/R (for lmbAp, from −533 to +3), Cp-xyl-F/R (for lmbCp, from −513 to −1), Dp-xyl-F/R (for lmbDp, from −581 to +3), Jp-xyl-F/R (for lmbJp, from −391 to +3), Kp-xyl-F/R (for lmbKp, from −896 to +3), Vp-xyl-F/R (for lmbVp, from −364 to −1), and Wp-xyl-F/R (for lmbWp, from −456 to +4). In respect to bldAp, +1 represents the start point of mature bldA. As for other promoters, +1 represents the translation starting point of the genes controlled by them. Promoter fragment and xylTE fragment were inserted into Pvu II site of the integrative vector pSET152 using Super Efficiency Fast Seamless Cloning kits (DoGene, China) (Hou et al., 2018b) to construct reporter plasmids pADPX, pBLPX, pUPX, pAPX, pCPX, pDPX, pJPX, pKPX, pVPX, pWPX, and pADPX2. Then introduced into S. lincolnensis NRRL 2936 or 1adpA respectively, to investigate the effects of AdpAlin on these targets. Referred to the method optimized by Hou et al. (2018b), Catechol dioxygenase activity analysis was carried out in triplication. Standard errors of the mean were calculated and the software GraphPad Prism 7.00 was used to draw the histogram of catechol dioxygenase activity.

# Electrophoretic Mobility Shift Assays (EMSAs)

The AdpAlin gene was amplified by PCR with primers ad-C-F/R and digested with Nde I/EcoR I. DNA fragment was cloned into corresponding sites of pET-28a (+) vector (Novagen, United States), and the resulting plasmid pADH was transformed into E. coli BL21 (DE3). Overexpression and purification of recombinant protein refer to the procedures described previously (Hou et al., 2018a). According to our experience, DNA probes with length around 200 bp are appropriate for EMSAs with

Kang et al. AdpAlin Positively Regulates Lincomycin Biosynthesis

AdpAlin. For the first round of amplification, primers adp-A-F/R, adp-B-F/R, adp-B-1-F/R, adp-B-F/adp-B-2-R, adp-B-F/madp-B-2-R, blp-A-F/R, blp-B-F/R, blp-N-F/R, mblp-B-F/blp-B-R, Up-A-F/R, Up-B-F/R, Up-C-F/R, mUp-B-F/R, Ap-A-F/R, mAp-F/R, Cp-A-F/R, Cp-B-F/R, Cp-C-F/R, Dp-A-F/R, Jp-A-F/R, mJp-A-F/R, Jp-B-F/R, Kp-B-F/R, mKp-B-F/R, and Vp-A-F/R in **Supplementary Table S1** were used to amplify DNA probes with putative AdpA binding sites, and primers nad-F/R were used to amplify DNA probe with no AdpA binding site as a negative control. Genes lmbV and lmbW share the same DNA probe. For the second round, amplified DNA fragments are used as templates with primer Biotin-linker<sup>∗</sup> to harvest DNA probes with biotin labeled at 5<sup>0</sup> terminal. EMSAs were performed as previously described (Liao et al., 2015) using chemiluminescent EMSA kits (Beyotime Biotechnology, China). AdpAlin of different concentrations (0, 1.6, 3.2, and/or 6.4 µM, respectively) interacted with 2.5 nM biotin labeled DNA probe in binding buffer TGB [20 mM Tris–HCl (Shize, China), 5% glycerol (Titan, China), and 0.1% BSA (Sangon, China), pH 7.5], and 200-folds excess of unlabeled probes were added as competitive assays.

# RESULTS

#### AdpAlin Positively Regulates Both Lincomycin Biosynthesis and Morphological Differentiation in S. lincolnensis

It has been shown that AraC/XylS family regulators control various metabolic pathways including antibiotic biosynthesis (Ibarra et al., 2008). There are about 30 AraC/XylS family regulators in S. lincolnensis, among which, AdpA is the most famous one. Based on this, we investigated the effects of AdpAlin on lincomycin biosynthesis and attempted to propose some innovative idea on this classic regulator. Alignment of AdpA from 26 Streptomyces species (**Supplementary Figures S1A,B**) showed that AdpA retained the conserved N-terminal ThiJ/PfpI/DJ-1-like (also referenced as GATase-1) dimerization domain and C-terminal AraC/XylS-type DNAbinding domain (DBD) (Ohnishi et al., 2005). The first 340 amino acids possessed an ortholog with over 90% identities, and the main diversity occurred at the tail end of C-terminus with a length of no more than 90 amino acids, after the conserved DNA binding domain (**Supplementary Figure S1A**). To infer the evolutionary history of AdpA, phylogenetic analysis was performed using a maximum likelihood method. The results showed that AdpAlin possessed an ortholog with 89% amino acid identity to AdpAsg, and thus we classified AdpAlin to be one member of the AraC/XylS family.

To investigate the effects of AdpAlin on lincomycin biosynthesis, AdpAlin null mutant was constructed and named as 1adpA. Lincomycin biosynthesis was significantly influenced by the non-functional AdpAlin. In medium FM2, lincomycin started to be produced on the second day in WT, reached a maximum bioassay of 30.10 µg/mL between the second and the fourth day, and maintained thereafter. However, bioassay of lincomycin in 1adpA remained undetectable throughout the entire 6 days (**Figure 1A**). Furthermore, the lawns of 1adpA on SMA medium exhibited a bald phenotype distinct from WT (**Figure 1B**). Deletion of AdpAlin blocked the sporulation and caused long, extended aerial hyphae when detected by SEM (**Figure 1C**). Complementation of AdpAlin in 1adpA strain (1adpA:adpA) restored both lincomycin biosynthesis and sporulation as expected (**Figure 1**) though lincomycin production in complemented strain did not restore to WT level, probably due to using the promoter ermE∗p. Moreover, the biomasses of WT, 1adpA, and 1adpA:adpA were measured at all the four detected days. The data showed that biomasses of the three strains have no significant differences at days 1 and 6, while at day 2, 1adpA had decreased biomass compared to WT and 1adpA:adpA, and at day 4, 1adpA:adpA had increased biomass compared to WT and 1adpA. These data suggested that AdpAlin is an important regulator of lincomycin biosynthesis in S. lincolnensis.

#### AdpAlin Directly Activates Transcription of the Structural Genes in the lmb Cluster

AdpA regulates more than 500 genes in S. griseus (Higo et al., 2012), and AdpA binding motifs have been well studied in other Streptomyces species such as S. griseus (Yamazaki et al., 2004), S. coelicolor (Kim et al., 2005), and S. lividans (Guyet et al., 2013). Additionally, Ming et al. have solved the complex structure of AdpA-DBD and target DNA in S. griseus (Yao et al., 2013). AdpA binding site is recognized as 5<sup>0</sup> -TGGCSNGWWY-3<sup>0</sup> (where S is G or C, W is A or T, Y is T or C, and N is any nucleotide), and G at position 2 and C at position 4 are more highly conserved than the other nucleotides in this motif (Yao et al., 2013). Lincomycin BGC was named as the lmb cluster, and the gene organization was shown in **Supplementary Figure S2A**. The lmb cluster contains 8 putative operons and the first genes of them are lmbA, lmbC, lmbD, lmbJ, lmbK, lmbV, lmbW, and lmbU, respectively. We looked into the nucleic acid sequence of the lmb cluster and searched with the conserved AdpA binding sequence, and no more than 3 mismatches in the last 3 bp were allowed. We found that putative AdpA binding sites were scattered in the upstream region of all the 8 putative operons (**Supplementary Figures S2B–F**). The numbers of putative AdpA binding sites in the upstream of lmbA, lmbC, lmbD, lmbJ, lmbK, lmbV, lmbW, and lmbU are 1, 6, 1, 3, 3, 2, 2, and 10, respectively. Therefore, we speculated that the entire biosynthesis process of lincomycin might be under the control of AdpAlin.

qRT-PCR analysis showed that the transcript level of lmbA, lmbC, lmbD, lmbJ, lmbV, and lmbW dramatically decreased in 1adpA with fold changes 129.41, 43.64, 60.30, 215.29, 11.95, and 301.20, respectively (**Figure 2A**). It was suggested that in 1adpA, lincomycin biosynthesis was blocked because of the decreased expression of structural genes lmbA, lmbC, lmbD, lmbJ, lmbV, and lmbW. Due to the low transcript level of lmbK (data not shown), we failed to calculate the relative expression in both WT and 1adpA. To investigate the regulation between AdpAlin

and lmbK, we cloned the 895 bp of DNA sequences upstream from lmbK translation starting site (TSS) and constructed reporting plasmid for catechol dioxygenase activity assay. The results demonstrated that AdpAlin got involved in regulating the transcription of lmbK (**Figure 2B**). Besides, transcriptions of lmbA, lmbC, lmbD, lmbJ, lmbV, and lmbW were activated by AdpAlin in catechol dioxygenase activity assay as well (**Figure 2B**). These data suggested that AdpAlin has primary effects on the structural genes in the lmb cluster.

Then we carried out EMSAs to investigate the interplay between AdpAlin and promotors of above lmb structural genes. DNA fragments containing putative AdpA binding sites were labeled with biotin and incubated with purified AdpAlin. Results of EMSAs showed that AdpAlin interacted with all of the 8 promoter regions containing putative AdpA binding sites (**Figure 2C**). DNA probe lmbJp-A served as part of the promoter regions of both lmbJ and lmbK, and similarly, DNA probe lmbVp-A served as the promoter regions of both lmbV and lmbW. For promoters containing more than one AdpA binding sites, such as genes lmbC, lmbJ, lmbK, lmbV, and lmbW, AdpAlin interacted with different putative AdpA binding sites and generated different forms of complexes (**Figure 2C**). DNA fragment without AdpA binding site, i.e., probe-neg, could not interact with AdpAlin (**Supplementary Figure S3**). To confirm the exact binding sites of AdpAlin with promoters, we deleted the putative AdpA binding sites in lmbAp-A, lmbJp-A and lmbKp-B, and EMSAs showed AdpAlin no longer interacted with these DNA probes (**Supplementary Figure S4**). Thus, we speculated that AdpAlin activates all of the 8 promoters in the lmb cluster by directly binding to putative AdpA binding sites.

#### AdpAlin Directly Activates Transcription of the CSR Gene lmbU

LmbU was recently reported by Hou et al. (2018a, 2019) to be a novel transcriptional regulator cited in the lmb cluster and positively regulate lincomycin biosynthesis. Here, we investigated the regulatory relationship between AdpAlin and lmbU. We analyzed the 770 bp of promoter region upstream from lmbU TSS, and found 10 putative AdpA binding sites where two of them are overlapped (**Supplementary Figure S2F**). Relative expression of lmbU significantly decreased by 26.26 folds in 1adpA compared with WT (**Figure 3A**). We cloned the 770 bp of DNA sequences upstream from lmbU

TSS to construct reporter plasmid for catechol dioxygenase assay. In accordance with qRT-PCR results, catechol dioxygenase assay demonstrated that AdpAlin remarkably activated lmbU promoter (**Figure 3B**). As displayed in **Supplementary Figure S2F**, lmbUp-A, lmbUp-B, and lmbUp-C are three DNA probes containing putative AdpA binding sites in the promoter region of lmbU. EMSAs indicated that AdpAlin directly bound to lmbUp-A, lmbUp-B, and lmbUp-C, separately (**Figure 3C**). Then, the putative AdpA binding site in lmbUp-B was deleted, and EMSA showed that AdpAlin could not bind to this DNA probe (**Supplementary Figure S4**), suggesting that AdpAlin activates the transcription of lmbU by directly binding to the lmbU promoter and thus gets involved in the activation of lincomycin biosynthesis.

#### AdpAlin Directly Activates the Expression of the Global Regulator Gene bldA

As mentioned above, deletion of AdpAlin in S. lincolnensis not only blocked lincomycin biosynthesis, but also significantly impaired the generation of spores (**Figures 1B,C**), which suggested that AdpAlin had complicated connections with genes outside the lmb cluster. Hou et al. (2018b) previously identified that BldA regulates morphological differentiation and lincomycin biosynthesis in S. lincolnensis. We analyzed the 799 bp of promoter region upstream from the mature bldA, three putative AdpA binding sites were found (**Figure 4A**). Relative expression of bldA exhibited a 7.29-fold decreased in 1adpA compared with WT (**Figure 4B**). The 799 bp of DNA sequences upstream from the mature bldA was cloned to construct reporter plasmid for catechol dioxygenase activity assay. And the results indicated that AdpAlin significantly activates the transcription of bldA in vivo (**Figure 4C**). BldAp-A and bldAp-B are two DNA probes containing putative AdpA binding sites marked in **Figure 4A**. We performed EMSAs of AdpAlin with bldAp-A and bldAp-B separately, and as displayed in **Figure 4D**, AdpAlin directly bound to DNA fragments containing putative AdpA binding sites in vitro. Thus, we speculated that AdpAlin participates in the lincomycin biosynthesis in S. lincolnensis through activating the transcription of bldA, and indirectly mediates the morphological differentiation.

Since AdpA has been identified to regulate the transcription of bldA in S. griseus (Higo et al., 2011), we further analyzed the bldA promoter from S. lincolnensis and S. griseus. The data showed that putative AdpA binding site and its flanking sequence in bldAp-A was highly conserved (**Supplementary Figure S5A**), and EMSA of AdpAlin with bldAp-A confirmed the binding (**Figure 4D**). However, the sequences upstream bldAp-A are various between the two species, which contains the AdpA binding sites in S. griseus, but not in S. lincolnensis (**Supplementary Figure S5A**). Further EMSA demonstrated that AdpAlin cannot bind to this fragment bldAp-N (**Supplementary Figure S5B**), confirming that it is not a functional AdpA binding site in S. lincolnensis. Moreover, putative AdpA binding site in bldAp-B, which has not

been studied before, was confirmed to be a AdpA binding site in S. lincolnensis (**Supplementary Figure S5C**).

# AdpAlin Positively Autoregulates Its Own Transcription via Directly Binding to Its Own Promoter

We carried out a detailed analysis of 610 bp of upstream DNA sequences from AdpAlin TSS and found 6 putative AdpAlin binding sites where 4 of them are overlapped in pairs (**Figure 5A**). To investigate the transcriptional regulation between AdpAlin and its own transcription, we cloned the 610 bp of DNA sequence upstream from AdpAlin TSS to construct a reporter plasmid for catechol dioxygenase activity assay. The results revealed that AdpAlin slightly activated its own transcription in vivo (**Figure 5B**). As displayed in **Figure 5A**, adpAp-A and adpAp-B are two DNA probes containing putative AdpAlin binding sites, and the results of EMSAs verified that AdpAlin directly interacted with the promoter region of AdpAlin in vitro (**Figure 5C**).

AdpA also has been identified to autoregulate its own transcription in S. griseus (Kato et al., 2005), then we compared AdpAlin promoter from S. lincolnensis with adpA promoter from S. griseus. In S. lincolnensis, the putative AdpA binding site and its flanking sequence in adpAp-A was high conserved (**Supplementary Figure S6A**) and the result of EMSA verified the binding (**Figure 5C**). However, the nucleic acid sequences upstream adpAp-A were highly diverse (**Supplementary Figure S6A**). Then we designed two probes, adpAp-B-1 and adpAp-B-2, which contain the unconserved AdpA binding motifs from S. griseus (Kato et al., 2005) and putative AdpA binding sites from S. lincolnensis, respectively. The results of EMSAs showed that AdpAlin can bind to adpAp-B-2, but not to adpAp-B-1 (**Supplementary Figures S6B,C**), indicating that two new AdpA binding sites are found in S. lincolnensis, and the differences between the two probes may allow them to respond to different regulatory mechanisms.

# DISCUSSION

In this study, we elucidated the effect of AdpAlin on lincomycin biosynthesis and morphological differentiation. There are 8 putative operons in the lmb cluster. Before this study, none of a regulator has been identified to directly bind to all the eight promoters in the lmb cluster and have an overall impact on the entire lincomycin biosynthesis progress. By deciphering the regulations on the 8 putative operons, we attempted to envision the transcriptional regulatory network on lincomycin biosynthesis. Lincomycin biosynthesis contains three main parts: formation of α-methylthiolincosaminide (MTL), formation of propylproline (PPL), and condensation and final methylation (**Supplementary Figure S2A**). AdpAlin activates lmbKp and lmbVp, therefore we speculate that AdpAlin directly activates the transcription of lmbK, lmbR, lmbO, and lmbN, so that AdpAlin positively regulates the biosynthesis of MTL structure (Sasaki et al., 2012; Lin et al., 2014). AdpAlin activates the transcription of lmbAp, lmbWp, and lmbUp, which means lmbB1, lmbB2, lmbX and lmbY are also activated by AdpAlin, thus we speculated that AdpAlin directly regulates the biosynthesis of PPL structure (Novotna et al., 2004, 2013; Pang et al., 2015; Jiraskova et al., 2016). In addition, AdpAlin activates the transcription lmbCp, lmbDp, lmbJp, and lmbVp, which means transcriptions of lmbT, lmbE, lmbF, and lmbG are activated by AdpAlin (Hola et al., 2003; Zhao et al., 2015; Kadlcik et al., 2017; Zhong et al., 2017; Zhang et al., 2018). Therefore, we inferred that AdpAlin functions as a primary activator of lincomycin biosynthesis and regulates the entire biosynthetic process. And this is the first case that AdpA directly activates the transcription of the overall structural genes in such a complicated antibiotic biosynthetic gene cluster (BGC). In addition, EMSAs of AdpAlin with targets showed that AdpAlin binds to different binding

sites with different affinities (**Figure 2C**). Based on these results, some strategies of genetic manipulations may be proposed for hyper-production of lincomycin, such as mutation, deletion or addition of AdpA binding sites in the promoter regions of target genes.

Besides directly participating in lincomycin biosynthesis, as a pleiotropic regulator, AdpAlin controls lincomycin biosynthesis by regulating other transcriptional regulators as well. LmbU has been identified to activate the transcription of lmbA, lmbC, and lmbJ, and repress the transcription of lmbK and lmbU itself (Hou et al., 2018a). In this study, we confirmed that the transcription of lmbU was activated by AdpAlin (**Figure 3**). As described by Hou et al. (2018b) there is a UUA codon in lmbU, and translation of lmbU is controlled by BldA. Existence of rare codon means very small changes of the tRNA could induce the significant change of protein amount (Chater, 2006). Besides, UUA codons also exist in lmbB2 and lmbY in the lmb cluster, indicating that LmbB2 and LmbY might be important regulatory targets during lincomycin biosynthesis. Furthermore, it has been showed that the adpA gene contains a UUA codon as well, on the other hand, transcription of bldA is regulated by AdpAlin (**Figure 4**), which may function as a feedback regulatory mechanism to keep the organism in balance. In this study, we speculated that AdpAlin, LmbU, and BldA formed a regulatory cascade that mediate lincomycin biosynthesis in S. lincolnensis (**Figure 6**). In addition, considering AdpA responds to the GBL-involved cascade regulation (Healy et al., 2009; Tan et al., 2015), bioinformatics analysis was performed and two GBL-signaling systems were found in S. lincolnensis (**Figure 6**). One system consists of the GBL receptor SLINC\_6539 (GenBank accession number ANS68763.1) and biosynthetic enzyme SLINC\_6540 (GenBank accession number ANS68764.1) which were highly homology with many receptors and enzymes in Streptomyces, whereas SLINC\_6539 and SLINC\_6540 had identities with ArpA (47%) and AsfA (70%) in S. griseus, respectively. The other system consists of SLINC\_5093 (GenBank accession number WP\_067437987.1) and SLINC\_5094 (GenBank accession number WP\_067437989.1) which were low similarities with other receptors and enzymes, while SLINC\_5093 and SLINC\_5094 had identities with ArpA (44%) and AsfA (33%) in S. griseus, respectively. But how these two GBL-signaling systems works to affect lincomycin biosynthesis will be needed further research.

In S. griseus, AdpA auto-depresses its own transcription (Ohnishi et al., 2005), whereas in S. lincolnensis, AdpAlin has a positive impact on its own transcription (**Figure 5B**). Differences of AdpA binding sites in S. lincolnensis and S. griseus indicated that AdpA homologs from different resource have specialized regulatory mechanism on their own transcription. Catechol dioxygenase activity showed that AdpAlin had a negatively effect on the AdpA binding site near the AdpAlin TSS (Data not shown). Thus, we speculated that AdpAlin bound to this putative site and prevented the AdpAlin promotor from being activated.

FIGURE 5 | AdpAlin positively regulates its own transcription. (A) Nucleic acid sequence of AdpAlin promoter region where +1 represents the translation starting site (TSS) of adpA. Putative AdpA binding sites are marked in red color with gray background. Probes for EMSA, adpAp-A and adpAp-B, are framed. (B) Catechol dioxygenase activity analysis of WT and 1adpA transformed with pADPX. ∗∗P < 0.01. (C) EMSAs of AdpAlin with 5<sup>0</sup> -biotin labeled DNA probes adpAp-A and adpAp-B. Solid triangles point to the bands of probes and hollow triangles point to the complexes of AdpAlin and DNA fragments. Concentrations of AdpAlin are 0, 1.6, 3.2, 6.4, 6.4, and 6.4 µM, respectively. Competitive assays were carried out with a 200-fold excess of unlabeled specific probe adpAp-A or adpAp-B (lane S) or with a 200-fold excess of unlabeled non-specific probe probe-neg (lane N).

As for another putative AdpA binding sites, we presumed it might recruit RNA polymerase after interacting with AdpAlin and thus the overall effect of AdpAlin on its own promoter appeared to be positive. In the natamycin producer S. chattanoogensis, AdpAch was an activator of natamycin biosynthesis, and 6 AdpA binding sites were identified in the scnRI-scnRII intergenic region (Du et al., 2011; Yu et al., 2018). It is notable that although the general effect of AdpAch on the transcription of scnRI is positive, AdpA binding site A and B serve as repression sites. Thus, we speculated that the varying amounts and locations of AdpA binding sites in the promoter region of AdpAlin targets exhibited different effects and constituted a complicated and subtle regulatory network of AdpA regulons.

In summary, we reported a lesser-known case that AdpAlin interacted with all of the 8 putative operons and activated the transcription of structural genes in the lmb cluster. Furthermore, we deduced AdpAlin, LmbU, and BldA in cascade regulation that controlled lincomycin biosynthesis. Based on these knowledge, more efforts should be devoted to complete the regulatory mechanism of lincomycin biosynthesis and to enhance to production of lincomycin.

#### DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the manuscript/**Supplementary Files**.

#### REFERENCES


# AUTHOR CONTRIBUTIONS

YK, BH, and HW designed the experiments. YK, YW, and RW carried out the experiments. YK, BH, XZ, JY, and HW analyzed the data. YK, JY, and HW wrote the manuscript. JY and HZ guided the work. All authors assisted with critical reading of the manuscript.

#### FUNDING

This study was supported by the National Natural Science Foundation of China (NSFC) (31900059), the China Postdoctoral Science Foundation (2019M650079), and the Research Program of State Key Laboratory of Bioreactor Engineering.

# ACKNOWLEDGMENTS

We thank Dr. Weihong Jiang for providing the plasmid for catechol dioxygenase activity assay and all the members of our laboratory.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.02428/full#supplementary-material




**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Kang, Wang, Hou, Wang, Ye, Zhu, Wu and Zhang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Recent Advances in Synthetic Biology Approaches to Optimize Production of Bioactive Natural Products in Actinobacteria

Lei Li<sup>1</sup>† , Xiaocao Liu1,2† , Weihong Jiang1,3 \* and Yinhua Lu<sup>4</sup> \*

<sup>1</sup> Key Laboratory of Synthetic Biology, CAS Center for Excellence in Molecular Plant Sciences, Institute of Plant Physiology and Ecology, Chinese Academy of Sciences, Shanghai, China, <sup>2</sup> School of Life Sciences, Henan University, Kaifeng, China, <sup>3</sup> Jiangsu National Synergetic Innovation Center for Advanced Materials, SICAM, Nanjing, China, <sup>4</sup> College of Life Sciences, Shanghai Normal University, Shanghai, China

#### Edited by:

Rustam Aminov, University of Aberdeen, United Kingdom

#### Reviewed by:

Linquan Bai, Shanghai Jiao Tong University, China Bertolt Gust, University of Tübingen, Germany

#### \*Correspondence:

Weihong Jiang whjiang@sibs.ac.cn Yinhua Lu yhlu@shnu.edu.cn; yhlu@shnu.edu †These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 01 September 2019 Accepted: 15 October 2019 Published: 05 November 2019

#### Citation:

Li L, Liu X, Jiang W and Lu Y (2019) Recent Advances in Synthetic Biology Approaches to Optimize Production of Bioactive Natural Products in Actinobacteria. Front. Microbiol. 10:2467. doi: 10.3389/fmicb.2019.02467 Actinobacteria represent one of the most fertile sources for the discovery and development of natural products (NPs) with medicinal and industrial importance. However, production titers of actinobacterial NPs are usually low and require optimization for compound characterization and/or industrial production. In recent years, a wide variety of novel enabling technologies for engineering actinobacteria have been developed, which have greatly facilitated the optimization of NPs biosynthesis. In this review, we summarize the recent advances of synthetic biology approaches for overproducing desired drugs, as well as for the discovery of novel NPs in actinobacteria, including dynamic metabolic regulation based on metabolite-responsive promoters or biosensors, multi-copy chromosomal integration of target biosynthetic gene clusters (BGCs), promoter engineering-mediated rational BGC refactoring, and construction of genome-minimized Streptomyces hosts. Integrated with metabolic engineering strategies developed previously, these novel enabling technologies promise to facilitate industrial strain improvement process and genome mining studies for years to come.

Keywords: actinobacteria, natural product, synthetic biology, dynamic regulation, BGC amplification, pathway refactoring, genome-minimized host

#### INTRODUCTION

As a kind of Gram-positive bacteria with high GC content, actinobacteria undergo complex morphological differentiation and secondary metabolism processes (Barka et al., 2016). Actinobacteria, particularly the Streptomyces genus, have been recognized as the main sources for microbial bioactive natural products (NPs), such as antibiotics, chemotherapeutics, immunosuppressants and anthelmintics, which make important contributions to health care and crop protection (Jensen et al., 2015; Genilloud, 2017; Nepal and Wang, 2019). With the rapid advances in genome sequencing and genome mining methods (Ziemert et al., 2016; Blin et al., 2019), a wealth of hidden NP biosynthetic gene clusters (BGCs) have been revealed by specialized software (i.e., antiSMASH) and are regarded as an untapped treasure trove for the discovery of novel bioactive compounds (Rutledge and Challis, 2015; Niu, 2018). For instance, a single Streptomyces genome usually harbors around 30 NP BGCs, approximately 10-fold more than

**21**

previously identified by bioactivity screening (Rutledge and Challis, 2015). However, the majority of NP BGCs in actinobacteria are silent or cryptic under laboratory culture conditions, and must be activated for the isolation and characterization of unknown compounds (Rutledge and Challis, 2015; Zarins-Tutt et al., 2016). Furthermore, production titers of many available actinomycete-derived drugs are still low for the economically viable industrial bioprocess. Construction of highly efficient microbial cell factories becomes increasingly critical for commercial application of desired NPs (Kim et al., 2016; Zhang et al., 2016).

With the advents of metabolic engineering and synthetic biology, genetic engineering of actinobacteria could address several major challenges associated with NP discovery, development and large-scale manufacturing. In the last three decades, a variety of genetic engineering strategies have been developed for strain development, including precursor engineering, BGC amplification, deletion of competing pathways, engineering of translational/transcriptional machineries as well as manipulation of pleiotropic/pathway-specific regulators (Baltz, 2016; Kim et al., 2016; Palazzotto et al., 2019). Interested readers are referred to the detailed reviews of systems biotechnology of actinobacteria (Weber et al., 2015; Baltz, 2016; Kim et al., 2016; Zhang et al., 2016; Palazzotto et al., 2019). It is worth noting that with breakthroughs in CRISPR-based genome editing methods, a series of novel enabling technologies have been developed, greatly facilitating the engineering of actinomycetal genomes (i.e., deletion, insertion, replacement and point mutation) as well as NP BGCs (i.e., cloning, editing, deletion and amplification) (Li et al., 2017a; Tong et al., 2018; Alberti and Corre, 2019). In this review, we briefly summarize the most recent synthetic biology approaches and discuss how these technologies enable the generation of microbial cell factories and the discovery of novel therapeutic drug leads. These include dynamic regulation based on metabolite-responsive elements, multiplex site-specific recombination (SSR) system-mediated BGC amplification (MSGE), systematical and rational BGC refactoring, as well as construction of genome-minimized Streptomyces hosts for NP overproduction and discovery. Together with traditional metabolic engineering strategies (Weber et al., 2015; Baltz, 2016), we believe that these newly developed tools will be widely applicable for actinobacteria, providing general strategies for (meta)genome mining-based novel NP discovery as well as for the overproduction of commercially important NPs.

# DYNAMIC PATHWAY REGULATION BASED ON METABOLITE-RESPONSIVE PROMOTERS OR REGULATORS

Dynamic metabolic regulation has proved to be an effective strategy to improve production titers of target compounds by balancing bacterial growth and biosynthesis of specific metabolites (Zhang et al., 2012; Dahl et al., 2013; Gupta et al., 2017). Generally, three different approaches – quorum sensing systems, metabolite-responsive promoters and protein/RNA-based biosensors – are used for autonomous control of metabolic pathway flux (Zhang et al., 2015; Polkade et al., 2016; Tan and Prather, 2017). These new concepts were initially developed in Escherichia coli or Saccharomyces cerevisiae, and the latter two have been extended to actinobacteria for the optimization of antibiotic biosynthetic pathways.

# Metabolite-Responsive Promoters

In the last three decades, a large variety of metabolic engineering strategies have been developed to optimize production of secondary metabolites in microbes. However, few approaches could enable the coordination between bacterial growth and biosynthesis of target compounds. Recently, Li et al. (2018) employed time-course transcriptome analysis to identify a series of antibiotic-responsible promoters with a transcription profile similar to the inducible promoters, when under the optimal conditions (**Figure 1A**). These dynamic responsive promoters could be used to efficiently optimize the expression of native actinorhodin and heterogeneous oxytetracycline (OTC) BGCs in Streptomyces coelicolor, subsequently improving the production titers of ACT and OTC by 1.3- and 9.1-fold, respectively, compared with constitutive promoters (Li et al., 2018). Xu et al. (2012) reported a metabolite-responsive promoter, which controlled the transcription of actAB encoding an antibiotic exporter in S. coelicolor. They found that the antibiotic ACT and its biosynthetic intermediates [e.g., (S)- DNPA] could relieve the repression of actAB by binding the transcriptional regulator ActR. That means that the actAB promoter could indirectly respond to intermediates or endproducts, thus synergistically regulating ACT biosynthesis and export. The metabolite-responsible promoter based strategy will achieve an autonomous induction of pathway regulation and provide a universal route for titer improvements of desired NPs in actinobacteria.

# Natural Product-Specific Biosensors

Genetically encoded biosensors are a kind of important synthetic biology tool for real-time detection of intracellular metabolites for specific readouts (Zhang et al., 2015). A typical biosensor is composed of three modules: (i) a signal input module, such as transcription factors (TFs) or riboswitches; (ii) a regulatory module, such as TF-dependent promoters; (iii) a signal output module, such as reporter genes (Mahr and Frunzke, 2016; Lim et al., 2018). Until now, three major categories of biosensors - fluorescence resonance energy transfer (FRET)-based sensors, TF-based sensors and riboswitches – have been widely used in metabolic engineering of microbial hosts (Zhang et al., 2015; Lim et al., 2018). As a kind of natural sensory protein, a classical TF has evolved to contain a ligand-binding domain for responding to environmental changes and a DNA-binding domain for regulating gene expression. Interestingly, many NP BGCs encode cluster-situated regulators (CSRs) in actinobacteria, such as TetRlike regulators and Streptomyces antibiotic regulatory proteins (SARPs; Liu et al., 2013; Romero-Rodriguez et al., 2015). These CSRs can be used to dynamically detect developmental state, population density or other environmental changes, and

FIGURE 1 | Dynamic pathway regulation balances flux between bacterial growth and antibiotic biosynthesis in actinobacteria. (A) Autoregulated fine-tuning of the expression of an antibiotic (e.g., oxytetracycline) BGC based on metabolite-responsive native promoters without specific transcription factors or additional inducers. BGC, biosynthetic gene cluster. (B) Antibiotic-responsive biosensors. This strategy is engineered from cluster-situated regulators (e.g., PamR2) that bind to their associated promoters upon interaction with the corresponding secondary metabolites (e.g., pamamycins). Placing antibiotic biosynthetic genes under such promoters enables antibiotic dependent gene expression.

thus determine the onset and production levels of secondary metabolites in actinobacteria. Recently, Rebets et al. (2018) developed a new antibiotic-specific whole-cell biosensor based on a TetR-like repressor for the development of antibioticoverproducing strains (**Figure 1B**). Briefly, the highly active macrodiolide antibiotic pamamycins BGC encodes a transporter PamW and a corresponding transcriptional repressor PamR2, which can be deactivated by binding to pamamycins (Rebets et al., 2015). Generally, PamW expression is controlled by PamR2 at low level of pamamycins. However, at very high levels of pamamycins, the native PamR2-based biosensors will reach the detection limit and cannot effectively regulate PamW expression (Rebets et al., 2018). The pamW promoter was directly used to control expression of the kanamycin resistance gene, thus generating the resistance-based G0 whole-cell biosensor system. After UV-induced mutagenesis, strains resistant to a highconcentration of kanamycin showed a significant increase in pamamycins production (up to 15–16 mg/L). Of note is the fact that the native pamamycins-responsive biosensor showed limited operating and dynamic ranges for further applications.

To overcome this obstacle, the G1 pamamycins biosensor was further developed by combining different promoters, varying the number and position of operators, as well as using diverse reporter genes. Using the new biosensor with higher sensitivity, three mutated strains were obtained, which could produce up to 30 mg/L of pamamycins. Furthermore, to overcome the low detection limit of the G1 biosensor, the binding affinity of PamR2-pamamycins was also efficiently decreased by designing a panel of PamR2 mutations. As expected, the G2 biosensor showed a better operating dynamic range than that of the G1 biosensor (Rebets et al., 2018). In fact, at least 17% of NP BGCs encode TetR-like regulators and putative transporters simultaneously, which provides possibilities for the development of diverse antibiotic-responsive biosensors (Rebets et al., 2018). With the aid of a metabolite-responsive biosensor system, it will become very convenient to assess the functional expression of NP BGCs after random mutagenesis, genetic manipulation or exposure to various cultivation conditions. In the future, TF-based biosensors will hold great promise for accelerating cell factory development for pharmaceutical production and the activation of silent BGCs for the discovery of novel compounds in actinobacteria (Kim et al., 2016; Zarins-Tutt et al., 2016; Sun et al., 2017).

### MULTIPLEX SITE-SPECIFIC GENOME ENGINEERING FOR NPs OVERPRODUCTION

In the last three decades, due to their broad host specificities, SSR systems have been widely applied to strain improvements, combinatorial biosynthesis, and heterologous expression of the entire BGCs in Streptomyces, Actinoplanes and other industrial actinobacteria (Baltz, 2012; Stark, 2017; Merrick et al., 2018). To date, at least ten SSR systems have been identified in actinobacteria, which are derived from bacteriophage CBG73463, R4, SV1, TG1, VWB, 8BT1, 8C31, 8Joe, 8K38-1 or 81/6 (Baltz, 2012; Fogg et al., 2017; Li et al., 2017a; Kormanec et al., 2019). In particular, 8BT1, 8C31 and TG1 systems have been widely used for cell factory development via multi-copy amplification of target genes or BGCs (Baltz, 2012). For instance, production titers of the antibiotic goadsporin were increased by 2.3-fold by a step-by-step introduction of two extra copies of the goadsporin BGC based on 8C31 and TG1 integration systems (Haginaka et al., 2014). However, this strategy requires repeated rounds of conjugal transfer and is limited by the number of selection markers. In 2011, Murakami and coworkers demonstrated that the ZouA-RsA/B-mediated recombination system could be used to achieve tandem amplification of the ACT gene cluster with 4–12 copies, resulting in a 20 fold increase in ACT production (Murakami et al., 2011). However, the engineered strains may be genetically unstable in the absence of antibiotic selection due to the presence of tandem amplification. In addition, the method requires the introduction of the ZouA-RsA/B system into both flanks of target BGCs in advance via two-round conjugal transfer before BGC amplification, which is time-consuming and labor-intensive. To address these limitations, our group has recently developed two novel enabling technologies for multi-locus chromosomal integration of target genes or BGCs, including MSGE and aMSGE (Li et al., 2017b, 2019b).

#### MSGE: Multiplex Site-Specific Genome Engineering Based on the "One Integrase-Multiple attB Sites" Concept

In the actinomycetal genome, there is typically no, or only one, native attB site for each SSR system (Baltz, 2012). In theory, one-step, multi-copy integration of target genes or BGCs could be achieved by introducing multiple artificial attB sites into the host chromosome in advance. Based on this "one integrasemultiple attB sites" concept, an innovative approach MSGE was developed for discrete BGC amplification (**Figure 2A**; Li et al., 2017b). Using the high-efficiency CRISPR/Cas9 genome editing method (Huang et al., 2015), we sequentially introduced three 8C31 and two 8BT1 attB sites into the genomic loci of deleted, non-target secondary metabolite BGCs in the industrial strain Streptomyces pristinaespiralis. Then, five extra copies of the pristinamycin II BGC were integrated into the modified chromosome in two steps using the 8C31 and 8BT1 compatible integration systems, which led to significantly improved PII titers (Li et al., 2015, 2017b). Importantly, the novel strategy was also extended to develop a series of powerful Streptomyces coelicolor heterologous expression hosts. Up to four copies of the chloramphenicol or YM-216391 BGCs were simultaneously integrated into these new chassis strains, thus resulting in increased production titers (Li et al., 2017b). Using the highly effective heterologous expression system, YM-216391 BGC was engineered to generate aurantizolicin and a hybrid compound 3, which exhibits significantly increased antitumor activity (Pei et al., 2018). Similarly, Myronovskyi et al. (2018) also constructed a panel of cluster-free, powerful Streptomyces albus chassis strains based on the "one integrase-multiple attB sites" concept. The production titers of a variety of bioactive compounds, including aloesaponarin II, cinnamycin, didesmethylmensacarcin and griseorhodin A, were significantly enhanced in these heterologous expression superhosts (Myronovskyi et al., 2018). It is worth noting that compared with the ZouA-RsA/B recombination system, MSGE-based engineered strains would be genetically stable in the absence of antibiotic selection due to discrete, sitespecific integration of target BGCs (Li et al., 2017b). However, when the MSGE method is used to amplify target BGCs with multiple copies in a single step, there is an upper limit to the number of integrated BGCs. For example, we found that the chloramphenicol or YM-216391 BGCs could be simultaneously inserted into S. coelicolor with up to 4 copies. The possible reason is that the high-order amplification of BGCs or the accumulation of target products places an excess burden on bacterial growth (Li et al., 2017b). In the near future, we believe that these versatile Streptomyces hosts will greatly accelerate (meta)genomic mining and combinatorial biosynthesis studies for novel bioactive NP discovery (Li et al., 2017a; Liu et al., 2018; Zhang et al., 2019).

# aMSGE: Advanced Multiplex Site-Specific Genome Engineering With Orthogonal Modular Recombinases

Although the MSGE method could be widely used to construct a variety of suitable chassis organisms for NP discovery and overproduction, two major limitations remains to be addressed. On the one hand, it will be difficult or even impossible to initially insert foreign attB sites into the chromosomes of genetically intractable industrial actinobacteria due to the lack of available replicative plasmids and also their low homologous recombination capability, such as spinosadproducing Saccharopolyspora spinosa and epoxomicin-producing Goodfellowiella coeruleoviolacea. On the other hand, repeated introduction of multiple artificial attB sites is still timeconsuming and labor-intensive, especially for the slow-growing actinobacteria. To overcome these bottlenecks, our group recently developed an advanced MSGE method (aMSGE) based on the "multiple integrases-multiple attB sites" concept (**Figure 2B**; Li et al., 2019b). In this improved method, native attB sites of different orthogonal SSR systems in the actinomycetal genome are simultaneously applied to multicopy integration of target genes or BGCs, rather than introducing foreign attB sites into the host chromosome. Accordingly, a plug-and-play amplification toolkit was designed and constructed, which contains 27 modular recombination plasmids with single or multiple SSR systems. Using this innovative technique, we achieved a high-efficiency introduction of the 5-oxomilbemycin A3/A4 BGC into the parental strain Streptomyces hygroscopicus with up to four extra copies, thus resulting in a significant increase in the production titers of 5-oxomilbemycin A3/A4 (Li et al., 2019b). Compared with previously developed metabolic engineering tools, the aMSGE method doesn't require the introduction of any genetic modifications before target gene or BGC amplification, which will considerably simplify and accelerate efforts to facilitate NP discovery and overproduction. The whole process for BGC amplification takes only ∼18 days for the construction of high-yield engineered strains (e.g., growth period is ∼6 days). More importantly, the aMSGE method could be applicable to genetically intractable actinobacteria for strain improvements. Given that SSR systems are widely distributed in a variety of microorganisms (Fogg et al., 2014; Stark, 2017), our newly developed methodology should be widely extended to establish more efficient industrial platforms for overproducing valuable chemicals and drugs. However, the aMSGE method could not be used for BGC amplification in actinobacteria without native attB sites.

# RATIONAL PATHWAY REFACTORING OF NP BIOSYNTHESIS

Generally, NP biosynthesis is under the control of highly complicated transcriptional, translational and metabolic regulation, which hampers the ability of systematic BGC engineering to maximize biosynthetic efficiency (Liu et al., 2013). Pathway refactoring provides an effective synthetic biological approach to decouple gene expression from complex native regulation and to achieve precise control of metabolite production by redesigning target BGCs in versatile surrogate hosts (Tan and Liu, 2017; Niu and Li, 2019). The key steps of BGC refactoring are to develop a set of well-characterized genetic control elements and high-efficiency DNA assembly methods (**Figure 3**).

In the last ten years, a series of synthetic regulatory elements have been identified to precisely regulate gene expression with wide dynamic ranges in actinobacteria, including constitutive or inducible promoters, ribosomal binding sites (RBSs), terminators and protein degradation tags (Myronovskyi and Luzhetskyy, 2016; Horball et al., 2018; Ji et al., 2019). For instance, the expression levels of 200 native or synthetic promoters and 200 artificial RBSs were systematically quantitated, which provides a universal toolbox of synthetic modular regulatory elements for the scalable and cost-effective optimization of NP biosynthetic pathway in different Streptomyces (Bai et al., 2015).

Next, a variety of in vitro or in vivo DNA assembly methods have been established to reconstruct target BGCs for diverse applications (Li et al., 2017a; Zhang et al., 2019). On the one hand, a panel of in vivo DNA assembly methods for single-marker or marker-free multiplexed promoter engineering of large BGCs have been developed on the basis of powerful homologous recombination capacity in S. cerevisiae, including DNA assembler (Shao et al., 2011), mCRISTAR (multiplexed Cas9 transformation-associated recombination) (Kang et al., 2016) and miCRISTAR (multiplexed in vitro Cas9-transformationassociated recombination) (Kim et al., 2019). For example, using the miCRISTAR strategy, the activation of a silent BGC led to the characterization of two bacterial cyclic sesterterpenes atolypene A and B (Kim et al., 2019). In another study, yeastmediated construction of a riboswitch-controlled pathway achieved a 120-fold increase in bottromycin productivity (Eyles et al., 2018). Intriguingly, multiple new bottromycin-related metabolites were also generated by using high-efficiency, flexible BGC modifications. On the other hand, a range of in vitro DNA assembly methods suitable for pathway refactoring have been developed, including modified Gibson assembly (Li et al., 2015), MASTER (methylation-assisted tailorable ends rational) ligation (Chen et al., 2013) and SLIC (Sequenceand Ligation-Independent Cloning) (D'Agostino and Gulder, 2018). Recently, an innovative DNA assembly method, ExoCET (Exonuclease Combined with RecET recombination), was also developed for large-size, multi-operon BGC refactoring (Song et al., 2018; Wang et al., 2018). The artificial 79-kb spinosad BGC with 7 artificial operons under the control of strong constitutive promoters achieved a 328-fold enhanced spinosad production compared to the native BGC (Song et al., 2018). As a simple and robust genetic platform, BGC refactoring will have potentially broad applications in combinatorial biosynthesis and antibiotic overproduction, as well as highthroughput activation of silent BGCs from either metagenomes or uncultured actinobacteria.

# CONSTRUCTION OF GENOME-REDUCED Streptomyces HOSTS FOR NPs DISCOVERY AND OVERPRODUCTION

Streptomyces species are capable of producing a wide range of secondary metabolites, such as polyketides, non-ribosomal peptides and terpenes, and possess the unique metabolic background needed for heterologous expression of NP BGCs from actinobacteria or other bacteria with high-GC content (Liu et al., 2018; Nepal and Wang, 2019). Within the streptomycetes, S. coelicolor, S. albus J1074 and Streptomyces lividans have been widely used as surrogate hosts for NPs discovery and overproduction (Beites and Mendes, 2015; Liu et al., 2018). With the continuous advances of novel bioinformatics tools and genetic manipulation techniques (Zhang et al., 2016, 2019;

hosts for activating silent BGCs or enhancing production titers of clinically important drugs. BGC, biosynthetic gene cluster.

Ziemert et al., 2016), a series of versatile Streptomyces chassis have been developed by deleting non-essential genomic regions, such as redundant BGCs, genome islands and insertion sequences (Liu et al., 2018; Huo et al., 2019; **Table 1**). Particularly, deletion of redundant BGCs are predicted to increase the supply of primary metabolite precursors, remove competing carbon and nitrogen sinks, and also facilitate the biosynthesis of heterologous BGCs. In addition, engineered strains without redundant BGCs will possess simple extracellular metabolite profiles for convenient structural characterization of novel bioactive compounds.

Along with the emergence of genome mining as a robust approach to discover novel drug leads, the development of reliable Streptomyces chassis for heterologous expression of cloned BGCs is becoming increasingly important (Huo et al., 2019; Nepal and Wang, 2019). As a model strain, S. coelicolor has been widely used to study the molecular regulation of antibiotic biosynthesis and morphological differentiation (Liu et al., 2013;


#### TABLE 1 | Characteristics of synthetic model Streptomyces chassis.

fmicb-10-02467 November 1, 2019 Time: 17:32 # 7

ACT, actinorhodin; CDA, calcium-dependent antibiotic; CPK, cryptic type I polyketide; RED, prodiginine; BGC, biosynthetic gene cluster.

Barka et al., 2016). In the last 10 years, a series of advanced genome editing tools and diverse synthetic regulatory elements have been developed for the optimization of heterologous pathways in S. coeliclor (Myronovskyi and Luzhetskyy, 2016; Alberti and Corre, 2019). In 2011, on the basis of the wild-type strain S. coelicolor M145, two versatile engineered hosts M1146 and M1152 were developed, which contained deletions in four BGCs, responsible for the biosynthesis of ACT, calciumdependent antibiotic (CDA), cryptic type I polyketides (CPK) and prodiginine (RED), or in combination with a point mutation in rpoB (Gomez-Escribano and Bibb, 2011). The productivity of the chloramphenicol or congocidine BGCs was significantly increased compared with the parental strain (Gomez-Escribano and Bibb, 2011). Nowadays, M1146 and M1152 have been widely used to heterologously express different types of compounds from fastidious original producers or metagenomic DNA (Huo et al., 2019; Nepal and Wang, 2019). Similarly, a genomereduced host S. lividans SBT5 was also developed by deleting BGCs responsible for ACT, CDA and RED as well as inserting the regulatory gene afsRScla (Bai et al., 2014). Furthermore, to specifically facilitate the study of Type III polyketides, an effective surrogate S. coelicolor host M1317 was constructed by removing all three endogenous Type III polyketide synthase (PKS) genes (gcs, srsA, rppA) on the basis of M1152, which could potentially increase precursor supply and prevent undesirable interference with expression of heterologous Type III PKS genes (Thanapipatsiri et al., 2015). As another important Streptomyces chassis, S. albus J1074 provides high success rates of heterologous BGC expression with rapid growth and high productivity (Nepal and Wang, 2019). Recently, on the basis of S. ablus J1074, 15 endogenous secondary metabolite BGCs were deleted to generate a powerful engineered host Del14 (Myronovskyi et al., 2018). The production yields of heterologously expressed metabolites in S. albus Del14 were higher than those in commonly used S. albus J1074 and S. coelicolor M1152. Furthermore, the authors also introduced 1-2 artificial 8C31 attB sites into the genomes of S. albus Del14, thus resulting in two another versatile engineered hosts, B2P1 and B4. The two powerful chassis could be easily used for BGC amplification of up to four copies, thus efficiently improving fermentation levels of the encoded compounds (Myronovskyi et al., 2018).

In addition to the construction of a model heterologous expression chassis, the genome-reduced strategy has been frequently used to develop non-model Streptomyces strains for the discovery and overproduction of important NPs (**Table 2**). Streptomyces sp. FR-008 is a fast-growing, potentially industrial production chassis, which produces the macrolide candicidin and other bioactive compounds. By deleting three endogenous polyketide genes, a mutant strain LQ3 was constructed with a stable and streamlined genome structure, which possibly allows for simple separation and purification of heterologously expressed compounds (Liu et al., 2016). As a producer of the anti-infective avermectin, Streptomyces avermitilis is already optimized to efficiently supply primary precursors. To construct a versatile industrial chassis for heterologous expression of secondary metabolite BGCs, a series of genomereduced S. avermitilis mutants were obtained by deleting the left subtelomeric region (∼2 Mb) that corresponds to the more variable genome regions (Komatsua et al., 2010). Herein, the SUKA5 and SUKA17 strains were highlighted, which had genome reductions of 17.9 and 18.5%, respectively. S. avermitilis SUKA5 strain has the deletions of both oligomycin BGC and the left subtelomeric region, and the SUKA17 strain is a derivative of SUKA5 with the additional deletions of three terpene BGCs (Komatsua et al., 2010). The feasibility and superiority of these two engineered hosts has been widely confirmed by the efficient production of more than 20 exogenous secondary metabolite BGCs (Komatsu et al., 2013; Ikeda et al., 2014). Similarly, 1.3-Mb and 0.7-Mb possible non-essential genomic regions were deleted in the natamycin-producing industrial strian

#### TABLE 2 | Characteristics of synthetic non-model Streptomyces chassis.

fmicb-10-02467 November 1, 2019 Time: 17:32 # 8


ACT, actinorhodin; BGC, biosynthetic gene cluster; PKS, polyketide synthase gene; NRPS, non-ribosomal peptide synthase gene; NA, not available.

Streptomyces chattanoogensis L10 by Cre-loxP recombination system, respectively, thus generating two efficient chassis L320 and L321 for the production of valuable polyketides (Bu et al., 2019). In another study, a newly engineered host ZXJ-6 was developed based on a salinomycin-producing industrial strain, Streptomyces albus BK3-25, by deleting the salinomycin BGC and chromosomal integration of a three-gene cassette for the biosynthesis of ethylmalonyl-CoA. It was successfully used for the heterologous expression of polyketide BGCs, such as the ACT BGC from S. coelicolor (Zhang et al., 2017). In the future, the genome-reduced strategy could be widely implemented in a variety of industrial actinobacteria for the improved production of various bioactive compounds.

# CONCLUDING REMARKS

Actinobacteria, especially bacteria from the genus Streptomyces, have long been employed as an important source of a wide range of novel bioactive small molecules (Barka et al., 2016; Newman and Cragg, 2016; Genilloud, 2017). To harness the production potential of actinobacteria, a variety of innovative metabolic engineering and synthetic biology strategies have been developed in the last 10 years, including dynamic metabolic regulation, BGC amplification, pathway refactoring and genomeminimized Streptomyces chassis (Beites and Mendes, 2015; Tan and Liu, 2017; Tan and Prather, 2017; Li et al., 2019a). We envision that NP discovery and development will be rapidly accelerated by the refactoring and amplification of whole biosynthetic pathways in combination with powerful heterologous expression platforms. Considering that a large number of BGCs are generally silent or expressed at very low levels, it is still important to develop diverse heterologous hosts and universal refactoring approaches to activate silent BGCs or boost production of secondary metabolites. Next, quorum sensing systems are widely distributed in actinobacteria (Polkade et al., 2016), which provide opportunities for bacterial growth-mediated dynamic regulation of metabolic pathways for enhanced production of target compounds. Furthermore, due to the complex regulation between primary and secondary metabolism, iterative application of the design-build-test-learn cycle will be necessary to overproduce different secondary metabolites (Liu et al., 2013; Kim et al., 2016; van der Heul et al., 2018). Finally, the advancements in the fields of genome sequencing, multi-omics and genome editing techniques are paving the way for systems metabolic engineering of industrial actinobacteria, including pathway engineering, regulatory circuit rewiring, host modification and enzyme engineering (Kim et al., 2016; Zhang et al., 2019). Given the requirement for high titers in commercial production, an integrated approach involving traditional mutagenesis screening and rational host/pathway engineering is required to systematically optimize the biosynthesis of target compounds. It is expected that these synthetic biology tools and metabolic engineering strategies presented in this review and future developments will play increasingly important roles in the discovery of novel drug leads, as well as yield improvement for large-scale manufacturing in actinobacteria.

# AUTHOR CONTRIBUTIONS

LL and XL wrote the manuscript. LL, WJ, and YL reviewed, edited, and approved its final version.

# FUNDING

We gratefully acknowledge the financial support from the National Mega-project for Innovative Drugs (2018ZX09711001- 006-012), the National Natural Science Foundation of China (31970083, 31770088, and 31630003), and the Science and Technology Commission of Shanghai Municipality (18ZR1446700).

# REFERENCES

fmicb-10-02467 November 1, 2019 Time: 17:32 # 9


cluster-specific secondary metabolites. ACS Synth. Biol. 6, 1026–1033. doi: 10. 1021/acssynbio.6b00353


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Li, Liu, Jiang and Lu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

fmicb-10-02467 November 1, 2019 Time: 17:32 # 10

# Regulation of Antibiotic Production by Signaling Molecules in Streptomyces

Dekun Kong1,2, Xia Wang1,2, Ju Nie1,3 and Guoqing Niu1,2 \*

<sup>1</sup> Biotechnology Research Center, Southwest University, Chongqing, China, <sup>2</sup> State Cultivation Base of Crop Stress Biology for Southern Mountainous Land, Academy of Agricultural Sciences, Southwest University, Chongqing, China, <sup>3</sup> College of Horticulture and Landscape Architecture, Southwest University, Chongqing, China

The genus Streptomyces is a unique subgroup of actinomycetes bacteria that are well-known as prolific producers of antibiotics and many other bioactive secondary metabolites. Various environmental and physiological signals affect the onset and level of production of each antibiotic. Here we highlight recent findings on the regulation of antibiotic biosynthesis in Streptomyces by signaling molecules, with special focus on autoregulators such as hormone-like signaling molecules and antibiotics themselves. Hormone-like signaling molecules are a group of small diffusible signaling molecules that interact with specific receptor proteins to initiate complex regulatory cascades of antibiotic biosynthesis. Antibiotics and their biosynthetic intermediates can also serve as autoregulators to fine-tune their own biosynthesis or cross-regulators of disparate biosynthetic pathways. Advances in understanding of signaling molecules-mediated regulation of antibiotic production in Streptomyces may aid the discovery of new signaling molecules and their use in eliciting silent antibiotic biosynthetic pathways in a wide range of actinomycetes.

Keywords: antibiotic biosynthesis, regulation, Streptomyces, hormone-like signaling molecule, antibiotic biosynthetic intermediate, elicitor

# INTRODUCTION

The genus Streptomyces, a unique subgroup of actinomycetes bacteria, is best-known for their ability to produce an enormous variety of bioactive secondary metabolites including antibiotics. The onset and production level of each antibiotic is subject to complex control by various environmental and physiological signals (Liu et al., 2013; Niu et al., 2016). Several excellent and comprehensive reviews have focused on the roles of signaling molecules in the regulation of antibiotic production of Streptomyces species (Takano, 2006; Sidda and Corre, 2012; Niu et al., 2016; Okada and Seyedsayamdost, 2017). Hormone-like signaling molecules are a group of small diffusible signaling molecules that can elicit antibiotic production and/or induce morphological differentiation at nanomolar concentrations. Typically, a hormone-like signaling molecule binds to its specific receptor, and exerts regulatory function through regulators at different hierarchical levels, including global regulators, pleiotropic regulators, and cluster-situated regulators (CSRs) (Liu et al., 2013; Niu and Tan, 2015; Wei et al., 2018). The most-studied regulatory system is the A-factor cascade that involves the signaling molecule A-factor and its receptor ArpA necessary for streptomycin and grixazone production in Streptomyces griseus

Edited by:

Yinhua Lu, Shanghai Normal University, China

#### Reviewed by:

Matthew Ian Hutchings, University of East Anglia, United Kingdom Weishan Wang, Institute of Microbiology (CAS), China

\*Correspondence:

Guoqing Niu niu062376@swu.edu.cn

#### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 29 October 2019 Accepted: 05 December 2019 Published: 19 December 2019

#### Citation:

Kong D, Wang X, Nie J and Niu G (2019) Regulation of Antibiotic Production by Signaling Molecules in Streptomyces. Front. Microbiol. 10:2927. doi: 10.3389/fmicb.2019.02927

(Horinouchi, 2007; Horinouchi and Beppu, 2007). Aside from the hormone-like signaling molecules, accumulating evidence suggests that antibiotics and their biosynthetic intermediates can also serve as autoregulators to modulate their own biosynthesis and as cross-regulators of disparate biosynthetic pathways (Niu et al., 2016). This review is not intended to be comprehensive, but to highlight recent findings on the regulation of antibiotic biosynthesis in Streptomyces by small molecules including hormone-like signaling molecules and antibiotics/their biosynthetic intermediates. A better understanding of the regulatory cascade of antibiotic production by signaling molecules in Streptomyces may aid the discovery of new signaling molecules and speed up their application in eliciting silent antibiotic biosynthetic pathways in actinomycetes.

### REPRESENTATIVE HORMONE-LIKE SIGNALING MOLECULES CONTROLLING ANTIBIOTIC BIOSYNTHESIS

Hormone-like signaling molecules are diffusible small molecules that can elicit antibiotic production and/or induce morphological differentiation at nanomolar concentrations. They exert their regulatory effect via specific receptor proteins that usually belong to the TetR family of transcriptional regulators (Takano, 2006; Niu et al., 2016). The TetR family regulators are characterized by a N-terminal helix-turn-helix (HTH) DNA-binding motif and a C-terminal ligand regulatory domain. They are widely distributed among bacteria and regulate diverse cellular processes in bacteria, especially antibiotic biosynthesis in Streptomyces (Cuthbertson and Nodwell, 2013). In Streptomyces species, A-factor, the first signaling molecule to be discovered, induces streptomycin production and morphological differentiation through a regulatory cascade involving the receptor ArpA, the pleiotropic regulator AdpA, and the CSR activator StrR in S. griseus (Horinouchi, 2007; Horinouchi and Beppu, 2007). Inspired by studies on the A-factor regulatory cascade, many more hormone-like signaling molecules have been identified, and great efforts has been directed to understanding signaling molecules-mediated regulatory cascades of antibiotic biosynthesis in Streptomyces.

#### Five Major Classes of Hormone-Like Signaling Molecules

Similar to the well-documented N-acyl-homoserine lactones (AHLs) in Gram-negative bacteria, hormone-like signaling molecules in the Gram-positive Streptomyces bacteria are diffusible low molecular weight chemical substances that can elicit antibiotic production and/or induce morphological differentiation at nanomolar concentrations (Swift et al., 1994; Takano, 2006). To date, hormone-like signaling molecules identified from Streptomyces are classified into five major classes, including γ-butyrolactones (GBLs), furans, γ-butenolides as well as PI factor and N-methylphenylalanyl-dehydrobutyrine diketopiperazine (MDD) (Niu et al., 2016). GBLs, furans and γ-butenolides are based on five-membered heterocyclic rings containing four carbons and one oxygen, while PI factor and MDD have quite different structures (Recio et al., 2004; Matselyukh et al., 2015). As most recent investigations are largely on GBLs and γ-butenolides, we herein only summarize recent findings on these two important classes. GBLs, the largest group of these signaling molecules, share a characteristic 2,3-disubstituted GBL core skeleton but differ in the length, branching and stereochemistry of the acyl side chain (Takano, 2006). So far, a total of 19 GBLs have been identified in streptomycetes, including the A-factor from S. griseus, eight butanolides (SCB1-8) from Streptomyces coelicolor, five virginiae butanolides (VBs A–E) from Streptomyces virginiae, IM-2 from Streptomyces lavendulae, factor 1 from Streptomyces viridochromogenes, and three Gräfe's factors from Streptomyces bikinensis and Streptomyces cyaneofuscatus (**Figure 1**; Niu et al., 2016; Sidda et al., 2016). To gain insight to the distribution of GBLs, an A-factor-deficient mutant of S. griseus was used as an indicator strain to screen against 203 actinomycete strains. Thirty of the strains tested showed distinct A-factor activity, indicating that A-factor was widely distributed among actinomycetes (Hara and Beppu, 1982). Earlier studies identified the butenolide synthase AfsA as a key enzyme for A-factor biosynthesis in S. griseus, and the AfsA homolog ScbA was required for the biosynthesis of the three SCBs in S. coelicolor (Takano et al., 2001; Hsiao et al., 2007). Among the eight butanolides identified in the model organism S. coelicolor, SCB1-3 were identified in the 2000s (Takano et al., 2000; Hsiao et al., 2009). Recently, the engineered strain S. coelicolor M1152 was found to overproduce GBLs SCB1–3 as well as five novel GBLs designated as SCB4– 8 (Sidda et al., 2016). The S. coelicolor M1152 is a derivative of S. coelicolor M145 lacking four biosynthetic gene clusters (BGCs) for the biosynthesis of actinorhodin (ACT), undecylprodigiosin (RED), calcium-dependent antibiotic (CDA) and coelimycin. It should be noted that the deletion of the coelimycin gene cluster included scbR2, which encodes a repressor of SCBs biosynthetic gene scbA, and thereby contributed to the overproduction of SCBs in S. coelicolor M1152 (Gomez-Escribano and Bibb, 2012; Sidda et al., 2016). In a BLAST survey, we found one to three afsAlike genes in each of nine Streptomyces genomes (Niu et al., 2016). In another study, BLAST search for homologs of AfsA from S. griseus within the actinobacteria genomes available from public databases showed that AfsA-like proteins are present in most actinomycetes (Ahmed et al., 2017). The widespread distribution of GBLs was further confirmed by using VB- and IM-2-deficient strains as indicators to examine VB and IM-2 distribution. A total of 10 strains of the 40 Streptomyces and 11 endophytic actinomycetes tested produce IM-2, and the same number of strains produced VB active compounds (Thao et al., 2017). These studies reinforce the notion that GBLs are the most common hormone-like signaling molecules among actinomycetes. More efforts should be directed to identifying chemical structures of these signaling molecules.

Representative γ-butenolides include two butenolides (SRB1 and SRB2) from Streptomyces rochei (Arakawa et al., 2012; Arakawa, 2014), the avenolide from Streptomyces avermitilis (Kitani et al., 2011), four butenolides from Streptomyces albus

J1074 (Nguyen et al., 2018), and three butenolides (SAB1-3) from Streptomyces ansochromogenes (Wang et al., 2018; **Figure 1**). Of special note is the avenolide that is necessary for triggering avermectin production in S. avermitilis (Kitani et al., 2011). The gene cluster responsible for avenolide biosynthesis consists of three genes (avaR1-3) that encode GBL receptor homologs, and two genes (aco and cyp17) that encode an acyl-CoA oxidase and a cytochrome P450 hydroxylase, respectively. Studies suggested that both the acyl-CoA oxidase Aco and the cytochrome P450 hydroxylase Cyp17 were required for the enzymatic production of avenolide. Furthermore, Aco/Cyp17 homologs were also found in the genomes of Streptomyces fradiae, Streptomyces ghanaensis, Streptomyces griseoauranticus, and S. albus J1074, suggesting that they have the capacity to produce avenolide or avenolidelike autoregulators (Kitani et al., 2011; Ahmed et al., 2017). Recently, Luzhetskyy and colleagues identified an avenolide-like compound 4 in a genetically modified strain of S. albus 1074 (Ahmed et al., 2017). In another study, Nihira and colleagues investigated the distribution of avenolide by using a S. avermitilis aco disruptant as a biosensor in a collection of 40 Streptomyces and 11 endophytic actinomycetes. They showed that 12 of the 51 actinomycetes strains exhibited avenolide activity with S. albus J1074 showing the highest activity (Thao et al., 2017). Metabolite profiling of a disruptant of the S. albus aco gene led to the identification of the compound 4 along with three other butenolides (compounds 2, 3, and 5). The four avenolidelike compounds showed different levels of avenolide activity in stimulating avermectin production in S. avermitilis (Nguyen et al., 2018). It is interesting to note that the four compounds have been isolated previously from marine-derived Streptomyces strains, though there are no reports on their connection with antibiotic biosynthesis in the native producers (Cho et al., 2001; Strand et al., 2014; Viegelmann et al., 2014; Igarashi et al., 2015). In a recent study, Tan and colleagues identified a putative BGC for

autoregulator biosynthesis in S. ansochromogenes (nikkomycin producer). Within the gene cluster, sabA encodes an AfsAlike enzyme, whereas sabP and sabD encode phosphatase and dehydrogenase enzymes, respectively. Heterologous expression of sabAPD in E. coli and Streptomyces led to the identification of three novel butenolides (SAB1, 2, and 3) (Wang et al., 2018). Since there are only limited numbers of natural GBLs and γ-butenolides identified in Streptomyces, these studies provide effective strategies for the discovery of new signaling molecules, which are normally produced in very small quantities. It is noteworthy that several butenolides have also been identified in Streptomyces species (Wang T. et al., 2014; de Oliveira et al., 2019), especially from marine-derived Streptomyces (Cho et al., 2001; Strand et al., 2014; Viegelmann et al., 2014; Igarashi et al., 2015). The roles of these butenolides in their native producers await further investigation.

### Receptors of Hormone-Like Signaling Molecules

The transmission of chemical signals starts with the binding of an autoregulator to its specific receptor protein, that usually belongs to the TetR family of transcriptional regulators. It is not uncommon that many Streptomyces genomes contain multiple genes for ArpA-like GBL receptors. Examples include ScbR, ScbR2, CprA and CprB in S. coelicolor, JadR2 and JadR3 in Streptomyces venezuelae, and AvaR1, AvaR2, and AvaR3 in S. avermitilis, SabR1 and SabR2 in S. ansochromogenes. It is noteworthy that most genes encoding these GBL receptors are closely linked both to each other and to the afsA-like genes, except that cprA and cprB are not linked to an afsA-like gene and therefore referred to as "orphan GBL receptor" genes (Onaka et al., 1998). To date, only a few identified receptors have been shown to interact with endogenous GBL molecules. In the model organism S. coelicolor, ScbR was characterized as a genuine receptor of SCBs (Takano et al., 2005; Gomez-Escribano et al., 2012). ScbR binds the scbR-scbA intergenic region to repress its own expression, while interacting with ScbA to activate scbA expression in response to SCBs (Takano et al., 2001). Another known target of ScbR is kasO, the CSR activator of the coelimycin biosynthetic pathway (Takano et al., 2005). In S. venezuelae, JadR3 was identified as a genuine receptor of SVB1, a GBL identical in structure to SCB3 of S. coelicolor. The SVB1 receptor JadR3 has dual activation and repression effects on jadomycin biosynthesis (Zou et al., 2014; Niu et al., 2016). In S. ansochromogenes, SabR1, the cognate receptor of SABs, was shown to repress the expression of sabA, sabR1, sabR2 and cprC by binding directly to their promoter regions. Interestingly, CprC, a homolog of CprA and CprB in S. coelicolor, was identified as an activator of the pleiotropic regulatory gene adpA, which in turn stimulated nikkomycin biosynthesis via activating the transcriptional initiation of the

activate the transcription of sanF-X through an unknown mediator.

CSR activator gene sanG and biosynthetic genes of the sanF-X operon (**Figure 2**; Pan et al., 2009; Wang et al., 2018). Of the three GBL receptors in S. avermitilis, AvaR1 acts as an avenolide receptor to control avenolide and avermectin production by directly repressing the transcription of aco, aveR, its own gene and the other two GBL receptor homologous genes (avaR2 and avaR3) (**Figure 3**; Zhu et al., 2017). ScbR2, JadR2, AvaR2 and SabR2 were identified as pseudo GBL receptors (Xu et al., 2010; Zhu et al., 2016; Wang et al., 2018), and the roles of these pseudo GBL receptors in the regulation of antibiotic biosynthesis will be discussed in a subsequent section. AvaR3 contains an extra 75 amino acid stretch that is not present in typical GBL receptors, and promotes avermectin production through a yet-unknown regulatory mechanism (Miyamoto et al., 2011). It is interesting to note that AvaR3 may represent a new subgroup of GBL receptors, and therefore further studies are needed to reveal its role in avermectin production.

# ANTIBIOTICS AS REGULATORY LIGANDS REGULATING ANTIBIOTIC BIOSYNTHESIS

Other than the hormone-like signaling molecules described above, antibiotics and their biosynthetic intermediates can also serve as regulatory ligands to regulate their own biosynthesis or cross-regulators of disparate biosynthetic pathways. In jadomycin biosynthesis, jadomycin and its biosynthetic intermediates actively participate in both feedback and feedforward cascade control of their own production in S. venezuelae. Such interplay makes jadomycin biosynthesis an illuminating model for regulatory studies of antibiotic production mediated by antibiotics and their biosynthetic intermediates (Niu et al., 2016). Aside from roles in regulating their own biosynthesis, antibiotics can also serve as autoregulators to control disparate antibiotic biosynthetic pathways. Typically, antibiotics exert their regulatory effect through CSRs that are associated with BGCs or pleiotropic regulators situated elsewhere in the genome (**Table 1**).

# Antibiotics and Their Biosynthetic Intermediates as Autoregulators of Their Own Biosynthesis

The first line of antibiotics as autoregulators came from studies on the roles of two atypical response regulators (ARRs) in antibiotic biosynthesis. The CSR activator JadR1, an OmpR-type ARR of S. venezuelae, activates the expression of jadomycin B (JdB) biosynthetic genes in the presence of a low concentration of JdB and its homologs, but high JdB concentrations cause the dissociation of JadR1 from its target promoters. Thus, JdB interacts with JadR1 directly in a dose-dependent manner to control JdB production in a feedback regulatory mechanism (Wang et al., 2009). Similarly, a NarL-type ARR RedZ was shown to promote RED biosynthesis in S. coelicolor by activating the CSR activator gene redD, and the DNA-binding activity of RedZ to its target gene was released by adding RED (Wang et al., 2009). Other than the end-product, studies also suggest that biosynthetic intermediates of antibiotics can also serve as autoregulatory molecules to modulate their own biosynthesis. For example, the TetR family regulator JadR<sup>∗</sup> can respond to early jadomycin intermediates [2,3-dehydro-UWM6 (DHU) and dehydrorabelomycin (DHR)] and release its repression on jadY

FIGURE 3 | Cascade regulation of avermectin biosynthesis. The complex regulation of avermectin production involves three GBL receptors. The avenolide receptor AvaR1 inhibits avenolide and avermectin production by directly repressing the transcription of aco, aveR, avaR1, avaR2, and avaR3. The pseudo GBL receptor AveR2 also acts as a repressor of avenolide and avermectin biosynthesis by binding to the same targets as AveR1. Another GBL receptor AvaR3 promotes avermectin production through an unknown regulatory mechanism. The DNA binding activities of AvaR1 and AvaR2 are modulated by avenolide, which are the enzymatic product of Aco and Cyp17.



JdB, jadomycin B; JdA, jadomycin A; DHU, 2,3-dehydro-UWM6; DHR, dehydrorabelomycin; RED, undecylprodigiosin; CHL, chlorothricin; DM-CHL, demethylsalicycloyl chlorothricin; deschloro-CHL, deschloro-chlorothricin; NOS, nosiheptide; NOS-AC, the intermediate of NOS; 2,4-DAPG, 2,4 diacetylphloroglucinol; MAPG, monoacetylphloroglucinol; ACT, actinorhodin; Cm, chloramphenicol.

transcription to ensure the timely supply of cofactors for JadG to convert DHR to the late intermediate jadomycin A (JdA) in jadomycin biosynthesis (Zhang et al., 2013).

Inspired by studies on the regulation of jadomycin biosynthesis in S. venezuelae, more antibiotics and their biosynthetic intermediates have been found to serve as autoregulatory molecules to modulate their own biosynthesis in many other streptomycetes. For illustration purposes, we herein only highlight several recent findings. For example, ChlF1, a TetR family regulator involved in chlorothricin biosynthesis of Streptomyces antibioticus, was found to repress the transcription of chlF1, chlG (encoding a major facilitator superfamily transporter), and chlK (encoding a type II thioesterase) but activates chlJ (encoding acyl-CoA carboxyl transferase) by binding to their promoter regions. However, ChlF1 was disassociated from its target promoters in the presence of chlorothricin and its biosynthetic intermediates (demethyl salicyloyl chlorothricin and deschloro-chlorothricin), which directly interacted with the regulator in a concentrationdependent manner (Li et al., 2016). In another study, CalR3, a TetR family regulator involved in calcimycin biosynthesis of Streptomyces chartreusis NRRL 3882, was found to repress the transcription of calR3 and its adjacent calT (encoding a putative transmembrane efflux pump protein of the MMPL family) by binding to their promoter regions. Similarly, both calcimycin and its biosynthetic intermediate (cezomycin) can act as ligands to dissociate CalR3 from its target promoters in a concentration-dependent manner (Gou et al., 2017; Wu et al., 2018). Other than TetR family regulators, Streptomyces antibiotic regulatory proteins (SARPs) family regulator was also found to regulate antibiotic biosynthesis in response to antibiotics and their biosynthetic intermediates. For example, NosP was found to regulate nosiheptide production in response to both the end-product nosiheptide (NOS) and biosynthetic intermediate (NOS-AC) (Li et al., 2018). It should be noted that such findings are not limited to the genus Streptomyces. One example comes from the actinomycete Amycolatopsis mediterranei, in which rifamycin B (the endproduct of rifamycin biosynthesis) serves as an extracellular signaling molecule to regulate rifamycin export in a feedback mechanism (Lei et al., 2018). Another study showed that 2,4-diacetylphloroglucinol and its biosynthetic intermediate could serve as autoregulatory ligands to regulate its own biosynthesis in Pseudomonas fluorescens (Yan et al., 2017). These studies suggested that antibiotics and their biosynthetic intermediates could function as autoregulators to regulate their own biosynthesis, mainly through the modulation of the binding activity of CSRs to their target genes within their cognate gene clusters.

### Antibiotics as Cross-Regulators of the Biosynthesis of Other Antibiotics

Earlier studies suggest that CSRs of one biosynthetic pathway can also control the biosynthesis of disparate antibiotic biosynthetic pathways. For example, RedZ, CSR of the red gene cluster, modulates the production of RED as well as that of ACT and CDA (Huang et al., 2005). In another study, the candicidin CSR FscRI was found to control the biosynthesis of candicidin as well as antimycin, the product of a disparate BGC in Streptomyces albidoflavus S4 (previously known as Streptomyces albus S4) (McLean et al., 2016; Li et al., 2019). Similar observations were also made with coordinated production of geldanamycin and elaiophylin by GdmRIII in Streptomyces autolyticus CGMCC0516 (Jiang et al., 2017), production of cephamycin C and clavulanic acid by CcaR in Streptomyces clavuligerus (Santamarta et al., 2002), and production of jadomycin and chloramphenicol by JadR1 in S. venezuelae (Wang et al., 2009). Recent years see growing evidence that antibiotics can also act as autoregulators regulating other biosynthetic pathways. The first examples to be described involve pseudo GBL receptors. In S. coelicolor, ScbR2, represses the same target, kasO, as the SCBs-interactive ScbR. Although scbR2 is linked to the gene cluster for coelimycin biosynthesis, ScbR2 also responds to the endogenous antibiotics ACT and RED, and thereby regulates the production of these antibiotics (Xu et al., 2010). In S. venezuelae, JadR2, a close homolog of ScbR2, modulates JdB biosynthesis via direct repression of jadR1. This repression is sensitive to the jadomycin end products. Likewise, although jadR2 is linked to the jad gene cluster for jadomycin biosynthesis, JadR2 could also

bind to the endogenous antibiotic chloramphenicol, the product of the distant cml biosynthetic gene cluster (Xu et al., 2010). These studies suggest that CSRs of one biosynthetic pathway can also control disparate antibiotic biosynthetic pathways in response to different antibiotics. The pseudo GBL receptor AvaR2 of S. avermitilis was as an important repressor of avermectin and avenolide biosynthesis. It directly repressed transcription of aveR, aco, its own gene and the other two GBL receptor homologous genes (avaR1 and avaR3) (**Figure 3**; Zhu et al., 2016). Interestingly, DNA-binding activity of AvaR2 can be modulated by endogenous avenolide in a concentrationdependent manner, indicating that avenolide serves as ligands for both the genuine GBL receptor AvaR1 and the pseudo GBL receptor AvaR2. Furthermore, DNA-binding activity of AvaR2 can also be modulated by exogenous antibiotics JdB and aminoglycoside antibiotics such as apramycin, hygromycin B, kanamycin and streptomycin (Zhu et al., 2016). It would be interesting to examine if JdB from S. venezuelae can serve as signals to modulate avermectin production in S. avermitilis under physiological conditions.

### SIGNALING MOLECULES IN THE IMPROVEMENT OF ANTIBIOTIC TITERS AND THE CONSTRUCTION OF GENETIC CIRCUITS

As mentioned above, signaling molecules serve as elicitors for antibiotic production in Streptomyces. It is reasonable that signaling molecules can be used to improve antibiotic production through exogenous addition. However, due to the fact that natural GBLs are produced in very small quantities, it is impractical to collect sufficient amount of GBLs for the improvement of antibiotic production. Synthetic GBLs and their analogs are then used for this purpose. Early studies showed that addition of chemically synthesized VB-C enhanced virginiamycin production in Streptomyces virginiae (Yang et al., 1995, 1996). Similarly, addition of chemically synthesized SCB1 promoted ACT production in S. coelicolor. Synthetic GBL analogs are also an alternative (Yang et al., 2009). In two recent studies, the addition of synthetic 1,4-butyrolactone enhanced validamycin A production in Streptomyces hygroscopicus 5008 and bitespiramycin production in Streptomyces spiramyceticus WS1-195, respectively (Tan et al., 2013; Gao et al., 2019). These studies showed that exogenous addition of signaling molecules can be used as an effective strategy to increase metabolic titers of antibiotics. However, this method has limited application due to the fact that there are only a few natural and synthetic analogs available. Therefore, it is necessary to expand the reservoir of both natural and synthetic signaling molecules. Furthermore, there is a growing interest in the construction of GBL-based genetic circuits in heterologous systems, which has been summarized by Takano and colleagues in 2015 (Biarnes-Carrera et al., 2015). Recently, orthogonal regulatory circuits based on the S. coelicolor GBL system have been constructed in E. coli (Biarnes-Carrera et al., 2018), reinforcing the promising applications of GBL systems from Streptomyces as a regulatory tool for synthetic biology in heterologous systems.

# SIGNALING MOLECULES IN THE DISCOVERY OF NOVEL NATURAL PRODUCTS

High-throughput DNA sequencing technologies have resulted in an explosion of microbial genome sequences. Genome sequencing of multiple well-known actinomycetes has revealed that they harbor a great number of BGCs that are predicted to direct the biosynthesis of diverse natural products (Nett et al., 2009). It has become clear that these enormous reservoir of uncharacterized BGCs can serve as an important source of novel bioactive compounds. However, most of these BGCs are not expressed efficiently or not at all under routine laboratory culturing conditions (Mao et al., 2018). Though the reason for the silence of these gene clusters is complicated and remains obscure, a lack of specific signaling molecules may be one of the contributing factors (Liu et al., 2013). Interestingly, antibiotics can be used as chemical elicitors to trigger the expression of these cryptic BGCs and expand the chemical diversity of natural products.

#### Antibiotics as Chemical Elicitors for the Discovery of Novel Natural Products

In recent years, growing evidence suggests that antibiotics, at subinhibitory concentrations, can potentiate antibiotic production in multiple Streptomyces species. For example, lincomycin at a sub-inhibitory concentration resulted in an elevated expression of the CSR activator gene actII-ORF4, and therefore increased ACT overproduction in S. coelicolor (Imai et al., 2015). A recent study suggests that lincomycin produced profound changes in gene expression profiles of S. coelicolor (Ishizuka et al., 2018). In another example, a sub-inhibitory concentration of JdB (the angucycline from S. venezuelae) induced early RED production and premature differentiation (formation of sporulating aerial mycelium) in S. coelicolor. Other angucyclines were also found to elicit similar phenotypes (Wang W. et al., 2014). Examples also include perturbation of antibiotics production by ribosome-targeting antibiotics (thiostrepton, spectinomycin, and chloramphenicol) in multiple Streptomyces species (Tanaka et al., 2017; Wang et al., 2017). Furthermore, antibiotics can also be used as chemical elicitors for the discovery of novel natural products. For example, S. lividans 1326 grown in the presence of lincomycin at a sub-inhibitory concentration produced abundant antibacterial compounds that were absent in cells grown in lincomycin-free medium. Some of these antibacterial compounds were revealed as novel congeners of CDA (Imai et al., 2015). In another case, polyether antibiotics, including promomycin and closely related salinomycin, monensin, and nigercin, at sub−inhibitory concentrations elicited antibiotic production in multiple Streptomyces strains (Amano et al., 2010, 2011). These studies suggest that antibiotics have the potential to

serve as chemical elicitors for the discovery of novel bioactive natural products.

# High-Throughput Elicitor Screening for the Discovery of Novel Natural Products

It is well-known that the cryptic gene clusters represent a treasure trove of new natural products, and various strategies and methods has been devised for the activation of these gene clusters (Liu et al., 2013; Niu et al., 2016; Niu, 2018; Niu and Li, 2019). Of special note is a chemogenetic method referred to as high-throughput elicitor screening approach (HiTES) (**Figure 4**). In this approach, a reporter gene is inserted into the BGC of interest to allow a rapid read-out for its expression. The resulting reporter strain is then screened against small molecule libraries to identify candidate elicitors (Seyedsayamdost, 2014). This approach was initially applied to activate two cryptic gene clusters in the Gram-negative bacterium Burkholderia thailandensis, leading to the identification of a new malleilactone analog. Interestingly, almost all elicitors identified from a library of 800 compounds were antibiotics (Seyedsayamdost, 2014). Furthermore, Seyedsayamdost and colleagues have employed this approach to identify 14 novel specialized products in S. albus J1074 (Xu et al., 2017). HiTES with S. albus J1074 using a library of ∼500 natural products identified two classical antibiotics (ivermectin and etoposide) as the best elicitors of a cryptic non-ribosomal peptide synthetase (NRPS) gene cluster (Xu et al., 2017). A major limitation of this method is that genetic manipulations and/or molecular biology approaches are required for the insertion of the reporter gene. To circumvent this problem, HiTES is then combined with bioactivity assays or imaging mass spectrometry (IMS) to identify novel antibiotics. In these approaches, a wild-type microorganism is subjected to HiTES. The resulting induced cultures are then screened directly for biological activity or subject to IMS (**Figure 4**). Application of these modified methods have uncovered two

novel specialized products in Streptomyces hiroshimensis (Moon et al., 2019b), and many other novel specialized metabolites in diverse microorganisms (Moon et al., 2019a; Xu et al., 2019). These studies suggest that HiTES represents a promising new avenue for the discovery of novel bioactive natural products in a high-throughput manner.

#### CONCLUSION AND PERSPECTIVES

Over the past several decades, significant advances have been made in understanding the regulation of antibiotic biosynthesis in Streptomyces. The emerging picture shows a complex interplay of various signals and regulatory proteins. It is interesting to note that both hormone-like signaling molecules and antibiotics/their biosynthetic intermediates actively participate in the regulation of antibiotic production. These small molecule regulators exert their function through the modulation of the binding activity of different regulators to their target genes. Further studies are needed to reveal the complex interplay of autoregulators and multiple GBL receptor homologs in antibiotic biosynthesis of different Streptomyces. However, the number of hormonelike signaling molecules identified from Streptomyces is still limited, most likely due to the fact that they are produced in very small quantities. BLAST searches for homologs of AfsA from S. griseus within the actinobacteria genomes available from public databases indicated that AfsA-like proteins are present in most actinomycetes (Niu et al., 2016; Ahmed et al., 2017). Similar observations have been made with a BLAST search for homologs of Aco and Cyp17 proteins from S. avermitilis (Ahmed et al., 2017). It is strongly believed that GBLs and γ-butenolide autoregulators are widely distributed among actinomycetes. It is therefore necessary to identify more hormone-like signaling molecules. Currently, several approaches have been used to discover new signaling molecules in Streptomyces, including deletion of gene encoding repressor of signaling molecule biosynthetic gene (Sidda et al., 2016), bioassay using disruptants of known signaling molecule biosynthetic genes as indicator strains (Thao et al., 2017; Nguyen et al., 2018), and heterologous

### REFERENCES


expression of putative signaling molecule biosynthetic genes in model microorganisms (Wang et al., 2018). These approaches will speed up the process of identifying new signaling molecules, and their signal transduction pathways. For autoregulators of antibiotics and their biosynthetic intermediates, more efforts should be geared toward understanding their effect on the finetuning regulation of antibiotic biosynthesis under physiological conditions. A better understanding of the complex regulatory pathways of signaling molecules in Streptomyces may be used to increase the metabolic titer of industrially and medically important antibiotics as well as activate silent antibiotic BGCs for the discovery of novel bioactive natural products in a wide range of actinomycetes.

# AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

# FUNDING

This work was supported in part by the National Natural Science Foundation of China (31870061), grants from the Chongqing Science and Technology Commission (cstccxljrc201904, cstc2018jcyjAX0066, and cstc2017jcyjAX0467), a startup grant from the Southwest University (SWU117015), and the Fundamental Research Funds for the Central Universities (SWU119034).

#### ACKNOWLEDGMENTS

We would like to thank all members of our lab for their helpful comments and discussion during the preparation of this manuscript. We also thank the two reviewers for their helpful comments. We sincerely apologize to those authors whose work we could not cite due to space limitations.


biosynthesis. Bioprocess Biosyst. Eng. 42, 1903–1913. doi: 10.1007/s00449-019- 02184-9


production superhost Streptomyces M1152. Org. Biomol. Chem. 14, 6390–6393. doi: 10.1039/c6ob00840b


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Kong, Wang, Nie and Niu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# New Molecular Tools for Regulation and Improvement of A40926 Glycopeptide Antibiotic Production in *Nonomuraea gerenzanensis* ATCC 39727

*Oleksandr Yushchuk1 , Andres Andreo-Vidal1 , Giorgia Letizia Marcone1 , Mervyn Bibb2 , Flavia Marinelli1 \* and Elisa Binda1*

*1Department of Biotechnology and Life Sciences, University of Insubria, Varese, Italy, 2Department of Molecular Microbiology, John Innes Centre, Norwich Research Park, Norwich, United Kingdom*

#### *Edited by:*

*Yvonne Mast, German Collection of Microorganisms and Cell Cultures GmbH (DSMZ), Germany*

#### *Reviewed by:*

*Max Julian Cryle, Monash University, Australia Marta Vaz Mendes, University of Porto, Portugal*

*\*Correspondence: Flavia Marinelli flavia.marinelli@uninsubria.it*

#### *Specialty section:*

*This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology*

*Received: 11 November 2019 Accepted: 06 January 2020 Published: 21 January 2020*

#### *Citation:*

*Yushchuk O, Andreo-Vidal A, Marcone GL, Bibb M, Marinelli F and Binda E (2020) New Molecular Tools for Regulation and Improvement of A40926 Glycopeptide Antibiotic Production in Nonomuraea gerenzanensis ATCC 39727. Front. Microbiol. 11:8. doi: 10.3389/fmicb.2020.00008*

Genome sequencing has revealed that *Nonomuraea* spp. represent a still largely unexplored source of specialized metabolites. *Nonomuraea gerenzanensis* ATCC 39727 is the most studied representative species since it produces the glycopeptide antibiotic (GPA) A40926 – the precursor of the clinically relevant antibiotic dalbavancin, approved by the FDA in 2014 for the treatment of acute skin infections caused by multi-drug resistant Gram-positive pathogens. The clinical relevance of dalbavancin has prompted increased attention on A40926 biosynthesis and its regulation. In this paper, we investigated how to enhance the genetic toolkit for members of the *Nonomuraea* genus, which have proved quite recalcitrant to genetic manipulation. By constructing promoter-probe vectors, we tested the activity of 11 promoters (heterologous and native) using the GusA reporter system in *N. gerenzanensis* and in *Nonomuraea coxensis*; this latter species is phylogenetically distant from *N. gerenzanesis* and also possesses the genetic potential to produce A40926 or a very similar GPA. Finally, the strongest constitutive promoter analyzed in this study, *aac(3) IVp*, was used to overexpress the cluster-situated regulatory genes controlling A40926 biosynthesis (*dbv3* and *dbv4* from *N. gerenzanensis* and *nocRI* from *N. coxensis*) in *N*. *gerenzanensis*, and the growth and productivity of the best performing strains were assessed at bioreactor scale using an industrial production medium. Overexpression of positive pathway-specific regulatory genes resulted in a significant increase in the level of A40926 production in *N. gerenzanensis*, providing a new knowledge-based approach to strain improvement for this valuable glycopeptide antibiotic.

Keywords: A40926, *Nonomuraea*, glycopeptide antibiotics, pathway-specific regulators, strain improvement

# INTRODUCTION

Research on glycopeptide antibiotics (GPAs) – drugs of "last resort" for treating severe infections caused by multi-drug resistant Gram-positive pathogens – has experienced a "renaissance" over the last decade (Marcone et al., 2018). Clinically important GPAs include two natural products (vancomycin and teicoplanin) and three second generation antibiotics (telavancin, dalbavancin, and oritavancin), which are semisynthetic derivatives of natural products endowed with an increased antimicrobial potency and superior pharmacokinetic properties. The urgent need for new potent antibiotics has driven much recent interest in GPAs. For example, chemical synthesis recently resulted in the generation of a plethora of vancomycin derivatives with novel modifications that show superior antimicrobial activities (Okano et al., 2017; Wu and Boger, 2019), while teicoplanin has been conjugated with nanoparticles resulting in increased activity against biofilm-forming pathogens (Armenia et al., 2018).

In contrast, genetic manipulation of GPA producers to yield novel potent derivatives is in its infancy. Recent work (Haslinger et al., 2015; Peschke et al., 2017; Schoppet et al., 2019) has revealed new details of the specificity and timing of non-ribosomal peptide synthesis, including chlorination and cross-linking steps, suggesting that the use of combinatorial biosynthesis to generate GPAs with completely novel oligopeptide scaffolds should be possible. In parallel, heterologous expression of enzymes involved in later stages of GPA biosynthesis (glycosylation, sulfation, acylation etc.) in known producers or *in vitro* has already generated novel GPA derivatives that could not be prepared easily by chemical synthesis (Banik and Brady, 2008; Banik et al., 2010; Yim et al., 2014).

In the meantime, genome sequencing has revealed the organization of GPA biosynthetic gene clusters (BGCs) in industrially valuable actinobacteria (D'Argenio et al., 2016), including long known GPA-producers (Vongsangnak et al., 2012; Truman et al., 2014; Kusserow and Gulder, 2017; Nazari et al., 2017; Adamek et al., 2018) as well as in novel producing strains (Thaker et al., 2013; Stegmann et al., 2014). Although the global regulation of GPA biosynthesis is still largely unexplored, the pathway-specific regulation controlling the expression of BGCs is being elucidated in model systems (Bibb, 2013). The roles of cluster-situated regulatory genes have been investigated in *Amycolatopsis balhimycina* (Shawky et al., 2007), *Nonomuraea gerenzanensis* (Lo Grasso et al., 2015; Alduina et al., 2018), and *Actinoplanes teichomyceticus* (Horbal et al., 2014b; Yushchuk et al., 2019), producing balhimycin, A40926 (the natural precursor of dalbavancin), and teicoplanin, respectively. Overexpression of the teicoplanin cluster-situated regulatory genes (*tei15\**, coding for a StrR-like transcriptional regulator, and *tei16\**, coding for a LuxR*-*type regulator) in *A. teichomyceticus* markedly increased teicoplanin production in the wild type strain, representing one of the most successful examples of using molecular tools for improving antibiotic production (Horbal et al., 2012, 2014b). In this work, we investigated the potential of molecular tools to improve the production of A40926, the dalbavancin precursor.

The BGC for A40926, named *dbv*, contains two genes encoding transcriptional regulators: *dbv3* and *dbv4* (Lo Grasso et al., 2015). *dbv3* encodes a LuxR-type regulator, which, however, is non-orthologous to the *tei* cluster encoded LuxR-regulator – Tei16\* (Yushchuk et al., 2019). *dbv4* codes for a StrR-like transcriptional regulator with close homologues in every GPA BGC (Yushchuk et al., 2019). The A40926-producing strain, recently re-classified as *N. gerenzanensis* (Dalmastri et al., 2016), belongs to a still poorly investigated genus of actinobacteria that was only recently identified as an untapped source of novel antibiotics and other bioactive metabolites (Sungthong and Nakaew, 2015). More recently, fully sequenced genomes of *N. gerenzanensis* (D'Argenio et al., 2016) and of the kistamicin producer *Nonomuraea* sp. ATCC 55076, previously classified as *Actinomadura parvosata* subsp. *kistnae* S382–8 (Kusserow and Gulder, 2017; Nazari et al., 2017), have confirmed the hidden potential of these uncommon actinomycetes as prolific producers of specialized metabolites. In this paper, we report that another member of this genus, *Nonomuraea coxensis* DSM 45129, which was isolated in Bangladesh in 2007 (Ara et al., 2007), has the genetic potential to produce A40926 or a very similar GPA; its BGC contains two regulatory genes, *nocRI* and *nocRII*, which are close homologs of *dbv3* and *dbv4*, respectively. Thus, we first developed the molecular tools to manipulate both *N. gerenzanensis* and *N. coxensis*, we then selected the strongest heterologous promoter to drive gene expression in *Nonomuraea* spp., and finally we overexpressed both native and heterologous cluster-specific regulatory genes in *N. gerenzanensis*, assessing the best performers at flask and bioreactor scale in industrial media. The overexpression of the positive pathway-specific regulators significantly increased the level of A40926 production in *N. gerenzanensis*, paving the way for knowledge-based strain improvement for the production of this valuable GPA.

# MATERIALS AND METHODS

#### Plasmids, Bacterial Strains, Antibiotics, and Culture Conditions

Plasmids and bacterial strains used in this work are summarized in **Table 1**. Compositions of media are given in ESM. Unless otherwise stated, all media components and antibiotics were supplied by Sigma-Aldrich, St. Louis, MO, United States. For routine maintenance, actinobacterial strains were cultivated on ISP3 or VM0.1 agar media supplemented with 50 μg/ml apramycin-sulfate when appropriate. For genomic DNA isolation, *N. gerenzanensis* ATCC 39727 and *N. coxensis* DSM 45129 were grown in 250 ml Erlenmeyer flasks containing 10 glass beads (ø5 mm) with 50 ml of liquid VSP medium on an orbital shaker at 220 rpm and at 30°C. Working cell banks (WCB) of *Nonomuraea* spp. were prepared as described previously (Marcone et al., 2014). *Escherichia coli* DH5α was used as a routine cloning host and *E. coli* ET12567 pUZ8002 as a donor for intergeneric conjugations. *E. coli* strains were grown at 37°C in LB liquid or agar media supplemented with 100 μg/ ml of apramycin-sulfate, 50 μg/ml of kanamycin-sulfate and 25 μg/ml of chloramphenicol when appropriate.

#### Generation of Recombinant Plasmids Construction of Promoter-Probe Vectors

To test the activity of different native *N. gerenzanensis* promoters, pSAGA (Koshla et al., 2019), where *gusA* (Myronovskyi et al., 2011) is expressed from *aac(3)IVp*, was chosen as a chassis. Genomic DNA, extracted from *N. gerenzanensis* according to the Kirby procedure (Kieser et al., 2000), was used as a template to amplify the putative promoter regions of: *BN4615\_P641* (coding for the DNA gyrase B subunit – GyrBng) 342 bp; *BN4615\_P8899* (coding for the RNA polymerase sigma factor RpoD – HrdBng), 524 bp; *BN4615\_P604* (coding for the singlestranded DNA-binding protein – Ssbng), 207 bp; *BN4615\_P1543* (coding for the SSU ribosomal protein S12p – RpsLng), 339 bp; *BN4615\_P7269* (coding for the rifamycin-resistant RNA polymerase subunit β – RpoBRng), 558 bp; and *BN4615\_P1539* (coding for the rifamycin-sensitive RNA polymerase subunit β – RpoBng), 493 bp. Amplicons were generated using Q5 High-Fidelity DNA Polymerase (New England Biolabs, Ipswich, MA, United States) according to the supplier's protocol and the oligonucleotide primers listed in **Table 2**. All of the amplicons were digested with *Bam*HI and *Eco*RV and cloned in pSAGA cleaved with the same endonucleases, thus replacing *aac(3)IVp* in front of *gusA* with each of the amplified promoter regions. The resulting recombinant plasmids were named pGBP (carrying *gyrB*ng*p*), pHBP (*hrdB*ng*p*), pSBP (*ssb*ng*p*), pRLP (*rpsL*ng*p*), pRBP1 (*rpoB*ng*p*), and pRBP2 (*rpoB*Rng*p*).

#### Construction of the *dbv3*, *dbv4*, and *nocRI* Overexpression Plasmids

The coding sequences of *dbv3* (2,635 bp) and *dbv4* (1,006 bp) were amplified from the A40Y cosmid (**Table 1**; Marcone et al., 2010a) using Q5 High-Fidelity DNA Polymerase and the dbv3\_F/R or dbv4\_F/R primer pairs (**Table 2**). The obtained amplicons were digested with *Eco*RI and *Eco*RV and cloned into pSET152A cleaved with the same enzymes. The resulting plasmids were named pSAD3 and pSAD4. To generate a vector for the overexpression of both *dbv3* and *dbv4*, the regions containing the coding sequences of both genes were amplified using the dbv4\_F/dbv3\_R primer pair and cloned into pSET152A in a similar fashion, generating pSAD3–4.

To construct the vector for overexpression of *nocRI* (the *dbv3* ortholog from *N. coxensis*), the coding sequence of *A3G7\_RS0138355* was amplified from the genomic DNA of *N. coxensis* isolated using the Kirby procedure (Kieser et al., 2000) using Q5 High-Fidelity DNA Polymerase and the orfR1\_F/R primer pair (**Table 2**). The obtained amplicon (2,661 bp) was digested with *Eco*RI and *Eco*RV and cloned into pSET152A cut with the same enzymes to generate pSAR1. All of the generated recombinant plasmids were verified by restriction endonuclease mapping and sequencing at BMR Genomics (University of Padua, Italy).

#### Conjugative Transfer of Plasmids Into *Nonomuraea* spp. and Verification of the Recombinant Strains

Conjugative transfer of plasmids into *N. gerenzanensis* was performed essentially as described previously (Marcone et al., 2010c). All recombinant plasmids were transferred individually into the non-methylating *E. coli* ET12567 pUZ8002 and the resulting derivatives used as donor strains for intergeneric conjugation. To prepare fresh vegetative mycelium of *N. gerenzanensis* prior to conjugal transfer, one vial of WCB was inoculated into 50 ml of VSP medium (250 ml Erlenmeyer TABLE 1 | Bacterial strains and plasmids used in this work.


#### *(Continued)*

#### TABLE 1 | Continued


flask with 10 ø5 mm glass beads) and incubated for 48 h on the orbital shaker at 220 rpm and at 30°C. The mycelium was collected by centrifugation (10 min, 3,220 × *g*), washed twice with sterile 20% v/v glycerol, resuspended in the same solution to a final volume of 20 ml, and stored at −80°C. 1 ml of mycelial suspension was mixed with approximately 109 of donor *E. coli* cells and the mixtures were plated on well dried VM0.1 agar plates supplemented with 20 mM of MgCl2. After 12–16 h of incubation at 30°C, each plate was overlaid with 1 ml of sterile deionized water containing 1.25 mg of apramycin-sulfate and 750 μg of nalidixic acid sodium salt. Transconjugants were selected as resistant to 50 μg/ml of apramycin-sulfate.

Spore suspensions of *N. coxensis* were prepared from lawns grown on ISP3 agar for 7 days. Spores from one plate were collected in deionized water and filtered through one layer of Miracloth (Merck KGaA, Darmstadt, Germany) to remove vegetative mycelial fragments. Then, spores were pelleted from a 50 ml suspension by centrifugation (15 min, 3,220 × *g*), resuspended in 1 ml of 15% v/v glycerol, and stored at −80°C. For conjugation, approx. 106 spores were mixed with 107 *E. coli* donor cells and plated on VM0.1 agar plates supplemented with 20 mM of MgCl2. The overlay for the selection of transconjugants was performed as described previously for *N. gerenzanensis*.

To verify the integration of promoter-probe vectors, an ~1 kbp internal fragment of *gusA* was amplified from the genomic DNA of recombinant *N. gerenzanensis* or *N. coxensis* strains using the gusA\_ver\_F/R primer pair (**Table 2**). To verify the integration of pSET152A, *aac(3)IV* was amplified using the aac(3)IV\_F/R primer pair (**Table 2**). To verify the integration of pSAD4, an ~1 kbp fragment of pSAD4 was amplified with the PAM\_seq\_F/dbv4\_R primer pair (**Table 2**), in which PAM\_seq\_F anneals within the *aac(3)IVp* sequence. Verification of pSAD3 and pSAD3–4 integration was made by amplification of an ~2 kbp fragment (for pSAD3) or an ~3 kbp fragment (for pSAD3–4), using the PAM\_seq\_F/dbv3\_seq\_R primer pair (**Table 2**), in which dbv3\_seq\_R anneals in the middle of *dbv3.* Finally, to verify the integration of pSAR1 an ~2 kbp fragment was amplified using the PAM\_seq\_F/orfR1mid\_ EcoRV\_R (**Table 2**) primer pair, in which orfR1mid\_EcoRV\_R anneals in the middle of *nocRI*.

#### β-Glucuronidase Activity Assay

β-Glucuronidase (GusA) activity in *Nonomuraea* strains growing on VM0.1 agar medium was assessed by adding, after 6 days of cultivation at 30°C, 10 μl drops of 5-bromo-4-chloro-3 indolyl-β-D-glucuronide (X-Gluc, Thermo Fisher Scientific, Waltham, MA, United States) 50 mg/ml in DMSO to the surfaces of the lawns. Chromogenic conversion of X-Gluc into the bluecolored 5,5′-dibromo-4,4′-dichloro-indigo was monitored after 1 h of incubation. For the quantitative measurements of GusA activity, *Nonomuraea* strains were grown in liquid media. One WCB vial of each of the strains was inoculated into a baffled 500 ml Erlenmeyer flask containing 100 ml of E26 (*N. gerenzanensis* strains) or of VSP (*N. coxensis* strains). After 72 h of cultivation, 10% v/v of this preculture was transferred into a baffled 500 ml Erlenmeyer flask containing 100 ml of FM2 (*N. gerenzanensis* strains) or ISP2 (*N. coxensis* strains). To induce *P21-cmt*-driven *gusA*-expression, cumate was added at the final concentration of 50 μM to cultures carrying pGCymRP21 24 h after inoculation. After 120 h of cultivation, mycelial lysates were prepared as previously reported by Horbal et al., 2013. Glucuronidase activity was measured as previously described (Myronovskyi et al., 2011; Horbal et al., 2013) using

#### TABLE 2 | Oligonucleotide primers used in this work.


*\*Restriction sites are underlined in primer sequence.*

a spectrophotometric assay following the conversion of the colorless *p*-nitrophenyl-β-D-glucuronide (Thermo Fisher Scientific, Waltham, MA, United States) into the colored *p*-nitrophenol at 415 nm using an Infinite 200 PRO microplate reader (Tecan, Switzerland). Glucuronidase activity was normalized to dry biomass weight as previously reported (Marcone et al., 2014). One unit of activity is defined as the amount of enzyme that is able to convert 1 μM of substrate in 1 min.

#### A40926 Production

One WCB vial was inoculated into 300 ml baffled flasks containing 50 ml of vegetative medium E26 with 10 glass beads (ø5 mm). Flask cultures were incubated for 72 h on a rotary shaker at 220 rpm and 30°C and then used to inoculate (10% v/v) 500 ml baffled Erlenmeyer flasks containing 100 ml of FM2 medium or a 3-l P-100 Applikon glass reactor (height 250 mm, ø130 mm) equipped with a AD1030 Biocontroller and AD1032 motor, and containing 2 l of the same production medium. Cultivations in FM2 in shake-flasks were conducted at 30°C and 220 rpm. Bioreactor fermentations were conducted at 30°C, with stirring at 450 rpm (corresponding to 1.17 m/s of tip speed) and 2 l/ min aeration rate. Dissolved oxygen (measured as % pO2) was monitored using an Ingold polarographic oxygen electrode. The pH values of culture broths were monitored using a pH meter. Foam production was controlled by adding Hodag antifoam (Hodag Chemical Corporation, Chicago, IL, United States) through an antifoam sensor. Samples were collected at regular cultivation time intervals and analyzed to estimate biomass (dry weight), glucose consumption (Diastix sticks, Bayer AG, Leverkusen, Germany), and A40926 production.

#### HPLC Analysis of Culture Extracts

A40926 was extracted from *Nonomuraea* spp. cultures as previously reported (Marcone et al., 2014). Chromatography was performed with a VWR Hitachi diode array L-2455 HPLC system with detection at 254 nm. The A40926 titers in the batch cultivations were estimated by injecting 50 μl of sample onto a 5 μm-particle-size Ultrasphere ODS (Beckman) HPLC column (4.6 by 250 mm) and eluting at a flow rate of 1 ml/ min with a 30 min linear gradient from 15 to 64% of phase B. Phase A was 32 mM HCOONH4 (pH 7) – CH3CN [90:10 (vol/vol)], and phase B was 32 mM HCOONH4 (pH 7) – CH3CN [30:70 (vol/vol)]. A volume of 50 μl of a pure sample of 200 μg/ml A40926 (Sigma-Aldrich, St. Louis, MO, United States) was used as an internal standard.

#### Tools for the Bioinformatics Analysis

Blastp was used to search for homologs (Altschul, 1990); protein sequence alignments were performed with Clustal Omega (EMBL-EBI, Sievers et al., 2011).

# RESULTS

#### Genetic Manipulation of *Nonomuraea coxensis*, a Novel Putative Producer of a A40926-Like Molecule

According to the 16S rRNA gene-based reconstruction of *Nonomuraea* phylogeny (Dalmastri et al., 2016), *N. coxensis* occupies a relatively distant position from *N. gerenzanensis*. Conversely, mining the partially sequenced genome of *N*. *coxensis*

(ASM37988v1), we found a close homolog of the *N. gerenzanensis* regulatory gene *dbv3* (locus *A3G7\_RS0138355* in the *N. coxensis* genome, coding for WP\_020547054.1 with 86.6% predicted amino acid sequence identity with Dbv3, ESM **Figure 1**). Dbv3 is a unique LuxR-regulator controlling the expression of the A40926 BGC and it is not closely related to the better-characterized family of Tei16\*-like regulators controlling teicoplanin biosynthesis, sharing only 32% of amino acid sequence identity with Tei16\* (Yushchuk et al., 2019). This *dbv3*-like gene (we named it *nocRI*) was found on a short *N. coxensis* contig (NZ\_KB904006) flanking a gene coding for a putative StrR-like transcriptional regulator apparently orthologous to *dbv4* (locus *A3G7\_*RS0138355, coding for WP\_026215141.1 with 94.39% of amino acid sequence identity with Dbv4); *dbv4* is the other known cluster-situated regulatory gene in the *dbv* BGC and we named the *N. coxensis* homolog *nocRII* (ESM **Figure 2**). AntiSMASH (Blin et al., 2019) analysis of *N. coxensis* genome revealed the presence of three short contigs (NZ\_KB904006, NZ\_ KB903995, NZ\_KB903969, ESM **Figure 3**) covering the majority of a *dbv-*like BGC, suggesting that *N. coxensis* might produce A40926 or a very similar GPA. Thus, we considered *N. coxensis* an interesting candidate for developing *Nonomuraea*targeted genetic tools.

Initially, we tried to transfer *φC31*-based integrative plasmids (**Table 1**) into *N. coxensis* by using the protocol of intergeneric conjugation optimized for conjugal transfer from a DNA-nonmethylating *E. coli* donor strain to *N. gerenzanensis* vegetative mycelium (Marcone et al., 2010c). Transconjugants were obtained at a very low frequency (*ca.* 1 × 10<sup>−</sup><sup>7</sup> ) and only for the relatively small pSAGA-based promoter-probe vectors (approx. 6 kbp). Increasing the amount of donor and recipient cells, changing the time of overlay, and adjusting medium composition (including increasing or decreasing MgCl2 concentration) did not allow the transfer of the larger pGUS-based promoterprobe vectors (such as pGUSmoeE5script or pGCymRP21, both more than 9 kbp). Since, unlike *N. gerenzanensis*, *N. coxensis* sporulates abundantly when grown on ISP3 agar medium (ESM **Figure 4**), we tried to use spores for conjugal transfer. This resulted in transfer rates of approximately 1 × 10<sup>−</sup><sup>3</sup> when 106 spores were mixed with 107 *E. coli* donor cells, regardless of plasmid size.

#### Using the GusA-Reporter System to Assess the Activity of Native and Heterologous Promoters in *Nonomuraea* Species

The set of *φC31*-based integrative plasmids transferred to *N. gerenzanensis* and *N. coxensis* were promoter-probe vectors utilizing the GusA reporter system and carrying a selection of native and heterologous promoters, the latter having been used previously to drive gene expression in streptomycetes or actinoplanetes (Horbal et al., 2013, 2014a; Makitrynskyy et al., 2013). Glucuronidase activity of the recombinant *Nonomuraea* strains was assessed qualitatively (on agar plates) and quantitatively (in cell lysates obtained from mycelium

grown in liquid medium). Both of the *N. coxensis* and *N. gerenzanensis* wild type strains did not display any glucuronidase activity (**Figure 1**). The heterologous promoters tested were: *aac(3)IVp* (in pSAGA) – the apramycin acetyltransferase gene promoter, derived from pSET152A (Horbal et al., 2013); *ermEp* (in pTEGA, Yushchuk et al., 2020a) – the erythromycin resistance gene promoter from pTES (Herrmann et al., 2012); *moeE5p* (in pGUSmoeE5script) – the *S. ghanaensis* moenomycin biosynthesis gene *moeE5* promoter (Makitrynskyy et al., 2013); *tei2p* (pGT2p, Yushchuk et al., 2020b) – the *A. teichomyceticus* teicoplanin resistance gene *tei2* promoter; and *P21*, a synthetic promoter fused with the cumate inducible *Pseudomonas putida* F1 *cmt* operon operator (in pGCymRP21, Horbal et al., 2014a). In parallel, we tested the activity of six native promoters derived from *N. gerenzanensis* house-keeping genes: *gyrB*ng*p* (in pGBP – the promoter of the DNA gyrase B subunit gene); *hrdB*ng*p*

(in pHBP – the promoter of the RNA polymerase σ-factor RpoD gene); *ssb*ng*p* (in pSBP – the promoter of the singlestranded DNA-binding protein); *rpsL*ng*p* (in pRLP – the promoter of the SSU ribosomal protein S12p gene); *rpoB*Rng*p* (in pRBP2 – the promoter of the rifamycin-resistant RNA polymerase subunit β gene); and *rpoB*ng*p* (in pRBP1 – the

cultivated in FM2 industrial medium in 500 ml Erlenmeyer flasks. Glucose consumption (filled rectangles), biomass accumulation (filled rhombi), pH (circles), and A40926 production were monitored every 24 h. Results given are mean values of three independent experiments. Error bars represent standard deviations. Statistical significance of the differences in A40926 production between the control and the recombinant strains was estimated using Welch's *t*-test: \**p* < 0.05; \*\**p* < 0.01; \*\*\**p* < 0.001.

promoter of the rifamycin-sensitive RNA polymerase subunit β gene). In contrast to the wild type strains, all of the recombinant derivatives grown on VM0.1 agar plates converted X-Gluc to its colored derivative 5,5′-dibromo-4,4′-dichloroindigo, albeit to different extents (**Figure 1**). *N. gerenzanensis* pSAGA+ and *N. coxensis* pSAGA+ , carrying *aac(3)IVp*, produced the most intensive color (**Figures 1A,B**), while the chromogenic conversion of X-Gluc in *N. gerenzanensis* pTEGA+ and *N. coxensis* pTEGA+ , carrying the *ermEp*, was only slightly visible. In strains carrying the inducible pGCymRP21, GusA activity was induced by the addition of 50 μM of cumate, proving inducible gene expression (**Figures 1A,B**). However, in *N. gerenzanensis* pGCymRP21+ , a basal level of GusA activity was detected even in the absence of cumate (**Figure 1A**). Contrary to this, no basal level of expression from *P21-cmt* was detected in *N. coxensis* pGCymRP21+ (**Figure 1B**).

When the glucuronidase activity present in cell lysates obtained from liquid cultures grown for 120 h (late exponential/ early stationary growth phase, see ESM **Figure 5**) was measured using a spectrophotometric assay, overall these quantitative results (normalized for the dry weight of the differently growing strains) correlated with those observed on agar plates: *aac(3)IVp* behaved as a strong promoter in both *N. gerenzanensis* and *N. coxensis*, surpassed only by *cmt-P21p* when induced by the addition of 50 μM cumate (**Figures 2A,B**). Also *rpsL*ng*p* and *hrdB*ng*p* proved to be strong promoters in both of the *Nonomuraea* spp. whereas the weakest was the *ermEp* (**Figures 2A,B**). Interestingly, the activity of *rpoB*ng*P* was higher than that of *rpoB*Rng*p* (**Figures 1**, **2**), consistent with previously reported data about differences in transcription levels of the two alleles (Vigliotta et al., 2004).

#### Knowledge-Based Generation of A40926 Overproducing Strains

Since *aac(3)IVp* was identified as the strongest constitutive promoter for *Nonomuraea* spp., we used it to overexpress *dbv3* and *dbv4* in *N. gerenzanensis*. Both genes were cloned into the integrative pSET152A vector (Horbal et al., 2013) yielding the recombinant vectors pSAD3 and pSAD4, respectively (**Table 1**). Benefiting from the neighboring positions of *dbv3* and *dbv4*, we also cloned them together, generating pSAD3–4, where *dbv4* was directly under the control of *aac(3)IVp*, but *dbv3* remained under the control of its native promoter (**Table 1**). Additionally, the *dbv3-*like *nocRI* from *N. coxensis* was cloned into pSET152A, generating pSAR1, and transferred to *N. gerenzanensis* to determine if it could improve A40926 production, and to assess possible cross-talk between the regulators from the two different *Nonomuraea* spp. All the *N. gerenzanensis* recombinant strains were grown for 120 h in parallel with the control strain carrying the empty vector and A40926 production was measured (**Figure 3**). The production of A40926 in *N. gerenzanensis* carrying pSET152A reached almost 250 mg/l after 120 h of cultivation (**Figure 3**). The four recombinant strains overexpressing the cluster-situated regulatory genes produced more antibiotic than the control strain (**Figure 3**). A40926 production in *N. gerenzanensis* pSAD3+ and pSAR1+ was comparable (*ca*. 500 mg/l), proving that *nocRI* had a similar impact in *N. gerenzanensis* as *dbv3.* The recombinant strain carrying pSAD3–4 vector produced slightly more, around 550 mg/l, whereas the best producer in these conditions was the strain overexpressing *dbv4*, which produced more than 650 mg/l.

#### Time Courses of A40926 Production in *Nonomuraea gerenzanensis* Recombinant Strains at Flask and at Bioreactor Scale

Although further investigations will be devoted to the expression of the heterologous *nocRI* and *nocRII in N. gerenzanensis*, in our strain improvement work we then focused on *N. gerenzanensis* strains overexpressing native regulators. Consequently, *N. gerenzanensis* strains containing pSET152A, pSAD3, pSAD4, or pSAD3–4 were grown for 192 h in parallel with the control strain carrying the empty vector using the previously optimized industrial medium FM2 (Marcone et al., 2010a, 2014) at flask scale. Samples for the analysis of dry weight, pH, glucose consumption, and A40926 production were collected at regular 24 h intervals. All three recombinant strains expressing the regulatory genes from *aac(3)IVp* accumulated detectable amounts of A40926 earlier than the empty vector control (**Figure 4**). A40926 production reached its peak after 120 h of growth for *N. gerenzanensis* pSAD3+ (*ca.* 500 mg/l) (**Figure 4B**) and pSAD4+ (nearly 650 mg/l) (**Figure 4C**), whereas in pSAD3–4+ the maximum productivity (*ca*. 600 mg/l) was delayed to 144 h (**Figure 4D**). The control strain produced *ca.* 300 mg/l after 144–168 h from inoculation (**Figure 4A**). Although glucose consumption was faster in the recombinant strains containing the cloned regulatory genes in comparison to the control, they accumulated less biomass than *N. gerenzanensis* pSET152A+ (**Figure 4**). Maximum biomass production was around 125 g/l (dry weight) in the overexpression strains versus the 175 g/l produced by the empty vector control strain. These data suggest that part of the consumed glucose was used by the strains carrying the regulatory genes under the control of the strong constitutive promoter *aac(3)IVp* to support antibiotic production at the detriment of biomass formation.

When the cultivation of recombinant strains was scaled up in 3 l vessel-bioreactors containing 2 l working volume of FM2, all of the strains produced significantly more A40926 and grew better than in flask culture with the exception of *N. gerenzanensis* pSAD3+ (**Figure 5**). The control strain with the empty vector grew and produced the antibiotic faster than at the flask level (**Figure 5A**). Maximum biomass (300 g/l dry weight, more than the double of that achieved in flasks) and A40926 production (nearly 400 mg/l) were reached after 120 h from inoculation; glucose was completely consumed within 96 h versus the 120 h needed at flask level. The recombinant strain pSAD4+ grew more (maximum biomass production of 220 g/l after 144 h of growth) than in flask culture, although less than the control strain in the bioreactor. Glucose was consumed faster than in the control strain and glucose concentration tended to zero at 48 h of fermentation (**Figure 5C**), and antibiotic production reached a peak of 700 mg/l after 168 h. The best performance in terms of A40926 productivity in the bioreactor was shown by the pSAD3–4+ strain, which grew better (nearly 250 g/l dry weight biomass) than in flask culture and produced the maximum concentration of the antibiotic (800 mg/l) after 168 h of cultivation (**Figure 5D**). Conversely, the pSAD3+ strain showed a reduced biomass production in comparison to all of the other strains, with consumption of glucose markedly delayed and A40926 production

cultivated in 2 l FM2 industrial medium in a 3 l fermenter. Glucose consumption (filled rectangles), biomass accumulation (filled rhombi), pH (circles), O2 level (filled circles), and A40926 production were monitored every 24 h.

starting very late (after 168 h of cultivation) and never exceeding 100 mg/l (**Figure 5B**). Microscopical observation of pSAD3+ strain showed that the mycelium was highly fragmented (data not shown), suggesting some kind of physiological stress resulting from *dbv3* overexpression. Some fragmentation of the pSAD3+ strain (as well as of the pSAR1 strain carrying the *dbv3-*like *nocRI* from *N. coxensis*) was also observed in flask culture, in contrast to the other strains which produced dense mycelial pellets; the fragmentation was less pronounced, indicating that scaling up in the bioreactor dramatically enhanced this effect. Interestingly, in the best performer at the bioreactor scale – the pSAD3–4+ strain – *dbv4* was expressed from the strong constitutive *aac(3)IVp* but *dbv3* was left under its endogenous promoter, suggesting that the balance between the expression level of the two cluster-situated regulatory genes is important for optimal improvement of A40926 production.

#### DISCUSSION

Apart from some reports on cultivating and manipulating the industrially valuable A40926 producer *N. gerenzanensis* (Stinchi et al., 2003, 2006; Marcone et al., 2010a,b,c; Alt et al., 2019) and the kistamicin producer *Nonomuraea* sp. ATCC 55076 (Greule et al., 2019), we are not aware of any other attempt to develop genetic tools for manipulating *Nonomuraea* spp. However, some other glycopeptide producers, like *A. teichomyceticus*, already possess well-developed toolkits for genetic manipulation which has greatly simplified investigations in these strains (Horbal et al., 2013; Yushchuk et al., 2016). Therefore, our first goal in this work was to develop a set of genetic tools for manipulating diverse species of *Nonomuraea*. To do this, we decided to work in parallel with the better-known A40926 producer, recently re-classified as *N. gerenzanensis* (Dalmastri et al., 2016), and with the little investigated *N. coxensis*, which was isolated in Bangladesh in 2007 (Ara et al., 2007). Although the available *N. coxensis* genome sequence is still incomplete, we could identify three contigs covering most of a *dbv*-like gene cluster including the *dbv*-like cluster-situated regulatory genes, which we named *nocRI* (*dbv3* homolog) and *nocRII* (*dbv4* homolog). The next step of our work will be additional sequencing to yield a properly annotated *N. coxensis* genome. Interestingly, during the course of our investigations, Waglechner et al. (2019), systematically screening the available sequences in genomic databases, also reported the presence of a BGC encoding for a A40926-like GPA in the genome of *N. coxensis*. In addition, the same authors reported that a newly isolated *Nonomuraea* sp. WAC01424 possess a BGC which could produce another A40926-related compound (Waglechner et al., 2019). Our preliminary analysis of this BGC suggests that it could be a sulfated A40926-like GPA, lacking the aliphatic side chains.

In the meantime, we tested both in *N. gerenzanensis* and *N. coxensis*, a set of heterologous and native promoters (the latter derived from a set of house-keeping genes in *N. gerenzanensis*) with the final goal of using them for driving gene expression in these strains. Besides the practical outcome of this screening (all of the generated promoter-probe vectors could be easily used as expression vectors simply by exchanging *gusA* for a gene of interest, offering a set of variable tools for gene expression), it is interesting to observe that the studied promoters had similar strengths in the two phylogenetically distant *Nonomuraea* species and that the native promoters from *N. gerenzanensis* worked similarly in *N. coxensis*. Consistently, we detected a comparable difference in the two strains between the strength of promoters driving the expression of rifamycinsensitive and rifamycin-resistant *rpoB* alleles of *N. gerenzanensis* (*rpoB*ng*p* and *rpoB*Rng*p*). Since we found the two *rpoB* alleles in *N. coxensis* draft genome (on a short genomic scaffolds KB904038 and KB904038), we might suppose that their manipulation could improve antibiotic production in this strain as already reported for *N. gerenzanensis* (Vigliotta et al., 2004). More generally, these vectors should be useful for the genetic manipulation of other members of the genus *Nonomuraea*, which have the potential to produce novel valuable specialized metabolites (Sungthong and Nakaew, 2015; Nazari et al., 2017).

As in the case of *Actinoplanes* spp. (Horbal et al., 2013), the strongest heterologous promoter was *aac(3)IVp*, although some *N. gerenzanensis* native promoters like *hrdB*ng*p* and *rpsL*ng*p* appeared to have comparable strength and merit further investigations. When we used *aac(3)IVp* to overexpress the cluster-situated regulatory genes *dbv3* and *dbv4* from the *N. gerenzanensis dbv* gene cluster and *nocRI* (*dbv3*-like) from the A40926-like BGC of *N. coxensis*, the recombinant *N. gerenzanensis* strains produced significantly more A40926 than the parental strain. The evidence that the heterologous expression of *nocRI* increased A40926 production in *N. gerenzanensis* confirmed its role in regulating the expression of a A40926-like BGC in *N. coxensis*. Additionally, it represents another case of cross-talk between regulators controlling GPA BGCs in producing actinomycetes (Spohn et al., 2014). Our next goal will be to investigate if and how (in which cultivation conditions) *N. coxensis* produces A40926 or a A40926 like molecule.

Previous work (Lo Grasso et al., 2015) reported that overexpression of *dbv3* (under the control of the thiostreptoninducible *tipA\** promoter in the integrative plasmid pIJ8600) in *N. gerenzanensis* increased A40926 production from 13 to 27 mg/l using the laboratory medium R3 (Lo Grasso et al., 2015). In this paper, we tested the real industrial potential of overexpressing not only *dbv3*, but also *dbv4*, and *dbv3* and *dbv4* together, cloning them under the strong constitutive *aac(3)IVp* promoter, scaling up their cultivation at bioreactor scale and using a previously optimized industrial medium where A40926 is produced in hundreds of milligrams per liter (Marcone et al., 2010a, 2014). At the bioreactor level, where strains could have a different performance from the flask-cultivation due to different mixing and mass transfer rates of nutrients and oxygen, the best performer was the strain carrying both *dbv4* under *aac(3)IVp* and *dbv3* under its own endogenous promoter. This strain produced nearly 800 mg/l of A40926, which is twice that of the parental strain grown under the same conditions. Conversely, the strain carrying only *dbv3* expressed from *aac(3)IVp* showed reduced production capacity and an altered phenotype particularly after scaling up from flask to bioreactor; in contrast, the *dbv4* overexpressing recombinant grew similarly under both conditions. It is widely recognized that any potentially higher producing mutant or derivative needs to be validated in a bioreactor-scale fermentation since unpredictable discrepancies in strain performance can occur during scaling up from flask culture (Lee and Kim, 2015). In the case of the *dbv3* overexpressing recombinant, we believe that the fragmented mycelial phenotype, which was much more apparent in the bioreactor, could be a specific consequence of the overexpression of the *dbv7* gene encoding a d,d-carboxypeptidase known as VanYn (Binda et al., 2012). The level of VanYn activity in cell extracts from the *dbv3*-carrying recombinant cultivated at bioreactor scale was found to be much higher than in the parental strain (unpublished data). Consistent with this, overexpression of VanYn altered the mycelial morphology in *N. gerenzanensis* as well as in heterologous hosts such as streptomycetes strains (Binda et al., 2013).

In conclusion, only a few GPAs are used in clinical practice and those produced by semi-synthesis from natural products, such as dalbavancin derived from A40926, are still quite expensive. Dalbavancin is the first antibiotic designated as a Qualified Infection Diseases Product by the FDA because of its potency, extended dosing interval, and unique dose regimen, but its cost largely exceeds that of first-generation GPAs and consequently its use in hospitals is still limited (Chiasson and White, 2016; Agarwal et al., 2018). Improving A40926-producing strains might lead to a decrease in the cost of dalbavancin. As demonstrated in this paper, A40926 production could be significantly enhanced by manipulating the expression of *dbv* cluster-situated regulators. An important and often-neglected aspect is testing the genetic stability and productivity of the selected recombinant strains in a fully developed industrial process at bioreactor level, which mimics the conditions of antibiotic large scale production. We were able to demonstrate here that the improvements we made to A40926 production levels in shake flasks were also achieved in the bioreactor, indicating the relevance of this approach to industrial-scale strain improvement.

#### REFERENCES


#### DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the article/**Supplementary Material**.

#### AUTHOR CONTRIBUTIONS

OY, EB, MB, and FM conceived and designed the experiments and wrote the paper. OY, AA-V, GM, and EB performed the experiments. OY, GM, and EB analyzed the data.

#### FUNDING

This work was supported by public grants "Fondo di Ateneo per la Ricerca" 2018 and 2019 to FM and GM and by Federation of European Microbiological Societies (FEMS) Research Fellowship 2015 to EB.

#### ACKNOWLEDGMENTS

Consorzio Interuniversitario per le Biotecnologie is acknowledged for its support to traveling and attending congresses for EB and GM.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb.2020.00008/ full#supplementary-material


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2020 Yushchuk, Andreo-Vidal, Marcone, Bibb, Marinelli and Binda. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# The Balance Metabolism Safety Net: Integration of Stress Signals by Interacting Transcriptional Factors in Streptomyces and Related Actinobacteria

#### Juan F. Martín\* and Paloma Liras

Área de Microbiología, Departamento de Biología Molecular, Universidad de León, León, Spain

#### Edited by:

Yvonne Mast, German Collection of Microorganisms and Cell Cultures GmbH (DSMZ), Germany

#### Reviewed by:

Andrew Hesketh, University of Brighton, United Kingdom Gang Liu, Institute of Microbiology (CAS), China

> \*Correspondence: Juan F. Martín jf.martin@unileon.es

#### Specialty section:

This article was submitted to Microbial Physiology and Metabolism, a section of the journal Frontiers in Microbiology

> Received: 04 November 2019 Accepted: 24 December 2019 Published: 22 January 2020

#### Citation:

Martín JF and Liras P (2020) The Balance Metabolism Safety Net: Integration of Stress Signals by Interacting Transcriptional Factors in Streptomyces and Related Actinobacteria. Front. Microbiol. 10:3120. doi: 10.3389/fmicb.2019.03120 Soil dwelling Streptomyces species are faced with large variations in carbon or nitrogen sources, phosphate, oxygen, iron, sulfur, and other nutrients. These drastic changes in key nutrients result in an unbalanced metabolism that have undesirable consequences for growth, cell differentiation, reproduction, and secondary metabolites biosynthesis. In the last decades evidence has accumulated indicating that mechanisms to correct metabolic unbalances in Streptomyces species take place at the transcriptional level, mediated by different transcriptional factors. For example, the master regulator PhoP and the large SARP-type regulator AfsR bind to overlapping sequences in the afsS promoter and, therefore, compete in the integration of signals of phosphate starvation and S-adenosylmethionine (SAM) concentrations. The cross-talk between phosphate control of metabolism, mediated by the PhoR–PhoP system, and the pleiotropic orphan nitrogen regulator GlnR, is very interesting; PhoP represses GlnR and other nitrogen metabolism genes. The mechanisms of control by GlnR of several promoters of ATP binding cassettes (ABC) sugar transporters and carbon metabolism are highly elaborated. Another important cross-talk that governs nitrogen metabolism involves the competition between GlnR and the transcriptional factor MtrA. GlnR and MtrA exert opposite effects on expression of nitrogen metabolism genes. MtrA, under nitrogen rich conditions, represses expression of nitrogen assimilation and regulatory genes, including GlnR, and competes with GlnR for the GlnR binding sites. Strikingly, these sites also bind to PhoP. Novel examples of interacting transcriptional factors, discovered recently, are discussed to provide a broad view of this interactions. Altogether, these findings indicate that cross-talks between the major transcriptional factors protect the cell metabolic balance. A detailed analysis of the transcriptional factors binding sequences suggests that the transcriptional factors interact with specific regions, either by overlapping the recognition sequence of other factors or by binding to adjacent sites in those regions. Additional interactions on the regulatory backbone are provided by sigma factors, highly phosphorylated nucleotides, cyclic dinucleotides, and small ligands that interact with cognate receptor proteins and with TetR-type transcriptional regulators. We propose to define the signal integration DNA regions (so called integrator sites) that assemble responses to different stress, nutritional or environmental signals.

These integrator sites constitute nodes recognized by two, three, or more transcriptional factors to compensate the unbalances produced by metabolic stresses. This interplay mechanism acts as a safety net to prevent major damage to the metabolism under extreme nutritional and environmental conditions.

Keywords: Actinobacteria, stress signals, overlapping binding sequences, transcriptional factors, sigma factors, cyclic nucleotides, safety net, signal integration nodes

#### COORDINATION OF PRIMARY AND SECONDARY METABOLISM IN Streptomyces: INTERACTION OF MULTIPLE TRANSCRIPTIONAL FACTORS AT SIGNAL INTEGRATION SITES

Evidence accumulated in the last decades shows that coordination of primary and secondary metabolism takes place by multiple interactions of pleiotropic and cluster situated transcriptional factors that work in networks and cascades (Martín and Liras, 2010; van Wezel and McDowall, 2011; Martín et al., 2012b; Liu et al., 2013; Romero-Rodríguez et al., 2016a,b). Some pleiotropic transcriptional factors, such as PhoP, GlnR, or MtrA, control, directly or indirectly, hundreds of reactions in the bacterial cells, and we will refer to them in this article as master regulators. Master transcriptional factors interact with upstream regions of cluster situated regulatory genes. This is the case of several transcriptional factors that regulate the act and red clusters in Streptomyces coelicolor. The promoter region of the actinorhodin regulatory gene actII-orf4 is recognized by at least nine transcriptional factors of different families, including GlnR, AfsS, AfsQ1, AdpA, AtrA, DasR, DraR, AbsA2, and AbsC (Floriano and Bibb, 1996; Ohnishi et al., 2005; Uguru et al., 2005; McKenzie and Nodwell, 2007; Rigali et al., 2008; Yu et al., 2012; He J.M. et al., 2016; Lewis et al., 2019). This phenomenon raises the question of how these transcriptional factors compete for binding to the actII-orf4 promoter region. In other words, these regions (hereafter named integrators sites) serve to integrate multiple signal cascades that respond to different nutritional and environmental stress signals in Streptomyces (**Figure 1**).

An important question is whether these different transcriptional factors interact in some way in the control of expression of a particular gene. Taking into account the large size of some transcriptional regulators (see below), it is likely that each of these transcriptional factors covers at least the major groove or a full turn of DNA (11 nucleotides). Moreover, some of these regulators, e.g., GlnR or PhoP, act as dimers or even oligomers and therefore cover a relatively large stretch of DNA. Mutations affecting these integrator sites alter not only the binding to one transcriptional factor but also to other interacting factors.

The interactions between transcriptional factors and regions upstream of some gene clusters are really elaborated and suggests that there is a fine tuning of the expression of important gene clusters by alternative transcriptional factors. There are many reports of putatively interacting transcriptional factors that affect the biosynthesis of secondary metabolites but the molecular evidence supporting those interactions is scarce. In this article we focus on the most relevant and best-known cases of overlapping interactions between transcriptional factors that allow us to get an insight into how the cells integrate inputs from environmental and nutritional stresses. Those studied for which there are experimental evidence of DNA binding and/or footprinting data are summarized in **Table 1**.

### MECHANISM OF CONTROL OF PHOSPHATE METABOLISM: THE PhoP MASTER REGULATOR

Phosphate is an essential nutrients of living beings. Phosphate starvation causes the slow-down of primary metabolism and triggers production of secondary metabolites in different microorganisms (Martín and Demain, 1980). Phosphate control of metabolism is mediated by the two-component system (TCS) PhoR-PhoP (Hutchings et al., 2004; Martín et al., 2012b) that controls not only primary metabolism but also the biosynthesis of secondary metabolites (e.g., actinorhodin and undecylprodigiosin) in Streptomyces lividans and S. coelicolor (Sola-Landa et al., 2003, 2005; Ghorbel et al., 2006). The PhoP/PhoR TCS has been also studied in Streptomyces natalensis, Streptomyces tsukubaensis, Streptomyces avermitilis, Streptomyces hygroscopicus var. geldanus, and orthologous sequences have been found in most Actinobacteria genomes. The genes encoding these TCS are well conserved in all Streptomyces (Martín et al., 2017, 2019; Ordóñez-Robles et al., 2017b; Martínez-Castro et al., 2018). The Streptomyces PhoR-PhoP system belongs to class IIIA of TCSs (Hutchings et al., 2004). PhoR is a protein sensor kinase with signal transducer activity (Cheung and Hendrickson, 2010) of 426 amino acids in S. lividans and S. coelicolor. It is membrane-anchored through a stretch of hydrophobic amino acids, and contains a cytoplasmic C-terminal region that putatively serves to interact with the phosphate starvation signal, although in some other TCS, the sensor kinase may have an extra-cytoplasmic signal sensor region (McKenzie and Nodwell, 2009). However, in Streptomyces species it is unclear if the phosphate limitation signal that interacts with PhoR is extracellular or if it is an internal signal molecule (Xp in **Figure 1**). In Bacillus subtilis it has been proposed that the signal molecule is an intracellular intermediate of teichoic acid biosynthesis. This intermediate is an inhibitor of the PhoR autokinase activity and is regulated by the phosphate concentration. Under phosphate limitation conditions the concentration of this intermediate decreases and, therefore, the autophosphorylation of PhoR

TABLE 1 | Well-known examples of interacting transcriptional factors in Actinobacteria<sup>1</sup> .


afsS gene are indicated. Positive regulation is indicated by arrows and negative regulations by black spheres. See text for additional details.

<sup>1</sup>Only examples of interaction confirmed by DNA-binding and/or footprinting assays are included. <sup>2</sup>CRP has been proposed to bind to the malEFG promoter but there is no experimental confirmation of this binding. See text for references and details.

and transfer of the phosphate group to PhoP are enhanced (Botella et al., 2014). PhoP is a member of the OmpR family of DNA-binding response regulators (223 amino acids). Under phosphate starvation conditions, the PhoR sensor protein kinase self phosphorylates, and then transfers its phosphate group to a receiver aspartate (D49) in PhoP, that causes a rearrangement of the protein structure resulting in an activation of the DNA-binding domain in the carboxy-terminal region of this transcriptional regulator. The phosphorylated form of PhoP binds specific sequences named PHO boxes and generally activates, but occasionally represses, the expression of genes of the Pho regulon. The Streptomyces PHO boxes are formed by direct repeat units (DRus) of 11 nucleotides of which seven nucleotides (GTTCACC) are well conserved (Sola-Landa et al., 2005; Rodríguez-García et al., 2007; Allenby et al., 2012; Santos-Beneit, 2015). Two DRus are required for binding of PhoP to DNA (Sola-Landa et al., 2008) in agreement with the binding of dimer forms of PhoP (Blanco et al., 2002).

# The PHO Regulon: PhoP-Regulated Genes

The core Pho regulon is formed by more than 100 genes, of which about 50 are well characterized (Santos-Beneit, 2015; Martín et al., 2017). Using chromatin immunoprecipitation, Allenby et al. (2012) found PhoP binding sites in 150 loci in S. coelicolor, not all of them corresponding to promoter regions of characterized genes. Phosphate depletion results in PhoPmediated induction of genes involved in scavenging, transport, and mobilization of phosphate, and in repression of genes for utilization of nitrogen sources. PhoP reduces expression of genes for aerobic respiration and activates nitrate respiration genes. This master regulator activates expression of genes for teichuronic acid formation and decreases that of genes for phosphate-rich teichoic acid biosynthesis. In S. coelicolor PhoP repressed several differentiation genes that affect development and indirectly antibiotic biosynthesis (Rodríguez-García et al., 2007; Allenby et al., 2012; Martín et al., 2017).

# Modification of PhoP Binding Affinity to Recognition Sequences by Acylation at Lysine Residues

The chemical structure of some regulatory proteins of the OmpR family, which includes PhoP, is known (Martínez-Hackert and Stock, 1997; Blanco et al., 2002, 2012; He X. et al., 2016). The DNA binding domain (DBD) of PhoP, and other OmpR-family regulators, includes a winged helix turn helix (HTH) structure in the carboxyl terminal region, that recognizes the PhoP binding sequence (PHO boxes) in the promoters of the target genes (Li et al., 2008). This DBD is highly positively charged and, therefore, has a strong affinity for the negative charged DNA backbone. It is known that several regulatory proteins are modified by acylation at the lysine rich region of their DBD (Li et al., 2008; Thao and Escalante-Semerena, 2011; Tiffert et al., 2011; Yao et al., 2014).

Particularly relevant for this review article is the acylation of residues in the lysine rich region of OmpR family regulators (Luo et al., 2004; Matsuzaki et al., 2005; Qin et al., 2016) that may alter the binding affinity to their recognition sequences (Banerjee et al., 2016). Recently, it was reported that the PhoP regulator of Saccharopolyspora erythraea is propionylated by an acyl transferase encoded by the acuA gene (Xu et al., 2018). Mass spectrometry analysis of the acylated PhoP revealed that the PhoP protein is propionylated at lysines K<sup>198</sup> or K203. The AcuA enzyme in other organisms utilizes acetyl-CoA as acetyl donor, but in this particular case appears to use propionyl-CoA (You et al., 2014).

Mutation of K<sup>198</sup> and K<sup>203</sup> to arginine, glutamate, or glutamine provided evidence that whereas mutation to arginine conserves the polarity of the DNA binding region, the mutation to glutamine (or glutamate) changes the charge of this region thus weakening the binding of the mutated PhoP to the DNA backbone. These observations were supported by electrophoretic mobility-shift assay (EMSA) which showed that the mutation to glutamate removed the binding ability of PhoP and, therefore, altered the control of expression of the Pho regulon genes (Xu et al., 2018). In conclusion, highly propionylated forms of PhoP decrease the binding affinity of PhoP to the PHO boxes in the DNA, and therefore weaken control of the Pho regulon.

Since the 1970s, it is known that erythromycin is formed by condensation of 7-three carbon units provided by one propionyl-CoA as starter unit and six methylmalonyl-CoA elongation units (Cortés et al., 1990). Propionate or propanol addition to Sac. erythraea cultures enhances erythromycin production (Guo et al., 2016). Interestingly, propionate addition to erythromycin producing cultures also enhances the propionylated form of PhoP thus reducing the intensity of phosphate control (Xu et al., 2018). This is another interesting aspect of cross-talk between phosphate regulation and carbon source utilization, particularly propionate. It is unclear, however, if acylation of Omp-type response regulators, other than PhoP, in Streptomyces may also produce similar alterations of their regulatory effects. Indeed, the GlnR nitrogen regulator is modified both by phosphorylation and acetylation (see below).

#### Modification of AfsR by Phosphorylation: Involvement in the Transduction of the S-Adenosylmethionine Concentration Signal

Expression of actinorhodin and undecylprodigiosin genes in S. coelicolor is controlled by the AfsR regulator that belongs to the large size SARP family (Horinouchi, 2003). AfsR (993 amino acids) contains a DBD in its N-terminal region and an internal ATP/GTP binding pocket in the middle of the protein. The AfsR regulatory effect is exerted on expression of the afsS gene, encoding a small protein (63 amino acids) that controls expression of the act and red gene transcriptional regulators (Floriano and Bibb, 1996; Lee et al., 2002) by binding upstream of these cluster specific regulatory genes. AfsR recognizes and binds the promoter of the afsS gene, and this binding is essential since no transcription of afsS occurs in mutants lacking AfsR. The synthesis of AfsS in turns activates the act and red transcriptional regulators, and therefore enhances

the biosynthesis of actinorhodin, undecylprodigiosin, and the calcium-dependent antibiotic (CDA).

AfsR is phosphorylated at serine/threonine residues by the AfsK kinase (Hong et al., 1991; Lee et al., 2002). Phosphorylation of AfsR increases significantly the binding affinity of AfsR to the afsS promoter and therefore enhances actinorhodin, undecylprodigiosin, and CDA production. In addition, phosphorylation of AfsR modulates the ATPase activity of this regulator, that is required to rearrange the closed RNA polymerase–AfsR complex in the afsS promoter to an open transcriptionally active form. Genes orthologous to afsK and afsR occur in several Streptomyces species genomes suggesting that the mechanism of regulation through this phosphorylation cascade is important in these actinobacteria. Noteworthy, the autophosphorylating activity of AfsK is inhibited by KbpA, a protein encoded by a gene located upstream of the afsK gene in S. coelicolor. The KbpA protein binds to the unphosphorylated AfsK and inhibits its self-phosphorylation at serine/threonine residues. In AfsK–AfsR "in vitro" reaction mixtures KbpA inhibits the transmission of phosphate from AfsK to AfsR (Umeyama and Horinouchi, 2001; **Figure 1**). In summary, KbpA acts as a negative regulator in the AfsK– AfsR phosphorylation cascade, an important mechanism for the control of secondary metabolism and differentiation in some Streptomyces species, e.g., sporulation in Streptomyces griseus. Disruption of kbpA increases the phosphorylation cascade flux and results in an increase of actinorhodin production in S. coelicolor.

An important advance was the finding that the AfsK/AfsR proteins serve to integrate the signal of S-adenosylmethionine (SAM) concentration level. Methionine is one of the major methyl donors in microbial metabolism. This amino acid regulates positively or negatively the biosynthesis of many antibiotics, including actinorhodin (Kim et al., 2003), although the methionine signal transduction cascade has not been elucidated until recently. The methylation reactions are mediated by SAM, an activated form of methionine (Fontecave et al., 2004). The stimulatory effect of SAM is not observed in mutants blocked in the genes afsR, afsS, or afsK, or in the specific regulator gene actII-orf4 (Jin et al., 2011). NMR studies of the serine/threonine kinase AfsK showed that SAM interacts with the carboxy-terminal moiety of this protein kinase and through this interaction increases the autophosphorylation of the AfsK protein in a dose-dependent mode (Jin et al., 2011). These authors propose that enhanced autophosphorylation of AfsK transmits the SAM signal leading to an increase in phosphorylation of AfsR, that in turns activates afsS, and therefore the production of actinorhodin and undecylprodigiosin (**Figure 1**).

#### Interplay of PhoP With Other Regulatory Factors: Binding to Overlapping DNA Sequences

Although phosphate control of actinorhodin and undecylprodigiosin biosynthesis in S. coelicolor is mediated by PhoP, no direct binding of PhoP to the region upstream of the translation start triplet of genes of the act and red gene clusters was observed (Martín, 2004; Santos-Beneit et al., 2009b; Allenby et al., 2012).

Rather, the phosphate control of these antibiotics biosynthesis is exerted by interference of PhoP and AfsR by binding to the same sequence in the promoter region of the asfS gene. Both phosphorylated PhoP and AfsR are competing positive regulators that integrate signals derived from different inputs (Lee et al., 2002; Santos-Beneit et al., 2009b, 2011b; **Figure 1**). Although both PhoP and AfsR overlap completely there is a preference for one particular nucleotide in the sequence, different for PhoP or AfsR. Indeed, nucleotides in the PHO and AfsR boxes at positions 2, 3, 4, and 6 are invariant but the nucleotide at position 5 may be a T or a G. When there is a G at position 5 it promotes binding of PhoP, whereas a T at this position promotes binding of AfsR. AfsR and PhoP compete not only in binding the PHO box upstream of afsS but also AfsR binds to PHO boxes in the upstream regions of pstSCAB, glnR, and phoRP (Santos-Beneit et al., 2009b, 2011b).

Additional examples of overlapping recognition sequences, by PhoP and other transcriptional factors, have been reported in the last decades. In one example PhoP controls directly cdaR expression, the cluster situated regulator of the CDA in S. coelicolor. In addition to PhoP, other master transcriptional factors, including AbsA2 and AfsQ1, bind the upstream sequence of cdaR (Allenby et al., 2012). Other interesting example is the regulation of the scbA gene, encoding the γ-butyrolactone synthase (D'Alia et al., 2011) by both PhoP and ScbR regulators (Allenby et al., 2012). Both scbA and scbR are expressed from a divergent promoter and expression of scbA is regulated by ScbR, while PhoP binding to the divergent promoter represses scbA expression. Therefore, there is a close connection of the regulation by the γ-butyrolactone system and the PhoP-mediated phosphate control in S. coelicolor (Allenby et al., 2012).

# Cross Regulation of Phosphate and Carbon Sources

Cross-talk regulation between phosphate and carbon sources has been described in Corynebacterium glutamicum and also in a few Streptomyces species. Initial studies in S. lividans on carbon source control of phosphate transport through the high affinity PstSCAB transport system reveal that formation and release of the PstS protein takes place only in cultures with high concentration of fructose, galactose, or mannose but not with other carbon sources (Díaz et al., 2005). The PstS phosphate binding protein is partially released from the PstSCAB complex located in the cells envelopes and this release seems to be due to glycosylation of the PstS protein (Díaz et al., 2005; Wehmeier et al., 2009). The released PstS protein becomes not functional and unable to bind phosphate (Esteban et al., 2008). Using Northern analysis of the pstSCAB gene and Western blot assays of the PstS protein in S. lividans and S. coelicolor, Esteban et al. (2008) observed that the pstS gene expression takes place in media containing glucose or fructose as carbon sources but does not occur in complex medium. A stretch of DNA upstream of the pstS translation start triplet, including 8 degenerated repeats of a sequence of 12 nucleotides, is involved in negative control

of the transcription from the pstS promoter by fructose. This stretch of DNA is adjacent to, but does not overlap with, the well-characterized PHO box recognized by PhoP, though both DNA regions affect expression of the pstS gene. However, further detailed analysis of PhoP and the putative role of the repeated 12 nucleotides sequence in regulation by carbon sources needs to be made. Similarly, in C. glutamicum the GlxR regulator, which is a cAMP-dependent transcriptional regulator, functionally similar to CRP in Enterobacteria, regulates phosphate transport by binding to a sequence upstream of the phosphate transporter gene pstS under phosphate limiting condition and activates its expression (Panhorst et al., 2011).

Another interesting example in S. coelicolor is the regulation of glpQ1 and glpQ2, that encode glycerophosphodiester phosphodiesterases, and are regulated by PhoP and also by glycerol and other carbon sources. PHO boxes have been identified in the promoter regions of glpQ1 and glpQ2 but the putative regulatory proteins recognizing glycerol or other carbon sources have not been studied yet (Santos-Beneit et al., 2009a). Increasing evidence indicates that coordination of phosphate and carbon source utilization, e.g., through the glycolysis have an important role in the metabolism of Streptomyces species. Recently, Ordóñez-Robles et al. (2017a) reported that in the tacrolimus producer S. tsukubaensis the use of glucose (but not maltose) as carbon source promotes drastically the expression of phoR–phoP, pstS, and other key genes of the Pho regulon. The findings that in S. tsukubaensis glucose stimulates significantly the genes involved in phosphate transport, and the similar effect of fructose in S. lividans indicates that an increase in the concentration of these hexoses in the culture medium favors the uptake of inorganic phosphate to balance the carbon and phosphate metabolism through the glycolysis, pentoses phosphate pathway, and TCA cycle. Reciprocally, inorganic phosphate affects many of the genes of the carbon metabolism and this regulation is mediated by PhoP (Rodríguez-García et al., 2007; Allenby et al., 2012), including enzymes involved in glycogen metabolism and gluconeogenesis (Thomas et al., 2012).

### REGULATION OF NITROGEN METABOLISM IN ACTINOBACTERIA: THE GlnR MASTER REGULATOR

One of the most interesting examples of cross-talk between transcriptional regulators is the overlapping binding of PhoP and GlnR in the promoters of several nitrogen metabolism genes. Streptomyces and other Actinobacteria use a variety of nitrogen sources for growth. These include ammonium ions, nitrate and nitrite, and complex organic nitrogen sources including urea, amino acids, peptones, and several other nitrogen-containing compounds, including polyamines (Krysenko et al., 2017). In addition, some Actinobacteria, e.g., Frankia species are able to utilize atmospheric nitrogen for the production of organic nitrogen compounds in the cells (Nouioui et al., 2019). These different nitrogen sources are converted to glutamate and glutamine that serve as amino donors in transamination reactions to form other nitrogen containing compounds. In Streptomyces species there are two glutamine synthetases, encoded by the genes glnA and glnII, that convert glutamate to glutamine but the major glutamine synthetase activity is that of GlnA (Fink et al., 2002). Other glutamine synthetase-like enzyme, e.g., GlnA3, is known to be involved in polyamine utilization but has no glutamine synthetase activity (Krysenko et al., 2017). The ammonium ions are transported in the cells by the amtB transporter, which is clustered with genes glnK and glnD. On the other hand, nitrate and nitrite ions are reduced to ammonium by the nitrate reductases NasA (Amin et al., 2012), and nitrite reductase NirB1B2C/D (Tiffert et al., 2011). The urea is metabolized by a system encoded by the ureABC cluster.

All nitrogen metabolism genes respond to the nitrogen availability in the culture (Tiffert et al., 2008). Two regulatory genes, glnR and glnRII, control expression of several nitrogen metabolism genes at the transcriptional level (Fink et al., 2002). GlnR (267 amino acids) contains the domains defining a response regulator but is not linked to a cognate sensor kinase (Wray et al., 1991; Fink et al., 2002) and is classified as an orphan response regulator (Hutchings et al., 2004). It has been shown that GlnR in Streptomyces species is atypical in its lack of phosphorylation of the D<sup>49</sup> residue (Hong et al., 2007); in PhoP this amino acid in its phosphorylated form activates the DBD but this activation does not seems to be required in GlnR. The protein region surrounding D<sup>49</sup> is phosphorylated in serine and threonine residues and is essential for maintaining the biologically active homodimer conformation (Lin et al., 2014). Nucleotide sequences for GlnR binding upstream of many different nitrogen metabolism genes were shown to correspond to 22 nucleotides forming a direct tandem repeat of 11 nucleotides each (Tiffert et al., 2008; Sola-Landa et al., 2013). Under nitrogen limitation conditions GlnR induces expression of glnA, glnII, and the amtB–glK–glD cluster, but represses the cluster of urea utilization genes.

#### Post-translational Phosphorylation or Acetylation of GlnR Affect Nitrogen Regulation in S. coelicolor

Mass spectrometry analysis of GlnR showed that this regulator is post-translationally phosphorylated at serine/threonine residues in six positions and acetylated at four lysine residues. The degree of phosphorylation correlates well with growth on nitrogen rich medium and is reduced, and even absent, in nitrogen limiting minimal medium. DNA binding studies using EMSA showed that the phosphorylated form of GlnR does not bind to its target sequences. Moreover, initial structural studies of GlnR indicated that the serine/threonine phosphorylated residues prevent formation of the active GlnR dimeric form (Amin et al., 2016). Overimposing the amino acid sequences of GlnR with the crystal structure of PhoP it was observed that the serine/threonine residues which are phosphorylated and the lysine residues which are acetylated in GlnR are conserved with respect to the same amino acids in PhoP. Regarding modification by acetylation, four lysine residues are acetylated in GlnR in nitrogen limiting medium but only one lysine is modified in nitrogen rich conditions. Therefore, it was proposed that the

modifications in GlnR are similar to those reported in PhoP except phosphorylation at D49. However, the lysine acetylation does not appear to respond to the nitrogen availability in the medium but rather it responds to carbon sources (Amin et al., 2016), as described above for lysine acylation in PhoP in response to propionate or propanol.

#### GlnR Regulates Secondary Metabolites Biosynthesis Through Pathway Specific Regulators

Detailed studies on nitrogen regulation mediated by GlnR were performed in the model actinobacteria S. coelicolor (Tiffert et al., 2011) and in Streptomyces venezuelae (Pullan et al., 2011). Other studies have been performed in Sac. erythraea (Yao et al., 2014), in Amycolatopsis mediterranei (Yu et al., 2007), and also in S. hygroscopicus var. jinggangensis, producer of validamycin (Qu et al., 2015). Transcriptional studies of a S. coelicolor mutant deleted "in frame" in glnR, confirmed that GlnR represses actinorhodin biosynthesis, and induces undecylprodigiosin formation. These observations were supported by DNA binding studies using purified GlnR. Noteworthy, it was found that GlnR binds directly to the promoter regions of genes for the pathway specific regulators ActII-orf4 and RedZ, respectively. He J.M. et al. (2016) identified a GlnR binding nucleotide sequence, 5-GTGAC-3, in the promoter regions of actII-orf4 and redZ genes that has similarity to the nucleotide sequence of the A fragment in GlnR binding site described by Tiffert et al. (2008). Mutation of nucleotides in this putative GlnR binding site supported the conclusion that GlnR binds to the promoters of actinorhodin and undecylprodigiosin gene clusters in a sequence which is slightly different from those identified in primary nitrogen metabolism genes. It is possible that to bind these different sequences GlnR needs cooperation with either a sigma factor or other transcriptional regulators. In the case of genes for nitrate–nitrite assimilation, the recognition by GlnR is promoted by NnaR, which might act as a GlnR co-activator for nirB1 expression (Amin et al., 2012).

In S. avermitilis, which produces the antihelminthic avermectin and the unrelated macrolide oligomycin, a mutant deleted in the glnR gene reduces drastically the production of avermectin but increased the biosynthesis of oligomycin (He J.M. et al., 2016). Transcriptional studies of the two regulatory genes involved in the control of avermectin, aveR, and oligomycin, olmR, indicate that GlnR in S. avermitilis binds to the promoter regions of these two regulatory genes. A very interesting aspect of this regulation in S. avermitilis is that the regulatory gene aveR contains also consensus binding sequences for PhoP (Yang et al., 2015; Martín et al., 2017), suggesting that there are regions upstream of the avermectin regulatory gene that integrate the response to different nutritional signals.

In summary, it may be concluded that GlnR is not only a master transcriptional regulator of nitrogen metabolism but it also coordinates primary metabolism and expression of secondary metabolites biosynthetic gene cluster through binding to the promoters of specific regulators of those gene clusters.

# CROSS TALK OF PHOSPHATE AND NITROGEN REGULATION IN Streptomyces

One of the most interesting recent development on interaction between master regulators is the cross-talk between GlnR and PhoP in the control of primary metabolism and also secondary metabolites biosynthesis in Actinobacteria.

PhoP and GlnR are related proteins of the OmpR response regulators family (Hutchings et al., 2004; Martín et al., 2012b). The expression of several genes involved in nitrogen metabolism is increased in a S. coelicolor1phoP mutant (Rodríguez-García et al., 2007). The control of PhoP over these genes is exerted in two ways. On the one hand, there is a binding of PhoP to the promoter of the glnR gene decreasing its expression when phosphate is limiting (Rodríguez-García et al., 2009; Allenby et al., 2012). In addition, PhoP binds directly to the promoters of at least six other genes related to nitrogen metabolism: glnA, glnII, the amtB–glnK–glnD operon, ureA, and some additional uncharacterized genes, interfering with the GlnR activation of these genes (Rodríguez-García et al., 2009; Sola-Landa et al., 2013). The interaction of PhoP with GlnR differs depending on the relative position of the sequences recognized by each of these proteins. In the glnA promoter, there is a direct competition between PhoP and GlnR because both regulators recognize overlapping sequences. In the amtB promoter, the PhoP and GlnR binding sequences are not fully overlapping, although they are very close and the binding of PhoP may alter the correct recognition of GlnR (Sola-Landa et al., 2013; **Figure 2**).

#### The Nitrogen Regulator GlnR Controls Phosphate Transport by the PstS System in A. mediterranei

The interplay between GlnR and PhoP in S. coelicolor is well studied, but it is unknown if these conclusions are valid for other actinobacteria. A relevant example is the interaction of these regulators in A. mediterranei, a strain of industrial importance that produces rifamycin. Zhang et al. (2018) cloned the pstSCAB cluster of A. mediterranei and demonstrated its involvement in phosphate transport under phosphate limiting conditions. A mutant disrupted in the pstS gene showed a decreased production of rifamycin and, interestingly, revealed different morphology (swollen, near spherical cells rather than filamentous hyphae).

The pstSCAB operon was expressed as a transcriptional unit in A. mediterranei (Zhang et al., 2018) as reported also in S. coelicolor (Esteban et al., 2008) and confirmed by RNAseq studies (Martín-Martín et al., 2019). Nitrate has a well-known stimulatory effect on rifamycin biosynthesis by A. mediterranei. Transcriptional analysis revealed that formation and release of the PstS protein was induced by nitrate and regulated positively by GlnR. Expression of pstSCAB in a glnR-null mutant of A. mediterranei revealed a fourfold decrease of pstS transcription even in the presence of added nitrate; in other words, the nitrate stimulation of pstS expression is dependent upon GlnR. DNA binding

studies with purified GlnR proved that this regulator binds to the upstream region of pstS, thus confirming that the high affinity phosphate transport system is under the control of GlnR in this actinobacteria. As indicated above, the pstSCAB cluster of many actinobacteria is strongly regulated by PhoP and, indeed, there is a PhoP binding sequence upstream of the pstSCAB genes in many bacteria. Therefore, the DNA sequences upstream of pstSCAB constitute an integrator site that combines signals of phosphate starvation and nitrogen availability (Zhang et al., 2018).

#### INTERPLAY OF THE DEVELOPMENTAL REGULATOR MtrA WITH GlnR: COMPETITION FOR THE GlnR BINDING SITE

In 2017, it was observed that the response regulator MtrA (master transcriptional regulator A), member of the MtrAB TCS has an important role in nitrogen regulation (Som et al., 2017a,b; Zhang et al., 2017). The mtrA gene was initially discovered in Mycobacterium tuberculosis, where it has an important role in cell development and antigen secretion, and was also characterized in C. glutamicum (Möker et al., 2004). In S. coelicolor the MtrA protein is well conserved compared to the homologous proteins of M. tuberculosis and C. glutamicum. Initial studies indicated that in S. coelicolor MtrA has an important role in cell development since mutation of this regulator results in the lack of mycelium formation (bald phenotype) and also alters secondary metabolism (Som et al., 2017a,b). These studies revealed that MtrA binds a 17 nucleotides sequence consisting in two repeats of six nucleotides, separated by a five nucleotides non-conserved stretch, that surprisingly, overlap with the GlnR binding site in several nitrogen metabolism genes.

#### Role of the Two Components System MtrAB in Nitrogen Metabolism in S. coelicolor

Recently, it was experimentally demonstrated that MtrA binds to one of the three putative GlnR binding sequences upstream of the amtB–glnK–glnD cluster (Zhu et al., 2019). Previously, three putative GlnR binding sites (named a1-b1, a2-b2, and a3-b3), each consisting in two binding repeats had been identified in the amtB promoter region (Wang et al., 2012; Li et al., 2018). More recently, Zhu et al. (2019) showed that the developmental regulator MtrA recognizes the a3-b3 binding sequence which also binds GlnR efficiently. This was confirmed by EMSA DNA binding assays using the a3-b3 sequence as probe (**Figure 2**). MtrA was also able to bind the a1-b1 and a2-b2 sites, although GlnR did not recognizes the a2-b2 site indicating that minor differences in nucleotides in these binding sequences determine the affinity of binding for MtrA versus GlnR. Additionally, binding of MtrA to GlnR boxes upstream of other nitrogen metabolism genes such as glnA, glnII, nirB1B2C/D, and ureA was also found. In all cases MtrA was bound to the GlnR recognition sites although the affinity of binding changes depending on the nucleotide sequences of the different GlnR boxes. It is worth to note that the a3-b3 site was also recognized by PhoP (**Figure 2**; Wang et al., 2012; Sola-Landa et al., 2013).

Similarly, competition experiments and EMSA assays on the nirB1B2C/D promoter using different amounts of MtrA and GlnR showed that the intensity of the DNA–protein

complexes varies depending on the relative amount of MtrA to GlnR. In other words, both proteins compete for the same nucleotide sequence, demonstrating that there is an overlapping binding of the two regulators to the upstream sequence of the nirB1B2C/D cluster.

In vivo binding of MtrA to GlnR recognition sequences in different nitrogen metabolism genes was shown by ChIPqPCR analysis using anti-Flag antibodies and a S. coelicolor 1mtrA mutant transformed with the Mtr-Flag fused gene (Zhu et al., 2019). Comparative transcriptional analysis of gene expression in the parental strain and the mtrAdeleted mutant showed that several nitrogen metabolism genes including the amtB–glnk–glnD cluster, glnII, nasA, and nirB1–nirB2–nirC/D were highly overexpressed in the mtrAdeleted mutant under nitrogen rich medium whereas the urea utilization cluster was not significantly upregulated under these conditions.

These results clearly indicate that under nitrogen rich conditions MtrA represses nitrogen metabolism genes, i.e., this is a regulation opposed to that of GlnR in minimal medium. Interestingly, comparative transcriptional experiments were also performed under nitrogen limiting conditions using the mtrAor the glnR-deleted mutant, compared to the parental strain (Zhu et al., 2019); the results clearly indicated that expression of several nitrogen metabolism genes respond to the GlnR concentration in the cells but not to the MtrA levels. In other words under nitrogen limiting conditions the positive regulatory effect of GlnR dominates over the MtrA repression.

Noteworthy, although GlnR does not regulates its own expression, binding of the purified MtrA protein to a sequence upstream of glnR was found. Indeed, there was no binding of purified GlnR to the upstream sequence of the glnR gene itself, although it is recognized by MtrA. These results are difficult to explain if it is assumed that both GlnR and MtrA recognized the same binding sequences. Probably other proteins interact with MtrA and modify its affinity compared to that of GlnR.

Similarly, NtrA of S. lividans regulates expression of several nitrogen metabolism genes under nitrogen rich conditions, as occurs in S. coelicolor (Zhu et al., 2019). Furthermore, the ntrA gene has been characterized in M. tuberculosis and C. glutamicum. Initial experiments using NtrA purified protein from these actinobacteria indicate that potentially MtrA is also able to regulate nitrogen metabolism genes both in Mycobacterium and in Corynebacterium. It is likely that a similar regulatory mechanism occurs in other actinobacteria. In summary, the wide distribution of genes encoding GlnR and MtrA in several actinobacteria suggests that these two master regulators play a very important role in nitrogen metabolism.

#### Interplay of Nitrogen and Carbon Regulatory Mechanisms Mediated by GlnR

Early studies on GlnR-mediated regulation of metabolism in Streptomyces suggested that this regulator is also involved in the control of carbon sources assimilation (Tiffert et al., 2008; Pullan et al., 2011).

An interesting recent development is the finding that GlnR interacts with metabolism of many carbon sources at the transport level (Liao et al., 2015). The actinobacteria use a large variety of carbon sources (Bertram et al., 2004; Titgemeyer et al., 2007). Some of them, e.g., glucose or glycerol, are preferred carbon sources but the utilization of many other hexoses, oligosaccharides, and polysaccharides is non-constitutive and is only induced when required. In Streptomyces species, glucose is transported by a well-known permease system (van Wezel et al., 2005). In contrast 87% of the total carbohydrate carbon sources are transported by ATP binding cassettes (ABC) transporter systems (van Wezel et al., 1997; Davidson et al., 2008). Studies on the control by GlnR of several ABC sugar transporters have been done in Sac. erythraea and in S. coelicolor (Liao et al., 2015). These authors identified a sequence similar to the GlnR-binding site in the upstream region of the malEFG operon of Sac. erythraea, that encodes an ABC transport system for maltose and maltodextrines (**Figure 3**). Immediately upstream of the malEFG operon in the complementary strand and in the opposite orientation are two genes malR and aglA, encoding the MalR repressor protein and an α-glucosidase involved in the cleavage of maltose and maltodextrins to produce glucose. Both in Sac. erythraea and in S. coelicolor two MalR binding sites were detected and confirmed in DNA binding assays in the intergenic region malEFG-malR, one of which overlap with the GlnR binding sequence (Liao et al., 2015; **Figure 3**). Mutants of Sac. erythraea or S. coelicolor deleted in the glnR gene showed impaired growth in media containing maltose as only carbon source but not in media supplemented with glucose. Transcriptional studies revealed that expression of the mal operon was dependent on the GlnR regulator, i.e., there is no expression of this operon in the glnR-deleted mutant. Binding of purified GlnR to the upstream region of the mal operon was shown by DNA binding assays. Transcription of the mal operon increases in response to nitrogen starvation, indicating that the nitrogen availability controls expression of the mal transport system (**Figure 3**).

In S. coelicolor 12 out of 37 putative ABC sugar transporters were found to contain GlnR binding sites. The same situation was found in A. mediterranei, Mycobacterium smegmatis, and three additional Streptomyces: S. griseus, S. venezuelae, and S. avermitilis (Liao et al., 2015). Similarly, GlnR binding sites were found in 8 of the 19 putative ABCtype sugar transporters in Sac. erythraea. It seems that in all these cases expression of the sugar binding subunits of the respective ABC transporters was dependent on the GlnR regulator.

Altogether these findings indicate that there is a cross-talk between the limitation of nitrogen sources and the transport of several sugars by ABC systems. This interplay between the nitrogen regulation and sugar transport system protects the cell against imbalances in nitrogen and carbon abundance by controlling the utilization of alternative sugar sources. When nitrogen is limiting the GlnR levels increase, and therefore favor the utilization of alternative nitrogen (Tiffert et al., 2011) and carbon sources (Liao et al., 2015).

#### POSSIBLE INTERACTION BETWEEN THE cAMP RECEPTOR PROTEIN AND THE GlnR NITROGEN REGULATOR

Glucose is the preferred carbon source by Streptomyces and many other bacteria, and there is a regulation by glucose of alternative non-preferred carbon sources utilization. This regulation in Enterobacteria is mediated by the cAMP receptor protein (CRP) and is dependent on the cAMP levels in the cells. Several studies have reported that in S. coelicolor, the carbon catabolite regulation of alternative carbon sources by glucose is not mediated by a CRP-homologous protein. Instead, carbon catabolite regulation is mediated by the glycolytic enzyme glucose kinase (Ikeda et al., 1984; Angell et al., 1994; van Wezel et al., 2007; Romero-Rodríguez et al., 2016a,b). However, in Sac. erythraea there is a CRP-like protein that seems to be involved in carbon catabolite control of utilization of alternative carbon sources. This regulation involves binding of the CRP-like protein to promoter regions of several ABC-type sugar transporters. Bioinformatic studies revealed that there are putative CRP binding sequences in the upstream regions of several ABC sugar transporters in Sac. erythraea and possibly also in other actinobacteria (Liao et al., 2015). This finding adds a new level of regulation by interaction of CRP with promoters of ABC sugar transporters, such as malEFG. Indeed, Liao et al. (2015) proposed that CRP binds to the same region recognized by GlnR in the intergenic region between malEFG and malR-aglA (**Figure 3**). However, the putative interaction at this level needs further experimental confirmation. In summary, the possible interaction between the CRP protein, the GlnR and the MalR regulator in the mal operon for maltose/maltodextrins metabolism, is another provocative example of interaction of several regulatory proteins on the promoter region of key sugar transporters and therefore constitutes an elaborate mechanism of control by interacting transcriptional regulators.

### INTERPLAY BETWEEN THE TWO COMPONENTS SYSTEM AFSQ1/Q2 AND THE NITROGEN REGULATOR GlnR

The two components system AfsQ1–AfsQ2 (of which afsQ1 encodes the response regulator) was initially found as a result of the observation that it increases the production of actinorhodin and undecylprodigiosin in S. coelicolor (Shu et al., 2009). In the opposite orientation of afsQ1–AfsQ2 is located the sigQ gene, encoding a putative sigma factor involved in antibiotic biosynthesis, that is regulated by AfsQ1 (Wang et al., 2013).

The observed regulation takes place only in minimal medium supplemented with 75 mM glutamate. Transcriptional studies using cell grown under these glutamate rich conditions revealed that the expression of actII-orf4, cdaR, and redZ genes was significantly reduced in the afsQ1-Q2 mutant. EMSA studies with purified AfsQ1 revealed that, indeed, this response regulator was able to bind to the actII-orf4, cdaR, and redZ promoters but not to the redD promoter region. The AfsQ1 regulator enhances the expression of sigQ but not that of the afsQ1–afsQ2 operon itself. In other words, there is a bidirectional transcription region that expressed afsQ1-Q2 in one orientation and sigQ in the other and AfsQ1 does not regulates directly its own expression. The AfsQ1 binding sequence was first identified in the promoters of sigQ and redZ and later also in actII-orf4 and cdaR. In all cases, except in actII-orf4, a 16 nucleotides sequence was protected by AfsQ1. This sequence, GTnAC-n6-GTnAC, consists of two repeated GTnAC nucleotides sequences separated by an intervening non-conserved six nucleotides stretch. Surprisingly, the sequence recognized by AfsQ1 in actII-orf4 does not adjust to the consensus five nucleotides repeated sequence indicating that there are factors, still unknown, that affect the specificity of the binding sequences recognized by AfsQ1 (Wang et al., 2013). These authors using a matrix based on the conserved 16 nucleotides binding sequence were able to recognize 17 sites that correspond to: (1) the upstream regions of nitrogen metabolism genes, such as glnA and nirB1; (2) the pstS gene involved in high affinity phosphate transport; and (3) several bidirectional promoters related to cell differentiation including whiD and bldM (Wang et al., 2013).

The putative AfsQ1-binding sequence contains the same five nucleotide core repeats that form part of the 16 nucleotides consensus sequence recognized by GlnR. EMSA studies with purified AfsQ1 revealed that this regulator binds to the promoter regions of several nitrogen metabolism genes but not to the promoter of glnR or glnII indicating that there are differences in the affinity of distinct regulatory proteins for nitrogen metabolism genes. Transcriptional studies using a mutant deleted in afsQ1–afsQ2 showed that AfsQ1 acts as a transcriptional repressor of several nitrogen assimilation genes under conditions of high glutamate concentration. Competition DNA binding experiments using both purified AfsQ1 and GlnR proteins lead to the conclusion that both regulators compete for the same nucleotide sequence in the upstream region of several nitrogen assimilation genes; in other words, AfsQ1 represses the expression of genes encoding enzymes for assimilation of alternative nitrogen sources in conditions of glutamate excess by competing with GlnR that would activate these promoters under nitrogen limitation conditions.

An important observation is the finding that AfsQ1 regulates expression of the pstSCAB cluster that is also strongly regulated by PhoP and AfsR (Santos-Beneit et al., 2009b; Martín et al., 2012a). Indeed, PhoP also regulates expression of several nitrogen assimilation genes as described above in this article. Taken together all this information suggests that these five master regulators PhoP, AfsR, AfsQ1, GlnR, and MtrA, interact in the fine tuning of regulation of key genes in carbon, nitrogen, and phosphate metabolism. This interaction protects the cell against unbalance metabolism due to lack or excess of some key nutrients, such as ammonium or glutamate (**Table 1** and **Figure 2**). Not all five regulators bind to the same integrative node; two, three, or more transcriptional factors may recognize certain sequences whereas combinations of other factors regulate expression of genes at alternative nodes.

# OVERLAPPING TRANSCRIPTIONAL FACTORS REGULATE EXPRESSION OF AtrA

Other interesting example of interplay between transcriptional factors is the cross regulation between PhoP and the AdpA regulator, that in turn responds to the γ-butyrolactone signaling mechanism. In the quorum sensing system the γ-butyrolactones are recognized by specific butyrolactone receptor proteins (Brps) (Horinouchi, 2002; Ohnishi et al., 2005) that when combined with its own cognate butyrolatone derepressed the binding of the Brps to the pleiotropic regulator AdpA (Takano, 2006). This occurs in different Streptomyces species (Higo et al., 2012; Salehi-Najafabadi et al., 2014) and has been studied with detail in Streptomyces roseosporus. The butyrolactone triggered signaling cascade in S. roseosporus is known to proceeded through the TetR-type regulator AtrA, that controls expression of the biosynthesis of the lipopeptide antibiotic daptomycin (Mao et al., 2015). Expression of AtrA is controlled by the pleiotropic regulator AdpA. Recently, Zheng et al. (2019) observed that expression of AtrA is regulated in addition by PhoP. Binding of these two regulators overlaps in the upstream region of the atrA gene as confirmed by DNA binding studies. Moreover, in turn PhoP indirectly controls expression of the adpA gene (Allenby et al., 2012). Therefore, PhoP controls AtrA at two different levels in a cumulative form. In conclusion, this is other excellent example of overlapping regulation exerted by two different master regulators, PhoP and AdpA, on the expression of the atrA gene. Since AdpA is regulated by the γ-butyrolactone quorum sensing system the overall mechanism connects the regulation by small ligands (Xu et al., 2010; Martín et al., 2019) with the regulation exerted by the master regulators of phosphate metabolism.

#### BINDING OF INTERACTING TRANSCRIPTIONAL FACTORS: THE CONCEPT OF SIGNAL INTEGRATIVE DNA REGIONS

Increasing evidence indicates that there are numerous examples of overlapping sequences that integrate nutritional and environmental stress signals in Streptomyces species (Romero-Rodríguez et al., 2018). As described above six or even more transcriptional factors may bind to a specific promoter region. Therefore, the concept of a metabolic safety net is of interest to understand how the interaction of these stress signals, through integrator DNA sites, protect the metabolism of the cells against

nutritional and environmental damage by achieving a balance of the key metabolic pathways.

The evidence that different master transcriptional factors interact on overlapping DNA sequences has been supported by studies on ChIP-on-chip analysis. This technique that uses antibodies against one of the binding proteins immunoprecipitates the DNA region bound by the specific factor targeted by the antibody and also results in coprecipitation of other interacting proteins (Allenby et al., 2012). This study indicated that there are many more binding sites for PhoP or other transcriptional factors, e.g., DasR (Swia¸tek-Połaty ´ nska ´ et al., 2015), than those expected from previous transcriptomic studies. The authors propose that interaction of PhoP with other transcriptional factors increase the affinity of PhoP for some poorly conserved PHO boxes. For example, Allenby et al. (2012) observed a massive binding of PhoP to the cpk gene cluster corresponding to internal sequences of cpkB and cpkC genes involved in coelimycin P1 biosynthesis (Gomez-Escribano et al., 2012). It is possible that this intense PhoP binding may be due to some interaction with some other factors that control those genes in the cpk gene cluster.

Moreover, these master regulators in turn control several other transcriptional factors at lower levels, that regulates differentiation, osmotic stress responses, oxygen metabolism, and secondary metabolite biosynthesis.

#### ADDITIONAL INTERTWINING LAYERS OF REGULATION REINFORCE THE SAFETY NET

In this article we have focused on interactions of transcriptional factors that control expression of different gene clusters in Streptomyces. These interactions constitute the backbone of the safety net but there are many other regulatory factors at distinct levels that provide different layers of control (**Figure 4**). These include sigma factors and anti-sigma factors (Paget, 2015), guanosine tetraphosphate, guanosine pentaphosphate, and novel cyclic guanine and adenine nucleotides, all of which act the transcriptional level, and ancillary subunits of the RNA polymerase. In this article we are concentrating on the interactions at the transcriptional level. Life of Actinobacteria is subject to frequent changes in nutritional and environmental conditions and this requires adaptation to the stress produced by those changes. To adjust to those changing conditions Actinobacteria have developed a variety of alternative sigma factors that recognize and activate adequate promoters in response to the novel conditions. Housekeeping genes that are strictly required for growth are transcribed by RNA polymerase that contain sigma factors of the Sig70 family, namely the HrdB factor; however, there are several paralogous genes similar to hrdB (e.g., hrdA, hrdC, and hrdD), whose function is still partially unclear. Sequencing of Streptomyces genomes has revealed that these actinobacteria contain many alternative sigma factors, 69 (66 in the chromosome and 3 in plasmids) in S. coelicolor (Bentley et al., 2002) and 62 in S. lividans (Rückert et al., 2015; Rebets et al., 2018), whereas Escherichia coli contains only 7 sigma factors and B. subtilis has 18. Alternative sigma factors play an important role on expression of genes involved in many aspects related to growth, differentiation, stress response, and secondary metabolites biosynthesis (Sun et al., 2017). Therefore, the interactions of these alternative sigma factors with promoter regions that also are regulated by the above described transcriptional factors provide an intertwining mechanism of control of gene expression. Several sigma factors are antagonized by anti-sigma proteins (19 in the S. lividans genome). The result is that the interaction of the cognate sigma factor with the RNA polymerase is blocked by binding of the corresponding antisigma factor. This adds an additional level of control on the interaction of transcriptional factors and sigma factors in the safety network. An additional protein that interacts with the RNA polymerase core enzyme is the ancillary protein RpoZ that in S. coelicolor is controlled by the phosphate regulator PhoP; mutant in rpoZ are impaired in cell differentiation and secondary metabolites biosynthesis (Santos-Beneit et al., 2011a).

Another group of molecules that interact with the RNA polymerase are the guanosine hyperphosphorylated molecules, namely guanosine tetra-phosphate (ppGpp) and guanosine pentaphosphate (pppGpp) These compounds initially characterized in E. coli have been found in many other bacteria including actinobacteria. These molecules are formed in response to carbon and nitrogen limitation, particularly in amino acid starvation caused be the absence of an essential amino acid (Dalebroux and Swanson, 2012) and constitutes the so called stringent response that limits transcription of several genes involved in vegetative growth (e.g., genes involved in ribosomal protein synthesis) and stimulates expression of stress response genes. The highly phosphorylated guanosine nucleotides are synthesized by a ppGpp synthetase encoded by the relA gene. In Enterobacteria the ppGpp is degrade by a ppGpp phosphohydrolase whereas in Streptomyces the relA gene appears to have both biosynthetic and degradative activities, being named relA/spoT (Martínez-Costa et al., 1998). Based on the information derived from S. coelicolor relA gene it was proposed that accumulation of ppGpp in response to nitrogen limitation increases biosynthesis of secondary metabolites (Bibb, 2005). Indeed, this increase is mediated by the extracytoplasmic factor SigT, thus providing an interesting example of interaction between sigma factors and stringent response (Feng et al., 2011). Mutants deleted in the relA gene showed greatly reduced transcription of the actII-orf4 and redD genes, which are required for expression of the act and red gene clusters (Chakraburtty and Bibb, 1997; Hesketh et al., 2001). Global transcriptomic studies comparing the S. coelicolor parental strain and the relA-null mutant showed that relA controls expression of 598 genes most of them involved in growth and differentiation and some of them affecting secondary metabolite biosynthesis (Hesketh et al., 2007). Among the impacted genes was the major vegetative sigma factor HrdB that is repressed following ppGpp accumulation. The biosynthesis of alternative sigma factors was also affected in the relA-deleted mutant (Hesketh et al., 2001). However, more recent information indicates that biosynthesis of ppGpp is not strictly required for triggering biosynthesis of secondary metabolites, even in some Streptomyces species such

as Streptomyces clavuligerus deletion of the relA-gene resulted in a 2–2.5-fold overproduction of cephamycin C and clavulanic acid, indicating that (p)ppGpp may influence positively or negatively antibiotic biosynthesis (Gomez-Escribano et al., 2008). The highly phosphorylated guanine nucleotides form part of a family of second messengers that have been paid more attention recently; these include in addition to ppGpp and pppGpp, cyclic adenosine 3<sup>0</sup> 5 0 -monophosphate (cAMP), cyclic guanosine 3<sup>0</sup> 5 0 -monophosphate (cGMP), and purine cyclic dimers as bis (3<sup>0</sup> ,50 )-cyclic di guanosine monophosphate (c-di-GMP) or cyclic dimer adenosine monophosphate (c-di-AMP) (Latoscha et al., 2019). There is still little information on these second messenger nucleotides in Streptomyces species. A putative adenylate cyclase encoded by the cyaA gene was found in S. coelicolor (Danchin et al., 1993). Bioinformatic analysis reveals that the S. coelicolor adenylate cyclase encoded by cyaA has the same domains as the adenylate cyclase of M. tuberculosis. This mycobacterial protein has been crystalized and has been proposed to have a role on acid pH recognition (Tews et al., 2005). The S. coelicolor cyaA-null mutant is unable to neutralize the acidic pH produced during growth of the culture and is bald. The aerial mycelium formation was restored when exogenous cAMP was added to the mutant culture (Süsstrunk et al., 1998). Interestingly, a CRP protein has been identified in S. coelicolor and found to bind cAMP in vitro (Derouaux et al., 2004). However, cAMP does not seem to play a role similar to that of carbon catabolite regulation by glucose of alternative carbon sources. Noteworthy, immunoprecipitation studies have shown that CRP co-precipitates with the upstream region of actII-orf4 and redD (Gao et al., 2012). cAMP has a CRP-dependent inducing effect on expression of these regulatory genes. These finding are consistent with the observation that CRP mutants are impaired in the biosynthesis of actinorhodin and undecylprodigiosin. In conclusion, it seems that the cAMP–CRP complex interacts with the upstream region of actII-orf4 and red and, therefore, regulates expression of the corresponding gene clusters. The possible mechanism, if any, of cAMP–CRP on carbon catabolite regulation in Streptomyces needs to be further studied (see the section "Possible Interaction Between the cAMP Receptor Protein and the GlnR Nitrogen Regulator").

Another modified nucleotide, c-di-GMP, has been studied recently in S. venezuelae (Al-Bassam et al., 2018). This nucleotide is synthesized by a dimeric guanylate cyclase using two GMP molecules as substrate. The c-di-GMP level is maintained by degradation catalyzed by a phosphodiesterase

to form a linear GMP di-nucleotide. Similar enzymes exist in S. coelicolor. Disruption of genes encoding these c-di-GMP biosynthetic and degradative enzymes impaired actinorhodin and undecylprodigiosin formation but the mechanism involved in this effect is unknown. Cell differentiation and spore pigmentation were also affected in the c-di-GMP-null mutant (Tran et al., 2011; Hull et al., 2012). The last of this group of modified nucleotides is bis (3<sup>0</sup> ,50 )-cyclic di adenosine monophosphate (c-di-AMP) that is formed by an adenylate cyclase from two ATP molecules. The genes forming c-di-AMP have been studied in B. subtilis and they are also present in actinobacteria including Streptomyces species (Witte et al., 2008). The intracellular balance of c-di-AMP is maintained by at least two different mechanisms, one of them involves secretion by a multidrug efflux pump and the second implicates hydrolysis by a phosphodiesterase. Interestingly, the gene encoding this enzyme, that occurs in other bacteria, is not present in Streptomyces genomes. The intracellular c-di-AMP levels affect many aspects of cell metabolism, including cell wall preservation and DNA integrity maintenance, but the major aspect regulated by this nucleotide is the ion homeostasis of the cell, particularly osmotic stress balance (Commichau et al., 2018). The interactions between the different second messenger nucleotides and phosphate control is still obscure. It is well known that synthesis of high levels of hyperphosphorylated guanine nucleotides results in a decrease of GTP pool in the cell and vice versa the level of GTP increases in the guanosine tetraphosphate null mutants. The cellular signals that regulate the levels of these modified nucleotides and their relationship with the global phosphate control need to be further investigated.

A further layer of interactions in the protection net is exerted by small ligands that bind to receptor proteins and caused distinct regulatory effects. These ligands are small organic molecules different from normal nutrients that are produced by bacteria and fungi in complex habitats and exert effect in the same species or between different species (Keller and Surette, 2006; Shank and Kolter, 2009; Straight and Kolter, 2009). Several of these ligands are themselves specialized metabolites that are secreted in small amount but exert potent communication effects on competing microorganisms. One group of these ligands are the so-called quorum sensing ligands (Bassler and Losick, 2006). One of the typical classes of quorum sensing molecules are the γ-butyrolactones that are well known in Streptomyces and related actinobacteria (Takano, 2006; Nodwell, 2014). The distinct γ-butyrolactones that differ in the modifications of their short fatty acid chains are recognized by γ-butyrolactone receptor proteins. Another important group of receptor proteins, that were initially designated pseudo-γbutyrolactone receptor proteins (Xu et al., 2010), and have been renamed specialized metabolites receptor proteins (Martín and Liras, 2019), do not recognize γ-butyrolactones but interact with different biosynthetic intermediates and natural products that act as signals in the bacterial intercommunication. In addition to the specialized metabolite receptor proteins other well-known molecules recognize small ligands; this is the case of the TetR-class of transcriptional regulators. The TetR cluster regulators have a ligand binding site in the carboxyterminal region that interact with diverse signal molecules (Cuthbertson and Nodwell, 2013). In conclusion, the number of reported examples of interactions between ligands and transcriptional factors is increasing and the overall picture is that these additional layers of regulation reinforce the safety net with a more intricate connections between the nods. In addition, translational and post-translational regulatory mechanisms also reinforce the safety net.

# Related Protection Nets in Other Bacteria

This article focuses on the protection net in Streptomyces species and related actinobacteria but equivalent protection nets exist in many other bacteria and in lower eukaryotes. However, the molecular mechanisms that protect the cells in response to nutrient starvation and environmental stresses are rather different from those observed in Streptomyces. For example, in E. coli there are only seven alternative sigma factors. The mechanism of control of the Pho regulon in E. coli is somehow different from that observed in Streptomyces species (Hsieh and Wanner, 2010; Antiqueira et al., 2012). In contrast to the strict dependence on the sensor histidine kinase PhoR in S. coelicolor (Fernández-Martínez et al., 2012), in E. coli the phosphorylation of the response regulator PhoB occurs also by acetyl-phosphate. Moreover, the phosphorylation of the PhoB regulator is also performed by histidine kinases of other twocomponents systems, as is the case of CreC histidine kinase (Wanner, 1993; Hsieh and Wanner, 2010). Interplay of regulatory systems has been also studied in B. subtilis and again there are differences with respect to the Streptomyces network; for example, the phosphate starvation response is mediated by a phosphorelay mechanism that involves at least three different steps in a cascade through the PhoR–PhoP TCS, the Spo0 system, and the ResD/ResP system (Hulett, 1996). In conclusion, it is clear that different bacteria have adapted their control mechanisms to the habitat in which they survive and the network of response mechanisms is adapted to those specific niches.

# CONCLUSION

All available evidence indicates that there are signal integration sites in the genome of Streptomyces that are recognized by interacting transcriptional factors.

These nucleotide sequences upstream of key genes constitute integrative nodes that serve to compensate the unbalance produced by the lack of a key nutrient or for its excess that would impair the cell metabolism. Probably there are several hundreds of these integrator sites; however, it remains to be confirmed that some of these sites serve to integrate signals for the major transcriptional factors such a PhoP, GlnR, MtrA, and so on, whereas other sites probably integrate two or a few signals as appears to be the case of the atrA promoter region that is bound by PhoP and AdpA. It will be of utmost interest to establish the hierarchy of the integration sites to get a clear idea of the role of these sites on regulation of overall metabolism,

particularly to advance our knowledge of the system biology of these actinobacteria.

Finally, an interesting conclusion is that mutations that initially may be designed to change a specific regulatory phenomenon that affects a single cluster may, however, alter the expression of many different clusters through the interplay of these major transcriptional factors. An intriguing question is whether the signal integration sites in the DNA correspond

#### REFERENCES


to open "unprotected DNA region" at difference of other DNA regions that may be protected by histone-like proteins.

#### AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.






expression of developmental regulatory genes. Front. Microbiol. 8:2013. doi: 10.3389/fmicb.2017.02013


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2020 Martín and Liras. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# sRNA scr5239 Involved in Feedback Loop Regulation of Streptomyces coelicolor Central Metabolism

Franziska Engel<sup>1</sup> , Elena Ossipova<sup>2</sup> , Per-Johan Jakobsson<sup>2</sup> , Michael-Paul Vockenhuber<sup>1</sup> \* and Beatrix Suess<sup>1</sup> \*

<sup>1</sup> Synthetic Genetic Circuits, Department of Biology, Darmstadt University Technology, Darmstadt, Germany, <sup>2</sup> Division of Rheumatology, Department of Medicine, Karolinska Institutet, Karolinska University Hospital, Solna, Sweden

#### Edited by:

Yvonne Mast, German Collection of Microorganisms and Cell Cultures GmbH (DSMZ), Germany

#### Reviewed by:

Bohdan Ostash, Ivan Franko National University of Lviv, Ukraine Antonio Rodríguez-García, Instituto de Biotecnología de León, Spain Paula Yagüe, University of Oviedo, Spain

#### \*Correspondence:

Michael-Paul Vockenhuber vockenhuber@bio.tu-darmstadt.de Beatrix Suess bsuess@bio.tu-darmstadt.de

#### Specialty section:

This article was submitted to Microbial Physiology and Metabolism, a section of the journal Frontiers in Microbiology

> Received: 04 October 2019 Accepted: 24 December 2019 Published: 23 January 2020

#### Citation:

Engel F, Ossipova E, Jakobsson P-J, Vockenhuber M-P and Suess B (2020) sRNA scr5239 Involved in Feedback Loop Regulation of Streptomyces coelicolor Central Metabolism. Front. Microbiol. 10:3121. doi: 10.3389/fmicb.2019.03121 In contrast to transcriptional regulation, post-transcriptional regulation and the role of small non-coding RNAs (sRNAs) in streptomycetes are not well studied. Here, we focus on the highly conserved sRNA scr5239 in Streptomyces coelicolor. A proteomics approach revealed that the sRNA regulates several metabolic enzymes, among them phosphoenolpyruvate carboxykinase (PEPCK), a key enzyme of the central carbon metabolism. The sRNA scr5239 represses pepck at the post-transcriptional level and thus modulates the intracellular level of phosphoenolpyruvate (PEP). The expression of scr5239 in turn is dependent on the global transcriptional regulator DasR, thus creating a feedback loop regulation of the central carbon metabolism. By post-transcriptional regulation of PEPCK and in all likelihood other targets, scr5239 adds an additional layer to the DasR regulatory network and provides a tool to control the metabolism dependent on the available carbon source.

#### Keywords: sRNA, Streptomyces, DasR, carbon metabolism, phosphoenolpyruvate, PEPCK

# INTRODUCTION

Streptomycetes are Gram-positive, filamentous, GC-rich, soil-dwelling bacteria. They are best known for their complex life cycle and their ability to produce a wide range of secondary metabolites. These include two thirds of all known antibiotics but also other bioactive compounds such as immunosuppressants, antivirals, and herbicides (Flärdh and Buttner, 2009). This biochemical potential is associated with a large genome size (usually 7–12 Mbp) and requires an equally large repertoire of regulatory proteins to control the network and adapt to changing environmental conditions (Bentley et al., 2002). Streptomyces coelicolor serves as a model organism for this large group of bacteria. Its regulatory repertoire includes ∼12% of its protein coding genes but also a large number of regulatory RNAs are known. The function of most of them, however, remains unknown (Pánek et al., 2008; Swiercz et al., 2008; D'Alia et al., 2010; Vockenhuber et al., 2011; Moody et al., 2013; Jeong et al., 2016; Setinova et al., 2017).

Small non-coding RNAs (sRNAs), approximately 50–500 nucleotides (nts) in length, are found in a broad range of bacteria and play an important role in the post-transcriptional regulation. Most small non-coding RNA act by base-pairing with their target mRNAs, which may affect both stability and/or translation of the target mRNA in a positive or negative manner (reviewed in Romby and Wagner, 2012). Depending on their genomic context, sRNAs are divided

**75**

into trans-encoded regulatory RNAs, which are encoded at different genomic locations from their target genes and often share only limited complementarity with them, and cis-encoded (antisense) regulatory RNAs. They are transcribed opposite to annotated genes and share complete complementarity with their targets. Most knowledge about sRNAs, their targets and regulatory networks has so far been derived from Gram-negative bacteria such as Escherichia coli and Salmonella (Frohlich et al., 2012; Guo et al., 2014; Porcheron et al., 2014; Kim et al., 2019). In contrast, only a few sRNAs in streptomycetes have been experimentally characterized so that their function is known. Examples include scr4677, which is thought to impact the actinorhodin production under specific growth conditions (Hindra et al., 2014) and scr3097, which in combination with a riboswitch influences rpfA expression post-transcriptionally (St-Onge and Elliot, 2017). Furthermore, there are the antisense RNA cnc2198.1 that regulates glutamine synthase glnA (D'Alia et al., 2010) and scr5239. The latter was identified using a deep sequencing approach (Vockenhuber et al., 2011). scr5239 expression is constitutive under several stress and growth conditions but dependent on the nitrogen supply. It is conserved in two thirds of all currently available Streptomyces genomes. The 159 nt long sRNA consists of five stem-loops P1–P5, of which stem P4 is involved in the interaction with both currently known target mRNAs. These targets – the genes for the methionine synthase metE and the agarase dagA – are crucial for both primary and secondary metabolisms, as they are important for methionine synthesis and the degradation and utilization of agar as a carbon source. Whereas S. coelicolor is the only known Streptomyces species that carries the agarase gene dagA, metE is conserved in a wide number of streptomycetes (Vockenhuber et al., 2011; Vockenhuber and Suess, 2012; Vockenhuber et al., 2015). Since non-coding RNAs are known to often control more than one target, and because of its remarkable conservation, we aimed to identify further targets of scr5239. Our previous studies indicated that in contrast to the majority of the characterized sRNAs to date, scr5239 did not induce degradation of the both validated target mRNAs metE and dagA (Vockenhuber et al., 2011; Vockenhuber et al., 2015). Therefore, we decided to carry out a proteomics study to identify new targets controlled by scr5239.

Here, we present the characterization of a new sRNA target that resulted from the proteomics study, the phosphoenolpyruvate carboxykinase (PEPCK, SCO4979). PEPCK is a key enzyme of the primary metabolism as it connects glycolysis with the tricarboxylic acid (TCA) cycle and is thought to catalyze the first step of gluconeogenesis in all organisms. In the presence of GTP it catalyzes the conversion of oxaloacetate (OAA) to phosphoenolpyruvate (PEP). This reaction is the rate-limiting step in the metabolic pathway that produces glucose from lactate and other precursors derived from the TCA cycle (Delbaere et al., 2004). Here we show that scr5239 controls PEPCK and thus the level of the metabolite PEP. scr5239 itself is controlled by DasR, one of the most important pleiotropic regulators of the primary and secondary metabolism in S. coelicolor. It controls more than 50 genes including the biosynthetic pathways for several antibiotics (Swi ˛atek-Połaty ´ nska ´ et al., 2015), developmental control (Rigali et al., 2006) or the response to cold shock stress (Nazari et al., 2013). We propose that scr5239 adds an additional layer to the DasR regulatory network by creating a feedback loop regulation for the control of the central metabolism.

# MATERIALS AND METHODS

# Cultivation of S. coelicolor

For growth on solid medium, 10<sup>8</sup> spores per 300 µl were pregerminated (Kieser et al., 2014), plated out on R2YE with cellophane overlays and incubated at 30◦C for 3–4 days. For growth in liquid medium, 10<sup>8</sup> spores per 50 ml were pregerminated and incubated at 28◦C under continuous shaking for 3–4 days in Jasenka medium (10% sucrose, 3% tryptic soy broth, 0,1% MgCl<sup>2</sup> and 0,1% yeast extract).

# Plasmids and S. coelicolor Strains

A list of all plasmids used may be found in **Table 1**. All integrating plasmids were constructed based on pAR933a (Rodríguez-García et al., 2005). It contains origin of replication for maintenance in E. coli, an apramycin resistance gene for selection in E. coli and S. coelicolor, an integrase gene and attachment site of the phiC31 phage and an origin of transfer. The plasmids used here all enable a stable integration of one single copy per chromosome (Kuhstoss et al., 1991), avoiding possible problems associated with the copy number of multicopy plasmids. The luciferase gene was excised from pAR933a using XbaI and SpeI and a 3x FLAGtag (for western blot detection) was inserted, thus converting the XbaI site to an MfeI site. Then, an insert containing two BsaI sites for golden gate cloning was cloned into the SpeI/MfeI digested vector resulting in pGold\_F3. E. coli ET12567/pUZ8002 was used to transfer the plasmids into S. coelicolor via intergeneric conjugation (Kieser et al., 2014). A list of S. coelicolor strains used may be found in **Table 2**.

#### Mass Spectrometry-Based Quantitative Proteomics

For quantitative liquid chromatography mass spectrometry (LC-MS/MS) analysis, the S. coelicolor M145 and the sRNA overexpression and deletion strains (scr5239+ and 1scr5239, respectively) where grown on solid R2YE medium as described above. Cells where harvested at the end of exponential growth when the mycelium just started to turn red. Cell lysis and whole proteome preparation were done as described in section "SDS-PAGE and Western Blot Analysis."

Proteins were precipitated from the lysates using ReadyPrepTM 2-D Cleanup Kit (Bio-Rad). Obtained protein pellets were suspended in dissolving buffer containing 2% SDS, 8 M urea, 25 mM HEPES, 1 mM DTT (pH 7.4) and the protein concentration was measured using Bradford assay (Bio-Rad). For the digestion, aliquots containing 200 µg of total protein amount were added onto Microcon-10 kDa Centrifugal Filter Units (Millipore) and processed using FASP protocol (Wisniewski et al., 2009) with modified digestion buffer containing 0.2% sodium deoxycholate, 0.25 M urea, in 100 mM HEPES, pH 7.6.

#### Engel et al. sRNA Regulates Central S. coelicolor Metabolism

#### TABLE 1 | List of plasmids used in this study.

fmicb-10-03121 January 18, 2020 Time: 17:49 # 3


After digestion, sodium deoxycholate was removed from the samples by precipitation in the presence of formic acid followed by centrifugation at 12,000 × g. Concentration of obtained peptides was measured using Lowry assay. Equal amount of protein digests (100 µg) from each sample were labeled with Tandem Mass Tag (TMT) 6-plex Isobaric Mass Tagging Kit (Thermo Fisher Scientific) according to the manufacturer's instructions. Labeled samples were dissolved in loading buffer (3% ACN, 0.1% FA), resulting in a final concentration of 10 µg/µl prior to LC-MS/MS analysis. Before analysis on the Q Exactive (Thermo Fischer Scientific), peptides were separated using an Agilent 1200 nano-LC system. Samples were trapped on a Zorbax 300SB-C18, and separated on a NTCC-360/100-5-153 (Nikkyo Technos, Ltd.) column using



a gradient of A (5% DMSO, 0.1% FA) and B (90% ACN, 5% DMSO, 0.1% FA), ranging from 5 to 37% B in 240 min with a flow of 0.4 µl/min. The Q Exactive was operated in a data-dependent manner, selecting top five precursors for fragmentation by HCD. The survey scan was performed at 70,000 resolution from 300 to 1700 m/z, with a max injection time of 100 ms and target of 1 × 10<sup>6</sup> ions. For generation of HCD fragmentation spectra, a max ion injection time of 500 ms and AGC of 1 × 10<sup>5</sup> were used before fragmentation at 30% normalized collision energy, 35,000 resolution. Precursors were isolated with a width of 2 m/z and put on the exclusion list for 70 s. Single and unassigned charge states were rejected from precursor selection.

Acquired MS raw files were searched against the UniProtKB/S. coelicolor database and filtered to a 1% FDR cut off. Ion mass tolerance of precursors equals ±10 ppm, whereas the fragments have a more limited mass tolerance of 0.02 Da for HCD-FTMS. The search algorithm assessed tryptic peptides with maximum one missed cleavage; carbamidomethylation (C), TMT 10-plex (K, N-term) as fixed modifications and oxidation (M) as variable modification. Only unique peptides in the data set were used for quantification, while reporter ions were quantified by Proteome Discoverer on HCD-FTMS tandem mass spectra (integration tolerance 20 ppm).

#### SDS-PAGE and Western Blot Analysis

A total of 1 ml liquid culture or 100 mg mycelium from solid medium were harvested and mixed with 1 ml 1× ZAP (50 mM NaCl, 50 mM Tris–HCl pH 8.0, 10% glycerol, 10 mM PMSF). The subsequent disruption of the cells was done using 200 µl glass beads (0.4 mm diameter) and the FastPrep-24 instrument (MP Biomedicals) for 4 × 1 min at 5.5 m/s. Mycelial debris was removed by centrifugation at 4◦C. Protein concentration of the supernatant was determined by Bradford assay (Bradford, 1976). Twenty to fifty µg of crude extract were separated by a 6% or 12% SDS-PAGE. 0,5% 2,2,2-trichloroethanol (TCE) in the polyacrylamide gels allowed fluorescent detection of proteins using the ChemiDocTM MP Imaging System (Bio-Rad). Gels were blotted using the Trans-Blot <sup>R</sup> TurboTM Transfer System (Bio-Rad). Membranes were blocked for 1 h in 2% ECLTM Blocking Agent (GE Healthcare) in TBS-T (20 mM

Tris base, 150 mM NaCl and 0.1% Tween 20), followed by incubation with a Monoclonal ANTI-FLAG <sup>R</sup> M2 antibody (Sigma-Aldrich #F1804; 1:50000 in 2% ECLTM Blocking Agent in TBS-T) for 1 h at RT and 3 × 10 min wash in TBS-T. Blots were developed using ECLTM Select Western Blotting Detection Reagent (GE Healthcare) and signals detected with the ChemiDocTM MP Imaging System. Protein bands were quantified using Image Lab Software 6.0.1. Total protein was used as loading control.

#### RNA Isolation

Total RNA was isolated following Vockenhuber et al. (2015). In brief, 1 ml culture or ∼100 mg mycelium were harvested and resuspended in 300 µl lysis buffer (10 mM sodium acetate, 150 mM sucrose, pH 4.8). Glass beads (200 µl, 0.4 mm diameter) and 300 µl acidic phenol were added. Cells were disrupted using a FastPrep-24 instrument (MP Biomedicals) for 4 × 1 min at 5.5 m/s. After phenol/chloroform extraction and ethanol precipitation, the RNA was resuspended in 500 µl ddH2O and the concentration was determined (usually 2–4 µg/µl). One hundred µg of total RNA were incubated with 30 U Turbo DNase (Ambion) for 1 h to remove residual DNA, subsequently precipitated and resuspended in 50 µl ddH2O. The yield was usually a concentration of 1–1.5 µg/µl, of which 1 µg was quality-checked on a 1% agarose gel.

# RT-qPCR

A total of 1 µg RNA was reverse-transcribed using the SuperScript IV Reverse Transcriptase (Thermo Fisher Scientific) and random hexamers according to the manufacturer's protocol. After transcription, the 20 µl reaction was filled up to 200 µl with ddH2O. mRNA levels were analyzed using 5 µl of 1:1 diluted cDNA product, the Fast SYBR Green PCR Master Mix (Applied Biosystems) and specific primers for the pepck (PEPCK\_fwd and PEPCK\_rev). The genes for hrdB (Primers: hrdB\_fwd and hrdB\_rev) and SCO1544 (Primers: 1544\_fwd and 1544\_rev) served as an endogenous control (Li et al., 2015). RT-qPCR was carried out on a StepOnePlus machine. All oligonucleotides used may be found in **Table 3**.

### β-Glucuronidase Measurement

The β-glucuronidase gene (gusA) was used to determine promoter activity in S. coelicolor (Myronovskyi et al., 2011). Here, 10–40 mg of protein crude extract (cell disruption protocol,



see section "SDS-PAGE and Western Blot Analysis") obtained from mycelium harvested from solid medium were used for the enzymatic reaction. The assay was performed as previously described (Rudolph et al., 2015).

### Measurement of PEP Concentration

The PEP Colorimetric/Fluorometric Assay Kit from Sigma-Aldrich was used to determine the PEP concentration in S. coelicolor according to the manufacturer's protocol. 2–4 mg of mycelium harvested from solid medium were used for the enzymatic reaction. Signals were quantified by fluorescence using a Tecan infinite 200 pro multiwell plate reader.

#### Northern Blot Analysis

A total of 30 µg total RNA was separated on 6% denaturing polyacrylamide gels and transferred to a positively charged nylon membrane (Hybond-N+, GE Healthcare) in a tank blotting device (Peqlab) at 4◦C. As probe for detection, 10 pmol oligonucleotide were radiolabeled at the 5<sup>0</sup> end using 5 µl γ- [ <sup>32</sup>P] ATP (∼3.3 pmol/µl, Hartmann-Analytic) and 1 µl T4 polynucleotide kinase (Roche) in the supplied buffer for 1 h at 37◦C and subsequently purified using Illustra MicroSpin G-25 columns (GE Healthcare). 25 µl radiolabeled oligonucleotides (approximately 300 kc.p.m./µl) where used as probe for each experiment. Signals were quantified by phosphoimaging using a FLA-5000 phosphoimager (Fujifilm life Science). Expression of scr5239 (probe: scr5239\_A) was normalized to the amount of 5S rRNA (probe: 5S\_A).

#### Bioinformatics RNAhybrid

RNAhybrid (Kruger and Rehmsmeier, 2006) is a tool for the prediction of microRNA/target duplexes. It predicts multiple potential binding sites of microRNAs in large target mRNAs by finding the energetically most favorable hybridization sites of a small RNA in a large RNA. To predict potential binding sites, we used RNAhybrid Version 2.2 with default parameters. Judging from past experience, the predicted mRNA/sRNA interaction sites are more reliable with this program than with other available prediction tools.

#### Target Explorer and PREDetector

For prediction of dre sites, an evaluation matrix ranking the individual nucleotides of the binding site according to their importance was required. For this purpose, the program Target Explorer was used (Sosinsky et al., 2003). To create the position matrix, the sequences of five different known dre sites were used. These can be found in **Supplementary Table S1**. The position matrix thus created is shown in **Supplementary Figure S1**. Using this matrix and PREDetector (Hiard et al., 2007), we predicted 15 new dre sites in the genome of S. coelicolor. The results, ranked according to the degree of matching with the matrix (score), can be found in **Supplementary Table S2**. The dre site upstream of scr5239 ranked fourth of all predicted sites.

# RESULTS

fmicb-10-03121 January 18, 2020 Time: 17:49 # 5

# Quantitative Proteomics Reveals 32 Proteins That Show scr5239-Dependent Changes in Expression

To identify new targets of scr5239, we compared the proteome of S. coelicolor wild type (M145) with its respective scr5239 deletion (1scr5239) and overexpression (scr5239+) strain. We applied quantitative proteomics based on the stable isotope labeling method to all three strains grown on R2YE solid medium and harvested in the transition phase. After protein extraction and fragmentation, the peptides where labeled with a different TMT labels for each strain and analyzed using LC-MS/MS (**Figure 1A**). This method has already been successfully applied in S. coelicolor to characterize differentiation and activation of the secondary metabolism (Rioseras et al., 2018).

A total of 1669 proteins were found to be expressed in all three strains (**Supplementary Table S3**). However, only 32 of them showed a significant change in expression level depending on scr5239, i.e., at least a two-fold change in mass spec. signal strength (1scr5239 vs. M145 or M145 vs. scr5239+) (**Figure 1B**). Functional analysis revealed that potential targets of the sRNA cluster in five main groups. The largest group, comprising about half of the identified annotated proteins (12), are metabolic enzymes. In addition to that, eight proteins cluster in the group of DNA-binding and regulatory proteins. Another three identified proteins are part of the CDA antibiotic biosynthesis cluster and two are involved in proteolysis. The remaining eight proteins are of unknown function. **Figure 1B** shows the detected changes in protein expression upon deletion or overexpression of the sRNA along with a functional analysis.

Since metabolic enzymes were the most abundant, we focused on potential targets within this group. Moreover, the fact that the previously identified targets of this sRNA, metE and dagA (Vockenhuber et al., 2011, 2015) play a role in central carbon and amino acid metabolism further encouraged us to prioritize metabolic enzymes. The candidate with the most interesting functionality here was SCO4979, which encodes the PEPCK. It connects glycolysis with the TCA cycle and catalyzes the ratelimiting step in the gluconeogenese. In consequence, it is of key importance for the central carbon metabolism. We therefore continued with this candidate.

#### PEPCK Protein and mRNA Levels Are Dependent on scr5239 Expression

To investigate the role of scr5239 in the regulation of PEPCK, we initially analyzed the mRNA and protein levels of PEPCK depending on the scr5239 level. To validate the quantitative proteomics data, we performed western blot analyses. We used an integrating plasmid and fused a 3x FLAG-tag to the C-terminus of PEPCK resulting in pPEPCK\_F3. We attached the tag to the C-terminus as sRNA/mRNA interactions usually take place in the 50UTR or close to the 5<sup>0</sup> end of the coding region of the mRNA. The plasmid was integrated into the genome of wild type (M145), 1scr5239 and scr5239+ into the attachment site of the phiC31 phage, thus resulting in strains FE01, FE02, and FE03, respectively. A list of all strains used and constructed across all experiments may be found in **Table 2**. We used the same growth conditions as in the TMT-based experiment and harvested the cells in the transition phase. The western blot analysis shown in **Figure 2A** effectively confirms the TMT data with a two-fold increase in PEPCK expression in FE02 (1scr5239) and approximately two-fold decrease in FE03 (scr5239+) compared to the wild type (FE01). To exclude an effect on the transcription of PEPCK, we replaced the wild type pepck promoter in pPEPCK\_F3 with the synthetic promoter SP4 (Horbal et al., 2018), thus generating the integrating plasmid pSP4\_PEPCK\_F3. Integration resulted in the respective strains FE04, FE05, and FE06 for M145, 1scr5239, and scr5239+, respectively. Western blot analyses of these strains show similar regulation, which indicates that the observed regulation is not promoter-dependent and therefore taking place on the posttranscriptional level (**Figure 2B**). mRNA levels of pepck were measured in M145, 1scr5239 and scr5239+. The RT-qPCR results revealed a four-fold increase of pepck-levels in 1scr5239 and a slight decrease (0.6-fold) in scr5239+ compared to the levels in M145 (**Figure 2C**).

Most known sRNAs base pair with the translational initiation region (TIR) and thus initiate mRNA degradation (Saramago et al., 2014). The results of the western blot and RTqPCR analyses indicate that the PEPCK regulation by scr5239 follows this mechanism.

#### Deletion of scr5239 Leads to an Increase of the Energy Donor Phosphoenolpyruvate in S. coelicolor

Next, we aimed to investigate the metabolic impact of PEPCK regulation through scr5239. We determined the steady state level of PEP, the product converted by PEPCK from OAA. M145, 1scr5239, and scr5239+ were grown and harvested under the same conditions as for the quantitative proteomics analysis. As shown in **Figure 3**, deletion of scr5239 dramatically modulates the cellular PEP level. In 1scr5239 we were able to detect an increase of 70% in PEP level compared to M145. Increasing the scr5239 level (scr5239+) did not significantly alter the PEP level (**Figure 3A**). **Figure 3B** illustrates the influence of the sRNA on PEPCK and consequently on the level of PEP.

Phosphoenolpyruvate is a key intermediate in carbohydrate metabolism and involved in glycolysis and gluconeogenesis. Its metabolization to pyruvic acid by pyruvate kinase generates ATP and is vitally important for the energy balance of the cell, as PEP has the highest-energy phosphate bond found in organisms. The influence the sRNA exerts on the cellular concentration of this key metabolite indicates its crucial role in the regulation of the primary metabolism.

#### Interaction of PEPCK and scr5239

To mechanistically characterize the regulation of PEPCK by scr5239, we analyzed the interaction between the sRNA and the target mRNA. We used RNAhybrid (Kruger and Rehmsmeier, 2006) to identify potential interaction sites. RNAhybrid predicted an 8 nt region within stem P4 of the sRNA to interact with the TIR

of the pepck mRNA (**Figure 4A**). In consequence, the interaction site with PEPCK predicted for scr5239 was in the same region of the sRNA as the binding sites of the other two targets dagA and metE.

To verify the predicted interaction site, we introduced mutations within the PEPCK sequence that disrupted potential base pairing. We used plasmid pPEPCK\_F3 and exchanged the PEPCK 50UTR including the ribosomal binding site (RBS) for an 50UTR of an unrelated gene (methionine synthase MetH), resulting in pPEPCK\_F3 M1. The methionine synthase MetH is not regulated by the sRNA (Vockenhuber et al., 2015). **Figure 4B** illustrates the relevant sequence for pepck mRNA. The construct pPEPCK\_F3 M1 was integrated into the genome of the three strains M145, 1scr5239, and scr5239+,

functional characteristics. Numbers on the x-axis refer to the SCO numbers of the expressed genes. The SCO number of PEPCK is highlighted in red.

deviation calculated from three independent experiments.

resulting in FE07, FE08, and FE09. **Figure 4B** shows that the expression levels of PEPCK are comparable for all three strains. PEPCK expression is not influenced by the scr5239 level as in the corresponding strains FE01, FE02, and FE03 that harbor a functional wild type PEPCK 50UTR (compare **Figure 2B**). Hence, the mutated PEPCK-F3 does not show dependency on the sRNA.

Furthermore, we mutated scr5239 and analyzed the influence on PEPCK expression. We introduced two mutations (U113A, G114C) into the predicted interaction site of the sRNA. U113 and G114 are central in the predicted binding motif and should prevent both interaction and, consequently, regulation. **Figure 4C** illustrates the relevant sequence of the mutated sRNA. We designed two variants of a plasmid carrying the PEPCK-F3 sequence and the sequence for scr5239. The former, pPEPC\_F3\_scr5239+, encoded for the wild type scr5239, the latter, pPEPC\_F3\_scr5239+\_M1, encoded for scr5239 with the described mutations. These constructs were integrated in 1scr5239 resulting in FE10 and FE11. The strains express the sRNA exclusively from the integrated plasmid (expression level of the scr5239 was confirmed by Northern blot, data not shown). Western blot analyses show that the PEPCK is more abundant in the strain with the mutated sRNA than in the strain with the wild type sRNA (**Figure 4C**), which indicates that the sRNA represses PEPCK expression more strongly than the mutated sRNA. These results suggest that the mutated sRNA is no longer able to bind pepck mRNA.

In sum, the results of the mutational analysis confirmed the predicted interaction site of scr5239 with the PEPCK mRNA.

#### Regulation of scr5239 by DasR

DasR is a global transcriptional regulator of the metabolism in S. coelicolor with more than 50 validated target genes. It is one of the most important pleiotropic regulators of the primary and secondary metabolism in this organism. It controls the N-acetylglucosamine metabolism, the monomer unit of the

FIGURE 3 | PEP levels in M145, 1scr5239, and scr5239+. (A) PEP levels were determined in S. coelicolor M145, 1scr5239, and scr5239+ and normalized to the levels in M145; error bars represent the standard deviation calculated from three independent experiments. (B) Schematic overview of the connection between scr5239, PEPCK, and PEP levels in S. coelicolor.

polymer chitin and the primary carbon and nitrogen source for S. coelicolor (Rigali et al., 2006; Colson et al., 2007). DasR binds to so called DasR responsive elements (dre sites) and prevents the transcription of downstream genes (Rigali et al., 2006; Colson et al., 2007; Rigali et al., 2008; Swi ˛atek-Połaty ´ nska ´ et al., 2015). After binding its ligand glucosamine-6-phosphate or

FIGURE 4 | Analysis of the interaction of scr5239 and PEPCK mRNA. (A) Predicted interaction site. The sequence of the sRNA is shown completely, that of the pepck mRNA only partially. The RBS (red) and the start codon (blue) of pepck are highlighted. The interaction site of dagA and metE is highlighted in red on the sRNA. (B) PEPCK was mutated at the predicted interaction site (PEPCmut-F3) and integrated in the genome of the M145, 1scr5239, and scr5239+. Mutated bases in the sequence are underlined. Western blot analysis using FE07, FE08, and FE09 and thus detecting PEPCmut-F3. (C) scr5239 was mutated at the predicted interaction site. Mutated bases in the sequence are underlined. The mutated and the wild type scr5239 in combination with PEPC-F3 were integrated into 1scr5239. Western blot analysis using FE10 and FE11 and thus detecting with PEPC-F3. Measurements were normalized to the respective strain expressing wild type scr5239; errors represent the standard deviation calculated from three independent experiments.

N-acetylglucosamine, respectively, DasR releases the DNA and transcription can take place. Using PREDetector software (Hiard et al., 2007), we identified a potential dre site upstream of scr5239. As the findings of this study clearly indicate a role of the scr5239 in the central metabolism, we decided to explore whether scr5239 is a direct target of DasR.

First, we analyzed whether scr5239 expression is dependent on DasR. We compared the expression of scr5239 in M145 with a strain lacking the dasR gene (BAP29; Rigali et al., 2006 and a strain overexpressing dasR (M145 [pFT241]; Schlicht, 2003). Northern blot analysis showed clear DasR-dependent changes in scr5239 expression (**Figure 5A**). Deletion or overexpression of DasR lead to significant increase and decrease of scr5239 expression, respectively. We then validated the significance of the dre site of scr5239 by reporter gene measurements. The scr5239 promoter along with the first 20 bases of the sRNA was fused to a gusA reporter gene (pGUS\_pscr5239). Two mutants were constructed, one with 6 nt of the dre site exchanged (pGUS\_pscr5239\_M1) and a second one that contained only the first 50 bp of the promoter with the dre site

deleted (pGUS\_pscr5239\_M2, **Figure 5B**, left). The constructs (pGUS\_pscr5239, \_M1, \_M2) were integrated into the genome of M145 resulting in the strains FE13, FE14, and FE15. The strains were grown to the transition phase and the GUS activity was measured. Both mutations lead to an approximately three-fold increase in 1-glucuronidase activity when compared to the wild type promoter (**Figure 5B**, right).

The Northern blot analysis and the reporter gene measurements of the promoter activity confirm that DasR regulates scr5239 transcription by interacting with the upstream dre site. However, please note that DasR, even when overexpressed, does not fully repress scr5239. Consequently, there are likely more factors that influence the expression rate of the sRNA in vivo.

# DISCUSSION

Starting from proteomics data, we confirmed pepck as a new target of scr5239 in S. coelicolor. Consequently, the sRNA has now three validated targets, which renders it the best characterized sRNA in streptomycetes to our knowledge.

The binding site of scr5239 in the pepck gene is located between the RBS and the start codon of the mRNA. In contrast, the binding sites of the other two known scr5239 targets are located within the open reading frame (dagA + 33 – + 52, metE + 7 – + 30). Furthermore, for dagA and metE, only the protein levels are changing while their mRNA levels stay constant. However, in the case of PEPCK, mRNA levels change in the same pattern as the protein levels, which is common for mRNA/sRNA interactions. The regulation described here thus follows the conventional mechanism described for the majority of sRNAs, mostly characterized in the well-studied organisms E. coli and Salmonella (Wagner and Romby, 2015). They regulate the expression of their target genes by blocking access to the RBS and/or start codon, which impedes translation and initiates mRNA degradation.

In the present study, we were able to show that scr5239 is involved in the regulation of the central metabolism of S. coelicolor. PEPCK is a key enzyme of the primary metabolism that catalyzes the conversion of OAA to PEP, the first and rate-limiting step of the gluconeogenesis (Delbaere et al., 2004). Since the regulation of glycolysis and gluconeogenesis is mutually exclusive, the control of PEPCK must be exactly adapted to the metabolic demand/situation of the cell. We show that scr5239 is involved in this regulation. It binds to the pepck mRNA leading to a reduced level of the enzyme. Consequently, a reduced level of the sRNA significantly increases the intercellular PEP level (see deletion studies). PEP can then be further converted into fructose-1,6-bisphosphate toward glucose but is also involved in carbon uptake and signaling through the phosphotransferase system (PTS) (Bruckner and Titgemeyer, 2002).

Most sRNAs are transcriptionally controlled by σ- or transcription factors. We were able to show that this is also the case for scr5239, which is regulated by the global transcription regulator DasR. We identified a functional dre site in the promoter of scr5239. Our findings match the results of previously performed ChIP-on-chip experiments and EMSAs (Swi ˛atek-Połaty ´ nska et al., 2015 ´ ). They showed that the dre site of scr5239 was one of the most tightly bound targets of DasR. Moreover, DasR binding to the dre site of scr5239 could be observed in the samples taken during vegetative growth and sporulation, which was only the case for a small fraction of the found targets. Furthermore, the genomic locations of DasR and scr5239 provide further insight. DasR is encoded in ORF5231, which puts protein and sRNA in a close evolutionary context. These findings indicate that scr5239 is at least one of the most preferred targets of DasR, thus suggesting a key role of the sRNA in the DasR regulatory network.

Small non-coding RNAs are often involved in forming (negative) feedback loops with the transcription factors they regulate (reviewed in Wagner and Romby, 2015). A feedback loop regulation can also be described for the scr5239 regulation by DasR. DasR is regulated by the PTS in streptomycetes with GlcNAc as preferred carbon source (Nothaft et al., 2003). If the supply of nutrients is sufficient (high level of GlcNAc that is converted via GlcN6-P into Fru-6P as input to glycolysis), DasR in its GlcN6-P-bound conformation cannot act as repressor, hence the sc5239 level is high. As a result, gluconeogenesis is switched off by binding and repressing pepck via scr5239,

the first and rate-limiting step of gluconeogenesis. If the nutrient supply deteriorates, DasR in its free form can act as repressor and the level of scr5239 decreases. In consequence, the PEP level increases, which initiates gluconeogenesis. However, PEP is not just important as intermediate for glycolysis and gluconeogenesis, but although for carbon uptake via the PTS. Increased PEP levels allow the cells to take up more GlcNAc through the PTS, which leads to a higher concentration of GlcN-6-P. GlcN-6-P in turn represses DasR, which then again leads to higher levels of scr5239. Through this feedback loop scr5239 can to influence its own repressor. A model of such feedback loop regulation including the DasR and scr5239 is depicted in **Figure 6**. Such regulatory feedback loops have already been described for sRNAs. An example is the transcription factor OmpR that activates the sRNAs OmpA and OmpB whereas these inhibit OmpR (Brosse et al., 2016).

The carbohydrate metabolism is often governed by complex regulatory networks. They allow the survival of bacteria in ever-changing habitats with fluctuating nutrient supplies by rapid adaptation of their metabolic capabilities. Such metabolic networks not only exploit transcription regulation but also include posttranscriptional regulation e.g., the control by sRNAs. Within the last years, sRNAs emerged as an important additional layer to fine-tune regulatory response. sRNAs are proposed to limit the response to substrate availability, set the threshold concentration or module the delay time for activation or shutdown of the system (Durica-Mitic et al., 2018). Prominent examples are the sRNA Spot42 that is involved in the carbon catabolite repression in E. coli or the sRNA CsrB and CsrC that control the RNA-binding protein CsrA – a posttranscriptional regulator responsible for the switch between glycolysis and gluconeogenesis. These examples impressively show that bacterial carbohydrate metabolism is controlled at all levels by large and densely interconnected regulatory networks. In Streptomyces, DasR is the major hub for the control of the central carbohydrate metabolism. We identified scr5239 as additional player in this regulatory network representing a switch for the cell to decide between gluconeogenese and glycolysis.

#### REFERENCES


# DATA AVAILABILITY STATEMENT

The datasets generated for this study can be found in the PRIDE repository project accession: PXD016811.

# AUTHOR CONTRIBUTIONS

BS, M-PV, and FE contributed to the conception, design of the study, and wrote the manuscript. FE and M-PV performed most of the experiments. EO and P-JJ were responsible for quantitative proteomics analysis.

# FUNDING

This work was funded by the Deutsche Forschungsgemeinschaft CRC902(A2), the Swedish Research Council (grant no. 2017- 02577), and Innovative Medicines Initiative (EU/EFPIA, ULTRA-DD, grant no. 115766).

# ACKNOWLEDGMENTS

The authors thank Dr. J. E. Weigand for fruitful discussions and for critically reading the manuscript.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2019.03121/full#supplementary-material

FIGURE S1 | dre site prediction matrix.

TABLE S1 | dre sites that were used to create the position martix.

TABLE S2 | Predicted dre sites.


repressing synthesis of Lpp, the major outer membrane lipoprotein. Genes Dev. 28, 1620–1634. doi: 10.1101/gad.243485.114


analyses of Streptomyces coelicolor reveal proteins and phosphoproteins modulating differentiation and secondary metabolism. Mol. Cell. Proteomics 17, 1591–1611. doi: 10.1074/mcp.RA117.000515


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2020 Engel, Ossipova, Jakobsson, Vockenhuber and Suess. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Stimulated Biosynthesis of an C10-Deoxy Heptaene NPP B2 *via* Regulatory Genes Overexpression in *Pseudonocardia autotrophica*

*Heung-Soon Park, Hye-Jin Kim, Chi-Young Han, Hee-Ju Nah, Si-Sun Choi and Eung-Soo Kim\**

*Department of Biological Engineering, Inha University, Incheon, South Korea*

Polyene macrolides, such as nystatin A1, amphotericin B, and NPP A1, belong to a large family of valuable antifungal polyketide compounds that are typically produced by soil actinomycetes. Previously, NPP B1, a novel NPP A1 derivative harboring a heptaene core structure, was generated by introducing two amino acid substitutions in the putative NADPH-binding motif of the enoyl reductase domain in module 5 of the NPP A1 polyketide synthase in *Pseudonocardia autotrophica*. This derivative showed superior antifungal activity to NPP A1. In this study, another novel derivative called NPP B2 was developed, which lacks a hydroxyl group at the C10 position by site-specific gene disruption of the P450 hydroxylase NppL. To stimulate the extremely low expression of the NPP B2 biosynthetic pathway genes, the 32-kb NPP-specific regulatory gene cluster was overexpressed *via* site-specific chromosomal integration. The extra copy of the six NPP-specific regulatory genes led to a significant increase in the NPP B2 yield from 0.19 to 7.67 mg/L, which is the highest level of NPP B2 production ever achieved by the *P. autotrophica* strain. Subsequent *in vitro* antifungal activity and toxicity studies indicated that NPP B2 exhibited similar antifungal activity but significantly lower hemolytic toxicity than NPP B1. These results suggest that an NPP biosynthetic pathway refactoring and overexpression of its pathway-specific regulatory genes is an efficient approach to stimulating the production of an extremely low-level metabolite, such as NPP B2 in a pathway-engineered rare actinomycete strain.

Keywords: *Pseudonocardia autotrophica*, polyene, regulatory genes, hemolytic toxicity, antifungal activity

#### INTRODUCTION

More than 45% of the bioactive compounds discovered from microbial secondary metabolites are derived from actinomycetes (Bu et al., 2019). A significant number of these secondary metabolites produced by actinomycetes were utilized further as the lead compounds in the field of medicine as clinically important anticancer, antibiotic, anti-inflammatory, antiviral, antiparasitic, and antioxidant drugs (Bérdy, 2005; Subramani and Aalbersberg, 2012; Manivasagan et al., 2014; Abdelmohsen et al., 2015). Among them are polyene macrolide antibiotics, such as nystatin,

*Edited by:* 

*Yinhua Lu, Shanghai Normal University, China*

#### *Reviewed by:*

*Patrick Caffrey, University College Dublin, Ireland Haoxin Wang, Shandong University, China*

> *\*Correspondence: Eung-Soo Kim eungsoo@inha.ac.kr*

#### *Specialty section:*

*This article was submitted to Microbial Physiology and Metabolism, a section of the journal Frontiers in Microbiology*

> *Received: 30 October 2019 Accepted: 07 January 2020 Published: 24 January 2020*

#### *Citation:*

*Park H-S, Kim H-J, Han C-Y, Nah H-J, Choi S-S and Kim E-S (2020) Stimulated Biosynthesis of an C10-Deoxy Heptaene NPP B2 via Regulatory Genes Overexpression in Pseudonocardia autotrophica. Front. Microbiol. 11:19. doi: 10.3389/fmicb.2020.00019*

**86**

amphotericin, candicidin, and pimaricin, which are potent antifungal compounds that are comprised typically of a polyketide core macrolactone ring with about 20–40 carbon atoms, including 3–8 conjugated double bonds (**Figure 1**; Caffrey et al., 2016). The major antifungal mechanism of these polyene antibiotics is believed to be the formation of ion channels *via* fungal ergosterol binding that mediates the leakage of cellular K+ and Mg2+, which leads to the death of fungal cells (Bolard, 1986; Neumann et al., 2016).

Over the last several decades, complete polyene biosynthetic gene clusters from nystatin, amphotericin, pimaricin, and candicidin have been isolated and characterized (Aparicio et al., 1999, 2003; Zotchev et al., 2000; Caffrey et al., 2001). Polyene compounds are biosynthesized typically by a giant enzyme complex called polyketide synthase (PKS), followed by further modification of the core macrolide ring by post-PKS modification enzymes, including P450 hydroxylases and glycosyltransferases (Kim et al., 2015). Subsequent attempts to generate a range of derivatives of these antifungal polyene macrolides were pursued to enhance the antifungal activity and/or reduce their intrinsic toxicities through both semisynthetic and biosynthetic pathway refactoring approaches (Byrne et al., 2003; Nedal et al., 2007; Silveira and Husain, 2007; Caffrey et al., 2016). A heptaene version of nystatin A1 called S44HP displayed considerably higher antifungal activity than the original tetraene nystatin A1, which is comparable to that of amphotericin B (Bruheim et al., 2004). Mannosyl-8-deoxyamphotericins produced by inactivation of the *amphL* gene and introduction of the pIJ02567-*nypY* construct in *Streptomyces*  *nodosus* exhibited lower hemolytic toxicity (Walmsley et al., 2017). In addition, highly water-soluble amphotericin B analogues were produced by the addition of extra sugar residues to amphotericin B (Caffrey et al., 2016). Recently, the discovery of C35deOAmB revealed that the antifungal activity was critical for binding to ergosterol, not for ion-channel formation (Gray et al., 2012), which generated toxicity-reduced amphotericin B derivatives (Cioffi et al., 2015; Muraglia et al., 2019). Overall, the engineering of polyene PKS to change the number of conjugated double bonds as well as post-PKS modifications for manipulation of the sugar and hydroxyl moieties is considered an important strategy for the generation of pharmacokinetically improved novel polyene derivatives (Bruheim et al., 2004; Gray et al., 2012; Caffrey et al., 2016; Walmsley et al., 2017; Muraglia et al., 2019).

NPP A1 is a disaccharide-containing polyene antifungal compound produced by a rare actinomycete, *Pseudonocardia autotrophica*, with an identical core macrolactone structure to nystatin A1 except for an additional *N*-acetyl-glucosamine (**Figure 1**; Lee et al., 2012). In the NPP A1 biosynthesis process, the macrolide backbone is synthesized by six type I PKSs, and the unique di-sugar moiety, mycosaminyl-(α1-4)-*N*-acetylglucosamine, is then attached by two glycosyltransferasesencoding *nppDI* and *nppY* and finally hydroxylated by P450 hydroxylase-encoding *nppL* (Kim et al., 2015). In terms of the biological activities, this polyene has superior water solubility and reduced hemolytic toxicity but shows slightly lower antifungal activity than nystatin A1 (Lee et al., 2012). Elucidation of the post-PKS modification steps of NPP A1 biosynthesis in

*P. autotrophica* highlighted the potential to develop novel NPP A1 derivatives (Kim et al., 2015). Several studies were conducted to develop NPP A1 derivatives to enhance the antifungal activity, resulting in two analogues, NPP A2 and NPP B1. NPP A2 lacks a C10 hydroxyl group that was generated by the inactivation of the P450 hydroxylase-encoding *nppL* gene in *P. autotrophica* (Kim et al., 2016). NPP B1 was generated by changing its core macrolactone structure from tetraene to heptaene through site-specific, two amino acid substitutions in the putative NADPH-binding motif of the enoyl reductase domain in module 5 of the NPP A1 polyketide synthase NppC (**Figure 1**; Kim et al., 2018). Interestingly, both NPP A2 and NPP B1 exhibited higher antifungal activities than those of NPP A1.

In this study, another NPP A1 analogue called NPP B2 was developed, which was generated by the inactivation of the polyketide enoyl reductase (ER) domain in the fifth module as well as the P450 hydroxylase-encoding *nppL* gene in the NPP A1 producing *P. autotrophica* strain (**Figure 1**). To overcome the extremely low level of NPP B2 production because of its pathway refactoring, the 32-kb NPP-specific regulatory gene cluster was overexpressed in the NPP B2 production strain, followed by isolation and characterization of the biological activities of NPP B2.

#### MATERIALS AND METHODS

#### Strains and Growth Conditions

*P. autotrophica* KCTC9441 purchased from the Korean Collection for Type Cultures was used for NPP production. The strain was grown routinely in ISP2 agar (malt extract 10 g, yeast extract 4 g, glucose 4 g, and agar 20 g/L) at 28°C for the sporulation and seed culture. YEME medium (yeast extract 3 g, peptone 5 g, malt extract 3 g, glucose 10 g, sucrose 340 g/L, and 5 mM MgCl2) was used to produce the NPP derivatives (Lee et al., 2012). *Candida albicans* ATCC 14053 was grown on YM medium (dextrose 10 g, peptone 5 g, yeast extract 3 g, Malt extract 3 g, and agar 20 g/L) at 30°C for 24 h. All *Escherichia coli* strains were incubated at 37°C in Luria-Bertani medium supplemented with the appropriate antibiotics where needed.

#### Inactivation of NppL and Overexpression of NPP-Specific Regulatory Gene Cluster

A *nppL* gene inactivation cassette, including the upstream and downstream regions, was constructed by PCR amplification using the following primer pairs: upstream region, DELL\_1F (5′-GAATTCCGTCCTGTACTCGTCGGT-3′) and DELL\_1R (5′-CTGCAGTCA TGACGCGTCCTCCGT-3′) and downstream region, DELL\_2F (5′-CTGCAGAC GCGGTCACGATGGCGC-3′) and DELL\_2R (5′-AAGCTTACCTGGCCGAGCAGATGG-3′) (Kim et al., 2015). The amplified fragments were digested with *Hind*III-*EcoR*I and ligated into pKC1132. The recombinants were selected on LB medium containing apramycin. The *nppL* gene-inactivation plasmid was introduced into the chromosome of the NPP B1 production strain, *P. autotrophica* ER5 mutant.

To overexpress the NPP-specific regulatory genes, the previously constructed pNPPREG, encompassing the 32-kb NPP-specific regulatory gene cluster, was integrated into the chromosome of the B2 production strain (Han et al., 2019). The recombination conjugants were selected on ISP2 medium containing apramycin.

#### Production and Purification of NPP Derivatives

NPP or its derivative production strains were inoculated in 300 ml of ISP2 medium containing the appropriate antibiotics at 30°C and 220 rpm for 72 h. The pre-cultures were added to 3 L YEME medium in a 5 L bioreactor for batch fermentation. After 48 h of cultivation, 150 g of Amberlite XAD16 resin (Sigma-Aldrich, USA) was added to the culture broth. After 24 h of resin addition, the mycelia and resin from the culture broth were separated and then extracted twice in 600 ml of *n*-butanol. The extract was concentrated using a vacuum evaporator, after which the concentrated extract was dissolved in methanol and loaded onto a column packed with a C18 reversed-phase silica gel (Daiso, Japan) along with methanolwater (30:70, v/v) to remove any residual sugar from the production media. The extracts with sugar removed were purified using a fraction collector (Interchim, France) on a gradient consisted of solvents A (water) and B (methanol): 30% B (v/v) (0–10 min) and 100% B (v/v) (100 min) at a flow rate of 20 ml/min. The fractions containing NPP or its derivatives with >80% purity were detected at 405 nm and analyzed by high performance liquid chromatography (HPLC). The column was equilibrated with 50% solvent A (0.05 M ammonium acetate, pH 6.5) and 50% solvent B (methanol); the flow rate was set to 1.0 ml/min using the following conditions: 0–3 min, 50–75% B; 3–30 min, 75–100% B; 30–33 min, 100–50% B; and 33–40 min, 50% B (Won et al., 2017).

#### Liquid Chromatography-Mass Spectrometry/Mass Spectrometry Analysis

NPP B2 showing >80% purity was analyzed using A Triple TOF 5600 + (AB Sciex, USA) coupled with Ultimate3000 (Thermo Scientific, USA). Mass spectrometry was operated in both positive and negative ion modes over a mass range from 50 to 2,000 *m*/*z* using an electrospray ionization source. The settings were nitrogen gas for nebulization at 50 psi, heater gas pressure at 50 psi, curtain gas at 25 psi, temperature of 500°C, and an ion spray voltage at 5,500 V in positive ion mode and −4,500 V in negative ion mode. The optimized declustering potential (DP) and collision energy (CE) were set to 60 and 10 eV in positive ion mode, and to −60 and −10 eV in negative ion mode, respectively. A sweeping collision energy setting at 35/−35 ± 15 eV was applied for collisioninduced dissociation (CID). Chromatographic conditions: solution A (0.1% formic acid in distilled water) and solution B (0.1% formic acid in acetonitrile) were used for elution and loaded onto Phenomenex Kinetex 1.7 μ C18 (2.1 mm × 150 mm, 1.7 μm). The flow rate was set to 0.4 ml/min using the following conditions: 0–1 min, 90% A; 1–5 min, 90–50% A; 5–18 min, 50–0% A; 18–25 min, 0% A; 25–27 min, 0–90% A; and 27–30 min, 90% A.

#### RNA Analysis by Quantitative Real-Time Polymerase Chain Reaction

RNA was prepared using the RNeasy Mini Kit (Qiagen, Germany). cDNA conversion was carried out using a PrimeScript 1ST strand cDNA Synthesis Kit (TaKaRa, Japan) according to the manufacturer's instructions. Real-time PCR was performed using TaKaRa SYBR Premix Ex Taq (Perfect Real Time) with a Thermal Cycler Dice Real Time System Single (code TP850; TaKaRa, Japan). **Supplementary Table S1** lists the primer pairs. The PCR conditions included activation for 10 min at 95°C, followed by 35 cycles of 30 s at 95°C, 30 s at 58°C, and 30 s at 72°C. The data were collected during each 72°C step, and melting curve analysis was performed at default settings ranging from 60 to 95°C. The relative level of amplified mRNA was normalized to the mRNA expression level of the housekeeping gene, *P. autotrophica hrdB*, which was amplified as an internal control using the primer pairs *hrdB*\_F (5′-GCGGTGGAGA AGTTCGACTA-3′) and *hrdB*\_R (5′-TTGATGACCTCGACCATGTG-3′) (Han et al., 2019).

#### *In vitro* Assays for Biological Activities

For *in vitro* antifungal assay, we adapted to the Clinical and Laboratory Standards Institute document M27-A3 (Wayne, 2008). After *C. albicans* was cultured in YM medium at 30°C for 24 h, the cultured solution was diluted with YM medium until the OD value is 0.3 at 530 nm. A working suspension was made by a 1:2,000 dilution with RPMI-1640 broth media (with glutamine and phenol red, without bicarbonate, Sigma-Aldrich, USA), which resulted in 5.0 × 102 to 2.5 × 103 cells per μl. Ten microliters of the DMSO containing polyene antibiotics at various concentrations (3.125–1600 μg/ml) were added to the working suspension of 990 μl, and then, the mixtures were incubated at 30°C without shaking for 48 h. The colorimetric change of the mixture from red to yellow indicated the growth of *C. albicans.* The minimum inhibitory concentration (MIC) values were determined by measuring the minimum concentration that changed color to yellow. The experiment was performed in the duplicate.

For *in vitro* hemolysis assay, we adapted from a previously reported method (Nedal et al., 2007). Briefly, defibrinated horse blood was purchased from Kisan Biotech (South Korea). The polyene compounds were then prepared to the following concentrations with DMSO: 1–200 μg/ml. A 50 μl of each polyene solution was added to 450 μl of 2.5% defibrinated horse blood buffered with RBC buffer (10 mM NaH2PO4, 150 mM NaCl, 1 mM MgCl2, and pH 7.4), which resulted in a 1:10 dilution of each concentration of polyene. The samples were then incubated at 37°C for 30 min. After incubation, the samples were centrifuged at 10,000 ×*g* for 2 min. Next, 100 μl of the supernatant from each sample was added to a 96-well plate, after which the absorbance was read at 540 nm using a microplate reader (TECAN, Switzerland). The percentage hemolysis of each sample was defined as (Abssample − Absnegative/Abspositive − Absnegative) × 100 (%) to calculate the minimum hemolysis concentration (MHC). The positive hemolysis sample was prepared by adding defibrinated horse blood 11.25 μl and DMSO 50 μl to distilled water 438.75 μl, and the negative hemolysis sample was prepared by adding defibrinated horse blood 11.25 μl and DMSO 50 μl to RBC buffer 438.75 μl.

#### RESULTS

#### Construction of the NPP B2 Production Strain

An *in vitro* antifungal assay confirmed that the NPP B1, a heptaene version of NPP A1, showed much higher antifungal activity than the tetraene NPP A1 (Kim et al., 2018). Therefore, we tried to generate pharmacokinetically improved heptaene NPP B1 derivative through the engineering of post-PKS modification. Based on the previous results that characterization of the P450 hydroxylases, which were a region-specific hydroxylation such as AmphL (amphotericin), NysL (nystatin), and NppL (NPP) (Byrne et al., 2003; Volokhan et al., 2006; Kim et al., 2016), the NPP B1 production strain was engineered by inactivation of the *nppL* gene, which was involved in NPP C10 region-specific hydroxylation (**Figures 2A,B; Supplementary Figure S1**). Inactivation of the *nppL* gene was performed successfully by homologous recombination using the pKC1132 *Streptomyces* suicide vector system, and the mutation was verified genetically by PCR product sequencing analysis (**Figure 2C; Supplementary Figure S2**). LC-MS analysis of purified NPP B2 contained a signal at *m*/*z* 1111.5807 for [C53H85N2O22] + (calculated mass of NPP B2 is 1110.57) confirming that NPP B2 had been produced, as expected in the constructed mutant strain (**Supplementary Figure S3**).

#### Stimulation of NPP B2 Production Through Overexpression of the NPP Pathway-Specific Regulatory Genes

Although the NPP B2 production strain was developed successfully by inactivation of the ER domain of module 5 followed by the P450 hydroxylase-encoding *nppL* gene from the NPP A1 production strain, the NPP B2 production level was reduced approximately 40-fold to 0.2 mg/L compared to the NPP A1 production level in the *P. autotrophica* wild type. Among the attempts to increase the production of NPP derivatives, a strategy was conducted to increase the production level of NPP B2 through overexpression of the NPP pathwayspecific regulatory genes. In previous studies, pNPPREG was constructed by cloning a 32-kb right-hand portion of BGC containing the six NPP-specific regulatory genes (*nppRI*–*nppRVI*) into a *Streptomyces* artificial chromosomal vector pSBAC (Han et al., 2019). To stimulate NPP B2 production, pNPPREG containing the entire regulatory genes was integrated into the chromosome of the NPP B2 production strain (**Figure 3A**). As a result, the level of NPP B2 production was increased significantly (approximately 39-fold to 7.67 mg/L) (**Figure 3B**).

from the NPP B2 production strain at 24 h; closed circle, transcripts from the NPP B2 production strain at 48 h; open triangle, transcripts from the NPP B2 over-production mutant at 24 h; closed triangle, transcripts from the NPP B2 over-production mutant at 48 h; house-keeping gene, *hrdB* (light purple); PKS genes, *nppA* (red), *nppC* (yellow), and *nppI* (emerald green); regulatory genes related to NPP B2 biosynthesis, *nppRI* (sky blue), *nppRII* (blue), *nppRIII* (dark blue), *nppRIV* (brown), *nppRV* (dark purple), and *nppRVI* (light green). All transcript measurements were performed in duplicate.

The transcription levels of the NPP B2 biosynthetic genes were analyzed to further understand the molecular basis responsible for enhanced NPP B2 production in the engineered NPP B2 over-production strain. qRT-PCR analysis revealed increased transcription of the PKS genes, such as *nppA* and *nppC*, in the NPP B2 over-production strain compared to the parental strain (**Figure 3C**). As expected, the transcription levels of all five NPP-pathway specific regulatory genes, except *nppRII,* were also stimulated (**Figure 3C**), suggesting that the enhanced transcription level of the refactored NPP B2 biosynthetic pathway genes is critical for the titer improvement.

#### *In vitro* Antifungal Activity and Hemolytic Toxicity

Based on successful NPP B2 yield improvement, the purified NPP B2 was evaluated for both *in vitro* antifungal activity and the hemolytic toxicity. Minimum inhibitory concentration (MIC) evaluation assays of *in vitro* antifungal activity using the colorimetric change in the RPMI-1640 media containing *C. albicans* were employed (Wayne, 2008). The MIC value of NPP B2 against *C. albicans* (1.0 μg/ml) was slightly higher than amphotericin B (0.25 μg/ml) and NPP B1 (0.5 μg/ml) (**Table 1; Supplementary Figure S4**). Interestingly, however, *in vitro* hemolytic toxicity evaluation, NPP B2 exhibited much reduced hemolytic toxicity than its parental NPP B1. The minimum hemolytic concentration (MHC) value for NPP B2 was measured as 80.18 μg/ml, whereas the MHC value of amphotericin B was 3.26 μg/ml under the conditions tested, indicating that the *in vitro* toxicity of NPP B2 was approximately 25-fold lower than that of amphotericin B (**Table 1; Supplementary Figure S5**). Moreover, the *in vitro* toxicity was 4.5-fold lower than the MHC value of NPP B1 (17.72 μg/ml). These results suggest that the absence of the hydroxyl moiety at the C10 position of NPP B2 could play an important role in controlling both antifungal acidity and hemolytic toxicity.

#### DISCUSSION

The genome mining approach of rare actinomycetes followed by the activation of its cryptic biosynthetic gene cluster (BGC) has



*a MIC, minimum inhibitory concentration (values resulting in no visible growth of C. albicans).*

*b MHC, minimum hemolytic concentration (values causing 90% hemolysis against horse blood cells ± percentage standard deviation).*

become an attractive strategy to screen and develop novel bioactive compounds (Choi et al., 2018). The *P. autotrophica* strain described here was classified originally as a polyene non-production strain, which was later proven to be a producer of a novel di-saccharidecontaining NPP A1 through culture optimization and whole genome sequencing (Lee et al., 2012). Although an NPP A1 BGC refactoring strategy generated a C10-deoxy NPP A1 (named NPP A2) and a heptaene version of NPP A1 (named NPP B1), their extremely low titers were major hurdles for further characterization of their biological activities.

Previously, several strategies, including overexpression of the pathway-specific regulatory gene, deletion of the global antibiotic downregulator, *in situ* screening of random mutants, co-culture system, and cultivation with xenobiotics, all failed to improve NPP B1 production (Kim et al., 2018). Recently, the entire cluster containing all six NPP pathway-specific genes in a pSBAC system followed by its re-integration into the *P. autotrophica* chromosome led to a significant increase in the NPP B1 titer (Han et al., 2019).

In this study, another novel derivative, called the NPP B2 production strain, was first generated by site-specific inactivation of the *nppL* gene in the NPP B1 production of the *P*. *autotrophica* mutant strain. C10-deoxy NPP B1 (NPP B2) was also generated based on a previous report that the C8-deoxy amphotericin produced by the inactivation of *amphL* in *S. nodosus* and C10-deoxy nystatin by the inactivation of *nysL* in *Streptomyces noursei* (Byrne et al., 2003; Volokhan et al., 2006). Owing to the extremely low titer of NPP B2, however, its biological characterization could not be pursued without strain improvement. Through the chromosomal integration of all six NPP pathwayspecific genes in the NPP B2 production strain, its titer was improved significantly (approximately 39-fold), which was sufficient to proceed for further biological assays, including antifungal and hemolytic toxicity assays.

This paper described for the first time the *in vitro* biological activities of NPP B2, a novel heptaene version of the NPP derivative. Although the *in vitro* antifungal activity of NPP B2 was higher than that of tetraene-type polyenes, such as nystatin A1, NPP A1, and A2, it showed slightly lower antifungal activity than other heptaene-type polyenes, including amphotericin B and NPP B1. Interestingly, the *in vitro* hemolytic activity of NPP B2 was approximately 25 times lower than those of amphotericin B and NPP B1, suggesting that the hydroxyl moiety at the C10 position of NPP could play a critical role in controlling both the antifungal activity and the hemolytic toxicity, probably by affecting its binding affinity to cholesterol and ergosterol. Overall, these results suggest that the combination of rational BGC refactoring and its genetic strain improvement approach is an efficient strategy to stimulate the production of an extremely low-level metabolite, such as NPP B2 in a rare actinomycetes strain.

#### DATA AVAILABILITY STATEMENT

The datasets generated for this study can be found in the EU108007.

#### AUTHOR CONTRIBUTIONS

H-SP, S-SC, and E-SK designed the experiments. H-SP, H-JK, and C-YH performed the experiments. H-SP, H-JN, and E-SK wrote the manuscript.

#### FUNDING

This study was supported by Agricultural Microbiome R&D Program, Ministry of Agriculture, Food and Rural Affairs,

#### REFERENCES


Republic of Korea [as part of the (multi-ministerial) Genome Technology to Business Translation Program] No. 918008-04. This work was also funded by National Research Foundation of Korea (NRF) (Project No. NRF-2017R1A2A2A05069859).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb.2020.00019/ full#supplementary-material


antibiotic biosynthesis in the nystatin producer *Streptomyces noursei* ATCC 11455. *Microbiology* 146, 611–619. doi: 10.1099/00221287-146-3-611

**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Copyright © 2020 Park, Kim, Han, Nah, Choi and Kim. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Disclosing the Potential of the SARP-Type Regulator PapR2 for the Activation of Antibiotic Gene Clusters in Streptomycetes

#### Janina Krause<sup>1</sup> , Ira Handayani1,2, Kai Blin<sup>3</sup> , Andreas Kulik<sup>1</sup> and Yvonne Mast1,4,5,6 \*

<sup>1</sup> Department of Microbiology/Biotechnology, Interfaculty Institute of Microbiology and Infection Medicine, Faculty of Science, University of Tübingen, Tübingen, Germany, <sup>2</sup> Research Center for Biotechnology, Indonesian Institute of Sciences (LIPI), Cibinong, Indonesia, <sup>3</sup> Novo Nordisk Foundation Center for Biosustainability, Technical University of Denmark, Lyngby, Denmark, <sup>4</sup> German Center for Infection Research (DZIF), Partner Site Tübingen, Tübingen, Germany, <sup>5</sup> Department of Bioresources for Bioeconomy and Health Research, Leibniz Institute DSMZ – German Collection of Microorganisms and Cell Cultures, Braunschweig, Germany, <sup>6</sup> Department of Microbiology, Technical University Braunschweig, Braunschweig, Germany

#### Edited by:

Rustam Aminov, University of Aberdeen, United Kingdom

#### Reviewed by:

Marie-Joelle Virolle, Centre National de la Recherche Scientifique (CNRS), France Jean-Luc Pernodet, Centre National de la Recherche Scientifique (CNRS), France Javier Santos Aberturas, John Innes Centre (JIC), United Kingdom Juan F. Martin, Universidad de León, Spain

> \*Correspondence: Yvonne Mast yvonne.mast@dsmz.de

#### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 29 October 2019 Accepted: 30 January 2020 Published: 18 February 2020

#### Citation:

Krause J, Handayani I, Blin K, Kulik A and Mast Y (2020) Disclosing the Potential of the SARP-Type Regulator PapR2 for the Activation of Antibiotic Gene Clusters in Streptomycetes. Front. Microbiol. 11:225. doi: 10.3389/fmicb.2020.00225 Streptomyces antibiotic regulatory protein (SARP) family regulators are well-known activators of antibiotic biosynthesis in streptomycetes. The respective genes occur in various types of antibiotic gene clusters encoding, e.g., for polyketides, ribosomally and non-ribosomally synthesized peptides, or β-lactam antibiotics. We found that overexpression of the SARP-type regulator gene papR2 from Streptomyces pristinaespiralis in Streptomyces lividans leads to the activation of the silent undecylprodigiosin (Red) gene cluster. The activation happens upon the inducing function of PapR2, which takes over the regulatory role of RedD, the latter of which is the intrinsic SARP regulator of Red biosynthesis in S. lividans. Due to the broad abundance of SARP genes in different antibiotic gene clusters of various actinomycetes and the uniform activating principle of the encoded regulators, we suggest that this type of regulator is especially well suited to be used as an initiator of antibiotic biosynthesis in actinomycetes. Here, we report on a SARP-guided strategy to activate antibiotic gene clusters. As a proof of principle, we present the PapR2-driven activation of the amicetin/plicacetin gene cluster in the novel Indonesian strain isolate Streptomyces sp. SHP22-7.

Keywords: actinomycetes, Streptomyces, antibiotic, regulator, SARP, silent gene cluster

#### INTRODUCTION

In 2018, the WHO warned that the dramatic increase of antibiotic resistances coupled with the scarcity of new antibiotics will lead to a global health crisis in the 21st century (World Health Organization, 2015). Even nowadays, infections that are caused by drug-resistant pathogens are suggested to account for 700,000 deaths worldwide annually (United Nations Foundation and the Wellcome Charitable Trust, 2016). According to the World Bank Group "by 2050, drug-resistant infections could cause global economic damage on par with the 2008 financial crisis" (Worldbank, 2016). Thus, there is a substantial need for new antibiotics in order to combat drug-resistances. Bacteria have long been recognized as a prolific source for antibiotics

(Newman and Cragg, 2016). Especially actinomycetes are potent producers of bioactive molecules, as they provide up to 70% of all medically important antibiotic agents (Tanaka and Omura, 1990). The capability to produce these natural compounds is genetically encoded in the actinomycetes genome, whereby the respective genes usually are organized as biosynthetic gene clusters (BGCs). In recent years, genomic analyses of actinomycetes have revealed the presence of numerous "silent" or "cryptic" BGCs, meaning that these clusters remain silent or are only weakly expressed under standard lab conditions. Indeed it is estimated that actinomycetes encode ∼10-times the number of secondary metabolites than anticipated from prior fermentation studies (Baltz, 2017). Thus, these microorganisms still hold the genetic potential to produce new bioactive compounds. Consequently, there are several attempts to activate silent gene cluster expression in order to find new antibiotics. However, most of these activation efforts are either (a) completely unspecific in terms of the BGC(s) to be activated [e.g., by adding general elicitors to the cell culture, co-cultivation approaches, or strain-cultivations following the "one strain-many compounds" (OSMAC) strategy] or (b) they are absolutely specific for the BGC of interest (e.g., heterologous expression of the BGC, introduction of an artificial promoter in front of the BGC, or manipulation of a cluster-situated regulator) as reviewed in Ochi and Hosaka (2013) and Zhu et al. (2014). Both approaches have their drawbacks, as there are either major analytical efforts to identify the product from the silent gene cluster [in terms of (a)] or tedious genetic engineering efforts to manipulate the producer strain [in terms of (b)]. Thus, it would be highly beneficial to have a more general activation strategy that targets a defined set of BGCs, which would tackle both issues. In a recent study from Martínez-Burgo et al. (2019) it has been shown that conserved pathway-specific activators can be used to activate BGC expression in a foreign Streptomyces strain. In this study the heterologous expression of the PAS-LuxR type regulator gene pimM from Streptomyces natalensis in Streptomyces clavuligerus led to the activation of clavulanic acid, cephamycin C, and tunicamycin production.

Here we demonstrate that Streptomyces Antibiotic Regulatory Protein (SARP)-type regulators can be used as activators of certain antibiotic gene clusters in actinomycetes and describe a genome-based approach to screen for SARP-activated gene clusters. SARPs have exclusively been found in actinomycetes, especially in streptomycetes, where they act as pathway-specific activators of secondary metabolite biosynthesis (Bibb, 2005). They are known to be associated with various antibiotic gene clusters, encoding type I- (Bate et al., 2002; Takano et al., 2005; Novakova et al., 2011) and type II-PKS derived polyketides (Lombó et al., 1999; Sheldon et al., 2002; Aigle et al., 2005; Novakova et al., 2011), ribosomally (Widdick et al., 2003; Wu et al., 2018) and non-ribosomally synthesized peptides (Ryding et al., 2002), hybrid polyketide-peptide compounds (Pulsawat et al., 2009; Suzuki et al., 2010; Xie et al., 2012; Salehi-Najafabadi et al., 2014; Mast et al., 2015; Ye et al., 2018), β-azachinones (Santamarta et al., 2002; Rodríguez et al., 2008; Kurniawan et al., 2014), and azoxy compounds (Garg et al., 2002). SARP genes usually are located within the BGC they are regulating. The encoded SARP gene products are characterized by a winged helix-turn-helix (HTH) DNA-binding motif at the N-terminus that binds to a conserved recognition sequence within the major groove of the target DNA (Wietzorrek and Bibb, 1997; Liu et al., 2013). The DNA recognition sequence constitutes direct heptameric repeat sequences followed by 4 bp spacers, which are often localized between the −10 and the −35 promoter element of the respective target DNA. Such a localization has already been described for the SARP type regulator AfsR from Streptomyces coelicolor, which binds to a recognition sequence 8 bp upstream of the −10 element (Tanaka et al., 2007). Also the SARP regulators Aur1PR4 from Streptomyces aureofaciens and FdmR1 from Streptomyces griseus bind to heptameric repeat sequences, which are located 8 bp upstream of the −10 region (Chen et al., 2008; Rehakova et al., 2013). ActII-ORF4 from S. coelicolor interacts with the −35 element for transcriptional activation (Arias et al., 1999). DnrI from Streptomyces peucetius and SanG from Streptomyces ansochromogenes bind to interaction sites that occur within the −35 element (Sheldon et al., 2002; He et al., 2010). It is suggested that in general the SARP binding site overlaps with the −35 region of the target promoter, which is a binding region of the majority of repressors but not activators. Thus, SARPdriven transcriptional activation has been proposed to occur via a novel mechanism (Tanaka et al., 2007). The C-terminal bacterial activation domain (BTAD) of the SARP protein activates the transcription of the target genes by recruiting the RNA polymerase (RNAP) to the respective promoter, where a ternary DNA–SARP–RNAP complex is formed allowing for transcriptional initiation (Tanaka et al., 2007). "Small" SARPtype activators only contain the HTH DNA binding and BTAD domain, whereas "large" SARPS carry additional domains at the C-terminal side of the protein. These domains include a domain of unknown function belonging to the P-loop NTPase family, and one or more copies of a tetratricopeptide repeat (TPR) motif (Liu et al., 2013). A typical "small" SARP-type activator is represented by PapR2, which has been identified as the major activator of pristinamycin biosynthesis in Streptomyces pristinaespiralis (Mast et al., 2015). A papR2 deletion mutant is unable to produce any pristinamycin, depicting that PapR2 is essential for pristinamycin biosynthesis (Mast et al., 2015). In contrast, overexpression of papR2 in S. pristinaespiralis leads to an increased pristinamycin production, which shows that PapR2 has an activating function (Mast et al., 2015). With the help of electromobility shift assays (EMSAs) and (quantitative) reverse transcription PCR [RT-(q)PCR] analysis the PapR2 target genes have been identified in the pristinamycin producer and a conserved PapR2 binding site was proposed (Mast et al., 2015).

In this study, we report on the potential of SARP-type regulators as genetic engineering devices for the activation of (silent) BGCs in actinomycetes. SARP-type regulators are present predominantly in actinomycetes with an abundance of 98% in the genus Streptomyces. We demonstrate that the SARPtype regulator PapR2 activates the silent undecylprodigiosin (Red) gene cluster in Streptomyces lividans. Additionally, we provide evidence for a PapR2-guided activation of a BGC in the poorly studied Indonesian strain isolate Streptomyces sp. SHP22-7 (SHP22-7), which yielded an increased production of the nucleoside antibiotic plicacetin.

#### MATERIALS AND METHODS

fmicb-11-00225 February 14, 2020 Time: 19:22 # 3

#### Bacterial Strains, Plasmids, and Cultivation Conditions

The bacterial strains and plasmids used in this study are listed in **Supplementary Table S1**. For routine cloning strategies Escherichia coli Novablue (Novagen) was used. S. lividans T7 (Fischer, 1996) and SHP22-7 (Handayani et al., 2018) were applied for antibiotic production analysis, generation of overexpression strains, and transcriptional analysis. Cloning procedures and strain cultivation were carried out as described before (Mast et al., 2015). For cultivation and isolation of RNA, Streptomyces strains were grown in 100 ml of R5 medium in 500-ml Erlenmeyer flasks (with steel springs) on an orbital shaker (180 rpm) at 28◦C (Kieser et al., 2000). For isolation of genomic DNA and protoplast transformation experiments, strains were grown in 100 ml of S-medium (Kieser et al., 2000). Kanamycin (50 µg/ml), apramycin (50 µg/ml), or thiostrepton (20 µg/ml) were used for selection when appropriate. For antibiotic production experiments with S. lividans, strains were grown in YEME medium as reported before (Mast et al., 2015). For antibiotic production experiments with SHP22-7, strains (SHP22-7papR2-OE; references: SHP22-7pRM4 and SHP22-7 WT) were grown in 50 ml NL410 medium consisting of glucose (10 g l−<sup>1</sup> ), glycerol (10 g l−<sup>1</sup> ), oat meal (5 g l−<sup>1</sup> ), soy flour (10 g l−<sup>1</sup> ), yeast extract (5 g l−<sup>1</sup> ), Bacto casamino acids (5 g l−<sup>1</sup> ), CaCO<sup>3</sup> (1 g l−<sup>1</sup> ), and distilled water (pH was adjusted to 7.0 with NaOH) as a preculture. After 3 days, 10 ml of preculture was transferred to 100 ml of sterile main culture medium NL19, consisting of mannitol (20 g l−<sup>1</sup> ), soy flour (20 g l−<sup>1</sup> ), and distilled water (pH adjusted to pH 7.5 with NaOH). Cells were grown for 168 h at 28◦C.

#### Molecular Cloning

Basic procedures for DNA manipulation were performed as described previously (Sambrook et al., 1989; Kieser et al., 2000). Primers used for PCR were obtained from MWG Biotech AG (MWG, Ebersberg, Germany) and are listed in **Supplementary Table S1**.

#### Construction of the papR2 Overexpression Strain SHP22-7papR2-OE

For papR2 overexpression experiments with SHP22-7, the papR2 gene was isolated as a NdeI/HindIII-fragment from plasmid pGM190/papR2 (Mast et al., 2015) and was cloned into the NdeI/HindIII restriction site of the integrative expression vector pRM4. In the resulting overexpression construct pRM4/papR2, the papR2 gene is under control of the constitutive promoter of the erythromycin resistance gene ermEp<sup>∗</sup> . pRM4/papR2 was transferred to SHP22-7 by protoplast transformation. Transformants were selected with apramycin (50 µg/ml), which resulted in the overexpression strain SHP22-7papR2- OE. Strain SHP22-7pRM4, harboring the empty pRM4 vector, was generated by protoplast transformation accordingly and served as a reference.

### PapR2 Protein Expression in S. lividans

For papR2 overexpression experiments SLpapR2-OE precultures were grown in 100 ml of YEME liquid medium for 2 days at 28◦C. Five milliliters of preculture was used as inoculum for 100 ml YEME liquid medium as main culture with thiostrepton (12.5 µg/ml) as inductor for gene expression (Mast et al., 2015). The main culture was cultivated for 3 days at 28◦C. PapR2 protein purification was carried out as reported before (Mast et al., 2015).

#### Spectrophotometrical Analysis for Red Detection

For Red detection, culture supernatant from SLpapR2-OE and SLpGM190 (reference) was treated as reported in Onaka et al. (2011) and absorption was measured with a Hitachi U-2000 spectrophotometer.

# Sample Treatment for SHP22-7 Bioassays and Compound Detection

For SHP22-7 compound detection and bioassay tests, 5 ml culture samples of SHP22-7papR2-OE (references: SHP22-7pRM4 and SHP22-7 WT) was extracted with 5 ml ethyl acetate for 30 min at RT. Ethyl acetate samples were concentrated in vacuo completely and then redissolved in 0.75 ml of methanol. Methanolic extracts were used for bioassays and high-performance liquid chromatography/mass spectrometry (HPLC–MS) analysis.

#### Bioassays

Antibiotic activity was analyzed in disc diffusion assays using Bacillus subtilis ATCC6051 as test organism. Thirty microliters of methanolic extract from three independent biological samples of SHP22-7 WT, SHP22-7pRM4, and SHP22-7papR2- OE, respectively, was pipetted on a filter disc, which was placed on a B. subtilis test plate. Five microliters of kanamycin (50 µg/ml) was applied as a positive control and 30 µl of methanol as negative control to test the functionality of the B. subtilis bioassay plates. The plates were incubated overnight at 37◦C. Antibiotic activity was quantified by measuring the diameter of the inhibition zone around the filter discs. The bioassay was carried out as 10 independent biological replicates.

#### HPLC and HPLC–MS Analysis for Amicetin/Plicacetin Detection

High-performance liquid chromatography analyses were performed with a HP1090M system with ChemStation 3D software rev. A.08.03 (Agilent Technologies, Waldbronn, Germany) on a Nucleosil C18 column (5 µm, 125 mm × 3 mm) fitted with a precolumn (20 × 3 mm) and with a flow rate of 850 µl min−<sup>1</sup> . Chromatography was done by linear gradient elution from 95.5 solvent A (water with 0.1% phosphoric acid) to 100% solvent B [acetonitrile (ACN)] over 15 min. The injection volume was 5 µl. Multiple wavelength monitoring was

performed at 210, 230, 260, 280, 310, 435, and 500 nm. UV-Vis spectra were measured from 200 to 600 nm. The evaluation of the chromatograms (210 nm only) was done by means of an in-house HPLC–UV–Vis database.

High-performance liquid chromatography–mass spectrometry analysis of amicetin/plicacetin was performed with an Agilent 1200 series chromatography system (binary pump, high performance autosampler, DAD-detector) coupled with an LC/MSD Ultra Trap System XCT 6330 (Agilent Technologies, Waldbronn, Germany). The sample (5 µl) was injected on a Nucleosil 100 C18 column (3 µm, 100 × 2 mm) fitted with a precolumn (3 µm, 10 × 2 mm) at a flow rate of 400 µl/min and a linear gradient from 100% solvent A (0.1% HCOOH in water) to 100% solvent B (0.06% HCOOH in ACN) over 15 min at 40◦C. UV–Vis-detection was done at 220, 260, 280, 360, and 435 nm. Electrospray ionization was performed in positive and negative ultra-scan mode (alternating) with a capillary voltage of 3.5 kV and a drying gas temperature of 350◦C. Detection of m/z values was conducted with Agilent DataAnalysis for 6300 Series IonTrap LC/MS Version 3.4 (Bruker Daltonik). Upon HPLC–MS analyses amicetin, plicacetin, and plicacetin isomer were identified by comparisons of their UV/visible spectra, retention times, and molecular masses with authentic standards, as m/z 617.1 [M-H] − and m/z 516.1 [M-H] <sup>−</sup>, respectively.

#### AntiSMASH Analysis

With the webtool antiSMASH whole genomes can be scanned for the occurrence of BGCs. Gene cluster similarity is given in % and indicates the number of similar genes to a known cluster. Genes are similar if a BLAST-alignment results in an e-value <10−<sup>5</sup> and the sequence identity is >30%. Additionally, the shortest alignment must enclose >25% of the sequence. If all genes of a known cluster can be found in the query cluster, the similarity of the sequences is 100%. The similarity lowers if less genes of the known cluster can be found in the query cluster (Medema et al., 2011).

#### PatScan Analysis

PatScan analysis (Blin et al., 2018) was performed with the S. lividans T7 genome (GenBank Accession Number ACEY00000000) and the PapR2 consensus sequence 5<sup>0</sup> -GTCAGSS-3<sup>0</sup> using the software at https: //patscan.secondarymetabolites.org/.

#### PapR2 Electromobility Shift Assays

For EMSAs with Red-specific promoter regions, 182 bp DNA fragments of the upstream regions of redP and redQ were amplified by PCR from genomic DNA of S. lividans T7 with primer pairs PredPfw/rv and PredQfw/rv, respectively (**Supplementary Table S1**). For EMSAs with the pliA promoter region, a 230 bp DNA fragment of the upstream region of pliA was amplified by PCR from genomic DNA of strain SHP22-7 as template and primer pair PpliAfw/rv (**Supplementary Table S1**). Promoter DNA amplificates included a 16 bp Cy5 adapter sequence, each at the 3<sup>0</sup> - and 5<sup>0</sup> -end, which was added via the respective primer sequences. The generated amplificates were used as templates in a second PCR approach together with a Cy5 primer (**Supplementary Table S1**) in order to conduct Cy5 labeling of the promoter regions. Promoter labeling and PapR2 EMSAs were carried out with variable concentrations of PapR2 protein sample as reported before (Mast et al., 2015). To verify the specificity of the PapR2-DNA binding, an excess of unlabeled, specific, and non-specific DNA, respectively, was added to the EMSA mixture as described previously (Mast et al., 2015). DNA bands were visualized by fluorescence imaging using a Typhoon TrioTM Variable Mode Imager (GE Healthcare).

#### Transcriptional Analysis by Reverse Transcription Analysis (RT-PCR)

SLpGM190 and SLpapR2-OE, as well as SHP22-7, SHP22- 7pRM4, and SHP22-7papR2-OE were each grown under papR2 overexpression conditions as described above. Thirty milliliters of each cell culture was harvested after 48 h. Cell disruption was carried out with glass beads (150–212 µm; Sigma) at 6,500 rpm, 1 × 20-30 s, using a Precellys Homogenizer (Peqlab). Total RNA was isolated as described previously (Sambrook et al., 1989) and served as the basis for RT-PCR experiments. DNA was removed by digestion with DNase (Thermo Fisher Scientific) and absence of DNA was verified via PCR analysis. RNA concentrations and quality were checked using a NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific). cDNA from 3 µg RNA was generated with random primers, reverse transcriptase, and cofactors (Fermentas). For RT-PCRs, primers were used that amplify cDNA of 200–250 bp from internal gene sequences. PCR conditions were 98◦C for 5 min, followed by 30 cycles of 95◦C for 30 s, 60◦C for 30 s, and 72◦C for 45 s, and a final cycling step at 72◦C for 5 min. As a positive control, cDNA was amplified from the 16S rRNA transcript, which is transcribed constitutively. To exclude DNA contamination, negative controls were carried out by using total RNA as a template for each RT-PCR reaction. At least three independent biological replicates have been tested.

#### Transcriptional Analysis by Real-Time qPCR

Real-time qPCR analysis was applied for quantitative cDNA determination. PCR reactions were run with SYBR <sup>R</sup> Green Supermix (BioRad) on an iQ5 Multicolor Real-Time-PCR Detection System (BioRad). The SYBR <sup>R</sup> Green Dye shows increased fluorescence when bound to double-stranded DNA. The fluorescence is measured at 494 and 521 nm and gives the proportional amount of generated dsDNA. cDNA was generated from cultures of SLpGM190 and SLpapR2-OE as described above. The primer pairs redPintfw/rv and redQintfw/rv, respectively (**Supplementary Table S1**), which amplify fragments of about 180 bp from internal gene sequences, were used together with cDNA as template in qPCR reactions. hrdB was used as housekeeping gene in each experiment in order to standardize the results by eliminating variation in RNA and cDNA quantity and quality. Each reaction mixture of 10 µL volume contained 5 µL SYBR <sup>R</sup> Green Supermix (BioRad), 3.85 µL nuclease-free water, 0.2 µL of each primer, and 0.75 µL template. PCR conditions were 98◦C for 5 min, followed by 35 cycles of 95◦C for 20 s, and 57◦C for 30 s. To determine amplification specificity, melting curve analyses were performed after the last cycle, showing in all cases one single peak. Results were analyzed using the 11CT-method (Livak and Schmittgen, 2001). Changes in gene expression are represented in relation to the data from samples of SLpGM190. Data are presented as the results from six independent biological replicates.

#### Database Analysis

fmicb-11-00225 February 14, 2020 Time: 19:22 # 5

FASTA sequences of SARP proteins were extracted from the antiSMASH database version 2 (Blin et al., 2019) by querying for all genes that hit the antiSMASH smCoG (secondary metabolite clusters of orthologous groups) profile SMCOG1041 (transcriptional regulator, SARP family). These hits were checked for the presence of at least one out of four SARP-related profiles from the PFAM database: PF00486.27 (Trans\_reg\_C, the HTH-style DNA binding domain), PF93704.16 (BTAD, the transcriptional activator), PF00931.22 (NB-ARC, a domain of unknown function found, e.g., in Saccharopolyspora erythraea SARPs), and PF13424.6 (TPR\_12, a TPR found in many larger SARPs). Sequences that did not hit at least one of these domains were discarded. Sequences were then annotated by which of the four profiles were hit and grouped by taxonomic order and BCG type.

#### RESULTS

### PapR2 Induces Expression of the Silent Red Gene Cluster in S. lividans

Streptomyces lividans is a widely used heterologous host strain, which under specific laboratory conditions does not exhibit production of the two pigmented secondary metabolites actinorhodin and undecylprodigiosin (Hu et al., 2002; Martinez et al., 2005; Rodríguez et al., 2013). In frame of analyzing the regulatory role of PapR2 in S. pristinaespiralis, the papR2 gene was heterologously expressed in S. lividans T7 for protein purification purposes (Mast et al., 2015). For overexpression experiments the S. lividans strain SLpapR2-OE was used, which harbors the papR2 gene under control of the thiostrepton-inducible promoter PtipA on the replicative medium-copy plasmid pGM190 (Mast et al., 2015). S. lividans strain SLpGM190 was used as a reference, containing the pGM190 empty vector. After 2– 3 days of growth in YEME liquid medium with thiostrepton as inductor for gene expression, the whole SLpapR2-OE culture, as well as the culture supernatant showed an intensive red pigment formation, which was not observed for samples of SLpGM190 (**Figure 1A**). This phenomenon was also observed on R5 agar with thiostrepton, where SLpapR2-OE mycelium was intensively red colored after 3–4 days of cultivation, whereas SLpGM190 mycelium was not (**Supplementary Figure S1**). S. lividans is known to harbor a Red BCG, which remains silent under normal growth conditions (Horinouchi et al., 1986). Red biosynthesis has mainly been studied in S. coelicolor; however, since S. coelicolor and S. lividans are very closely related species, knowledge on Red biosynthesis and regulation can be transferred to S. lividans (van Wezel et al., 2000; Lewis et al., 2010).To investigate, whether the red color of the SLpapR2-OE cultures originates from the formation of the Red metabolite, we performed spectrophotometrical analysis. A pH shift was carried out with the culture supernatant of SLpapR2-OE and the absorption maxima of the sample was determined by using a Hitachi U-2000 spectrophotometer. Spectrophotometrical analysis led to the detection of the Red-specific spectral absorption maxima (Onaka et al., 2002) at 533 and 468 nm under acidic and basic conditions, respectively (**Figure 1B**), which proved that the SLpapR2-OE samples contained Red (**Figure 1C**). These data suggested that the overexpression of papR2 in S. lividans induced Red biosynthesis.

#### PapR2 Mimics the Function of the S. lividans SARP-Type Regulator RedD

In S. lividans Red biosynthesis is under control of the SARPtype regulator RedD, which directly activates the Red biosynthetic genes (Takano et al., 1992; White and Bibb, 1997). An amino acid sequence comparison using BLASTP revealed that PapR2 and RedD are highly similar to each other (44% identity, 55% similarity) (Mast et al., 2015). This amino acid sequence similarity was even higher for the HTH motif of the protein (66% identity, 75% similarity). Thus, we suspected that PapR2 may substitute for the function of RedD and activates Red biosynthesis in S. lividans. In order to identify potential SARP-type binding motifs within the S. lividans Red BGC, the genome was analyzed with the bioinformatic tool PatScan, which allows for the identification of specific sequence patterns in a given genome sequence (Blin et al., 2018). PatScan analysis was performed with the previously described PapR2 consensus motif (5'-GTCAGSS-3') (Mast et al., 2015) as sequence pattern on the S. lividans genome sequence. Thereby, two highly conserved PapR2-like motifs were identified within the intergenic region of the Red-specific biosynthetic genes redP (SCO5888) and redQ (SCO5887) with each 100 and 96.5% identity, respectively (**Figure 2A**). redP encodes a 3 ketoacyl-acyl carrier protein synthase, whereas redQ codes for an acyl carrier protein, both of which have been shown to be involved in Red biosynthesis (Mo et al., 2008). To analyze the functionality of the identified SARP motifs, EMSAs were performed with the PapR2 protein and the upstream regions of redP (PredP, 182 bp) and redQ (PredQ, 182 bp), respectively, harboring the PapR2 consensus sequence. EMSAs showed that PapR2 specifically binds to the PredP and PredQ fragment, respectively (**Figure 2B**), revealing the functionality of the identified SARP-type motifs.

To confirm the regulatory effect of PapR2 on the transcription of the Red BGC, we performed RT-PCR and quantitative qPCR experiments. For these studies SLpapR2-OE, as well as the reference strain SLpGM190 were grown in R5 medium. After 72 h of cultivation samples were harvested for RNA isolation. For each strain six biological replicates were performed. Isolated RNA was used as a template in RT-PCR experiments as a negative control (**Figure 3A**), whereas cDNA was used together with 16S primers as a positive control (**Figure 3B**). For red gene-specific transcriptional analysis, isolated RNA was used as template for RT-PCR experiments with the primer pairs redPintfw/rv

and redQintfw/rv, which annealed to internal parts of redP and redQ, respectively. The transcriptional analysis revealed that there are stronger signals for the redP and redQ cDNA amplificates in SLpapR2-OE samples compared to SLpGM190 samples (**Figures 3C,D**, respectively). In order to quantify the amount of redP and redQ RT-qPCR was performed. Based on the threshold cycle it was calculated that redP and redQ transcripts were increased to 57- and 492-fold, respectively, in samples of SLpapR2-OE compared to samples of SLpGM190 (**Supplementary Figure S2**). Thus, transcriptional analyses demonstrated that PapR2 activates the transcription of the Red biosynthetic genes redP and redQ. Due to the intensive Red production of strain SLpapR2-OE together with the data obtained from PapR2 EMSA studies and transcriptional analysis, we propose that overexpressed PapR2 in S. lividans takes over the regulatory function of RedD and activates the transcription of the Red biosynthetic genes, which leads to the formation of the red colored secondary metabolite Red.

#### PapR2 Expression Improves Antibiotic Activity of SHP22-7

In order to study the general application of SARP-type regulators for activation of antibiotic biosynthesis, we exemplary tested strain SHP22-7 as a host for papR2 expression. SHP22-7 is a novel strain isolate from a soil sample of the unique desert island Enggano, Indonesia (Handayani et al., 2018). SHP22-7 shows broad-spectrum antibacterial activity against Gram-positive and Gram-negative bacteria, including E. coli, Staphylococcus carnosus, Micrococcus luteus, and B. subtilis. The SHP22-7 genome has been sequenced recently and antiSMASH analysis led to the identification of 25 potential secondary metabolite gene clusters (Handayani et al., 2018; **Table 1**). Four of the gene clusters (clusters 3, 6, 10, and 15) contain SARP genes, which makes SHP22-7 a good candidate strain for a SARPguided activation. For papR2 expression studies in SHP22-7, the papR2 gene was cloned into the integrative plasmid pRM4 under control of the constitutive ermE<sup>∗</sup> promoter, resulting in construct pRM4/papR2. The pRM4 vector was used as expression plasmid in these analyses to avoid addition of thiostrepton as inductor (see above), which would influence subsequent antibacterial bioassays. The plasmid was transferred to SHP22- 7 by protoplast transformation. The resulting expression strain SHP22-7papR2-OE, as well as the two reference strains SHP22- 7pRM4, which harbors the empty pRM4 vector, and the SHP22-7 wild-type (WT) were used for antibacterial bioassay studies. SHP22-7papR2-OE, SHP22-7pRM4, and SHP22-7 WT were each cultivated in NL19 medium and samples were taken at 168 h. Methanolic culture extracts were applied for antibacterial bioassays using B. subtilis as test organism. Filter discs with kanamycin and methanol served as positive and negative control, respectively. Bioassays were carried out as 10 independent biological replicates. Thereby, significantly larger inhibition zones against B. subtilis were observed on average with extracts from SHP22-7papR2-OE compared to extracts from SHP22- 7pRM4 or SHP22-7 WT (**Figure 4A**). The extract of SHP22- 7papR2-OE caused a zone of inhibition of 11.7 ± 2.1 mm, whereas the extracts of SHP22-7pRM4 and SHP22-7 WT yielded smaller

inhibitions zones of 8.3 ± 2.6 and 7.3 ± 2.7 mm, respectively (**Figure 4B**). These data showed that papR2 expression in SHP22- 7 leads to a significantly improved antibiotic activity.

#### PapR2 Activates Transcription of Cluster 9 in SHP22-7

To identify the substance that is responsible for the antibiotic activity of SHP22-7 but avoid elaborative chemical analytics in the first place, we screened the SHP22-7 genome sequence for the occurrence of PapR2-like consensus sequences by using the PatScan tool. Here, we only considered motifs that were located within intergenic regions of genes from suggested SHP22- 7 BGCs. The motif search led to the identification of PapR2 like consensus sequences within 5 of the 25 predicted BGCs from SHP22-7. These included clusters 6 and 10 (NRPS-like gene cluster), cluster 15 (type 2 PKS-butyrolactone-like gene cluster), as well as clusters 3 and 9 ("other" type of gene cluster) (**Table 1**). In order to find out which of the five gene clusters is activated by PapR2, comparative RT-PCR analysis was carried out with SHP22-7papR2-OE, SHP22-7pRM4, and SHP22-7 WT in four independent biological replicates. Cells were grown under antibiotic production conditions and samples were harvested at 168 h of cultivation. Two samples were collected for each strain and growth time point, whereby one sample was used for RT-PCR analysis and the second one for HPLC– MS analysis (see the section "SARP-Type Regulatory Genes Are Widespread in BGCs From Diverse Actinobacteria"). RT-PCR was carried out with RNA isolated from cell pellets of SHP22- 7papR2-OE, SHP22-7pRM4, and SHP22-7 WT, respectively. Isolated RNA was used as a template in RT-PCR experiments as a negative control (**Figure 5A**), whereas cDNA was used together with 16S primers as a positive control (**Figure 5B**). For cluster specific transcriptional analysis, cDNA was generated with primer pairs cl3fw/rv, cl6fw/rv, cl9fw/rv, cl10fw/rv, and cl15fw/rv (**Supplementary Table S1**), each aligning to a predicted biosynthesis gene of clusters 3, 6, 9, 10, and 15, respectively. The transcriptional analysis revealed that there is a stronger signal for the cluster 9 amplificate in the SHP22-7papR2-OE samples compared to the samples of SHP22-7pRM4 and SHP22-7 WT, where there is nearly no signal at all (**Figure 5C**). This difference in signal intensity could not be observed for clusters 3, 6, 10, and 15 (**Figure 5C**). Thus, these data suggested that PapR2 activates the transcription of cluster 9 in SHP22-7papR2-OE.

### Cluster 9 Resembles an Amicetin Gene Cluster

In order to deduce for what type of metabolite cluster 9 encodes for, we analyzed the cluster sequence region in detail. The antiSMASH output predicted a 75% similarity of cluster 9 to the plicacetin/amicetin gene cluster from Streptomyces vinaceusdrappus NRRL 2363 (**Table 1**). The cluster similarity describes the number of genes with a similarity above the ClusterBlast threshold of 30% sequence similarity at over 25% coverage. Manually cluster analysis by sequence alignments yielded a 100% sequence similarity since the sequence region from amiA–amiD was not recognized by antiSMASH to be part of the amicetin BGC (**Figures 6A** vs. **B**). Plicacetin (**Figure 6C**) and amicetin (**Figure 6D**) are disaccharide pyrimidine nucleoside antibiotics with a broad-spectrum antibacterial (especially against Mycobacterium tuberculosis) and antiviral activity. They act as peptidyl transferase inhibitors and thus inhibit protein synthesis. Amicetin consists of the two deoxysugar moieties, D-amosamine and D-amicetose, as well as cytosine, p-aminobenzoic acid (PABA), and a terminal methylserine moiety, whereby the latter moiety is missing in plicacetin (Zhang G.G. et al., 2012; Korzybski et al., 2013). The gene cluster analysis of SHP22-7 revealed that the PapR2-like motif is located directly in front of the orf pliA (**Figure 6B**) and shows a rather weak sequence identity of 65.5% to the PapR2 consensus motif (**Figure 2A**). The same motif is present upstream of amiA in S. vinaceusdrappus. The pliA gene presents 99.75% gene nucleotide sequence identity to amiA of S. vinaceusdrappus, which translates to a 100% amino acid sequence identity among the predicted gene products. In S. vinaceusdrappus amiA is the first gene of the amicetin BCG and encodes a putative 4-amino-4-deoxychorismate lyase, which is suggested to catalyze the conversion from 4-amino-4 deoxychorismate to PABA (Zhang G.G. et al., 2012). Remarkably, all genes in the amicetin gene cluster of S. vinaceusdrappus and the amicetin-like gene cluster of SHP22-7 are organized in one direction, suggesting a unidirectional transcription. In this context, it would make sense that regulatory activation targets the promoter of the first gene of the unidirectional BGC. In order to

find out if PapR2 can bind to the pliA promoter region, EMSAs were performed with the PapR2 protein and the pliA upstream region containing the PapR2 consensus sequence. However, no shifted band was detected in these assays (data not shown). Thus, it might be that the motif is not functional at all and does not constitute a SARP binding motif. It could also be that due to the less conserved PapR2 consensus sequence, the motif is not functional in such an in vitro assay. The presence of a SARP consensus sequence would hint for a pathway-specific regulatory gene located within the amicetin(-like) BGC. However, a SARPtype regulatory gene has not been identified in any of the amicetin(-like) clusters from SHP22-7 nor S. vinaceusdrappus (**Table 1**; Zhang X. et al., 2012). Overall, six SARP genes have been identified in total within the SHP22-7 genome (**Supplementary Table S2**). All of them are part of BGCs (clusters 3, 6, 10, and 15). Cluster 15 harbors three SARP genes, whereas the other BGCs each contain one SARP gene (**Table 1**). Furthermore, putative SARP binding motifs have been found within the promoter regions of the BGCs (**Table 1** and **Supplementary Table S2**).

It might be possible that one of these SARP-type regulators plays a role in trans-activating amicetin cluster transcription. So far, regulation of amicetin biosynthesis is not understood. In S. vinaceusdrappus three genes [orf(-3), orf(-2), amiP] have been identified, which encode for putative transcriptional regulators, whereby orf(-3) and orf(-2) seem not to be part of the amicetin BGC and amiP codes for TetR-like transcriptional regulator, which usually function as repressors of antibiotic biosynthesis (Zhang G.G. et al., 2012). Thus far, it cannot be excluded that also trans-acting regulator(s) are involved in the regulation of amicetin biosynthesis.

#### PapR2 Activates Plicacetin Production in SHP22-7

To investigate if an amicetin(-like) antibiotic is produced by SHP22-7, we analyzed methanolic extracts of cell culture samples from SHP22-7papR2-OE by HPLC/MS. To analyze if production is influenced by papR2 expression, extracts from samples of


TABLE 1 | Secondary metabolite gene clusters of Streptomyces sp. SHP22-7 as predicted by antiSMASH 4.0 with an indication of the presence of cluster-situated SARP genes (count), predicted SARP proteins with amino acid sequence homology to PapR2 (I = identity, S = similarity), as well as identified SARP-type motifs.

+, cluster contains gene/motif; −, cluster does not contain gene/motif.

SHP22-7pRM4 and SHP22-7 WT served as references. For these analyses we used the second culture samples obtained from SHP22-7papR2-OE, SHP22-7pRM4, and SHP22-7 WT, respectively (see above). Based on comparisons with an inhouse substance database, HPLC analysis revealed the presence of amicetin [retention time (RT) 4.0 min], plicacetin (RT 4.8 min), and a plicacetin isomer (RT 5.6 min) in all three samples (**Figure 7A**). The identity of the compounds was verified by MS/MS analysis (amicetin m/z 617.1 [M-H]; plicacetin m/z 516.1 [M-H]; plicacetin isomer m/z 516.1 [M-H]) (**Figure 7B**; for HRMS data see **Supplementary Figures S3**, **S4**). Comparisons between the HPLC spectra from samples of SHP22-7papR2-OE, SHP22-7pRM4, and SHP22-7 WT displayed that peak intensities were especially increased for plicacetin (mAU 605) and the plicacetin isomer (mAU 259) in samples of SHP22-7papR2-OE compared to samples of SHP22-7pRM4 [plicacetin (mAU 120), plicacetin isomer (mAU 97)] and SHP22-7 WT [plicacetin (mAU 66), plicacetin isomer (mAU 57)] (**Figure 7C**). Thus, it could be shown that papR2 expression in SHP22-7 activates explicitly plicacetin biosynthesis, whereas amicetin biosynthesis seems not to be affected. Overall, with this result it could be confirmed that PapR2 induced cluster 9 transcription, which resulted in an increased plicacetin production. Thus, it can be concluded that the increased bioactivity of SHP22-7papR2-OE samples against B. subtilis arises from the increased production of the nucleoside antibiotic plicacetin. However, since no direct interaction of the PapR2 regulator with the amicetin promoter region could be shown by EMSA analysis, it cannot be deduced if the activation effect is a direct or an indirect one. The SARP consensus motif harbors the central TCA triad, which is also present in the Pho

box (GTTCACC), resembling the target site of the phosphate control two-component system PhoP/PhoR (Martin and Liras, 2020). This sequence region is also known to be bound by the large size SARP regulator AfsR (Martin and Liras, 2020). Additionally, the central TCA triad can be present in binding motifs recognized by the nitrogen regulator GlnR or the DmdR1 (Flores and Martín, 2004; Martin and Liras, 2020). Cross-talk between different transcriptional regulators via the interaction of the same binding sites have been shown before (Martín et al., 2011; Martin and Liras, 2020). Thus, it is also possible that transcriptional activation of the amicetin BGC is an effect from multiple regulatory interactions.

# SARP-Type Regulatory Genes Are Widespread in BGCs From Diverse Actinobacteria

Of the top 10 genera containing SARP-type regulators in the antiSMASH database 2 (Blin et al., 2019), 9 belong to the phylum Actinobacteria. Broken up by genus, 98% of Streptomyces (611/625), 81% of Nocardia (78/96), 100% of Salinispora (72/72), 100% of Micromonospora (62/62), and 97% of Amycolatopsis (38/39) genomes harbor SARP-type regulatory genes, showing their prevalence in filamentous Actinobacteria. On the other hand, only 42% of Mycobacterium (115/276) species have SARP-type regulators. Outside of Actinobacteria, mainly Proteobacteria have hits for the SMCOG1041 profile, but mostly lack hits for the HTH motif domain or the BTAD (transcriptional activator domain) [e.g., 6% (79/1236) of Pseudomonas hit the smCoG profile, but none of them contain a BTAD match]. Thus, SMCOG1041 profiles from Proteobacteria may not represent typical SARP-type regulators.

Of the total of 6525 proteins containing a hit against any of the four SARP-related PFAM domains, 3289 (50%) are from the order Streptomycetales (**Figure 8**). These again break down into 47% "small" SARPs (only containing the HTH and BTAD domains) and 36% "large" SARPs (also containing the NB-ARC and/or TPR domains). The remaining proteins miss the HTH and/or BTAD domains, likely an artifact of bad sequencing data in published draft genomes. The overrepresentation of SARP-type regulators in Streptomycetales may also be explained by the higher abundance of available genomes in the database. Pseudonocardiales cover 17% of the SARP-type proteins (1080/6525). 31% (330/1080) are "small" SARPs and 39% (421/1080) "large" SARPs. In Micromonosporales, accounting for 13% (822/6526) of the dataset, 30% (244/822) are "small" SARPs, and 40% (330/822) are "large" SARPs (**Figure 8**). Interestingly, Micromonosporales is the only order to contain a significant amount (22%, 184/882) of SARP-type proteins only containing a hit against BTAD without hitting the HTH domain, suggesting a different subfamily of transcriptional activator. Furthermore, SARP genes were found to be present in various different types of BGCs with a prominent abundance in NRPS

profile of all three strains. Peaks in the HPLC spectrum representing amicetin are marked with a, plicacetin marked with b, and plicacetin isomer marked with c (A). Mass spectra of plicacetin (Upper), amicetin (Middle), and a plicacetin isomer (Lower) (B) isolated from SHP22-7 WT. Graphical representation of production levels of plicacetin and plicacetin isomer in each strain. n = 10; <sup>∗</sup> means significance (C).

and PKS gene clusters (**Figure 9**), which however might also be associated with the higher frequency of these cluster types in the database.

NB-ARC domain of unknown function, and R is the tetratricopeptide repeat

DISCUSSION

domain.

Regulation is the screw plug to unlock the biosynthetic potential of natural compound producers. Here, we showed that heterologous expression of the SARP-type regulatory gene papR2 in the foreign host S. lividans leads to transcriptional activation of the silent Red BGC. Since SARP regulators show a comparable protein architecture and bind to similar recognition sequences at the DNA, they can substitute for their regulatory functions as exemplified for PapR2 and RedD. This is also underpinned by a previous study, where it has been shown that overexpression of the SARP gene vlml from the valanimycin producer Streptomyces viridifaciens in a redD mutant of S. coelicolor M512 restores Red production, demonstrating that vlmI can complement a redD mutation (Garg and Parry, 2010). Besides having gained indications for such kind of cross-regulation, our data also contribute to a better understanding of the regulation of Red biosynthesis in general. So far, there have been only bioinformatic predictions on potential regulatory binding regions within the Red BGC (Iqbal et al., 2012). Here, we provide first experimental evidence by EMSA studies and (q)RT-PCR experiments for the SARP-binding capability to promoter regions (PredP, PredQ) of the Red BGC. These data clearly show that redP and redQ are direct targets of a SARP-driven regulation during Red biosynthesis.

In the literature SARPs often are designated as "pathwayspecific" transcriptional activators, which means that they control

the expression of an individual BCG. However, SARPs indeed can have activating effects on the biosynthesis of different substances. One example occurring in nature is provided by the SARPtype regulator CcaR from S. clavuligerus, which activates the cephamycin gene cluster, where ccaR is part of, as well as the adjacent clavulanic acid gene cluster (Pérez-Llarena et al., 1997). Thus, the term "pathway-specific" is not accurate for SARPs and should better be replaced by "cluster-situated" as suggested previously (Huang et al., 2005; Liu et al., 2013). Our results show that SARPs have the potential to control different BGCs, when present in different producer organisms: The SARP regulator PapR2 activates the transcription of the corresponding pristinamycin gene cluster in S. pristinaespiralis (Mast et al., 2015) but it also affects different antibiotic BGCs, such as Red and plicacetin, when expressed in foreign strains, such as S. lividans or SHP22-7, respectively. Especially, in SHP22-7 papR2 expression significantly improved plicacetin production and as a result also led to an improved production of a so far not further characterized plicacetin derivative. This derivative is dissimilar from the plicacetin isomer since it shows a different retention time in HPLC analysis (∼6.5 min) (**Figure 9**) and a smaller mass of 481 m/z (data not shown). If PapR2-driven plicacetin gene cluster activation is a direct or an indirect regulatory effect is unclear at the moment since no direct interaction of the PapR2 regulator with the plicacetin gene cluster could be demonstrated. Furthermore, no SARP gene is present in the plicacetin gene cluster. However, SARP genes have been found in four other SHP22-7 BGCs, which would allow the possibility that one of these regulators may act as the natural activator of amicetin/plicacetin biosynthesis. A similar untargeted effect

has been observed when the PAS-LuxR-type regulator PimM from S. natalensis was expressed in S. clavuligerus, which led to an improved production of cephamycin, clavulanic acid, and tunicamycins (Martínez-Burgo et al., 2019). For tunicamycin production the regulatory effect upon pimM expression is unclear. pimM expression improved tunicamycin production without affecting tunicamycin gene cluster transcription. Besides, no pathway-specific activator has been identified within the tunicamycin BGC. In this study, the authors speculated that PimM may exert its effect on tunicamycin production, e.g., due to a positive influence on precursor supply (Martínez-Burgo et al., 2019). Furthermore, they found that PimM from S. natalensis shows some similarity to a PimM-like regulator, encoded by a gene of the S. clavuligerus genome (Martínez-Burgo et al., 2019). Thus, also in these analyses the non-native PimM regulator may have occupied the regulatory role of a homologous natural regulator and along this path provoked antibiotic production. Interspecies cross-regulation has also been shown for the LuxRfamily type (LAL) regulator PikD. In Streptomyces. venezuelae PikD regulates the expression of pikromycin. The heterologous expression of the pikD homologous genes rapH and fkbN from Streptomyces hygroscopicus in S. venezuelae increased the production of the antibiotic pikromycin. LAL-family regulators resemble SARP-type regulators in the ATP-binding motif at the N-terminus and in the DNA-binding motif to some extent, which in both regulator types consists of an HTH-motif (Mo and Yoon, 2016).

The fact that SARP genes are widely distributed among actinomycetes, where they occur in many different types of secondary metabolite gene clusters makes them good candidates

to be used as engineering tools for the activation of BCGs. Especially in Streptomyces they are predominant occurring with a ∼100% abundance. Interestingly, there might be a phylogenetic grouping of different SARP regulator types from different actinobacterial genera. If so this would raise the question if there is a co-evolution of certain regulator genes with their corresponding gene clusters. Besides that, the statistical analysis revealed that there is a different abundance of different types of SARP regulators, belonging either to the small SARPs (<400 amino acids) with only HTH and BTAD domains (e.g., ActII-ORF4 or RedD) or the large SARPs, which contain an additional NTPase domain and/or a conserved C-terminal TPR domain of unknown function (∼1000 residues) (Liu et al., 2013). Overall small SARP type regulators are more abundant in actinomycetes than the large ones (**Figure 8**). Besides, more experimental data are available on small SARP-guided regulations, and the unclear function of the additional domains included in the large SARPs likely causes further constraints in the function of the regulatory activity. Thus, we propose that especially representatives from the group of small SARP regulators are good candidates to be used for activation approaches. As outlined above, SARP regulators can bind to recognition sequences, which occur at various positions within the promoter of the target genes. The variety of binding sites may also reflect the diversity of SARP-type activators. Thus, it would be interesting to bioinformatically group the SARP regulators by taking into account their DNA-binding domains, which may provide a better picture of the different SARP subsets.

That SARP expression can lead to the activation of silent gene cluster expression has been shown in a recent study. Here, the aim was to activate some of the more than 20 cryptic gene clusters from Streptomyces sp. MSC090213JE08 (Du et al., 2016). In this study the authors combined an OSMAC approach with the expression of several native SARP genes from Streptomyces sp. MSC090213JE08. Thereby, four of the seven generated recombinant SARP expression strains produced nine metabolites that were hardly detected in the control strains. Expression of one of the SARP genes (SARP-7) in Streptomyces sp. MSC090213JE08 led to the production of the novel polyene-like substance ishigamide (Du et al., 2016). This study successfully showed the potency of a SARP-guided silent gene cluster activation. However, the drawback of this approach is that several conditions (number of different culture media) need to be tested and a set of genetic manipulations (cloning of each individual SARP gene) has to be done in order to provoke cluster activation. Further experimental setup is then linked to untargeted laborious compound purifications and analytics. Indeed, this is also the main problem of several other efforts to activate secondary metabolite synthesis in actinomycetes as outlined above. Thus, based on our gained knowledge we propose a targeted SARP-guided strategy for the activation of BGCs in actinomycetes. Our strategy involves (1) prioritization of strains with SARP genes and SARP binding motifs in the BGC. In our experimental setup we focus on SARP genes that encode for predicted proteins with high similarity to PapR2 (>55% amino acid sequence similarity), as well as the occurrence of a PapR2 consensus motif within the promoter region(s) of the SARPgene containing BGCs, (2) introduction of a SARP-expression construct [in our approach this means heterologous expression of the PapR2 regulator with the help of the pGM190/papR2 and/or pRM4/papR2 expression construct(s)], (3) comparative biological and chemical analyses of SARP-activated expression samples with samples from non-manipulated strains. In our study we focus on PapR2 as the activator brick, however, of course any other type of SARP regulator with a known consensus sequence might be used as the basis for such an activation approach. For sure there might be limitations in such an activation strategy, e.g., transcriptional activation may fail due to SARP-specificity reasons or a lack of a broader set of well-characterized SARP regulators to be tested as candidate elicitors. However, here we disclose a screening idea or a kind of dragnet investigation, which does not aim to cover all possible SARP-regulated clusters but highlights the most probable ones to be activated upon SARP expression. Our strategy has two main advantages: (1) In contrast to completely unspecific cluster activation efforts, such as addition of general elicitors, co-culture approaches or the OSMAC strategy, our approach is more targeted, as it focuses only on a defined set of BGCs, namely those ones that are bioinformatically predicted to be under SARP control. Applying combinatory bioinformatics, such as antiSMASH and PatScan allows to directly identify the associated BGC. In addition, the detection of the activated compound is more straightforward if the outline of the structure can be deduced from the cluster sequence. (2) In contrast to activation strategies that are absolutely specific for the respective BGC, e.g., heterologous expression of the BGC, introduction of artificial promoters in front of the BGC, or the manipulation of clustersituated regulators, no major effort to manipulate the genome is necessary with our procedure since it only involves cloning of one SARP gene-containing expression construct into the respective strain(s) of interest. These major benefits make the activation of BGCs by SARPs a promising strategy to be applied on putative antibiotic producers.

#### DATA AVAILABILITY STATEMENT

The complete genome sequence of S. lividans T7 has been deposited at DDBJ/ENA/GenBank under the accession number ACEY00000000. The main genome scaffold sequence of SHP22-7 has been deposited at DDBJ/ENA/GenBank under the accession number QXMM00000000. Raw sequencing data are available under SRA accession number PRJNA489221.

# AUTHOR CONTRIBUTIONS

YM generated the strains SLpGM190 and SLpapR2-OE, carried out the bioinformatic analyses, and designed, supervised, and coordinated the study. KB performed the statistical analyses. JK created the vector pRM4/papR2 while the vector was inserted in SHP22-7 by IH. IH tested the bioactivity. JK performed all qualitative and quantitative transcriptional analyses. AK and IH performed the HPLC analyses. YM, JK, IH, and KB wrote the manuscript.

#### FUNDING

We gratefully acknowledge the funding received from the BMBF German–Indonesian Cooperation project NAbaUnAk. The study was supported by a DAAD scholarship given to JK and a scholarship from the Indonesian Ministry for Research and Technology for IH.

#### ACKNOWLEDGMENTS

fmicb-11-00225 February 14, 2020 Time: 19:22 # 15

We thank Vera Kübler for performing the EMSAs. We also thank Regina Ort-Winklbauer from the University of Tübingen for

#### REFERENCES


excellent technical assistance. We acknowledge help from Harald Gross, Nico Ortlieb, and Dorothee Wistuba from the University of Tübingen for High-Resolution Mass Spectrometric analysis. Finally, we thank Puspita Lisdiyanti and Shanti Ratna Komala from the Indonesian Institute of Sciences for providing us with the Indonesian strain isolate SHP 22-7.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2020.00225/full#supplementary-material

sp. strain SHP22-7, a new species isolated from mangrove of Enggano island. Indonesia. Microbiol. Resour. Announc. 7, 1–3. doi: 10.1128/MRA.013 17-18


Actinobacteria. Front. Microbiol. accepted 10:3120. doi: 10.3389/fmicb.2019. 03120



Zhu, H., Sandiford, S. K., and van Wezel, G. P. (2014). Triggers and cues that activate antibiotic production by actinomycetes. J. Ind. Microbiol. Biotechnol. 41, 371–386. doi: 10.1007/s10295-013-1309-z

**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The reviewer JA declared a past co-authorship with one of the authors AK to the handling Editor.

Copyright © 2020 Krause, Handayani, Blin, Kulik and Mast. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The Streptomyces coelicolor Small ORF trpM Stimulates Growth and Morphological Development and Exerts Opposite Effects on Actinorhodin and Calcium-Dependent Antibiotic Production

#### Edited by:

Yvonne Mast, German Collection of Microorganisms and Cell Cultures GmbH (DSMZ), Germany

#### Reviewed by:

Marie-Joelle Virolle, Centre National de la Recherche Scientifique (CNRS), France Angel Manteca, Universidad de Oviedo Mieres, Spain

> \*Correspondence: Alberto Vassallo alberto.vassallo@unipa.it

#### Specialty section:

This article was submitted to Antimicrobials, Resistance, and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 16 October 2019 Accepted: 30 January 2020 Published: 19 February 2020

#### Citation:

Vassallo A, Palazzotto E, Renzone G, Botta L, Faddetta T, Scaloni A, Puglia AM and Gallo G (2020) The Streptomyces coelicolor Small ORF trpM Stimulates Growth and Morphological Development and Exerts Opposite Effects on Actinorhodin and Calcium-Dependent Antibiotic Production. Front. Microbiol. 11:224. doi: 10.3389/fmicb.2020.00224 Alberto Vassallo1,2 \*, Emilia Palazzotto<sup>3</sup> , Giovanni Renzone<sup>4</sup> , Luigi Botta<sup>5</sup> , Teresa Faddetta<sup>1</sup> , Andrea Scaloni<sup>4</sup> , Anna Maria Puglia<sup>1</sup> and Giuseppe Gallo<sup>1</sup>

<sup>1</sup> Laboratory of Molecular Microbiology and Biotechnology, STEBICEF Department, University of Palermo, Palermo, Italy, <sup>2</sup> Laboratory of Microbial and Molecular Evolution, Department of Biology, University of Florence, Sesto Fiorentino, Italy, <sup>3</sup> Novo Nordisk Foundation Center for Biosustainability, Technical University of Denmark, Lyngby, Denmark, <sup>4</sup> Proteomic and Mass Spectrometry Laboratory, ISPAAM, National Research Council, Naples, Italy, <sup>5</sup> Dipartimento di Ingegneria, Università di Palermo, Palermo, Italy

In actinomycetes, antibiotic production is often associated with a morpho-physiological differentiation program that is regulated by complex molecular and metabolic networks. Many aspects of these regulatory circuits have been already elucidated and many others still deserve further investigations. In this regard, the possible role of many small open reading frames (smORFs) in actinomycete morpho-physiological differentiation is still elusive. In Streptomyces coelicolor, inactivation of the smORF trpM (SCO2038) – whose product modulates L-tryptophan biosynthesis – impairs production of antibiotics and morphological differentiation. Indeed, it was demonstrated that TrpM is able to interact with PepA (SCO2179), a putative cytosol aminopeptidase playing a key role in antibiotic production and sporulation. In this work, a S. coelicolor trpM knock-in (Sco-trpMKI) mutant strain was generated by cloning trpM into overexpressing vector to further investigate the role of trpM in actinomycete growth and morpho-physiological differentiation. Results highlighted that trpM: (i) stimulates growth and actinorhodin (ACT) production; (ii) decreases calcium-dependent antibiotic (CDA) production; (iii) has no effect on undecylprodigiosin production. Metabolic pathways influenced by trpM knockin were investigated by combining two-difference in gel electrophoresis/nanoliquid chromatography coupled to electrospray linear ion trap tandem mass spectrometry (2D-DIGE/nanoLC-ESI-LIT-MS/MS) and by LC-ESI-MS/MS procedures, respectively. These analyses demonstrated that over-expression of trpM causes an over-representation of factors involved in protein synthesis and nucleotide metabolism as well as a

down-representation of proteins involved in central carbon and amino acid metabolism. At the metabolic level, this corresponded to a differential accumulation pattern of different amino acids – including aromatic ones but tryptophan – and central carbon intermediates. PepA was also down-represented in Sco-trpMKI. The latter was produced as recombinant His-tagged protein and was originally proven having the predicted aminopeptidase activity. Altogether, these results highlight the stimulatory effect of trpM in S. coelicolor growth and ACT biosynthesis, which are elicited through the modulation of various metabolic pathways and PepA representation, further confirming the complexity of regulatory networks that control antibiotic production in actinomycetes.

Keywords: small open reading frame, trpM, actinorhodin production, Streptomyces coelicolor, cytosol aminopeptidase, calcium-dependent antibiotic, primary and secondary metabolism

#### INTRODUCTION

Bacteria belonging to the phylum Actinomycetales are widely recognized as a very prolific source of biologically active natural compounds, such as antibiotics, immunosuppressants, and herbicides. As reviewed by Palazzotto et al. (2019), different approaches have been adopted to increase production of these molecules in actinomycetes, ranging from metabolic engineering strategies – driving the nutrient catabolism toward an increased supply of precursors – to the modification of the expression of regulators taking directly part in natural product biosynthesis. Besides being studied as a model for investigation of bacterial differentiation, the filamentous Gram-positive bacterium Streptomyces coelicolor is a model organism for the study of polyketide antibiotic production in Actinomycetales. S. coelicolor produces different biologically active metabolites whose biosynthesis has been widely documented – including the blue-pigmented Type II polyketide actinorhodin (ACT), the redpigmented alkaloid undecylprodigiosin (RED), and the cyclic lipoundecapeptide calcium-dependent antibiotic (CDA) (Liu et al., 2013). In S. coelicolor, many aspects of regulatory circuits controlling antibiotic production have been already elucidated although many others still deserve further investigations. As an example, Xu et al. (2019) demonstrated that ACT biosynthesis is affected by a number of genes spread along the chromosome that have been never previously associated with production of ACT. Also, supplementation of specific nutriments exerts a control on S. coelicolor morphological and physiological differentiation: as an example, Palazzotto et al. (2015) demonstrated that the supplementation of L-tryptophan (L-Trp) promotes sporulation and stimulates the production of CDA – that contains proteinogenic and non-proteinogenic amino acids, including L-Trp and D-Trp – and the production of ACT – that does not contain any Trp in its structure. In this regard, it is noteworthy that in S. coelicolor the expression of genes involved in L-Trp biosynthesis (i.e., trp genes) is not repressed by Trp supplementation (Hu et al., 1999; Palazzotto et al., 2016). In addition, trp genes are organized either as gene clusters (i.e., trpC1MBA and trpC2D2GE2) or single genes (i.e., trpE3, trpE1 and priA/trpF) spread in the genome with trpC2D2GE2 localized within the CDA biosynthetic gene cluster. So, this gene organization probably allows S. coelicolor to express a subset of trp genes independently from the others and in response to specific metabolic needs (Xie et al., 2003).

trpM (SCO2038) is part of the trpC1MBA locus (Hu et al., 1999; Palazzotto et al., 2016). Because of its small size (just 64 codons), trpM and the corresponding protein can be listed as a small open reading frame (smORF) and a small open reading frame-encoded protein (SEP), respectively. smORFs and SEPs have been extensively ignored so far and discovered mostly serendipitously. Nonetheless, they are known to take part in different important cell processes in bacteria – e.g., spore formation, cell division, membrane transport, regulation of enzymatic activities and signal transduction – and nowadays they are gaining more and more attention (Storz et al., 2014; Chu et al., 2015; Chugunova et al., 2018; Delcourt et al., 2018). It was previously demonstrated that TrpM is involved in L-Trp biosynthesis (Palazzotto et al., 2016). Indeed, a trpM-knockout mutant strain showed an impaired growth on minimal medium but a normal growth was restored upon addition of L-Trp or its precursors (i.e., L-serine and indole) to the medium. These results were corroborated through a proteomic investigation that compared the trpMknockout mutant and wild type strains, showing that the list of differentially abundant proteins included some components directly involved in L-Trp biosynthesis (Palazzotto et al., 2016). Moreover, the inactivation of trpM significantly affected ACT biosynthesis, since the trpM-knockout mutant produced 10-fold less ACT than the WT strain. In addition, a pull-down assay using immobilized His-tagged TrpM allowed to identify the putative cytosol L-leucine-aminopeptidase PepA belonging to M17 metalloprotease family (SCO2179), the ribosomal protein S1 (SCO1998) and the ribosomal protein S2 (SCO5624) as probable interacting proteins of this SEP. TrpM and PepA interaction was also demonstrated by bacterial two-hybrid assay (Palazzotto et al., 2016).

In this work, a trpM knock-in mutant strain was obtained to characterize the role of TrpM in both morphological differentiation and antibiotic biosynthesis of S. coelicolor; results from different comparative experiments on this mutant and wild type strains allowed us to suggest a possible molecular model explaining its mode of action.

# MATERIALS AND METHODS

fmicb-11-00224 February 17, 2020 Time: 16:57 # 3

#### Bacterial Strains, Plasmids, and Cultivation Conditions

All Streptomyces coelicolor strains and plasmids used in this work are listed in **Table 1**. Besides, Escherichia coli TOP10 (Invitrogen), E. coli S17-1 (Simon et al., 1983), and E. coli BL21-AI (Invitrogen) were used as described below.

The Escherichia coli strains were cultivated in LB medium (Sambrook and Russell, 2001) supplemented with apramycin (50 µg/mL) and ampicillin (100 µg/mL) in the case of strains carrying pIJ8600/pIJ8600:trpM and pRSET-B/pRSET-B:pepA, respectively, at 37◦C and 200 rpm. For S. coelicolor cultures, minimal medium [NaNO<sup>3</sup> (1 g/L), MgSO47H2O (0.5 g/L), KCl (0.5 g/L), KH2PO<sup>4</sup> (1 g/L), glucose (10 g/L), trace element solution (1 mL/L), pH 7 as adjusted before sterilization] was used; trace element solution contained FeSO47H2O (1 g/100 mL), ZnCl<sup>2</sup> (1 g/100 mL), and biotin (0.1 g/100 mL). Glucose and trace element solution were added upon sterilization, and solid media were prepared adding bacto agar (15 g/L) to the recipes reported. If not otherwise indicated, 1.5 × 10<sup>7</sup> spores of S. coelicolor strains were spread on solid minimal medium, and incubated at 30◦C, for 7 days. In the case of cultures used for RNA, protein and metabolite extraction, a disc of cellophane (Cellophane Membrane Backing, Bio-Rad, United States) was placed on the surface of medium to facilitate mycelium harvesting.

#### Construction of Recombinant Streptomyces coelicolor Strains

DNA manipulation, purification, ligation, restriction analysis, gel electrophoresis and transformation of E. coli were performed according to standard techniques (Sambrook and Russell, 2001). trpM and pepA were amplified from genomic DNA of S. coelicolor M145 using the couple of primers trpM\_exp\_F/trpM\_exp\_R and pepA\_exp\_F/pepA\_exp\_R, respectively, which are reported in **Table 2**. Taq DNA Polymerase Recombinant (Invitrogen) was used in both cases, and standard conditions indicated by the manufacturer were adopted. Purified PCR products containing trpM and pepA genes were digested

TABLE 1 | List of strains and plasmids used in this work.


TABLE 2 | List of primers used in this work.


\*Underlined sequence corresponds to restriction sites.

with restriction enzymes (NdeI and BamHI in case of trpM, while BamHI and HindIII were used for pepA), and ligated into previously restricted pIJ8600 and pRSET-B vectors, respectively (**Tables 1**, **2**). Ligation mix was then used for transformation of One Shot Chemically Competent E. coli TOP10 cells (Invitrogen). The identity of all DNA fragments amplified by PCR was confirmed by DNA sequencing. The pIJ8600 and pIJ8600:trpM plasmids were used to transform chemically competent E. coli S17-1 cells, which were used as donor ones to transform S. coelicolor M145 through a conjugation-based protocol (Kieser et al., 2000). Integration of either pIJ8600 and pIJ8600:trpM plasmids in the attB8C31 site of the chromosomal DNA of S. coelicolor was verified through Southern blotting. The 1480 bp-long restriction fragment derived from digestion of pIJ8600 with StyI was labeled using the Dig High Prime DNA Labeling and Detection Starter kit I (Roche), and used as probe. Chromosomal DNA of S. coelicolor strains was digested with BamHI and blotting was performed according to the manufacturer's instructions and Sambrook and Russell (2001).

#### Transcriptional Analysis

RNA was extracted using the Illustra RNAspin Midi RNA Isolation Kit (GE Healthcare, United States) and according to the protocol provided by the manufacturer. A two-step protocol was applied for qRT-PCR and the High-Capacity cDNA Reverse Transcription Kit with RNase Inhibitor (Applied Biosystems) was firstly used to convert RNA in cDNA. Thus, Power SYBR Green PCR Master Mix (Applied Biosystems) was used for relative RNA quantification. To this purpose, primers reported in **Table 2** were used, and hrdB was chosen as internal standard. Two biological replicates and three technical replicates were used for each condition.

#### Scanning Electron Microscope Observations

Agar block samples (1 cm × 1 cm × 0.5 cm) were cut from agar-medium cultures, washed three times with phosphatebuffered saline (PBS) for 3 min, and then fixed using 4% v/v glutaraldehyde for 5 min, under gentle agitation. Upon removing of glutaraldehyde solution, samples were washed with 15% v/v ethanol for 3 min, and incubated at 65◦C until obtaining a thin slice. Samples were sputter-coated with gold to avoid electrostatic charging under the electron beam and examined by Scanning Electronic Microscopy (SEM) (Phenom ProX, PhenomWorld).

### Bacterial Growth Kinetics

fmicb-11-00224 February 17, 2020 Time: 16:57 # 4

An amount of '10<sup>8</sup> spores of S. coelicolor M145 strains carrying pIJ8600:trpM (Sco-trpMKI) and pIJ8600 (Sco-EV) were inoculated in 25 mL of J medium (Kieser et al., 2000), respectively, using 250 mL baffled flasks, and incubated for 30 h (30◦C, 200 rpm in an orbital shaker). Cultures were centrifuged at 3000 × g for 15 min, and pellets were washed twice with sterile water. Finally, pellets were resuspended in 50 mL of sterile water, and 3 mL of the obtained suspensions were inoculated in 200 mL of minimal medium (MM), using 1 L baffled flasks, and incubated for 4 days (30◦C, 200 rpm in an orbital shaker). Every 12 h, three aliquots of 1 mL were sampled for dry weight determination. Thus, they were centrifuged, decanted, dried at 65◦C for 24 h, and finally weighed.

#### Spore Counting and Antibiotic Assays

An amount of '1.5 × 10<sup>7</sup> spores of Sco-trpMKI and Sco-EV were spread on six MM agar plates, and incubated at 30◦C, for 7 days. Three of them were used for spore harvesting and a subsequent serial dilution counting, while the others were used for determination of ACT and undecylprodigiosin production. Antibiotic assays were performed as described by Kieser et al. (2000). In particular, total ACT and undecylprodigiosin (RED) were sequentially extracted by treating cultivations with a 1 N KOH aqueous solution and a 0.5 N HCl/methanol solution, respectively. Insoluble matter, resulting as pellet after 1 N KOH treatment and centrifugation (13000 x g, for 15 min), was washed twice with 1 M Tris–HCl, pH 7.5, before treatment with a solution of 0.5 N HCl/methanol 50:50 v/v (Scaffaro et al., 2017). The amount of ACT and RED, whose identity was confirmed by visible absorption spectra (Horinouchi and Beppu, 1984), was spectrophotometrically evaluated at 640 and 530 nm as the mean value measured for three independent cultivations. Concentrations of ACT and RED were determined using molar extinction coefficients of pure compounds. They are ε<sup>640</sup> = 25320 M−<sup>1</sup> x cm−<sup>1</sup> and ε<sup>530</sup> = 100500 M−<sup>1</sup> x cm−<sup>1</sup> for ACT and RED, respectively. CDA production was evaluated according to Kieser et al. (2000) and Palazzotto et al. (2015).

# Proteomic Analysis

Proteins were extracted from the mycelium of Sco-trpMKI and Sco-EV according to Puglia et al. (1995), and proteomic investigation was conducted by 2D-DIGE as described by Palazzotto et al. (2016), except that 1.3 was chosen as threshold value for the identification of differentially represented proteins. ANOVA test was used to assess the statistical significance of protein abundance fold change, using a p ≤ 0.05 to consider differentially represented proteins. Proteins were extracted from three biological replicas for each strain and 2D-DIGE procedure was conducted with two technical replicas with a total of six replicas per strain. Identification of protein material present in each differentially represented spot was performed by nanoLC-ESI-LIT-MS/MS analysis, which was performed with an LTQ XL mass spectrometer (Thermo Fisher Scientific, United States) equipped with a Proxeon nanospray source connected to an Easy nanoLC (Thermo Fisher Scientific, United States) (Lirussi et al., 2012). Peptide mixtures were resolved on an Easy C18 column (10 – 0.075 mm, 3 µm) (Thermo Fisher Scientific, United States) as previously reported (Palazzotto et al., 2016).

MASCOT search engine version 2.2.06 (Matrix Science, United Kingdom) was used to identify protein spots from an updated NCBI non-redundant database (downloaded January 2018) also containing protein sequences for S. coelicolor A3(2), using mass spectrometric data. Database searching was performed selecting trypsin as proteolytic enzyme, a missed cleavages maximum value of 2, Cys carbamidomethylation as fixed modification, Met oxidation and N-terminal Gln conversion to pyro-Glu as variable modifications, respectively. Candidates with at least 2 assigned peptides with an individual peptide expectation value less than 0.05, which corresponds to a confidence level for peptide attribution greater than 95%, were further evaluated by the comparison with their calculated mass and pI values, using the experimental data obtained from electrophoresis. Protein assignment was always associated with manual verification. Finally, in case of multiple protein identifications, unambiguous protein identity was assigned according to an emPAI ratio criterion calculated between the two most abundant protein species (i.e., emPAI 1st/emPAI 2nd > 1.50) (Shinoda et al., 2010). The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (Perez-Riverol et al., 2019) partner repository with the dataset identifier PXD015937.

# Metabolic Profile Determination

Metabolic profiles of Sco-trpMKI and Sco-EV were investigated through LC-ESI-MS/MS analysis in multiple reaction monitoring (MRM) mode coupled with protein precipitation and extraction of metabolites using organic solvents. All procedures reported below were carried out at Centro di Ingegneria Genetica (CEINGE) – Biotecnologie avanzate (Naples, Italy). Samples were lysed in 250 µL of a buffer containing 10 mM NH4CO3, 10 mM NaF, 7 M urea, 75 mM NaCl, and the suspension thus obtained was homogenized. Cellular lysates were fivefold diluted with a cold solution of acetonitrile/methanol (50:50 v/v) containing 0.1% v/v acetic acid. Samples were sonicated for 10 min, and centrifuged at 12000 rpm for 10 min, at 4 ◦C. Supernatants were dried and resuspended in 500 µL of a solution of acetonitrile/methanol/water (40:40:20 v/v/v) containing 0.1% v/v acetic acid. Upon centrifugation at 12000 rpm for 10 min, supernatants were filtered using 0.22 µm centrifugal filters and analyzed by LC-ESI-MS/MS. Quantitative analysis was performed by the external calibration method. Standard solutions were prepared dissolving 1 mg of each analyte in 1 mL of 5% v/v acetonitrile (1000 ppm), and serial dilutions of them were used for the generation of calibration curves. 1 µL of supernatants obtained from

samples were analyzed with a 6420 Triple Quadrupole System (SCIEX, United States) coupled to a HPLC 1100 Series Binary Pump (Agilent, United States). Analytes were separated using a Kinetex 5 µm C18 analytical column (100 mm × 2.1 mm) (Phenomenex). The mobile phase was generated by mixing eluent A (0.1% v/v acetic acid containing 3 mM ammonium acetate) and eluent B (50% v/v acetonitrile, 50% v/v 2 propanol and 0.1% v/v acetic acid), at a flow rate of 0.3 mL/min. The elution gradient ranged from 5% to 95% of eluent B in 7 min. Tandem mass spectrometry was performed using a turbo ion spray source operating in negative mode and MRM mode was used for the selected analytes. Metabolites were tuned for ionization polarity, optimal declustering potential, precursor and daughter product ions, and collision energy. Metabolites were extracted from two biological replicates of both strains and three technical replicates of each one were processed. Student's t-test was applied to verify the statistical significance of metabolite abundance fold change, using a p ≤ 0.05 to consider differentially represented metabolites.

### Over-Expression of PepA in Escherichia coli

One Shot Chemically Competent E. coli BL21-AI cells (Invitrogen) were transformed with either pRSET-B:pepA or pRSET-B, with the former used to obtain a 6xHis-tagged recombinant PepA and the latter to have a control protein extract. In particular, an amount of 50 mL of LB medium was inoculated with 2.5 mL of an overnight-grown culture of E. coli BL21-AI carrying pRSET-B:pepA. The culture was incubated (37◦C, at 200 rpm in an orbital shaker) until OD<sup>600</sup> reached a value of 0.8. Then, a sterile solution of the inducer L-arabinose was added to have a final concentration of 0.1% w/v, and induction of pepA over-expression was performed at 30◦C, for 3 h. E. coli BL21-AI carrying pRSET-B was used as negative control, and aliquots from all cultures were collected as further control before adding the inducer. Cultures were centrifuged, and the pellet resuspended in 2 mL of 50 mM Tris–HCl, pH 8. Then, suspensions were sonicated with 3 pulses (10 sec each and 4 as output control) in an ice bath and centrifuged (30 min, 7000 × g, at 4◦C). Supernatants containing the water-soluble (WS) protein fraction of the cell lysate were directly used for leucine aminopeptidase assay measurements. Bradford reagent was used to determine WS protein concentration (Bradford, 1976).

#### Leucine Aminopeptidase Assay

An aliquot of 400 µg of WS proteins was added to the reaction mix (0.25 mM L-leucine-p-nitroanilide, 1 mM CoCl<sup>2</sup> or MnCl2, 50 mM Tris–HCl, pH 8) in a final volume of 500 µL. After incubation at 60◦C for 10 min, a volume of 10% v/v acetic acid was added, and the solution was incubated at 100◦C, for 5 min. Samples were finally centrifuged at 13000 × g, for 5 min, at 4◦C, and the absorbance of supernatants was measured at 405 nm (Kuo et al., 2003; Song et al., 2013). This assay was performed with three different technical replicates.

# RESULTS

# Construction and Characterization of a trpM Knock-in Mutant Strain

For the construction of a S. coelicolor trpM knock-in mutant strain (thereof named Sco-trpMKI), trpM was PCR-amplified and cloned within the multiple cloning site of the integrative expression vector pIJ8600 under the control of the thiostreptoninducible promoter (PtipA) (Murakami et al., 1989; Kieser et al., 2000). This plasmid was delivered to S. coelicolor through interspecific conjugation, and the correct integration in the attB 8C31 site was verified by PCR, sequencing and Southern blotting. Analogously, a strain having the empty vector pIJ8600 integrated in its genome (indicated as Sco-EV) was constructed and used as control during all the following experiments. Since qRT-PCR showed that the expression of trpM in Sco-trpMKI was already twofold greater than that in Sco-EV (**Supplementary Figure S1**), thiostrepton was not added to the growth medium as inducer to avoid the introduction of any possible negative effects on antibiotic production and on protein expression pattern as described by Wang et al. (2017) and Chiu et al. (1999), respectively.

During solid-medium growth, Sco-trpMKI had a faster growth in comparison to Sco-EV since the former showed an evident substrate mycelium after 24 h, which was not present in the latter (**Supplementary Figures S2A,B**, respectively). After 48 h of growth, both strains showed developed hyphae with few and faint septal constrictions that were more evident in Sco-trpMKI (**Supplementary Figures S2C,D**). Spore chains were clearly visible in the 72 h-old Sco-trpMKI (**Supplementary Figure S2E**) as well as quite lagging in Sco-EV (**Supplementary Figure S2F**). Nonetheless, after 120 h of growth both ScotrpMKI and Sco-EV had a comparable phenotype with no major morphological differences (**Supplementary Figures S2G,H**). This finding is in good agreement with the faster and higher biomass-yielding growth kinetics in liquid medium cultures, in which Sco-trpMKI had a more rapid and abundant production of biomass up to 72 h (**Figure 1**). Accordingly, a higher alkalinization of the growth medium was observed in

Sco-trpMKI that was attributed to conversion of NaNO<sup>3</sup> (the sole nitrogen source in the growth medium) into ammonia (Fischer et al., 2012). This suggested that Sco-trpMKI had an increased utilization of nitrogen consistently with its higher growth rate, further confirming that the expression of the extra copy of trpM influenced the growth rate accelerating morphological development progress. However, trpM expression did not affect the final amount of spores and undecylprodigiosin (**Figure 2**), but exerted a positive effect on the production of the polyketide antibiotic ACT, both in liquid and on solid media, with an amount increment ranging from 2.5 to 3 times in comparison to Sco-EV (**Figure 3**). On the other hand, the expression of the extra copy of trpM had a negative effect on CDA, with a more than fivefold decrement of CDA production compared to Sco-EV (**Figure 4**).

#### Proteomic and Metabolic Profiles of the trpM Knock-in Mutant Strain

A 2D-DIGE proteomic analysis (**Supplementary Figure S3**) coupled with nanoLC-ESI-LIT-MS/MS was performed to infer metabolic pathways influenced by the expression of trpM in Sco-trpMKI. This investigation led to the identification of 40 differentially represented proteins, with 23 and 17 over- and down-represented ones in Sco-trpMKI compared to Sco-EV, respectively (**Table 3**, **Supplementary Figure S4A** and **Supplementary Tables S1**, **S2**). According to a KEGG orthology-based classification, most of them are involved in

translation, carbon metabolism and folding-sorting-degradation of proteins (**Table 3**). Above all, the over-representation of proteins involved in protein synthesis, DNA synthesis and energy metabolism was consistent with the improved growth of Sco-trpMKI. On the other hand, the effect of trpM expression on growth was associated with a downrepresentation of several glycolytic enzymes as well as with the simultaneous over- and down-representation of SCO5042 (fumarate hydratase) and SCO4827 (malate dehydrogenase),

#### TABLE 3 | Differentially represented proteins in Sco-trpMKI in comparison to Sco-EV.


<sup>1</sup>Positive and negative values stand for up- and down-representation in Sco-trpMKI, respectively. <sup>2</sup>Functional classification is based on KEGG orthology.

respectively, which take part into tricarboxylic acid (TCA) cycle. Interestingly, the serine hydroxymethyltransferase (GlyA) involved in interconversion of serine and glycine was downrepresented in the Sco-trpMKI strain, thus revealing that TrpM can affect the biosynthesis of serine, which is in turn a direct precursor of Trp.

Notwithstanding these proteomic determinations, similar intracellular levels for serine and Trp were ascertained in Sco-trpMKI and Sco-EV strains by quantitative metabolite measurements (**Table 4** and **Supplementary Tables S3**, **S4**). Indeed, this analysis showed a down-representation of tyrosine and phenylalanine (**Table 4** and **Supplementary Table S3**) competing with Trp for metabolic intermediates. In general, most differentially abundant compounds were related to: (i) amino acid, (ii) carbon metabolism intermediates; (iii) nucleotide metabolism intermediates (**Supplementary Figure S4B**); most of these compounds were down-represented in Sco-trpMKI strain. Exceptions were histidine, inosine, AMP, methylmalonate and succinate, which were over-represented in the Sco-trpMKI in the respect of Sco-EV (**Table 4** and **Supplementary Table S3**). This observation, together with the down-representation of compounds belonging to carbon metabolism pathways (i.e., glycolysis and TCA cycle), depicted a metabolic profile that was in good agreement with that deriving from proteomic results (**Supplementary Figure S5**).

#### Enzymatic Activity of PepA

Different experimental data suggested functional relations between PepA and TrpM. In particular, Palazzotto et al. (2016) identified PepA as a putative TrpM-interacting protein through pull-down experiments, and then confirmed this protein-protein interaction in vitro by an adenylate cyclase-based bacterial twohybrid assay. In this work, PepA was down-represented in ScotrpMKI in comparison with Sco-EV (**Table 3** and **Supplementary Tables S1**, **S2**); on the contrary, a higher expression of this gene was observed by qRT-PCR (**Supplementary Figure S6**). This finding suggested the presence of putative post-transcriptional and/or post-translational regulatory mechanisms controlling the final representation of the active protein, as already globally described elsewhere (Jayapal et al., 2008; Jeong et al., 2016).

As reported in the UniProtKB database, the protein product of pepA is a putative cytosol aminopeptidase (entry identifier Q9S2Q7) belonging to the peptidase M17 family. In order to have an experimental confirmation of the PepA function, pepA was cloned in the expression vector pRSET-B and over-expressed in E. coli as His-tagged protein. Purification of His-tagged PepA

TABLE 4 | Differentially represented metabolites in Sco-trpMKI in comparison to Sco-EV. Compound Differential abundance<sup>1</sup> P (Student's t-test) Functional classification<sup>2</sup> Aspartic acid −3.23 2.76 × 10−<sup>2</sup> Amino acid metabolism Arginine −1.37 4.65 × 10−<sup>2</sup> Phenylalanine −1.73 3.31 × 10−<sup>2</sup> Tyrosine −1.53 2.45 × 10−<sup>6</sup> Histidine NA NA Glutamic acid −1.37 3.86 × 10−<sup>2</sup> Glutamine −1.69 1.26 × 10−<sup>2</sup> Glucosamine 6-phosphate −2.26 2.81 × 10−<sup>2</sup> Glycolysis; Pentose phosphate pathway 3-phosphoglycerate −1.91 4.79 × 10−<sup>2</sup> Fructose 6-phosphate −1.67 3.43 × 10−<sup>2</sup> Glucose 6-phosphate −3.13 4.20 × 10−<sup>3</sup> 6-phosphogluconate −1.84 4.62 × 10−<sup>2</sup> 2-oxoglutarate −1.63 2.78 × 10−<sup>2</sup> TCA cycle Fumarate NA NA Citrate −1.36 4.20 × 10−<sup>2</sup> Methylmalonate 1.70 4.91 × 10−<sup>2</sup> Succinate 1.72 4.59 × 10−<sup>2</sup> Malate −1.90 4.21 × 10−<sup>2</sup> Oxalate NA NA AMP NA NA Nucleotide metabolism CMP −2.97 2.33 × 10−<sup>2</sup> UMP −3.37 4.64 × 10−<sup>3</sup> Uracil −2.15 4.91 × 10−<sup>2</sup> GMP NA NA Inosine 2.04 4.54 × 10−<sup>2</sup> Guanosine −1.34 1.41 × 10−<sup>2</sup>

<sup>1</sup>Positive and negative values stand for up- and down-representation in Sco-trpMKI, respectively. NA, not applicable. <sup>2</sup>Functional classification is based on KEGG orthology.

under native conditions was not successful, because of the low amount of the soluble protein, due to the undesired formation of inclusion bodies. Thus, WS protein fractions obtained from the pepA-overexpressing E. coli strain and from the strain carrying the vector pRSET-B were used for comparative enzymatic assay experiments, using the latter as control. In agreement with Song et al. (2013), PepA activity was not revealed in the protein extraction in presence of CoCl2, while the use of MnCl<sup>2</sup> notably highlighted for the first time that PepA exerted the predicted aminopeptidase activity (**Figure 5**).

#### DISCUSSION

The results described in this work revealed that trpM boosts the morpho-physiological differentiation of S. coelicolor. This positive effect was inferred from the accumulation of biomass in liquid cultures and by the mycelium morphology on solid growth medium observed in Sco-trpMKI in comparison to Sco-EV. In agreement, SCO0409 product – known as spore-associated protein A or SapA – is deregulated as shown by differential proteomic analysis. This protein, along with SapB, is expressed in aerial hyphae and is localized on the external surface of spores, contributing to the shaping of spores and to their hydrophobicity (Guijarro et al., 1988; Willey et al., 1991).

The synthesis of the three major secondary metabolites of S. coelicolor – i.e., ACT, undecylprodigiosin and CDA – was studied in the trpM knock-in strain. It revealed a stimulatory effect only on ACT production. This finding is consistent with the results described by Palazzotto et al. (2016) concerning a trpM-knockout mutant strain, whose production of ACT was impaired upon cultivation on minimal medium. These results suggest that trpM expression could influence ACT biosynthesis through an indirect regulation of central carbon metabolism. Indeed, the biosynthesis of antibiotics in S. coelicolor is greatly affected by precursor supply. In the trpM knock-in strain, the relative abundances of two TCA cycle enzymes – i.e., malate dehydrogenase (SCO4827) and fumarate hydratase class II (SCO5042) – suggested a differential abundance of TCA intermediates. Accordingly, the metabolite profile analysis revealed the accumulation of succinate, which can be related also with the accumulation of methylmalonate, a metabolic intermediate of the branched-chain amino acid catabolism providing precursors for the biosynthesis of type II polyketide antibiotics, like ACT (Stirrett et al., 2009). However, the exact metabolic routes possibly leading to an increment of malonyl-CoA, which is an extending unit precursor of the ACT backbone (Revill et al., 1995; Hopwood, 1997; Rawlings, 1999) were not inferred from the proteomic and metabolic profiles of Sco-trpMKI strain.

On the other hand, the decrement of CDA production may be related to the altered amounts of amino acids observed in Sco-trpMKI strain. Indeed, the biosynthesis of CDA is dependent on amino acid availability. In particular, the peptide moiety of CDA includes residues of serine, threonine, tryptophan, glutamic acid, hydroxyphenylglycine, aspartic acid, glycine, and asparagine (Hojati et al., 2002). Although tyrosine is not part of CDA backbone directly, it is known that CDA biosynthesis intersects tyrosine metabolism since the metabolic intermediary compound 4-hydroxyphenylpyruvate – a product of tyrosine catabolism that is also involved in CDA precursor biosynthesis – exerts a positive effect on the regulatory protein HpdR. This regulator in turn controls genes required for tyrosine catabolism and CDA biosynthesis (Liu et al., 2013). In this work, both glutamic and aspartic acid, which are present in CDA backbone, were down-represented in Sco-trpMKI reinforcing the hypothesis that CDA yield was strongly affected by amino acid availability. However, it is noteworthy that Trp levels – along with some of its precursors anthranilate, chorismate and phenylpyruvate – seemed to be not limiting for CDA biosynthesis, since these chemical species were not differentially represented in ScotrpMKI in comparison to Sco-EV.

In the trpM knock-in mutant, the decreased intracellular amount of tyrosine and phenylalanine – sharing different biosynthetic steps with Trp – parallels the decrement of arginine, glutamate and glutamine: these three amino acids derive from TCA cycle intermediates oxaloacetate (i.e., arginine) and alphaketoglutarate (i.e., glutamate and glutamine) and, moreover, biosynthesis of arginine is directly linked to glutamate/glutamine through ammonium assimilation (Lu, 2006; Lewis et al., 2011). On the other hand, the decreased amount of these amino acids is coupled with increment of histidine whose biosynthesis intersects different metabolic pathways, including those involved in Trp, purine and pyrimidine biosynthesis, as well as in C1 metabolism (Schwentner et al., 2019). In fact, Trp, purine, pyrimidine and histidine biosyntheses share the phosphoribosyl pyrophosphate (PRPP), a metabolite intermediate originating from ribose-5P. Tetrahydrofolate (THF) is a key cofactor in C1 metabolism; a metabolic source for the generation of loaded THF molecules is the reaction of serine hydroxymethyltransferase (GlyA – SCO4837) that interconverts glycine and serine. This last amino acid is a Trp precursor required in the last biosynthetic step of condensation with indole. Interestingly, a differential abundance was observed in ScotrpMKI strain for GlyA and FolD (SCO4824, a bifunctional protein with methylenetetrahydrofolate dehydrogenase and methenyltetrahydrofolate cyclohydrolase activities), with the

latter enzyme being required in biosynthesis of THF derivatives. Additionally, THF is involved in the synthesis of pantothenate that is a key precursor for the biosynthesis of essential cofactor coenzyme A (CoA) and carrier proteins having a phosphopantetheine prosthetic group such as polyketide synthases and non-ribosomal peptide synthetases (Leonardi and Jackowski, 2007). Altogether these results suggest that trpM overexpression stimulates metabolic circuits that are interconnected with each other and with Trp biosynthesis but whose regulation and functions still deserve further investigations. The alteration of amino acid biosynthetic pathways due to trpM over-expression could be also linked to the observed alkalinization of medium in liquid culture of Sco-trpMKI. Indeed, glutamate/glutamine and arginine play a key role in nitrogen assimilation (Lu, 2006; Lewis et al., 2011) and their down-representation could account for an increased amount of free ammonium causing medium alkalinization. In addition, it has been recently reported that different Streptomyces species, including S. coelicolor, are able to produce ammonia in presence of higher levels of glycine, exerting (i) an antimicrobial activity and (ii) reinforcing the efficiency of other antibiotics through perturbation of membrane permeability in target cells (Avalos et al., 2019). At this regard, it would be interesting to clarify the role of GlyA, which converts serine in glycine, and its putative interplay with TrpM.

The altered metabolism of amino acids could also indirectly contribute to increase ACT biosynthesis in two different manners. Indeed, on one hand, the augmented synthesis of some amino acids (e.g., histidine) could cause the occurrence of oxidative stress in Sco-trpMKI and, at this regard, it has been proposed that ACT could be a mean to reduce oxidative metabolism in S. coelicolor (Esnault et al., 2017). On the other hand, same authors proposed that metabolite shortage (like that of aspartic acid, arginine, phenylalanine, tyrosine, glutamic acid and glutamine in Sco-trpMKI) could act as a trigger as well. In particular, considering that trpM expression is positively correlated with ACT biosynthesis, its possible role in limiting oxidative stress through control of ACT concentration deserves further investigation: structural studies of TrpM could clarify, for example, if the two cysteine residues in position 19 and 24 are part of a 2Fe-2S cluster able to sense redox stress and induce conformational changes in TrpM structure. If this is the case, the experimental proof reported by Palazzotto et al. (2016) about the ability of TrpM to form dimers would suggest that a 2Fe-2S cluster could form upon dimerization, providing the four cysteine residues needed.

The differential proteomic analysis revealed that proteins involved in protein synthesis were more abundant in Sco-trpMKI strain. These ones included ribosomal proteins and chaperonins. The over-representation of GroEL (SCO4296) was in agreement with its well-documented involvement in metabolism during stationary phase, in morpho-physiological differentiation and in production of secondary metabolites, such as ACT (Kim et al., 2008). Moreover, proteins taking part in nucleotide metabolism and in replication and repair of DNA were more abundant as well, mirroring, at molecular level, the higher accumulation of biomass in Sco-trpMKI strain. As it is inferred from metabolite profile analysis, Trp biosynthesis does not seem dramatically altered in Sco-trpMKI, in opposition to the findings obtained for the trpMknockout mutant strain (Palazzotto et al., 2016). Indeed, in the latter, some proteins involved in Trp biosynthesis (AroF and TrpE) or its utilization (TrpS) were more abundant, suggesting a compensatory effect probably due to Trp shortage. In ScotrpMKI we did not observe up-regulation of proteins involved in Trp biosynthesis, neither an accumulation of Trp or its precursors (i.e., anthranilate and chorismite), suggesting that the trpM knock-in mutation does not exert dramatic effects on Trp metabolism as trpM-knockout mutation does. Main differences regarding these two mutant strains are reported in **Table 5**.

In addition, the relationship between stress response proteins and streptomycete development is well known since it has been described by Puglia et al. (1995). In the trpM knockin strain there was also an accumulation of stress response proteins. Analogously, a stimulatory role of Trp on S. coelicolor morphological and physiological differentiation was described in Palazzotto et al. (2015) and, among the proteins accumulated in the strain upon Trp supplementation into growth medium, proteins related to stress response were observed, including also oxidative stress proteins. Anyhow, it is still unclear how the up-regulation of trpM can exert the pleiotropic effects observed.

As reported by Palazzotto et al. (2016), TrpM interacts with PepA. In addition, qRT-PCR analysis showed that pepA was more expressed in Sco-trpMKI strain, while proteomics showed that the corresponding protein was less abundant therein. These results suggest the presence of post-transcriptional and/or posttranslational regulatory mechanisms that regulate PepA levels. It would be very interesting to find out if this putative regulation is actually due to the TrpM-PepA interaction; this point deserves further investigations.

In this work, pepA was cloned and over-expressed in E. coli to study the putative enzymatic activity of the corresponding protein. Here, we demonstrated for the first time that PepA has an enzymatic activity that corresponds to the predicted one, shedding new lights on the role of TrpM-PepA interaction. It will be important to know whether or not TrpM affects PepA activity upon binding and, if this is the case, whether

TABLE 5 | Main differences between Sco-trpMKI and the trpM-knockout mutant strains in comparison to their corresponding control strains (Sco-EV and S. coelicolor M145, respectively).


<sup>1</sup>+, yield increment/stimulation; −, yield decrement/inhibition; =, no difference observed; N.D., not determined. \*Palazzotto et al. (2016).

the effect is positive or negative. It is well documented that bacterial differentiation can be regulated through the proteasemediated selective degradation of regulators (Gottesman and Maurizi, 1992; Baker and Sauer, 2006). For instance, in S. coelicolor, the ATP-dependent serine proteases ClpP have a role in differentiation; in addition, protease inactivation was shown to induce both "bald" phenotype and lack of pigmentated antibiotics (i.e., ACT and undecylprodigiosin) (de Creìcy-Lagard et al., 1999; Bellier and Mazodier, 2004). Although PepA does not belong to the same protease class of ClpP proteins, it is interesting to note that its inactivation caused an enhancement of sporulation and production of ACT, thus suggesting a key role in S. coelicolor differentiation (Song et al., 2013; Xu et al., 2019).

Many genes involved in diverse cellular processes different from ACT production – such as metabolism of amino acids, carbohydrate, cell wall and DNA – or having unknown functions were shown to affect ACT production and/or actinomycete development (Gehring et al., 2004; Xu et al., 2019). In this work, TrpM was shown to play a role in regulating S. coelicolor growth and differentiation, specifically stimulating ACT production. In particular, our experimental data highlighted the stimulation of a regulatory circuit involving amino acid, cofactor and protein biosynthesis. In this circuit, a possible role could be exerted by the interplay between TrpM and the aminopeptidase PepA. Further studies are necessary to deeply understand the specific molecular mode of action of TrpM, explaining if and how TrpM-PepA interaction has a direct influence on the growth, antibiotic biosynthesis and developmental program of S. coelicolor. To such purpose, it will be important to determine the temporal expression profile of both TrpM and PepA and to identify cellular targets of the aminopeptidase activity of PepA, also considering that the augmented expression of trpM and the inactivation of pepA (Song et al., 2013) both positively influence ACT production in S. coelicolor. Furthermore, even the putative interaction between TrpM and the ribosomal protein S2 (SCO5624) will deserve a specific investigation: in fact, it was previously identified as a putative interacting protein of TrpM (Palazzotto et al., 2016) and it was interestingly overrepresented in Sco-trpMKI as well. Finally, considering that TrpM is conserved among Streptomyces species, it should be investigated whether it is able to exert a positive effect even on other secondary metabolite biosynthesis, especially polyketides. Indeed, this chemical class includes several compounds with antibiotic, antitumoral and immunosuppressant activity. So, the development of recombinant strains with altered expression of TrpM and/or PepA could be promising to enhance industrial production of these molecules.

#### DATA AVAILABILITY STATEMENT

Proteomic data have been submitted to the PRIDE database under the project accession PXD015937.

# AUTHOR CONTRIBUTIONS

AV, EP, GR, LB, and TF performed the experiments. AS and GG supervised the experiments and analyzed the results. AV, AP, and GG designed the experiments. AP and GG conceived the project. AV wrote the manuscript with the help of AS and GG. All authors read and approved the manuscript.

#### FUNDING

This work was partly supported by MIUR FFABR 2017 and University of Palermo FFR 2018 to GG.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2020.00224/full#supplementary-material

FIGURE S1 | qRT-PCR analysis showing the relative quantification (RQ) of trpM expression in Sco-trpMKI and Sco-EV.

FIGURE S2 | SEM images of Sco-trpMKI and Sco-EV after 24, 48, 72, and 120 h of growth on solid MM. (A) Sco-trpMKI 24 h. (B) Sco-EV 24 h. (C) Sco-trpMKI 48 h. (D) Sco-EV 48 h. (E) Sco-trpMKI 72 h. (F) Sco-EV 72 h. (G) Sco-trpMKI 120 h. (H) Sco-EV 120 h.

FIGURE S3 | Representative 2D-proteome map of whole protein extracts obtained from Sco-trpMKI. Differentially represented proteins listed in Supplementary Tables S1, S2 are labeled.

FIGURE S4 | Distribution into functional classes of differentially represented proteins (A) and metabolites (B) when Sco-trpMKI and Sco-EV strains were compared. Percentages indicate the relative abundance of each functional class.

FIGURE S5 | Metabolic pathway map showing over-represented proteins (red), down-represented proteins (light green), over-represented metabolites (maroon), and down-represented metabolites (dark green) in Sco-trpMKI in comparison to Sco-EV. Only proteins involved in the labeled metabolic pathways are reported. Proteins are indicated with the corresponding genomic locus (i.e., SCOXXXX), while metabolites with their KEGG identifier (i.e., CXXXXX). This map was generated using the web-based tool Interactive Pathways Explorer v3 (iPath3) (Darzi et al., 2018).

FIGURE S6 | qRT-PCR analysis showing relative quantification (RQ) of pepA expression in Sco-trpMKI and Sco-EV.

TABLE S1 | Description, abundance profile, functional classification and mass spectrometry identification parameters of differentially represented proteins. Functional classification is based on KEGG orthology.

TABLE S2 | Protein identification details of spots observed as differentially represented.

TABLE S3 | Description, abundance profile and functional classification of differentially represented metabolites. Functional classification is based on KEGG orthology.

TABLE S4 | Description, abundance profile and functional classification of not differentially represented metabolites. Functional classification is based on KEGG orthology.

# REFERENCES

fmicb-11-00224 February 17, 2020 Time: 16:57 # 12



**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2020 Vassallo, Palazzotto, Renzone, Botta, Faddetta, Scaloni, Puglia and Gallo. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The Application of Regulatory Cascades in Streptomyces: Yield Enhancement and Metabolite Mining

Haiyang Xia<sup>1</sup> , Xiaofang Li<sup>1</sup> , Zhangqun Li<sup>1</sup> , Xinqiao Zhan<sup>1</sup> , Xuming Mao1,2 \* and Yongquan Li1,2 \*

1 Institute of Biopharmaceuticals, Taizhou University, Taizhou, China, <sup>2</sup> Institute of Pharmaceutical Biotechnology, School of Medicine, Zhejiang University, Hangzhou, China

Streptomyces is taken as an important resource for producing the most abundant antibiotics and other bio-active natural products, which have been widely used in pharmaceutical and agricultural areas. Usually they are biosynthesized through secondary metabolic pathways encoded by cluster situated genes. And these gene clusters are stringently regulated by interweaved transcriptional regulatory cascades. In the past decades, great advances have been made to elucidate the regulatory mechanisms involved in antibiotic production in Streptomyces. In this review, we summarized the recent advances on the regulatory cascades of antibiotic production in Streptomyces from the following four levels: the signals triggering the biosynthesis, the global regulators, the pathway-specific regulators and the feedback regulation. The production of antibiotic can be largely enhanced by rewiring the regulatory networks, such as overexpression of positive regulators, inactivation of repressors, fine-tuning of the feedback and ribosomal engineering in Streptomyces. The enormous amount of genomic sequencing data implies that the Streptomyces has potential to produce much more antibiotics for the great diversities and wide distributions of biosynthetic gene clusters in Streptomyces genomes. Most of these gene clusters are defined cryptic for unknown or undetectable natural products. In the synthetic biology era, activation of the cryptic gene clusters has been successfully achieved by manipulation of the regulatory genes. Chemical elicitors, rewiring regulatory gene and ribosomal engineering have been employed to crack the potential of cryptic gene clusters. These have been proposed as the most promising strategy to discover new antibiotics. For the complex of regulatory network in Streptomyces, we proposed that the discovery of new antibiotics and the optimization of industrial strains would be greatly promoted by further understanding the regulatory mechanism of antibiotic production.

Keywords: antibiotic production, regulatory cascades, rewiring regulatory network, unlocking cryptic metabolites, Streptomyces

# INTRODUCTION

Streptomyces, Gram-positive mycelial bacteria with high GC content, shows a complex morphological differentiation and belongs to actinobacteria (Hopwood, 2019). It is reported that 61% of so far discovered microorganism-derived bioactive substances are produced by actinobacteria (mainly Streptomyce). Most of the clinically essential drugs containing antibiotic and

#### Edited by:

Yvonne Mast, German Collection of Microorganisms and Cell Cultures GmbH (DSMZ), Germany

#### Reviewed by:

Gang Liu, Institute of Microbiology (CAS), China Gong-Li Tang, Shanghai Institute of Organic Chemistry (CAS), China

#### \*Correspondence:

Xuming Mao xmmao@zju.edu.cn Yongquan Li lyq@zju.edu.cn

#### Specialty section:

This article was submitted to Microbial Physiology and Metabolism, a section of the journal Frontiers in Microbiology

> Received: 31 October 2019 Accepted: 26 February 2020 Published: 24 March 2020

#### Citation:

Xia H, Li X, Li Z, Zhan X, Mao X and Li Y (2020) The Application of Regulatory Cascades in Streptomyces: Yield Enhancement and Metabolite Mining. Front. Microbiol. 11:406. doi: 10.3389/fmicb.2020.00406

**124**

other bio-active agents are derived from Streptomyces (Waksman et al., 2010). In the past decades, scientists have made great advances to elucidate the regulatory mechanisms related to antibiotic production in Streptomyces. Generally, the antibiotic production is stringently and elaborately regulated by pyramidal transcriptional regulatory cascades, including signaling pathways, global regulators, pathway-specific regulator (PSR), and feedback regulation. This interweaved networks can determine the production levels of antibiotic under specific culture condition (Chater, 2016).

With the accumulation of genome data, the number of predicted secondary biosynthetic gene clusters (SBGs) on genome is much more than that of the products has been identified. Generally, most of these gene clusters have been defined as cryptic ones because of unknown or undetectable secondary metabolites. Scientists proposed that the Streptomyces has been greatly underestimated for the capability to produce diversity of natural products (Baltz, 2017, 2019). Up to now, Streptomyces strains are still taken as the most promising candidates for novel antibiotic discovery. In the synthetic biology era, manipulation of regulatory genes has been employed to activate of some cryptic gene clusters.

The production of antibiotic in Streptomyces can be largely enhanced by rewiring the regulatory network. There are several reviews on the advances about the elucidation of regulatory networks and the activation of cryptic gene clusters in Streptomyces (Ochi, 2017; Onaka, 2017; Wei J. et al., 2018; Palazzotto et al., 2019). Antibiotics or specialized metabolites usually means same for secondary metabolites produced by Streptomyces. Here, the term antibiotic is used to stand for the natural products. In this review, we will focus on the production enhancement and new antibiotic discovery by manipulation of the regulatory networks in Streptomyces. We proposed that the systematic rewiring of regulatory networks in Streptomyces would play a critical role in drug discovery and production enhancement in the near future.

#### THE REGULATORY CASCADES OF ANTIBIOTIC PRODUCTION IN Streptomyces

It is commonly recognized that antibiotic production is stringently regulated by pyramidal transcriptional regulatory cascades in Streptomyces. Recent years, there were several reviews on the regulatory networks of Streptomyces (Romero-Rodriguez et al., 2018; van der Heul et al., 2018; Wei J. et al., 2018; Palazzotto et al., 2019). So here we won't go into detail of the regulatory networks involved in antibiotic production. The regulatory cascades of antibiotic production will be briefly presented from the following four levels.

The first one is the onset of antibiotic production, which is triggered by the coupled receptors of Streptomyces "hormones" or other signals. Among them, A-factor, the chemical structural characteristic γ-butyrolactone (GBL), is the first Streptomyces hormone reported as a signal triggering streptomycin production in Streptomyces griseus in 1967. Avenolide was identified as a new type of butenolide hormone regulating avermectin production in Streptomyces avermitilis (Kitani et al., 2011). The production of the methylenomycin is induced by methylenomycin furan (MMF), a furan-type autoregulator in Streptomyces coelicolor (Xu G. et al., 2010). It was reported that 84.1% actinomycetes probably use either GBL (64.1%) or butenolide (24%) to control the antibiotic production (Thao et al., 2017). A-factor, coupled with its receptor ArpA, controls transcription of the master positive regulator AdpA for morphological differentiation and streptomycin production in S. griseus (Higo et al., 2012).

The second level is the global regulators which bring pleiotropic effects on the lower level of regulators. The global regulators affect more than one metabolic pathways and may not directly affect any specific biosynthetic gene clusters (BGCs). They also respond to a variety of chemical or physiological signals, e.g., nutrient limitation, the concentration of chitin or N-acetylglucosamine (GlcNAc) in the medium, cell wall damage, heat shock or pH shift. Two-component systems (TCSs, consisted with a membrane-bound histidine kinase, which senses specific environmental stimuli, and a cognate regulator) play crucial roles in sensing extracellular signals. Typical TCSs mediate responses to the cellular signal, mainly through regulating the transcription of downstream genes. Moreover, TCSs, as the most abundant pleiotropic regulators, are involved in the dynamic control of the biosynthesis of secondary metabolites in Streptomyces. E.g., PhoP/PhoR can regulate antibiotic production and morphological differentiation (Yepes et al., 2011; Rodriguez et al., 2013). Both antibiotic production and GlcNAc uptake via the phosphotransferase system are directly regulated by DasR, a GntR-family allosteric regulator, with the GlcNAc as a ligand (Fillenberg et al., 2016). The major checkpoint for the onset of secondary metabolism is the accumulation of GlcNAc through the autolytic degradation of the vegetative mycelium. The antibiotic production is triggered by DasR regulon (Rigali et al., 2008). AdpA, which is ubiquitously distributed in streptomycetes as a member of the AraC/XylS family regulators, affects transcription of hundreds of genes involved in morphological differentiation and antibiotic biosynthesis (Higo et al., 2012; Guyet et al., 2014). ArpA, the receptor of A-factor, regulated the transcription of adpA in S. griseus. The transcription level of adpA is dynamic controlled by the interaction between ArpA and A-factor. As A-factor reached a threshold concentration, it binds to ArpA and releases the repression of adpA transcription (Wei J. et al., 2018). WblA and AtrA also have been extensively investigated as the global regulators involved in the biosynthesis of antibiotic. As a global repressor, WblA affects doxorubicin (DXR), tautomycetin, and daptomycin biosynthesis in their natural host strains (Noh et al., 2010; Nah et al., 2012; Kim et al., 2014). However, WblA activates natamycin biosynthesis in Streptomyces chattanoogensis (Yu et al., 2014).

The third level is pathway-specific regulators (PSRs), which are taken as the master switches of antibiotic production. PSRs are situated in the BGCs of secondary metabolites and directly regulate the transcription of the biosynthetic genes. So they have been called the "cluster-situated regulators (CSRs)." SARP (Streptomyces antibiotic regulatory proteins), LAL (large ATP-binding regulators of the LuxR family) and PAS-LuxR

family regulators usually belong to PSRs (Martin and Liras, 2012). The ActII-ORF4 and RedD of S. coelicolor and DnrI of S. peucetius are typical SARPs family activators (Song et al., 2011). The LAL family regulators, which comprise an N-terminal ATP/GTP-binding domain and a C-terminal LuxR family DNAbinding domain, usually activate the biosynthesis of antibiotic in Streptomyces. PimR, RapH, NysRI, AveR, and SlnR, which are typical LAL regulators, are located in type I polyketide BGCs in Streptomyces (Anton et al., 2004; Sekurova et al., 2004; Kuscer et al., 2007; Guo et al., 2010; Zhu et al., 2017). The whole process of avermectin biosynthesis is controlled by direct interaction between AveR and all promoters of ave cluster (Guo et al., 2010). The PAS-LuxR family regulator is characterized with an N-terminal PAS sensory domain and a C-terminal LuxR type DNA binding domain. PAS domain probably function as responsor to light, redox potential, oxygen, overall energy level of a cell, and small ligands (Taylor and Zhulin, 1999). For the cytosol localization, proteins containing PAS domains can sense internal signals and other environmental factors which cross the cell membrane. PimM, a PAS-LuxR family regulator, positively regulates pimaricin production in Streptomyces natalensis (Anton et al., 2007). The orthologs of PimM have been represented in most of the reported BGCs of antifungal polyketides like amphotericin (AmphRIV), candicidin (FscRI), nystatin (NysRIV), and filipin (PteF) gene clusters (Santos-Aberturas et al., 2011). PSRs also contain other family repressor or activator. The production of antibiotic can be positively or negatively regulated by TetR, MarR, LysR, and IclR family regulators (Molina-Henares et al., 2006; Martin and Liras, 2012; Cuthbertson and Nodwell, 2013; Zhang et al., 2015; Guo et al., 2018).

The fourth level is the feedback regulation which is brought by antibiotic and/or intermediates to coordinate antibiotic production and transport. Evidences have shown that antibiotic functions as signals to regulate the production of antibiotic besides as feedback substances for the enzymatic reactions. Antibiotic, as ligand for proper regulator, affects the final production in Streptomyces. The expression of antibiotic biosynthetic genes was modulated by the RedZ and undecylprodigiosin complex (Wang L. et al., 2009). The activity of AtrA, which regulates primary and secondary metabolism, is reduced by lidamycin of Streptomyces globisporus and actinorhodin (ACT) of S. coelicolor (Li et al., 2015). The biosynthesis of jadomycin is dynamically modulated by the interaction among jadomycin B, chloramphenicol, JadR1 and JadR2 in Streptomyces venezuelae (Wang L. et al., 2009; Xu D. et al., 2010). Daunorubicin (DNR) biosynthesis is regulated by three DNA binding regulatory proteins (DnrI, DnrN, and DnrO). The DNA binding activity of DnrO can be modulated by Rhodomycin D, a glycosylated precursor of DXR (Jiang and Hutchinson, 2006). Simocyclinone and its precursors inhibit the binding activity of SimReg1 to several promoter regions of simocyclinone biosynthesis genes and SimReg1 encoding gene (Horbal et al., 2012). As a GBL receptor-like protein, PapR5, which is the major regulator of pristinamycin biosynthesis, may sense pristinamycin or intermediate(s) of the pathway (Mast et al., 2015). SsaA can activate sansanmycin biosynthesis by binding to five different regions within the sansanmycin BGC. The sansanmycins A and H inhibit DNA-binding activity of SsaA in a concentration-dependent manner (Li et al., 2013). The rifamycin B, the end product of rifamycin biosynthesis, can relieve the repression of RifQ on the transcription of the rifamycin efflux pump (RifP) (Lei et al., 2018). Transporters may affect product maturation. The ABC transporters NysG and NysH involved in the biosynthetic export of nystatin in Streptomyces noursei. Deletion of either transporter gene caused ca. 35% reduction in the levels of fully formed nystatin and led to concomitant enrichment of a deoxy precursor. The authors proposed that a NysGH complex might preferentially export fully formed nystatin and that this activity would increase the efficiency of the last biosynthetic step through product removal (Sletta et al., 2005). For both class I and II lantibiotics, the function of a dedicated ABC transporter, generally termed 'LanT,' is essential for the formation of the final product (Gebhard, 2012).

Signals and regulators, dynamic fluctuating on the metabolic state of the cell and its environment, are extensively involved in regulation of secondary metabolism. The abundances of nitrogen, phosphate and carbon sources are nutrimental signal for the onset of secondary metabolism. It has been reported that GlnR and PhoR–PhoP was involved in the type of regulation depending on the carbon, nitrogen, and phosphate supply. For the similarity of binding sites, PhoP and GlnR showed competitive binding to target genes in some cases (Romero-Rodriguez et al., 2018). In Saccharopolyspora erythraea, the GlnR regulon is not only involved in nitrogen metabolism, but it also appears to control ABC-type transporters for uptake of carbon sources (Liao et al., 2015). In S. avermitilis, AveR positively regulates avermectin production and negatively affects oligomycin biosynthesis, respectively (Guo et al., 2010). AveI, an AtrA-like regulator, regulates production of avermectin, oligomycin, melanin, and morphological differentiation by directly regulating the transcription of ave, olm, melC1C2, ssgRD, wblI and genes in primary metabolism, including substrates transport, the metabolism of amino acids, lipids, and carbohydrates (Liu et al., 2019). Usually more than twenty secondary metabolic pathways exist in most of the Streptomyces (Nett et al., 2009). There is complex cross-talk regulation among different biosynthetic clusters and between the primary and secondary metabolism. With the deep understanding the regulatory mechanism of the antibiotic production, there is a consensus that the complicated and interweaved regulatory networks contribute to the dynamic antibiotic production in Streptomyces.

# ENHANCEMENT OF ANTIBIOTIC PRODUCTION IN Streptomyces BY REWIRING THE REGULATORY NETWORKS

As mentioned above, the regulatory networks constitute major bottlenecks to over produce target antibiotics. The manipulation of regulatory genes can contribute to overcome these bottlenecks

to turn on the expression of gene clusters for antibiotic production. Various strategies have been taken to manipulate regulatory genes to achieve the optimal antibiotic production in both the native and/or heterologous hosts.

# Enhancing Antibiotic Production by Overexpression of Positive Regulator Genes

The regulators also can be defined as positive and negative regulators according their effect on the antibiotic production. The positive regulators (activators) can promote the biosynthesis of antibiotics. But the negative ones (repressors) can repress the biosynthesis of antibiotics (Martin and Liras, 2010). Since the positive regulators activate the transcription of antibiotic BGCs, they can be manipulated to enhance the production of antibiotic in Streptomyces. The titer improvement can efficiently and simply be achieved by over-expression of genes encoding activators with proper promoters. As listed in **Table 1**, overexpression of genes encoding LAL family regulators, such as MilR, NemR, and AveR, has been used to increase production of milbemycin in S. bingchenggensis BC04, nemadectin in S. cyaneogriseus subsp. non-cyanogenus NMWT1 and avermectin in S. avermitilis, respectively (Guo et al., 2010; Zhang Y. et al., 2016; Li et al., 2019). Overproduction of nikkomycin has been achieved by engineering of the CSR activator gene sanG with different constitutive promoters (Liu et al., 2005). Overproduction of oxytetracycline (OTC) has been achieved by overexpression of the CSR activator gene otcR as tandem copies under the control of a constitutive SF14 promoter (Yin et al., 2015). Similar strategy has also been used to overproduce tacrolimus (FK506) in Streptomyces tsukubaensis NRRL18488 by overexpression of bulZ (Ma et al., 2018). Other examples include LysR family regulator for ascomycin production in S. hygroscopicus var. ascomyceticus (Song et al., 2017), PAS-LuxR family regulator for wuyiencin production in Streptomyces wuyiensis CK-15 (Liu et al., 2014), and Crp/Fnr family regulator for leinamycin production in Streptomyces atroolivaceus (Huang et al., 2016).

The expression level of positive regulator is not always correlated to the production of antibiotic. E.g., overexpression of milR with a strong constitutive promoter led to decrease of milbemycin production in S. bingchenggensis (Zhang Y. et al., 2016). These showed that the threshold of the over-expressed regulator was key point to determine the production of antibiotic in Streptomyces.

#### Enhancing Antibiotic Production by Removal of Repressor Genes

The TetR and LysR family regulators are widely distributed in the genome of Streptomyces. Most of TetR and LysR regulators function as repressors. More details of TetR family regulators have been reviewed elsewhere (Cuthbertson and Nodwell, 2013). Deletion of TetR family repressors has been used to increase avermectin production in S. avermitilis (He et al., 2014), calcimycin production in Streptomyces chartreusis NRRL 3882 (Gou et al., 2017), and pristinamycin production in Streptomyces pristinaespiralis (Meng et al., 2017).

WblA, a WhiB-like protein, also widely distributes among actinomycetes (Kim et al., 2014; Huang et al., 2017). WblA affects morphological development and antibiotic biosynthesis. It generally functions as a global repressor of antibiotic biosynthesis, such as DXR biosynthesis in S. peucetius, tautomycetin biosynthesis in Streptomyces sp. CK4412, and daptomycin biosynthesis in Streptomyces roseosporus. Deletion of wblA leads to overproduction of pikromycin in S. venezuelae (Woo et al., 2014), daptomycin in S. roseosporus (Huang et al., 2017). Other examples include deletion of genes encoding GntR family regulators for platensimycin and platencin overproduction in Streptomyces platensis (Smanski et al., 2009).

The nsdA, a gene negatively affecting Streptomyces differentiation, had been proved a pleiotropic negative regulatory gene in S. coelicolor (Li et al., 2006). It plays a negative role in sporulation, morphological differentiation and antibiotic synthesis. The overproduction of ACT, calcium-dependent antibiotic (CDA), and methylenomycin was detected in a nsdA mutant. The nsdA homologous genes have been found to conservatively distribute in Streptomyces. The milbemycin A4 and nanchangmycin production increased in nsdA mutant of S. bingchengensis by 1.5-fold and 9-fold, respectively (Wang X. J. et al., 2009). The natamycin production can be increased 1.9-fold by deletion of nsdA gene in Streptomyces lydicus A02 (Wu et al., 2017).

γ-Butyrolactone receptors, like the A-factor receptor ArpA, repress transcription of adpA to affect the production of streptomycin in S. griseus. Deletion of the GBL receptors probably can promote antibiotic production. The validamycin production of mutants with deletion of arpA homologs were increased by 26% (1shbR1) and 20% (1shbR3) in S. hygroscopicus 5008, respectively (Tan et al., 2015). Deletion of avaR1, the avenolide (a novel butenolide-type autoregulator) receptor encoding gene, increased production of avermectin B1a approximately 1.75 times compared with the parent strain in a high-producing S. avermitilis strain (Wang et al., 2014).

#### Enhancing Antibiotic Production by Manipulation of Feedback and Transport

Genes encoding exporters, which are responsible for the secretion of antibiotic, often situate in their BGCs. The majority of BGClinked antibiotic exporters belongs to various subgroups like the ATP-binding cassette (ABC) superfamily and major facilitator superfamily (MFS) of transporters. It is commonly assumed that BGC-linked exporters function in the secretion of the antibiotic made by BGC. Toxic end-products can be pumped out of the cells by exporters to realize more durable and sustainable productivity.

Importantly, even in the absence of toxicity, the ability to secrete the end-products may play key roles in affecting the expression of the BGCs. ActA (ActII-ORF2) and ActB (ActII-ORF3) are MFS and RND (root-nodulation-division) exporters, respectively, involved in ACT production in S. coelicolor. ActA and ActB are crucial players in the sophisticated feed-forward mechanism, by which secretion of the end-products leads to the transcriptional activation of the BGCs in the cells (Tahlan et al., 2007; Xu et al., 2012). The double actAB mutant could

#### TABLE 1 | Examples of antibiotic production enhancement in Streptomyces by regulatory gene manipulation.


\*Heterologous expression.

only produce one fifth of ACT compared to WT. The activation of ACT production occurs in two waves where expression of key act genes, including the actAB operon, is initially induced by a biosynthetic intermediate. The ACT production grows into full induction only once the accumulating ACT is secreted out of the cells.

The genes encoding for the AvtAB ABC transporter are adjacent and co-expressed with the BGC for avermectin in S. avermitilis. In the WT strain, deletion of either avtA or avtB had no effect on avermectin production, while over-expression of avtAB led to the increased level of avermectin B1a with two-fold. But the production level of oligomycin A, another product from

S. avermitilis, was found unaltered. The production promotion effects of avtAB could be specific to avermectin in S. avermitilis (Qiu et al., 2011). Co-overexpression of three OTC resistance genes, including otrA (encoding a ribosomal protection protein), otrB and otrC (encoding two efflux proteins), led to 179% increase of OTC production in Streptomyces rimosus M4018 (Yin et al., 2017).

Feed-forward mechanisms may explain the regulation of BGCs for the actinobacterial ribosomally synthesized and posttranslationally modified peptides (RiPPs), planosporicin and microbisporicin. The multiple ABC transporter genes in these BGCs seem to modulate their production and self-immunity (Foulston and Bibb, 2010; Sherwood et al., 2013). GouM, the MFS transporter, is responsible for the secretion of gougerotin outside of Streptomyces graminearus (Wei et al., 2014). The overexpression of BotT, a putative efflux pump encoded in the bottromycin BGC, increased bottromycin production about 20 times in a heterologous host (Huo et al., 2012).

Export of antibiotic is important for the producer to reduce the intracellular antibiotic concentration, which can relieve selftoxicity. In Amycolatopsis mediterranei, 1rifQ mutant brought overexpression of RifP. The accelerated export of rifamycin may reduce the intracellular rifamycin concentration, relieve other possible feedback inhibition of rifamycin biosynthesis and finally lead to more than two-fold improvement of rifamycin B production (Lei et al., 2018). Overexpression of DrrC, which provide self-resistance to DNR and DXR, achieved 5.1-fold increase in DXR production in S. peucetius ATCC 27952 (Malla et al., 2010).

For there are many transporters help to transport end products or intermediates out of the cell, this can help to remove the feedback caused by the antibiotics or intermediates. Transporters enhance the efflux of the self-produced antibiotics, which can be an important strategy for self-protection from self-toxicity. Many antibiotic BGCs comprise functional exporter genes (Rees et al., 2009; Severi and Thomas, 2019). Also some exporters encoding genes for detoxification are outside of the BGCs. By systematical investigation, it has proved that more than 25 groups of genes contributed the efflux of molecules involved in natamycin biosynthesis (Shan et al., 2019).

#### Enhancing Antibiotic Production by Ribosome Engineering

Classical strain improvement is usually achieved by multiple rounds of random mutagenesis and repetitively screening. Although the outcome is fruitful, the weakness of this strategy is time intensive and laborious. A ribosome engineering approach has been designed to obtain antibiotic overproducing strains by corelatively screening for mutants with proper drug resistant level, e.g., streptomycin resistance mutations (Hosoya et al., 1998). It has been proven to be an effective rational strategy for strain improvement. The percentage of overproducing mutants among drug resistance mutants can be up to 2–20% in a species dependent way, which is much higher than that of random mutagenesis. S12-mutated variants, which have mutations in the rpsL gene with resistance to streptomycin, stabilize the open A-site conformation and prevent streptomycin from binding (Holberger and Hayes, 2009). Screening of streptomycin resistant mutations has been demonstrated a useful tool for strain improvement in many Streptomyces strains (Ochi, 2007). The undecylprodigiosin production in S. lividans was activated in RpsL mutants with K88E, L90K, and R94G substitutions (Okamoto-Hosoya et al., 2003b). The industrial strain of S. albus, which had RpsL (K88R) and RpoB mutation, enhanced salinomycin production by 1.5-fold (Tamehiro et al., 2003). In S. coelicolor, over 10-fold ACT improved in RpsL (K88E and P91S) strains than the original strain (Okamoto-Hosoya et al., 2003a). Both the K88E mutation and an insertion mutation at GI92 led to substantially higher levels of ACT in S. coelicolor strains (Wang G. et al., 2009). Selection of mutations in the rpsL gene are widely used to improve the production in a variety of industrial antibiotic-producing strains (Tamehiro et al., 2003; Beltrametti et al., 2006).

Another method of "ribosome engineering" incorporates the inactivation of the rsmG gene, which encodes an S-adenosylmethionine (SAM)-dependent 16S rRNA methyltransferase. Recombinants with deletion of rsmG gene resulted in overproduction of ACT in S. coelicolor (Nishimura et al., 2007).

Previous reports showed that ribosome engineering was an effective strategy for yield improvement in Streptomyces. From the reference, ribosome engineering has been employed for yield improvement in more than fifty Streptomyces strains (Zhu et al., 2019). As a rational and cost-effective approach, ribosome engineering could be adapted to speed up strain development for those without clear genetic background. If compared with other direct genetic manipulation approach, the ribosome engineering still be with laborious screening to overcome the phenotype uncertainty of resistant mutants.

The above data show that manipulation of regulatory cascades can be an efficient way to improve the production of antibiotic. It is obvious that the balance and synergy between primary and secondary metabolism is very important for the overproduction of antibiotic. Rewiring regulatory network combined with metabolic engineering will be a more powerful way to enhance the production of antibiotic in Streptomyces.

#### DISCOVERING NOVEL ANTIBIOTICS THROUGH ACTIVATING THE CRYPTIC GENE CLUSTERS BY UNLOCKING THEIR REGULATORY NETWORK

The genome sequencing of streptomycetes demonstrated that each genome has the genetic capacity with 20–40 distinct gene clusters for secondary metabolites. They can produce far more compounds than reported previously (Nett et al., 2009). Since many of these gene clusters are expressed at low levels under laboratory conditions, they are called cryptic gene clusters. Several strategies have been designed to activate these clusters and discover novel molecules. The feasible approaches include uses of signals probes, ribosome engineering, regulatory unlocking

and heterologous expression (Laureti et al., 2011; Ochi, 2017; Myronovskyi and Luzhetskyy, 2019). For this review is about the regulation in Streptomyces, the heterologous expression will not be discussed here.

# Activation of Cryptic Gene Clusters by Chemical or Physical Signals

Though Streptomyces has the potential to produce over 20 secondary metabolites, most BGCs are cryptic or silent. As we know from the regulatory cascades, there are many signaling factors affecting the expression of secondary metabolic gene, and many culture dependent methods have been developed (Yoon and Nodwell, 2014).

The one strain many compounds (OSMACs) approach has been previously applied to explore the secondary metabolic potential of different strains with altering a single parameter in the growth conditions or eliciting a stress response (Bode et al., 2002). Heat shock and ethanol shock are two widely applied measures. S. venezuelae produces negligible amounts of jadomycin when it is cultured at 27◦C without heat/ethanol shock. The yield of jadomycin B reach 25 µg/ml after 12 h by shifting the temperature from 27 to 42◦C. Similarly, cultures with 6% ethanol enhanced yields of jadomycin to as high as 30 µg/ml at 27◦C (Doull et al., 1993, 1994). Another compound that is produced in response to the heat/ethanol shock is validamycin A (VAL-A) by Streptomyces hygroscopicus. VAL-A rapidly accumulated at relatively higher fermentation temperatures (37◦C, 40◦C, and 42◦C). But the production was very low at lower temperatures (28◦C, 30◦C, 33◦C, and 35◦C) (Liao et al., 2009). Several other stress responses have been explored. New metabolites can be produced by Streptomyces parvulus upon increasing the hydrostatic pressure during fermentation (Bode et al., 2002). S. coelicolor can produce ectoine and 5 hydroxyectoine under high salt conditions and temperature stress (Bursy et al., 2008). Methylenomycin production can be activated in S. coelicolor by either alanine growth-rate-limiting conditions and/or acidic pH shock (Hayes et al., 1997).

As we know that the autoregulators can trigger the antibiotic production in Streptomyces. These classes of naturally produced chemical probes like GBLs can be used to elicit the expression of cryptic SBGs. Goadsporin, a compound isolated from Streptomyces sp. TP-A0584, was shown to stimulate the production of prodiginine antibiotic in S. lividans, and promote pigment production and morphogenesis on 36 streptomycetes (Onaka et al., 2001).

In Streptomyces, antibiotic production, which is usually coupled with the onset of development, is triggered by the interaction between DasR and the accumulation of GlcNAc after autolytic degradation of the vegetative mycelium. So GlcNAc can be used as a signal chemical to activate pathways for secondary metabolite biosynthesis. Several Streptomyces species were examined for their antibiotic production on MM plates (25 mM mannitol as the sole carbon source) with or without GlcNAc (50 mM). GlcNAc had a stimulating effect on antibiotic production on S. clavuligerus, S. collinus, S. griseus, S. hygroscopicus, and S. venezuelae (Rigali et al., 2008).

By investigating small molecules that perturb secondary metabolism, Justin group had screen out 19 compounds from 30,569 small molecules of the Canadian Compound Collection for their ability to alter the pigmentation of S. coelicolor. The ARC2 series (ARC2, ARC3, ARC4, and ARC5) were structurally related to triclosan, a synthetic antibiotic (Craney et al., 2012). Particularly, ARC2 altered the secondary metabolite output in all of the tested streptomycetes (Craney et al., 2012; Ahmed et al., 2013). These probes appear to have the potential to widely use as active elicitors for mining secondary metabolite.

Since Streptomyces is the soil dwelling microbe, it was reported that antibiotic overproduction could be promoted by 2- to 25-fold when the rare earth, scandium (Sc), added at a low concentration (10–100 mM) to cultures of S. coelicolor A3(2) (ACT producer), Streptomyces antibioticus (actinomycin producer), and S. griseus (streptomycin producer). Scandium was also effective in activating the dormant ability to produce ACT in S. lividans (Kawai et al., 2007). The rare earth elements, scandium and/or lanthanum, can activate the expression of nine genes belonging to nine secondary metabolite–BGCs by 2.5- to 12-fold in S. coelicolor A3(2). Several compounds can only be detected with HPLC in the rare earth-treated cultures (Tanaka et al., 2010).

Combined-culture is a co-culture method to activate secondary metabolism in Streptomyces (Onaka, 2017). The biosynthesis of red pigment by S. lividans TK23 can be influenced by co-culture with Tsukamurella pulmonis TP-B0596. It was proved that the biosynthesis of cryptic natural products in Streptomyces species can only be induced by living mycolic acid-containing bacteria (MACB). ∼90% of Streptomyces species, which isolated from soil samples collected in the Hokuriku district of Japan, show changes in secondary metabolism in combined-culture with T. pulmonis (Onaka et al., 2011).

Binding to a specific receptor, like the A factor, is proposed as possible mechanisms of activation of secondary metabolite biosynthesis by some chemical probes. The perturbation of host metabolism can be another reason to activate cryptic clusters. The ARC2 series, similar to triclosan, repress fatty acid biosynthesis by inhibiting the enoyl reductase FabI and change the flux of precursor molecules to antibiotic biosynthesis. Quorum sensing molecules could also important factors to induce the secondary metabolite production in actinomycetes. The results from co-cultivation probably due to physical cell to cell interactions, small molecule-mediated communication or metabolite precursor supply (Abdelmohsen et al., 2015; Jones et al., 2017). Though a great many of products are discovered through activation of cryptic gene clusters by chemical or physical signals, it still lacks clearly explanation of the molecular mechanism triggering the expression of the secondary metabolic genes. But for this approach is non-genetic dependent, it can be developed to high-throughput screening model to investigate more strains easily.

# Activation of the Cryptic Gene Clusters by Manipulation of the Regulatory Genes

The transcription processes of BGCs are usually modulated by specific regulatory gene clusters, including activators and

repressors, which activate or repress biosynthesis, respectively. The identification of PSRs genes offers possibility to activate the desired BGCs either by inactivation of the repressor or overexpression of the activator genes (Baral et al., 2018).

TetR family of transcription regulators, which participate in the regulatory pathways associated with the efflux of antibiotic, cell–cell signaling, antibiotic biosynthesis, biofilm formation, etc., is one of the most predominant families of transcription factors in the prokaryotic system (Bhukya and Anand, 2017). The TetR family members usually serve as repressors which possess a N-terminal DNA-binding domain and a C-terminal signal reception domain (Cuthbertson and Nodwell, 2013). The inactivation of a TetR transcriptional repressor gene arpRII in Streptomyces argillaceus promoted the activation of the cryptic gene cluster arp, and discovery of argimycins P, a natural product in polyketide alkaloid family (Ye et al., 2018). ScbR, another TetRlike transcriptional repressor, has been confirmed to directly control the expression of the cryptic type I polyketide BGC by binding at two different positions of kasO promoter in S. coelicolor (Takano et al., 2005; Xu D. et al., 2010; Bhukya and Anand, 2017). GdmRIII, a TetR family transcriptional regulator, was connected with the biosynthetic pathways of geldanamycin and elaiophylin in Streptomyces autolyticus CGMCC0516 (Jiang et al., 2017). Moreover, it has been proved that the biosynthetic pathways including jadomycin, kinamycin, and auricin, were activated after removal of TetR family repressors, JadR/JadR2 (Zhang et al., 2013), AlpW (Bunet et al., 2011), and SCO1712 (Lee et al., 2010), respectively.

LuxR superfamily is another predominant class of regulators associated with quorum sensing (Baral et al., 2018). Contrary to the TetR family, LuxR-like proteins usually act as activators during secondary metabolic regulation. LuxR proteins consist of an N-terminal DNA binding domain and a C-terminal ligand-binding domain, where ligand binding induces homodimerization and subsequent binding of protein to DNA to initiate transcription (Maddocks and Oyston, 2008). LuxR proteins, which are abundant among Actinobacteria, Proteobacteria, and Firmicutes, play significant roles in improvement of antibiotic production and activation of cryptic gene clusters after overexpression. The LuxR family members, GdmRI and GdmRII, are positive regulators required for geldanamycin biosynthesis in S. hygroscopicus 17997, and inactivation of them resulted in blocked production of geldanamycin (He et al., 2008). The synthesis of β-lactam antibiotic thienamycin was first observed when ThnI protein of LuxR family was induced in Streptomyces cattleya (Rodriguez et al., 2008). Furthermore, the LAL–proteins are large ATPbinding regulators constituting a noteworthy branch of LuxR family (De Schrijver and De Mot, 1999). The LAL–subfamily members usually contain an additional ATP binding domain at the N-terminus, which is responsible for the interaction with the inducers, maltotriose and ATP (Danot, 2001; Baral et al., 2018). Series of studies have confirmed that LAL regulators play important roles in activating the silent gene clusters. Overexpression of LAL-type regulatory factor samR in Streptomyces ambofaciens activated a 150 kb cryptic gene cluster, and promoted the discovery of a glycosylated macrolide product with anti-cancer activity (Laureti et al., 2011). Similarly, LAL-type transcriptional regulators of GdmRI and GdmRII in S. hygroscopicus are positive to geldanamycin biosynthesis (Martin et al., 2019). Furthermore, the LAL-like 63-amino acid protein AfsS and its homologs have been shown to relate closely with the induction of avermectin in S. avermitilis (Lee et al., 2000), pikromycin in S. venezuelae (Maharjan et al., 2009), and DXR in S. peucetius (Parajuli et al., 2005).

Manipulation of the regulators in cryptic gene clusters probably the most convenient way to activate the pathway. But for the uncertainty of regulatory network, the success of activation is case by case.

# Activation the Cryptic Gene Clusters by Ribosome Engineering

Ribosome engineering, first proposed by Kozo Ochi, focuses on modifying ribosomal or RNA polymerases whose structural and functional changes related to the synthesis of secondary metabolites (Ochi et al., 2004). The introduction of specific mutations into the ribosome elements triggers structural changes to ribosome with affected protein synthesis, which ultimately influences secondary metabolism of the microorganism. Targeted strains are screened by mutations in antibiotic resistance acting on ribosome without special requirement of equipment or genetic background of the strains. Isoindolinomycin (Idm), an unprecedented bioactive polyketide with a novel isoindolinonecontaining tetracyclic skeleton, was discovered by activation of cryptic gene through screening rifampicin-resistant (rif) mutants from actinobacteria strains (Thong et al., 2018). Piperidamycins production was induced in a gentamicin-resistant mutant of Streptomyces mauvecolor, whereas the wild strain does not produce these metabolites at any detectable level on various media (Hosaka et al., 2009; Baral et al., 2018). Ribosome engineering now is developed into an important technology for improving secondary metabolites with commercial value and stimulating new natural product production, and also contributes to increase in structural diversity of bioactive compounds. Mutations in these ribosomal genes have also been shown to activate the expression of silent or poorly expressed BGCs in S. griseus (Tanaka et al., 2009).

Most above strategies to activate cryptic gene clusters are by manipulating the original host. With developing of large DNA cloning, heterologous expression is extensively used to overexpress target clusters in super host strains (Yamanaka et al., 2014; Greunke et al., 2018; Choi et al., 2019). There are series of reviews on this aspect. Also some heterologous expression was solved with overexpression of regulatory genes by proper promoters in the target clusters (Fazal et al., 2019; Myronovskyi and Luzhetskyy, 2019; Nepal and Wang, 2019; Xu and Wright, 2019).

#### CONCLUSION AND PERSPECTIVES

Streptomyces, which has complex morphological development and cell differentiation, can adapt to diverse environments by ingenious regulation and produce variety of secondary

metabolites. The regulation of antibiotic production has been extensively investigated in the past years. For the complicated interactions among primary and secondary metabolic pathways, elucidation of the regulatory network in antibiotic production still was a tough task for scientist. With further recognizing the secondary metabolism regulatory cascades, novel proposed strategies can greatly promote the production enhancement and drug discovery in Streptomyces. For there are several reviews on the regulation of antibiotic biosynthesis, strategy to enhance production of antibiotic in Streptomyces and new compounds discovery in actinobacteria by Nodwell group and van Wezel group (van Wezel et al., 2009; Yoon and Nodwell, 2014; Rigali et al., 2018; van der Heul et al., 2018). Here we drafted a review on the cascades of regulation and its application in production enhancement and compound discovery (**Figure 1**). We proposed to give an extensive review on the regulation of antibiotic production in Streptomyces.

The complex hierarchical regulation constitutes a major bottleneck to over produce target antibiotic. On the basis of understanding the regulation of antibiotic biosynthesis, rewiring the regulatory network is much more efficient to optimize antibiotic producers than the classical random mutagenesis methods. It has mentioned above that the enhancement of antibiotic titers can be achieved by overexpression of positive regulator, inactivation of negative regulator, tuning feedback and ribosomal engineering. Combination of different strategies to manipulate regulatory genes can achieve higher antibiotic production in both the native and/or heterologous host. It is necessary to achieve maximal level of antibiotic production by systematically rewiring the regulatory network.

With the accumulation of explosive genome sequencing data, it has been estimated that Streptomyces species harbor a huge unexploited potential to produce novel natural products. A new discovery approach-microbial genome mining has been well developed in the past years (Baltz, 2019). How to efficiently activate cryptic gene clusters is still challenging to scientist. Because the interweaved regulatory networks in most cases are not well studied or understood. Elucidation of the regulatory network is crucial to discover new antibiotics.

For efficient manipulation of the regulatory genes, multi-loci simultaneous editing technology is necessary to be invented. The recent application of CRISPR-Cas9 dependent serials genome editing system provides a new opportunity for rapid rewiring regulatory network in Streptomyces (Tong et al., 2019). For precisely tuning the transcription level of regulators, a panel of quantitative promoters should be defined in Streptomyces. This would be practical for tuning the expression of secondary metabolic pathway. Though there are many publications on promoter engineering in Streptomyces. Still serial quantitative promoters are necessary for predictable engineering with synthetic biology strategy in future.

#### AUTHOR CONTRIBUTIONS

HX, XL, ZL, and XZ prepared material for the manuscript. HX, XM, and YL wrote the manuscript. All authors read and approved the final manuscript.

# FUNDING

This study was funded by the Natural Science Foundation of Zhejiang Province to HX (LY19C010002).

#### ACKNOWLEDGMENTS

The assistance of the colleagues of Institute of Biopharmaceuticals during preparing manuscript is gratefully acknowledged.

precursor or efflux of antibiotics.

# REFERENCES




differentiation in Streptomyces chattanoogensis L10, is positively regulated by AdpAch. Appl. Environ. Microbiol. 80, 6879–6887. doi: 10.1128/aem.01849-14


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2020 Xia, Li, Li, Zhan, Mao and Li. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Regulation of Bottromycin Biosynthesis Involves an Internal Transcriptional Start Site and a Cluster-Situated Modulator

Natalia M. Vior, Eva Cea-Torrescassana, Tom H. Eyles, Govind Chandra and Andrew W. Truman\*

Department of Molecular Microbiology, John Innes Centre, Norwich, United Kingdom

Bottromycin is a ribosomally synthesized and post-translationally modified peptide (RiPP) produced by several streptomycetes, including the plant pathogen Streptomyces scabies. There is significant interest in this molecule as it possesses strong antibacterial activity against clinically relevant multidrug resistant pathogens and is structurally distinct from all other antibiotics. However, studies into its efficacy are hampered by poor yields. An understanding of how bottromycin biosynthesis is regulated could aid the development of strategies to increase titres. Here, we use 5<sup>0</sup> -tag-RNA-seq to identify the transcriptional organization of the gene cluster, which includes an internal transcriptional start site that precedes btmD, the gene that encodes the bottromycin precursor peptide. We show that the gene cluster does not encode a master regulator that controls pathway expression and instead encodes a regulatory gene, btmL, which functions as a modulator that specifically affects the expression of btmD but not genes up- or downstream of btmD. In order to identify non-cluster associated proteins involved in regulation, proteins were identified that bind to the main promoter of the pathway, which precedes btmC. This study provides insights into how this deceptively complex pathway is regulated in the absence of a pathway specific master regulator, and how it might coordinate with the central metabolism of the cell.

Keywords: Streptomyces, bottromycin, RiPP, antibiotic, regulation, transcription, RNA-seq, natural product

# INTRODUCTION

Infectious diseases are responsible for one-fifth of deaths worldwide, and antimicrobial resistance is becoming an increasingly serious threat to global public health, which makes the development of new antibiotics a pressing need (Laxminarayan et al., 2016; Martens and Demain, 2017). The effective treatment of multidrug resistant bacteria requires the identification of compounds with either unusual chemical scaffolds or novel molecular targets, capable of overcoming multiple and widespread resistance mechanisms (Laxminarayan, 2014; Newman and Cragg, 2016; Martens and Demain, 2017). One antimicrobial compound of great interest is bottromycin, a peptide antibiotic with promising activity against multi-drug resistant Gram-positive bacterial pathogens (Shimamura et al., 2009; Tacconelli and Magrini, 2017). Bottromycin was first isolated in 1957 from cultures of Streptomyces bottropensis (Waisvisz et al., 1957), and then more recently as a product of

#### Edited by:

Yinhua Lu, Shanghai Normal University, China

#### Reviewed by:

Weishan Wang, Institute of Microbiology (CAS), China Di Sun, Jiangsu Normal University, China

> \*Correspondence: Andrew W. Truman andrew.truman@jic.ac.uk

#### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 07 August 2019 Accepted: 06 March 2020 Published: 26 March 2020

#### Citation:

Vior NM, Cea-Torrescassana E, Eyles TH, Chandra G and Truman AW (2020) Regulation of Bottromycin Biosynthesis Involves an Internal Transcriptional Start Site and a Cluster-Situated Modulator. Front. Microbiol. 11:495. doi: 10.3389/fmicb.2020.00495

the plant pathogen Streptomyces scabies (Crone et al., 2012) and other Streptomyces species (Gomez-Escribano et al., 2012; Hou et al., 2012; Huo et al., 2012). Bottromycin is a ribosomally synthesized and post-translationally modified peptide (RiPP). In its biosynthetic pathway, a precursor peptide (BtmD) undergoes a complex and unprecedented series of modifications catalyzed by enzymes encoded in the bottromycin (btm) gene cluster (**Figure 1**) (Crone et al., 2012, 2016; Gomez-Escribano et al., 2012; Huo et al., 2012; Franz et al., 2017; Schwalen et al., 2017; Sikandar et al., 2019).

Bottromycin is a structurally novel compound that selectively blocks the binding of aminoacyl t-RNAs to the A-site in the 50S subunit of bacterial ribosomes, ultimately leading to protein synthesis inhibition (Otaka and Kaji, 1976, 1981, 1983). This makes bottromycin an interesting lead for development, as it attacks a molecular target that is currently unexploited in the clinic. However, bottromycin is produced in low yields, where titres in wild type producers are below 1 mg/L (Huo et al., 2012). These low production levels complicate both drug development and further biosynthetic studies of these molecules (Huo et al., 2012; Crone et al., 2016). An understanding of how bottromycin biosynthesis is regulated may enable the rational improvement of yield by modifying pathway regulation, especially as the pathway is inefficient in laboratory conditions (Crone et al., 2016; Eyles et al., 2018).

It is very well documented that altering the regulatory networks that control secondary metabolite biosynthesis often leads to significant yield increases, either by overexpressing positive regulators, inactivating repressors, or altering higher levels of regulation in the producer strain (Makitrynskyy et al., 2013; Vior et al., 2014; Yoo et al., 2015; Huang et al., 2016). RiPP biosynthesis, especially in the case of lantibiotics, is often regulated by classic two-component systems or orphan response regulators (Bierbaum and Sahl, 2009; Cooper et al., 2010), but this is not a mechanism common to all RiPP pathways. Systems involving sigma-anti-sigma factor pairs, SARP, LuxR, or TetR regulators have also been documented (Izawa et al., 2013; Flinspach et al., 2014; Zhang et al., 2014; Fernández-Martínez et al., 2015), and more diverse regulatory systems are likely to be found in the future, especially because many RiPP gene clusters do not encode any obvious pathway specific regulators (Bartholomae et al., 2017). The btm gene cluster in S. scabies encodes one potential regulatory protein, BtmL (**Figure 1**) (Crone et al., 2012). This protein is conserved across all characterized bottromycin gene clusters (Crone et al., 2012; Gomez-Escribano et al., 2012; Huo et al., 2012), but nothing is known about how this putative regulator controls bottromycin biosynthesis, nor whether additional regulators have critical roles in btm cluster regulation.

In this work, we apply transcriptomic, proteomic and metabolomic techniques in combination with qRT-PCR and reporter activity experiments in order to obtain key details on the regulation of bottromycin biosynthesis. We show that BtmL is not a master regulator of biosynthesis and instead specifically enhances expression levels of the precursor peptide gene btmD. This occurs in conjunction with a transcriptional start site for btmD that is internal to the preceding gene in the cluster. We show that this pathway is surprisingly complex and provide evidence into how it is regulated in the absence of a pathway specific master regulator.

# RESULTS

#### BtmL Specifically Modulates the Expression of the Precursor Peptide Gene btmD and Is Independent of Cobalt Levels in the Medium

The btm gene cluster in S. scabies encodes a single putative regulator, BtmL, a 20.5 kDa protein that contains a C-terminal domain of unknown function (DUF2087 or PF09860), which has been associated with putative transcriptional regulators and is proposed to bind nucleic acids (Rigden, 2011). To date (July 2019), over 6,500 DUF2087-containing proteins have been sequenced and deposited in Genbank, of which more than 2,000 have additional DNA-binding domains. In silico analysis of BtmL using both Phyre2 (Crone et al., 2012; Kelley et al., 2015) and I-TASSER (Yang et al., 2015) predicted that the N-terminus of the protein has structural homology with SmtB-ArsR-like repressors, and therefore would feature a winged helix-turn-helix (wHTH) domain characteristic of this family of transcriptional regulators

(Busenlehner et al., 2003; Osman and Cavet, 2010; Chakravorty and Merz, 2014). However, no canonical wHTH domain was detected by sequence analysis and the protein appears to lack the conserved residues characteristically involved in ArsR-family metal binding (Osman and Cavet, 2010; Crone et al., 2012). Despite this, previous empirical evidence showed that addition of cobalt(II) to the production medium increased bottromycin yields (Crone et al., 2012), inferring that the ArsR-like structure of BtmL may control cluster expression via a metal-binding mechanism. This hypothesis was especially compelling given that several genes in the btm cluster encode class B radical SAM methyltransferases, which are cobalamin-dependent enzymes (Bauerle et al., 2015).

To investigate the role of BtmL in bottromycin biosynthesis, a mutant S. scabies strain carrying an in-frame deletion of btmL was generated (Crone et al., 2012) and its ability to produce bottromycin was assessed using liquid chromatography - mass spectrometry (LC-MS). Inactivation of btmL did not abolish bottromycin biosynthesis, but we could observe a moderate and consistent decrease in production levels to approximately 40% of wild type (WT) levels (**Figure 2A**). This result suggests that btmL acts as a positive modulator of bottromycin biosynthesis, but is not the master activator of the pathway. WT levels of bottromycin production were restored in 1btmL upon in trans complementation with a copy of btmL under the control of the constitutive promoter ermE∗p (Bibb et al., 1985), confirming that the phenotype was due to the deletion of btmL. Surprisingly, when that same btmL expression construct was introduced in the WT strain (generating strain WT + L), there was no increase in bottromycin levels (**Figure 2A**).

To evaluate quantitatively whether btmL could be acting as a transcriptional regulator, qRT-PCR experiments were carried out with RNA samples from both the WT and 1btmL strains (**Figure 2B**). The btm gene cluster is organized in two divergent groups of genes: btmA and btmB on one strand and btmCbtmM on the other strand (**Figure 1**). We therefore measured the expression of btmB from the smaller block, and btmC, btmD (the gene encoding the precursor peptide), btmE and btmL from the rest of the cluster. Expression values of these genes were normalized using the expression of hrdB, which encodes the principal sigma factor of RNA polymerase. A time course experiment determined that the optimal time to measure btm gene expression was at 72 h post-inoculation of bottromycin production medium (**Supplementary Figure S1**). At this time point, it could be observed that deletion of btmL has no significant effect on most of the genes tested, with one notable exception: btmD, the gene encoding the bottromycin precursor peptide. Transcription of btmD is reduced to approximately 40% of WT levels, which correlates with the reduction in bottromycin production (**Figure 2B**). Furthermore, qRT-PCR analysis of the WT + L strain revealed that transcription of btmD is significantly increased in this strain, confirming that BtmL is a specific and positive regulator of btmD. Interestingly, upstream and downstream genes (btmB, btmC, and btmE) are only slightly overexpressed or not overexpressed at all, suggesting the presence of an alternative operon encompassing only btmD.

After confirming that btmL is involved in bottromycin regulation, we assessed whether cobalt(II) levels influence pathway productivity, and whether this effect was mediated by btmL. Bottromycin production was measured in S. scabies WT and 1btmL cultured in bottromycin production medium (BPM) supplemented with different concentrations of CoCl<sup>2</sup> (**Figure 3A**). In absolute terms, bottromycin production increased upon addition of cobalt(II), reaching a maximum at 15–25 µg/mL CoCl<sup>2</sup> after which production steadily decreased

FIGURE 2 | Effect of btmL on bottromycin production and btm gene transcription. (A) LC-MS analysis of bottromycin production in S. scabies WT, 1btmL (1L), 1L complemented with a copy of btmL (1L + L) and the WT strain overexpressing btmL (WT + L). Error bars represent the standard deviation of the average production values in biological triplicates, which are normalized by culture growth. (B) qRT-PCR analysis of the transcription of representative btm genes in the WT (purple bars), 1btmL (yellow bars) and WT + L (green bars) at 72 h of growth in BPM. Expression values are relative to the expression of the target gene in the WT, which was set at 1. Transcription of hrdB was used to normalize the expression levels across samples. Error bars represent standard error of the mean from triplicate experiments, and the asterisk represents a statistically significant difference in btmD expression between wt and the mutant ∆btmL (p-value < 0.001).

and represented on the secondary axis. Error bars represent the standard deviation of biological triplicate data. The concentration of cobalt used for the qRT-PCR experiment in (B) appears highlighted with a green square. (B) qRT-PCR analysis of the expression of btm genes in the WT (purple bars) and 1btmL mutant (yellow bars) at 72 h of growth in BPM with and without added cobalt. Expression values are relative to the expression of the target gene in the WT without added cobalt, which was set at 1. Transcription of hrdB was used to normalize the expression levels across samples and error bars represent standard error of the mean from triplicate experiments.

until almost disappearing at 200 µg/mL CoCl2. However, the increase in production simply corresponded to a proportional increase in mycelial growth upon addition of up to 25 µg/mL of cobalt(II) in the medium (**Figure 3A**). At higher cobalt(II) concentrations mycelial growth was sustained, so the drop in production was probably due to a general inhibitory effect of excess metal in the medium, a phenomenon reported for the biosynthesis of other antibiotics (Rogers and Birnbaum, 1974; Abbas and Edwards, 1990). The growth-dependent production increase was also noticeable in 1btmL, which followed a similar pattern, albeit at a reduced level of production. Moreover, a comparison of qRT-PCR data for the expression levels of several pathway genes in the presence or absence of added cobalt(II) (**Figure 3B**) showed that these are nearly identical in either condition for both the WT and 1btmL strains, discarding any specific regulatory effect of this metal. This result is in accordance with the observation that btmL lacks the conserved metal-binding residues characteristic of ArsR-SmtB regulators (Osman and Cavet, 2010; Crone et al., 2012).

These results ruled out cobalt(II)-dependent pathway regulation by BtmL, which is instead a cobalt-independent transcriptional modulator that selectively increases btmD transcript levels in order to generate high levels of the precursor peptide in a feed-forward loop. Other positive loop systems have been reported in the biosynthesis of several RiPPs like microbisporicin and planosporicin, where the autoinduction mechanism relies of the detection of small quantities of precursor peptide by a sigma-antisigma complex (Sherwood and Bibb, 2013; Fernández-Martínez et al., 2015), or cinnamycin, where production is launched after onset of immunity to the compound in the producer strain (O'Rourke et al., 2017).

# Overexpression of Exporter BtmA Has a Moderate Positive Effect on Bottromycin Production

The discordance between the increase in btmD expression and the lack of change in bottromycin production in the WT + L strain (**Figure 2A**) led us to postulate that the potential increase in available BtmD might not be effectively channeled by the pathway and exported out of the cell, leading to increased degradation of pathway intermediates and generation of shunt metabolites. Supporting this hypothesis, previous work showed that overexpression of botT, the major facilitator superfamily (MFS) transporter gene (Kumar et al., 2016; Quistgaard et al., 2016) in the bottromycin cluster from Streptomyces sp. BC16019 increased bottromycin production in a heterologous host, although it still did not reach wild type production levels (Huo et al., 2012). We therefore assessed whether overexpressing the homolog of this gene (btmA, 88% identity to botT) in the S. scabies btm gene cluster might increase the efficiency of the pathway and self-resistance to potentially toxic levels of the antibiotic.

Expression of a second copy of btmA under the control of the constitutive promoter ermE∗p provided a 25% increase in bottromycin production with respect to the WT strain (**Figure 4A**), an effect that was maintained, but not improved, when it was expressed in WT + L (generating WT + L + A). When the quantification was extended to other bottromycin related metabolites described previously (Crone et al., 2016; Eyles et al., 2018) this pattern was still observed for the total set of mature bottromycins (those including the main posttranslational modifications) (**Figure 4A**). However, in the case of the pathway

intermediates and shunt metabolites the trend was not present, with WT + L and WT + A producing more of them than the WT and the double overexpression strain WT + L + A (**Figure 4B**). This result indicates that efficiency can be increased to some extent by improving export of the molecule, but this is not the major bottleneck for bottromycin production. The main limiting factor for bottromycin production is likely to be the post-translational maturation of the precursor peptide instead, as we previously showed that the bottromycin pathway inefficiently stalls at numerous biosynthetic steps, even in the WT strain (Crone et al., 2016).

# Mapping of the Transcriptional Organization of the btm Cluster Identifies a Secondary Transcriptional Start Site for btmD

Given that btmD appears to be situated within a larger operon, the specific increase in btmD transcription following btmL overexpression was unexpected and prompted us to investigate the transcriptional organization of the btm gene cluster. This cluster contains 13 genes, where 11 (btmC-M) are arranged in the same orientation and appear tightly clustered, with btmE to btmJ physically overlapping and btmK starting 8 bp after the btmJ stop codon. To assess the possibility of co-transcription of the remaining genes, RT-PCR experiments were performed on 72 h cultures of both WT and 1btmL strains with specific primer pairs for the intergenic regions btmA-B, btmC-D, btmD-E, and btmK-L (**Figure 5A**). These showed co-transcription of all of the tested regions except btmA-B, indicating that genes btmC-M co-transcribe and can behave as a single 14.5 kb operon, whereas btmA and btmB are transcribed separately. The gene cluster was analyzed for putative terminators using WebGeSTer (Mitra et al., 2011). Consistent with the RT-PCR results, terminators were predicted spanning up to 90 bp away from the stop codon of btmA (1G = −19.5), between btmA and btmB (1G = −19.43) and 149 bp away from the stop codon of btmM (1G = −19.27) (**Figure 5A**).

This evidence for btmC-M co-transcription did not exclude the possibility of alternative operons, and the specific increase in btmD transcription following btmL overexpression hinted toward an alternative transcription start site (TSS) for btmD that would explain how it is differentially regulated from its surrounding genes. In order to map the TSSs for the whole cluster we employed 5<sup>0</sup> -tag-RNA-seq (also called tagRNAseq) (Innocenti et al., 2015) that enables the identification of TSSs in an untargeted, genome-wide fashion. This powerful technique has the advantage of differentially distinguishing primary transcripts (those generated by an RNA polymerase and therefore coming from true TSSs) from processed transcripts, which arise upon degradation or RNase mediated cleavage of the original transcripts at specific processing sites (PSs). In this particular case, with an operon spanning over 10 kb, degradation or post-transcriptional modification of the resulting transcript could also account for the difference between btmD expression and the remainder of the btmC-M operon.

5 0 -tag-RNA-seq reads mapped to the S. scabies 87-22 genome clearly showed the presence of three TSSs in the btm cluster: one preceding btmB, one before btmC, and an extra one preceding btmD and within the coding region of the upstream gene, btmC (**Figure 5B**). Unexpectedly, no TSS was found before btmA, in contrast with what was expected from the

#### FIGURE 5 | Continued

fmicb-11-00495 March 24, 2020 Time: 16:1 # 7

location of predicted transcription terminators -solid line- and attenuators -dashed line- in the cluster). (B) Transcription start site mapping of the btm cluster using 5 0 -tag-RNA-seq. The plots (generated in IGB) show the mapping of reads with TSS tags in the btm cluster in the WT (top) and 1btmL (bottom) strains. Sharp peaks corresponding to mapped reads precede btmB, btmC, and btmD, along with some non-specific transcription initiation in the btmD region. The vertical axis in the plots represents read counts and the horizontal axis maps to the gene cluster shown in (A). The direction of transcription from the identified TSSs is represented with black arrows in the cluster. (C) Precise mapping of TSS and potential promoter regions in the btmB-C intergenic region. btmB and btmC start codons are marked in green, and the TSSs identified for btmB and btmC using both 5<sup>0</sup> tag-RNA-seq and 5<sup>0</sup> -RACE are labeled in red. −10 and −35 promoter regions for btmC (predicted with BPROM) and btmB (annotated manually) are represented in light blue. (D) Mapping of btmD TSS (red) within the coding sequence of btmC (shadowed in gray, with its stop codon highlighted in bold).

intergenic RT-PCR result (**Figure 5A**). To independently confirm the existence of these TSSs, we performed 5<sup>0</sup> -RACE (Rapid amplification of cDNA ends) experiments for btmBTSS, btmCTSS, and btmDTSS, which yielded identical results to the 5<sup>0</sup> tag-RNAseq (**Figures 5C,D**). btmCTSS is in position -32 with respect to the btmC start codon, and a prediction of putative σ70 promoters in the btmB-btmC intergenic region using BPROM (Solovyev and Salamov, 2011) identified a nearly canonical −35 sequence (TTGACT) and a poorly conserved −10 region (TCCTACGGT) that are consistent with the location of the btmCTSS (**Figure 5C**). It was not possible to confidently identify −10 and −35 boxes for the btmBTSS using BPROM or other prediction tools, so potential sequences were identified manually (**Figure 5C**). The btmDTSS is located 348 bp away from the start of btmD and 215 bp before the end of btmC coding sequence (**Figure 5D**), but we could not identify any suitable −10 or −35 regions for this TSS, either manually or with predictive tools. Unexpectedly, btmDTSS appears both in the WT and in 1btmL, albeit with lower intensity in 1btmL (**Figure 5B**), which indicates that although btmL does affect the transcription of btmD, it is not essential for the expression of this alternative transcript.

Interestingly, the 5<sup>0</sup> tagRNA-seq results showed that at 72 h of growth there is an extremely high level of transcription of some btm genes, in particular btmD. Transcription initiation preceding this gene seems rather promiscuous with some background noise (**Figure 5B**). Strikingly, an analysis of the full transcriptome to assess levels of gene expression as reads per kilobase per million reads (RPKM) and transcripts per million (TPM) (Wagner et al., 2012), showed that btmD and btmC are the 4th and 9th most highly transcribed genes in the genome at this time point (**Supplementary Datasheet S1**). Transcription after btmD drops considerably, showing that even upon overexpression of btmL, expression of btmE did not increase in a comparable way to btmD, in accordance with the qRT-PCR results (**Figure 2B**). The transcriptomic data also shows that all downstream genes in the operon are transcribed at lower levels than their preceding genes. This may be explained by a transcription attenuator downstream of btmD that likely serves to dampen transcription of downstream genes in the cluster. Accordingly, a putative terminator sequence between btmD and btmE (1G = −21.22) is predicted by WebGeSTer (Mitra et al., 2011).

#### Genome-Wide Analysis of Transcriptional Start Sites in S. scabies

In RiPP biosynthesis, it appears to be beneficial to evolve mechanisms to selectively increase precursor peptide production in relation to catalytic proteins, given that precursor peptides are structural rather than catalytic. The mechanisms by which this happens are poorly understood, especially in pathways where the precursor peptide is apparently co-transcribed with upstream genes encoding catalytic proteins. The secondary transcriptional start site for btmD provides a previously unobserved mechanism by which this can occur. To assess whether this strategy was apparent in other RiPP gene clusters in S. scabies, we assessed their transcriptional organization using the 5<sup>0</sup> tagRNAseq dataset. This showed that for all of the putative RiPP clusters identified in the S. scabies genome, the precursor peptide gene always has its own TSS (**Supplementary Figures S2, S3**). This happens even when the gene is preceded by genes predicted to encode tailoring enzymes (**Supplementary Figure S2**). Notably, in the case of SCAB\_32021, which is tightly clustered with a preceding gene encoding a putative methyltransferase, its dedicated TSS overlaps with the end of this gene, in a similar scenario to btmC and btmD (**Supplementary Figure S2**). This suggests that having a dedicated TSS could be a widespread strategy for obtaining the appropriate stoichiometry of precursor peptide and tailoring enzymes in those cases where the precursor peptide gene is not at the beginning of an operon.

#### Reporter Assays Show Strong Transcription From the btmC Promoter Yet Negligible Transcription From the Putative btmD Promoter

To further characterize the main promoter regions in the pathway, promoter activity was measured using the β-glucuronidase reporter gene gusA (Myronovskyi et al., 2011). A 414 bp fragment containing the whole intergenic region btmB-btmC, including btmCTSS but not its RBS, was cloned in vector pIJ10742 (Feeney et al., 2017) to create a transcriptional fusion with gusA. This construct was introduced in S. scabies WT, 1btmL, and WT + L to test whether the different genetic backgrounds would affect promoter activity. As a positive control, vector pIJ10741 carrying gusA under the control of the constitutive promoter ermE∗p (Feeney et al., 2017) was used, while empty pIJ10742 with no promoter was employed as a negative control. btmCp proved to be an exceptionally strong promoter in both in solid and in liquid culture (**Figure 6**), and at 72 h of growth its activity was comparable to ermE∗p. This result correlates with the high levels of expression of btmC observed in the transcriptomic analysis. The same gusA reporter experiment was performed with a 545 bp region that was predicted to contain the promoter preceding btmDTSS, but no reporter activity could be detected in any of the genetic backgrounds tested (**Figure 6**).

Therefore, further reporter constructs were made that progressively extended the region tested until it included the whole coding sequence of btmC (but not btmCp), but these provided identical negative results (**Supplementary Figures S4, S5**). This result pointed to the possibility that transcription from btmDTSS is not driven from its own promoter, but rather from btmCp, where some sort of secondary structure might help enhance transcription of this particular gene. To test this, we assessed the promoter activity of the whole region (btmCp, btmC, and the intergenic region before btmD) and compared its promoter activity to that of the original btmCp reporter. Two promoter versions were generated, one containing the WT sequence, and another one where several synonymous point mutations were introduced in the region of btmDTSS via yeast refactoring (**Supplementary Figure S4**) in order to disrupt possible DNA secondary structures in this region. Although these new transcriptional fusions were more active than all the other sequences lacking btmCp, they showed drastically reduced activity when compared to the original btmCp reporter, indicating that the additional DNA sequence dampened the promoter activity of btmCp. The mutated btmDTSS region did not significantly affect this phenotype (**Supplementary Figures S4, S5**).

### Identification of Potential Pathway Regulators Using DNA Affinity Capture Assays

Our transcriptional analysis combined with the data from S. scabies 1btmL and WT + L show that the master regulator(s) of the btm cluster are not encoded by the btm cluster and are therefore unlikely to be pathway-specific. Identification of such regulators could aid in improving the yield of bottromycin. Given the difficulty in predicting binding sequences for transcriptional activators in streptomycetes without extensive transcriptomic and metabolic datasets (Iqbal et al., 2012; Hwang et al., 2019), the intergenic region btmB-C was used as a probe in a DNAbinding protein capture experiment. This region was amplified using biotinylated primers and the resulting fragment was immobilized on streptavidin coated magnetic beads. The beads were then incubated with lysates of 72 h cultures of S. scabies WT grown in BPM. As a negative control, the same promoter region was incubated with lysates of cultures grown in GYM medium, where the cluster is not expressed and therefore there is no bottromycin production (**Supplementary Figure S6**). After incubation and several washes with competitor DNA to eliminate proteins binding non-specifically to the probe, the remaining proteins were eluted, washed again and analyzed using quantitative proteomics. A total of 321 proteins were identified in the 2 sets of samples, 120 of which appeared in both conditions, whereas 177 where present exclusively in the BPM samples and only 24 were specific to GYM samples. These results were filtered to identify annotated regulatory proteins or uncharacterized proteins binding to btmCp specifically in bottromycin production conditions. This reduced the candidates to 21 proteins (**Supplementary Table S3**), of which three were annotated as regulators, while the remainder were uncharacterized proteins.

Based on the intensity of the signal obtained or the biological relevance of the hit, we selected four promising candidates for further genetic analysis to assess their involvement in bottromycin biosynthesis: MtrA (SCAB\_55281), GlnK (SCAB\_61751), a putative regulatory protein (SCAB\_ 85931) and an uncharacterized protein that binds specifically to btmCp under production conditions (SCAB\_51451). MtrA forms a three-component system with the histidine kinase MtrB and LpqB, a lipoprotein that is involved in signal transduction (Hoskisson and Hutchings, 2006; Nguyen et al., 2010). This system is widely conserved in Actinobacteria, where it has been shown to regulate both development and antibiotic production in Streptomyces venezuelae and in Streptomyces coelicolor (Som et al., 2017; Zhang et al., 2017). SCAB\_61751 encodes a homolog

of GlnK, a signal transduction protein with a central role in nitrogen metabolism (Thomas et al., 2000). SCAB\_ 85931 is annotated as a putative regulatory protein and contains domains characteristic of RsbR-like anti-anti-sigma factors. Rsb proteins form part of the "stressosome," which responds to bacterial stress and ultimately leads to the activation of the alternative sigma factor σ B (Lee et al., 2004). SCAB\_51451 contains no conserved domains. In order to test the role of these genes in bottromycin biosynthesis, each of them was constitutively overexpressed under the control of ermE∗p in S. scabies WT using the integrative plasmid pIB139-RBS, a derivative of pIB139 (Wilkinson et al., 2002). The potential involvement of the MtrAB two-component system prompted us to also express mtrB, which encodes the cognate histidine kinase. Despite the evidence provided by DNA binding, the bottromycin titres were comparable to the WT under the conditions tested (**Supplementary Figure S7**).

# DISCUSSION

An understanding of genetic regulation is required to inform rational approaches to overproduce or engineer the biosynthesis of medicinally promising specialized metabolites. In the case of bottromycin and other RiPPs, their ribosomal origin enables the rapid generation of analogs by precursor peptide mutagenesis, which can lead to the rapid generation of libraries of bioactive analogs (Yang et al., 2018). A distinct challenge for the regulation of RiPP biosynthesis is to enable the expression of a sufficiently large amount of the precursor peptide to support RiPP production, given that this is the substrate for the pathway, while producing other pathway proteins in catalytic quantities. Numerous genetic strategies have evolved that enable this differential production of the precursor peptide. Most commonly, precursor peptide genes are encoded at the beginning of RiPP operons, where they can be followed by imperfect transcriptional terminators (sometimes called "attenuators") that dampen transcript levels of the following genes (Severinov et al., 2007; Foulston and Bibb, 2010; O'Rourke et al., 2017). Alternatively, some RiPP gene clusters contain multiple precursor peptide genes (Li et al., 2010), while others feature a single precursor peptide containing multiple core peptides (Leikoski et al., 2013; Santos-Aberturas et al., 2019).

There is substantial interest in the RiPP antibiotic bottromycin due to its novel structure, unique molecular target and its activity toward multi-drug resistant pathogens, including methicillin-resistant Staphylococcus aureus (MRSA) and vancomycin-resistant enterococci (VRE) (Shimamura et al., 2009; Tacconelli and Magrini, 2017). Identification of the bottromycin biosynthetic gene cluster (Crone et al., 2012; Gomez-Escribano et al., 2012; Huo et al., 2012) showed that the precursor peptide gene, btmD, appeared to be encoded in the middle of a larger operon. In this study, we used 5<sup>0</sup> -tag-RNA-seq and 5<sup>0</sup> -RACE to reveal that the bottromycin gene cluster features an internal transcription start site that enhances transcription of btmD in relation to the remainder of the gene cluster (**Figure 5B**). The use of 5<sup>0</sup> -tag-RNA-seq provides confidence that this is a primary transcript and not simply the result of degradation of a longer transcript. In bottromycin production conditions, btmD is one of the most highly expressed genes in the genome. The presence of an internal TSS (btmDTSS) to boost precursor peptide transcript levels has not previously been reported but our data indicate that it may be a common regulatory feature in RiPP biosynthesis. Evidence for this was provided by an analysis of other RiPP gene clusters in the S. scabies genome, which showed that their precursor peptides also have dedicated TSSs, in some cases overlapping with the coding sequence of the preceding genes, as with btmD. The occurrence of intragenic TSS is not a rare phenomenon, and recent TSS mapping work in Streptomyces clavuligerus has revealed the presence of 155 intragenic TSSs in its genome out of a total of 2659 identified (Hwang et al., 2019) and even higher numbers have been reported for other bacteria (Wade, 2015; Brito et al., 2017).

The gene cluster encodes one putative transcriptional regulator, BtmL (**Figure 1**), but deletion and over-expression experiments showed that this does not function as a master regulator of biosynthesis (**Figure 2**) and instead functions to increase transcript levels of btmD. Notably, btmDTSS is still present in S. scabies 1btmL (**Figure 5B**), indicating that BtmL does not solely control this regulatory feature. Unexpectedly, a gusA reporter experiment did not reveal any promoter activity associated to btmDTSS, indicating that transcription from this TSS might instead be driven from btmCp. This could also explain why mobility shift assays with purified BtmL using this region as probe were unsuccessful (data not shown). However, a transcriptional fusion containing the entire region from btmCp to the region before btmD provided lower GusA reporter activity than only btmCp. This unexpected result indicates that a complex and unusual system is functioning to regulate btmD gene expression, and further supports the theory that btmDTSS does not have its own dedicated promoter. These data highlight the need for further studies to elucidate the precise role of internal TSSs in RiPP biosynthesis.

Despite these null results, the importance of btmDTSS is supported by our prior work that showed that altering the region preceding btmD practically abolishes bottromycin production (Eyles et al., 2018), either by deletion of btmC or by swapping the positions of btmC and btmD in the gene cluster. Similarly, DNA affinity capture with the btmB-C region did not pull down BtmL (**Supplementary Table S3**). Therefore, the mechanism by which BtmL modulates btmD transcription remains cryptic. One possible hypothesis would be the interaction with a cis-regulatory region in the 5<sup>0</sup> untranslated region (UTR) of btmD mRNA. Transcripts with UTRs bigger than 100–150 bp are usually considered "extended-leadered mRNAs" and are likely candidates to contain such a regulatory feature (Brito et al., 2017; Hwang et al., 2019). The btmD 5'UTR (347 bp) is within this category, but searches in this sequence against the Rfam database for any known RNA structures did not yield a clear result.

We sought to use our knowledge of pathway expression to increase bottromycin titre, firstly by over-expressing btmL either alone or along with btmA, which encodes a MFS transporter. While this increased btmD transcription (**Figure 2B**), it did not lead to higher bottromycin yields, indicating that btmD levels are

not the rate limiting factor for pathway expression. Similarly, in S. scabies, export by BtmA is not a bottleneck in this pathway. We had previously shown that the bottromycin pathway produces a significant amount of side-products in both S. scabies and a heterologous expression system (Crone et al., 2016; Eyles et al., 2018). Pathway refactoring in this heterologous system showed that constitutive high expression of the btm cluster leads to high yields of shunt metabolites and low titres of mature products, while controlled expression using a riboswitch increases overall production with higher ratios of mature bottromycins versus shunt metabolites. This is consistent with a study that refactored the bottromycin gene cluster from Streptomyces sp. BC16019 (Horbal et al., 2018), which also concluded that high levels of transcription did not fully correlate with increases in production. Promoter choice is a key factor for optimizing production of RiPPs, and can also depend on the timing of gene expression, such as in the heterologous production of telomestatin in Streptomyces avermitilis (Amagai et al., 2017).

There is surprisingly little known about the regulation of specialized metabolites that do not contain master regulators in their gene clusters. A rare example is erythromycin biosynthesis in Saccharopolyspora erythraea, which is regulated by BldD, a key regulator of actinomycete development, as well as two other regulatory genes which are not located within its biosynthetic cluster (Chng et al., 2008; Kirm et al., 2013; Wu et al., 2014). We sought to identify the master regulators of bottromycin biosynthesis using a DNA affinity capture experiment with the intergenic region between btmB and btmC, which contains the major promoters that control the pathway. This provided a series of promising candidate proteins, including the well-characterized regulatory proteins MtrA and GlnK. However, over-expression of these genes also did not lead to increased bottromycin production in S. scabies. This negative result does not disprove the involvement of these regulators, in particular in the case of MtrA and GlnK. As a member of classical two-component system the activity of the response regulator MtrA is dependent on its phosphorylation state (Kenney, 2002; Desai and Kenney, 2017). Overexpression of the gene, if it is not appropriately phosphorylated, might not be enough to reveal a phenotype. GlnK is a PII family signal transducer protein that modulates the activity of multiple biosynthetic and regulatory proteins in nitrogen metabolism (Arcondéguy et al., 2001; Gerhardt et al., 2015; Shimizu, 2016). It would make sense that nitrogen levels would exert an effect on bottromycin biosynthesis, due to the peptidic nature of bottromycin and in fact a paralog of this protein has been shown to affect secondary metabolite production in S. coelicolor (Waldvogel et al., 2011). These proteins also rely on post-translational modifications to properly function, which adds a layer of complication in their regulatory functions that may dampen the effect of their overexpression. Further studies will be necessary to fully characterize the role of these regulators, as well as the uncharacterized proteins identified in this work, on bottromycin production.

In summary, we have revealed that the regulation of bottromycin biosynthesis is surprisingly complex and features interplay between global regulatory proteins and a cluster-situated modulator, BtmL. Precursor peptide transcription is enhanced in relation to surrounding genes by BtmL and a cryptic internal transcriptional start site. This single-gene internal transcript makes the precursor peptide gene one of the most highly transcribed genes in the S. scabies genome. A wider assessment of the S. scabies genome shows that this strategy could represent a widespread regulatory mechanism for the expression of RiPP precursor peptide genes.

### MATERIALS AND METHODS

#### Chemicals and Molecular Biology Reagents

Unless otherwise specified, antibiotics, media components and chemical reagents used in this work were purchased from Sigma-Aldrich (United Kingdom) with the exception of soy flour, which was purchased from Holland & Barret (United Kingdom). Enzymes and molecular biology kits were purchased from New England Biolabs and Promega Healthcare, respectively.

#### Bacterial Strains and Culture Conditions

The following strains were used in this work: S. scabies DSM 41658, both the wild-type (WT) and the mutant strain 1btmL (1L), whose construction was described previously (Crone et al., 2012). Additionally, Escherichia coli K-12 strain DH5α (Invitrogen) was used for plasmid propagation and DNA manipulation, and the methylation deficient strain E. coli ET12567 containing pUZ8002 (Paget et al., 1999) was used for intergeneric conjugal transfer of genetic material to S. scabies, which was performed following standard procedures (Kieser et al., 2000). E. coli culture media used in this work are described in Sambrook et al. (1989). Several Streptomyces culture media were used: mannitol soya flour medium (SFM) (Kieser et al., 2000) was used for Streptomyces propagation and conjugations, and instant potato mash agar [2% Smash (Premier Foods), 2% agar] was used to grow S. scabies for spore harvesting. GYM medium (0.4% glucose, 0.4% yeast extract, and 1.0% malt extract, in Milli-Q water) and bottromycin production medium (BPM: 1% glucose, 1.5% soluble starch, 0.5% yeast extract, 1.0% soy flour, 0.5% NaCl, and 0.3% CaCO3, in Milli-Q water) were used for bottromycin production. Production experiments were performed as follows: 30 µL of concentrated spores were used to inoculate 10 mL of GYM medium in 50 mL flasks and were incubated for 48 h at 30◦C and 250 rpm. 250 µL of seed culture were used to inoculate 10 mL BPM in 50 mL Falcon tubes covered with foam bungs instead of caps. Alternatively, 1 mL of seed culture was used to inoculate 50 mL of BPM in 250 mL flasks containing a spring. When necessary, cultures were supplemented with appropriate concentrations of CoCl<sup>2</sup> (25 µg/mL unless otherwise stated). Triplicate production cultures were incubated for 5–6 days at 28◦C and 230 rpm, at which point samples were collected and processed immediately, or frozen at −20◦C until further processing and analysis. When antibiotic selection was necessary, culture media were supplemented with the appropriate antibiotics at the following final concentrations: kanamycin at 50 µg/mL, apramycin at 50

µg/mL, hygromycin at 50 µg/mL, chloramphenicol at 25 µg/mL, and nalidixic acid at 25 µg/mL.

#### Gene Expression Plasmids Construction

All the primers used for gene amplification and generation of the following constructs are listed in **Supplementary Table S1**. The gene expression vectors used in this work were pIB139- RBS-btmD (Crone et al., 2016) and pIJ10257 (Hong et al., 2005), which are both integrative plasmids containing the constitutive ermE<sup>∗</sup> promoter. pIB139-RBS-btmD is a derivative of pIB139 (Wilkinson et al., 2002) that carries the btmD gene preceded by a ribosome binding site (RBS) and an NdeI restriction site installed to facilitate cloning and improve gene expression (Crone et al., 2016). This plasmid was linearized using restriction enzymes NdeI and EcoRI, releasing btmD and allowing the introduction of the gene of interest, either via ligation of an EcoRI/NdeI treated PCR fragment or by Gibson assembly, following published protocols (Crone et al., 2012). The ligation method was used to introduce btmL and generate pIB139-RBS-btmL, which was used to complement the mutant 1btmL and to overexpress this gene in the WT strain (WT + L). The Gibson assembly method was used to introduce mtrA, mtrB, glnB, SCAB\_85931, and SCAB\_51451, into pIB139-RBS for gene overexpression the WT strain. An empty version of pIB139-RBS was generated to use as a control in production experiments, via Gibson assembly, using the NdeI/EcoRI linearized vector and primers pIB-RBS\_fw and pIB-RBS\_rv, carrying the RBS sequence. pIJ10257 was used for the expression of btmA. This gene was amplified using primers btmA-start and btmA-end, digested with NdeI and HindIII and ligated into NdeI/HindIII digested pIJ10257. Correct construction of all expression plasmids was confirmed using colony PCR and sequencing, after which they were transferred to E. coli ET12567/pUZ8002 cells for conjugation into S. scabies (WT or 1btmL as appropriate). pIB139 integrates in the φC31 phage integration site, and pIJ10257 integrates in the φBT1 site, which allowed for the simultaneous overexpression of two genes when necessary. Correct integration of the plasmids was verified by colony PCR using ermEp\_chk1 in combination with the reverse amplification primer for each of the genes. As a control for the production experiments, a strain carrying the corresponding empty vector was generated in each case.

#### LC-MS Analysis of Bottromycin Production

Bottromycin production culture samples (1 mL) were extracted with an equal volume of methanol, mixed with shaking for a minimum of 10 min. The mixtures were then centrifuged for 4 min at 13,000 rpm to pellet cellular material and other particulate contaminants. 2 µL of the resulting supernatants were injected onto a Phenomenex Kinetex 2.6 µm XB-C18 column (50 mm × 2.1 mm, 100 Å) attached to a Shimadzu Nexera X2 UHPLC and eluted with a linear gradient of 5 to 95% acetonitrile (ACN) in water + 0.1% formic acid (FA) over 6 min, with a flow-rate of 0.6 mL/min. MS data were obtained in positive mode using a Shimadzu IT-TOF mass spectrometer coupled to the UHPLC and analyzed using LabSolutions software (Shimadzu). Bottromycin production was plotted in peak area units. To normalize production values across samples, culture growth was quantified by measuring DNA concentration with an adaptation of the Burton diphenylamine colorimetric assay (Zhao et al., 2013).

#### Isolation of Total RNA

Two milliliter samples of S. scabies liquid cultures were harvested after 72 h incubation in conditions as described above. Samples were washed with an equal volume of RNAlater (Thermo Fisher Scientific) and stored at −80◦C until further processing. RNA was then extracted following previous protocols (Crone et al., 2016), resuspending the mycelium in 1 mL of RLT buffer from the RNeasy Kit (Qiagen) and homogenizing the sample in lysing matrix B tubes (MP Biomedicals) using a FastPrep instrument and a program of 3 × 30 s pulses at 6 m/s with 1 min cooling intervals on ice. The lysates were then centrifuged at 13,000 rpm and 700 µL of supernatant from each sample were then transferred to spin tubes from the RNeasy Kit to undergo purification following manufacturer instructions. Chromosomal DNA contamination was eliminated with oncolumn DNase I treatment (Qiagen) and a further cleanup step using TURBO DNA-free Kit (Ambion, Invitrogen). RNA concentration in the samples was quantified measuring A<sup>260</sup> using a Nanodrop. In the case of the samples for 5<sup>0</sup> tag-RNAseq, further quantity and integrity measurements were performed with RNA ScreenTape (Agilent).

#### RT and qRT-PCR Analyses

Both RT-PCR and qRT-PCR analyses were carried out using 250 ng of total RNA as template in a two-step protocol. The first step consisted of cDNA synthesis with the QuantiTect Reverse Transcription Kit (Qiagen), as detailed in the manufacturer's instructions. For RT-PCR analyses the cDNA was then used as a template for PCR reactions using Taq polymerase and specific primers for the regions to test (**Supplementary Table S1**). The amplification conditions were as follows: initial denaturation at 95◦C for 3 min followed by 33 cycles of 95◦C for 30 s, 58–62◦C for 30 s and 72◦C for 40–60 s, with a final extension step at 72◦C for 5 min. The resulting RT-PCR products were separated in 2% agarose gels and stained with ethidium bromide for visualization. For qRT-PCR analyses the aforementioned cDNA was used as template in quantitative reactions with the SensiFAST SYBR No-ROX Kit (Bioline) following the manufacturer's instructions. The reactions were run in a Bio-Rad CFX96 thermocycler and the amplification protocol was a 2-step cycling PCR program: 1 cycle at 95◦C for 2 min followed by 40 cycles of 5 s at 95◦C and 30 s at 60◦C. An additional melting curve step was used at the end of the reaction to assess the specificity of the amplified products. The qRT-PCR results were analyzed with CFX Manager software (Bio-Rad). In both cases, negative control samples of the cDNA synthesis step with no retrotranscriptase were included, in order to control for the presence of contaminating chromosomal DNA in the RNA samples. Primers used in these analyses (**Supplementary Table S1**) were designed with a preference for

17–23 mers and T<sup>m</sup> ∼ 65◦C with the help of Vector NTI Advance 11.5 (Invitrogen) and Primer3 software (Untergasser et al., 2012) and validated used the online tool NetPrimer (Premier Biosoft). In the specific case of qRT-PCR primers, these were designed to amplify fragments of ∼100 bp and their efficiency was tested using serial dilutions of chromosomic DNA as template. For both RT-PCR and qRT-PCR, primers for hrdB, encoding a housekeeping sigma factor, were used as an internal control to assess the quality of RNA and in the latter case to normalize gene expression levels.

#### 5 0 tag-RNA-Seq (tagRNA-Seq)

Library Construction and Sequencing Total RNA samples of S. scabies WT and 1btmL were extracted as previously described and, following quality control assays to ensure their integrity, they were submitted to Vertis Biotechnologie AG (Germany) for the construction and sequencing of tagRNA-seq libraries in a protocol adapted from the technique described in Innocenti et al. (2015). Prior to library construction, rRNA was depleted in the samples using the Ribo-Zero rRNA Removal Kit for bacteria (Epicenter). The remaining material was labeled sequentially as follows: sequence tag CTGAAGCT was ligated to transcripts presenting 5 0 -monophosphate groups (processed transcripts). The samples were then treated with RNA 5<sup>0</sup> polyphosphatase (50PP; Epicenter) to convert the 5<sup>0</sup> -triphosphate groups of primary transcripts into 5<sup>0</sup> -monophosphate ends amenable for ligation with the alternative tag sequence TAATGCGC. Once labeled, the samples were used as template for first-strand cDNA synthesis using random hexameric primers. After fragmentation and RNA clean up, Illumina TruSeq sequencing adapters were ligated in a strand specific manner to the 5<sup>0</sup> and 3<sup>0</sup> ends of the cDNA fragments. The cDNA was then amplified with a proof-reading enzyme to enrich the samples. At this point, it is possible to specifically PCR amplify the 5<sup>0</sup> -ends which carried the two tag sequences, but that would mean losing the information relating to the rest of the transcriptome. Therefore, a full transcriptome enrichment (for fragments that had Illumina adapters on both ends) was carried out in order to preserve the full coverage of the transcriptome in our libraries. The resulting material was purified with the Agencourt AMPure XP Kit (Beckman Coulter Genomics) and analyzed by capillary electrophoresis. The cDNA preparations were pooled in equimolar amounts and size selected (240–450 bp range) and the pooled libraries were sequenced on an Illumina NextSeq 500 system using 75 bp read lengths. RNA-seq data has been deposited in the ArrayExpress database (accession number E-MTAB-8236).

#### Data Analysis

Sequencing reads were first sorted according to their tags in order to generate independent fastq files for each of the datasets (see **Supplementary Table S2** for a summary of the sequencing results). Once sorted, the reads were trimmed to remove the tag sequences and aligned to S. scabies 87-22 genome (RefSeq NC\_013929.1) using Bowtie2 (Langmead and Salzberg, 2012), which yielded SAM alignment files for each of the fastq files. Downstream processing of the SAM alignment files was performed using a series of Perl scripts supported by the Bioperl (Stajich et al., 2002) toolkit. Using the SAM alignments as input, the number of reads mapping to each nucleotide position of the S. scabies genome was calculated and saved in a coverage file. The coverage information was then integrated with a feature table including the coordinates of all of the gene coding sequences in the genome, in order to calculate the number of reads per gene, or gene counts. These gene counts were used to make an estimation of gene expression levels by calculating RPKM and TPM values for each gene (Wagner et al., 2012) (**Supplementary Datasheet S1**). For mapping and visualization of the transcription start sites (TSSs) in the bottromycin cluster, the SAM files containing the sequences labeled with 5<sup>0</sup> tags were transformed into wig files suitable for visualization in the Integrated Genome Browser (IGB) software (Freese et al., 2016) plotted against the S. scabies genome. These files were normalized to eliminate biases due to difference in sequencing depth using the normalizeQuantiles function in the limma package (Ritchie et al., 2015) of R (R Core Team, 2014). Given the high amount of processed transcripts present in the btmC-D region of the bottromycin cluster, a "wig minus wig" file was generated in which processed transcript reads were subtracted from the primary transcript reads. The regions mapped in this filtered file were considered to be true TSSs.

#### 5 0 -RACE Experiments

The TSSs identified in the 5<sup>0</sup> tag-RNA-seq experiment were validated using a 5<sup>0</sup> -RACE system for rapid amplification of cDNA ends (Invitrogen), using the manufacturer's instructions (version 2.0). 1 µg of total RNA from S. scabies WT and 1btmL harvested after 72 h of growth at 28◦C and 230 rpm was used to carry out cDNA synthesis with specific primers for each of the promoters tested (**Supplementary Table S1**). The cDNA was purified and treated with terminal deoxynucleotidyl transferase (TdT) to add poly(dC) tails to its 3<sup>0</sup> ends. After an initial PCR amplification of the tailed fragments with the 5<sup>0</sup> -RACE abridged anchor primer and subsequent amplifications with the universal amplification primer (both provided by the kit) and specific nested primers (**Supplementary Table S1**), defined amplification products were observed. These products were gel purified and submitted for Sanger sequencing to confirm the position of the TSSs.

#### gusA Transcriptional Fusions Construction and β-Glucuronidase Reporter Activity Assays

Putative promoter regions of btmC (btmCp, 414 bp) and btmD (btmDp, 545 bp) and extended promoter regions spanning different lengths of btmC were amplified from S. scabies genomic DNA or from a refactored bottromycin cluster containing 33 synonymous point mutations in the btmDTTS region (**Supplementary Figure S4**) with primers containing NdeI and XhoI restriction sites (**Supplementary Table S1**). These were verified by sequencing, and then ligated or

assembled into NdeI/XhoI digested pIJ10742, which contains a promoterless copy of the reporter gene gusA (Feeney et al., 2017). These plasmids were introduced via intergeneric conjugation into S. scabies (WT, 1btmL, and WT + L) using E. coli ET12567/pUZ8002, where they integrated in the φBT1 phage integration site. Hygromycin resistant exconjugants were analyzed by colony PCR with primers pGUS\_chk\_fw and pGUS\_chk\_fw. β-glucuronidase assays in solid and liquid medium were carried out as described previously (Sherwood and Bibb, 2013). In the case of the liquid assays, reporter activity was represented as Miller units/mg protein. In both cases, pIJ10742 carrying no promoter was used as a negative control, and pIJ10741, carrying gusA under the control of ermE∗p (Feeney et al., 2017), was included as a positive control of promoter activity.

#### DNA Affinity Protein Capture Assay Sample Preparation

The intergenic region btmB-btmC (414 bp) was PCR amplified using primers probe\_BC\_fw2\_b and probe\_BC\_rv2 to generate a 5 <sup>0</sup> biotinylated probe. 40 µg of this probe were inmobilised onto 10 mg of streptavidin magnetic beads (Dynabeads <sup>R</sup> MyOneTM Streptavidin T1, Invitrogen, United Kingdom), following the manufacturer's instructions. Protein extracts from S. scabies wild type were obtained from 500 mL cultures in either BPM or GYM incubated for 72 h at 28◦C and 250 rpm. Cell pellets were harvested by centrifugation for 15 min at 7,000 rpm, resuspended in binding buffer (20 mM TrisHCl pH7.5, 1 mM EDTA, 100 mM NaCl, 10% glycerol, and 1 mM DTT) supplemented with protease inhibitors (cOmpleteTM Protease Inhibitor Cocktail, Roche, United Kingdom) and lysed by sonication with a Vibracell sonicator (Sonics & Materials Inc., United States) using 150 × 2 s pulses at 40% amplitude alternated by 5 s rest on ice. Binding assays were performed using a modification of the protocol reported by Bekiesch et al. (2016) using 60 mg of total protein in a 10 mL final volume of the aforementioned binding buffer supplemented with 0.1 mg/mL of salmon sperm DNA (Invitrogen, Germany). Proteins were eluted twice in 250 µL binding buffer containing 2 M NaCl and the binding assay was repeated twice for each sample, generating 1 mL of final eluate per sample. These eluates were then acidified to pH3 with trifluoroacetic acid (TFA) and applied to an C4 SPE column (OMIX C4, Agilent). The samples were washed twice with 0.1% TFA in water and three times with 0.1% FA before eluting in three fractions (200 µL 30% ACN and 0.1% FA, 200 µL 30% ACN and 200 µL 70% ACN) that were pooled and kept frozen on dry ice until analysis.

#### Sample Analysis

After purification and clean up, duplicate samples for each condition were analyzed by LC-MS/MS on an Orbitrap-FusionTM mass spectrometer (Thermo Fisher, United Kingdom) equipped with an UltiMateTM 3000 RSLCnano System using an Acclaim PepMap C18 column (2 µm, 75 µm × 500 mm, Thermo). Samples were loaded and trapped using a pre-column which was then switched in-line to the analytical column for separation. Peptides were eluted with a gradient of 5–40% ACN in water/0.1% FA at a rate of 0.5% min−<sup>1</sup> . The column was connected to a 10 µm SilicaTipTM nanospray emitter (New Objective, United States) for infusion into the mass spectrometer. Data dependent analysis was performed using a CID/HCD fragmentation method with the following parameters: positive ion mode, orbitrap MS resolution = 60 k, mass range (quadrupole) = 300–1500 m/z, AGC target 2e<sup>5</sup> , MS2 in ion trap, threshold 1e<sup>4</sup> , isolation window 1.6 Da (quadrupole), charge states 2–6, MS2 top20, AGC target 1.5e<sup>4</sup> , max inject time 200 ms, dynamic exclusion 1 count, 60 s exclusion, exclusion mass window ±7 ppm. MS scans were saved in profile mode while MS2 scans were saved in centroid mode. Raw files were processed with MaxQuant (version 1.5.3.30) (Tyanova et al., 2016). The peak lists were searched against a S. scabies protein database downloaded from Uniprot.org (25.05.2016) with 16846 entries together with the MaxQuant contaminants database (249 entries) using an in-house Mascot Server (2.4.1, Matrix Science, United Kingdom) with trypsin with 2 missed cleavages, carbamidomethylation (C) as fixed and oxidation (M), acetylation (protein N-terminus), and deamidation (N,Q) as variable modifications. Mass tolerances were 6 ppm for precursor ions and 0.6 Da for fragment ions. Mascot search results were imported into the Scaffold software (Proteome Software Inc., United States) to probabilistically validate protein identifications derived from the MS/MS sequencing results using the X!Tandem (Craig and Beavis, 2003) and ProteinProphet algorithms (Nesvizhskii et al., 2003). Validation parameters were set to 95% protein probability and 95% peptide probability.

# DATA AVAILABILITY STATEMENT

The datasets generated for this study can be found in the ArrayExpress Archive of Functional Genomics Data (accession number E-MTAB-8236).

# AUTHOR CONTRIBUTIONS

NV performed genetic manipulation experiments, LC-MS analyses, RNA extraction, and gene expression analyses. NV, EC-T, and TE performed promoter activity assays. GC and NV performed RNA-seq data processing and analysis. AT devised and supervised this work. NV and AT wrote the manuscript.

# FUNDING

Financial support was provided by a Royal Society University Research Fellowship (AT), Biotechnology and Biological Sciences Research Council (BBSRC) MET and MfN Institute Strategic Programme grants (BB/J004596/1 and BBS/E/J/000PR9790) to the John Innes Centre (AT and NV), BBSRC grant BB/M003140/1 (AT and NV), a European Commission Erasmus+ traineeship (EC-T), and a Ph.D. studentship (BBSRC Doctoral Training Program grant BB/M011216/1) (TE). Open access publication fees are paid by a United Kingdom Research Council grant to the John Innes Centre.

#### ACKNOWLEDGMENTS

fmicb-11-00495 March 24, 2020 Time: 16:1 # 14

We thank Dr. Gerhard Saalbach (John Innes Centre Proteomics Facility) for his help with the proteomic analysis of the DNA binding protein capture samples.

#### REFERENCES


#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2020.00495/full#supplementary-material


and translational regulatory elements encoded in the Streptomyces clavuligerus genome. Nucleic Acids Res. 47, 6114–6129. doi: 10.1093/nar/gkz471



**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2020 Vior, Cea-Torrescassana, Eyles, Chandra and Truman. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# SrrB, a Pseudo-Receptor Protein, Acts as a Negative Regulator for Lankacidin and Lankamycin Production in Streptomyces rochei

Yuya Misaki 1,2†, Shouji Yamamoto2†‡, Toshihiro Suzuki 2‡, Miyuki Iwakuni <sup>2</sup> , Hiroaki Sasaki <sup>2</sup> , Yuzuru Takahashi <sup>2</sup> , Kuninobu Inada<sup>3</sup> , Haruyasu Kinashi <sup>2</sup> and Kenji Arakawa1,2 \*

<sup>1</sup> Unit of Biotechnology, Graduate School of Integrated Sciences for Life, Hiroshima University, Higashi-Hiroshima, Japan, <sup>2</sup> Department of Molecular Biotechnology, Graduate School of Advanced Sciences of Matter, Hiroshima University, Higashi-Hiroshima, Japan, <sup>3</sup> Natural Science Center for Basic Research and Development, Hiroshima University, Higashi-Hiroshima, Japan

Streptomyces rochei 7434AN4, a producer of lankacidin (LC) and lankamycin (LM), carries many regulatory genes including a biosynthesis gene for signaling molecules SRBs (srrX), an SRB receptor gene (srrA), and a SARP (Streptomyces antibiotic regulatory protein) family activator gene (srrY). Our previous study revealed that the main regulatory cascade goes from srrX through srrA to srrY, leading to LC production, whereas srrY further regulates a second SARP gene srrZ to synthesize LM. In this study we extensively investigated the function of srrB, a pseudo-receptor gene, by analyzing antibiotic production and transcription. Metabolite analysis showed that the srrB mutation increased both LC and LM production over four-folds. Transcription, gel shift, and DNase I footprinting experiments revealed that srrB and srrY are expressed under the SRB/SrrA regulatory system, and at the later stage, SrrB represses srrY expression by binding to the promoter region of srrY. These findings confirmed that SrrB acts as a negative regulator of the activator gene srrY to control LC and LM production at the later stage of fermentation in S. rochei.

Keywords: Streptomyces, regulatory cascade, pseudo-receptor, antibiotic, biosynthesis

# INTRODUCTION

Secondary metabolites production is strictly controlled by small diffusible signaling molecules that constitute signaling-molecule/receptor regulatory systems in Streptomyces species (Bibb, 2005; Takano, 2006; Horinouchi and Beppu, 2007; Martín and Liras, 2019). The most-studied signaling-molecule/receptor system is A-factor/ArpA in Streptomyces griseus for streptomycin and grixazone production (Ohnishi et al., 1999, 2005). In the absence of A-factor, ArpA protein specifically binds to the promoter region of the target activator gene adpA and represses its transcription. When A-factor reaches a critical concentration, A-factor/ArpA complex dissociates from the promoter region of adpA, leading to the onset of adpA transcription. Then, the gene product of adpA binds to its targets (AdpA-regulons) to activate streptomycin and grixazone production and morphological differentiation (Ohnishi et al., 1999, 2005). Gene sets involved in the signaling-molecule-dependent regulatory pathways for

#### Edited by:

Evi Stegmann, University of Tübingen, Germany

#### Reviewed by:

Yihua Chen, Institute of Microbiology (CAS), China Juan F. Martin, Universidad de León, Spain

> \*Correspondence: Kenji Arakawa karakawa@hiroshima-u.ac.jp

†These authors have contributed equally to this work

#### ‡Present address:

Shouji Yamamoto, Department of Bacteriology I, National Institute of Infectious Diseases, Tokyo, Japan Toshihiro Suzuki, Department of Fermentation Science, Faculty of Applied Biosciences, Tokyo University of Agriculture, Tokyo, Japan

#### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 17 December 2019 Accepted: 30 April 2020 Published: 09 June 2020

#### Citation:

Misaki Y, Yamamoto S, Suzuki T, Iwakuni M, Sasaki H, Takahashi Y, Inada K, Kinashi H and Arakawa K (2020) SrrB, a Pseudo-Receptor Protein, Acts as a Negative Regulator for Lankacidin and Lankamycin Production in Streptomyces rochei. Front. Microbiol. 11:1089. doi: 10.3389/fmicb.2020.01089 Misaki et al. Function of a Pseudo-Receptor SrrB in Antibiotic Production

secondary metabolite production are listed in **Table 1** (Arakawa, 2018; Xu and Yang, 2019); e.g., streptomycin and grixazone production in S. griseus (Hara and Beppu, 1982; Onaka et al., 1995; Ohnishi et al., 1999, 2005), lankamycin (LM) and lankacidin (LC) (**Figure 1**) in Streptomyces rochei (Arakawa et al., 2007, 2012; Yamamoto et al., 2008; Suzuki et al., 2010), tylosin in Streptomyces fradiae (Bate et al., 1999, 2002; Stratigopoulos and Cundliffe, 2002), coelimycin P-1 in Streptomyces coelicolor (Takano et al., 2001, 2005; Hsiao et al., 2009; Gottelt et al., 2010; Li et al., 2015), actinorhodin and undecylprodigiosin in Streptomyces coelicolor (Xu et al., 2010; Wang et al., 2011), virginiamycin in Streptomyces virginiae (Kondo et al., 1989; Kinoshita et al., 1997; Kawauchi et al., 2000), jadomycin in Streptomyces venezuelae (Yang et al., 1995; Wang and Vining, 2003; Wang et al., 2009; Xu et al., 2010; Zou et al., 2014), kinamycin in Streptomyces ambofaciens (Aigle et al., 2005; Bunet et al., 2008, 2011), and avermectin in Streptomyces avermitilis (Kitani et al., 2011; Wang J. B., et al., 2014; Zhu et al., 2016).

Streptomyces rochei 7434AN4 produces two structurally unrelated polyketide antibiotics, LM and LC (**Figure 1A**) and carries three linear plasmids pSLA2-L, -M, and -S (Kinashi et al., 1994). Together with the biosynthetic genes for LM and LC, many regulatory genes including a biosynthetic gene for signaling molecules SRBs (**Figure 1B**) (srrX), six tetR-type repressor genes (srrA, srrB, srrC, srrD, srrE, and srrF), and three SARP (Streptomyces antibiotic regulatory protein) family activator genes (srrY, srrZ, and srrW) are located on giant linear plasmid pSLA2-L (210,614 bp) (Mochizuki et al., 2003). Our group revealed that SRBs/SrrA complex dissociates from the promoter region of srrY, leading to the activation of LC production (Yamamoto et al., 2008), whereas the gene product of srrY further activates a second activator gene srrZ to produce LM (Suzuki et al., 2010). In addition, mutation of an additional receptor gene srrB greatly increased the production of both LC and LM (Arakawa et al., 2007). This finding suggested that srrB negatively regulates LC and LM production, however, its functional role has not been clarified.

In this study, we extensively investigated the function of srrB by analyzing antibiotic production and transcription, the results of which indicated that SrrB acts as a negative regulator by binding to the promoter region of the activator gene srrY to control LC and LM production at the later stage of fermentation in S. rochei.

# MATERIALS AND METHODS

#### Bacterial Strains and DNA Manipulation

S. rochei strain 51252 carrying only the linear plasmid pSLA2-L was used as the parent strain (Kinashi et al., 1994). All strains, plasmids, and oligonucleotides used in this study are listed in **Table 2**. Streptomyces strains were grown in YM medium (0.4% yeast extract, 1.0% malt extract, 0.4% D-glucose, pH 7.3) for antibiotic production and RNA isolation. Escherichia coli strains were grown in Luria-Bertani (LB) medium supplemented with ampicillin (100µg/ml), apramycin (50µg/ml), and/or chloramphenicol (25µg/ml) when necessary. For protoplasts preparation, Streptomyces strains were grown in YEME medium (Kieser et al., 2000). Protoplasts were regenerated on R1M plates (Zhang et al., 1997). DNA manipulations for E. coli (Sambrook et al., 1989) and Streptomyces (Kieser et al., 2000) were performed according to the standard protocols. PCR amplification was done on a 2720 Thermal Cycler (Thermo Fisher Scientific, Waltham, MA, USA) with KOD-Plus- DNA polymerase (Toyobo, Osaka, Japan).

#### Construction of Mutant and Plasmid Construction of srrA and srrB Double Mutant

The target plasmid pKAR3014 that carries in-frame deletion of srrA in E. coli-Streptomycesshuttle vector pRES18 (Ishikawa et al., 1996) was constructed as described previously (Arakawa et al., 2007). Targeted mutagenesis was performed as follows. Plasmid pKAR3014 was transformed into protoplasts of S. rochei strain KA07 (srrB mutant), and thiostrepton-resistant strains were obtained. Among these transformants, single-crossover strains were selected by Southern hybridization. Some single-crossover colonies were continuously grown in YEME liquid medium to facilitate a second crossover. Finally, thiostrepton sensitive strains were selected as double crossover strains, to obtain a strain TS03 (srrAB mutant). Gene disruption was checked by Southern hybridization analysis using DIG DNA Labeling and Detection Kit (Roche Diagnostics GmbH, Mannheim, Germany).

#### Construction of srrY and srrB Double Mutant

The target plasmid pKAR3055 that carries in-frame deletion of srrY in pRES18 was constructed as described previously (Yamamoto et al., 2008). This plasmid was transformed into protoplasts of S. rochei strain KA07, and an srrB-srrY double mutant KA64 was constructed in a similar manipulation as above mentioned.

#### Construction of in vivo srrB Expression Plasmid

The srrB gene was amplified using cosmid A8 (Mochizuki et al., 2003) and primers, srrB-8600f1 and srrB-8600r1. The resulting PCR product was digested with NdeI and XbaI and cloned into pIJ8600, an E. coli-Streptomyces shuttle vector carrying a tipA promoter (Sun et al., 1999), to obtain pKAR3065.

This plasmid was introduced into strain 51252, and transformants were cultured for 24 h at 28◦C in YM liquid medium with 10µg/ml apramycin. Thiostrepton (10µg/ml as final concentration) was added at 24 h to induce srrB expression. After cultivation for additional 24 h, the broth filtrate was extracted twice with equal volume of ethyl acetate. The combined organic phase was dried with Na2SO4, and concentrated in vacuo to obtain crude extracts.

#### Construction of srrB Overexpression Plasmid in E. coli

The srrB-coding sequence was PCR amplified using the template cosmid A8 (Mochizuki et al., 2003) and primers, KAR7903OE and KAR7902OE. The amplified fragment was digested with BglII and EcoRI and cloned into pET32b(+), a (His)6-tagged expression vector, to obtain pKAR3036.


TABLE 1 | Gene sets of signaling molecule(s)/receptor/pseudo-receptor for secondary metabolite production in Streptomyces species.

ND, Not determined.

#### Isolation and Analysis of Metabolites

The 48-h cultures of S. rochei strains were harvested, and the supernatant was extracted twice with equal volume of ethyl acetate. The crude extracts were purified by Sephadex LH-20 chromatography (1 × 40 cm, GE Healthcare, Chicago, IL) with methanol. Then the fractions containing antibiotics were purified by silica gel chromatography with chloroform-methanol (80:1–10:1, v/v). NMR spectra were recorded on an ECA-500 spectrometer (JEOL, Tokyo, Japan) equipped with a field gradient accessory. Chloroform-d and methanol-d<sup>4</sup> were used as solvents. Chemical shifts were recorded in δ value based on the solvent signals (δ<sup>C</sup> = 77.0 in CDCl3, δ<sup>C</sup> = 49.0 in CD3OD, and δ<sup>H</sup> = 3.30 in residual CH3OH) or an internal standard tetramethylsilane (δ<sup>H</sup> = 0). High resolution ESI-MS spectra were measured by a LTQ Orbitrap XL mass spectrometer (Thermo Fisher Scientific). The <sup>1</sup>H- and <sup>13</sup>C-NMR assignments for lankamycin (**1**), lankacidin C (**2**), lankacidin A (**3**), lankacidinol A (**4**), isolankacidinol (**5**), and lankacidinol (**6**) have already been reported (Suzuki et al., 2010; Arakawa et al., 2011; Yamamoto et al., 2018).

#### SRB Assay

Two strains KA61 (1srrY) and KA64 (1srrY1srrB) were cultured at 28◦C for 30 h, and the supernatant (60 ml) was acidified to pH 3 and extracted with equal volume of ethyl acetate twice. The combined organic phase was concentrated in vacuo. Appropriately diluted culture extract (100 µl) was added to the fresh culture (5 ml) of strain KA20, an srrX-deficient strain, and cultured at 28◦C for 36 h to restore LM and LC production.

#### Time-Course Analysis

S. rochei strains were grown in YM liquid medium and harvested at various time periods at 12-36 h. Cells were used for measurement of dry cell weight (dcw) and isolation of total RNA, while the culture supernatant was for measurement of antibiotic production.

#### Measurement of DCW

Cultures were collected at various time periods and centrifuged at 5,000 rpm for 10 min. The resulting pellet was washed twice with 10.3% sucrose, and then placed in a 60◦C dry oven until the weight reaches to a constant value.

#### RNA Preparation and Reverse Transcription-PCR (RT-PCR)

S. rochei strains were cultured at 28◦C in YM liquid medium for various time periods. Total RNAs was extracted from cells with a TRI reagent (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions. Trace amounts of DNA were removed with RNase-free DNase I (Takara, Kyoto, Japan). The concentration of purified RNA was determined by UV absorbance at 260 nm using Ultrospec 3300 pro spectrometer (GE Healthcare). The cDNAs were synthesized using Transcriptor Reverse Transcriptase (Roche Diagnostics). Each reaction mixture contained 1 µg of total RNA and 0.08 A<sup>260</sup> units random primer. Each mixture was sequentially treated at 85◦C for 5 min, at 25◦C for 10 min, and 55◦C for 45 min for the cDNA synthesis. The 16S rRNA was used as an internal standard (Lane, 1991;Turner et al., 1999).

#### 5 ′ Rapid Amplification of cDNA Ends (5′ RACE)

Transcriptional start site (TSS) of srrB was determined using 5 ′ RACE System, Version 2.0 (Invitrogen, Carlsbad, CA, USA). Total RNA was prepared from a 24-h culture sample of parent strain. One microgram of total RNA was converted to the cDNA using specific primer srrB-GSP1, and the resultant was treated with ribonuclease and purified through spin column to afford cDNA. A homopolymeric tail was then added to the 3'-end of cDNA using terminal deoxynucleotidyl transferase and dCTP. PCR was performed with poly C tailed cDNA as a template using abridged anchor primer and

inner specific primer RT79-R2. TSS was determined from nucleotide sequence of amplified PCR product using ABI PRISM 310 Genetic Analyzer (Life Technologies, Carlsbad, CA, USA).

#### Overexpression and Purification of SrrB Protein

E. coli BL21(DE3)pLysS was used as hosts for plasmid pKAR3036. Cells were grown in LB liquid medium supplemented with <sup>100</sup>µg/ml ampicillin and 25µg/ml chloramphenicol at 37◦<sup>C</sup> to OD<sup>600</sup> = 0.6 and then were induced with 1 mM isopropyl β-thiogalactopyranoside (IPTG). Cultivation was continued for 12 h at 16◦C, and then cells were harvested and disrupted by SONIFER 250 ultrasonic homogenizer (Branson Ultrasonics Corporation, Danbury, CT, USA). The (His)6-fusion protein was purified by Ni2+-nitrotriacetic acid agarose (Qiagen GmbH, Hilden, Germany) according to the manufacture's protocol. After dialysis with PBS buffer (137 mM NaCl, 8.1mM Na2HPO4·12H2O, 2.68 mM KCl, 1.47 mM KH2PO4), the (His)6 tagged SrrB protein was treated with enterokinase (Novagen, Madison, WI, USA), and the (His)6-tag peptide upstream of the N-terminal SrrB was removed by Enterokinase Cleavage Capture Kit (Novagen) according to the manufacture's protocol. The protein was analyzed by SDS-PAGE with 15% polyacrylamide gel. The protein concentration was determined according to the methods of Bradford using Bio-Rad protein assay (Bio-Rad, Hercules, CA, USA) with bovine serum albumin as a standard.

#### TABLE 2 | Bacterial strains, plasmids, and oligonucleotides used in this study.


# Preparation of DNA Probes and Gel Shift Assay

The srrB probes for gel shift assay were prepared as follows. For preparation of probe B1, a 564-bp DNA fragment containing the upstream region of srrB was amplified using pKAR3004 as a template and primers SRRBf3 and SRRBr3 (positions −81 to +483 from TSS of srrB; nt 140,677-141,240 of pSLA2-L). For preparation of probe B2, a 386-bp DNA fragment containing the internal region of srrB was amplified using pKAR3004 as a template and primers SRRBf1E and SRRBr4 (positions +574 to +959 from TSS of srrB; nt 140,201-140,586 of pSLA2-L).

Probe B1 was then 3′ -end labeled with [γ-<sup>32</sup>P]ATP (GE Healthcare) and T4 polynucleotide kinase (Toyobo). The reaction mixture contained the binding buffer (20 mM Tris-HCl [pH 8.0], 100 mM NaCl, 1 mM dithiothreitol, 0.1 mg of bovine serum albumin and 5% glycerol), 0.5 nM labeled DNA and 2µM SrrA protein. SrrA protein was prepared as reported previously (Yamamoto et al., 2008). When necessary, synthetic SRB1 [(1– ′R)-isomer; **Figure 1B**] (Arakawa et al., 2012) was added to the reaction mixture. For competition experiment, unlabeled probes B1 and B2 were used at a final concentration of 200 nM. The reaction mixture was incubated at 26◦C for 30 min, and subjected to electrophoresis at room temperature on a native 4.5% polyacrylamide gel in 0.5 × TBE buffer (46 mM Tris base, 46 mM boric acid, 1 mM EDTA). The <sup>32</sup>P-labeled DNAs were detected by autoradiography.

Preparation of srrY probes for gel shift assay was described previously (Yamamoto et al., 2008). To analyze the effect of SRB on the binding of SrrA and SrrB, various concentration of synthetic SRB1 [(1′R)-isomer; **Figure 1B**] (Arakawa et al., 2012) was added to the reaction mixture. In order to evaluate the effect of endogenous metabolites in S. rochei and other antibiotics on the binding of SrrB, the following compounds (1 mM) were separately added to the reaction mixture; LC, LM, chlorotetracycline, kanamycin, and ampicillin.

# DNase I Footprinting

The method used for DNase I footprinting analysis for the upstream region of srrY was described previously (Yamamoto et al., 2008). For the upstream region of srrB, the primer SRRBf3 was 5′ -end labeled using [γ-<sup>32</sup>P]ATP (GE Healthcare) and T4 polynucleotide kinase (Toyobo), and then PCR reaction was performed with unlabeled primer SRRBr3 and pKAR3004 as a template to afford a 564-bp product containing the upstream region of srrB (positions −81 to +483 from TSS of srrB; nt 140,677-141,240 of pSLA2-L). Binding reaction mixture (50 µl) contained 10 nM labeled DNA, 20 mM Tris-HCl (pH8.0), 1 mM MgCl2, 100 mM NaCl, 1 mM dithiothreitol, 0.1 mg/ml BSA, 5% glycerol, and various concentrations of SrrA. The binding reaction mixture was incubated for 30 min at 25◦C, and then a mixture was treated with DNase I (Roche Diagnostics) solution [1 ng in 50 µl of 5 mM MgCl<sup>2</sup> and 5 mM CaCl2] for 2 minutes at room temperature. The reaction was terminated by 100 µl of phenol-chloroform. The aqueous fraction containing DNAs was precipitated by ethanol and separated on a 5% polyacrylamide gel containing 6 M urea. The labeled DNAs were detected by autoradiography. Sequencing ladders were generated by Maxam-Gilbert sequencing of the labeled DNA used for binding reaction.

# Comparative Sequence Analysis

Alignment of Amino acid sequences of the pseudo-receptors including SrrB was performed by BioEdit version 7.2.5 software (https://bioedit.software.informer.com/) (Hall, 1999) (**Figure S1A**). Phylogenetic tree was constructed by the neighborjoining algorithm of MEGA X version 10.1.5 software (Kumar et al., 2018) (**Figure S1B**).

# RESULTS

#### SrrB Acts as a Negative Regulator for Lankacidin and Lankamycin Production

The gene product of srrB belongs to the TetR-type transcriptional regulator family proteins, which contains a helix-turn-helix DNA binding motif at the N-terminal region (**Figure S1**). We previously reported overproduction of LM and LC in the srrB mutant KA07 based on TLC bioautography (Arakawa et al., 2007). In this study, we performed comparative metabolite analysis of the srrB mutant and its parent strain 51252. As shown in **Figures 2A,B**, the srrB mutant KA07 accumulated larger amount of compounds **1**-**6** compared with the parent strain 51252. Namely, KA07 produced 6-folds of lankamycin (**1**) (**Figure 2A**) and 9.9, 25, 4.2, and 5.7-folds of lankacidin C (**2**), lankacidinol A (**4**), iso-lankacidinol (**5**), and lankacidinol (**6**), respectively (**Figure 2B**). To investigate the effect of SrrA on antibiotic production, we further analyzed two mutants, an srrA mutant KA12 and an srrA-srrB double mutant TS03 (**Figure S2**). KA12 produced about 40% of metabolites when compared with the parent, while TS03 overproduced metabolites **1**-**6** at the same level with the srrB mutant KA07 (**Figures 2A,B**). These results confirmed the following two aspects; srrA mutation causes a slight decrease of the metabolic titer, whereas SrrB acts as a negative regulator for lankacidin and lankamycin production in S. rochei.

To determine the role of srrB in the regulation of lankacidin and lankamycin production, we further performed time-course analysis of metabolite profile, growth curve, and transcription in the parent and three mutants (KA07, KA12, and TS03) at various time periods. As shown in **Figure 2C**, all strains grew in a similar proportion, indicating that overproduction in KA07 and TS03 was due to srrB mutation but not to cell growth difference. The time-course of antibiotic production was analyzed by the titer of **2**, a major product among lankacidin derivatives (**2**-**6**; **Figure 1A**). As shown in **Figure 2D**, **2** was detected after 18 h, and its titer at 48-h growth in the srrB deficient strains, KA07 and TS03, were 9.0- and 7.2-times of 51252, respectively, which agrees with the overproduction profiles in KA07 and TS03 in **Figures 2A,B**. To confirm the negative regulatory property of SrrB in vivo, overexpression of SrrB in S. rochei was carried out. The intact srrB gene was introduced into plasmid pIJ8600, an E. coli-Streptomyces shuttle plasmid with a thiostreptoninducible tipA promoter, to give pKAR3065. We tested antibiotic production in the S. rochei 51252 recombinants containing either the empty vector pIJ8600 or the srrB overexpression plasmid (pKAR3065). Compared with the control recombinant S. rochei

FIGURE 2 | Metabolite profiles and cell growth of four S. rochei strains; 51252 (parent), KA12 (1srrA), KA07 (1srrB), and TS03 (1srrA-srrB). (A) TLC analysis of the crude extract of S. rochei strains. All strains were grown at 28◦C for 48 h. The left panel represents the TLC plate under ultraviolet irradiation (254 nm). The right panel represents the TLC plate after baking with anisaldehyde-H2SO4. TLC plates were developed with chloroform-methanol = 15:1 (v/v). (B) HPLC analysis of metabolites produced by S. rochei strains. The crude extracts were applied on a COSMOSIL Cholester column (4.6 × 250 mm, Nacalai Tesque) and eluted with a mixture of acetonitrile-10 mM sodium phosphate buffer (pH 8.2) (3:7, v/v) at a flow rate of 1.0 ml/min. (C) Time-course growth of S. rochei strains. Symbols represent

FIGURE 2 | each dry cell weight (DCW; g/l); strain 51252, blue circles and line; strain KA12, red diamonds and line; strain KA07, green squares and line; TS03, purple triangles and line. Results are representative of at least three independent experiments. (D) Time-course production of 2 in S. rochei strains. Symbols represent each production yield of 2 (µg/ml); strain 51252, blue circles and line; strain KA12, red diamonds and line; strain KA07, green squares and line; TS03, purple triangles and line. Results are representative of at least three independent experiments. (E) Effect of srrB overexpression on antibiotic production. Thiostrepton (10µg/ml) was added to the 24-h culture of S. rochei 51252 recombinants harboring either pIJ8600 (control) or pKAR3065 (intact srrB), and then the cultures were further incubated at 28◦C for 24 h. (1) TLC analysis of crude extracts. Crude extracts were separated by TLC [eluent; chloroform-methanol = 15:1 (v/v)] and detected by baking with anisaldehyde-H2SO4. Lane 1, LM and LC standard; lane 2, recombinant harboring pIJ8600 (control); lane 3, recombinant harboring pKAR3065 (intact srrB). (2) HPLC analysis of crude extracts. The crude extracts were applied on a COSMOSIL Cholester column (4.6 x 250 mm, Nacalai Tesque) and eluted with a mixture of acetonitrile-10 mM sodium phosphate buffer (pH 8.2) (3:7, v/v) at a flow rate of 1.0 ml/min.

51252/pIJ8600, the S. rochei 51252/pKAR3065 recombinant significantly reduced antibiotic production (**Figure 2E**). These results clearly indicated that SrrB acts as a negative regulator for antibiotic production in S. rochei.

In our preliminary experiment, gel shift assay indicated that SrrA and SrrB could bind to the upstream region of srrX, a gene responsible for SRB biosynthesis. This finding suggested that the signaling molecule receptor/pseudo-repressor repress the transcription of srrX. Two strains were used to evaluate the comparative yield of SRB; an srrY single mutant KA61 and an srrY-srrB double mutant KA64 (**Figure S3**), both of which are unable to produce LC or LM due to a mutation on the major activator srrY. The yields of SRBs were evaluated by a help of bioassay using an srrX-deficient mutant as described previously (Arakawa et al., 2012). One-eighth of the crude extract of KA64 contained an equivalent amount of SRBs to that of KA61 (**Figure S4**), suggesting that srrB negatively controls the titer of SRBs.

#### srrB and srrY Are Expressed Under the SRB/SrrA Regulatory System, and SrrB Then Represses srrY Expression at the Later Stage

To analyze the role of srrB in the SRB/SrrA regulatory system in S. rochei, we performed comparative transcriptional analysis of the selected regulatory genes in the parent and three mutants (srrA, srrB, and srrA-srrB). Transcription of srrY in the parent appeared around 18 h and diminished after 32 h (**Figure 3A**), while that in the srrB mutant KA07 continued until the later stage (**Figure 3C**). On the other hand, srrB transcription in the parent appeared around 16 h and prolonged until the late stage of fermentation (**Figure 3A**). In the srrA mutant KA12, transcription of both srrY and srrB appeared at 12 h or earlier (**Figure 3B**), whereas srrY transcription in the srrA-srrB mutant TS03 was detected through all time periods (12–36 h) (**Figure 3D**). These findings together with our previous result (Yamamoto et al., 2008) showed that SRB/srrA regulatory system controls transcription of both srrY and srrB, and SrrB represses srrY transcription at the later stage.

#### SrrA Binds to the Promoter Region of srrB

A transcriptional start point (TSS) of srrB was determined be 401-bp upstream of its translational start codon by 5′ - RACE (**Figure 4** and **Figure S5**). To determine whether SrrA binds to the upstream promoter region of srrB (srrBp), gel shift assay was performed using <sup>32</sup>P-labeled probe B1 (nt 140,677-141,240 of pSLA2-L) that contained the upstream region of srrB (**Figure 5A**). SrrA protein was overexpressed in E. coli as described previously (Yamamoto et al., 2008). A band shift of probe B1 was observed in the presence of SrrA protein in a concentration-dependent manner (**Figure 5B**). Competition experiments using unlabeled probes B1 and B2 (nt 140,201-140,585 of pSLA2-L) (**Figure 5C**) were performed to determine the specific binding of SrrA to the region of probe B1. A band shift disappeared in the large excess of unlabeled probe B1, whereas probe B2 gave no effect on band shift (**Figure 5C**). Addition of SRB led to dissociation of SrrA from probe B1 (**Figure 5C**), indicating that the srrB transcription is controlled by SRB/SrrA regulatory system.

DNase I footprinting experiment was performed to identify the SrrA binding sequence in the upstream of srrB. Positions −36 to −11 of the non-template strand was protected by SrrA (**Figure 5D**). The protected region overlapped with a possible srrB promoter containing a palindromic sequence (asterisks in **Figure 4B**), whose sequence well-matched with the SrrAbinding sequence of srrY (Yamamoto et al., 2008). Taking account of transcriptional analysis above mentioned, SrrA binds to the upstream regions of both srrB and srrY to repress their transcription at the early growth phase (∼16 h).

#### SrrB Represses srrY Transcription at the Later Stage of Fermentation by Binding to the Promoter Region of srrY

To analyze whether SrrB binds to the promoter region of srrY (srrYp), gel shift assay was performed using probe Y1 (positions −452 to +100 from TSS of srrY) (**Figure 6A**) containing the promoter region of srrY, which was constructed previously (Yamamoto et al., 2008). SrrB protein was overexpressed in the E. coli BL21(DE3)pLysS/pKAR3036 recombinant with IPTG induction and purified by Ni-NTA agarose (**Figure S6**). The addition of SrrB protein gave a shifted band of probe Y1 in a concentration-dependent manner (**Figure 6B**). Competitive experiments (**Figure 6C**) revealed that SrrB specifically binds to probe Y1, not to probe Y2 (positions +101 to +333 from TSS of srrY). The pseudoreceptors hitherto studied are insensitive to endogenous signaling molecules and interact with endogenous antibiotics (Martín and Liras, 2019; Xu and Yang, 2019) (details are described in Discussion Section). We tested the effects of various signaling-molecule/antibiotics on the binding of SrrB to srrYp through gel shift analysis by using endogenous signaling molecule SRB, endogenous antibiotics (LC and LM),

and other exogenous antibiotics (**Figures 6D,E**). Dissociation of SrrB from srrYp could be caused by SRB, however, a higher concentration of 1 mM was required (500-fold excess against SrrB protein). The sensitivity of SrrB against SRB was 50 fold lower compared with the signaling molecule receptor SrrA (**Figure 6D**). Dissociation of SrrB from srrYp was not caused by endogenous antibiotics LC and LM in S. rochei and neither by exogenous antibiotics (chlorotetracycline, kanamycin, ampicillin) at even 1 mM concentration (500-fold excess against SrrB protein) (**Figure 6E**).

dues no treatment of transcriptase. Red arrow indicates transcription of srrY, while blue arrow does transcription of srrB.

We further performed DNase I footprinting experiment to identify the SrrB binding sequence in the upstream of srrY. As shown in **Figure 6F**, positions −61 to −4 of non-template strand were protected by SrrB. Although SrrB covers larger upstream region (58 bp) than SrrA does (28 bp; blue letters in **Figure 6F**), both SrrA and SrrB could bind to the promoter region of srrY (**Figure 4B**).

#### DISCUSSION

In this study, we revealed that srrB acts as a negative regulator by binding to the promoter region of the target gene srrY to repress LC and LM production in S. rochei. Expression of srrB is controlled by SRB/SrrA regulatory system.

TetR-type receptors have a conserved DNA-binding helixturn-helix motif in the N-terminus and a ligand-binding pocket in the C-terminus (Yu et al., 2010). Particularly, the signaling molecule receptors and the pseudo-receptors constitute one of a major class of TetR-type regulators (**Figure S1**). The signaling molecule receptors have a helix-turn-helix DNAbinding motif in the N-terminus and a ligand-binding Trp residue in the C-terminus. It is noteworthy that the signaling molecule synthase and its cognate receptor gene pairs usually locate adjacently on the genome (Biarnes-Carrera et al., 2015), which allows us to predict signaling molecule/receptor systems

and green highlights indicate SrrB and Orf80 coding sequences, respectively. (B) Comparison of the binding sequences for SrrA and SrrB. The confirmed SrrA- and SrrB-binding sequences are shown as blue and red underlines, respectively. Possible SrrA-binding sites at upstream of srrW and srrX are deduced from sequence data. For comparison of consensus sequence, SrrA-binding sites at the upstream of srrY (SrrA-srrY) are shown as green. Bases identical with SrrA-srrY are shown in boldface letters. Complementary bases are indicated as asterisks. The center of palindromes is shown as a vertical dashed line.

in Streptomyces species (Niu et al., 2016). On the other hand, the pseudo-receptors also have a conserved DNA-binding motif like the signaling molecule receptors (**Figure S1A**), however, their location has no relationship with the signaling molecule synthase genes. Many of them act as negative regulators for antibiotic production; for example, TylQ for tylosin production in Streptomyces fradiae (Stratigopoulos and Cundliffe, 2002), BarB for virginiamycin in Streptomyces virginiae (Matsuno et al., 2004), ScbR2 for coelimycin P-1 in Streptomyces coelicolor (Gottelt et al., 2010), AlpW for orange pigment kinamycin in Streptomyces ambofaciens (Bunet et al., 2008), and AvaR2 for avermectin in Streptomyces avermitilis (Zhu et al., 2016) (**Table 1**).

In general, the pseudo-receptors are insensitive to endogenous signaling molecules. BarB has no binding affinity to virginia butanolides in S. virginiae (Matsuno et al., 2004). Surprisingly, ScbR2 from S. coelicolor does not bind to the signaling molecules SCB1-3 but binds to two endogenous antibiotics, actinorhodin and undecylprodigiosin (Xu et al., 2010), and exogenous antibiotic jadomycin (Wang W., 2014). In S. venezuelae, JadR2 binds to endogenous jadomycin and chloramphenicol as ligands, (Xu et al., 2010). Thus, ScbR2 and JadR2 bind to multiple antibiotics, and coordinate their biosynthesis (Xu et al., 2010; Zou et al., 2014). In S. avermitilis; AvaR2 binds to the endogenous signaling molecule avenolide, but not to oligomycin and avermectin (Zhu et al., 2016). Its mutational analysis revealed that AvaR2 plays a negative regulatory role in avermectin production and cell growth (Zhu et al., 2016). In S. rochei, SrrB-srrYp complex was disrupted by endogenous signaling molecule SRB at 1 mM concentration, although its minimum dissociation concentration for SrrB was 50-fold higher than that for SrrA, the SRB receptor. SrrB showed no binding activity to endogenous polyketide antibiotics LM or LC in S. rochei and neither to exogenous antibiotics including aromatic polyketide chlorotetracycline, aminoglycoside antibiotic kanamycin, and β-lactam antibiotic ampicillin even at 1 mM concentration (500-fold excess against SrrB). Thus, functions of the pseudo-receptors are variable in Streptomyces species.

The possible regulatory pathway in S. rochei is shown in **Figure 7**. At the early growth phase, SrrA represses transcription

of both srrY and srrB (panel A). When SRB reaches a critical concentration, SrrA-SRB complex dissociates from both promoter regions to induce expression of srrY and srrB (panel B). Then SrrB represses srrY transcription at the later stage fermentation (panel C), suggesting a transient expression of srrY by two receptor proteins SrrA and SrrB in S. rochei. A similar regulatory pathway was proposed for kinamycin production in S. ambofaciens (Bunet et al., 2008) although

and 2,000 nM recombinant protein. To the reaction mixture, various concentration of synthetic SRB1 [(1′ R)-isomer; Figure 1B) (Arakawa et al., 2012) was added. (Continued) FIGURE 6 | (E) Effect of endogenous metabolites in S. rochei and other antibiotics on the binding of SrrB. To the same reaction mixture described for panel D, various compounds including SRB1, LC, LM, chlorotetracycline, kanamycin, and ampicillin (each 1 mM) were separately added. (F) DNase I footprint analysis of SrrB-binding site on the upstream of srrY. Probe Y1 was end labeled on the non-template strand. Each reaction mixture contained 2 nM labeled DNA and SrrB (100 nM). Sequencing ladders were generated by Maxam-Gilbert sequencing of the labeled probe Y1. Capital letters at right side indicate SrrB-binding sequences, among which blue letters indicate SrrA-binding sequences.

its ligand has not yet been identified. In the early stage of growth, the signaling molecule receptor AlpZ represses both transcription of alpV (an SARP-type activator gene) and alpW (a pseudo-receptor gene). When an unidentified ligand interacts with AlpZ, this protein dissociates from the promoter regions in both alpV and alpW, leading to kinamycin production. At the later stage of fermentation, AlpW represses alpV transcription again to cease kinamycin production. Another interesting features in the S. rochei regulatory pathway is the presence of srrY-srrC cistron (**Figure 7**). The srrC mutant showed no sporulation, suggesting that srrC acts as a positive regulator for morphological differentiation (Arakawa et al., 2007). As shown in **Figure 7**, srrB negatively regulates the transcription of both srrY and srrC, which leads to transient controls for antibiotic production and morphological differentiation, respectively.

The srrB mutation increased the titers of antibiotics as well as SRBs. This result well agreed with our preliminary gel shift assay that both SrrA and SrrB bind to the upstream region of SRB biosynthesis gene srrX (data not shown). Large excess of SRBs has no effect on antibiotic overexpression in S. rochei (Arakawa et al., 2012), hence, exact mechanism of the srrX repression by SrrB at the later stage remains to be clarified.

Manipulation of regulatory genes often causes activation of "silent" secondary metabolites (Olano et al., 2008; Zerikly and Challis, 2009; Rutledge and Challis, 2015; Arakawa, 2018). In S. ambofaciens, a mutant of the pseudo-receptor gene alpW accumulated kinamycin (Bunet et al., 2011). A mutant of the repressor gene ksbC accumulated β-carboline compound kitasetaline in Kitasatospora setae (Aroonsri et al., 2012). To our surprise, azoxyalkene compound KA57-A accumulated in a triple knockout mutant of S. rochei that have mutations on two biosynthetic gene clusters for LC, LM, and srrB (Kunitake et al., 2015). The genome sequence of the S. rochei chromosome has been determined to be 8.36 Mb in size, and at least 35 secondary metabolites gene clusters are coded on the chromosome (Nindita et al., 2019). Comprehensive mutational analysis on various regulatory genes may lead to activate silent secondary metabolite gene clusters in S. rochei, which is in progress in our laboratory. In conclusion, we have extensively characterized the role of the pseudoreceptor SrrB for antibiotic production in S. rochei. Further understanding and manipulation of the regulatory system in Streptomyces will lead to a natural product discovery with notable biological activities.

#### DATA AVAILABILITY STATEMENT

The datasets generated for this study are available on request to the corresponding author.

### AUTHOR CONTRIBUTIONS

YM, SY, TS, HK, and KA designed the experiments. YM, SY, TS, MI, HS, YT, KI, and KA performed the experiments. YM, SY, TS, MI, HS, and KA analyzed the data. YM, SY, TS, HK, and KA wrote the manuscript with input from all of the authors. All authors approved the final version of the manuscript.

#### FUNDING

This work was supported by Grants-in-Aid for Scientific Research on Innovative Areas (23108515, 25108718, 17H05446, and 19H04659 to KA) from Ministry of Education, Culture, Sports, Science and Technology of Japan (MEXT), Grants-in-Aid for Scientific Research (B) (16H04917 to KA) from the Japan

#### REFERENCES


Society for the Promotion of Science (JSPS), and JSPS A3 Foresight Program.

# ACKNOWLEDGMENTS

We would like to thank Dr. D. Kajiya and Mrs. T. Amimoto (N-BARD, Hiroshima University) for measurement of the high resolution mass spectra.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2020.01089/full#supplementary-material


Natl. Acad. Sci. U.S.A. 108, 16410–16415. doi: 10.1073/pnas.11139 08108


**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2020 Misaki, Yamamoto, Suzuki, Iwakuni, Sasaki, Takahashi, Inada, Kinashi and Arakawa. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Molecular Dynamics to Elucidate the DNA-Binding Activity of AlpZ, a Member of the Gamma-Butyrolactone Receptor Family in Streptomyces ambofaciens

Cláudia M. Vicente<sup>1</sup> \* † , Jean-Michel Girardet<sup>2</sup> , Laurence Hôtel<sup>1</sup> and Bertrand Aigle<sup>1</sup> \*

<sup>1</sup> Université de Lorraine, INRAE, DynAMic, Nancy, France, <sup>2</sup> Université de Lorraine, INRAE, IAM, Nancy, France

#### Edited by:

Yvonne Mast, German Collection of Microorganisms and Cell Cultures GmbH (DSMZ), Germany

#### Reviewed by:

Shigeru Kitani, Osaka University, Japan Wenqing Chen, Wuhan University, China

#### \*Correspondence:

Cláudia M. Vicente claudia.vicente@insa-toulouse.fr Bertrand Aigle bertrand.aigle@univ-lorraine.fr

#### †Present address:

Cláudia M. Vicente, TBI, CNRS, INRAE, INSA, Université de Toulouse, Toulouse, France

#### Specialty section:

This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology

Received: 08 December 2019 Accepted: 18 May 2020 Published: 02 July 2020

#### Citation:

Vicente CM, Girardet J-M, Hôtel L and Aigle B (2020) Molecular Dynamics to Elucidate the DNA-Binding Activity of AlpZ, a Member of the Gamma-Butyrolactone Receptor Family in Streptomyces ambofaciens. Front. Microbiol. 11:1255. doi: 10.3389/fmicb.2020.01255 Signaling molecule receptors play a central role in quorum sensing and in the coordination onset of specialized metabolite biosynthesis in Streptomyces due to their dual function in signal detection and gene expression control through DNA-binding in the promoter region of their target genes. In Streptomyces ambofaciens the alp biosynthetic gene cluster includes the signaling molecule receptor AlpZ that negatively regulates through a complex regulatory cascade the expression of key genes involved in the kinamycin antibiotic production until its cognate ligand, a yet unidentified signaling molecule, prompts its release from target promoters. Here we use an original molecular dynamics method to evaluate the DNA-binding properties of AlpZ to its target DNA sequence and the impact the signaling molecule has on the interaction. It is the first time this approach is used to characterize a regulator from the γ-butyrolactone receptor family. The observed K<sup>D</sup> in the nanomolar range indicates that AlpZ-DNA constitute a particularly stable complex. The signaling molecule ably disturbs this binding while kinamycin has no effect on the activity of AlpZ. Regulator size was determined and found to be considerably large regarding protein sequence, indicating that AlpZ regulates gene expression by binding the DNA as a homodimer, and structural modeling comparison with closely related γ-butyrolactone receptors supports this conclusion.

Keywords: molecular dynamics, DNA-binding, signaling molecule receptor, quorum sensing, kinamycin

# INTRODUCTION

Bacterial quorum sensing is a well-established phenomenon by which bacteria produce and sense chemical signals to communicate with neighboring cells. These diffusible signaling molecules are able to trigger a plethora of activities such as genetic exchange, antibiotic production, motility, virulence, biofilm formation, among others. The signaling molecule accumulates extracellularly during cell growth and when the concentration reaches a certain threshold it is detected by a receptor that elicits a downstream signal transduction cascade, triggering a specific gene expression program. An array of signaling molecules are used for cellular communication. For example, the acylhomoserine lactones (AHL) are the representatives and most studied chemical signals in Gram-negative bacteria, while Firmicutes mainly use peptides as signaling molecules

**170**

(Waters and Bassler, 2005). Streptomyces, ubiquitous filamentous Gram-positive bacteria characterized for their complex life cycle, linear chromosome and their ability to produce a wide variety of specialized metabolites including antibiotics, herbicides, immunosuppresssants and anticancer agents, display a different profile of signaling molecules.

The vast majority of signaling molecules produced by Streptomycetes and identified to date are classified into three major groups, γ-butyrolactones (GBL) (Yamada et al., 1987; Ohnishi et al., 1999; Takano et al., 2000; Takano, 2006), 2-alkyl-4-hydroxymethylfuran-3-carboxylic acids (AHFCA) (Corre et al., 2008) and γ-butenolides (Kitani et al., 2011; Arakawa et al., 2012). Two other less studied signaling molecules with quite different structures exist, the PI factor and the N-methylphenylalanyldehydrobutyrine diketopiperazine (Recio et al., 2004; Matselyukh et al., 2015). These molecules are widespread in the genus, often controlling both specialized metabolism and morphological differentiation (McCormick and Flärdh, 2012). Many of these molecules are active at nanomolar concentrations.

Gamma-butyrolactones constitute the largest and most studied group of signaling molecules in Streptomyces, and the first bacterial signaling molecule to be discovered was the A-factor of Streptomyces griseus (Khokhlov et al., 1967), a γbutyrolactone type signaling molecule that governs a vast regulon of genes involved in both the morphological differentiation and specialized metabolism (Horinouchi et al., 2001; Horinouchi, 2002). Specialized metabolite biosynthetic pathway expression is a much regulated process (Liu et al., 2013). The onset of production is tightly controlled by a regulatory network comprised of different transcriptional regulators incorporating various environmental and physiological signals (Bibb, 2005; van Wezel and McDowall, 2011; Martín and Liras, 2012). Cluster-situated regulators (CSRs) play an important role, directly controlling gene transcription within the specialized metabolite biosynthetic gene cluster [although some CSRs have been described to cross regulate other biosynthetic gene clusters (Vicente et al., 2015; McLean et al., 2019)], while pleiotropic regulators modulate the biosynthesis of several specialized metabolites and often morphological development as well. Some of these transcriptional regulators are central to the quorum sensing phenomenon, acting as receptors of the diffusible signaling molecules, and usually belong to the large and widely distributed TetR family of transcriptional regulators (Ramos et al., 2005). A typical TetR-family GABR (γ-butyrolactone receptor) is comprised of two functional domains, a N-terminal helix-turn-helix DNA-binding domain that can interact with specific DNA target sequences in promoter regions, and a C-terminal signaling molecule-binding domain that interacts with cognate ligands (Cuthbertson and Nodwell, 2013). The promoter-bound signaling molecule receptor prevents transcription, and binding of the cognate signaling molecule leads to its release from the DNA allowing gene expression (Willey and Gaskell, 2011; Sidda and Corre, 2012).

The type II polyketide synthase (PKS) alp biosynthetic gene cluster of Streptomyces ambofaciens is responsible for the production of the antibiotic kinamycin and its regulatory network has been previously elucidated (Pang et al., 2004; Aigle et al., 2005; Bunet et al., 2008, 2011). Several forms of kinamycin have been identified (Bunet et al., 2011) and for the purpose of this work will be referred to as kinamycin. The cluster is duplicated (two identical copies) due to its location in the terminal inverted repeat (TIR) sequences at both ends of the chromosome and was initially identified as comprising 27 genes including three genes that compose the minimal PKS (alpA, alpB, and alpC) and five regulatory genes. Recent reports indicate that the cluster size could be extended to include a 24 gene region containing 6 genes responsible for the synthesis of the diazo functional group (Wang et al., 2015). Of the regulator genes, alpT, alpU, and alpV are members of the SARP (Streptomyces antibiotic regulatory proteins) family (Wietzorrek and Bibb, 1997), and alpW and alpZ belong to pseudo-GABR and GABR family, respectively (Matsuno et al., 2004; Nishida et al., 2007). Pseudo-GABRs are paralogs of GABR but are unable to interact with the cognate GBL. In the regulatory cascade (**Supplementary Figure S1**), the positive regulator AlpV activates kinamycin biosynthesis promoting the expression of the minimal PKS-encoding genes. During the initial growth phase, the negative regulator AlpZ specifically binds and prevents expression of the operon alpXW and of alpV as well as its own, hindering antibiotic biosynthesis particularly through the repression of alpV. The binding sites of AlpZ, as well as those of AlpW, have been characterized (Bunet et al., 2008, 2011) and show high homology to previously defined ARE (autoregulatory elements) motifs (Folcher et al., 2001). Following synthesis and build-up of the yet unknown S. ambofaciens signaling molecule, it binds to the regulator AlpZ releasing it from the target promoters and allowing gene expression that leads to kinamycin production (Bunet et al., 2008). At this stage late negative regulator AlpW accumulates and blocks alpV expression thus switching off kinamycin biosynthesis, at which point the signaling molecule concentration also decreases and based on transcription analysis data it is hypothesized that the negative control exerted by AlpZ on the expression of both alpV and alpXW resumes (Bunet et al., 2008, 2011). Although it remains unidentified, the signaling molecule responsible for the quorum sensing regulation of kinamycin production in S. ambofaciens has been characterized to some extent. Initial studies predict that it is not a γ-butyrolactone type signaling molecule, based on its physical properties such as its resistance to alkaline conditions and also to high temperatures (Bunet et al., 2008), but rather perhaps a AHFCA-like molecule. Furthermore, an afsA-like gene responsible for its biosynthesis was identified close to the alp cluster (unpublished data). AfsA-like enzymes have been shown to be key in both γ-butyrolactones and AHFCA synthesis (Hsiao et al., 2007; Kato et al., 2007; Corre et al., 2008).

As in the case of kinamycin biosynthesis, these signaling molecules and their receptors constitute one of the most extended regulatory systems to elicit the biosynthesis of specialized metabolites and/or morphological differentiation in a coordinated fashion in Streptomyces (Polkade et al., 2016). Several signaling molecules have been identified so far with more expected to exist (Thao et al., 2017). For this reason, signaling molecule receptors are central to modulate specialized metabolite biosynthesis and can be used as a strategy for cryptic gene

cluster activation (Aigle and Corre, 2012; Sidda and Corre, 2012), shaping regulatory systems and to develop expression control tools (Bowyer et al., 2017; Biarnes-Carrera et al., 2018).

gold surface, allowing the accurate measurement of the oscillating orientation change of the DNA in dynamic response.

In this study, we address the role of AlpZ, a negative regulator and receptor of an unknown signaling molecule in S. ambofaciens that controls kinamycin biosynthesis, using a novel biophysical-based approach that allows a detailed depiction of the regulator's DNA-binding activity and how it is affected by the signaling molecule.

# RESULTS

# Regulator AlpZ Has High Affinity to Its Target Sequence

A recently developed technology was used to perform molecular dynamics studies and characterize the putative signaling molecule receptor AlpZ, the switchSENSE <sup>R</sup> . It is based on the electric properties of negatively charged double stranded DNA sequences, that when grafted to a gold microelectrode and subjected to alternating electric potentials are either attracted or repelled from the surface. Real-time tracking of the fluorescence emission from the fluorescent dye attached to the dsDNA nanolevers that depends on its proximity to the gold surface, allows to assess variations in the switching movement. To determine the affinity of AlpZ to its target sequences through molecular dynamics, specific DNA sequences were designed, the dsoverhangs. These contain both the target dsDNA sequence and a ssDNA region complementary to the biochip attached nanolevers (**Figure 1**). The AREV sequence located in the promoter region of the alpV gene (i.e., AREV) was used as a model of the known recognized sequences in the gene cluster (Bunet et al., 2008) and to create the dsoverhang nanoAREV. Functionalization of the dsoverhangs on the biochip was verified before each assay through increased fluorescence signal detection.

First, different concentrations of protein from 1.6 to 55.6 µM were used to assess the DNA-binding activity of AlpZ to the nanoAREV. Results show a strong affinity and remarkable stable binding of AlpZ to its target sequence. Indeed, in these conditions binding rate constant kON values reached 1.46 to 4.43 × 10<sup>4</sup> M−<sup>1</sup> .s−<sup>1</sup> (**Table 1**), respectively. Notably, with increasingly higher protein concentrations the kON remains unchanged suggesting a saturation phenomenon occurs easily (**Figure 2A** left panel, **Table 1**). Furthermore, in these conditions no dissociation was observed even after a long period of time in dissociation favoring conditions (**Figure 2A** right panel). To better analyze the binding properties of AlpZ to the nanoAREV, assays were then performed using a lower concentration of protein (50 nM). It was observed that in these conditions protein binding occurs more slowly, with a kON of 1.38 × 10<sup>5</sup> M−<sup>1</sup> .s−<sup>1</sup> (**Figure 2B** left panel, **Table 1**). Additionally, a slow dissociation is finally observed, although an even longer period of time is necessary to reach a steady state. This dissociation is however, enough to determine the binding rate constant kOFF, which is of 1.05 × 10−<sup>4</sup> s −1 .

TABLE 1 | Kinetic parameters of AlpZ with target DNA sequence.


The dissociation constant K<sup>D</sup> of AlpZ to its target model sequence was found to be 37.1 nM (**Figure 2B** right panel, **Table 1**). These results indicate that the AlpZ-DNA complex is substantially stable.

To ensure AlpZ specificity in binding DNA during the assays, an unspecific dsoverhang was designed, the nanoControl. Instead of the AREV sequence it contains the AREU sequence originally located in the promoter region of alpU from the kinamycin gene cluster. It has been previously demonstrated that this sequence, which contains a weakly conserved ARE motif, is neither recognized nor bound by the regulator (Bunet et al., 2008). Molecular dynamic analysis using the established conditions and the nanoControl indicate that AlpZ does not bind to the dsoverhang. There is no observed association and consequently no dissociation (**Figure 2C**), confirming that AlpZ recognizes and specifically binds its target DNA sequences.

#### Sizing and Protein Structure Modeling Suggest AlpZ Functions as a Homodimer

The used molecular dynamics approach enables the measurement of protein friction, ultimately allowing protein size determination through hydrodynamic diameter (DH) measurements. As the experimental set-up uses dsoverhangs, the movements of the biochip's nanolevers with the dsoverhangs used for the analysis have increased friction and lower dynamic response (DR) compared to those of the nanolevers alone. It is therefore necessary to first determine the relative size of the dsoverhang (different from the actual D<sup>H</sup> as the dsoverhangs

do not have a globular shape that is assumed in the model for size calculation). By comparing the switching dynamics of the dsoverhang to those of the reference nanolevers, the nanoAREV was shown to have 3.65 nm in relative size (**Figure 3A**), identical to the size of the nanoControl, as expected since the dsovehangs have the same length (data not shown).

Interestingly, we determined that AlpZ has a D<sup>H</sup> of 5.27 nm (**Figure 3B**). AlpZ is a relatively small protein with 237 aa and a theoretical MW of 25.7 kDa, however, this result suggests its size is considerable. The only other related γbutyrolactone receptor with structural data available is CprB from Streptomyces coelicolor A3(2) (Natsume et al., 2004). AlpZ and CprB belong to the GABR and pseudo-GABR family, respectively and show significant similarity (almost 30% identity and CprB has 26.4 kDa, and moreover similar predicted structures). Although it has been demonstrated that CprB binds DNA as a dimer of dimers (Bhukya et al., 2014), further biophysical modeling analysis using the switchANALYSIS software and the structural data of the apo-form of CprB (Natsume et al., 2004) predicted its D<sup>H</sup> to be 5.21 nm, a very similar size to that of AlpZ. This result indicates that the form of AlpZ that binds the nanoAREV could be a homodimer. Furthermore, using a threading/fold recognition method (Zhang, 2009; Yang and Zhang, 2015) to modelize AlpZ it was possible to predict its single molecule structure. The best model has a relatively good C-score of −0.60 and is structurally very close to the homodimer structure of the TetR family regulator TylP from Streptomyces fradiae (Ray et al., 2017), showing a TM-score of 0.87 and coverage of 0.822

(**Figure 4**). This result further strengths the hypothesis that AlpZ is active as a homodimer.

#### Signaling Molecule Drastically Disrupts AlpZ-DNA Complex

It has been previously reported that a yet unidentified S. ambofaciens signaling molecule is able to affect the DNA binding by AlpZ, effectively modulating the regulation of kinamycin biosynthesis (Bunet et al., 2008). To obtain signaling molecule containing extracts, the S. ambofaciens 11alpW11alpID mutant strain was used as it lacks the late regulator alpW leading a constitutive production of signaling molecule (unpublished data). This strain is also deficient in kinamycin production [loci alpIABCD deleted, in which the alpABC locus encodes the minimal PKS (Pang et al., 2004)], allowing therefore analysis without the putative interference of the biosynthetic pathway products (i.e., kinamycins or its intermediates). Extracts obtained from the culture supernatant were routinely checked for the presence of signaling molecule using electrophoretic mobility shift assays (EMSA), as seen in **Figure 5**.

To evaluate the effect of signaling molecule on AlpZ and its binding to the nanoAREV, an assay was designed to accurately detect protein release from the dsoverhang. The correct binding of AlpZ to the dsoverhang is first verified, and then an extract containing signaling molecule is added and molecular dynamics analysis are performed. Size analysis confirmed that the signaling molecule drastically alters the DNA binding function of AlpZ. The protein that was steadily bound to the dsoverhang (**Figure 6A**) shows a size of 5.7 nm, and is released when signaling molecule is present, changing the detected size to that of the dsoverhang and reference with 3.5 nm (**Figure 6B**). Additionally, kinetics analysis also demonstrates that the impact of the signaling molecule can be significant. Concentrated extracts of 0.5 mg/ml induce an almost immediate release, showing a considerably high kOFF of 1.95 × 10−<sup>1</sup> s −1 (**Figure 7A**), whereas more diluted samples lead to a fast but steady dissociation event with no protein molecule remaining bound to the dsoverhangs after just 5 minutes (**Figures 7B,C**). The observed effect is manifestly concentration dependent, and even small modifications of concentration affect the kOFF values. Extract samples from the 11alpW11alpID strain at 25 and 50 µg/ml display kOFF values of 1.04 × 10−<sup>2</sup> and 1.79 × 10−<sup>2</sup> s −1 , respectively. Other solutions were previously proven to have no effect on the DNA-binding activity of AlpZ, including the used culture medium and purified molecules of the three subclasses of γ-butyrolactones, the A-factor, SCB1 and IM-2 (Bunet et al., 2008), confirming the observed phenomenon is specifically related to the unknown S. ambofaciens signaling molecule.

The interactions involving the three players AlpZ, target DNA and signaling molecule are clearly visible with a conformational analysis. The assay was designed and performed to analyze AlpZ binding to the nanoAREV dsoverhang and subsequently adding PE40 buffer (control) and signaling molecule containing extract to observe its effects on the association, followed by a conformational analysis (**Figure 8**). As in every performed assay, six electrodes are used, the first two of which are the controls without neither ligand nor analytes, and the following four electrodes are used for the assay itself (**Figure 8a**). As expected, the functionalization of the dsoverhangs (the first step of the assay), already has a negative impact on the dynamic response (**Figure 8b**). When AlpZ is introduced the dynamic response further decreases, as the protein tightly binds to the dsoverhangs (**Figure 8c**). The complex protein-DNA is very stable and not altered by solutions like the buffer PE40 (**Figure 8d**). However, when in the presence of signaling molecule containing extracts the dynamic response increases to the original levels upon release of the dsoverhangs by AlpZ (**Figure 8e**). During the entire assay, the control electrodes remain unchanged, confirming that no unspecific protein binding occurs to the nanolevers.

# Kinamycin Does Not Impact AlpZ DNA-Binding Activity

Some regulator-receptors have been demonstrated to bind the product of their biosynthetic gene clusters in Streptomyces, such

as JadR2 (Xu et al., 2010).To analyze if kinamycin could also act as a ligand of AlpZ, we performed molecular dynamic assays in the presence of purified kinamycin D and the results were analyzed (**Figure 9**). No dissociation of AlpZ was observed when kinamycin was used, confirming the regulator does not respond to kinamycin as a signal to modulate transcriptional control. Although the S. ambofaciens signaling molecule remains elusive, these results support the initially proposed classification of AlpZ as a true GABR.

#### DISCUSSION

The TetR regulator AlpZ plays a key role in the regulatory network of kinamycin biosynthesis in S. ambofaciens ATCC 23877, coupling it to the production of a yet unknown signaling molecule. As a negative transcriptional regulator, it exerts its control by binding specific promoter located sequences and hence blocking gene expression. An in-depth molecular characterization of this regulator is crucial to understand how it responses to target promoters and triggering signals.

The present study indicates that AlpZ has a strong affinity for its target DNA sequences and forms very stable complexes upon binding, as seen by a K<sup>D</sup> in the nanomolar range. This result is also in good agreement with the EMSA experiments performed. Signaling molecule receptors such as GABRs have long been known for their affinity to DNA targets (Miyake et al., 1989; Onaka et al., 1995), though recent characterization of the binding activity of CprB to its target DNA shows a considerably higher K<sup>D</sup> in the micromolar range (Biswas et al., 2015). The stable AlpZ-DNA complex explains how the regulator achieves a steady control on gene expression and on subsequence kinamycin biosynthesis. Protein size was determined through molecular dynamics, an approach validated in previous studies by comparing it with other more recurrent methods (Blocquel et al., 2017). The determined size of DNA-bound AlpZ is hence consistent and nonetheless unexpectedly large, suggesting that the regulator could be in a homodimer form. A model of AlpZ structure obtained through a threading/fold recognition prediction method shows high similarities with that of TylP, also a homodimer (Ray et al., 2017), strengthening the hypothesis that AlpZ binds its target DNA as an homodimer.

following binding of 50 nM of AlpZ (c), in presence of PE40 buffer (d), and signaling molecule-containing extract (e). Above the graph are representations of the assay conditions and kinetics on the nanolevers, nanoAREV dsoverhang, with AlpZ (red circles), in the presence of buffer (+buffer) and signaling molecule (star). Vertical bars show dynamic response in the 6 electrodes present in the biochip. In each case the first two electrodes are the controls with nanolevers to verify the absence of unspecific binding to this region, and the next four electrodes were subjected to the different conditions. Assay was performed with 50 nM of AlpZ and a concentrated signaling molecule containing extract (0.5 mg/ml). Signal expressed in dynamic response units (dru). Error bars represent the standard deviation.

Only one trigger has been identified to actively interfere with the AlpZ-DNA complex, the yet unidentified S. ambofaciens signaling molecule. When present, the signaling molecule prompts a rapid AlpZ dissociation and the slightest amount variation is detected and impacts the dissociation rate. Consistent data were found both with molecular dynamics and sizing assays, substantiating this conclusion.

It is unclear if the DNA release is a result of an interference mechanism (e.g., competition for the same binding site), or rather the induction of structural conformational changes on the regulator hindering its DNA-binding capability. Based on

previous studies, the latter hypothesis seems more probable, as TetR regulators appear to shift between mutually exclusive conformational states, either binding DNA or their cognate ligand of choice (Ramos et al., 2005; Bhukya et al., 2014), unlike the LuxR receptor regulator in proteobacteria whose structure has been resolved in complex with both the target DNA and the cognate pheromone (Zhang et al., 2002). Some bacterial compounds have been described to function as both antibiotics and signaling molecules (Beyersmann et al., 2017), however, that does not seem to be the case of kinamycin in regard to AlpZ. The DNA-binding activity of AlpZ is not impacted by kinamycin, indicating the regulator is not able to bind the antibiotic in contrast with other regulators that have been described to bind intermediates or products of their gene cluster, such as TylP (Ray et al., 2017), MphR(A) regulator of erythromycin biosynthesis (Zheng et al., 2009), SimR in Streptomyces antibioticus (Le et al., 2009) and the regulator of tetracyclin production TetR (Hinrichs et al., 1994). However, we hypothesize that one of the other regulators present in the alp cluster, AlpW, could perhaps be able to bind kinamycin or one of its biosynthetic pathway intermediates, as it has been described to be the case for other pseudo-GABRs, such as JadR2 and jadomycin (Xu et al., 2010).

The interactions between AlpZ-DNA complex and signaling molecule constitute a remarkably sensible system for gene regulation and induction of kinamycin biosynthesis, where even small concentration changes are sensed and impact protein-DNA complex stability. The promoter region of the afsA-like gene identified in the proximity of the alp cluster has been shown to contain an ARE motif (unpublished). It would be interesting to explore the role of AlpZ in the expression regulation of the afsA-like gene. One could envision that the AlpZ-AREafsA-like complex would activate gene expression for a positive feedback loop of induction of production of signaling molecule. Even though the nature of this signaling molecule continues to elude us, this work clearly demonstrates that the response of AlpZ to the extract of S. ambofaciens is in no way linked to kinamycin and broadens the knowledge on AlpZ regulatory role, possibly paving the way for similar autoregulator receptors. Efforts are underway for the purification and identification of the S. ambofaciens signaling molecule. The discovery of new signaling molecules in Streptomyces is impaired by the fact that these compounds are typically produced in very small amounts. Moreover, purification scale can be highly variable, avenolide in S. avermitilis required 2000 L of fermentation broth whereas S. coelicolor furans were identified using only 40 squared Petri dishes (Corre et al., 2008; Kitani et al., 2011). Approaches being used include the traditional large-scale fermentation method and specific capture using GABRs (Yang et al., 2005; Zou et al., 2014).

Although several studies exist on autoregulator receptors, there is still a long way to go to identify the cognate ligands of all TetR regulators identified to date (Cuthbertson and Nodwell, 2013). The present work constitutes the first described study of a signaling molecule receptor using an innovative molecular dynamics approach and provides a further step on the way to identifying the signaling molecule in S. ambofaciens. Furthermore, understanding how these regulators control gene expression and respond to the trigger of signaling molecules will ultimately contribute to the development of new approaches for specialized metabolite discovery and production improvement, as well as advancing other application tools.

#### MATERIALS AND METHODS

#### Strains, Media, and Growth Conditions

Strains used in this study are listed in **Table 2**. Streptomyces strains were grown in R2 medium (Kieser et al., 2000) and Bacillus subtilis was grown in LB. Kinamycin production was


assessed on R2 medium as described previously (Pang et al., 2004; Aigle et al., 2005).

# Construction of the Mutant Strain 11alpW11alpID

To obtain a strain of S. ambofaciens producing the signaling molecule controlling the biosynthesis of kinamycin but unable to synthetize the antibiotic itself, we first made an in-frame deletion in the 11alpW mutant strain (Bunet et al., 2011) of the locus alpID on both chromosomal arms. This locus includes the alpA, alpB and alpC genes encoding the minimal PKS and was replaced by a scar. The strategy was based on the REDIRECT system (Gust et al., 2003) and was carried out as described previously (Pang et al., 2004; Bunet et al., 2008). Only the start and stop codons of alpI and alpD, respectively remained after deletion and gene replacement was confirmed by Southern blot and PCR analysis using the CK1 and CK2 primer pair (**Supplementary Table S1**, data not shown).

### Extraction of AlpZ Signaling Molecule

To study the effect of S. ambofaciens signaling molecule on AlpZ DNA-binding crude extracts were prepared as follows. The mutant strain 11alpW11alpID was grown in 50 ml R2 medium in 250 ml flasks and incubated at 30◦C until late exponential-stationary transition phase. Supernatant was collected and extracted twice with 1 volume of ethyl acetate, dried in a rotavapor, and dissolved in 100 µl methanol-H2O 1:1 and stored at −20◦C.

# DNA-Protein Binding Assays

DNA binding assays were performed by EMSA as described previously (Bunet et al., 2008). Briefly, AREV DNA probe was obtained by PCR using the primers listed in **Supplementary Table S1** and directly labeled with digoxigenin using the DIG Oligonucleotide 3'-End Labeling kit, 2nd Generation (Roche Applied Science). Binding assays were performed with AlpZ protein, purified as described previously (Bunet et al., 2008). Labeled probes (0.4 ng) were incubated at 30◦C for 10 min with AlpZ protein (80–160 nM) in binding buffer (20 mM HEPES pH7.6, 1 mM EDTA, 10 mM (NH4)2SO4, 1 mM DTT, 0.2% Tween 20, 30 mM KCl) containing 50 µg/ml poly(dI-C) in a 20 µl final volume. When necessary, S. ambofaciens 11alpW11alpID extracts (1 µl) were added after the incubation period and incubated for further 10 min to verify the presence of signaling molecule. Binding reactions were analyzed with 5% native PAGE and run in 0.5X TBE buffer. DNA was then transferred onto a positively charged nylon membrane (Amersham Hybon-N+) by electroblotting, then fixed by UV crosslinking, detected with antidigoxigenin antibodies, and developed by chemiluminescence with the CDP-StarTM reagent (Roche Applied Science).

### Binding Kinetics and Hydrodynamic Diameter Measurements

A switchSENSE <sup>R</sup> DRX instrument and MPC-48-2-R1-S biochips (Dynamic Biosensors GmbH, Martinsried, Germany) were used to characterize the binding kinetics and protein size changes (Knezevic et al., 2012; Langer et al., 2013).

Specific dsDNA sequences coupled with ssDNA overhang regions complementary to the ssDNA sequences tethered to the biochip (nanolevers) called dsoverhangs, were designed. To prepare the dsoverhangs 2 µM of the two comprising ssDNA sequences (listed in **Supplementary Table S1**) were annealed by heating at 95◦C for 10 min and slowly cooled to room temperature in PE40 buffer pH 7.4 (10 mM Na2HPO4/NaH2PO4, 40 mM NaCl, 0.05% Tween20, 50 µM EDTA, 50 µM EGTA). For the analysis of protein affinity to the dsoverhang, 150 µl of AlpZ at a range of concentrations were injected with a flow rate of 30 µl/min, and dissociation was measured with running buffer at a flow rate of 30 µl/min over 120 min to 4 h. Assays to assess the effect of signaling molecule on the binding of AlpZ to DNA started with a preparation step with dsoverhang (the ligand, either nanoAREV or nanoControl) and 150 µl of AlpZ (the analyte) at 50nM with a flow rate of 30 µl/min without dissociation, followed by a sandwich type of capture injecting 140 µl of signaling molecule containing extract (second analyte) at different concentrations (dilutions performed in PE40) with a flow rate of 20 µl/min. All assays used dsoverhangs at 500 nM and were performed at 30◦C, combining both association/dissociation and size measurements and used PE40 as running buffer. A regeneration solution (alkaline solution pH 13) was used to remove the ligand after each assay.

Binding rates constants (kON and kOFF) and dissociation constants (KD) were determined using real-time measurements of the switching dynamics. Protein size was estimated by comparing the switching dynamics of bound protein with those of bare DNA and with the lollipop biophysical model (Langer et al., 2013). Analysis was performed with the switchANALYSIS software from Dynamic Biosensors and using at least three independent assays.

# Protein Structure Modeling

Structure prediction was performed using a threading/fold recognition method and the server I-TASSER (Zhang, 2009; Yang and Zhang, 2015). After predicting secondary and solvent accessibility, structure templates with the highest significance in threading alignments are identified with LOMETS (Wu and Zhang, 2007). Structure models are predicted and ranked based on the C-score, a confidence score to estimate the quality of models predicted by I-TASSER. It is calculated based on the significance of threading template alignments and the convergence parameters of the structure assembly simulations, and values vary in a range of [−5, 2] where models with high confidence show higher C-score values (Zhang, 2008; Yang et al., 2015).

# Kinamycin D Purification

A solution of kinamycin D was purified by semipreparative HPLC on an Agilent 1100 instrument equipped with an Agilent Zorbax Eclipse RP-C18 column (21 × 100 mm, 5 µm) monitoring the absorbance at 420 nm as described previously (Bunet et al., 2011).

#### DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the article/**Supplementary Material**.

#### AUTHOR CONTRIBUTIONS

CV and LH performed the experiments. CV, J-MG, and BA analyzed the data. CV and BA conceived the experiments and wrote the manuscript. BA supervised the project and obtained the funding. All authors have read and approved the final manuscript.

#### FUNDING

This work was funded by the French National Research Agency through Project ANR MiGenIs (ANR-13-BSV6-0009) and the

### REFERENCES


French National Institute for Agricultural Research (INRA) and Laboratory Excellence ARBRE (ANR-11-LABX-0002-01). CV was also supported by the AgreenSkillsPlus Program (FP7- 609398.0000) and the Région Grand Est.

#### ACKNOWLEDGMENTS

Real-time biomolecular interactions were investigated with the switchSENSE <sup>R</sup> technology available on the ASIA platform (Université de Lorraine-INRAE, https://a2f.univ-lorraine.fr/ asia/).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2020.01255/full#supplementary-material

in Streptomyces ambofaciens by the autoregulator receptor AlpZ and its specific ligand. J. Bacteriol. 190, 3293–3305. doi: 10.1128/JB.01989-07




**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2020 Vicente, Girardet, Hôtel and Aigle. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.