# ENGINEERING THE MICROBIAL PLATFORM FOR THE PRODUCTION OF BIOLOGICS AND SMALL-MOLECULE MEDICINES

EDITED BY : Dipesh Dhakal, Eung-Soo Kim and Mattheos Koffas 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 978-2-88963-253-4 DOI 10.3389/978-2-88963-253-4

#### 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

# ENGINEERING THE MICROBIAL PLATFORM FOR THE PRODUCTION OF BIOLOGICS AND SMALL-MOLECULE MEDICINES

Topic Editors: Dipesh Dhakal, Sun Moon University, South Korea Eung-Soo Kim, Inha University, South Korea Mattheos Koffas, Rensselaer Polytechnic Institute, United States

Citation: Dhakal, D., Kim, E.-S., Koffas, M., eds. (2019). Engineering the Microbial Platform for the Production of Biologics and Small-Molecule Medicines. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-253-4

# Table of Contents


Pin Yu, Qing-Ting Bu, Yi-Li Tang, Xu-Ming Mao and Yong-Quan Li


Fu-Xing Niu, Xin He, Ya-Qin Wu and Jian-Zhong Liu


Ruiquan Qi, Blaine A. Pfeifer and Guojian Zhang

	- *Melanin Production in* Bacillus thuringiensis Wenjun Yang, Lifang Ruan, Jiangming Tao, Donghai Peng, Jinshui Zheng and Ming Sun

*120 Recombinant Protein Expression System in* Corynebacterium glutamicum *and its Application*

Min Ju Lee and Pil Kim

*134 Engineering of the Filamentous Fungus* Penicillium chrysogenum *as Cell Factory for Natural Products*

Fernando Guzmán-Chávez, Reto D. Zwahlen, Roel A. L. Bovenberg and Arnold J. M. Driessen


Rui Pan, Xuelian Bai, Jianwei Chen, Huawei Zhang and Hong Wang

*194 A Review of the Microbial Production of Bioactive Natural Products and Biologics*

Janette V. Pham, Mariamawit A. Yilma, Adriana Feliz, Murtadha T. Majid, Nicholas Maffetone, Jorge R. Walker, Eunji Kim, Hyo Je Cho, Jared M. Reynolds, Myoung Chong Song, Sung Ryeol Park and Yeo Joon Yoon

# Editorial: Engineering the Microbial Platform for the Production of Biologics and Small-Molecule Medicines

Dipesh Dhakal <sup>1</sup> \*, Eung-Soo Kim<sup>2</sup> \* and Mattheos Koffas <sup>3</sup> \*

*<sup>1</sup> Department of Life Science and Biochemical Engineering, Sun Moon University, Asan, South Korea, <sup>2</sup> Department of Biological Engineering, Inha University, Incheon, South Korea, <sup>3</sup> Department of Chemical and Biological Engineering, Rensselaer Polytechnic Institute, Troy, NY, United States*

Keywords: microbial cell factories, metabolic engineering, synthetic biology, heterologous production, biologics and small molecule medicines

**Editorial on the Research Topic**

#### Edited by:

*Weiwen Zhang, Tianjin University, China*

#### Reviewed by:

*Guang Zhao, Qingdao Institute of Bioenergy and Bioprocess Technology (CAS), China Long Liu, Jiangnan University, China*

\*Correspondence:

*Dipesh Dhakal medipesh@gmail.com; dipeshdhakal@sunmoon.ac.kr Eung-Soo Kim eungsoo@inha.ac.kr Mattheos Koffas koffam@rpi.edu*

#### Specialty section:

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

> Received: *07 August 2019* Accepted: *20 September 2019* Published: *09 October 2019*

#### Citation:

*Dhakal D, Kim E-S and Koffas M (2019) Editorial: Engineering the Microbial Platform for the Production of Biologics and Small-Molecule Medicines. Front. Microbiol. 10:2307. doi: 10.3389/fmicb.2019.02307* **Engineering the Microbial Platform for the Production of Biologics and Small-Molecule Medicines**

Microorganisms are the prominent sources of valuable products ranging from large (e.g., proteins, carbohydrate polymers, nucleic acids, even cells) to small molecules (e.g., microbial metabolites, signaling molecules, growth factors, etc.). Most of these small molecules termed as "secondary metabolites (SM)s" are inessential to the producer for their growth and development. However, these SMs have significant applications in human and animal health (Demain, 2000; Bhan et al., 2013; Wang et al., 2016; Dhakal and Sohng, 2017). Besides, several biologics pharmaceutical ingredients extracted from animals, plants, and microorganisms such as antibodies, vaccines, receptor modulators or replacement/modulators of enzymes are applied for human welfare (Kinch, 2005; Lacana et al., 2007). The host microorganisms engineered for the production of such small molecular medicines or relatively complex biologics are termed as "microbial cell factories (MCF)." Recently, metabolic engineering approaches are developed for engineering of metabolism and biosynthetic pathways in these MCFs for better performance (Davy et al., 2017; Choi et al., 2018). The papers published in this Research Topic have attempted to explore the current state of the art of microbial engineering along with its diverse approaches.

Pham et al. have summarized the biological activities and applications of a variety of small molecular medicines and biologics. The manuscript has reviewed the diverse microbial systems used for the production of these biomolecules along with the versatile engineering strategies of such microbial platforms. Generally, each of the microbial strains can produce multiple compounds, but it can produce only subsets of these compounds under specific growth conditions. Therefore, variations in cultivation parameters can elicit the production and discovery of new SMs. For example, by changing cultivation parameters such as temperature, salinity, aeration, and even by altering the shape of the flasks, the production profile from a microbial platform can be altered (Bode et al., 2002). Pan et al. have provided comprehensive information regarding the exploration of structural diversity of microbe secondary metabolites using one strain many compounds (OSMAC) approach. They have presented the role of variation in medium, cultivation conditions, use of epigenetic modifiers, and co-cultivation in the discovery of novel secondary metabolites from diverse microbial sources utilizing OSMAC approach (Pan et al.).

Escherichia coli is reported as the most common cell factory for the production of both small molecules and biologics. The clear understanding of its physiological and genetic characteristics, fast-growth even in minimal salts medium, and availability of easy genetic manipulation techniques has established it as first-choice production host (Liu et al., 2015). Also, systems metabolic engineering approaches that combine knowledge of systems biology, synthetic biology, and evolutionary engineering into the traditional metabolic engineering, has facilitated the development of E. coli as a robust production host for heterologous expression of small molecules and complex biologics (Choi et al., 2019). So, different metabolic engineering approaches utilizing E. coli as microbial platform have been presented in this Research Topic. Wang et al. have reviewed different aspects of terpenoid production using E. coli including the metabolic engineering and genome engineering approaches. Li et al. utilized product resistance and targeted metabolic engineering for the production of equol in E. coli. Similarly, the combination of product tolerance, evolution engineering, and modular co-culture was utilized for the production of pinene (Niu et al.). The metabolic engineering approach utilizing gene over-expression cassette for enhanced production of nucleotide diphosphate (NDP)-sugars was utilized for generating the salicylate glucoside and other glycosylated variants (Qi et al.). Similarly, the gene-silencing approach was employed for enriching the titer of 3′ -phosphoadenosine-5′ -phosphosulfate (PAPS), which is donor substrate for sulfation of natural product (NP) precursors. Hence, by inhibiting the degradation of PAPS mediated by repression of PAPS reductase (cysH) and optimization of different sulfate donors significantly enhanced the production titer of naringenin-7-sulfate (Chu et al.). Ribosomally synthesized and post-translationally modified peptides (RiPPs) are special class of NPs with diverse structures and bioactivities, and thus possess a complex biosynthetic mechanism. Different aspects for heterologous expression of RiPPs in E. coli have been reviewed by Zhang et al..

Due to the presence of endotoxins in products obtained from Gram negative bacteria as E. coli, some of the non-lethal Grampositive bacteria including the native producer strains such as actinobacteria or heterologous hosts [generally recognized as safe (GRAS)] such as Bacillus and Corynebacterium are used as excellent cell factories in industries. Actinobacteria are characterized as the most prominent producers of thousands of bioactive molecules, particularly small molecular medicines such as commercially available antibiotics and anticancerdrugs (Dhakal et al., 2017; Rangseekaew and Pathom-aree, 2019). In some cases, NPs from these actinomycetes are cryptic or not produced in a significant amount. Thus, precise metabolic engineering can be employed in a native host or genetically tractable alternative heterologous hosts for significant production. Li et al. performed whole genome sequencing of the producer strain, analyzed the genome data by computational tools and isolated nocardamine utilizing genome mining of Streptomyces atratus SCSIOZH16. Peng et al. used S. lividans as platform organism and optimized the host for higher heterologous expression of foreign biosynthetic gene cluster (BCG) by modulation by a number of global positive and negative regulatory genes, and genes encoding drug efflux pumps. Further the optimized strain was used for production of NPs of diverse nature such as actinorhodin, murayaquinone, hybrubins, piericidin A1, dehydrorabelomycin, and actinomycin D. Generally, the production of SMs in Streptomyces is controlled by a complex regulatory network that involves pathway-specific, pleiotropic, and global regulators, which tune the expression level of biosynthetic genes in response to a variation in diverse physiological and environmental conditions (van Wezel and McDowall, 2011). Hence, the engineering of such regulation cascades by activators and repressors have significant role in determining the productivity of target molecules. Yu et al. identified AdpAch, as a bidirectional pleiotropic regulator of natamycin biosynthesis in S. chattanoogensis L10. Subsequently, the production titer of natamycin was enhanced by mutating the AdpAch-binding sites, that had an inhibitory effect. Recently, the application of precise genetic engineering based on clustered regularly interspaced short palindromic repeats (CRISPER) and its associated protein (Cas9) has enabled the multiplexed genome engineering of actinomycetes including Streptomyces. Tao et al. have reviewed the application of CRISPR/Cas9 based genome editing in Streptomyces for discovery, characterization, and production of NPs. The recent advances in heterologous expression of RiPPs in Streptomyces have been presented by Zhang et al..

Bacillus species has an ability to adapt to varying environmental conditions and capacity for high production yield (Pham et al.), hence they are crucial industrial microorganisms. Further, the application of recent advances in metabolic engineering, enzyme/pathway engineering along with the synthetic biological tools have contributed to ameliorate the production titer from these microorganisms. Yang et al. utilized enzyme engineering of homogentisate dioxygenase for production of enhanced production of melanin. Similarly, a metabolic engineering approach was utilized for enhanced heterologous production of 2-deoxy-scyllo-inosose in Bacillus subtilis. Unlike E.coli and Bacillus, Corynebacterium has significant ability to utilize a variety of carbon sources (Heider and Wendisch, 2015). C. glutanicum is established as a major industrial producer of proteins, including biologics and enzymes as well as utilized in the production of diverse secondary metabolites as carotenoids, terpenes, and flavonoids. Lee and Kim have reviewed different crucial aspects of recombinant protein expression systems in C. glutanicum and its applications.

Fungi is the second largest kingdom of microorganism after bacteria. They are established as a promising source of bioactive natural products containing unique chemical compounds against various diseases (Singh et al., 2019). Ever since Penicillium notatum was identified as a source of penicillin, there has been immense interest in the exploration of the potential of fungal species for their capacity to produce versatile NPs with biotechnological and pharmaceutical applications. Guzmán-Chávez et al. have summarized on the engineering aspects of the P. chrysogenum for establishing it as a sustainable cell factory for NPs. They have provided the comprehensive summary about the basic biosynthetic logic of such NPs and various rational strategies for activation of biosynthetic gene clusters by optimizing culture parameters or targeted genetic engineering Guzmán-Chávez et al. In addition to the bacteria and fungi, the yeast strain such as Saccharomyces cerevisiae is successfully employed for the production of both bulk and fine chemicals (Kavšcek et al., 2015 ˇ ). The different aspects of biosynthesis and prospects of metabolic engineering for the production of terpenoids in S. cerevisiae are summarized by Wang et al..

Taken together, all these papers illustrate the applicability of engineering of microbial platforms for the production of small molecular medicines to complex biologics. However, in case of all of these microbial cell factories (native, engineered, or heterologous) the industrial scale titer, yield, and productivity is generally difficult to achieve. The major constraint is unavailability of abundant information about their metabolic behavior, unavailability of appropriate genetic engineering tools, or complication in redesigning appropriate flux balance for diverting primary metabolites to target molecules (Bhan et al., 2013). Recently the application of large-scale genome sequencing, gene expression profiling, in silico metabolic modeling and simulation, and enzyme/pathway engineering has eased the rational approaches for metabolic engineering. Particularly, the traditional approach of single strain/pathway specific "try and test" approach is replaced by the application of systems metabolic engineering approach that utilizes integration of strain selection/development, pathway design/engineering, and enzyme selection/engineering for efficient production of target molecules. In addition, the application of tools for generating artificial genetic circuits/metabolic pathways incorporating efficient promoters, RBS, terminators, etc, or multiplexed genome engineering utilizing CRISPR/ Cas9 for gene knock-in/knock out, or activation/repression has advanced the engineering approaches of these MCFs to the next level.

#### REFERENCES


In future, it can be expected that it can be feasible to generate the super host with minimized genome and enriched metabolic pathway centered on particular class of molecules. Such super hosts can be engineered by introducing the synthetic genome to attain the designers' strain for specific target. The burgeoning development in both genetic studies as well as computational approaches such as artificial intelligence (AI) has great prospects for simulating the connection between the genomics and metabolomics to generate the intelligence in these super hosts, so that they can sense the environment condition, and respond rationally.

#### AUTHOR CONTRIBUTIONS

DD wrote the manuscript. E-SK and MK revised and corrected the manuscript. The final draft of the manuscript was finalized and approved for publication by all the authors.

#### ACKNOWLEDGMENTS

The Editors would like to thank all authors that participated in this Research Topic in Engineering the Microbial Platform for the Production of Biologics and Small-Molecule Medicines. We are grateful all reviewers and editorial team members, who has contributed for success of this Research Topic. We are also grateful to support by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MEST) (NRF-2017R1D1A1B03036273) to DD.


**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 Dhakal, Kim and Koffas. 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.

# Bidirectional Regulation of AdpAch in Controlling the Expression of scnRI and scnRII in the Natamycin Biosynthesis of Streptomyces chattanoogensis L10

#### Pin Yu1,2,3, Qing-Ting Bu1,2, Yi-Li Tang1,2, Xu-Ming Mao1,2 and Yong-Quan Li1,2 \*

1 Institute of Pharmaceutical Biotechnology, Zhejiang University, Hangzhou, China, <sup>2</sup> Zhejiang Provincial Key Laboratory for Microbial Biochemistry and Metabolic Engineering, Hangzhou, China, <sup>3</sup> College of Life Sciences, Zhejiang University, Hangzhou, China

#### Edited by:

Eung-Soo Kim, Inha University, South Korea

#### Reviewed by:

Yinhua Lu, Shanghai Institutes for Biological Sciences (CAS), China Yasuo Ohnishi, The University of Tokyo, Japan

> \*Correspondence: Yong-Quan 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: 18 December 2017 Accepted: 09 February 2018 Published: 02 March 2018

#### Citation:

Yu P, Bu Q-T, Tang Y-L, Mao X-M and Li Y-Q (2018) Bidirectional Regulation of AdpAch in Controlling the Expression of scnRI and scnRII in the Natamycin Biosynthesis of Streptomyces chattanoogensis L10. Front. Microbiol. 9:316. doi: 10.3389/fmicb.2018.00316 AdpA, an AraC/XylS family protein, had been proved as a key regulator for secondary metabolism and morphological differentiation in Streptomyces griseus. Here, we identify AdpAch, an ortholog of AdpA, as a "higher level" pleiotropic regulator of natamycin biosynthesis with bidirectional regulatory ability in Streptomyces chattanoogensis L10. DNase I footprinting revealed six AdpAch-binding sites in the scnRI–scnRII intergenic region. Further analysis using the xylE reporter gene fused to the scnRI–scnRII intergenic region of mutated binding sites demonstrated that the expression of scnRI and scnRII was under the control of AdpAch. AdpAch showed a bi-stable regulatory ability where it firstly binds to the Site C and Site D to activate the transcription of the two pathway-specific genes, scnRI and scnRII, and then binds to other sites where it acts as an inhibitor. When Site A and Site F were mutated in vivo, the production of natamycin was increased by 21% and 25%, respectively. These findings indicated an autoregulatory mechanism where AdpAch serves as a master switch with bidirectional regulation for natamycin biosynthesis.

Keywords: bidirectional regulation, AdpA, natamycin biosynthesis, Streptomyces chattanoogensis L10, pathway-specific gene

# INTRODUCTION

The secondary metabolic process in Streptomyces is regulated by a complex regulatory network involving pathway-specific, pleiotropic, and global regulators which respond to a variety of physiological and environmental condition alterations (van Wezel and McDowall, 2011; Liu et al., 2013). The best characterized is the A-factor regulatory cascade in which AdpA is the most important transcriptional factor for the secondary metabolism (Horinouchi, 2002; Ohnishi et al., 2005). In early culture stages, the transcription of adpA in Streptomyces griseus is repressed by ArpA, the receptor protein for A-factor (Onaka and Horinouchi, 1997). When A-factor reaches a critical concentration, it binds to ArpA and confers the conformational change of ArpA (Ohnishi et al., 1999). This results in dissociation of ArpA from the adpA promoter, in turn switching on the expression of adpA (Ohnishi et al., 1999). The induced AdpA then activates the transcription of

various genes related to secondary metabolism such as strR, the pathway-specific regulatory genes for streptomycin in S. griseus (Retzlaff and Distler, 1995; Tomono et al., 2005).

AdpA is a member of the AraC/XylS family proteins (Gallegos et al., 1997). It has been suggested to form a dimer through the N-terminal portion which belong to the ThiJ/PfpI/DJ-1 family (Yamazaki et al., 2004; Ohnishi et al., 2005). To date, a number of AdpA orthologs have been described as having essential roles in the secondary metabolism in many Streptomyces species, such as Streptomyces lividans (Guyet et al., 2013), Streptomyces coelicolor A3(2) (Takano et al., 2001; Nguyen et al., 2003), Streptomyces ansochromogenes (Pan et al., 2009), Streptomyces avermitilis (Komatsu et al., 2010), Streptomyces hygroscopicus 5008 (Tan et al., 2015), and Streptomyces clavuligerus (López-García et al., 2010).

Typically, AdpA is regarded as an activator for downstream regulated genes, except itself which is proved to be negatively auto-regulated by binding to its own promoter region (Kato et al., 2005b; Hara et al., 2009). The molecular mechanism of transcriptional activation begins as a dimer of AdpA binds to the target sites with consensus sequences which then recruit RNA polymerase to the promoter for transcriptional initiation (Yamazaki et al., 2004; Kato et al., 2005a). For different target genes, AdpA showed a different number of binding sites in the promoter regions. For example, there are two AdpA-binding sites in the promoter of strR (Tomono et al., 2005), whereas there are three AdpA-binding sites for regulation of ssgA (Yamazaki et al., 2003a). However, the precise regulation mechanism how the AdpA binds to multiple sites to activate transcription has not been experimentally determined. Based on the importance of AdpA in the biosynthesis of the secondary metabolism, it is necessary to elucidate details of its regulatory mechanisms.

Natamycin, an antifungal polyene macrolide antibiotic, is synthesized by a type I polyketide synthase gene cluster. Previous analysis of the gene cluster of natamycin in Streptomyces chattanoogensis L10 revealed the existence of 17 open-reading frames, including two pathway-specific genes, scnRI and scnRII (Du et al., 2011a). These two genes showed high sequence identity to pimR and pimM of Streptomyces natalensis, respectively (Antón et al., 2007; Santos-Aberturas et al., 2012). Gene disruption of scnRI resulted in a large decrease in the expression of biosynthetic genes, indicating its role as a pivotal activator for the biosynthesis of natamycin (Du et al., 2011a). scnRII, adjacent but divergently transcribed transcriptional regulatory genes, was shown to act as a second positive regulator for natamycin production (Du et al., 2009). We also had proved that AdpAch controls the production of natamycin, but the detailed relationship among AdpAch, ScnRI, and ScnRII had not been well characterized (Du et al., 2011a).

Here, we reveal the sophisticated regulatory characteristics of AdpAch in the natamycin biosynthesis of S. chattanoogensis L10. AdpAch acts as a "higher level" pleiotropic regulator for transcription of the two divergently transcribed pathway-specific genes, scnRI and scnRII. In this regulatory process, AdpAch shows a bi-stable regulatory ability, where it firstly acts as an activator, then a repressor. Moreover, natamycin production was enhanced by mutating the AdpAch-binding sites which had an inhibitory effect. This work not only advances the understanding of detailed regulatory mechanism of AdpA, but also provides a potential target for the enhancement of other antibiotic production levels by manipulating the regulatory network.

# RESULTS

# AdpAch Identified as a "Higher Level" Pleiotropic Regulator for Natamycin Biosynthesis

In our previous study, the biosynthetic gene cluster of natamycin has been cloned and characterized in S. chattanoogensis L10. Within this there are two divergently transcribed genes, scnRI and scnRII, encoding proteins that resemble pathway-specific regulators (Du et al., 2009, 2011a). Although the functions of these two regulators have been well characterized, an important question remains as to whether there are multiple levels of control in the biosynthesis of natamycin. Based on our previous study that AdpAch affected the transcription of these two pathway-specific genes (Du et al., 2011a), we speculated that AdpAch may act as a "higher level" pleiotropic regulator for regulating the natamycin biosynthesis.

To test this hypothesis, electrophoretic mobility shift assays (EMSAs) were applied. As shown in **Figure 1**, retardation was readily detected upon the addition of 50 pM AdpAch with the probe RI–RII, while the addition of 50- to 100-fold excess of unlabeled specific PCR product reduced the proportion of the labeled promoter-containing fragment (**Figure 1**). These data clearly demonstrate that AdpAch could specifically bind to the scnRI–scnRII intergenic region and could control the expression of these two pathway-specific genes.

0, 50, and 100 pM, respectively. Lanes 4 and 5, 50- and 100-fold excess of unlabeled specific PCR product was added into binding reactions.

# DNase I Footprinting Assay Reveals Six AdpAch-Binding Sites in the scnRI–scnRII Intergenic Region

To identify the exact DNA sequences that AdpAch protected in the scnRI–scnRII intergenic region, DNase I footprinting assays, in absence or presence of purified recombinant AdpAch, were performed. In our previous studies, we had determined the transcription start site (TSS) of the two pathway-specific genes, scnRI and scnRII (Du et al., 2011a). As seen in **Figure 2A**, at a lower AdpAch protein concentration of 100 pM, the DNA strands of the scnRI–scnRII intergenic region showed two protected regions, Site C and Site D, extending from positions −69 to −44 and −106 to −74 relative to the TSS of scnRI. When increasing the protein concentration to 500 pM, another four protected regions (Sites A, B, E, and F) were observed. With respect to the scnRI TSS, the AdpAch-binding Site A locates at positions +8 to +54, Site B at positions −20 to +2, Site E at positions −161 to −114, and Site F at positions −283 to −259 (**Figure 2B**). The six AdpAch-binding sites were spread over the scnRI–scnRII intergenic region. Notably, Site A was located downstream of the scnRI TSS, while Site B overlapped the −10 region of the scnRI promoter. Site F was located downstream of the scnRII TSS, and Site E overlapped the −35 region of the scnRII promoter. This data suggest that AdpAch might have a negative regulatory ability for the expression of these two pathway-specific genes. Additionally, the results from the DNase I footprinting assay also reveal that AdpAch may have higher affinity to Site C and Site D than to the others.

# The Consensus AdpAch-Binding Sequence in the AdpAch-Binding Sites

The orthologs of AdpAch identified in S. griseus and S. coelicolor have been reported to have the consensus binding sequence, 5-TGGCSNGWWY-3 (S: G or C; W: A or T; Y: T or C; N: any nucleotide) (Yamazaki et al., 2004). After alignment of these six protected regions, we also found that there were highly conserved AdpAch-binding sequences in each binding site (**Figure 3A**). To further study the roles of these consensus sequences in the AdpAch-binding ability, EMSAs were carried out using the probes containing either the sequences of wild-type (wt) binding sites or the mutated sites (**Figure 3A**). As shown in **Figure 3B**, no binding shift was detected for the mutated sites A–F when compared with their corresponding wt targets. Taken together, these data demonstrated that AdpAch indeed

overlined.

showing the predicted AdpAch-binding sites. The TSS is marked by a bent arrow, the AdpAch-binding sites are underlined, and the −10 and −35 regions are

consensus sequences are in bold, and these consensus sequences are changed with an EcoRI site indicated with underlines. (B) EMSAs for determination of AdpAch binding to mutated sequences. Probes A–F contained the fragment of Sites A–F as shown in A, respectively. Probes mA–mF contained the fragment of corresponding mutated sites. The amounts of AdpAch protein used were 50 and 100 pM.

has six binding sites in the scnRI–scnRII intergenic region and the consensus sequence is essential for the binding activity of AdpAch.

# AdpAch Has Differing Affinities for Different Binding Sites

In the DNase I footprinting analysis, Site C and Site D were occupied with a lower concentration of AdpAch than the other sites. This suggests that there may be affinity differences for AdpAch between the six binding sites. To test this possibility, competitive EMSAs with 50- to 100-fold excess of unlabeled fragments of six AdpAch-binding sites were used to compete with each labeled fragment. As shown in **Figure 4A**, 100-fold excess of unlabeled S<sup>B</sup> 0 (Site B) and S<sup>F</sup> 0 (Site F) could not completely abolish AdpAch complex formation with the labeled probe S<sup>A</sup> (Site A). However, the same amount of unlabeled S<sup>C</sup> 0 (Site C), S<sup>D</sup> 0 (Site D), and S<sup>E</sup> 0 (Site E) outcompeted the labeled probe SA. This result indicated that AdpAch binds to Site A more tightly than Site B and Site F, but less tightly than Site C, Site D, and Site E. Following this way, we could conclude that Site B has less affinity for AdpAch than others, except for Site F (**Figure 4B**), which was the weakest affinity among the six binding sites (**Figure 4F**), and Site D was the strongest affinity of these six sites (**Figure 4D**). The affinity of Site E for AdpAch was between that of Site C and Site A (**Figures 4A,C,E**). Therefore, we determined the affinity of AdpAch to different binding sites in the following order: Site D > Site C > Site E > Site A > Site B > Site F.

#### Promoter-Probe Assays of the AdpAch-Binding Sites in the scnRI–scnRII Intergenic Region

The binding sites of AdpAch in the scnRI–scnRII intergenic region were adjacent to either the scnRI or the scnRII start codon. This raised the possibility that this intergenic region might harbor a bidirectional promoter allowing AdpAch to regulate transcriptions of the divergently transcribed flanking genes, scnRI and scnRII (**Figure 2B**). To investigate the promoter activities of the two pathway-specific genes with each of the AdpAch-binding sites, we used the promoter-probe plasmid pIJ8601 carrying the xylE gene, encoding catechol 2,3-dioxygenase, as the reporter. As shown in **Figure 5A**, the transcriptional profiles of scnRI were severely decreased when the AdpAch-binding Site C and Site D were mutated. Conversely, its transcriptional activity was increased when Site A and Site B were mutated and remained almost unchanged when Site E and Site F were mutated. For the promoter activity of scnRII, we did not detect any consistent differences when Sites A, B, and C were mutated, but mutation in the Sites D and E resulted in a large decreases of up to 70 and 40%, respectively, compared to those of the wt. The mutation in Site F resulted in a statistically significant increase

fmicb-09-00316 February 28, 2018 Time: 16:16 # 4

(**Figure 5B**). These findings indicated that expressions of scnRI and scnRII are both under the control of AdpAch, which has a completely different regulatory ability (activation or inhibition) when binding to different binding sites.

# Effect of Mutated AdpAch-Binding Sites on Natamycin Production in Vivo

There have been some reports where effects upon DNA-binding sites were found in vitro that failed to be exhibited in vivo. In order to test this possibility and reveal the function of the six AdpAch-binding sites in natamycin biosynthesis in vivo, a series of mutants were constructed as described in Experimental procedures. As shown in **Figure 6A**, compared to the WT strain, the level of natamycin production in the R-mA (mutation in Site A) and R-mF (mutation in Site F) had increased by 21 and 25%, respectively. However, the constructed strains of R-mC (mutation in Site C), R-mD (mutation in Site D), and R-mE (mutation in Site E) showed up to 31, 42 and 15% reductions, respectively. The natamycin production of R-mB (mutation in Site B) mutant exhibited almost no change. This finding indicated that the AdpAch-binding Sites A and F play negative roles for natamycin biosynthesis, while the functions of the Sites C, D, and E were positive. Quantitative real-time PCR (qRT-PCR) analysis showed that the promoting effect of site mutation on natamycin production was due to alteration of the pathway-specific genes at the transcriptional level (**Figure 6B**).

# DISCUSSION

Streptomyces spp. have developed complicated mechanisms to adapt to altered circumstances (Santos-Beneit et al., 2009; Yu et al., 2012). Among these mechanisms, the multiple levels of regulation in controlling the expression of the genes responsible for the formation of the secondary metabolism are drawing increased attention. In this study, we focused on the regulatory network of natamycin biosynthesis in S. chattanoogensis L10,

an industrial strain for natamycin production. In our previous study, we determined that gamma-butyrolactones (GBLs) serve as quorum-sensing signaling molecules for activating natamycin production in S. chattanoogensis L10 (Du et al., 2011b), and ScnRII acts as a positive regulator by directly binding to the promoters of natamycin biosynthetic genes (Du et al., 2009) where ScnRI acts as a positive regulator for the transcription of scnRII (Du et al., 2011a). However, the deletion of scnRI did not result in a complete halt of the transcription of scnRII (our unpublished data). This is quite different from the function of PimR in S. natalensis where the deletion of pimR almost completely destroys the transcription of pimM (Antón et al., 2004; Santos-Aberturas et al., 2012). As the regulation of antibiotic biosynthesis involves numerous transcription factors (McKenzie and Nodwell, 2007; van Wezel and McDowall, 2011), participation of other regulator(s) is possible, in the regulation of scnRII.

With AdpAch being able to regulate the expression of both of the pathway-specific genes, scnRI and scnRII, it provides a possible explanation that there is a coordinate regulation in

controlling expression of scnRII by AdpAch and ScnRI. This regulatory pattern may occur in following steps. Firstly, AdpAch binds to the scnRI–scnRII intergenic region and activates both transcription of scnRI and scnRII. Then ScnRI also binds to the scnRI–scnRII intergenic region which, in turn, promotes the transcriptional level of scnRII. However, these two genes were not completely controlled by AdpAch. Trace expression of scnRI was observed in the adpAch mutant, and then ScnRI would promote the transcription of scnRII (Du et al., 2011a). Notably, a certain amount of AdpAch is required for binding to the scnRI–scnRII intergenic region (∼50 pM). This is why we did not detect the shifted band with low concentration AdpAch (∼1 pM) in the binding reaction of our previous study (Du et al., 2011a).

In most cases, AdpA acts as an activator for the target genes, except for itself where it exhibits an autorepression (Kato et al., 2005b). In this study, we concluded from promoterprobe assays in vivo that AdpAch could not only regulate both

pathway-specific genes, but also displayed completely opposite regulatory abilities in control of them. The AdpAch-binding Site C and Site D were involved in activating the transcription of scnRI, while AdpAch binding to Sites A and B resulted in repression. For the promoter activity of scnRII, mutation in the Site C and Site D resulted in a decrease of transcriptional profiles, while a mutation in the Site F led to a statistically significant increase. A similar phenotype was observed in S. ansochromogenes where transcription of sanG decreased when Site I and Site V were mutated but increased when other three AdpA-L-binding sites were mutated (Pan et al., 2009). However, when combinations of binding site mutations were carried out, the promoter activities were not in accordance with our predictions. For example, mutations in both Sites E and F reduced the transcriptional level of scnRII (data not shown). Based on the short distances between the AdpAch-binding sites which are spread over the scnRI–scnRII intergenic region, there may be complicated interactions between different AdpAch dimmers to explain this.

With further analysis using competitive gel shift assays, we could conclude that AdpAch binds to Sites A–F with the following affinities: Site D > Site C > Site E > Site A > Site B > Site F (**Figure 4**). These data are consistent with the footprinting assay where the regions of Site C and Site D were previously protected at a lower AdpAch protein concentration (**Figure 2A**). This gives a hint that the regulatory ability of AdpAch may occur in a growth phase-dependent manner. In the early stage, AdpAch firstly binds to the Site C and D to recruit RNA polymerase to the promoter and initiates the transcription of scnRI and scnRII. This in turn triggers natamycin production (**Figure 7**). When AdpAch is accumulated to a certain critical level, it will bind to other binding sites located near the TSS. A DNA loop may be formed via the interaction between different AdpAch dimers, thus preventing RNA polymerase from access to the promoter of the pathway-specific genes (**Figure 7**). Reduced transcription of the pathway-specific genes will result in a low rate of natamycin production.

The discovery of this bidirectional regulation of AdpAch in the control of natamycin biosynthesis reveals an artful adaptive mechanism in microbial cells. Microorganisms produce molecules with antibiotic activity and expel them into the environment, presumably enhancing their ability to compete with their neighbors (Berdy, 2005; Hopwood, 2007). However, most of these molecules are toxic to the producer (Mak et al., 2014; Moody, 2014). Mechanisms must exist to ensure that antibiotic production reaches a reasonable level. The proposed model of AdpAch in **Figure 7** may provide a fresh mechanistic insight into how S. chattanoogensis controls the production level of natamycin via AdpAch. However, further work will be needed to prove the proposed model and the detailed mechanism of how AdpAch responds to the signal of natamycin. In all, the complicated regulatory network involving AdpAch, ScnRI, and ScnRII helps advance our understanding of the molecular regulation mechanisms of antibiotic biosynthesis and provides an effective strategy to help improve yields in industrial strains.

#### MATERIALS AND METHODS

#### Media, Plasmids, Strains, and Growth Conditions

All plasmids and bacterial strains used in this study are listed in **Table 1**. General techniques for the manipulation of nucleic acids and bacterial growth were carried out according to the standard protocols as previously described (Kieser et al., 2000). Escherichia coli DH5α was the general cloning host. Vectors used were pSET152, pIJ8660, pTA2. S. chattanoogensis L10 strains were grown at 28◦C on YMG agar for sporulation and at 30◦C

and the presumptive manner of AdpAch-binding forms a dimmer. Stage 1: AdpAch binds to the Site C and Site D when the concentration of AdpAch is low in the early growth stage. Stage 2: When the concentration AdpAch reaches a high level, AdpAch binds to Site A, Site B, Site E, and Site F, most of which located downstream of the −35 sequences of the corresponding genes. A DNA loop may be formed via interaction between different AdpAch dimmers, thus preventing RNA polymerase from access to the promoter.

TABLE 1 | Bacterial strains and plasmids used in this work.

fmicb-09-00316 February 28, 2018 Time: 16:16 # 8


in YEME medium (3 g/l yeast extract, 3 g/l malt extract, 5 g/l tryptone, 10 g/l glucose) for natamycin production.

#### Electrophoretic Mobility-Shift Assays (EMSAs)

His-AdpAch, histidine-tagged protein was purified from the soluble fractions of E. coli BL21 (DE3) harboring the plasmids pET32a-adpAch, as previously described (Du et al., 2011a). The Bradford reagent (Bio-Rad) was used to determine the protein concentration. For probe preparation, all primers used in this study are listed in Supplementary Table S1. The EMSA DNA probe RI–RII (517 bp) spanning the entire scnRI–scnRII intergenic region was amplified by PCR using primer pair RI–RII-F and RI–RII-R. The PCR product was then cloned into a pTA2-vector (TOYOBO) to generate the plasmid pT-RI–RII. The biotin-labeled probe RI–RII was made with 5<sup>0</sup> -biotin-labeled M13 universal primer pair using pT-RI–RII as a template by PCR

amplification. The probes A (295 bp), B (281 bp), C (294 bp), D (282 bp), E (288 bp), F (284 bp), mA (295 bp), mB (281 bp), mC (294 bp), mD (282 bp), mE (288 bp), and mF (284 bp) were prepared following the above-mentioned method. In the EMSAs assay, 1 ng of the probe was incubated with varying quantities of AdpAch, at 25◦C for 30 min in the buffer (20 mM Tris, pH 7.5, 5% glycerol, 0.01% BSA, 50 µg ml−<sup>1</sup> sheared sperm DNA). For the competition assay, 100 times of excessive un-labeled probes and non-specific DNA were added to the reaction buffer, respectively. Reactions were displayed on 5% acrylamide gels for separation in 0.5× TBE buffer. EMSA gels were then electro-blotted onto the nylon membrane and UV-fixed by UV crosslinker. Labeled DNA was detected with streptavidin-HRP and BeyoECL plus (Beyotime, China) as described by the manufacturer.

# DNase I Footprinting Assay

DNase I footprinting assay was performed as previously described (Mao et al., 2009). Firstly, AdpAch protein was ultra-filtered with YM-10 (Millipore) for 10 kD cut-off and eluted in 20 mM Tris buffer, pH 7.5. Then, FAM-labeled probe was amplified using 5<sup>0</sup> -(6-FAM)-labeled M13 universal primers from plasmid pT-RI-RII, followed by gel recovery. About 50 ng of fluorescently labeled probe was added to the reaction mixture to a final volume of 50 µl. After binding of the AdpAch protein to 5<sup>0</sup> -(6-FAM)-labeled probe (30◦ c, 30 min), 0.01 U of DNase I (Promega) was added for 1 min at 30◦C, followed with equal volume of 100 mM EDTA to stop the reactions and extracted by phenol/chloroform. After precipitation with 40 µg of glycogen, 0.75 M ammonium acetate (NH4Ac), and ethanol, the digested DNA mixture was loaded into ABI 3130 DNA sequencer with Liz-500 DNA marker (MCLAB). DNA sequencing ladder was prepared according to Thermo Sequenase Dye Primer Manual Cycle Sequencing Kit (USB).

# Alterations of the Consensus Sequence for AdpAch-Binding Sites

The consensus sequence of AdpAch-binding sites A–F was replaced by the sequence of EcoRV restriction sequence sites using overlapping primers (Supplementary Table S1). The PCR product was then cloned into a pTA2-vector (TOYOBO). The resulted plasmids were used as template for PCR to amplify mutated probes using 5<sup>0</sup> -biotin-labeled M13 universal primers, and the binding ability was measured by EMSAs.

# Construction and Analysis of Transcriptional Fusions to the xylE Reporter Gene

For xylE fusions, the xylE gene was PCR amplified with the primers xylE-F and xylE-R. This fragment was digested with NdeI and NotI, and introduced into the likewise-digested pIJ8660 (Sun et al., 1999) to construct pIJ8601. To probe scnRIp and scnRIIp activities with the mutation of AdpAch-binding sites, the wt and mutated promoter regions were amplified by PCR using upstream primers carrying a BamHI site listed in Supplementary Table S1. These promoter fragments were cloned into BamHI-cut pIJ8601 and transferred by conjugation into S. chattanoogensis L10. Plasmid-containing strains were grown on YEME medium for 24 h. Cell pellets from 1 ml culture samples were kept on ice and measured immediately. Assays of catechol 2,3-dioxygenase were performed as previously described (Kieser et al., 2000).

### Mutational Analysis of the AdpAch-Binding Sites on Natamycin Biosynthesis

The 1.8 kb DNA fragment containing the sequence of scnRI– scnRII intergenic region was amplified by PCR using primers scnRI-F and scnRII-R. The resulted 1.8 kb sequence was used as template to amplify the DNA fragment for construction of mutated AdpAch-binding sites in vivo using overlapping primers (Supplementary Table S1), then PCR product was purified and ligated into pKC1139. The resulting plasmids containing DNA fragment of mutated sites was conjugated by E. coli ET12567/pUZ8002 into S. chattanoogensis L10. The mutants were selected by replica plating for apramycin-sensitive colonies and they were used as template for PCR with primer pairs RI-RII-F and RI-RII-R. The amplified sequences were digested with EcoRV to confirm the mutants.

# Determination of Natamycin Production by HPLC Analysis

Natamycin production was confirmed by HPLC analysis with the Agilent 1100 HPLC system. HC-C<sup>18</sup> column (5 µm, 4.6 by 250 mm) was used with UV detector set at 303 nm. Mobile phase and gradient elution process were as described previously (Du et al., 2009).

# AUTHOR CONTRIBUTIONS

PY, Q-TB, and Y-LT performed the experiments. X-MM assisted with the primary data analysis. Y-QL supervised the project and revised the manuscript. All authors reviewed the manuscript.

# FUNDING

This work was supported by the National Key Research and Development Program (2016YFD0400805), the National Natural Science Foundation of China (Nos. 31520103901 and 31470212), and the China Postdoctoral Science Foundation (2016M601953).

# ACKNOWLEDGMENTS

We gratefully thank Dr. Chris Wood, a native English biologist for his critical reading of this manuscript.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb.2018. 00316/full#supplementary-material

# REFERENCES

fmicb-09-00316 February 28, 2018 Time: 16:16 # 10



**Conflict of Interest Statement:** 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 © 2018 Yu, Bu, Tang, 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 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.

# To Construct an Engineered (S)-Equol Resistant E. coli for in Vitro (S)-Equol Production

Hailiang Li1,2,3† , Shaoming Mao<sup>3</sup>† , Huahai Chen1,2, Liying Zhu<sup>2</sup> , Wei Liu<sup>2</sup> , Xin Wang<sup>2</sup> \* and Yeshi Yin1,2 \*

<sup>1</sup> Key Laboratory of Comprehensive Utilization of Advantage Plants Resources in Hunan South, College of Chemistry and Bioengineering, Hunan University of Science and Engineering, Yongzhou, China, <sup>2</sup> State Key Laboratory Breeding Base for Zhejiang Sustainable Pest and Disease Control, Institute of Plant Protection and Microbiology, Zhejiang Academy of Agricultural Sciences, Hangzhou, China, <sup>3</sup> Hunan Provincial Key Laboratory for Forestry Biotechnology, College of Life Science and Technology, Central South University of Forestry and Technology, Changsha, China

#### Edited by:

Mattheos Koffas, Rensselaer Polytechnic Institute, United States

#### Reviewed by:

Haoran Zhang, Rutgers, The State University of New Jersey, United States Gyoo Yeol Jung, Pohang University of Science and Technology, South Korea

#### \*Correspondence:

Xin Wang xxww101@sina.com Yeshi Yin yinyeshi@126.com †These authors have contributed equally to this work.

#### Specialty section:

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

> Received: 22 February 2018 Accepted: 15 May 2018 Published: 04 June 2018

#### Citation:

Li H, Mao S, Chen H, Zhu L, Liu W, Wang X and Yin Y (2018) To Construct an Engineered (S)-Equol Resistant E. coli for in Vitro (S)-Equol Production. Front. Microbiol. 9:1182. doi: 10.3389/fmicb.2018.01182 (S)-equol is one of the major metabolites of daidzein that is produced by human and animal gut bacteria. Most of the physiological functions of soybean isoflavones, such as anti-oxidative activity, anti-cancer activity, and cardiovascular protection have been ascribed to (S)-equol. However, only 30–50% people contain this kind of equol-producing bacteria, and therefore are able to convert daidzein to (S)-equol. Administration of (S)-equol may be more beneficial than soybean isoflavones. The aim of this study was to construct an engineered (S)-equol resistant Escherichia coli to enhance (S)-equol production in vitro. First, transposon mutagenesis libraries were constructed and screened to isolate the (S)-equol resistant mutant E. coli strain BL21 (ydiS) in order to overcome the inhibitory effects of (S)-equol on bacterial growth. Bacterial full genome scan sequencing and in vitro overexpression results revealed that the ydiS gene was responsible for this resistance. Second, the (S)-equol-producing genes Ldznr, L-ddrc, L-dhdr, and L-thdr of Lactococcus strain 20–92 were synthesized and cloned into compatible vectors, pETDuet-1 and pCDFDuet-1. These plasmids were subsequently transformed into BL21 (DE3) and its mutant BL21 (ydiS). Both engineered BL21 (DE3) and BL21 (ydiS) could use daidzein as substrate to produce (S)-equol under both anaerobic and aerobic conditions. As expected, engineered BL21 (ydiS) had faster growth rates than BL21 (DE3) when supplemented with high concentrations of (S)-equol. The yield and the daidzein utilization ratio were higher for engineered BL21 (ydiS). Interestingly, engineered BL21 (ydiS) was able to convert daidzein to (S) equol efficiently under aerobic conditions, providing a convenient method for (S)-equol production in vitro. In addition, a two-step method was developed to produce (S)-equol using daidzin as substrate.

Keywords: (S)-equol production, (S)-equol resistance, soybean isoflavone, transposon mutagenesis, ydiS gene

#### INTRODUCTION

Soy isoflavones have multiple health benefits due to their anti-carcinogenic, anti-oxidant, and anti-atherosclerotic properties (Xiao et al., 2017). These chemicals also interact with the estrogen receptor, enabling them to act as weak to moderate phytoestrogens (Nielsen and Williamson, 2007). Interestingly, a variety of studies have suggested that the clinical effectiveness of isoflavones might

**19**

be due to their metabolites (Setchell et al., 2002; Sarkar and Li, 2003; Cobb et al., 2006; Cooke, 2006; Jackman et al., 2007). In 2002, Setchell et al. (2002) proposed the "Equol Hypothesis," which posits that daidzein is converted to (S)-equol by gut bacteria in certain individuals, and that it is the equol that accounts for the noted health benefits of soy isoflavones.

Equol (7-hydroxy-3-[4-hydroxyphenyl]-chroman) was first isolated from equine urine in 1932 (Marrian and Haslewood, 1932) and was also identified 50 years later in human urine as a metabolite of soy isoflavones (Axelson et al., 1982). Equol is not present in soybeans, but it is produced naturally in the human gut by daidzein, a major isoflavone predominantly found in soybean, by intestinal bacteria (Setchell and Clerici, 2010). Equol exhibits stronger anti-oxidant and estrogenic activities than daidzein (Hwang et al., 2003; Kinjo et al., 2004; Turner et al., 2004; Rufer and Kulling, 2006) and has been demonstrated to act as vasorelaxant (Jackman et al., 2007), along with having anti-inflammatory properties (Blay et al., 2010), which have both been observed previously in soy isoflavones (Chacko et al., 2005; Hall et al., 2008). However, only 30–50% of the human population can produce equol (Low et al., 2005; Ozasa et al., 2005; Setchell and Cole, 2006; Hall et al., 2007). This suggests that health effects of functional foods supplemented with (S)-equol could be more beneficial than daidzein. Several studies have demonstrated that a diet supplemented with natural (S)-equol alleviates menopausal symptoms, such as hot flushes and crow's feet wrinkles (Aso et al., 2012; Oyama et al., 2012).

Currently, the majority of equol production is performed by chemical synthesis, although production of (S)-equol via bacterial fermentation may have several advantages over chemical synthesis. Specific intestinal bacteria are responsible for the conversion of daidzein to (S)-equol, such as Coriobacteriaceae sp. and Lactobacillus sp. (Setchell and Clerici, 2010). The Hishigaki group has cloned and identified a gene cluster responsible for converting daidzein to (S)-equol from an equol-producing strain Lactococcus 20–92 (Shimada et al., 2011, 2012). The gene product of L-dznr is responsible for converting daidzein into (R)-dihydrodaidzein; the L-ddrc gene product converts (R)-dihydrodaidzein into (S)-dihydrodaidzein; the L-dhdr gene product converts (S)-dihydrodaidzein into trans-tetrahydrodaidze; and the L-thdr gene product converts trans-tetrahydrodaidzein into (S)-equol (Shimada et al., 2011, 2012). Lee et al. (2016) constructed a recombinant Escherichia coli BL21 strain which can produce (S)-equol in vitro. However, Vázquez et al. (2017) reported that isoflavone-derived compounds like (S)-equol have the ability to inhibit the growth from many bacteria species. The aim of this study was to obtain an (S)-equol resistant host E. coli, which can be engineered for (S)-equol production by co-expressing the equol-producing genes L-ddrc, L-dznr, L-dhdr, and L-thdr. As a result, a putative oxidoreductase gene ydiS was identified to be responsible for the (S)-equol resistance. An engineered equol-producing bacterial strain was constructed using an (S)-equol resistant mutant [E. coli BL21 (ydiS)] to coexpress the equol-synthesis genes. A two-step method was utilized to convert diadzin to (S)-equol under aerobic conditions. All results of this study have been summarized in a schematic diagram (**Figure 1**).

# MATERIALS AND METHODS

#### Chemicals and Reagents

fmicb-09-01182 June 1, 2018 Time: 11:17 # 3

Daidzin, daidzein and (S)-equol were purchased from Daicel Chiral Technologies Co., Ltd. (Shanghai, China). The antibiotics and isopropyl-D-thiogalatopyranoside (IPTG) were ordered from Sangon Biotech Bio (Shanghai, China). The restriction enzymes and ligation kit were purchased from TaKaRa Bio (Dalian, China).

## Bacteria Strains, Plasmids, and Growth Conditions

Detailed information regarding the strains and plasmids used in this study is listed in Supplementary Table 1. In brief, E. coli strains DH5α and BL21 (DE3) were ordered from TaKaRa Bio (Dalian, China) and Transgene Biotech (Beijing, China), respectively. Mariner transposon plasmid pFAC (Wong and Mekalanos, 2000) and a dap auxotroph E. coli strain WM3064 (Saltikov and Newman, 2003) were obtained from Dr. Gao's laboratory (Zhejiang University, Hangzhou, China). DH5a E. coli (pRK2013) was ordered from Biomedal S. L.<sup>1</sup> . The genes L-dznr (GenBank accession number: AB558141.1), L-ddrc (GenBank accession number: AB694972.1), L-dhdr (GenBank accession number: AB592970.1), L-thdr (GenBank accession number: AB592969.1), and bgl (GenBank accession number: JQ957567.1) were synthesized and then sub-cloned the pUC57 vector by GenScript Biotechnology (Nanjing, China). The ydiS gene was PCR amplified using primers ydiS-F: 5-ATG TCG GAT GAC AAA TTT GAT GCC A-3, and ydiS-R: 5-ATC GCG CCA ACG AGG GAA TTA-3. The ydiT gene of BL21 (ydiS) was synthesized and sub-cloned into the pRSFDuet-1 vector by GenScript Biotechnology (Nanjing, China). E. coli compatible vectors pRSFDuet-1, pETDuet-1, and pCDFDuet-1 were acquired from Merck Millipore (Germany). E. coli strains were grown at 37◦C in Lennox broth (LB) or LB agar. When required, 50 µg/mL carbenicillin, 50 µg/mL streptomycin, and 15 µg/mL kanamycin were added to the broth or plates. When required, an anaerobic chamber (anaerobic workstation AW 500, Electrotek Ltd., United Kingdom) was employed to minimize oxygen exposure.

#### Transposon Mutagenesis Library Screening

In this study, the transposon of the pFAC plasmid, consisting of a transposable element flanked by two inverted repeats of 27 bps (5<sup>0</sup> -aca ggt tgg ctg ata agt ccc cgg tct-3<sup>0</sup> ) and a gentamycin resistance cassette in the middle (aacC1: 534 bp) was used. A gene encoding the hyperactive mariner transposase, and a gene encoding β-lactamase (bla) were included in this plasmid (Withers et al., 2014). In theory, the promotor of gentamycin (PGm) transferred together with the mariner transposon, causing adjacent genes to be overexpressed or repressed, dependent on the transcript directions for the gentamycin promotor and its downstream genes. Transposon mutagenesis was prepared via conjugation utilizing pFAC plasmid-carrying E. coli WM3064 as the donor strain and BL21 (DE3) as the recipient strain. Transfer of plasmids from WM3064 to BL21 (DE3) were performed via tripartite conjugations using the helper plasmid pRK2013. In brief, bacterial E. coli WM3064, BL21 (DE3), and DH5α (pRK2013) were incubated in LB media at 37◦C overnight, 500 µL of each bacterium was then mixed together in a 2 mL tube. After centrifuging, the bacterial pellet was resuspend using LB media and transferred onto a dry LB plate (supplemented with 2,6-diaminopimelic acid) in three compact droplets. After incubation for ∼6 h, the bacteria were gathered and streaked onto LB plates supplemented with gentamicin (15 µg/mL). Bacterial colonies were then seeded into 96-well plates containing LB media supplemented with 200 µg/mL (S)-equol. The OD<sup>600</sup> was detected using a Model 680 microplate reader (Bio-Rad, United States) and SP-2000UV spectrometer (Shanghai Spectrum Instruments Co., Ltd.). The equol-resistant character of the five clones was further verified using a tube culture method. Chromosomal DNA of these (S)-equol resistant mutants was isolated using an OMEGA Genomic DNA Extraction Kit (Omega, United States). Taxa identification was performed using by 16s rRNA sequencing and BLAST analysis. The mutant E. coli strain BL21 (ydiS) was then selected for full genome sequencing using Illumina Hiseq at Majorbio Bio-Pharm Technology Co., Ltd., Shanghai, China. The draft genome sequence data of BL21 (ydiS) has been deposited in NCBI as Accession Number PIYU00000000.

## Detection of Bacterial Growth Rate

For static culture, 30 mL of LB media were added to a 100 mL flask bottle. After autoclaving, bacteria, antibiotics and chemical reagents were then added into the bottle. Bacterial density (OD600) was measured every 3 h. For culturing under shaking conditions, a real-time detection instrument Microscreen-16 (Gering Instrument Manufacturing (Tianjin) Co., Ltd., Tianjin, China) was used. 40 mL of LB media were added to the 50 mL measure bottle, and 400 rpm (equivalent to 100 rpm in the general shake incubator) was utilized for stirring. The optical absorption value was measured at 30-min intervals. Optical absorption was detected at OD850, and a conversion factor between OD<sup>850</sup> and OD<sup>600</sup> was calculated using E. coli before the experiment.

#### Batch Culture Fermentation for Equol Production

Batch culture fermentations were cultured without shaking at 37◦C, under both anaerobic and aerobic conditions. Briefly, a basic growth medium, LB (Qingdao Hope Bio-Technology Co., Ltd., Qingdao, China), was used to assess the utilization of daidzin or daidzein in the engineered equol-producing E. coli. For fermentation, bacteria were aliquoted into 50 mL flask bottles containing 20 mL of culture media supplemented with either daidzin or daidzein. Twenty microliters of IPTG (25 mg/mL) was added to each bottle to induce gene expression. Samples then were collected at 48 and 72 h after IPTG induction to detect equol and daidzein via HPLC. For production of

<sup>1</sup>http://lifescience.biomedal.com/

(S)-equol from daidzin, a two-step fermentation was attempted in this study. For the first step, daidzin was transformed to daidzein using E. coli (pRSFDuet-1-bgl) under fermentation 72 h. The fermentation liquid was then collected after centrifugal separation, and the upper liquid was used for preparing a new LB media (LB-D). For the second step, DDDT-BL21 (ydis) was inoculated into the LB-D media to detect the equol production.

#### HPLC Detection

fmicb-09-01182 June 1, 2018 Time: 11:17 # 4

Identification of equol and daidzein was performed using HPLC according to a previously described method with some modification (Decroos et al., 2005). In brief, 1 mL of each sample was extracted three times with 1 mL acidic ether, then the ether fractions were combined, evaporated to dryness and resuspended in 200 µL of methanol and stored at −20◦C until analysis. HPLC analysis was performed using a Waters e2695 system. Fifteen microliter aliquots of each sample were injected and separated using a SunFireTM C18 5 µm column (4.6 mm × 205 mm). The temperature was set at 30 ± 2 ◦C and the flow rate was maintained at 0.8 mL/min. Elution was isocratic with a mobile phase consisting of 0.01% formic acid:methanol:acetonitrile (50:20:30). Equol was detected at 205 nm; daidzein at 254 nm. Calibration curves for the quantification of daidzein and equol were constructed using pure standards obtained from Daicel Chiral Technologies Co., Ltd. (Shanghai, China).

#### Statistical Analysis

SPSS Software (version 20.0; SPSS Inc., United States) and the Student's t-test was employed in this study. P < 0.05 was considered to be statistically significant.

#### RESULTS

#### Inhibitory Effects of (S)-Equol on Host E. coli BL21 (DE3)

Previously, (S)-equol was shown to inhibit the growth of representative human gut bacteria (Vázquez et al., 2017). However, the inhibitory effects of the fermentation product (S)-equol on host bacterial E. coli BL21 (DE3) requiring further investigation. In this study, an E. coli strain was engineered to coexpress the four equol-producing genes L-ddrc, L-dznr, L-dhdr, and L-thdr, which originated from an equol-producing bacterial Lactococcus strain 20–92 (Supplementary Figure 1). In order to evaluate the equol-producing activity of the engineered E. coli, 50 µg/mL (∼200 µM) daidzein was added to LB culturing media and (S)-equol production was detected under both anaerobic and aerobic conditions at different time points after IPTG induction. As demonstrated in Supplementary Figures 2A,D, the metabolites from the engineered E. coli had a similar HPLC peaks to the reference standard for (S) equol. The LC-MS results further verified that the equol peak detected by HPLC had the same molecular weight as the (S) equol reference standard (Supplementary Figures 2B,C,E,F). Previous studies have reported that the equol produced by gut bacteria is (S)-equol (Setchell et al., 2005), and that the metabolite produced by Lactococcus strain 20–92 is (S)-equol (Shimada et al., 2011, 2012), the metabolite produced by the engineered E. coli was likely to be (S)-equol (Supplementary Figure 3). However, bacterial growth rates were inhibited during fermentation (Supplementary Figure 4); bacterial density decreased 30 and 37% under anaerobic and aerobic conditions, respectively, after IPTG was added as an inducer for 24 h. In addition, there was almost no bacterial growth observed even cultured for 72 h (Supplementary Figure 4). In order to clarify which compounds had inhibitory effects on the growth of BL21 (DE3), bacterial plates were prepared using different compounds. The fermentation product (S)-equol did inhibit BL21 (DE3) growth both under both anaerobic and aerobic conditions (Supplementary Figure 5), however, daidzein did not cause inhibition. In order to further verify the inhibitory effects of daidzein and equol on bacterial growth, static liquid culture and shake culture experiment were done under aerobic conditions at 37◦C. Supplementary Figure 6 illuminated that equol has the ability to inhibit BL21 (DE3) growth which was dependent on the equol concentration. However, daidzein not inhibit growth and may have slightly promoted bacterial growth for both BL21 (DE3) and BL21 (G2) (Supplementary Figures 6A,B).

## Screening of (S)-Equol Resistant BL21 (ydiS) Mutants

The feedback inhibitory effects of the fermentation products could prevent high yields of (S)-equol product, transposon mutagenesis libraries were constructed and screened in order to develop (S)-equol resistant bacteria. In summary, thounds of bacterial clones grown on LB+Gm plates from the mutant library. 93 clones were then randomly picked up and then seeded in a 96-well plate to evaluate their equol resistance. The growth rate was monitored at different time points using a microplate reader. Among screened clones, five clones were identified as growing faster than BL21 (DE3) in the presence of 200 µg/mL (S)-equol (data not shown). Twenty-milliliter tubes containing 5 mL of LB and 5 µL of (S)-equol were further used to verify the equol-resistance of these clones (Supplementary Figure 7). 16s rRNA gene sequencing and BLAST analysis were then used for taxonomy analysis of these clones. Mutant E. coli strain BL21 (DE3)\_G2 [renamed as BL21 (ydiS)] was selected for further verification and study. The bacterial density of BL21 (ydiS) was higher than BL21 (DE3) under both static and shaken culture conditions in the presence of 50 or 100 µg/mL (S)-equol (**Figure 2**). This BL21 (ydiS) strain was then selected for equol production and has been deposited into the China General Microbiological Culture Collection Center (CGMCC No. 14219).

#### Identification of Equol Resistant Gene in BL21 (ydiS)

In order to identify the transposon insert site and infer the equol resistant mechanism of BL21 (ydiS), full genome scanning and reverse PCR sequencing were performed. The sequencing results revealed that the PFAC transposon was inserted at 307 bp upstream of the ydiS gene. The direction

FIGURE 2 | (S)-equol resistance of BL21 (DE3) mutant. Growth rates under static (A) and shaking (B) culture conditions for the BL21 (DE3) and mutant BL21 (ydiS) were compared when different concentration of (S)-equol were added. At each timepoint, two duplications were measured for each sample, means and standard deviations (SD) were calculated.

of PFAC Gm promotor is same as the ydiS gene (**Figure 3A**). Overexpression of ydiS and its downstream gene ydiT may have contributed to the equol resistance for BL21 (ydiS). YdiS and ydiT genes were then cloned into pRSFDuet-1, respectively. When the OD<sup>600</sup> reached ∼0.6, 5 µL of IPTG (25 mg/mL) was then added to induce foreign protein expression under aerobic condition. As indicated in **Figure 3B**, strains that overexpressed ydiS had faster growth rates than pRSF-Duet BL21 (DE3) under both static and shaking culture conditions when supplemented with 100 µg/mL (S)-equol. However, overexpressed ydiT did not growth faster than the pRSF-Duet BL21 (DE3) when 100 µg/mL (S)-equol was supplemented (Supplementary Figure 8). Together, these results indicated that ydiS gene was responsible for the equol resistance in the BL21 (ydiS) strain.

# Comparison of the Equol-Producing Activity of BL21 (DE3) and BL21 (ydiS)

No equol production was detected under shaking at 200 rpm (data not shown) and this is consistent with previous reports (Lee et al., 2016). Static culture conditions were utilized to detect the equol production in this study. In order to verify that the mutant strain BL21 (ydiS) could grow faster and provide higher yields of (S)-equol than BL21 (DE3) during fermentation, the same plasmids pETDuet-1-L-ddrc-dznr and pCDFDuet-1-L-dhdr-thdr were transformed into BL21 (DE3) [DDDT-BL21 (DE3)] and BL21 (ydiS) [DDDT-BL21 (ydiS)], respectively. In 50 mL flasks, 20 mL of LB, carbenicillin (final concentration 50 µg/mL), streptomycin (final concentration 50 µg/mL) and two different concentrations of daidzein (5 µg/mL or 50 µg/mL) were added. Before adding IPTG,

fmicb-09-01182 June 1, 2018 Time: 11:17 # 5

the bacterial density was adjusted to OD<sup>600</sup> = 0.6. After being induced for 48 and 72 h under both anaerobic and aerobic conditions, the bacterial density, daidzein, and equol concentration were measured. The DDDT-BL21 (ydiS) was not much better than DDDT-BL21 when 5 µg/mL daidzein was added as the substrate. In contrast, the growth rate, equol yield, and daidzein utilization ratio of DDDT-BL21 (ydiS) was greater than DDDT-BL21 (DE3) when 50 µg/mL daidzein was added as substrate both under anaerobic and aerobic conditions (**Figures 4A–C**; Supplementary Figures 9A–C). The daidzein utilization ratio of DDDT-BL21 (ydiS) reached 90% under aerobic condition after IPTG induction, while the ratio for DDDT-BL21 was only about 27% (**Figure 4C**). In addition, the higher yield of (S)-equol for BL21 (ydiS) was not only due to faster growth rates, but also because BL21 (ydiS) produced more (S)-equol than BL21 (DE3) (**Figure 4D** and Supplementary Figure 9D).

#### Production of (S)-Equol Using Daidzin as the Fermentation Substrate

As daidzin is easily collected from soybean meal, using daidzin as fermenting substrate is extremely convenient. As such, a glycoside hydrolysis gene bgl 1269 (Gang et al., 2012) was cloned into the compatible vector pRSFDuet-1. The pRSFDuet-1-bgl plasmid was then transformed into BL21 (ydiS) [PRSF-bgl- BL21 (ydiS)] to verify the gene's function. After a 72 h induction, daidzin or soybean meal was converted into daidzein under aerobic conditions, with the metabolites from the engineered E. coli had a similar HPLC peaks to the reference standard for daidzein (Supplementary Figure 10). Interestingly, after transforming pETDuet-1-L-ddrcdznr, pCDFDuet-1-L-dhdr-thdr and pRSFDuet-1-bgl into BL21 (ydiS), the transformed bacteria could not convert daidzin to (S)-equol (data not shown). In order to overcome this hurdle, a two-step method was utilized to transfer daidzin to (S)-equol. First, PRSF-bgl-BL21(ydiS) was used to convert daidzin to daidzein, then the fermentation supernatant was used as substrate for DDDT-BL21 (ydiS) to produce (S)-equol. After first step fermentation, ∼0.17 mmol/L daidzein was produced from 50 µg/mL daidzin (Supplementary Figure 11A). At this point, if the water prepared for LB media was fully replaced with the fermentation supernatant, (S)-equol still was not detected after fermentation (data not shown). However, if 10% of the water was replaced with fermentation supernatant when preparing LB media, ∼0.017 mmol/L (S) equol could be detected after the two-step fermentation under both aerobic and anaerobic conditions (Supplementary Figure 11B).

FIGURE 4 | Comparison (S)-equol production between DDDT-BL21 (ydiS) and DDDT-BL21 (DE3) under aerobic conditions. (A) Change in bacterial density; (B) Comparison of (S)-equol production of DDDT-BL21 (DE3) and DDDT-BL21 (ydiS) measured by HPLC; (C) Comparison of the daidzein utilization ratio of DDDT-BL21 (DE3) and DDDT-BL21 (ydiS); (D) Comparison of (S)-equol production per bacterium for DDDT-BL21 (DE3) and DDDT-BL21 (ydiS). At each timepoint, two duplications were measured for each sample, means and standard deviations (SD) were calculated. The Student's t-test was employed in this study, and P < 0.05 was considered to be statistically significant.

# DISCUSSION

fmicb-09-01182 June 1, 2018 Time: 11:17 # 7

Recently, the gut microbiota has become an intensely researched topic (Garrett, 2017). The microbes in the gut have been recognized as important for proper digestive functions, allowing for a variety of dietary components to be metabolized (Koppel et al., 2017), development of the host immune system (Pamer, 2017), as well as for their impacts on some diseases and infections (Boulange et al., 2016). Dietary components, especially polyphenols have been extensively used as functional food components. Previous research has demonstrated that gut microbiota contribute to polyphenol metabolism and affect its bioavailability (Stevens and Maier, 2016); (S)-equol and soy isoflavones are typical examples. (S)-equol is the metabolite transferred from soybean meal by gut microbes (Setchell and Clerici, 2010), and the equol hypothesis infers that the main functions of soy isoflavones are due the metabolite product (S)-equol (Setchell et al., 2002). Furthermore, the safety of (S)-equol has been tested (Liu et al., 2016), indicating that further study of the function and molecular mechanism of (S)-equol is important. Heemstra et al. (2006) have developed chemical methods to synthesize (S)-equol in vitro, but natural (S)-equol obtained by microbial fermentation is more attractive, especially as it is produced in vivo. Considering the vital function of (S)-equol, its production could be important for broad applications. In this study, we constructed an engineered E. coli mutant BL21 (ydiS) that could convert higher concentration of daidzein to (S)-equol under aerobic conditions (**Figure 4**). As daidzin is the main form that is found in soybean meal, (S)-equol production from daidzin was attempted by coexpressing the genes bgl, L-ddrc, L-dznr, L-dhdr, and L-thdr, ultimately producing (S)-equol using only a two-step method (Supplementary Figure 11). Coexpression of all five genes in a single system was not sufficient to convert daidzin to (S)-equol (data not shown). This phenomenon could have several explanations: (1) glucose generated by bgl conversion may inhibit the enzymatic activity. Not only various microbial β-glucosidases reported previously are strongly inhibited by glucose [11–13], intracellular alpha-L-rhamnosidase activity from Pseudoalteromonas sp. also affected by the monosaccharides concentration [8]. In addition, during the two-step fermentation equol could be detected 10% but not when it was fully replaced with fermentation supernatant further supporting this hypothesis. (2) host cells were too old to overexpress other genes after converting daidzin to daidzein. (3) the DDRC, DZNR, DHDR and THDR enzymes were deactivitated during the conversion of daidzin to daidzein. Regardless, the fermentation parameters and process need further adjustment.

Polyphenols have inhibitory activity on bacterial growth (Vázquez et al., 2017), which presents challenges when utilizing high yield fermentation to obtain polyphenol products (Chouhan et al., 2017). Although many antibiotic resistance genes have been identified (Liu and Pop, 2009), few studies have investigated polyphenol resistance. In this study, an (S)-equol resistant mutant was generated through a transposon mutagenesis screen. Sequencing and overexpression results revealed that ydiS, a putative oxidoreductase gene, was responsible for (S)-equol resistance (**Figure 3**). Although Bayer et al. reported that complex I NADH oxidoreductase gene (snoD) in Staphylococcus aureus affected the susceptibility of thrombin-induced platelet microbicidal protein 1 (Bayer et al., 2006), the equol resistance mechanism of the putative oxidoreductase ydiS gene requires further investigation. The inhibitory effects of (S)-equol on bacterial growth may be due to its antioxidant function, as the potential oxidoreductase ydiS gene product may be counteract redox active of (S)-equol, thereby granting equol resistance. Schrettl et al. (2010) have reported similar phenomenon, which they hypothesize that the primary mechanism of gliotoxin inhibits Aspergillus fumigatus growth may be via antioxidant activity. Gliotoxin exposure up-regulates several antioxidant-related proteins and elevates superoxide dismutase activity. Moreover, reactive oxygen species production also increases after exposure to gliotoxin. However, glutathione (GSH) levels were significantly elevated in Aspergillus nidulans 1gliT compared to wild-type (Carberry et al., 2012). The gliT gene encoded a gliotoxin oxidoreductase exhibits a gliotoxin reductase activity, and overexpression of GliT confers protection against exogenous gliotoxin in A. nidulans and Saccharomyces cerevisiae (Schrettl et al., 2010).

In summary, a putative oxidoreductase gene ydiS was identified to be responsible for (S)-equol resistance. As a result, an engineered equol-producing bacterial strain was constructed using an (S)-equol resistant mutant [E. coli BL21 (ydiS)] to coexpress the equol-synthesis genes. A two-step method was constructed to convert diadzin to (S)-equol under aerobic conditions, providing a new method for (S)-equol fermentation and production. In addition, the method used in this study may be useful for screening resistant host cells as an alternative method for production of anti-bacterial components, such as antibiotics and antibacterial peptides. Recently, herbal medicinal remedies have been gaining increased attention, often being combined with probiotics for therapeutic care. However, the inhibitory effects of herbs on probiotics may prevent their application; therefore, screening for polyphenol resistance genes and probiotics engineering could be beneficial for their combined use.

#### AUTHOR CONTRIBUTIONS

YY and XW conceived and designed the experiments. YY, HL, SM, and HC performed the experiments. YY, HL, LZ, and WL analyzed the data. YY, HL, and XW wrote the paper.

# FUNDING

This study was funded by the National High Technology Research and Development Program of China (863, No. 2015AA020701), the National Nature Science Foundation (NSFC, Nos. 31100097 and 21606079), and the State Key Laboratory Breeding Base for Zhejiang Sustainable Pest and Disease Control (No. 2010DS700124-ZZ1604).

#### ACKNOWLEDGMENTS

fmicb-09-01182 June 1, 2018 Time: 11:17 # 8

We thank Xiaodan Wu from the Analysis Center of Agrobiology and Environmental Sciences, Institute of Agrobiology and Environmental Sciences, Zhejiang University for LC-MS analysis.

#### REFERENCES


We would like to thank LetPub (www.LetPub.com) for providing linguistic assistance during the preparation of this manuscript.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.01182/full#supplementary-material


**Conflict of Interest Statement:** 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 © 2018 Li, Mao, Chen, Zhu, Liu, Wang and Yin. 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 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-09-01182 June 1, 2018 Time: 11:17 # 9

# Genome Sequencing of Streptomyces atratus SCSIOZH16 and Activation Production of Nocardamine via Metabolic Engineering

#### Yan Li1,2, Chunyan Zhang1,2, Chengxiong Liu<sup>3</sup> , Jianhua Ju1,2 \* and Junying Ma1,2 \*

<sup>1</sup> CAS Key Laboratory of Tropical Marine Bio-resources and Ecology, Guangdong Key Laboratory of Marine Materia Medica, RNAM Center for Marine Microbiology, South China Sea Institute of Oceanology, Chinese Academy of Sciences, Guangzhou, China, <sup>2</sup> College of Life Sciences, University of Chinese Academy of Sciences, Beijing, China, <sup>3</sup> Hubei Key Laboratory of Natural Products Research and Development, College of Biological and Pharmaceutical Sciences, China Three Gorges University, Yichang, China

#### Edited by:

Dipesh Dhakal, Sun Moon University, South Korea

#### Reviewed by:

Yinhua Lu, Shanghai Institutes for Biological Sciences (CAS), China Jae Kyung Sohng, Sun Moon University, South Korea

#### \*Correspondence:

Jianhua Ju jju@scsio.ac.cn Junying Ma majunying@scsio.ac.cn

#### Specialty section:

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

> Received: 03 April 2018 Accepted: 24 May 2018 Published: 13 June 2018

#### Citation:

Li Y, Zhang C, Liu C, Ju J and Ma J (2018) Genome Sequencing of Streptomyces atratus SCSIOZH16 and Activation Production of Nocardamine via Metabolic Engineering. Front. Microbiol. 9:1269. doi: 10.3389/fmicb.2018.01269 The Actinomycetes are metabolically flexible microorganisms capable of producing a wide range of interesting compounds, including but by no means limited to, siderophores which have high affinity for ferric iron. In this study, we report the complete genome sequence of marine-derived Streptomyces atratus ZH16 and the activation of an embedded siderophore gene cluster via the application of metabolic engineering methods. The S. atratus ZH16 genome reveals that this strain has the potential to produce 26 categories of natural products (NPs) barring the ilamycins. Our activation studies revealed S. atratus SCSIO ZH16 to be a promising source of the production of nocardamine-type (desferrioxamine) compounds which are important in treating acute iron intoxication and performing ecological remediation. We conclude that metabolic engineering provides a highly effective strategy by which to discover drug-like compounds and new NPs in the genomic era.

Keywords: Streptomyces atratus ZH16, in-frame deletion, metabolic engineering, siderophore, nocardamine

#### INTRODUCTION

Microbially produced natural products (NPs) are one of the most important classes of compounds known to mankind having a vast assortment of applications in medical and agricultural sectors (Bérdy, 2005). It has been estimated that most major classes of antibiotics and over 70% of anti-cancer small molecule therapeutics are microbial NPs, their derivatives, or related congeners/analogs (Newman and Cragg, 2016). With the recently noted rise in resistance to antibiotics and cancer chemotherapeutics, it has become increasingly obvious that novel bioactive NPs are urgently needed to ensure the success of new drug discovery and development initiatives. Genomics and metabolomics have played central roles in ensuring that these needs get met.

Over the last two decades significantly improved technologies for genome sequencing have made it much easier to sequence full microbial genomes. Additionally, studies of microbial genomes have made clear that many microorganisms have far greater potential to produce specialized metabolites than previously thought (Rutledge and Challis, 2015). It has been estimated

that the genomes of species from Streptomyces family members, the largest bacterial genus house on average, about 20 biosynthetic gene clusters (BGCs) coding for NPs (Nett et al., 2009). However, over 80% of these gene clusters are typically orphaned under normal laboratory culturing conditions (Nett et al., 2009; Baltz, 2017). Consequently, metabolic engineering and genome mining methods have increasingly been applied to discover secondary metabolites whose corresponding BGCs are normally silent; such BGCs are also sometimes considered "orphan BGCs" to convey the absence of a correlatable NP/s (Zerikly and Challis, 2009; O'Connor, 2015; Baltz, 2016).

Microbial siderophores biosynthesis can generally be classified into two main pathways: non-ribosomal-peptide synthetase (NRPS)-dependent and siderophore synthetase super-family (Barry and Challis, 2009); both pathways are exploited by a range of genera belonging to the Actinomycetes (Wang et al., 2014). Many siderophores are NRPS dependent family NPs, such as griseobactin (Patzer and Braun, 2010), coelichelin (Challis and Ravel, 2000), oxachelin (Sontag et al., 2006), as well as tsukubachelin (Kodani et al., 2011, 2013), peucechelin (Kodani et al., 2015), and chlorocatechelins (Kishimoto et al., 2014). Nocardamine is the representative one of the siderophore synthetase super-family (Stoll et al., 1951; Hossain et al., 1983; Ueki et al., 2009). Nocardamines (also called desferrioxamines), composed of alternating dicarboxylic acid and diamine units, originally isolated as antibacterial metabolites from a Nocardia strain (Stoll et al., 1951). The BGC responsible for desferrioxamines G<sup>1</sup> and E in Streptomyces coelicolor A3(2) was investigated by Barona-Gómez et al. (2004). Among their findings was that the des operon contained a subset of four genes coding for the production of various desferrioxamines (Barona-Gómez et al., 2004); their production was found to be regulated by both iron concentrations and by an irondependent regulatory protein-IdeR (Günter et al., 1993; Ueki et al., 2009).

Streptomyces atratus SCSIO ZH16 is a deep sea-derived Streptomyces that predominantly produces ilamycins under standard laboratory conditions; the biosynthesis of ilamycins has been elucidated in our previous studies (Ma et al., 2017). By applying a combination of Frameplot 3.0 beta (Ishikawa and Hotta, 1999) and AntiSMASH 3.0 (Weber et al., 2013), two online software systems, we were able to predict that up to 26 BGCs are housed within the genome of S. atratus SCSIO ZH16; we envisioned that the majority of these are orphan clusters. Accordingly, we applied metabolic engineering methods to activate these putative orphan/silent clusters en route to the production of new compounds with potential applications in drug discovery and bioremediation. Here we report: (i) the complete genome sequence of S. atratus SCSIO ZH16 as well as a comparative analysis to get further insights into genetic elements involved in biosynthesis of NPs, (ii) the construction of in-frame deletion mutant S. atratus SCSIO ZH16S and S. atratus SCSIO ZH16NS, and (iii) the identification and structural characterization of nocardamine. Our study highlights the enabling power of metabolic engineering to generate new NPs encoded by orphan gene clusters and also validates the engineered S. atratus ZH16NS as a promising nocardaminebased siderophore producer.

# EXPERIMENTAL SECTION

# General Experimental Section

All bacteria, plasmids and primers used in this work are listed in Supplementary Tables S1, S2. The antibiotics and reagents were purchased from Sangon Biotech Co., Ltd. (Shanghai, China), the PCR polymerase and related reagents were purchased from Takara Biotechnology Co., Ltd. (Dalian, China), gel recycle and PCR recycle kits were purchased from Omega Bio-tek Inc. (Norcross, GA, United States). All solvents were analytical or chromatographic grade and purchased from Guangzhou Chemical Reagent Factory (Guangzhou, China) and Thermo Fisher Scientific Inc. (Waltham, MA, United States).

Column chromatography (CC) was carried out using normal phase silica gel (100–200 mesh, Jiangyou, China) and reverse phase C18 silica gel (40–63 µm, Merck, Germany). Medium-pressure liquid chromatography was performed with a CHEETAH 100 automatic flash chromatography system (Bonna-Agela, China) with an ODS-A flash column (S-50 µm, 12 nm; 100 mm × 20 mm, YMC, Japan). Semipreparative HPLC was carried out using an Agilent 1260 liquid chromatography system with diode array detector (DAD) (Agilent, United States) and YMC-Pack ODS-A column (250 mm × 20 mm, 5 mm, YMC, Japan). NMR spectra were performed with an Advance 700 MHz spectrometer (Bruker, Germany). High-resolution mass spectral data was obtained from a MaXis quadrupole-time-of-flight mass spectrometer (Bruker, Germany). All the gene amplification was performed with Eppendorf mastercycler pro PCR equipment (Eppendorf, Germany).

# Genome Sequencing Bioinformatic Analysis

The collection, identification, and genome sequencing of S. atratus SCSIO ZH16 has been previously described (Ma et al., 2017). BGCs and their related ORFs were analyzed by antiSMASH 3.0<sup>1</sup> and Frame Plot 3.0 beta<sup>2</sup> , respectively. Furthermore, functional gene annotations and sequence alignments were carried out using Basic Local Alignment Search Tool<sup>3</sup> .

# In-Frame Deletion of Ilamycin Genes

To obtain NPs encoded by other orphan/silent gene clusters using metabolic engineering methods in S. atratus SCSIO ZH16, genetic engineering mutants with clean metabolic background were constructed. IlaS has been identified as a large nonribosomal peptide synthetase responsible for the incorporation of

<sup>1</sup>https://antismash.secondarymetabolites.org/#!/start

<sup>2</sup>http://www0.nih.go.jp/∼jun/cgi-bin/frameplot.pl

<sup>3</sup>https://blast.ncbi.nlm.nih.gov/Blast.cgi

amino acid building blocks to form the full-length heptapeptide. IlaN encoding a cytochrome P450 monooxygenase and IlaO encoding a prenyltransferase were involved in the biosynthesis of L-3-nitrotyrosine and N-(1,1-dimethyl-1-allyl)-tryptophan building blocks, respectively. In-frame gene deletions were achieved by following the REDIRECT protocol (Gust et al., 2003). The S. atratus SCSIO ZH16 genomic cosmid library was constructed as previously reported (Ma et al., 2017). The apramycin resistance gene oriT/aac(3)IV fragment was obtained by using specific primers that contain additional SpeI restriction sites, and used to replace the target genes in the cosmids 2- 10E or 4-07H (Supplementary Table S2). Restriction digests of mutant cosmids with SpeI and subsequent relegation predictably abolished the apramycin resistance gene oriT/aac(3)IV fragment. The second round of PCR-targeting was performed to replace the kanamycin resistance gene on SuperCos I with another apramycin resistance gene oriT/aac(3)IV fragment obtained by a primer pair ARK (Zhang et al., 2013). The constructed mutant cosmids were introduced into non-methylating Escherichia coli ET12567/pUZ8002 and then transferred into S. atratus SCSIO ZH16 by conjugation. Because the strain was sensitive to apramycin, exconjugants were grown on solid apramycin containing ISP-4 medium to select for the chromosomal integration of the inactivation constructs. To ensure loss of the target gene from the chromosome, exconjugants were replicaplated once onto antibiotic-free ISP-2 plates. Single colonies were again replica-plated onto apramycin-containing ISP-2 plates and antibiotic-free ISP-2 plates. Apramycin-sensitive clones were evaluated by PCR to ensure proper generation of the desired mutant clones. Two in-frame deletion mutants, S. atratus SCSIO ZH16S and S. atratus SCSIO ZH16NS, were obtained using this method.

#### Fermentation and Isolation

The S. atratus SCSIO Zh16 wild-type, S. atratus SCSIO ZH16S, and S. atratus SCSIO Zh16NS mutant strains were cultured in 250 mL flasks containing 50 mL Am2ab liquid media consisting of 0.5% soluble starch, 2% glucose, 0.2% yeast extract, 0.2% peptone, 0.5% soybean meal, 0.05% MgSO4.7H2O, 0.05% KH2PO4, 0.4% NaCl, 0.2% CaCO3, and 3% crude sea salt. Then they were incubated at 28◦C on a rotary shaker at 200 rpm for 7 days. Fermentation broths were extracted by 80 mL butanone and the solvent was removed under reduced pressure to give crude extracts that were dissolved in 0.6 mL methanol. Each extract was ultimately subjected to HPLC analysis. The HPLC analysis was carried out with a reversed phase column SB-C18, 5 µM, 4.6 × 150 mm (Aglient, United States) with UV detection at 210 and 285 nm under the following program: solvent system (solvent A, water supplemented with 0.1% trifluoroacetic acid; solvent B, acetonitrile supplemented with 0.1% trifluoroacetic acid); 2–98% solvent B (linear gradient, 0–30 min), 98% solvent B (30–35 min), 98–2% solvent B (35.0–35.1 min), 2% solvent B (35.1–40 min); flow rate was set at 1 ml/min.

The large-scale fermentation of S. atratus SCSIO ZH16NS was performed using a two-stage fermentation process. The spores grown on ISP-2 supplemented with 3.0% crude sea salt plates were incubated in a total of 63 flasks (250 mL

volume) containing 50 mL Am2ab liquid media at 28◦C on a rotary shaker at 200 rpm for 60 h. The 25 mL seed cultures were then transferred into 1 L flasks containing 200 mL Am2ab medium with supplemental 3.0% XAD-16 resin. The resin-containing mixtures were then cultured at 28◦C on a rotary shaker at 200 rpm for 8 days. Aliquots (10 mL) were removed from each flask on a daily basis and analyzed by HPLC. After 8 days of growth fermentation broths (28 L) were centrifuged to ensure separation of supernatant, mycelium and resin which were extracted sequentially with twofolds volume butanone (3x), acetone (3x), and EtOH (3x), respectively.

The following HPLC analyses, all extracts (butanone, acetone, and EtOH) were combined, solvent removed in vacuo and the remaining extracts subjected to silica gel CC using gradient elution with a CHCl3-MeOH mixture (100:0, 98:2-1, 98:2-2, 96:4- 1, 96:4-2, 94:6-1, 94:6-1, 92:8, 9:1,85:15, 8:2,1:1, 0:100) to give 13 fractions (Fr.A1–Fr.A13). Fr.A7 was purified by MPLC with ODS column, which was eluted from 0 to 60% solvent B (A: H2O, B: CH3CN) over the course of 60 min to obtain 14 fractions (Fr.J1–Fr.J14). Fr.J2-Fr.J3 were combined and further purified by Sephadex LH-20 chromatography eluted by MeOH and semipreparative HPLC eluted from 5 to 20% solvent B over 10 min and 20% solvent B over 5 min at a flow rate of 2.5 mL/min using a detection wavelength of 210 nm.

#### Structural Elucidation

Nocardamine: white acicular crystal, <sup>1</sup>H and <sup>13</sup>C NMR data (**Table 1**). HRESIMS: m/z 601.3556 ([M+H]<sup>+</sup> calcd 601.3553), m/z 623.3402 ([M+Na]<sup>+</sup> calcd 623.3384).

# RESULTS AND DISCUSSION

#### Genome Sequencing and Annotation of Streptomyces atratus SCSIO ZH16

Genome sequence information is playing a progressively more important role in NPs discovery as well as studies to elucidate NP biogenesis. Many antibiotic producing strains have been sequenced and examinations of their genomic data have revealed

TABLE 1 | The <sup>1</sup>H NMR and <sup>13</sup>C NMR data of nocardamine [In CDCl<sup>3</sup> and MeOD (1:1), 700 MHz <sup>1</sup>H NMR and 175 MHz13C NMR in δ ppm].


their full biosynthetic potentials which are often far exceeded initial expectations. To better understand the full secondary metabolic potential of S. atratus SCSIO ZH16, its genome was sequenced using a combination of 2nd-generation 454 and Illumina HiSeq 4000 sequencing technologies and 3rd-generation PacBio sequencing technology at Shanghai Biozeron Co., Ltd. The assembled genome, along with elucidation of its GC content enabled its classification as a Streptomyces. The S. atratus SCSIO ZH16 genome is 9,641,288 bp long and consists of a linear chromosome with an average GC content of 69.5% (**Figure 1**). The genome contains 9245 coding sequences, 18 rRNA genes and 69 tRNA genes for transfer of all 20 amino acids (Supplementary Table S3). The genome sequence of S. atratus SCSIO ZH16 has been deposited in the Genbank database with the accession number of CP027306.

To elucidate the gene clusters encoded in its genome, the assembled genome sequence was subjected to analysis for secondary metabolite BGCs using online antiSMASH software (see footnote text 1) (Weber et al., 2013) and Frameplot 3.0 beta (Ishikawa and Hotta, 1999). These analyses revealed 26 gene clusters within the S. atratus SCSIO ZH16 genome, including six NRPS, four PKS (Type I, Type II, and Type III), four hybrid PKS-NRPS, four terpene, three bacteriocin, two siderophore and three other categories BGCs (Supplementary Table S4) indicating the great potential of the strain to produce an array of secondary metabolites. Significantly, only one gene cluster responsible for ilamycin biosynthesis has been characterized and elucidated from the wild-type strain (Ma et al., 2017). This strain will serve as a target for further genome mining of secondary metabolite BGCs.

# Construction of Ila Gene Cluster In-Grame Deletion Mutants and the Discovery of a New Peak in the Genetic Engineered Mutant

Secondary metabolites may be overlooked due to low production levels, a large metabolic background or unpropitious culture conditions (Scherlach and Hertweck, 2009); indeed such considerations have inspired the term "orphan" instead of "silent" BGCs – it is not that a BGC is completely inactive, rather its product simply has not been identified. In our previous report, we have identified IlaS as a large non-ribosomal peptide synthetase responsible for the incorporation of amino

acid building blocks to form the full-length heptapeptide of ilamycins (Ma et al., 2017). To enable the production of potential secondary metabolites from S. atratus SCSIO ZH16, an inframe deletion mutant S. atratus SCSIO ZH16S (**Figure 2**), devoid of the 1.5 kb ilaS gene was constructed to abolish production of the ilamycins, a predominant product of wild-type S. atratus SCSIO ZH16 (**Figure 3**, i). Subsequent HPLC analyses revealed that, indeed, ilamycins biosynthesis was abolished upon deletion of ilaS (**Figure 3**, ii). However, the ilamycins precursor N-(1,1-dimethyl-1-allyl)-tryptophan generated by IlaO from tryptophan was identified in the fermentation extracts of the 1ilaS mutant strain. The generation of a clean background strain for genome mining was not possible. We have demonstrated that IlaN (a cytochrome P450 monooxygenase) is involved in the biosynthesis of L-3-nitrotyrosine building block, and that IlaO (a prenyltransferase) is responsible for

the biosynthesis of N-(1,1-dimethyl-1-allyl)-tryptophan building block (Ma et al., 2017). Therefore, we sought to further reduce the metabolic background of a genetically engineered S. atratus SCSIO ZH16 by carrying out further gene deletions. The in-frame deletion mutant S. atratus SCSIO ZH16NS was generated via deletion of an 8-kb fragment spanning from ilaN to ilaS (**Figure 2**). This mutant strain failed to produce ilamycins as well as the previously noted tryptophan-derived ilamycin precursor (**Figure 3**, iii) and consequently served as an excellent starting strain for the genome mining of S. atratus SCSIO ZH16.

In order to analyze the metabolomic differences between genetic engineered mutants and their wild type predecessor S. atratus SCSIO ZH16, HPLC chromatograms of extracts from the fermentations of the three strains were carefully compared. Notably, a new peak was identified uniquely in the S. atratus SCSIO ZH16S and S. atratus SCSIO ZH16NS mutants. HRESIMS analysis of the signal generating species revealed a low molecular weight NP with [M+H]<sup>+</sup> = 601.3556, [M+Na]<sup>+</sup> of 623.3402 (Supplementary Figure S1); on the basis of these data this mutant specific compound was assigned a molecular formula of C27H48N6O9.

### Fermentation, Isolation, and Structural Elucidation of Nocardamine From S. atratus SCSIO ZH16NS

To isolate and elucidate the structure of the newly generated NP, large-scale fermentation of S. atratus SCSIO ZH16NS was carried out by using 28 L of Am2ab liquid media with a two-step fermentation process as previously reported (Ma et al., 2017). A compound peak retention time of 13.2 min was apparent when using detection at 210 nm; this signal correlated perfectly to the species originally identified

on analytical scale fermentations/analysis with S. atratus SCSIO ZH16NS. After several rounds of silica column isolation and medium pressure preparative HPLC, 13.3 mg of analytically pure compound was obtained. For elucidating the structure of the purified compound, the NMR analysis was carried out. The NMR data revealed that <sup>1</sup>H and <sup>13</sup>C NMR data (**Table 1** and Supplementary Figures S2, S3) which were consistent with the nocardamine spectral data reported in the literature (Yuan et al., 2010). So the compound was identified as nocardamine (also called desferrioxamine E) (**Figure 4**).

#### Bioinformatic Analysis of the Nocardamine Gene Cluster and the Proposed Biosynthetic Pathway of Nocardamine in S. atratus SCSIO ZH16

Further to identify the gene cluster likely responsible for nocardamine biosynthesis, all clusters encoded in the genome of S. atratus SCSIO ZH16 were surveyed. Of the 26 unidentified gene clusters, there are two siderophore gene clusters, while cluster 11 showed a high degree of similarity to the previously published desferrioxamine B BGC (Barona-Gómez et al., 2004; Ueki et al., 2009). Accordingly, we assigned this cluster as the

nocardamine (desferrioxamine E) BGC. The detailed annotation of the nocardamine gene cluster in S. atratus SCSIO ZH16 was postulated and found to contain a noc operon with a subset of four genes, enough to code for the biosynthesis of nocardamine (desferrioxamine E). More rigorous BLAST analysis showed that the four genes nocABCD in S. atratus SCSIO ZH16 encode for pyridoxal decarboxylase (nocA gene), putative monooxygenase (nocB gene), N-acetyltransferase (nocC gene), and IucA/IucC family siderophore biosynthesis protein (nocD gene) having 83, 77, 60, 73% identity with desA, desB, desC, desD in S. coelicolor M145, respectively (**Table 2**). The BGC for desferrioxamine has been reported or characterized in another two Streptomyces strains, Streptomyces avermitilis K139 (Ueki et al., 2009), and S. pristinaespiralis HCCB10218 (Li et al., 2015). Gene functions and organization were identical in each desferrioxamine BGC. The location and organization of the nocardamine gene cluster in the chromosome of S. atratus ZH16 were shown in **Figures 5A,B**.

Based on the proposed functions of the genes in noc gene cluster, the biosynthetic pathway of nocardamine was proposed as follows: firstly, L-lysine was decarboxylated to yield cadaverine (**2**) by L-2,4-diaminobutyrate decarboxylase encoded by nocA. Secondly, cadaverine is hydroxylated at an amino group by monooxygenase encoded by nocB, to form N-hydroxycadaverine (**3**). Then, N-hydroxy-cadaverine was condensed with a succinyl-CoA to generate the N-hydroxy-N-succinylcadaverine (HSC, **4**). Finally, three HSC units were catalyzed to form nocardamine by LucA/Iuc family siderophore biosynthesis protein encoded by nocD. The proposed biosynthetic pathway of nocardamine in S. atratus SCSIO ZH16 was shown in **Figure 5D**.

#### The Possible Regulation of Nocardamine Production in S. atratus SCSIO ZH16

The production of nocardamine is carefully controlled by a regulatory protein termed iron-dependent regulatory protein (IdeR) (Günter et al., 1993). To investigate whether the production of nocardamine is also regulated by an IdeR ortholog, three ideR genes from Streptomyces davawensis strain JCM 4913, Mycobacterium tuberculosis H37Rv, and S. coelicolor A3(2) were used as probes for BLAST searching within the genome sequence of S. atratus SCSIO ZH16. BLAST results indicated


ID/SI: Identity/Similarity; <sup>a</sup>Streptomyces coelicolor M145; <sup>b</sup>Streptomyces griseus subsp. griseus NBRC 13350.

that a ideR ortholog located at 5.1 Mbp of S. atratus SCSIO ZH16 chromosome genome has the highest identity (87%) and coverage (88%) with that from the S. davawensis strain JCM 4913, and has 86 and 80% identity with that from the S. coelicolor A3(2) and M. tuberculosis H37Rv, respectively. These results indicated that the production of nocardamine in S. atratus SCSIO ZH16 is likely regulated by ideR homologues. In order to further determine the binding motif of a putative IdeR regulator, the upstream sequence of the operon nocABCD was analyzed by comparison the binding motif to those previously reported in the literature (Günter et al., 1993; Ueki et al., 2009). These comparisons indicated that a 17 base pair sequence "TTAGGTTAGGCTAACCT" has the same IdeR binding motif reported in S. avermitilis K139 (Ueki et al., 2009), which is also a nocardamine producer. The regulatory function of IdeR in the production of nocardamine by S. atratus SCSIO ZH16 will be characterized in forthcoming publications. The location of the ideR gene in the chromosome of S. atratus SCSIO ZH16 was shown in **Figure 5A**.

# CONCLUSION

The deep sea-derived S. atratus SCSIO ZH16 has a linear genome chromosome whose average GC content is 69.5%. Twenty-six gene clusters housed in the genome endow the strain a potential target for genome mining of bioactive NPs. An orphan BGC (noc) coding for a siderophore has been activated via metabolic engineering; this entailed the construction and ensuing metabolite analyses of the mutant strains of S. atratus SCSIO ZH16S and S. atratus SCSIO ZH16NS. These mutants result from efforts to knock out selected portions of the wellstudied ilamycin gene cluster. The in-frame deletion mutant of S. atratus SCSIO ZH16NS with a relative clean metabolic background, provides excellent opportunity to further mining other orphan gene cluster encoding NPs. The siderophore was structurally elucidated using a combination of HRESIMS and NMR analyses and shown to be previously reported nocardamine. By virtue of its excellent metal ion complexation abilities, nocardamine can be used as an iron carrier to relieve metal toxicity. This work highlights the notion that shifting metabolic flux of an NP producing wild-type strain away from the predominant product pathway may enable the production of new metabolites that, otherwise, are simply not attainable. We posit that nocardamine production (noc activation) is enabled at the expense of ilamycin biosynthesis by virtue of engineered shifting of the S. atratus metabolic flux. Studies to further dissect this system of engineered cluster activation in the unique marine-derived microbe will be published in due course.

#### AUTHOR CONTRIBUTIONS

fmicb-09-01269 June 13, 2018 Time: 11:11 # 8

JM and JJ designed the experiments and revised the manuscript. YL analyzed the BGC sequences, performed the mass fermentation of S. atratus SCSIO ZH16NS, isolated the compound, and prepared the draft manuscript. CZ constructed two in-frame deletion mutants of S. atratus SCSIO ZH16S and S. atratus SCSIO ZH16NS and wrote part of the draft manuscript. CL analyzed the NMR data. All authors reviewed the manuscript.

#### FUNDING

This work was financially supported by grants from National Natural Science Foundation of China (31270134, U17062606,

#### REFERENCES


and 81741166) and Guangdong Natural Science Foundation (2016A030312014).

#### ACKNOWLEDGMENTS

We thank Dr. Xiao, Mr. Li, and Ms. Sun in the analytical facility center of South China Sea Institute of Oceanology for acquisition of all NMR and MS data.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.01269/full#supplementary-material


biosynthetic gene clusters. Nucleic Acids Res. 52, 1231–1234. doi: 10.1093/nar/ gkv437


**Conflict of Interest Statement:** 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 JKS and handling Editor declared their shared affiliation.

Copyright © 2018 Li, Zhang, Liu, Ju and Ma. 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 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.

# CRISPR/Cas9-Based Editing of Streptomyces for Discovery, Characterization, and Production of Natural Products

Weixin Tao, Anna Yang, Zixin Deng and Yuhui Sun\*

Key Laboratory of Combinatorial Biosynthesis and Drug Discovery (Ministry of Education), Wuhan University School of Pharmaceutical Sciences, Wuhan, China

Microbial natural products (NPs) especially of the Streptomyces genus have been regarded as an unparalleled resource for pharmaceutical drugs discovery. Moreover, recent progress in sequencing technologies and computational resources further reinforces to identify numerous NP biosynthetic gene clusters (BGCs) from the genomes of Streptomyces. However, the majority of these BGCs are silent or poorly expressed in native strains and remain to be activated and investigated, which relies heavily on efficient genome editing approaches. Accordingly, numerous strategies are developed, especially, the most recently developed, namely, clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR associated (Cas) system reveals remarkable higher accuracy and efficiency for genome editing in various model organisms including the Streptomyces. In this mini review, we highlight the application of CRISPR/Cas9 based approaches in Streptomyces, focus on the editing of BGCs either in vivo or in vitro, as well as target cloning of large-sized BGCs and heterologous expression in a genetically manipulatable host, for discovery, characterization, reengineering, and production of potential pharmaceutical drugs.

#### Edited by:

Eung-Soo Kim, Inha University, South Korea

#### Reviewed by:

Yinhua Lu, Shanghai Institutes for Biological Sciences (CAS), China Sisun Choi, Inha University, South Korea

> \*Correspondence: Yuhui Sun yhsun@whu.edu.cn

#### Specialty section:

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

> Received: 30 March 2018 Accepted: 04 July 2018 Published: 24 July 2018

#### Citation:

Tao W, Yang A, Deng Z and Sun Y (2018) CRISPR/Cas9-Based Editing of Streptomyces for Discovery, Characterization, and Production of Natural Products. Front. Microbiol. 9:1660. doi: 10.3389/fmicb.2018.01660 Keywords: natural product, Streptomyces, biosynthetic gene cluster, genome editing, CRISPR/Cas9

### INTRODUCTION

Streptomyces species are known for the most prolific antibiotic producers and have provided a large number of clinical drugs during past decades. However, discovery of natural product (NP) drugs from these talented bacteria has suffered a blow after the Golden Age of NP discovery in 1950s–1960s, that is severely influenced by high-throughput screening of synthetic libraries and the low efficiency of traditional top-down screening strategies (Li and Vederas, 2009). Recently, great advances in next-generation sequencing technologies and computational resources reacquaint microbial genomes and are regarded as a huge reservoir of untapped NP biosynthetic gene clusters (BGCs; Rutledge and Challis, 2015; Weber and Kim, 2016; Kim et al., 2017); moreover, a vast majority of uncultured microorganisms in environments provide limitless possibilities for NP drugs discovery (Banik and Brady, 2010; Katz et al., 2016). For Streptomyces, the most gifted bacteria are supposed to possess 20–50 BGCs in a single genome, that greatly exceed the identified compounds (Challis, 2014; Baltz, 2017). Nevertheless, most of BGCs are silent or poorly expressed in native hosts under conventional laboratory culture conditions. To activate these cryptic BGCs, high-efficient approaches for genome editing and BGC engineering garner widespread attention and become a rapidly advancing field for NP drugs discovery (Hsu et al., 2014; Rutledge and Challis, 2015; Choi and Lee, 2016; Jakociunas et al., 2016; Li et al., 2017a; Ren et al., 2017; Zou et al., 2018).

Compared with other model organisms, like Escherichia coli and Saccharomyces cerevisiae, Streptomyces strains show poverty in genetic manipulation and most are recalcitrant for genome editing. In Streptomyces, recombinase-mediated homologous recombination has been commonly used for genome editing; however, the related protocols are often laborious and timeconsuming (Gust et al., 2003; Komatsu et al., 2010; Fernandez-Martinez and Bibb, 2014; Li et al., 2017a). Until recently, application of clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR associated (Cas) system, especially the CRISPR/Cas9 system, has greatly facilitated high-efficiency genome editing (Jinek et al., 2012; Choi and Lee, 2016). Likewise, CRISPR/Cas9-based genome editing approaches have greatly accelerated insights into Streptomyces derived NP drugs. In this mini review, we summarize the recent developments and challenges of CRISPR/Cas9-based approaches for editing BGCs of Streptomyces; moreover, cloning and assembly of intact BGCs for heterologous expression are also emphasized.

#### CRISPR/Cas9 ADVANCES THE GENOME EDITING

CRISPR/Cas system functions as adaptive immune system in numerous bacteria and archaea, of which RNAs harboring "spacer" sequence from previously exposed bacteriophages help Cas proteins recognize and cleave the specific exogenous DNA (Barrangou et al., 2007; Grissa et al., 2007; Horvath and Barrangou, 2010). Since CRISPR/Cas system exhibits higher specificity and accuracy on sequence targeting, it has become excellent choice for precision genome editing (Jinek et al., 2012). CRISPR/Cas9, a type II CRISPR/Cas system, originally employs CRISPR RNA (crRNA) and trans-activating crRNA (tracrRNA) to form crRNA-tracrRNA duplex and then assists Cas9 nuclease to recognize and cleave target DNA harboring trinucleotide protospacer adjacent motif (PAM) and a 5' end of 20 nucleotides complementary to the spacers (Deltcheva et al., 2011; Jinek et al., 2012; Hsu et al., 2014; Nishimasu et al., 2014). System reprogramming that fuses crRNA and tracrRNA into a synthetic single guide RNA (sgRNA) greatly facilitates preparation of transcripts and significantly promotes the application of CRISPR/Cas9 system (Jinek et al., 2012; Hsu et al., 2014). Reprogrammed CRISPR/Cas9 system has since been successfully used in a variety of organisms, including S. cerevisiae (DiCarlo et al., 2013), Drosophila melanogaster (Gratz et al., 2013), Caenorhabditis elegans (Friedland et al., 2013), plants (Jiang et al., 2013), and human embryos (Baltimore et al., 2015).

## IN VIVO STRATEGIES FOR GENOME EDITING IN STREPTOMYCES

Streptomyces are of utmost importance for novel NP drugs discovery, of which the investigating process relies heavily on high-efficiency genome editing. In Streptomyces, classic genome editing commonly achieves through homologous recombination with a suicide or temperature-sensitive or self-replicative plasmid, and requires intensive and time-consuming screening process. The application of CRISPR/Cas9 system for genome editing in Streptomyces started in 2015, and since then related approaches have been tremendously developed. As shown in **Table 1**, diversified approaches are widely used to edit or refactor BGCs for NP drugs discovery and characterization.

Cobb et al. (2015) first introduced CRISPR/Cas9 system for genome editing in Streptomyces. The pCRISPomyces-2 system equips a codon-modified Cas9 nuclease driven by a strong promoter, a sgRNA expression cassette, and a 2 kb homology repair template (HRT). It first specifically generates a double-strand break (DSB) at target site by Cas9 nuclease under the guidance of sgRNA harboring a custom-designed spacer, and then repairs the resulting chromosome break by homology-dependent repair (HDR) system in the presence of HRT and introduces chromosomal deletions ranging from 20 bp to 31 kb with an efficiency ranging from 70 to 100% (**Figure 1A**). Multiplex genome editing may be achieved by equipping multiplex sgRNA cassettes and corresponding repairing templates in a pCRISPomyces system, and excision of 31 kb BGC of undecylprodigiosin (Red) in Streptomyces lividans 66 has thus successfully obtained.

Slightly afterward, three different groups successively applied CRISPR/Cas9 system in Streptomyces for diverse applications. The pKCcas9dO system by Lu group similarly revealed high editing efficiency of CRISPR/Cas9 system in Streptomyces coelicolor M145 for single gene/BGC deletion, as well as multiplex genes/BGCs deletions (Huang et al., 2015). Besides, a point mutation editing strategy that CRISPR/Cas9 cleaves chromosome DNA at specific site guided by synthetic sgRNA, and then the HDR in S. coelicolor helps repair DSB in assistance of HRT with designed point mutation (AAG of 262–264 nucleotides in rpsL was changed to GAA), has performed to convert Lys88 to Glu in rpsL (**Figure 1A**). Tong et al. (2015) have thoroughly investigated editing efficiency when repairing Cas9 generated site-specific DSBs by non-homologous end joining (NHEJ) system in S. coelicolor A3(2). It revealed an incomplete NHEJ system in S. coelicolor that lacking a core component LigD, and led to randomly sized deletions around target site. Reconstitution of this defective NHEJ system by complementing Streptomyces carneus derived ScaligD has increased editing efficiency up to 77% and qualified the mutations to 1– 3 bp deletion/insertion/substitution in most cases (**Figure 1B**). Moreover, high precision genome editing efficiency near 100% achieved when supplying the HRT. In the study, CRISPRi using a catalytically inactive Cas9 nuclease (dCas9) has also been developed, to target promoter region or open reading frame of actIORF1 for reversible regulation of actinorhodin production in S. colicolor. In the same year, Sun group developed an extraordinary CRISPR/Cas9-CodA(sm) combined system, using CodA(sm), the D314A mutant of cytosine deaminase to convert 5-fluorocytosine to toxic 5-fluorouracil, as an efficient counterselection approach to select for progenies lost recombinant plasmid, which greatly accelerates screening process (Zeng et al., 2015). Besides, a most significant feature that differs from above three systems is application of a segregationally unstable sti<sup>−</sup> pIJ101-derived shuttle vector. The behavior of self-replication


**39**

actinomycetes.

fmicb-09-01660 July 20, 2018 Time: 14:58 # 3

with high copy number of about 50 per chromosome of this deliver vector can produce large amounts of single-strand plasmid DNA and plenty of template DNAs, resulting in dramatically high efficiency of double cross-over recombination and frequency of target mutant (Zeng et al., 2015). All the above reports facilitate rapid progress for genome editing in Streptomyces, since CRISPR/Cas9 helps select against wildtype sequence in the presence of HRTs. CRISPR/Cas9 system also enables activation of cryptic BGCs in Streptomyces. Zhao group utilized CRISPR/Cas9-mediated knock-in strategy for efficient and precise insertion of constitutive promoters upstream of main biosynthetic operons or pathway-specific activators, and triggered production of novel NPs of different classes in multiple Streptomyces species (**Figure 1A**; Zhang et al., 2017).

To date, CRISPR/Cas9 system has been applied for genome editing in Streptomyces for 3 years, the high specificity and efficiency made it the most attractive technology in that field. Its application has now extended to many non-model Streptomyces strains, like Streptomyces formicae from the African fungus-growing plant-ant Tetraponera penzigi (Qin et al., 2017), Streptomyces rimosus with distinctive chromosome terminal and core regions (Jia et al., 2017), Streptomyces sp. SD85 from tropical mangrove sediments (Low et al., 2018), and some rare actinomycetes like Actinoplanes sp.SE50/110 (Wolf et al., 2016) and Micromonospora chersina (Cohen and Townsend, 2018). However, in vivo application of this fascinating technology in Streptomyces is confined to the strains are genetically tractable, missing out on a vast amount of precious BGC resources from genetically intractable strains or yet uncultured strains. In that

case, acquiring and refactoring of intact BGCs for heterologous expression in a genetically tractable surrogate host could be alternatively considered.

# CRISPR/Cas9-MEDIATED BGCS CLONING AND REFACTORING FOR HETEROLOGOUS EXPRESSION

#### Cloning of Large-Sized BGCs

A variety of cryptic BGC awakening approaches, like pathwayspecific/global regulator manipulation, promoter refactoring, and ribosome engineering, have been used for NPs discovery in Streptomyces. However, most require genetic manipulation of native strains thus are constrained in genetically intractable strains or BGCs from environmental DNA (eDNA; Rutledge and Challis, 2015; Weber et al., 2015; Zhang et al., 2016). Strategies for heterologous expression of BGCs in a genetically manipulatable host can perfectly circumvent above bottleneck, but cloning and editing of large-sized BGCs (sometimes over 100 kb) remain challenging. For cloning large-sized DNAs, classic strategies generally utilize randomly digested genomic libraries; however, the screening process is always laborious and it is arduous for packaging intact BGCs over 100 kb in a single vector. Previous precision cloning strategies often utilize restriction enzymes (REs) to release target BGCs that subsequently acquired by coupling diverse DNA capturing strategies. Linear–linear homologous recombination (LLHR) uses RecE/T mediated homologous recombination for direct capture of REs generated genome segments and is widely used for direct cloning of NP BGCs from Streptomyces (Fu et al., 2012; Nah et al., 2017). Gibson assembly coupling REs cleavage is also used for capturing BGCs, and accordingly, the conglobatin cluster has target cloned by Leadlay group (Zhou et al., 2015). Transformation-associated recombination (TAR) cloning uses homologous recombination in S. cerevisiae to capture REs generated BGC segments, and has employed for cloning BGCs of taromycin A (Yamanaka et al., 2014), alterochromide (Ross et al., 2015), and thiotetronates (Tang et al., 2015). However, these REs-dependent approaches are severely constrained for broader application since appropriate RE cutting sites do not regularly exist close to BGC terminals. CRISPR/Cas9 system perfectly addresses the limitation, that target cleaves the DNA guided by a synthetic sgRNA, allowing target cloning of large-sized BGCs. Wang et al. (2015) tentatively applied CRISPR/Cas9 system as REs in vitro to linearize a large vector (22 kb) and subsequently seamlessly assembled with a small DNA using Gibson assembly (Wang et al., 2015). For precision acquiring large-sized DNAs harboring NP BGCs, Jiang et al. (2015) developed Cas9-assisted targeting of chromosome segments (CATCH), which allows target cloning of intact BGCs up to 100 kb that cleaved by CRISPR/Cas9 at specific sites guided by custom-designed sgRNAs and subsequent target captured by Gibson assembly (Jiang et al., 2015). Simultaneously, Lee et al. (2015) combined CRISPR/Cas9 with TAR cloning that employs homologous recombination in yeast to target capture CRISPR/Cas9 released chromosomal segments and dramatically accelerated capture efficiency of TAR cloning up to 32% (Lee et al., 2015). Soon after, CRISPR/Cas9 system coupling TAR cloning was further applied to construct even megabase-sized DNA segments. Zhou et al. (2016) developed Cas9-facilitated homologous recombination assembly (CasHRA), which cointroduces large circular DNAs into S. cerevisiae and release the target DNA segments by CRISPR/Cas9 for subsequent assembly by homologous recombination. It provides an alternative for assembly of large-sized BGCs over 100 kb, using DNAs obtained from cosmid libraries of Streptomyces or eDNA. However, it involves assembly steps and tends to be time-consuming.

## CRISPR/Cas9-Mediated Editing of BGCs

Acquiring of intact BGCs of interest is the first step to heterologously investigate the novel NP drugs. Editing of acquired BGCs is generally required for successful heterologous expression. Routine strategies for BGCs editing are always constrained by difficulty of handling large-sized DNAs, and are always laborious. λ-Red recombination mediated PCRtargeting has often used for editing BGC by creating gene replacements/deletions; however, residues like antibiotic selection markers or FRT sequence remain at editing sites, and unintended recombination may raise from repetitive sequences of such modular PKS or NRPS genes (Gust et al., 2003; Yamanaka et al., 2014). λ-Red recombination also enables promoter refactoring or domains/modules exchange for characterization of NPs biosynthesis (Nguyen et al., 2006; Du et al., 2013). Recently, a more facile promoter refactoring approach based on homologous recombination in S. cerevisiae has been developed by Brady group, that enables multiplex promoter refactoring in a single TAR reaction (Montiel et al., 2015). Based on this, production of eDNA-derived indolotryptoline antiproliferative agents, lazarimides A and B, was activated. Nevertheless, the refactoring rate is relatively low. Homologous recombination in yeast could be greatly improved if specific DSBs are introduced at recombination sites (Storici et al., 2003; Storici and Resnick, 2006; Lee et al., 2015). Accordingly, Brady group developed multiplexed-CRISPR-TAR (mCRISTAR) approach, which introduces CRISPR/Cas9 system to specifically create DSBs across target recombination sites (Kang et al., 2016). With mCRISTAR, multiplex CRISPR/Cas9 generated operon fragments can be reassembled with synthetic promoter cassettes by homologous recombination, and are capable of achieving four promoters exchange simultaneously in a single round using one auxotrophic marker selection, with efficiency up to 80% (**Figure 1C**). General applicability of mCRISTAR has been validated by applying to activate three different cryptic BGCs coding for tetarimycin, lazarimide, and AB1210, indicating a powerful and promising technology for discovery of novel NP drugs from cryptic BGCs resource.

In contrast to the above in vivo strategies for BGC editing based on homologous recombination in E. coli or S. cerevisiae, Sun group developed a new in vitro CRISPR/Cas9-mediated editing (ICE) system for high-efficient BGCs editing (Liu et al., 2015). ICE system allows a complete in vitro operating process with normal molecular operations, which cleaves BGCs at specific sites guided by synthetic sgRNAs and ligates the blunt ends

that are repaired by T4 polymerase, to create gene in-frame deletion/replacement/insertion mutations (**Figure 1D**). With ICE system, BGCs of tetronate RK-682 and dithiolopyrrolone holomycin were readily edited (Liu et al., 2015), especially for Tü 3010, a particular thiotetronate antibiotic, various gene in-frame deletions were rapidly constructed and accordingly deciphered biosynthesis of this exceptional thiotetronate structure (Tao et al., 2016). Soon afterward Lu group utilized a similar in vitro approach that coupling CRISPR/Cas9 system with Gibson assembly to refactor the bacterial artificial chromosome vector harboring BGC of pristinamycin II for following multiplexed sitespecific genome engineering in Streptomyces (**Figure 1D**; Li et al., 2017b). The above two examples indicate that in vitro application of CRISPR/Cas9 could be of wide applicability for BGCs editing, especially coupling the subsequent heterologous expression of BGCs for NP drugs discovery, characterization, and engineering. Nevertheless, optimization of in vitro strategies for multiplex loci refactoring of BGCs is of great necessity, and coupling of CRISPR/Cas9 system with Gibson assembly may preliminarily address the problem.

### CONCLUSION

In conclusion, CRISPR/Cas9 system has proved to be a powerful technology for genome editing or BGC refactoring due to the

#### REFERENCES


outstanding features, like higher sequence specificity, artificial guided targeting, and high editing efficiency. Its applications of genome editing specialized for Streptomyces are still relatively narrower in range, especially for the strains little studied. Thus, more efficient and convenient CRISP/Cas tools are of urgent requirement. For instance, diversified CRISPR/Cas systems like Cpf1 (Zetsche et al., 2015; Fonfara et al., 2016; Li et al., 2016), the newly identified class 2 type V CRISPR/Cas protein, xCRISPR/Cas9 (Hu et al., 2018), the most recently evolved CRISPR/Cas9 system with broad PAM compatibility, and even the CRISPR/Cas systems from Streptomyces (Choi and Lee, 2016) could be introduced for diverse applications in Streptomyces, to advance the researches on NP drugs and open a new era for NP drugs discovery.

#### AUTHOR CONTRIBUTIONS

WT and YS wrote the manuscript. All authors revised and approved the manuscript.

# FUNDING

This research was funded by the National Natural Science Foundation of China (Grant Nos. 31770069 and 31300061).

clusters. Appl. Microbiol. Biotechnol. 97, 6383–6396. doi: 10.1007/s00253-013- 4836-7



**Conflict of Interest Statement:** 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 SC and handling Editor declared their shared affiliation.

Copyright © 2018 Tao, Yang, Deng and Sun. 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.

# Metabolic Engineering of Escherichia coli for Enhanced Production of Naringenin 7-Sulfate and Its Biological Activities

Luan L. Chu<sup>1</sup> , Dipesh Dhakal<sup>1</sup> , Hee J. Shin<sup>1</sup> , Hye J. Jung1,2, Tokutaro Yamaguchi<sup>2</sup> \* and Jae K. Sohng1,2 \*

<sup>1</sup> Department of Life Science and Biochemical Engineering, Sun Moon University, Asan, South Korea, <sup>2</sup> Department of BT Convergence Pharmaceutical Engineering, Sun Moon University, Asan, South Korea

Flavonoids are one of the predominant groups of plant polyphenols, and these compounds have significant effects on human health and nutrition. Sulfated flavonoids have more favorable attributes compared to their parent compounds such as increased solubility, stability, and bioavailability. In this research, we developed a microbial system to produce sulfated naringenin using Escherichia coli expressing a sulfotransferase (ST) from Arabidopsis thaliana (At2g03770). This wild-type strain was used as a model system for testing clustered regularly interspaced short palindromic repeats (CRISPR) interference (CRISPRi) metabolic engineering strategies. Using synthetic sgRNA to mediate transcriptional repression of cysH, a gene encoding 3<sup>0</sup> -phosphoadenosine-5 0 -phosphosulfate (PAPS) ST, which is involved in sulfur metabolism, resulted in an increase in intracellular PAPS accumulation by over 3.28-fold without impairing cell growth. Moreover, naringenin 7-sulfate production by engineering E. coli with its cysH gene repressed in the open reading frame through CRISPRi was enhanced by 2.83-fold in compared with the wild-type control. To improve the efficiency of biotransformation, the concentration of SO2<sup>−</sup> 4 , glucose, and substrate were optimized. The bioproductivity of naringenin 7-sulfate was 135.49 µM [∼143.1 mg (47.7 mg L−<sup>1</sup> )] in a 3-L fermenter at 36 h. These results demonstrated that the CRISPRi system was successfully applied for the first time in E. coli to develop an efficient microbial strain for production of a sulfated flavonoid. In addition, antibacterial and anticancer activities of naringenin 7-sulfate were investigated and found to be higher than the parent compound.

Keywords: sulfotransferase, CRISPRi, E. coli, PAPS, metabolic engineering

# INTRODUCTION

Flavonoids are major natural phenolic compounds found in plants (Buer et al., 2010). One of the most predominant citrus flavonoids is naringenin and it was found to exhibit various biological effects on human health and nutrition. Naringenin showed anti-estrogenic, antioxidant (Bugianesi et al., 2002), anti-obesity and anti-diabetic activities (Hossain et al., 2016). It has been demonstrated to control non-alcoholic steatohepatitis and associated inflammation (Jadeja and Devkar, 2014). Like flavonoids, most of the naringenin in nature accumulates in a glycosylated form (Gattuso et al., 2007), which improves solubility, storage, and stability of the parent compounds

#### Edited by:

Biswarup Mukhopadhyay, Virginia Tech, United States

#### Reviewed by:

Blaine Pfeifer, University at Buffalo, United States Alvaro R. Lara, Universidad Autónoma Metropolitana, Mexico

#### \*Correspondence:

Tokutaro Yamaguchi yamaguchi@sunmoon.ac.kr Jae K. Sohng sohng@sunmoon.ac.kr

#### Specialty section:

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

> Received: 29 March 2018 Accepted: 04 July 2018 Published: 27 July 2018

#### Citation:

Chu LL, Dhakal D, Shin HJ, Jung HJ, Yamaguchi T and Sohng JK (2018) Metabolic Engineering of Escherichia coli for Enhanced Production of Naringenin 7-Sulfate and Its Biological Activities. Front. Microbiol. 9:1671. doi: 10.3389/fmicb.2018.01671

(De Bruyn et al., 2015). In addition to glycosylation, many the sulfate conjugate of flavonoids discovered in the different type of plants (Barron et al., 1998). However, the biological and physiological properties of sulfated flavonoids have not been studied (Totta et al., 2005; Wang et al., 2014).

Sulfation plays important roles in detoxification of xenobiotics (Yi et al., 2006; Chen et al., 2015) and regulating the activity of animal hormones (Geyer et al., 2017). Enzymatic sulfation is catalyzed by a family of sulfotransferases (STs) that transfer the sulfonate group of 3<sup>0</sup> -phosphoadenosine-5<sup>0</sup> -phosphosulfate (PAPS) to a hydroxyl group or amino group of acceptor compounds with the parallel formation of 3<sup>0</sup> -phosphoadenosine-5 0 -phosphate (PAP) (Paul et al., 2012). Although the sequences of many STs have been deposited to various genome databases of plant (Gidda and Varin, 2006; Marsolais et al., 2007), bacteria (Hossain et al., 2011), and mammalian (Wong et al., 2010; Shimohira et al., 2017), the enzymatic and microbial synthesis of sulfonated natural products still are limited. One of the major problems of in vitro sulfation reactions is the use of the costly and unstable sulfate donor, PAPS, which impedes industrial scale-up of the process. Moreover, while chemical synthesis of sulfated compounds is tedious and time-consuming in terms of involving multiple steps (Zhang et al., 2012), PAPS has a poor availability in the cytosol of microbial cells like Escherichia coli and Streptomyces for production or modification of natural compounds (Sekowska et al., 2000; Nishikawa et al., 2017). Recently, genetic engineering has been used for production or post-modification of the natural compound via improving the availability of the common precursor in the biosynthetic pathway. For example, metabolic engineering of E. coli in the superpathway of methionine and S-adenosyl-L-methionine (SAM) biosynthesis, lead to improved SAM availability, and finally increased anthocyanin O-methylation (Cress et al., 2017). The sugar pathway was overexpressed to improve the cytoplasmic pool of NDP-sugars and subsequently expressed glycosyltransferase was used for the biosynthesis of glycosylated flavonoids in E. coli (Kim et al., 2015; Pandey et al., 2016). However, the same approach has not been applied to the biosynthesis of sulfated flavonoids.

Genetic editing using the clustered regularly interspaced short palindromic repeat (CRISPR) system has been widely used in diverse organisms including bacteria (Huang et al., 2015; Wu et al., 2015), fungal kingdoms (Wang et al., 2017), plant (Khan et al., 2017), animals (Gandhi et al., 2017), and human cell lines (Bertomeu et al., 2017). In the type II CRISPR, an RNA-guide DNA endonuclease (Cas9) is targeted to specific DNA sequences through a chimeric guide RNA (gRNA), which is a fusion between a precursor CRISPR RNAs (crRNA) and trans-activating CRISPR RNAs (tracrRNAs) (Huang et al., 2015). This gRNA is capable of recognizing sequences target sites for marker-free integration or gene disruption that are followed by the protospacer-adjacent motif (PAM) sequence NGG, N being any nucleotide (Cress et al., 2015). Recently, CRISPR interference (CRISPRi) has been developed which contains a mutation in the Cas9 protein (D10A and H840A) resulted in a lack of endonuclease activity (dCas9), but DNA binding capability remained for inhibition of transcription (Bikard et al., 2013). CRISPRi has been applied to down-regulation of genes in certain pathways via metabolic engineering of E. coli for production of value natural compounds (Wu et al., 2015; Liang et al., 2016; Gao et al., 2017). However, it has not been reported to regulate a gene in E. coli sulfur metabolism.

In this study, we expressed a ST from Arabidopsis thaliana (At2g03770) for biosynthesis of sulfated naringenin in E. coli BL21 (DE3). Naringenin has been shown to be substrate specificity of At2g03770 (Hashiguchi et al., 2013). Furthermore, we employed a CRISPRi system as a tool for improving PAPS availability by knockdown PAPS ST (cysH) in the sulfur metabolism (**Figure 1**), resulted in enhancement the final products. This is the first research that a metabolic engineering approach for conjugating a sulfate moiety generated in the cytoplasm of E. coli. The biosynthesized compound was also tested for its potential bioactivities against various pathogens and cancer cell lines.

## MATERIALS AND METHODS

## Chemicals and Reagents

Restriction enzymes, shrimp alkaline phosphatase, T<sup>4</sup> polynucleotide kinase, and T<sup>4</sup> DNA ligase were obtained from New England Biolabs (Hertfordshire, United Kingdom). Standard naringenin, adenosine 3<sup>0</sup> -phosphate 5<sup>0</sup> -phosphosulfate lithium salt hydrate (PAPS), adenosine 5<sup>0</sup> -triphosphate disodium salt hydrate (ATP), adenosine 5<sup>0</sup> -diphosphate (ADP), deuterium oxide (D2O), and dimethyl sulfoxide-d<sup>6</sup> (DMSO-d6) were purchased from Sigma-Aldrich (St. Louis, MO, United States). Isopropyl-β-D-thiogalactopyranoside (IPTG) was obtained from GeneChem Inc. (Daejeon, South Korea). All other chemicals were of the highest grade commercially available.

#### Plasmid and Strains Constructions

All the E. coli strains and plasmids used in this study are supported in **Table 1**. PCR primers utilized for gene amplification and cloning were synthesized by GenoTech Corp. (Daejeon, South Korea). At2g03770 were codon-optimized for E. coli, synthesized, cloned into the plasmid vector pUC57 by General Biosystems, Inc. (NC, United States). The BamHI and XhoI sites were added to the 5<sup>0</sup> and 3<sup>0</sup> ends of the gene. The synthesized gene was restricted by BamHI and XhoI digestion, subcloned into the expression vector pET28a (+) to create the pET28a(+)-At2g03770. Then, the recombinant plasmid was transformed into E. coli BL21 (DE3) to form the strain used for production (wild-type) (**Table 1**).

Plasmids used for CRISPRi/dCas9-mediated transcriptional repression was obtained from Addgene (Plasmid #65006) (Supplementary Figure S1) and constructed as previously reported (Cress et al., 2015). The specific targeting spacer of cysH from the E. coli K12 genomic DNA was identified near promoter (cysH1) and open reading frame region (cysH2) to prevent RNAP binding and elongation, respectively. The primer pairs crRNA1-Fw/Rv and crRNA2-Fw/Rv were used to construct CRISPRi-1 and CRISPRi-2 listed in Supplementary Table S1. Both primers were synthesized, phosphorylated with T<sup>4</sup> polynucleotide kinase, and annealed (Cress et al., 2017).

cysIJG, sulfite reductase; adk, adenylate kinase; APS, adenosine 5<sup>0</sup> -phosphosulfate; PAP, 3<sup>0</sup> -phosphoadenosine-5<sup>0</sup> -phosphate.



The products insert were then ligated into BsaI-digested, dephosphorylated, gel-purified CRISPRi plasmid backbone. E. coli DH5α was used for cloning experiments. All CRISPRi plasmid arrays possessing synthetic the specific targeting spacer were verified by colony PCR with primer pairs cPCR-Fw/Rv (Supplementary Table S2) and sequencing. Plasmids were constructed with CRISPRi-1 and CRISPRi-2 were transformed into the wild-type strain through calcium chloride and a heat-shock method (Dagert and Ehrlich, 1979), forming variant S1 and S2 strains (**Table 1**).

# Culture Conditions and Recombinant Protein Expression

Wild-type E. coli was cultured in Luria Bertani (LB) liquid medium. The sample was kept in a 250 mL flask on a 37◦C with

shaking incubator 200 rpm. Kanamycin (50 µg mL−<sup>1</sup> ) was used for plasmid selection and maintenance. A total of 250 µL of the pre-inoculum of wild-type E. coli was transferred to 50 mL fresh LB liquid medium containing an appropriate amount of antibiotic and then incubated at 37◦C with 200 rpm shaking. Protein expression was induced by the different concentration of IPTG (0.1, 0.5, and 1.0 mM) when the optical cell density at 600 nm (OD600 nm) reached 0.6–0.8. The cell growth was continued at 20◦C for 20 h and harvested by centrifugation at 842 × g for 15 min. The sample was washed twice with 50 mM Tris–HCl (pH 7.5) buffer containing 10% glycerol and prepared for sonication in 1 mL of the same buffer. Following centrifugation at 13,475 × g for 30 min at 4◦C, the resulting soluble and insoluble protein fractions were analyzed by 12% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE).

# Analysis of Cell Growth and mRNA-Expression Levels

The seed culture of the wild-type and S1, S2 strains were transferred to 50 mL fresh LB medium in a 250 mL flask and OD600 nm of each flask was diluted to 0.1. All the culture containing chloramphenicol (34 µg mL−<sup>1</sup> ) or kanamycin (50 µg mL−<sup>1</sup> ) were growth under sharking at 37◦C with 200 rpm shaking. After 6 h, 0.1 mM IPTG was added to induce the CRISPRi system and the growth was continued at 20◦C with 200 rpm shaking for 48 h.

Total RNA was isolated by using RNeasy Plus Mini Kit (Qiagen, United States). The QuantiTect Reverse Transcription Kit (Qiagen, United States) was used to synthesize the cDNA from 1 µg total RNA sample. Power SYBR Green Master Mix (Thermo Fisher Scientific, United States) was performed by quantitative real-time PCR (qRT-PCR) on StepOnePlus real-time PCR system (Applied Biosystems, United States). The 16S rRNA gene was used to an endogenous control with primer pairs 16S rRNA-Fw/16S rRNA-Rv (Supplementary Table S2). The comparative 2−Ct experiment was used to calculate relative gene expression (Livak and Schmittgen, 2001) by using wild-type as the reference sample with primer pairs cysH-Fw/cysH-Rv. The results were analyzed using StepOne Software v2.3 (Applied Biosystems, United States).

# Intracellular PAPS Collection and Quantification

Three recombinant strains including wild-type, S1 and S2 were grown, induced with 0.1 mM IPTG and expressed in 50 mL LB medium at 20◦C with 200 rpm shaking for 12 h. Then, the cells were harvested by centrifugation, suspended, and incubated in 250 mL flasks containing 50 mL of M9 minimal salt medium pH 7.4 at 28◦C for 12 h. After that, the samples were taken, chilled immediately on ice and centrifuged at 842 × g at 4◦C for 15 min. One milliliter of 6% perchloric acid was used for cell lysis and 0.3 mL of 3 M potassium carbonate was added to neutralize while vortexing the sample (Fowler et al., 2009). After centrifugation removed the cell residue, the supernatant containing was collected, filtered through a 0.22-µm syringe filter and injected into UFLC-PDA to determine the products yield. The determination of microbial biomass was carried out by dry cell weight using a 0.22-µm syringe filter. An aliquot of 1 mL cell culture was filtered, washed with distilled water, and dried in a conventional oven. DCW was represented by the weight difference between empty membranes and those with cell residues (Wu et al., 2015).

# Sulfated Naringenin Production and Extraction

The wild-type and S1, S2 strains harboring heterologous a ST and CRISPRi were first expressed in 50 mL fresh LB medium by 0.1 mM IPTG at 20◦C for 12 h. Next, the cells were harvested by centrifugation, washed twice with 100 mM phosphate buffer pH 7.4, and incubated at 28◦C with shaking in 250 mL flasks containing 50 mL M9 minimal salt medium pH 7.4 (Yan et al., 2005). Appropriate antibiotics and IPTG were added at the same time. Subsequently, 100 µM substrate was supplemented to the same samples and kept on 48 h. We carried out fermentation in 3-L of optimal media under an optimal condition for large-scale production of naringenin derivatives. The processing of fermentation was followed as previously described (Chu et al., 2016). Furthermore, the culture samples were extracted with twice volume of ethyl acetate. The extracts were concentrated by freezing rotary evaporator, suspended in methanol, and then injected into UFLC-PDA to determine the products yield.

# HPLC Analysis

The UFLC-PDA system containing a reversed-phase column [Mightysil RP-18 GP 250-4.6 (5 µm) Cica-Reagent; Kanto Chemical Co., Inc., Japan] was used to separate the samples. The mobile phase containing solvent A (HPLC-grade water consisting 0.05% TFA) and solvent B (100% HPLC-grade methanol) was maintained at 30◦C. A calibration PAPS and naringenin standards were created with various concentration.

For quantification of PAPS, the intracellular PAPS was analyzed at a UV absorbance of 254 nm (Uesugi et al., 1997; Xu et al., 2012). The mobile phase including solvents A and B was maintained at 1 mL min−<sup>1</sup> for 10 min. The program was followed by 10% B for 1 min, 20% B for 1 min, 30% B for 30 s, 35% B for 30 s, 40% B for 1 min, 50% B for 4 min, 90% B for 1 min, and 10% B for 1 min.

Quantification of production was based on the peak areas obtained at 290 nm. The conversion percentage of the substrate was determined after integrating substrate and product peaks. Solvents A and B were run at 1 mL min−<sup>1</sup> for a 30 min. The gradient of the mobile phase was carried out and followed by 20% B for 5 min, 50% B for 5 min, 70% B for 5 min, 90% B for 5 min, and 10% B for 10 min.

The purification of naringenin derivatives was performed by preparative HPLC (Shimadzu, Tokyo, Japan) with C<sup>18</sup> column [YMC-Pack ODS-AQ (150 mm × 20 mm I.D., 10 µm)] connected to a UV detector at 290 nm using a 40 min binary program with 5% B for 5 min, 40% B for 5 min, 40% B for

5 min, 90% B for 5 min, and 10% B for 10 min at a flow rate of 10 mL min−<sup>1</sup> .

# Mass Spectrometry and Nuclear Magnetic Resonance

High-resolution quadruple time-of-flight electrospray ionization-mass spectrometry (HR–QTOF ESI/MS) analysis was carried out with an ACQUITY column coupled an SYNAPT G2-S (UPLC, Waters Corp., Billerica, MA, United States). A reversed-phase column [Acquity BEH C<sup>18</sup> 2.1 mm × 100 mm (1.7 µm)] was used to separate the samples. ESI/MS detection of the samples: positive ion mode ESI+, acquisition range: 50–1,400 m/z, capillary voltage: 2.5 kV, cone voltage: 30 V, source temperature: 120◦C, desolvation gas temp: 600◦C, cone gas flow: 20 L h−<sup>1</sup> , desolvation gas flow: 800 L h−<sup>1</sup> . The analyses were performed at a flow rate of 0.35 mL min−<sup>1</sup> using the same mobile phase with a gradient of 30% B for initial, 90% B for 4 min, 100% B for 3 min, 100% B for 2.5 min, and 37.5% B for 2.5 min. MassLynx software version 4.1 (Waters Corp.) was used for analysis the chromatograms.

The purified products were dried, lyophilized, and recorded on Bruker Biospin 300 MHz spectrometer in DMSO-d<sup>6</sup> for proton <sup>1</sup>H-NMR.

# Antibacterial Activity

Nine Gram-positive bacteria and six Gram-negative bacteria listed in **Table 2** were used for testing the antibacterial of naringenin and its derivative. All strains were cultured in LB medium at 37 ± 1 ◦C. We applied the paper disc diffusion method using ampicillin (current antibiotic standard) as an antibacterial positive control for screening the antibacterial agent (Rios et al., 1998). Each inoculates consisted of 10<sup>6</sup> colony forming units (CFU mL−<sup>1</sup> ) and was spread on MHA plates for bio-assay. Sterile filter paper discs (6 mm in diameter) consisting of 10 µL of 100 mM of compounds dissolved in MeOH were spotted on the agar surface. The plates were incubated at 37 ± 1 ◦C and checked for 36 h.

# Anticancer Activities

Three cancer cell lines containing A375SM melanoma, MCF-7 breast cancer, AGS gastric cancer, and 267B1 cellosaurus were observed from the Korean Cell Line Bank (KCLB, Seoul, South Korea). Minimum essential medium (MEM) added 10% fetal bovine serum (FBS) (Gibco, Grand Island, NY, United States) was used for culture the cell lines at 37◦C in a humidified 5% CO<sup>2</sup> incubator. Cells seeded at 2 × 10<sup>3</sup> cell well−<sup>1</sup> in 96-well plates (SPL Life Sciences, Gyeonggi, South Korea) were treated with each compound in serial dilution (400, 200, 100, and 50 µM) for 72 h. Taxol, a current anticancer agent, was also carried out to provide an additional point of comparison. We examined cell growth using an MTT colorimetric assay (Lee et al., 2017).

# Statistical Analysis

Values are mean ± standard deviation (SD). All the results were the average of three independent experiments with SD (n = 3). Differences with p-values <0.05 and <0.005 were indicated a considered significant.

# RESULTS

# Escherichia coli Expression of Recombinant A. thaliana SULTs

To optimize conditions for the expression of recombinant ST At2g03770 from A. thaliana, we induced with various IPTG

TABLE 2 | Inhibition zone diameter (mm) of naringenin and naringenin 7-sulfate against nine Gram-positive bacteria and six Gram-negative bacteria.


−, no inhibition zone; +, inhibition zone detected.

concentrations (0.1, 0.5, and 1.0 mM) under 20◦C in LB liquid medium. The SDS-PAGE analysis of the soluble and insoluble fraction of At2g03770 indicated the highest expression level as determined at 0.1 mM IPTG. The recombinant protein, At2g03770 (37.72 kDa), was overexpressed in E. coli BL21 (DE3) and obtained in the soluble fraction (Supplementary Figure S2).

#### Construction of CRISPRi System in E. coli

Gene cysH was selected for studying CRISPRi system on the sulfur metabolism of E. coli. The knockdown of cysH might to the increase of the intracellular PAPS pool enhancing production of sulfated substrates. Therefore, two 66 bp complementary offset oligonucleotides containing 30 bp protospacer (target) sequence were designed (Supplementary Table S1) and inserted to pCRISPathBrick, resulted in CRISPRi-1 and CRISPRi-2 systems formation (**Table 1**). The length of the cysH gene is 735 bp, while the length of the target sequence is 30 bp. To transcriptional interference, cysH1 was designed 56 bp upstream from start codon (from position −76 to −57), after the PAM sequence AGG (from position −58 to −56). At the same time, cysH2 was designed 10 bp downstream of the start codon (from position 40 to 11), after PAM sequence AGG (from position 41 to 43) (**Figure 2A**). The unique properties of E. coli BL21 genomic of each two protospacer were confirmed via nucleotide BLAST<sup>1</sup> . The plasmids were assembled and then biotransformation into E. coli DH5α. Colony PCR (cPCR) was performed with cPCR-Rv and cPCR-Fw primers (Supplementary Table S2). The cPCR products were confirmed through 2% agarose gel with 50 bp DNA ladder marker (ELPIS-Biotech. Inc., South Korea). The length of amplifying obtained from clones of pCRISPathBrick is 85 bp, while a 66 bp increase in PCR obtained from clones of CRISPRi-1 and CRISPRi-2 (Supplementary Figure S3). The plasmids CRISPRi-1 and CRISPRi-2 were transformed into wild-type strain, forming production strains S1 and S2, respectively (**Table 1**).

# Effect of CRISPRi System on Cell Density and Gene Expression

We determined and compared the growth curves of wild-type, S1, and S2 strains by using a spectrophotometer (Thermo Fisher Scientific, United States) via the OD600 nm. The data showed that the S1 and S2 had to resemble with the wild-type on the rate of cell growth (**Figure 2B**), indicating that cysH gene inhibition by CRISPRi did not affect the cell growth. In addition, the relative qRT-PCR data of knockdown S1 and S2 strains were analyzed through cysH mRNA quantification using 16S rRNA as an endogenous control and wild-type as the reference sample. 16S rRNA, known as a housekeeping gene, is one of the most commonly used endogenous control in E. coli (Zhou et al., 2011). The data showed qRT-PCR cycle threshold values for 16S rRNA gene expressed at almost similar levels in all three samples containing wild-type, S1, and S2 with 23.852 ± 0.371, 22.558 ± 0.031, and 21.732 ± 0.115, respectively. This results demonstrated that the suitability of 16S rRNA gene for the specific case of this experiment (Supplementary Figure S4). Importantly, the value of relative expression of cysH gene transcription in wild-type was 1, while this figure for S1 and S2 were 0.312 and 0.081, respectively (**Figure 2C**). The result indicated that both the prevent RNA polymerase binding (CRISPRi-1) and elongation (CRISPRi-2) have high efficiency of the reducing transcriptional expression level of cysH.

#### The Knockdown of cysH Gene Increased PAPS and Naringenin Derivatives

The intracellular PAPS in three strains were extracted using the method described above. All the samples consisting authentic ADP, ATP, and PAPS standard were determined by using UHPLC-PDA coupled HR-QTOF ESI/MS. UHPLC-PDA analysis showed the ADP standard appeared at retention time t<sup>R</sup> ∼ 0.83 min, while t<sup>R</sup> of ATP and PAPS standard were 2.71 and 1.27 min. Interestingly, the intracellular PAPS was detected at t<sup>R</sup> ∼ 1.27 min in the cytosol of wild-type as well as mutant strains (Supplementary Figure S5). These peaks were confirmed by HR-QTOF ESI/MS and the results were shown in Supplementary Table S3 and Supplementary Figures S6, S7. These results suggested that UHPLC-PDA coupled HR-QTOF ESI/MS not only the distinguishing between PAPS from potential interferents ADP and ATP but also the detection of the intracellular PAPS in the cytosol of E. coli. Finally, the intracellular concentration of PAPS was analyzed and shown in **Figure 2D**. While the figure for S1 and S2 strains went up to 40.58 and 74.85 µM g−<sup>1</sup> DCW, the concentration of PAPS for wild-type obtained a 22.85 µM g−<sup>1</sup> DCW at the same time. This result demonstrated that cysH is the most required target to inhibit the PAPS consumption.

Moreover, we used the three recombinant strains to produce the sulfated derivative from naringenin. The UFLC-PDA chromatograms of extract from the whole cell of all strains showed a new peak at retention time t<sup>R</sup> ∼ 18.248 min (P1) in comparison with naringenin standard at t<sup>R</sup> ∼ 19.979 min under UV absorbance at 290 nm (Supplementary Figure S8). These peaks were further analyzed by HR-QTOF ESI/MS. The found mass of naringenin was ∼273.0780 [M + H]<sup>+</sup> m/z<sup>+</sup> equivalent to molecular formula C15H13O<sup>5</sup> with λmax ∼ 287 nm, for which calculated mass was ∼273.0763 (Supplementary Figure S9A). The found mass of P1 at ∼353.0330 with λmax: 277; 335 nm, corresponding to the exact mass of the mono-sulfate derivative of naringenin with molecular formula C15H12O8S for [M + H]<sup>+</sup> m/z<sup>+</sup> ∼ 353.0331 (Supplementary Figure S9B). The structural identifies of the product could be verified via NMR in the future experiment. The percentages of bioconversion of naringenin to mono-sulfated naringenin were 17.2% in the wildtype strain, while the figure for both S1 and S2 strains showed an enhancement to 32.47 and 48.67% at the same time (**Figure 2E**). These results demonstrated that S1 and S2, both mutant strains include the knockdown of the cysH gene were used for the improvement of naringenin derivative. It might be true that S2

<sup>1</sup>https://blast.ncbi.nlm.nih.gov

could be a good recombinant host system produce derivatives product form naringenin.

# Regulating the Concentration of Inorganic Sulfate and Glucose in Medium

For regulation of the concentration MgSO4, we used the M9 minimum media including the various concentrations of MgSO<sup>4</sup> (2, 5, 10, 15, and 20 mM) in comparison with the M9 medium without MgSO4. The cell growth and substrate conversion were taken out at 12 h intervals. While the production was noticeably low in the medium without MgSO4, the maximum bioconversion of substrate obtained 98.34% at 48 h with OD<sup>600</sup> ∼ 9.94 in M9 medium with 10 mM MgSO<sup>4</sup> when 100 µM naringenin was added (**Figures 3A,B**). Subsequently, the production was carried out by using an M9 medium with 10 mM MgSO<sup>4</sup> consisting of the various concentrations of glucose (2, 4, 6, 8, and 10%). The data showed that nearly 100% of 100 µM naringenin was converted to its sulfated derivatives during the addition 2, 4, and 6% glucose at 48 h with OD<sup>600</sup> ∼ 9.94, 10.73, and 9.57, respectively (**Figures 3C,D**). The M9 medium supplementation of 10 mM MgSO<sup>4</sup> and 2% glucose were selected for further optimizing the concentration of substrate.

### Bioconversion With Different Naringenin Concentration and Scale-Up by Fermentation

Five various concentration of naringenin (200, 400, 600, 800, and 1,000 µM) were supplied for biocatalytic reaction system with the S2 strain. The OD<sup>600</sup> and conversion rate of naringenin to its derivative were monitored at 12-h intervals. The highest production obtained 189 µM at 48 h with OD<sup>600</sup> ∼ 10.24 when 250 µM was fed under M9 medium including 10 mM MgSO<sup>4</sup> and 2% glucose (**Figures 4A,B**).

Finally, these optimal conditions were applied for the bioconversion process into the 3-L fermenter (Biotron, South Korea). When OD<sup>600</sup> ≥ 6, 0.1 mM IPTG was induced and the temperature decreased to 20◦C. After 12 h induction, 250 µM (∼264 mg 3-L) of naringenin was fed to

the culture under the pH and temperature were maintained at 7.4 and 30◦C, respectively. The samples were taken at 12 h interval and measured by UFLC-PDA. UFLC-PDA analysis revealed that sulfated naringenin was obtained to 135.49 µM (∼143.1 mg 3-L) at 36 h with OD<sup>600</sup> ∼ 51.06 (**Figure 4C**).

# Structural Elucidation of Sulfated Naringenin

Previous study exhibited that even though no sulfating activity toward 5-, 3<sup>0</sup> and 4<sup>0</sup> -hydroxyflavone, At2g03770 showed the catalytic activity of 7-hydroxyflavone (Hashiguchi et al., 2013). We re-confirmed the structural compound by <sup>1</sup>H-NMR of naringenin standard and purified sulfated product at 300 MHz in DMSO-d6. The <sup>1</sup>H-NMR spectrum of sulfated naringenin displayed the absence of OH− group signal at δ = 10.83 ppm for C-7 in comparison with the <sup>1</sup>H-NMR spectrum of naringenin (Supplementary Figure S10). Moreover, the <sup>1</sup>H-NMR spectrum of this compound exhibited a lower shift at H-6 and H-8 (Supplementary Table S4), indicating that the hydroxyl group at 7 positions of naringenin was substituted by a sulfate group. These data also agreed well with naringenin 7-sulfate obtained in the fungus Cunninghamella elegans NRRL 1392 (Ibrahim, 2000). Based ppm all the results, we could be identified that product was naringenin 7-sulfate. While naringenin was colorless compound, naringenin 7-sulfate was obtained as a yellowish solid (**Figure 4D**).

# Antibacterial and Anticancer of Compounds

Results of disc diffusion assays displayed that naringenin showed only antibacterial activity against Salmonella enterica ATCC 14028 when 10 µL of 100 mM compound was used. In contrast, naringenin 7-sulfate exhibited broad-spectrum antibacterial activity against not only Gram-positive bacteria containing Staphylococcus aureus CCARM 0205, S. aureus CCARM 0204, S. aureus CCARM 3634, Proteus hauseri NBRC 3851, Micrococcus luteus KACC 13377, but also negative– positive bacteria consisting of S. enterica ATCC 14028, Pseudomonas aeruginosa KACC 10232, and Enterobacter cloacae subsp. dissolvens KACC 13002. Noticeably, naringenin 7-sulfate exhibited antibacterial activity against M. luteus KACC 13377 with a zone of inhibition values of 9.5 ± 0.18 mm. Moreover, we detected a zone of inhibition against S. aureus CCARM 0205 (MSSA) by naringenin 7-sulfate and ampicillin seems to share the most similarity with values of 7.0 ± 0.08 and 7.2 ± 0.09 mm. Interestingly, while naringenin and ampicillin did not show antibacterial activity against P. aeruginosa KACC 10232, naringenin 7-sulfate displayed a zone of inhibition values of 7.2 ± 0.12 mm (**Table 2** and Supplementary Figure S11). These results indicated that sulfation of naringenin at hydroxyl

group of C-7 position could be advantageous for intensifying its antibacterial activity against various bacteria.

Furthermore, the cell viability data showed that naringenin 7-sulfate exhibited good anticancer activities where naringenin did not have anticancer activities against all cell lines. Cell viability of A375SM, MCF-7, AGS, and 267B1 treated 400 µM of naringenin 7-sulfate reduced approximately 48.85, 35.96, 41.78, and 71.20% (p < 0.05), respectively, in comparison with normal test (NT) (**Figure 5**). These results suggest that naringenin 7-sulfate was relatively less cytotoxic to non-cancer cell than cancer cell lines. In addition, the data showed naringenin 7-sulfate was less cytotoxic than naringenin in the non-cancer cell at 400 µM. As expected, taxol exhibited highly strong anticancer activity at nanomolar ranges of concentration. Taxol also inhibited A375SM and AGS cancer cell lines a little better than 267B1 non-cancer cell line with inhibition of 65.47, 63.40, and 49.90% at 100 nM, respectively (**Figure 5**). In conclusion, naringenin 7-sulfate could be the more potent anticancer agent with lower cytotoxicity against non-cancer cell than naringenin.

#### DISCUSSION

Sulfotransferases can be found in various life form from prokaryotes to eukaryotes (Suiko et al., 2017). In a model plant organism, A. thaliana, 18 different STs classified into seven groups are present (Klein and Papenbrock, 2004). Among all the A. thaliana STs, At2g03770 has been exploited as catalysis the sulfation for a variety of flavonoids (Hashiguchi et al., 2013). In this research, we produced 17.2 µM naringenin derivatives via overexpressed At2g03770 in E. coli. Although the yield of substantial amount of At2g03770 product appears in insoluble fraction might be led to the lower harvest of sulfated product (Supplementary Figure S2), the result indicated that heterologous expression STs from A. thaliana could be efficient produce sulfated compounds in whole cell biotransformations. Moreover, metabolic engineering in E. coli also might be applied for the enhanced production of prospective compounds extensively.

In E. coli sulfur metabolism, operon cysUWAsbp codes for a sulfate permease, an ATP binding cassette (ABC)-type transporter (Sekowska et al., 2000), carried the sulfate into the internal cell. Subsequently, ATP sulfurylase coded by operon cysDN, catalyzed an adenylylation of sulfate in the form of APS. After being activated by APS kinase cysC, APS is phosphorylated to PAPS (Hatzios and Bertozzi, 2011). PAPS is the universal sulfate donor in the reaction of sulfation performed by STs on the bacterial metabolite. On another hand, consumption of PAPS in sulfate assimilation pathway start catalyzed with the cysH-encoded PAPS ST. This reaction is continuously reduced for the biosynthesis of essential reduced sulfur metabolites (**Figure 1**). Therefore, CRISPRi has been introduced into background strain consist of expressed ST At2g03770, targeting strategy to inhibit PAPS consumption and subsequently increase

sulfated naringenin. This is the first time the production of sulfated flavonoids has published with the CRISPRi system in E. coli.

CRISPRi recently has applied to the enhanced biosynthesis of flavonoid (Cress et al., 2015) and O-methylated anthocyanin (Cress et al., 2017) through transcription regulation of metabolic pathway in E. coli. Compared to the traditional gene deletion methods, CRISPRi showed the great tool for targeted gene inhibition in microorganisms. Firstly, the construction of CRISPRi plasmid is simple and saving time because dCas9 protein and sgRNA expressed in one vector (Supplementary Figure S1). The efficiency of two different nucleotide sequence targeted on the cysH gene was not the same. The proportion suppression of spacer sequence bound to open reading frame region (cysH2) was 1.34-fold higher than the figure for spacer sequence bound to near promoter (cysH1), at 91.9 and 68.8%, respectively (**Figure 2C**). The reason behind the variation in gene expression level could be the corresponding to the distance of cysH2 and cysH1 to transcription start site, 10 bp of cysH2 in compared with 56 bp of cysH1 (**Figure 2A**). Moreover, the

regulation of sulfate metabolic pathway by CRISPRi system result in improving PAPS pool without affecting cell growth (**Figure 2B**). Ultimately, CRISPRi system not only reduces the metabolic burden associated with high-copy overexpression of heterologous ST At2g03770, but also increase production of target molecules. In the whole cell biocatalysis system, production of naringenin derivatives increased by up to 2.83-fold (**Figure 2E**). This result suggested that S2 strain not only led to the accumulation of intracellular PAPS but also improved the efficiency of sulfation of naringenin.

Recently, the analysis of polar molecules such as ATP, ADP, and PAPS was carried out via liquid chromatography-mass spectrometry (Johnsen et al., 2011; Dowood et al., 2016). In this study, UHPLC-PDA coupled HR-QTOF ESI/MS allowed the detection PAPS, although the quantity of PAPS was lowed in the cytoplasm of E. coli. The method not only is high accuracy but also simple to perform with general solvent and saving time. Moreover, this method was very significant for the separation between PAPS and nucleoside triphosphates as ATP and ADP, because these compounds were the same physical properties (Dowood et al., 2016).

The cultivation of the wild-type and S1, S2 strains was the initial phase of rapid cell growth at 37◦C, followed by a growth arrest phase at 20◦C, and the biosynthesized compound produced at 28◦C. The growth-arrested E. coli induced by using low temperature led to improved the conformational quality and the solubility of STs from A. thaliana and CRISPRi systems (Vera et al., 2007; Cress et al., 2015), which associated with an increased sustained viability and an extended production phase (Rosano and Ceccarelli, 2014). This experiment has a significant impact on the optimal production of the target compound. We decided to the production sulfated naringenin in M9 medium containing MgSO<sup>4</sup> and glucose as sulfur and carbon sources, respectively. Almost 100% of conversion rate from a substrate to its derivative was obtained in media consisting of 10 mM MgSO<sup>4</sup> and 2–6% glucose, indicating that sulfated compound all accumulated in the extracellular fraction. One possible reason behind this result could be organic anion molecules as sulfated naringenin has been eliminated into extracellular by E. coli multidrug resistance ATP binding cassette transporters (Chang and Roth, 2001). On the other hand, sulfated naringenin including the negative charged SO2<sup>−</sup> 4 led to might not to cross the cell membrane of E. coli. Furthermore, even though the media lacking MgSO<sup>4</sup> resulted in the cell growth was low, the production still obtained around 4.42% (**Figures 3A,B**). The reason could be because of when the growth medium absent inorganic sulfur, E. coli can induce a series of sulfate starvation-inducible genes led to utilize organosulfur compound as a source of sulfur (van der Ploeg et al., 2001). Moreover, the media absence glucose also caused the production was low as well as the cell growth inefficient (**Figures 3C**, **4D**). These results demonstrated that both SO2<sup>−</sup> 4 and glucose are an essential factor produce sulfated naringenin by At2g03770-expressed in E. coli cells. However, the high concentration of SO2<sup>−</sup> 4 (above 15 mM) and glucose (over 6%), as well as more than 0.5 mM of the substrate, could be a reason for inhibition of sulfation product in E. coli whole cell (**Figures 3**, **4**). Finally, the engineered E. coli S2 strain has been applied successfully for the large-scale production of naringenin 7-sulfate, which obtained at 135.49 µM [∼143.1 mg (47.7 mg L−<sup>1</sup> )] in a 3-L fermenter (**Figure 4C**). These results indicate that the system is efficient and could be applied for other flavonoids to generate the libraries of molecules with various sulfation approaches.

We also evaluated the antibacterial activity of naringenin and its derivative. Five Gram-positive bacteria S. aureus CCARM 0205, S. aureus CCARM 0204, S. aureus CCARM 3634, P. hauseri NBRC 3851, M. luteus KACC 13377 and three Gram-negative bacteria S. enterica, P. aeruginosa, E. cloacae were sensitive with naringenin 7-sulfate. This result indicated that negatively charged sulfate group might be improved to naringenin for antibacterial activity, however, the mechanism of this compound against bacterial pathogens have been not reported. In addition, naringenin derivative exhibited the most potential anticancer activity against three tested cancer cell lines. The compound showed the similarity in features between its and taxol the less cytotoxic to non-cancer cell line in comparison with cancer cell lines (**Figure 5**). This is the first report of the activity of naringenin 7-sulfate against A375SM, MCF-7 and AGS cancer cell lines. The previous research has shown that the substituent at the C-7 position of naringenin containing thiophenecarboxylate, phenyl carbonate, isobutyrate, methyl benzoate, and allyloxy inhibited effects on HCT116 human colon cancer cell line via block G1 cell cycle progression by interaction with cyclin-dependent kinase 2 (CDK2) (Yoon et al., 2013). This possible mechanism behind the anticancer activity of naringenin 7-sulfate against three tested cancer cell lines, however, the exact mechanisms of action of this compound must confirm in further studies.

In summary, we targeted to improving production of sulfated flavonoids in engineered E. coli. This is the first report on CRISPRi mediated inhibition in the sulfur metabolism of E. coli. Repression of key reduced sulfur metabolite enzyme cysH by over 91%, causing increase intracellular PAPS accumulation and enhancement of naringenin 7-sulfate over 3.28- and 2.83 fold compared to control, respectively. Further media culture optimization led to obtained at 135.49 µM [∼143.1 mg (47.7 mg L−<sup>1</sup> )] in a 3-L fermenter. This engineered E. coli opened prospects for the biosynthesis of the sulfated flavonoids. In addition, naringenin 7-sulfate exhibited antibacterial activity against both Gram-positive and Gram-negative bacteria. This compound could be also used as a cancer drug when shown to anticancer activity against A375SM melanoma, MCF-7 breast, and AGS gastric cancer cell lines.

#### AUTHOR CONTRIBUTIONS

LC designed, performed the majority of the experiment work, analyzed data, and wrote the manuscript. DD and TY helped in analyzing data. HS and HJ did the majority of anticancer activities. JS and LC were responsible for the original concept and supervised the work. All authors read and approved the final manuscript.

# FUNDING

fmicb-09-01671 July 25, 2018 Time: 16:39 # 12

This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MEST) (NRF-2017R1A2A2A05000939).

#### REFERENCES


#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.01671/full#supplementary-material

pathway characterization with CRISPRi-guided multiplexed metabolic tuning. Biotechnol. Bioeng. 115, 661–672. doi: 10.1002/bit.26486


sulfotransferases from Arabidopsis, AtST4a (At2g14920) and AtST1 (At2g03760). Planta 225, 1233–1244. doi: 10.1007/s00425-006-0413-y


**Conflict of Interest Statement:** 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 © 2018 Chu, Dhakal, Shin, Jung, Yamaguchi and Sohng. This is an openaccess 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-09-01671 July 25, 2018 Time: 16:39 # 13

# Enhancing Production of Pinene in *Escherichia coli* by Using a Combination of Tolerance, Evolution, and Modular Co-culture Engineering

Fu-Xing Niu, Xin He, Ya-Qin Wu and Jian-Zhong Liu\*

Institute of Synthetic Biology, Biomedical Center, Guangdong Province Key Laboratory of Improved Variety Reproduction in Aquatic Economic Animals and South China Sea Bio-Resource Exploitation and Utilization Collaborative Innovation Center, School of Life Sciences, Sun Yat-sen University, Guangzhou, China

α-Pinene is a natural and active monoterpene, which is widely used as a flavoring agent and in fragrances, pharmaceuticals, and biofuels. Although it has been successfully produced by genetically engineered microorganisms, the production level of pinene is much lower than that of hemiterpene (isoprene) and sesquiterpenes (farnesene) to date. We first improved pinene tolerance to 2.0% and pinene production by adaptive laboratory evolution after atmospheric and room temperature plasma (ARTP) mutagenesis and overexpression of the efflux pump to obtain the pinene tolerant strain Escherichia coli YZFP, which is resistant to fosmidomycin. Through error-prone PCR and DNA shuffling, we isolated an Abies grandis geranyl pyrophosphate synthase variant that outperformed the wild-type enzyme. To balance the expression of multiple genes, a tunable intergenic region (TIGR) was inserted between A. grandis GPPSD90G/L175<sup>P</sup> and Pinus taeda Pt1Q457L. In an effort to improve the production, an E. coli-E. coli modular co-culture system was engineered to modularize the heterologous mevalonate (MEV) pathway and the TIGR-mediated gene cluster of A. grandis GPPSD90G/L175<sup>P</sup> and P. taeda Pt1Q457L . Specifically, the MEV pathway and the TIGR-mediated gene cluster were integrated into the chromosome of the pinene tolerance strain E. coli YZFP and then evolved to a higher gene copy number by chemically induced chromosomal evolution, respectively. The best E. coli-E. coli co-culture system of fermentation was found to improve pinene production by 1.9-fold compared to the mono-culture approach. The E. coli-E. coli modular co-culture system of whole-cell biocatalysis further improved pinene production to 166.5 mg/L.

Keywords: pinene biosynthesis, *Escherichia coli,* tolerance engineering, directed evolution, chemically induced chromosomal evolution, modular co-culture

# INTRODUCTION

α-Pinene is a natural and active monoterpene, which is widely used in flavorings, fragrances, insecticides, pharmaceuticals, and fine chemicals (Breitmaier, 2006; Behr and Johnen, 2009; Kirby and Keasling, 2009; Gandini and Lacerda, 2015). It was recently produced as a candidate renewable jet fuel due to its favorable energy content, cold weather properties, and high octane/cetane

#### *Edited by:*

Dipesh Dhakal, Sun Moon University, South Korea

#### *Reviewed by:*

Shuai Qian, Rice University, United States Vlada B. Urlacher, Heinrich Heine Universität Düsseldorf, Germany

> *\*Correspondence:* Jian-Zhong Liu lssljz@mail.sysu.edu.cn

#### *Specialty section:*

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

> *Received:* 30 April 2018 *Accepted:* 28 June 2018 *Published:* 31 July 2018

#### *Citation:*

Niu F-X, He X, Wu Y-Q and Liu J-Z (2018) Enhancing Production of Pinene in Escherichia coli by Using a Combination of Tolerance, Evolution, and Modular Co-culture Engineering. Front. Microbiol. 9:1623. doi: 10.3389/fmicb.2018.01623

numbers (George et al., 2015). The main source of pinene is turpentine, a by-product of the wood pulp industry (Behr and Johnen, 2009). However, this extraction from plants is tedious and inefficient and requires substantial expenditure of natural resources due to low content (Chang and Keasling, 2006). Therefore, there is much interest in developing biotechnologies for pinene production from renewable resources by engineering microorganisms. Similar to other monoterpenes, α-pinenes are biosynthesized from the C5 intermediates isopentenyl diphosphate (IPP) and dimethylallyl diphosphate (DMAPP) via geranyl diphosphate synthase (GPPS). The head-to-tail condensation produces geranyl diphosphate (GPP, C10), which is, in turn, cyclized by pinene synthase (PS) to produce either αor β-pinene. Escherichia coli (Yang et al., 2013; Sarria et al., 2014; Tashiro et al., 2016) and Corynebacterium glutamicum (Kang et al., 2014) have been engineered to produce α-pinene. α-Pinene (5.4 mg/L) has been produced in engineered E. coli through the introduction of a heterologous mevalonate (MEV) pathway and α-pinene synthase (Pt30) from Pinus taeda (Yang et al., 2013). The combinatorial expression of Abies grandis GGPS-PS fusion proteins enhanced pinene production (32 mg/L) in E. coli (Sarria et al., 2014). The directed evolution of α-pinene synthase (Pt1) from P. taeda increased α-pinene productivity. E. coli plasmid-expressing the evolved α-pinene synthase (Pt1Q457L) from P. taeda, MEV pathway enzymes, IPP isomerase and A. grandis GGPS produced the highest levels of pinene (140 mg/L) in a flask culture to date (Tashiro et al., 2016). The coexpression of native 1-deoxy-d-xylulose-5-phosphate synthase (Dxs) and isopentenyl diphosphate isomerase (Idi) with P. taeda PS and A. grandis GPPS in C. glutamicum yielded a pinene level of 27µg/g cell dry weight (Kang et al., 2014).

However, the production level of pinene is much lower than that of hemiterpene (isoprene) (Whited et al., 2010) and sesquiterpenes (farnesene) (Zhu et al., 2014) to date. Pinene is highly toxic to E. coli. E. coli growth is inhibited by 0.5% pinene (Dunlop et al., 2011). The inherent tolerance of E. coli may limit the production potential. It was demonstrated that increasing the tolerance of E. coli by overexpressing the efflux pump AcrBDFa (YP\_692684) from Alcanivorax borkumensis significantly enhanced limonene production (Dunlop et al., 2011). Another reason for the lower yield may be that PS has a lower expression level and/or lower enzymatic activity in E. coli. Thus, we first combined tolerance engineering with directed evolution of the enzyme to improve pinene production in E. coli.

Recently, there has emerged a new modular co-culture engineering approach for engineering microorganisms. Modular co-culture engineering approaches divide a complete biosynthetic pathway into separate serial modules, which are introduced into different strains to accommodate individual modules for achieving designed biosynthesis (Zhang and Wang, 2016). The advantages of using modular co-culture engineering include the following: (1) reducing the metabolic burden on each host strain; (2) providing diversified cellular environments for functional expression of the different pathway genes; (3) reducing the undesired interference of different pathways; (4) easily balancing the biosynthetic pathway between individual pathway modules by simply changing the strain-to-strain ratio; (5) high-efficiency utilization of complex materials containing multiple active substrates; and (6) supporting the plug-and-play biosynthesis of various target products (Zhang and Wang, 2016). Thus, modular co-culture engineering was also used to enhance pinene production in E. coli.

# MATERIALS AND METHODS

### Strains, Plasmids, and Primers

The bacterial strains and plasmids used in this study are listed in **Table 1**. The primers used in this study are listed in Supplementary Table 1.

# Genetic Methods

pMEVI was derived from pJBEI-6409 (Alonso-Gutierrez et al., 2013), which was obtained from Addgene. pJBEI-6409 contains six genes of the MEV pathway (atoB from E. coli, HMGS, and HMGR from Staphylococcus aureus and MK, PMK, and PMD from Saccharomyces cerevisiae, idi from E. coli, GPPS from A. grandis, and limonene synthase gene (LS) from Mentha spicata). The GPPS-LS gene cluster was removed from pJBEI-6409 to obtain pMEVI. The fusion gene cluster of the codon-optimized GPPS and PS from A. grandis with a (GSG)<sup>2</sup> linker was synthesized by Suzhou GENEWIZ, Inc. (Suzhou, China) and ligated into pQE30 to obtain pQE-GPPS-L-PS. The GPPS-PS gene cluster from pQE-GPPS-L-PS was inserted into the BamHI/XhoI sites of pMEVI to obtain pMEVIGPS. The evolved codon-optimized Pt1 (Pt1Q457L) from P. taeda was synthesized by Suzhou GENEWIZ, Inc. (Suzhou, China) and ligated into pQE30 to obtain pQE-Pt1Q457L. The PS gene of pQE-GPPS-L-PSDNA shuffling was replaced with the Pt1Q457L gene to obtain pQE-GPPSMUT-L-Pt1Q457L .

The acrB and acrAB were amplified from E. coli and inserted into pZEABP to obtain pZEA-acrB and pZEA-acrAB, respectively. The P. putida KT2440 ttgB, P. putida KT244 mexF, and A. borkumensis acrBDFa were amplified from pBbA5K-EPL11, pBbA5K-EPL14, and pBbA5K-EPL95 and inserted into pZEABP to obtain pZEA-ttgB, pZEA-mexF, and pZEA-acrBDFa, respectively.

The TIGR-mediated GPPSMUT-Pt1Q457L gene cluster was cut from pQE-GPPSMUT-TIGR-Pt1Q457L with EcoRI/HindIII and then cloned into EcoRI/HindIII-digested pHKKF3T5b to obtain pHKKF3T5b-GPPSMUT-TIGR-Pt1Q457L. The MEVI operon was cut from pMEVI with EcoRI/XhoI and then cloned into EcoRI/SalI-digested pP21KF3T5b to obtain pP21KF3T5b-MEVI. Chromosomal integration was carried out by direct transformation as described by Chen et al. (2013). Chemically induced chromosomal evolution (CIChE) of the above construct was carried out by subculturing the resulting strains in 5 mL Super Optimal Broth (SOB) medium with increasing concentrations of triclosan in 15 mL culture tubes, as described by Chen et al. (2013). The strains were grown to the stationary phase in 1µM triclosan for pP21KF3T5b-GPPSMUT-TIGR-Pt1Q457L or 0.25µM for pHKKF3T5b-MEVI. Fifty milliliters of the culture were subcultured into a new culture tube, in TABLE 1 | Strains and plasmids used in this study.


which the triclosan concentration was doubled to 132 or 32µM and allowed to grow to the stationary phase. The process was repeated until the desired concentration was reached. The recA gene of the CIChE strain was then deleted by the markerless deletion approach using the isopropyl β-D-1-thiogalactopyranoside (IPTG)-inducible ccdB as a counterselectable marker (Wei et al., 2016).

Gene replacement of the native promoter of E. coli acrAB and the integration of ttgB from P. putida KT2440 were carried out by the CRISPR-Cas method as described by Jiang et al. (2015). To enhance specificity and reduce off-target effects, the cas9 on pCas (Jiang et al., 2015) was site-directed mutated into cas9(K848A/K1003A/R1060A) as described as Slaymaker et al. (2016) to obtain pCas<sup>∗</sup> . To easily assemble the sgRNA sequence using the BglBrick standard method, the BglII site in the sgRNA plasmid pTargetF was first removed, and then a BglII site was added in the front of EcoRI site to obtain the sgRNA plasmid pTargetB.

### Adaptive Laboratory Evolution for Improving Pinene Tolerance

A 1-mL culture of logarithmic phase E. coli was collected by centrifugation, washed twice with saline, and diluted to a cell concentration of 10<sup>6</sup> to 10<sup>7</sup> with physiological saline. Then, atmospheric and room temperature plasma (ARTP) mutagenesis was performed using an ARTP mutation system (ARTP-IIS, Tmaxtree Biotechnology Co, Ltd, Wuxi, China) with the following parameters: (1) the radio frequency power input was 100 W; (2) the flow of pure helium was 10 standard liters per min; (3) the distance between the plasma torch nozzle exit and the slide was 2 mm; and (4) the different treatment times were selected (10, 20, 40, 60, 80, 100, and 120 s). Ten microliters of the aforementioned cell dilution were evenly scattered on the slide and subjected to ARTP mutagenesis. After treatment, the slide was washed with LB medium (10 g/L tryptone, 5 g/L yeast extract, 10 g/L NaCl), transferred to 5 mL of LB medium with 0.5% pinene in a 15 mL falcon tube, and cultivated at 30 ◦C and 200 rpm for 24 h. The cultures were serially passed into fresh medium (initial OD<sup>600</sup> of 0.2) daily. Continuously repeating this transfer procedure at 0.5% pinene until OD<sup>600</sup> at 24 h did not increase further, the culture was then sequentially transferred to a pinene concentration of 1.0%, 1.5% and 2.0%. The cultures were frozen and stored at −80◦C at every pinene concentration.

The cultures of 2.0% pinene stored at −80◦C were transferred by the IPP/FPP sensor plasmid pPrstA-GFP. Single colonies were inoculated in individual wells of a 48 deep-well microplate (4.6 mL) containing 600 µL of LB medium and incubated at 30◦C and 200 rpm for 24 h on a Multitron shaker (Infors). The cells were harvested by centrifugation at 14000 × g for 2 min and then resuspended with 0.6 mL (137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 2 mM KH2PO4, pH 7.4). Then, 200 µL of the bacterial culture was transferred into a 96-well plate in which the OD<sup>600</sup> and fluorescence were read with the excitation at 485 nm and emission at 528 nm using a SynergyNeo2 multi-mode reader (SynergyNeo2, BioTek, USA).

# Generating Random Mutagenesis Libraries Using Error-Prone PCR and Screening

The random mutagenesis libraries of the fusion gene cluster of AgGPPS-AgPS after optimization of the first 18 codons using the 6AA method (Boë et al., 2016) were constructed through errorprone PCR. The gene cluster of AgGPPS-AgPS was amplified from pQE-GPPS6AA-L-PS using the primers EcoRI-GPPS/HindIII-PS. The error-prone PCR reaction mixture consisted of 5 mM MgCl2, 0.3 mM MnCl2, 0.2 mM each of dATP and dGTP, 1 mM each of dCTP and dTTP and Tag DNA polymerase. The PCR product was digested by EcoRI/HindIII, ligated into the EcoRI/HindIII sites of pQE30, and then transferred into the lycopene-producing strain E. coli LYCOP to generate the mutant library.

The mutant library was plated on LB agar with ampicillin and IPTG. The plates were incubated at 30◦C overnight. The mutant plasmid was isolated from the whiter colony and then transferred into component E. coli BW25113 (PT5-dxs, pMEVI). The pinene productions of them were analyzed in a shake flask.

#### Generating Random Mutagenesis Libraries Using DNA Shuffling and Screening

DNA shuffling experiments were performed by the following steps: parental template preparation, DNase I digestion, primerless PCR and PCR with primers. The mutant plasmids from the 7 colonies resulted from error-prone PCR and were used as the template to amplify the gene cluster fragments with the primers EcoRI-GPPS/HindIII-PS. Following purification, 2 µg of the eight PCR products was mixed and treated with 0.02 U of DNaseI in 100 µL of the 10 × DNaseI buffer on ice for 2 min and terminated by the loading buffer containing SDS. The purified fragments of 50–300 bp were used in the primer-less PCR reactions to reassemble into full-length genes. The primerless PCR reaction mixture contained 0.5 mM each dNTP, 10 × Taq buffer and 0.5 µL Taq DNA polymerase (Takara). The PCR reaction conditions were as follows: 95◦C for 1 min, 35 cycles of 94◦C for 30 s, 45◦C for 30 s, 72◦C for 3 min, and final incubation at 72◦C for 8 min. The PCR products with the correct size were purified and subjected to PCR amplification using the same conditions with the primers EcoRI-GPPS/HindIII-PS. Finally, the mutated PCR products of the full-length gene were digested by EcoRI/HindIII, ligated into the EcoRI/HindIII sites of pQE30, and transferred into the lycopene-producing strain E. coli LYCOP to generate the mutant library.

The mutant library was plated on LB agar with ampicillin and IPTG. The plates were incubated at 30◦C overnight. Single colonies with a whiter color were inoculated in individual wells of a 48 deep-well microplate (4.6 mL) containing 1 mL of LB medium and incubated at 30◦C and 200 rpm on a Multitron shaker (Infors). After 8 h, the cultures were induced with 1 mM IPTG and overlaid with 20% dodecane to trap pinene. After induction, the cultures were incubated at 30◦C and 200 rpm for 48 h. The pinene concentration in individual wells was assayed using the concentrated sulfuric acid method as follows. One hundred microliters of the dodecane layer were mixed with 200µL sulfuric acid, then inoculated for 5 min in

boiling water, and the absorbance of the reaction solution at 450 nm was determined using a spectrophotometer (Shimadzu, Japan).

#### Creating TIGR Libraries and Screening

TIGRs were synthesized using PCR to assemble the oligonucleotides into chimeric DNA sequences as described by Pfleger et al. (2006) and Li et al. (2015). Briefly, 40 mmols of an equimolar oligonucleotide (A, B, C, and D in Supplemental Table 1) mixture was added to a mixture containing 2.5 units of Primer Star DNA Polymerase (Takara, Dalian, China). The assembly was conducted over 35 cycles of PCR for 10 s at 98◦C, 30 s at 72◦C, and 20 + 5 s/cycle at 72◦C. The assembly products were purified using a nucleotide removal column and amplified using the end-specific primers TIGRs-F(X)/TIGRs-R(A) and then cloned into the SacI/SalI sites of pQE-GPPSMUT-Pt1 Q457L to obtain the plasmid libraries pQE-GPPSMUT-TIGRs-Pt1Q457L . The plasmid libraries were transferred into component E. coli BW25113 (PT5-dxs, pMEVI) to generate the mutant library.

The TIGR library was plated on LB agar with ampicillin. The plates were incubated at 30◦C overnight. Single colonies were inoculated in individual wells of a 48 deep-well microplate (4.6 mL) containing 1 mL of LB medium and incubated at 30◦C and 200 rpm on a Multitron shaker (Infors). After 8 h, the cultures were induced with 1 mM IPTG and overlaid with 20% dodecane to trap pinene. After induction, the cultures were incubated at 30◦C and 200 rpm for 48 h. The pinene concentration in individual wells was assayed using the above concentrated sulfuric acid method.

#### Pinene Biosynthesis in Shake Flasks

For pinene fermentation production, a single colony was inoculated into 5 mL of LB medium in a falcon tube, which was cultured overnight at 37◦C. The overnight seed culture was then inoculated into 50 mL of SBMSN medium with a starting OD<sup>600</sup> of 0.1. SBMSN medium (pH 7.0) containing the following (g/L): sucrose 20, peptone 12, yeast extract 24, KH2PO<sup>4</sup> 1.7, K2HPO<sup>4</sup> 211.42, MgCl2·6H2O 1, ammonium oxalate 1.42, and Tween-80 2. The main cultures were then incubated at 37◦C and 200 rpm until an OD<sup>600</sup> of 0.8 was reached. Then, the cultures were induced with 1 mM IPTG and overlaid with 20% dodecane to trap pinene. After induction, the cultures were incubated at 30◦C and 130 rpm for 72 h.

#### Co-culture of *E. coli* PINE and MEVI for Pinene Production

E. coli PINE and MEVI cells were first separately grown in 5 mL SBMSN medium in a falcon tube at 37◦C overnight. The overnight culture was inoculated into 50 mL of SBMSN medium with a starting OD<sup>600</sup> of 0.1 and incubated at 37◦C and 200 rpm until an OD<sup>600</sup> of approximately 6.0 was reached. The cultures were then incubated at 20◦C and 200 rpm for 16 h. For pinene biosynthesis using co-cultures, the E. coli PINE culture and the desired amount of the E. coli MEVI culture were inoculated into the 30 mL SBMSN medium with a starting OD<sup>600</sup> of 0.1. The mixed culture was culture at 37◦C and 200 rpm until an OD<sup>600</sup> of 0.8 was reached. Then, the cultures were overlaid with 20% dodecane to trap pinene, and were incubated at 30◦C and 130 rpm for 72 h.

## Whole-Cell Biocatalysis for Pinene Production

A single colony of E. coli PINE and MEVI was separately inoculated into 5 mL of SBMSN medium in a falcon tube, which was cultured overnight at 37◦C. The overnight cultures were then inoculated into 50 mL SBMSN medium with a starting OD<sup>600</sup> of 0.1. The cultures were then incubated at 37◦C and 200 rpm until an OD<sup>600</sup> of approximately 6.0 was reached. Then, the cultures were incubated at 20◦C and 200 rpm for 16 h. Finally, the E. coli PINE culture was mixed with the E. coli MEVI culture at the inoculation ratio of 2:1. The mixed cells were harvested by centrifugation (6000 × g at 4◦C) and washed twice with cooled phosphate buffer (0.1 M, pH 7.0).

For biocatalysis, the above cells were resuspended in 10 mL phosphate buffer (0.1 M, pH 7.0) containing 20 g/L of sucrose, 10 mM MgCl<sup>2</sup> and 5 mM MnCl<sup>2</sup> to form the cell suspension (OD<sup>600</sup> = 30). The reaction mixture was overlaid with 20% dodecane. The catalysis was performed for 28 h at 30◦C and 130 rpm.

## GC Analysis

Five hundred microliters of the dodecane layer was placed in a 1.5-mL microcentrifuge tube and centrifuged at 25,000 × g for 1 min, and 50 µL of dodecane was diluted in 450 µL of ethyl acetate spiked with the internal standard limonene (10 µg/L). The samples were analyzed by GC-FID by using a standard curve of α-pinene (Sigma Aldrich). The GC-FID (Techcomp GC7900, Techcomp Ltd, China) was used with a TM-5 column (30 m × 0.32 mm × 0.50µm). The inlet temperature was set to 300◦C, with the flow at 1 mL/min, the oven at 50◦C for 30 s, ramp at 4 ◦C/min to 70◦C, and ramp at 25◦C/min to 240◦C.

### Quantitative Real-Time PCR (qRT-PCR)

The total RNA from E. coli cells grown for 24 h in shake flasks was isolated using an RNA extraction kit (Dongsheng Biotech, Guangzhou, China), following the manufacturer's instructions. The first-strand cDNA was synthesized using an All-in-OneTM First-Strand cDNA Synthesis kit (GeneCopoeia, Guangzhou, China). The qRT-PCR was perfor1med with the All-in-OneTM qPCR Mix kit (GeneCopoeia) on an iCycler iQ5 Real Time PCR system (Bio-Rad Laboratories, California, USA). The template was 100 ng of cDNA. The PCR conditions were as follows: 95◦C for 10 min, followed by 45 cycles of denaturation at 95◦C for 10 s, annealing at 60◦C for 20 s, and extension at 72◦C for 15 s. The primers for qRT-PCR are presented in Supplementary Table 1. The data were analyzed by the 2−11Ct method described by Livak and Schmittgen (2001) and normalized by cysG gene expression.

Gene copy numbers were measured by qPCR on genomic DNA isolated from the appropriate CIChE strains. qPCR was performed as described above. The primers QPt1F/QPt1R and QHF/QHR (Supplementary Table 1) were used to measure the copy number of Pt1 and HMGS, respectively.

# Statistical Analysis

All experiments were conducted in triplicate, and the data were averaged and presented as the means ± standard deviation. Oneway analysis of variance followed by Tukey's test was used to determine significant differences using the OriginPro (version 7.5) package. Statistical significance was defined as p < 0.05.

# RESULTS

### Tolerance Engineering to Improve Pinene Production

To improve pinene tolerance, E. coli cells harboring pPrstA-GFP were treated with ARTP and then serially transferred into LB medium supplemented with increased concentrations of pinene of 0.5, 1.0, 1.5, and 2.0%. The culture was transferred daily. After the adaptive evolution at 2.0% pinene, the culture was streaked on LB plates for isolated colonies. It has been demonstrated that the IPP/FPP sensor plasmid pPrstA-GFP has been successfully used to test the intracellular IPP/FPP concentration and to screen the library with higher IPP/FPP concentrations (Dahl et al., 2013; Shen et al., 2016). Thus, we also used it to screen the library. Of the 670 clones, 14 strains with higher fluorescence strength (Supplementary Figure 1) were selected for further shake flask analysis. As shown in **Figure 1**, E. coli YZ-3 produced the highest level of pinene (7.3 ± 0.2 mg/L), which was 31% higher than the starting strain E. coli BW25113 (PT5-dxs).

To improve pinene production, we investigated effects of efflux pumps on pinene production. Dunlop et al. (2011)reported that expressing some efflux pumps significantly improved pinene tolerance. Thus, we tested whether pumps that improved pinene tolerance also enhanced its production. As shown in **Figure 2A**, expressing native AcrB, AcrAB, or TtgB (NP\_743544) from Pseudomonas putida KT2440 in E. coli YZ-3 using plasmid

resulted in increased pinene production. However, expressing A. borkumensis AcrBDFa (YP\_692684) or P. putida KT2440 MexF (NP\_745564) from did not improve pinene production. Therefore, we first replaced the native promoter of E. coli YZ-3 acrAB operon with the strong P37 promoter to obtain E. coli YZ-3-A, resulting in an increase in pinene production to 8.1 ± 0.2 mg/L from 7.3 ± 0.2 mg/L (**Figure 2B**). Then, we integrated the ttgB from P. putida KT2440 under the control of the P37 promoter in E. coli YZ-3-A to obtain E. coli YZ-3-A-T. The modification further improved pinene production to 9.1 ± 0.2 mg/L (**Figure 2B**). These results indicate that overexpressing some efflux pumps (E. coli acrAB and Pseudomonas putida KT2440 ttgB), whcih improved pinene tolerance, also enhanced its production.

To further improve pinene production, we isolated a mutant resistant to an inhibitor of biosynthetic pathway after ARTP mutagenesis. Isolating a mutant resistant to an inhibitor of biosysnthetic pathway is a common strategy used for strain improvement. In E. coli, the important precursors IPP and DMAPP are produced by the 1-deoxy-D-xylulose-5-phosphate (DXP) pathway. Fosmidomycin is the DXP pathway inhibitor that inhibits 1-deoxy-D-xylulose-5-phosphate reductoisomerase (Dxr) and methylerythritol phosphate cytidyltransferase (IspD) of the DXP pathway (Zhang et al., 2011). Genes involved in the DXP pathway are essential for E. coli growth. The wildtype E. coli YZ-3-A-T can grow in the presence of 2% pinene (Supplementary Figure 2A), but does not grow in the presence of 35µM fosmidomycin (Supplementary Figure 2B). After ARTP mutagenesis, cells grow well in the presence of 35µM fosmidomycin (Supplementary Figure 2C). Overexpression of dxr or ispD in E. coli improved the fosmidomycin tolerance (Zhang et al., 2011). This indicates that the fosmidomycin resistant mutants may show higher level of Dxr and IspD. Screening the fosmidomycin resistant mutants will increase the probability to obtain a mutant with higher IPP flux. Thus, to increase the probability to obtain a mutant with higher IPP flux, we screened the fosmidomycin resistant mutants using the IPP/FPP sensor. E. coli YZ-3-A-T cells harboring pPrstA-GFP were treated with ARTP. After ARTP mutagenesis, the cells were transferred into the LB medium supplemented with 35µM fosmidomycin and 2.0% pinene. A total of 720 clones were screened for analyzing fluorescence strength in deepwell microplate cultures (Supplementary Figure 3). Twenty-one strains with higher fluorescence strength were selected for further shake flask analysis. As shown in **Figure 3**, Strain No. 19, which was denoted as E. coli YZFP, produced the highest level of pinene, which reached 9.9 ± 0.1 mg/L. In fact, our quantitative real-time PCR analysis also demonstrates that the dxs, dxr and ispD of the DXP pathway in E. coli YZFP showed higher transcription level than the wild-type strain (Data not shown, will be published in another paper).

To characterize the pinene tolerance, the growth of the above strains were compared in different concentrations of pinene. **Figure 4A** shows the growths of these strains in the presence of 2% pinene. The starting strain did not grow well in the presence of 2% pinene. The above engineered strains did grow well in the presence of 2% pinene. The maximum cell densities of the

three engineered strains were similar. The growth rate of E. coli YZFP was higher than that of the other engineered strains. These results indicate that the engineered strains have higher pinene tolerance than the starting strain. However, the maximum cell densities of the three engineered strains were lower than that of the starting strain in the absence of pinene (**Figure 4B**). The reason may be that the three engineered strains produced higher level of IPP than the starting strain. IPP is toxicity to E. coli. We also investigated the genetic stability of E. coli YZFP. The strain can also grow well in the presence of 2% pinene and the level of pinene production remained constant after 20 rounds of subculturing in absence of selective pressure (data not shown).

## Evolution Engineering to Improve Pinene Production

The lower expression level and/or lower enzymatic activity of GPPS and PS in E. coli may result in the lower yield of pinene production. Sarria et al. (2014) compared GPPSs and PSs from A. grandis and P. taeda and found that the combination of GPPS and PS from A. grandis was most suitable for pinene production. Thus, we first optimized the first 48 nucleotide sequences of A. grandis GPPS with the 6AA method to increase the expression level of the A. grandis GPPS-PS gene cluster in E. coli. The 6AA method substitutes all Arg, Asp, Gln, Glu, His, and Ile codons with the synonymous codon having the highest single-variable logistic regression slope (CGT, GAT, CAA, GAA, CAT, and ATT, respectively), while the other 14 amino acids were not changed from the wild-type gene sequence (Boë et al., 2016). The 6AA optimization increased pinene production from 5.6 ± 0.1 mg/L to 6.4 ± 0.3 mg/L (**Table 2**).

Because pinene shares the same 5-carbon precursors IPP and DMAPP with carotenoids, a lycopene-producing strain E. coli LYCOP (Chen et al., 2013) was used to screen the error-prone


Data represent the means of three replicates and standard deviations.

PCR mutant libraries of the GPPS-PS cluster from A. grandis after 6AA optimization. The higher the activity of the GPPS-PS cluster, the lower the intracellular precursor levels for lycopene biosynthesis, thereby reducing the pigmentation of the E. coli. Of approximately 1 500 colonies, 7 colonies with a whiter color were observed. The mutant plasmids were isolated from the 7 colonies and then were co-transferred with the MEV pathway plasmid pMEVI into E. coli BW25113 (PT5-dxs). The pinene productions of them were analyzed in a shake flask, and the results are presented in **Figure 5A**. The strains harboring the mutant gene cluster produced higher pinene by 7.8–85.7% than that with the wild-type gene cluster. To increase the gene cluster activity, the 7 mutant gene clusters were used for DNA shuffling.

Because the colonies of E. coli LYCOP harboring the above mutant gene cluster became a faint color, it is difficult to discriminate these colonies by using the above carotenoid-based method. A more sensitive and quantitative screening method is needed. It is known that monoterpene can hydrate readily in the presence of acid catalysts, such as H2SO<sup>4</sup> (Robles-Dutenhefner et al., 2001). As a result, the initial reaction solutions turn yellow and then brown. After reaction with concentrated sulfuric acid in boiling water for 5 min, it was observed that the color of the reaction solution become darker as the pinene concentration increases and the absorbance at 450 nm is linearly related with pinene concentration (Supplementary Figure 4). Thus, the concentrated sulfuric acid method can quantitatively predict pinene concentrations.

After DNA shuffling, the mutant plasmids were transferred into E. coli LYCOP harboring pMEVI. Fifty colonies with a whiter color were used for assays of pinene production in a shake flask using the concentrated sulfuric acid method. The results are presented in **Figure 5B**. E. coli LYCOP harboring the mutant gene cluster produced higher pinene (6.5–10.1 mg/L) than those

with the wild-type gene cluster. The mutant plasmid with the highest pinene production was isolated from strain No. 15 and then was co-transferred with pMEVI into E. coli BW25113 (PT5 dxs). E. coli BW25113 (PT5-dxs) harboring the mutant plasmid and pMEVI produced 12.4 ± 0.2 mg/L of pinene (**Table 2**). The mutant plasmid with the highest pinene production was then sequenced. The two amino acid mutants (D90G and L175P) were observed in the CDS of GPPS from A. grandis. No mutant in the CDS of PS from A. grandis was observed. It has been reported that (-)-α-pinene synthase (Pt1) from P. taeda has the lowest K<sup>m</sup> for GPP among the known PSs (Phillips et al., 2003). Tashiro et al. engineered an E. coli with the highest yield of pinene so far using the pinene synthase mutant (Pt1Q457L) from P. taeda (Tashiro et al., 2016). Thus, replacing A. grandis PS in the mutant plasmid with P. taeda pinene synthase mutant gene (Pt1Q457L) yielded pQE30-GPPSmut-L-Pt1Q457L . E. coli BW25113 (PT5-dxs) harboring pQE30-GPPSmut-L-Pt1Q457L produced a higher level of pinene (15.2 ± 0.2 mg/L) than those harboring pQE30- GPPSmut-L-AgPS, which achieved 12.4 ± 0.2 mg/L (**Table 2**).

The unbalanced expression of multiple genes may overburden the cell and cause accumulation of toxic metabolic intermediates, resulting in reduced product titers. Pfleger et al. (2006) developed a combinatorial engineering approach for coordinating the expression of cascade enzymes. For this purpose, libraries of tunable intergenic regions (TIGRs) are generated that encode mRNAs with diverse secondary structures with RNase cleavage sites. The TIGR approach was applied to balance the gene expression of the MEV pathway using the TIGR approach, resulting in a 7-fold increase in mevalonate production. Moreover, our previous paper demonstrated that the TIGR approach was more efficient compared to protein fusion for coordinating expression (Li et al., 2015). Thus, we constructed a library of TIGRs to balance the expression of A. grandis GPPSD90G/L175P and P. taeda Pt1Q457L. The library of TIGRS was inserted between GPPSD90G/L175P and P. taeda Pt1Q457L to yield a series of operons. The functional operons from the libraries were screened by using the concentrated sulfuric acid method. A total of 768 colonies were used for the assay of pinene production in deep-well microplate cultures using the concentrated sulfuric acid method (Supplementary Figure 5). Forty-three strains with higher OD<sup>450</sup> were selected for further shake flask analysis. As shown in **Figure 6**, strain No. 6 produced the highest level of pinene (17.6 ± 0.2 mg/L). Thus, the TIGR-mediated plasmid was recovered from strain No. 6 and sequenced (Supplementary Table 2). We also retransformed the plasmid back to the host strain E. coli BW25113 (PT5-dxs) and checked the pinene production. The resulting strain produced the same level of pinene (17.9 ± 0.1 mg/L), indicating that the pinene production improvement is the result of TIGR-mediated optimization.

#### Modular Co-culture Engineering to Improve Pinene Production

To take advantage of emerging co-culture engineering approaches to improve overall pinene biosynthesis in E. coli, the complete biosynthetic pathway was divided into the following two modules: the upstream module of the MEV pathway and the downstream module of the TIGR-mediated gene cluster of A. grandis GPPSMut and P. taeda Pt1MUT (**Figure 7**). The two

modules were integrated into the chromosome of the pinene tolerance strain E. coli YZFP and then then evolved to a higher gene copy number by triclosan induction, respectively.

**Figure 8A** shows the results of pinene production in CIChE strains of the TIGR-mediated gene cluster of A. grandis GPPSMut and P. taeda Pt1MUT without the MEV pathway. The maximum pinene production was obtained by the CIChE strains resistant to 32µM triclosan. Thus, the recA gene of the CIChE strain resistant to 32µM triclosan was deleted to obtain E. coli PINE. We determined the GPPS-Pt1 gene copy number in E. coli PINE. The copy number reached approximately 60 in the CIChE strain, which is the equivalent copy number of a high copy plasmid. **Figure 8B** shows the results of IPP/FPP concentration of the CIChE strains of the MEV pathway measured by the IPP/FPP sensor (pPrstA-GFP). As shown in **Figure 8B**, the maximum IPP/FPP production was obtained by the CIChE strains resistant to 0.5µM triclosan. Thus, the recA gene of the CIChE strain resistant to 0.5µM triclosan was deleted to obtain E. coli MEVI. We also determined the MEV pathway gene copy number in E. coli MEVI. The copy number reached approximately 4 in E. coli MEVI.

Zhou et al. (2015) demonstrate that the modular co-culture engineering can be applicable to isoprenoids because their scaffold moleculars can generally permeate membranes. To demonstrate IPP can also cross cell membranes, we cultured E. coli (pPrstA-GFP) with the cell-free culture broth of E. coli MEVI and measured fluorescence strength. After addition of the cell-free culture broth of E. coli MEVI, E. coli (pPrstA-GFP) showed higher fluorescence strength (Supplementary Figure 6). Moreover, the E. coli MEVI: PINE co-culture produced higher level of pinene than E. coli PINE (**Figure 9**). These results indicate that IPP produced by E. coli MEVI diffused into E. coli PINE and was subsequently converted into pinene. We then optimized the E. coli MEVI: PINE co-culture system to further improve pinene production. To this end, different inoculation ratios between E. coli MEVI and PINE were investigated. As shown in **Figure 9**, the highest pinene production of 64.9 ± 0.9 mg/L was achieved when E. coli MEVI and PINE were inoculated at a ratio of 1:2. Compared with the mono-culture strategy using E. coli PINE harboring pMEVI, the pinene production was increased by 1.9 fold (from 22.3 ± 0.2 mg/L to 64.9 ± 0.9 mg/L). To test if all of IPP produced by E. coli MEVI were converted into pinene by E. coli PINE, we measured the IPP concentration in the broth of the co-culture and the E. coli MEVI after 28 h using the IPP sensor plasmid. The results showed that about 57.8% of IPP were converted by E. coli PINE (Supplementary Table 3). Thus, we introduced pQE-GPPSMUT-TIGR-Pt1Q457L to overexpress the pinene biosynthetic pathway and checked the pinene production. The co-culture system after introducing the pinene biosynthetic pathway into E. coli MEVI produced higher level of pinene (60.2 ± 0.2 mg/L) than the control strain (52.1 ± 0.1 mg/L) harboring the empty plasmid (Supplementary Table 4). The result also demonstrates that not all of IPP can be converted in the co-culture system.

Biotechnological approaches for chemicals production can be broadly classified into fermentation and biocatalysis. In biocatalysis, cell growth and production phase are separated. In comparison to the fermentation bioprocess, whole-cell biocatalysis is an attractive method due to its great efficiency and relative simplification of downstream processing (Lin and Tao, 2017). The whole-cell biocatalysis processes comprise the following two stages: growth and conversion of the substrates. After the cells are cultured, they are harvested and washes with a buffer solution and suspended in the buffer for biocatalysis. Thus, the E. coli-E. coli modular co-culture system of whole-cell biocatalysis was used to further enhance pinene production. As shown in **Figure 10**, the highest pinene production of 166.5 ± 0.3 mg/L was achieved by the whole-cell biocatalyst after 28 h. The pinene titer obtained by the whole-cell biocatalysis was 2.6-fold higher than that produced by the fermentation process.

#### DISCUSSION

It has been reported that E. coli growth is inhibited by 0.5% pinene (Dunlop et al., 2011). We first improved pinene tolerance from 0.5 to 2.0% and pinene production by adaptive laboratory evolution after ARTP mutagenesis. In fact, improvements in tolerance are not sufficient to guarantee an increase production. Our results also demonstrate this point. Overexpreesion of A. borkumensis acrBDFa or P. putida KT2440 mexF that improved pinene tolerance did not improved pinene production (**Figure 2A**). To obtain a mutant with higher level of pinene production, we used the IPP/FPP sensor pPrstA-GFP to screen the mutants tolerant to 2% pinene. Tolerance engineering has also successfully been used to improve the production of limonene (Dunlop et al., 2011), amorphadiene (Zhang et al., 2016), olefin (Mingardon et al., 2015), n-octane (Foo and Leong 2013). Although the level of pinene production reported in literatures did not inhibit growth, higher tolerance is beneficial to pinene production. Thus, the 2% pinene tolerant strain E. coli was used the parent strain in this study. To further improve pinene

production, we then expressed the efflux pumps in the pinene tolerant strain E.coli and subsequently selected a mutant resistant to fosmidomycin after ARTP mutagenesis. The pinene tolerant strain E. coli YZFP with higher level of pinene production was obtained through a two-step screening process. There is no directed evidence to prove the improved pinene production is the result of improved pinene tolerance.

Our study demonstrates that the overexpression of some efflux pumps improved pinene tolerance and production. Many groups also reported that overexpression of efflux pumps enhanced biofuel tolerance. Dunlop et al. (2011) reported that the overexpression of efflux pumps, such as A. borkumensis AcrBDFa, P. putida KT2440 MexF, P. putida KT2440 TtgB or E. coli AcrB, enhanced pinene tolerance. However, they did not investigate the effects of the pumps on pinene production. Our results demonstrate that overexpression of E. coli AcrAB and P. putida KT2440 TtgB enhanced pinene production (**Figure 2**). Overexpression of A. borkumensis AcrBDFa or P. putida KT2440 MexF did not improved pinene production (**Figure 2A**). However, Dunlop et al. (2011) reported that overexpression of A. borkumensis AcrBDFa enhanced limonene tolerance and yield. Overexpression of tolC together with ABC family transporters (macAB) or MFS family transporters (emrAB or emrKY) was found to improve amorphadiene titer by more than 3-fold (Zhang et al., 2016). Overexpression of the native and evolved acrB improved olefin tolerance and production (Mingardon et al., 2015). Evolved AcrB variants with improved tolerance to pinene and n-octane have also been reported by Foo and Leong (2013). Taken together with these previous studies, our results show that a combination of the adaptive laboratory evolution with overexpression of some efflux pumps can improve pinene tolerance and production.

In this study, we reported a high-throughput screening method, which is known as the concentrated sulfuric acid method, for recombinant E. coli that overproduce pinene. We successfully applied the concentrated sulfuric acid method to

FIGURE 9 | Effect of the inoculation ratio of E. coli PINE and MEVI on pinene production in the co-culture system. OD600 (Gray bars), Pinene concentration (White bars). 0:1, only E. coli PINE; 0:1 (pMEVI), only E. coli PINE (pMEVI); others, the E. coli MEVI: PINE co-culture system with different inoculation ratio. The data represent the means of three replicates and error bars represent standard deviations.

screen the DNA shuffling library of the GPPS-PS gene cluster and the library of the TIGR-mediated GPPS-Pt1 gene cluster. Because limonene has the same properties as pinene, the concentrated sulfuric acid method can also be used to screen mutants for limonene production. Although the carotenoid-based method has been successfully used to screen isoprene synthase variants (Emmerstorfer-Augustin et al., 2016), the carotenoid-based method has a limitation when the colony has a faint color.

GPPS and PS have been identified as a major limiting factor in pinene production (Yang et al., 2013; Sarria et al., 2014; Tashiro et al., 2016). After directed evolution of the A. grandis GPPS-PS gene cluster using error-prone PCR and DNA shuffling, pinene production was increased by 1.2-fold (**Table 2**). Two amino acid mutants were observed in the CDS of A. grandis GPPS. However, no mutant was observed in the CDS of A. grandis PS. Tashiro et al. evolved P. taeda Pt1 and constructed a recombinant E. coli with the highest pinene yield reported in literatures using the evolved variant (Tashiro et al., 2016). Using the A. grandis GPPSMut - P. taeda Pt1MUT gene cluster resulted in an increase in pinene production by 22.6% compared to using the A. grandis GPPSMut - PS gene cluster (**Table 2**).

GPPS and PS are inhibited by their substrate (GPP) or product (pinene) (Sarria et al., 2014). To overcome GPPS inhibition by GPP, GPPS was fused to PS, resulting in improved pinene production (Sarria et al., 2014; Tashiro et al., 2016). Our previous paper demonstrated that the TIGR approach was more efficient compared to protein fusion for coordinating expression (Li et al., 2015). This study shows that using the TIGR-mediated gene cluster led to an increase in pinene production by 15.8% compared with the fused gene cluster (**Table 2**).

In the present study, an E. coli-E. coli co-culture system was engineered to modularize the MEV and heterologous biosynthetic pathway. The MEV pathway and heterologous biosynthetic pathway (the A. grandis GPPSMut -P. taeda Pt1MUT gene cluster) was engineered in the pinene tolerance strain E. coli YZFP, respectively. The best co-culture system was found to improve pinene production by 1.9-fold compared to the mono-culture system. The modular co-culture can distribute the metabolic burden and allow for modular optimization by simply changing the strain-to-strain ratio. The E. coli-E. coli modular coculture system has been successfully used to improve 3-aminobenzoic acid (Zhang and Stephanopoulos, 2016), flavonoid (Jones et al., 2016), muconic acid (Zhang et al., 2015), and perillyl acetate (Willrodt et al., 2015), etc. In fact, the critical issue for modular co-culture engineering is the mass transfer of the pathway intermediate (IPP). It has been demonstrated that isoprenoids scaffold molecules can cross cell membranes (Zhou et al., 2015). Our results also demonstrate that IPP can cross cell membranes and secreted to the extracellular medium (Supplementary Figure 6 and **Figure 9**). Moreover, our results showed that the pinene tolerance strain E. coli YZFP (pPrstA-GFP) had higher fluorescence strength than the parent strain harboring pPrstA-GFP after addition the cell-free broth of E. coli MEVI (Supplementary Figure 6), indicating that E. coli YZFP shows greater membrane permeability than the parent strain. Our results demonstrate that there were still some IPP not to be converted into pinene by E. coli PINE (Supplementary Tables 3, 4). Moreover, Overexpression of the pinene biosynthetic pathway in E. coli MEVI enhanced pinene production in the E. coli MEVI-E. coli PINE co-culture system (Supplementary Table 4). However, overexpression of the pinene biosynthetic pathway in E. coli PINE did not enhance pinene production (Supplementary Table 4). Increasing the inoculation ratio of E. coli PINE and E. coli MEVI from 2:1 to 2.5:1 or 3:1 did not enhanced pinene production (**Figure 9**). These results indicate that the IPP transportation may be a key factor for further improving pinene production. Transporter engineering strategies have successfully been used to enhance the secretion of the pathway intermediates, improving production (Boyarskiy and Tullman-Ercek, 2015; Kell et al., 2015; Zhang et al., 2015). Thus, appropriate metabolite transporters engineering strategies may be used to further improve pinene production of the E. coli-E. coli co-culture system.

This study also demonstrated that whole-cell biocatalysis further improved pinene production by 1.6-fold compared to the fermentation process. The whole-cell biocatalysis has also successfully been used in many biotechnological production (Tao et al., 2011; Lin et al., 2015; Kogure et al., 2016; Chen et al., 2017; Lin and Tao, 2017). Kogure et al. (2016) also reported that the significantly higher shikimate productivity (141.3 g/L) was achieved by the whole-cell biocatalysis compared to that (78.8 g/L) achieved by the fed-batch fermentation accompanying cell growth. The pinene production improvement may be resulted from higher cell density (OD<sup>600</sup> of 30) and the growth-arrested cells used in the whole-cell biocatalysis.

#### CONCLUSIONS

Pinene tolerance and production were first improved via adaptive laboratory evolution and efflux pump overexpression. Through error-prone PCR and DNA shuffling, a GPPS variant was screened, which outperformed the wild-type enzyme. To balance the expression of multiple genes, a TIGR was inserted between A. grandis GPPSD90G/L175P and P. taeda Pt1Q457L. To construct an E. coli-E. coli co-culture system to modularize the MEV and heterologous biosynthetic pathway, the MEV pathway and

#### REFERENCES


heterologous biosynthetic pathway (the A. grandis GPPSMut -P. taeda Pt1MUT gene cluster) was integrated into the chromosome of the pinene tolerance strain E. coli YZFP and then evolved to a higher gene copy number by CIChE, respectively. The E. coli-E. coli modular co-culture system of whole-cell biocatalysis resulted in the highest pinene production of 166.5 mg/L. Our results demonstrate that the E. coli-E. coli modular co-culture system of the whole-cell biocatalysis is a promising approach for the production of pinene.

#### AUTHOR CONTRIBUTIONS

F-XN performed all of the experimental works. XH and Y-QW performed the pinene assay. J-ZL designed the study and wrote the manuscript. All the authors read and approved the final manuscript.

#### FUNDING

This work was funded by the National Natural Science Foundation of China (Grant NO. 21276289), the Natural Science Foundation of Guangdong Province (NO. 2015A030311036), the Project of the Scientific and Technical Program of Guangdong Province (NO. 2015A010107004) and the Project of the Scientific and Technical Program of Guangzhou (NO. 201607010028).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.01623/full#supplementary-material


**Conflict of Interest Statement:** 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 © 2018 Niu, He, Wu and Liu. 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.

# Heterologous Production of Microbial Ribosomally Synthesized and Post-translationally Modified Peptides

#### Yi Zhang<sup>1</sup> , Manyun Chen<sup>1</sup> , Steven D. Bruner <sup>2</sup> and Yousong Ding<sup>1</sup> \*

<sup>1</sup> Department of Medicinal Chemistry, Center for Natural Products, Drug Discovery and Development, College of Pharmacy, University of Florida, Gainesville, FL, United States, <sup>2</sup> Department of Chemistry, University of Florida, Gainesville, FL, United States

Ribosomally synthesized and post-translationally modified peptides, or RiPPs, which have mainly isolated from microbes as well as plants and animals, are an ever-expanding group of peptidic natural products with diverse chemical structures and biological activities. They have emerged as a major category of secondary metabolites partly due to a myriad of microbial genome sequencing endeavors and the availability of genome mining software in the past two decades. Heterologous expression of RiPP gene clusters mined from microbial genomes, which are often silent in native producers, in surrogate hosts such as Escherichia coli and Streptomyces strains can be an effective way to elucidate encoded peptides and produce novel derivatives. Emerging strategies have been developed to facilitate the success of the heterologous expression by targeting multiple synthetic biology levels, including individual proteins, pathways, metabolic flux and hosts. This review describes recent advances in heterologous production of RiPPs, mainly from microbes, with a focus on E. coli and Streptomyces strains as the surrogate hosts.

Keywords: RiPPs, heterologous expression, precursor peptide, processing enzymes, synthetic biology, *E. coli*, *Streptomyces*

# INTRODUCTION

Ribosomally synthesized and post-translationally modified peptides (RiPPs) are a large group of natural products with a high degree of structural diversity and a wide variety of bioactivities (**Figure 1A**; Arnison et al., 2013). So far, over 20 different families of RiPPs have been discovered, each carrying unique chemical features (Ortega and van Der Donk, 2016). A biosynthetic logic for RiPPs has emerged and can be simplified as the post-translational modification (PTM) of ribosomally synthesized precursor peptides (**Figure 1B**; Arnison et al., 2013). An ever-growing list of PTMs expand chemical functionality and often impart metabolic and chemical stability upon precursor peptides. One precursor peptide usually contains the leader peptide (in rare cases Cterminal, named as follower peptide) N-terminal to the core peptide. The leader peptide binds to and guides biosynthetic enzymes for PTMs on the core peptide and is eventually removed from the modified core peptides by proteases. The entire sequence of the core peptide is generally retained in the final structures of RiPPs and can carry multiple variable sites. As such, the separation of substrate recognition and catalysis enables a concise RiPP biosynthetic route, possessing an evolutionary advantage of accessing high chemical diversity at low genetic cost.

*Edited by:*

Dipesh Dhakal, Sun Moon University, South Korea

#### *Reviewed by:*

Johannes Koehbach, The University of Queensland, Australia Christian W. Gruber, Medizinische Universität Wien, Austria

> *\*Correspondence:* Yousong Ding yding@cop.ufl.edu

#### *Specialty section:*

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

> *Received:* 07 June 2018 *Accepted:* 17 July 2018 *Published:* 07 August 2018

#### *Citation:*

Zhang Y, Chen M, Bruner SD and Ding Y (2018) Heterologous Production of Microbial Ribosomally Synthesized and Post-translationally Modified Peptides. Front. Microbiol. 9:1801. doi: 10.3389/fmicb.2018.01801

As a consequence of their ribosomal origin, the chemical structures of RiPPs are more predictable from genomic data than other families of natural products, making RiPPs an attractive target of genome-driven natural product discovery efforts. Compared to conventional "top-down" approaches, the starting point of the genome-driven approach is genome sequences that have exponentially grown over the past decade. Many specialized bioinformatic tools have been developed for identifying RiPPs biosynthetic gene clusters, such as AntiSMASH (Weber et al., 2015), PRISM (Skinnider et al., 2017), SMURF (Khaldi et al., 2010), and more recently RODEO (Tietz et al., 2017). However, there are many technical challenges to translate the identified clusters into chemical entities, rendering the genome-driven approach far from being a panacea for accessing the chemical space that natural products occupy (Luo et al., 2014). Indeed, the diversity and complexity of PTMs, which are often essential for bioactivity of RiPPs, are not readily identifiable on the core peptides as our understanding of biosynthetic enzymes, particularly their substrate specificity and regio-, stereo-, and chemo-selectivity, remains limited (Arnison et al., 2013). On the other hand, the structural determination of RiPPs is often challenged with their no-to-low isolation yields from samples collected from the field or cultured under laboratory conditions (Smith et al., 2018). Over the past decade, many approaches have been developed to address this critical, major issue of the genome-driven approach, including the activation of silent biosynthetic gene clusters (e.g., modification of fermentation methods and engineering of original producers), heterologous expression using a genetically tractable surrogate host, and in vitro reconstruction (Chiang et al., 2011; Abdelmohsen et al., 2015; Reen et al., 2015; Ren et al., 2017). Among them, heterologous expression of RiPPs in surrogate hosts, commonly Escherichia coli and Streptomyces strains, has so far been one of the most successful methods to elucidate cryptic gene clusters and discover new RiPPs (Ortega and van Der Donk, 2016). Furthermore, heterologous production can effectively harvest the promiscuity of RiPP biosynthetic systems to produce designed analogs through genetic engineering of precursor peptides. Importantly, many emerging strategies have been developed to improve the success of heterologous production of RiPPs over the past several years, mainly focusing on the manipulation of individual proteins, pathways, metabolic flux and hosts (**Figure 2**). Herein, this review describes the details of these strategies ensuring and expanding the heterologous expression approach to discover and develop RiPPs. Representative examples of heterologous expression of each major family of RiPPs were summarized in **Table 1**. Of note, thousands of antimicrobial peptides have been isolated from a variety of organisms (Deng et al., 2017), and this manuscript excluded their heterologous production in discussions.

#### MANIPULATION OF COMPONENTS OF RiPP BIOSYNTHETIC PATHWAYS

A RiPP gene cluster commonly comprises of all essential genes for the production of RiPP. Manipulation of the pathway-specific components allows precise and rational improvement of RiPP production and minimizes potential perturbation of the holistic metabolism of the heterologous host. Detailed information regarding the function, timing, specificity and regulation on the pathway can also be extracted via this approach. From a synthetic biology standpoint, here we use representative examples to describe different strategies used to manipulate RiPP biosynthetic pathways for successful heterologous expression.

# Promoter Engineering to Control Gene Transcription

Altered transcription levels of biosynthetic genes are commonly observed when they are introduced into heterologous hosts. Genetic engineering of a biosynthetic gene cluster by the introduction of one or more constitutive or inducible promoters has proved very effective for the heterologous production of different RiPP families. Importantly, a number of wellcharacterized promoters of commonly used hosts (e.g., E. coli and Streptomyces strains) (De Mey et al., 2007; Li et al., 2015; Myronovskyi and Luzhetskyy, 2016) have been available to enable this synthetic biology approach. For example, lichenicidin is a two-component lantibiotic produced by Bacillus licheniformis I89, and its heterologous production from the native gene cluster in E. coli BLic5 led to a significantly lowered yield compared with the native producer (**Table 1**; Caetano et al., 2011a,b). By contrast, driving the expression of each biosynthetic gene by a strong T7 promoter resulted in a yield of lichenicidin up to 100 times higher than B. licheniformis I89 (Kuthning et al., 2015). In another example, Staphylococcus warneri ISK-1 produces a lantibiotic nukacin ISK-1 (Sashihara et al., 2000) but the heterologous expression of its gene cluster in S. carnosus TM300 and Lactobacillus plantarum ATCC 14917<sup>T</sup> failed to produce any natural product (Aso et al., 2004). Aso et al. addressed this problem through the identification of a cognate response activator and by driving the cluster expression with a nisininducible promoter PnisA (**Table 1**; Aso et al., 2004). Likewise, the utilization of a proper promoter was also essential for the successful production of a macrocyclic peptide telomestatin (**Table 1**). Initially, a xylose-inducible promoter (xylAp) was used to drive the expression of its gene cluster in the highly engineered Streptomyces avermitilis SUKA17 (Komatsu et al., 2013) but yielded no targeted molecule. It was later speculated that the transcription of the gene cluster should be activated during the late logarithmic phase of cell growth. Accordingly, the replacement of xylAp with the olmRp promoter led to the production of telomestatin in S. avermitilis SUKA17 (Amagai et al., 2017), clearly indicating the essentiality and importance of temporal control of gene expression in the successful production of natural products. Other remarkable examples of applying constitutive or inducible promoters to promote the success of RiPP heterologous expression include the complete refactoring of the cyanobactin patellamide pathway for its expression in E. coli Rosetta2 (DE3) (Donia et al., 2006), the use of inducible araPBAD promoter to drive the entire operon of a lasso peptide in E. coli BL21 (DE3) (Metelev et al., 2013), increased production of thiopeptides GE2270 and lactazole A

in Streptomyces hosts after introduction of the constitutive ermE<sup>∗</sup> promoter (Flinspach et al., 2014) and by a strong promoter (Hayashi et al., 2014), respectively (**Table 1**). Of note, the Link group constructed an expression system with two orthogonally inducible promoters to permit a separate control of the production and the export/immunity of lasso peptide MccJ25 in E. coli (**Table 1**, **Figure 3**). This elegant design enabled highthroughput screening of saturation mutagenesis libraries of the ring and β-hairpin tail regions of MccJ25 to obtain new insights to its structure-activity relationship (Pan and Link, 2011).

# RBS Substitution to Optimize Translation Efficiency

A ribosomal binding site (RBS) is critical in initiating the translation of many downstream genes. Its efficiency depends on the core Shine-Dalgarno (SD) sequence, the surrounding secondary structure, and the spacing between the SD sequence and the start codon AUG. Upon translation initiation, the 3′ sequence of the 16S rRNA complementarily pairs with the SD sequence in the RBS. Over millions of years of evolution, microbes have created and utilized a diverse set of RBSs to control protein translation (Omotajo et al., 2015), which is also employed to regulate the production of secondary metabolites. As such, RBSs are an important component part of synthetic biology applications including the heterologous production of RiPPs and other families of natural products (Bai et al., 2015). For example, the incorporation of optimized E. coli RBSs has proven to be an efficient way to significantly increase the yields of multiple lasso peptides, including astexin–1,−2, and−3 (Maksimov et al., 2012; Maksimov and Link, 2013), capistruin (Pan et al., 2012), and caulosegnin (**Table 1**) (Hegemann et al., 2013a). In a more inclusive example, Hegemann et al. cloned the gene clusters of lasso peptides from various sources into the expression vector pET41a, and included a strong E. coli RBS in the intergenic region between their precursor gene(s) and the genes encoding

processing enzymes (**Table 1**; Hegemann et al., 2013b). This design increased the production yields of almost all expressed lasso peptides by 1.8- to 84.5-folds, although the deletion of extra precursor peptides might also contribute to the yield improvement in some cases (Hegemann et al., 2013b).

# Optimization of the Catalytic Performance of Processing Enzymes

RiPP biosynthesis recruits a rapidly expanding list of functionally diverse enzymes to furnish structural and functional diversity (Arnison et al., 2013). The reactions of some RiPP biosynthetic enzymes require cofactors/co-substrates that may not be (or insufficiently) available in the surrogate host, leading to suboptimal production of targeted RiPPs. Therefore, optimal heterologous expression of RiPPs sometimes can be achieved by targeting cofactors/co-substrates of essential processing enzymes. For instance, NisB is a dehydratase involved in the biosynthesis of the food preservative nisin and its catalytic function requires glutamyl-tRNAGlu as a co-substrate, uncommon to RiPP processing enzymes (Ortega et al., 2016). Accordingly, increasing the cellular availability of Microbispora sp. 107891 glutamyltRNAGlu in E. coli was attempted to enhance the catalytic activity of MibB, a homolog of NisB involved in the biosynthesis of NAI-107. This study led to the production of NAI-107 analogs containing up to seven dehydrations, in contrast to nearly no dehydration when having no expressed Microbispora sp. 107891 glutamyl-tRNAGlu (**Table 1**) (Ortega et al., 2016). In a more pronounced example, the Schmidt group found that the addition of cysteine (5–10 mM) to the culture media, along with minor process changes, increased the yield of cyanobactin patellins by 150-folds (**Table 1**; Tianero et al., 2016). It was proposed that sulfide derived from cysteine specifically modulates the substrate preference of cyanobactin processing enzymes, enabling posttranslational control of product formation in vivo. Moreover, elevating the availability of the isoprene precursor, which is required by the pathway-specific prenyltransferase (Mcintosh et al., 2011), gave rise to an additional ∼18-fold increase of patellin yield in E. coli (**Table 1**).

# Codon Optimization to Enhance Heterologous Expression

Due to the different abundance of tRNAs in various hosts, each organism has its own codon preference. Thus, codon optimization of biosynthetic genes proves to be a good strategy to achieve optimal heterologous expression. For example, the biosynthetic genes of geobacillin I, a nisin analog encoded by the thermophilic bacterium Geobacillus thermodenitrificans NG80- 2, were codon-optimized before their introduction to E. coli for heterologous expression (Garg et al., 2012). Likewise, genes cylLL, cylLS, and cylM encoding the enterococcal cytolysin were synthesized with codon optimization for use in E. coli (Tang and Van Der Donk, 2013). Notably, in the heterologous expression of patellamides in E. coli, much lower yield was observed with vectors that were not codon-optimized (Schmidt et al., 2005).

# Manipulation of Pathway-Specific Regulators

Despite the brevity of RiPP biosynthetic logic (**Figure 1B**), their gene clusters often encode components for precursor peptides, processing enzymes, resistance mechanism and regulators, the same as other families of natural products (e.g., polyketides and nonribosomal peptides) (Ortega and van Der Donk, 2016). Targeting any of these components, particularly the regulators of RiPP biosynthetic pathways, can favor the success of RiPP TABLE 1 | Selected successful examples of heterologous expression of different RiPP families<sup>a</sup> .


(Continued)

#### TABLE 1 | Continued


<sup>a</sup>Entries were arranged first by the alphabetical order of the names of RiPP families and then chronically by the year of the publication.

heterologous production. A comprehensive review on generegulatory mechanisms operating in RiPPs biosynthesis was recently reported elsewhere (Bartholomae et al., 2017). We highlighted here an example about the essentiality of a pathwayspecific regulator to successful RiPP heterologous expression. The biosynthetic gene cluster of thiopeptide GE2270 (pbt) from Planobispora rosea ATCC 53733 previously failed to express the natural product in several Streptomyces hosts (**Table 1**; Tocchetti et al., 2013). In a recent report, Flinspach et al. revealed that the expression of PbtR, a TetR family of transcriptional regulator, is essential to the successful heterologous production in S. coelicolor M1146 (Flinspach et al., 2014).

## Engineering Resistance Mechanisms to Improve RiPP Productivity

Natural products are known to possess biological activities that target organisms in the same environmental niches, thereby offering survival benefits (Behie et al., 2017). To avoid self-toxicity, the producers accordingly evolve many different types of resistance mechanisms (e.g., transporters, chemical modification and target modification), often embedded in the natural product gene clusters (Jia et al., 2017; Almabruk et al., 2018). Expectedly, resistance mechanisms can offer a way to regulate the production of natural products, including RiPPs. For example, the biosynthesis of the lantibiotic nisin in Lactococcus lactis requires the dehydratase NisB, the cyclase NisC, the ABCtype transporter NisT, and the protease NisP, which together convert the precursor peptide NisA into the final product (Cheigh and Pyun, 2005). NisT forms a protein complex with NisB, C and P to effectively export bioactive nisin after its formation. Indeed, no secreted nisin was detected from the medium of a L. lactis mutant lacking the nisT gene, while the expression of nisABCP in this strain resulted in a considerable growth inhibition due to the intracellular accumulation of nisin (**Table 1**; Van Den Berg Van Saparoea et al., 2008). This example illustrates the necessity of a resistance mechanism to protect RiPP native producers. The same is likely true to surrogate hosts. For example, the ABC transporter MdnE was reported to be crucial for the successful production of a unique RiPP family, cyanobacterial tricyclic microviridins, in E. coli (**Table 1**; Weiz et al., 2011). In this case, MdnE might also act as a scaffold protein to guide the biosynthesis (Weiz et al., 2011). In another example, the multidrug transporter BotT of a bottromycin biosynthetic pathway is key to produce this antibiotic peptide in the surrogate host (Huo et al., 2012). Overexpression of the botT gene driven by a strong PermE<sup>∗</sup> promoter in S. coelicolor host enhanced the production titer by 20 times compared to the control with the unmodified cluster. In addition to transporter genes, other resistance-imparting genes can also be used to boost the heterologous production of RiPPs. For instance, the heterologous production of the bacteriocin enterocin A (EntA) was accomplished by fusing a Sec-dependent signal peptide (SPusp45) with mature EntA and coexpressing the EntA immunity gene entiA (**Table 1**; Jiménez et al., 2015). EntiA protects the producing strain by forming a strong complex with the receptor protein, mannose phosphotransferase system, to avoid the toxicity. These manipulations led to a 4.9-fold higher production of EntA than the native producer (Jiménez et al., 2015).

## Engineering Precursor Peptides to Produce RiPPs and Their Analogs

For the successful heterologous expression of RiPPs, one common hurdle is the lack of proper peptidases in the surrogate host to remove the leader peptide after finishing modifications on the core peptide (Bindman et al., 2015). Indeed, a number of RiPP gene clusters do not encode a protease dedicated to the removal of the leader peptide. The sequences of the linkers between the leader and core peptides also provide limiting information for the identification of such a protease from the genomes of native producers. To address this issue, the digestion site of a well-characterized, commercially available protease, such as GluC (Tang and Van Der Donk, 2012; Zhao and Van Der Donk, 2016;), trypsin (Himes et al., 2016), and C39 protease domain of the ABC transporter (Wang et al., 2014), can be engineered into the linker for the in vitro proteolytic release of the leader peptide from the matured precursor peptides isolated from heterologous hosts. As another approach, the van der Donk group genetically incorporated unnatural amino acids (UAAs) hydroxyl acids in the first position of a lanthipeptide by using a pyrrolysyl-tRNA synthetase-tRNAPyl CUA pair in E. coli (**Table 1**; Bindman et al., 2015). The installation of hydroxyl acid leads to an ester linkage between the leader and core peptides, which is readily cleavable by simple hydrolysis.

The majority of RiPPs precursor peptides comprise of the leader peptide region for the interactions with processing enzymes and the core peptide region that becomes the final products after chemical modification and proteolytic removal of the leader peptide (Arnison et al., 2013). The core region often carries multiple sequence variations that are tolerated by processing enzymes in modifications, providing opportunities to expand the chemical diversity of RiPPs. Indeed, genetic engineering of the core peptides of multiple RiPP families has led to impressive successes in exploring new chemical space for therapeutic applications. Two strategies have commonly been employed to diversify core peptide sequences, including singlesite saturation mutagenesis (Young et al., 2012) and multiplesite sequence randomization (Ruffner et al., 2015; Yang et al., 2018). The first strategy is advantageous to screen small-size libraries but can miss desirable mutants that require multiple mutations on the core peptides. By contrast, the second strategy in principle explores the broadest chemical space covered by large libraries (e.g., 10<sup>6</sup> -10<sup>9</sup> members), which is favored in drug discovery and development research. However, the success of this strategy depends on all three following factors, (1) the expression of all precursor peptide mutants in the host, (2) the proper processing of all mutants to generate large numbers of RiPP analogs, and (3) the high throughput screening methods to identify desirable compounds. In one recent example, Ruffner et al. employed the second strategy to randomly mutate the core peptide (TSIAPFC) of cyanobactin trunkamide (**Table 1**; Ruffner et al., 2015), whose processing enzymes are known to exhibit unusually relaxed sequence selectivity (Sardar et al., 2015). They prepared three double mutant libraries (XXIAPFC, TSXXPFC, and TSIXPXC) and a quadruple mutant library (XSXXPXC) in E. coli using the degenerate codon NNK. From the double mutant libraries (theoretically, 1,200 unique sequences in each library), they randomly screened a total of 460 clones, found 260 full-length precursor peptides, and detected 150 trunkamide analogs, giving a 33% success rate. The quadruple mutant library had the potential to produce 160,000 different sequences. The authors assessed the quality of this library by screening randomly picked 96 clones, found 65 full-length precursor peptides, and detected nine trunkamide analogs. The lower success rate (9.4%) of the quadruple mutant library may correlate with the selectivity of processing enzymes. In this regard, the van der Donk group recently leveraged the remarkable substrate tolerance of a lanthipeptide synthetase ProcM to generate a genetically encoded lanthipeptide library (Yang et al., 2018). They first randomized 10 positions of the core peptide of the precursor peptide ProcA2.8 (**Table 1**) (AACXXXXXSMPPSXXXXXC) using the NWY codon that encodes eight amino acids, leading to a 1.07 × 10<sup>9</sup> library. Limited by the transformation efficiency of E. coli, they obtained ∼10<sup>6</sup> clones, 99.7% of which produced unique peptide sequences. Screening of 33 randomly selected clones led to identify 33 cyclized samples, illustrating the impressive versatility and substrate flexibility of ProcM. The authors then screened all 10<sup>6</sup> lanthipeptides using a cell survival-based high

throughput assay and identified one potent inhibitor of HIV p6 protein (Yang et al., 2018). In addition to the use of E. coli as a host to produce mutated RiPPs, both yeast display and phage display have recently been used to generate libraries of 10<sup>6</sup> lanthipeptides for screening for new bioactive analogs (Urban et al., 2017; Hetrick et al., 2018). These two well-characterized platforms can find more applications in expanding the chemical space of other RiPP families by the sequence randomization strategy.

In addition to 20 proteinogenic amino acids, a variety of UAAs can be used to expand the chemical diversity of RiPPs (Young and Schultz, 2010). This strategy has demonstrated its success with multiple RiPP families, including lantipeptide (Nagao et al., 2005; Oldach et al., 2012; Bindman et al., 2015; Kuthning et al., 2016; Lopatniuk et al., 2017; Zambaldo et al., 2017), lasso peptide (Piscotta et al., 2015), cyanobactin (Tianero et al., 2012), and sactipeptide (Himes et al., 2016). However, these unnatural RiPP analogs showed no significant improvement of their bioactivities possibly due to the relatively small extent of chemical expansion brought by a single UAA on a single position. However, coupled with the directed evolution of targeted core peptides, e.g., multiple-site randomization as described above, this strategy can generate new-to-nature RiPP analogs with enhanced structural and functional diversity.

RiPP precursor peptides physically separate their molecular recognition and catalysis sites for the processing by enzymes. Capitalizing on this distinct feature, a chimeric leader peptide strategy was recently developed to produce RiPP hybrids (Burkhart et al., 2017). Specifically, the leader peptides for the binding of thiazoline-forming cyclodehydratase, thioetherformation AlbA involved in the biosynthesis of sactipeptide, and lanthipeptide dehydratases NisB/C and ProcM were fused to allow sequential interactions with multiple processing enzymes of different RiPP families (**Figure 4**). As such, the engineered core peptides were received a combination of chemical transformations to produce unnatural peptide products, providing a generally applicable strategy to unlock the vast chemical space afforded by a variety of RiPP biosynthetic machinery (Burkhart et al., 2017).

## MANIPULATION OF SURROGATE HOSTS FOR THE PRODUCTION OF RiPPs

### Optimization of Culture Conditions

Screening a wide array of fermentation conditions, e.g., temperature, pH, shaking speed, nutrient levels, and trace metals, has routinely been practiced for the optimal production of target products. For example, Knappe et al. heterologously expressed the gene cluster of lasso peptide capistruin in E. coli and achieved a yield of 0.2 mg/L in the defined medium M20, which was 30% of its native producer Burkholderia thailandensis E264 in the same medium (**Table 1**) (Knappe et al., 2008). Surprisingly, no capistruin was produced when culturing transformed E. coli in commonly used LB medium. In another example, after testing a variety of conditions, the co-expression of Fe-S cluster biogenesis genes and lowered shaking speed together led to the significantly improved expression of subtilosin A in E. coli (Himes et al., 2016).

# The Use of Suitable Hosts for the Production of RiPPs

An ideal host for the heterologous expression of natural products usually requires a clean background and high compatibility with the target biosynthetic gene cluster. More specifically, the ideal heterologous host would be able to supply abundant biosynthetic precursors from its primary metabolism while maintaining a relative clean secondary metabolic background, and also be capable of recognizing exogenous genetic parts, thus allowing access to the vast biosynthetic potential of the host. In this regard, E. coli has become one of the most popular heterologous hosts, and produced many RiPP families, e.g., cyanobactins, lantipeptides, lasso peptides, microviridins, and sactipeptides (Donia et al., 2006; Weiz et al., 2011; Metelev et al., 2013; Himes et al., 2016; Kuthning et al., 2016). On the other hand, the RiPP gene cluster from a high G+C producer is often expressed in a host with a relatively comparable genetic background. For example, the lantibiotic cinnamycin is produced by several Streptomyces strains and its gene cluster from S. cinnamoneus cinnamoneus DSM 40005 was successfully expressed in S. lividans to produce this peptidic antibiotic (Widdick et al., 2003). In another study, S. lividans TK23 and S. avermitilis SUKA17 were used as the hosts to produce thioviridamide (**Table 1**, **Figure 5A**; Izawa et al., 2013; Izumikawa et al., 2015). Interestingly, the expression of its gene cluster in S. avermitilis SUKA17 led to the production of a novel thioviridamide derivative, JBIR-140, further demonstrating the significant influence of a surrogate host on RiPP production.

In addition to the widely used hosts like E. coli and Streptomyces strains, several uncommon microorganisms have also been characterized as suitable RiPP heterologous hosts. With the consideration of available substrates and comparable genetic backgrounds, these heterologous hosts are often from the same family of the native producers of the target RiPPs. For example, when expressing the clusters of the lantibiotics microbisporicin and planosporicin and the thiopeptide GE2270 from Microbispora coralline, Planomonospora alba and Planobispora rosea, respectively, Nonomuraea sp. ATCC 39727, which is in the same family of the above native producers, acted as a viable host to produce corresponding RiPPs, but not multiple other tested Streptomyces strains (**Table 1**, **Figure 5B**; Foulston and Bibb, 2010; Sherwood et al., 2013; Tocchetti et al., 2013).

In recent years, fungal RiPPs have attracted increasing attentions given the availability of a number of fungal genomes in public domain (Hallen et al., 2007; Ding et al., 2016; Nagano et al., 2016; Ramm et al., 2017). To realize the chemical and functional potential of fungal RiPPs, their heterologous expression systems have to be established. In this regard, commonly used fungal strains can be initial targets in the development. Encouragingly, several biosynthetic genes of ustiloxin, the first filamentous fungal RiPP, were successfully expressed in Aspergillus oryzae, greatly facilitating the understanding of the macrocyclic formation and its entire biosynthetic pathway (**Table 1**, **Figure 5C**; Ye et al., 2016). In a more recent example, the partial reconstitution of the

FIGURE 4 | A chimeric leader peptide strategy to produce unnatural RiPP hybrids. By properly designing the concatenated leader peptides, recognition and processing by multiple enzymes from unrelated RiPP pathways could be realized. By using this method, a thiazoline-forming cyclodehydratase was combined with biosynthetic enzymes from the sactipeptide and lanthipeptide families to create new-to-nature hybrid RiPPs, demonstrating the feasibility of the strategy.

biosynthesis of one dodecapeptide omphalotin A, which is ribosomally produced by the basidiomycete Omphalotus olearius, was succeeded in Pichia pastoris strain GS115, but not E. coli (Ramm et al., 2017). This work further shed light on a novel biosynthesis mechanism for a RiPP in which a self-sacrificing enzyme, methyltransferase OphMA, bears its own precursor peptide.

Cyclotides are a family of plant-derived RiPPs that are characterized by a head-to-tail cyclic peptide backbone and a cystine knot arrangement of disulfide bonds. These peptidic compounds possess a wide range of bioactivities (e.g., protease inhibition, anti-microbials, and cytotoxicity) and are good carriers of other bioactive peptides, both of which are attractive to pharmaceutical research. Recently, the heterologous production of cyclotides were successfully achieved by coexpressing a select asparaginyl endoprotease and its precursor peptide in planta, using Nicotiana benthamian, tobacco, bush bean, lettuce, and canola as hosts (Poon et al., 2018). Interestingly, alternative strategies such as intein-mediated protein trans-splicing (Jagadish et al., 2013) and sortase-induced backbone cyclization (Stanger et al., 2014) have also been developed to produce cyclotides in bacterial and yeast expression systems, in which the asparaginyl endoprotease is not employed for the cyclization.

#### CONCLUSION AND FUTURE PERSPECTIVES

Harnessing the biosynthetic prowess of RiPPs via heterologous expression has witnessed several exciting advances in recent years. As described above, due to the conciseness of the biosynthetic route, the cloning and mobilization of the RiPP gene clusters typically do not constitute a major hurdle for the heterologous production of RiPPs. However, the functional expression of biosynthetic genes in surrogate hosts could be complicated by many less-predictable factors, such as the availability of protein cofactors, promoter recognition, product toxicity, protein–protein interaction, and imbalanced protein dosage. On the other hand, with E. coli and Streptomyces strains serving as the most common hosts in the heterologous expression of RiPPs, the ever-increasing number of synthetic biology tools developed for these systems can be applied to overcome these challenges. In addition, in vitro characterization of RiPP biosynthesis and in silico prediction can be coupled to streamline and improve the outcomes of heterologous expression efforts. We are optimistic that a small set of highly developed hosts will be available as generally applicable platforms for rapid and robust sampling of the vast chemical space of RiPPs from bacteria, fungi, and even plants in future.

# AUTHOR CONTRIBUTIONS

YZ, MC, SB, and YD planned, wrote and reviewed the manuscript.

# REFERENCES


#### FUNDING

This work was partly supported by America Cancer Society Institutional Research Grant (YD), the Department of Medicinal Chemistry at the University of Florida and NIH NIGMS (1R35GM128742).

#### ACKNOWLEDGMENTS

We would like to thank lab members of the Ding and Bruner groups for many discussions of this topic. We also thank Prof. Hendrik Luesch for informative suggestions. Due to length limitations, here we were unable to include many other excellent studies in the heterologous production of RiPPs. We apologize to the authors whose studies were not cited.


by multigene assembly. Appl. Microbiol. Biotechnol. 99, 6351–6361. doi: 10.1007/s00253-015-6557-6


system from a ruminant bacterium. Cell Chem. Biol. 23, 246–256. doi: 10.1016/j.chembiol.2015.11.014


**Conflict of Interest Statement:** 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 © 2018 Zhang, Chen, Bruner and Ding. 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.

# Engineering Heterologous Production of Salicylate Glucoside and Glycosylated Variants

Ruiquan Qi<sup>1</sup> , Blaine A. Pfeifer1,2,3 \* and Guojian Zhang1,2,3 \*

<sup>1</sup> Department of Chemical and Biological Engineering, University at Buffalo, The State University of New York, Buffalo, NY, United States, <sup>2</sup> Key Laboratory of Marine Drugs, Chinese Ministry of Education, School of Medicine and Pharmacy, Ocean University of China, Qingdao, China, <sup>3</sup> Laboratory for Marine Drugs and Bioproducts, Qingdao National Laboratory for Marine Science and Technology, Qingdao, China

Salicylate 2-O-β-D-glucoside (SAG) is a plant-derived natural product with potential utility as both an anti-inflammatory and as a plant protectant compound. Heterologous biosynthesis of SAG has been established in Escherichia coli through metabolic engineering of the shikimate pathways and introduction of a heterologous biosynthetic step to allow a more directed route to the salicylate precursor. The final SAG compound resulted from the separate introduction of an Arabidopsis thaliana glucosyltransferase enzyme. In this study, a range of heterologous engineering parameters were varied (including biosynthetic pathway construction, expression plasmid, and E. coli strain) for the improvement of SAG specific production in conjunction with a system demonstrating improved plasmid stability. In addition, the glucoside moiety of SAG was systematically varied through the introduction of the heterologous oliose and olivose deoxysugar pathways. Production of analogs was observed for each newly constructed pathway, demonstrating biosynthetic diversification potential; however, production titers were reduced relative to the original SAG compound.

#### Edited by:

Dipesh Dhakal, Sun Moon University, South Korea

#### Reviewed by:

Joong-Hoon Ahn, Konkuk University, South Korea Prakash Parajuli, University of Maryland, Baltimore, United States

#### \*Correspondence:

Blaine A. Pfeifer blainepf@buffalo.edu Guojian Zhang guojianz@buffalo.edu; zhangguojian@ouc.edu.cn

#### Specialty section:

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

> Received: 21 July 2018 Accepted: 03 September 2018 Published: 20 September 2018

#### Citation:

Qi R, Pfeifer BA and Zhang G (2018) Engineering Heterologous Production of Salicylate Glucoside and Glycosylated Variants. Front. Microbiol. 9:2241. doi: 10.3389/fmicb.2018.02241 Keywords: salicylate, salicylate 2-O-β-D-glucoside, metabolic engineering, E. coli, analog

### INTRODUCTION

Plants have dedicated metabolism for the production of salicylate and a glycosylated version, salicylate 2-O-β-D-glucoside (SAG), which is often stored intracellularly until external stress is encountered (Vlot et al., 2009; Rivas-San Vicente and Plasencia, 2011). At which point, the reversion of SAG to salicylate allows the bioactivity of the latter compound to combat various biological threats to the plant system. Salicylate is also a central component of aspirin and, as such, SAG has the potential to possess similar anti-inflammatory properties.

These various bioactivities of SAG prompted us to explore its production through a heterologous bacterial host. Leveraging the knowledge and prior studies associated with engineering the shikimate pathway of Escherichia coli (Lin et al., 2014), we generated a production host supportive of high titer levels of salicylate (>1 g/L) (Ahmadi et al., 2016). This work included the introduction of an Irp9 salicylate synthase gene from Yersinia enterocolitica, which streamlined metabolism toward this precursor (**Figure 1**; Pelludat et al., 2003; Kerbarh et al., 2005). The introduction of a glucosyltransferase gene (ugt74f1) from Arabidopsis thaliana enabled conversion to the final SAG compound (Ahmadi et al., 2016).

In the work presented herein, we were interested in improving SAG production through the application of various heterologous production parameters that spanned the E. coli production strain and several expression plasmids and associated components. In addition, based upon previous work by us and others toward glycodiversification of heterologous natural products, we tested the expanded analog potential of the original SAG compound through the systematic incorporation of two isomeric deoxysugar pathways.

# MATERIALS AND METHODS

#### Plasmids and Strains

All cloning procedures were completed in E. coli TOP10 through which all recombinant plasmids were transferred and propagated. The irp9 gene from Y. enterocolitica genomic DNA, the pgm and galU genes from E. coli K-12 MG1655, and a codon-optimized ugt74F1 (salicylate glucosyltransferase gene) from A. thaliana were amplified by PCR with primers listed in **Supplementary Table S1**. Plasmids pMKA-41 and pGEX-UDP bearing these genes were constructed as described previously (Ahmadi et al., 2016) and used as the templates for the PCR reactions conducted in the current work. The PCR products were gel-purified and then digested with restriction enzyme pairs NheI/SalI (for irp9), XbaI/SalI (for galU), and NdeI/SalI (for pgm and ugt74F1). Digested irp9, pgm, galU, and ugt74F1 were separately ligated into similarly digested pET28a to yield pET28-irp9, pET28 pgm, pET28-galU, and pET28a-ugt74F1. Plasmid pET28-pgm was digested with XbaI/SalI and the insert transferred to an SpeI/SalI digested pET28-galU to yield pET28-galU-pgm. In the same way, the XbaI/SalI digested ugt74F1 fragment from pET28-ugt74F1 was inserted into SpeI/SalI digested pET28-irp9 to construct pET28-irp9-ugt74F1. Plasmid pET28-irp9-ugt74F1 was digested with XbaI/SalI and the insert ligated to SpeI/SalI digested pET28-galU-pgm to generate pRQS1. The pRQS1 galUpgm-irp9-ugt74F1 cassette featured in **Figure 2** was digested with NheI/BsiWI and ligated into pETcoco-1 for subsequent digestion and transfer of the same cassette (using XbaI/SalI) to pBAD33, yielding pRQS2 and pRQS3, respectively. A full list of plasmids and strains are presented in **Supplementary Table S2** and detailed plasmid maps are provided in **Supplementary Figures S1**, **S2**.

The double-knockout BW mutant (BW23) was constructed as described in previous work (Ahmadi et al., 2016) through deletion of pheA and tyrA, with these deletions improving both SA and SAG production. The λDE3 prophage was integrated into the BW23 chromosome through co-infection using a λDE3 Lysogenization Kit (Novagen), yielding BW23(DE3). By doing so, the new host is equipped with the λDE3 recombinant phage DNA encoding for the T7 RNA polymerase, therefore, allowing the expression of the galU-pgm-irp9-ugt74F1 cassette though the T7 promotor within the pRQS1 and pRQS2 constructs. The pRSQ plasmids were transformed into corresponding strains through the standard heat-shock transformation protocol (**Supplementary Table S3**), and the resulting strains were stored as 20% glycerol stocks at −80◦C.

The gene fragment for irp9 was liberated from pET28 irp9 via XbaI/SalI digestion and then ligated into SpeI/SalI digested pET28-galU-pgm to construct pET28-galU-pgmirp9, which was then XbaI/SalI digested and ligated into pET21c to generate pRQS4. Plasmids pGJZ1, 2, 3, and 4

(Zhang et al., 2015), containing genes from the oleandomycin, chromomycin, and urdamycin A polyketide biosynthetic pathways, were used to produce two pairs of deoxysugar pathways for oliose and olivose. These four plasmids were integrated with codon-modified urdGT to construct pGJZ1-GT, pGJZ2-GT, pGJZ3-GT, and pGJZ4-GT (**Supplementary Figure S3**) which were, respectively, co-transferred with pRQS4 into BL21(DE3) (**Supplementary Figure S4** and **Supplementary Table S4**) to provide the complete biosynthetic pathways for SAG analogs.

# Culture Conditions and Medium Components

The bacterial culture medium and associated chemical and analytical components were obtained from Sigma-Aldrich (St. Louis, MO, United States) or Thermo Fisher Scientific (Waltham, MA, United States). The DNA-manipulating agents, including restriction enzymes, T4 DNA ligase, Phusion High-Fidelity PCR Master Mix, and associated reagents were purchased from New England Biolabs (Ipswich, MA, United States). PCR primers (**Supplementary Table S1**) were obtained from Eurofins Genomics (Huntsville, AL, United States).

Respective glycerol stocks of producing strains from **Supplementary Table S3** were used to initiate overnight 3 mL cultures at 37◦C with shaking in lysogeny broth (LB) medium prior to inoculating (1% v/v) 25 mL of M9Y medium which is formulated with (per liter): Na2HPO4·7H2O (12.8 g); KH2PO<sup>4</sup> (3 g); NaCl (0.5 g); NH4Cl (1 g); yeast extract (1 g), glycerol (10 g), glucose (2.5 g), MgSO4·7H2O (246.5 mg), and CaCl2·2H2O (14.7 mg) supplemented with micronutrients including (per liter) vitamin B1 (2.0 mg), H3BO<sup>3</sup> (1.25 mg), NaMoO4·2H2O (0.15 mg), CoCl2·6H2O (0.7 mg), CuSO4·5H2O (0.25 mg), MnCl2·4H2O (1.6 mg), and ZnSO4·7H2O (0.3 mg). After inoculation, cultures were incubated at 30◦C with shaking for 2 days with induction initiated using 200 µM isopropyl β-D-1-thiogalactopyranoside (IPTG) and/or 3 mg/mL arabinose when the culture OD600nm reached 0.4–0.6. As needed, plasmid selection in both liquid and solid medium was maintained with 100 mg/L ampicillin, 50 mg/L kanamycin, and 20 mg/L chloramphenicol.

#### Plasmid Stability Assay

Salicylate 2-O-β-D-glucoside producing strains were cultured in 25 mL production medium containing appropriate antibiotics at 37◦C and 250 rpm until the OD600nm reading reached 0.4–0.6. At this point, cultures were cooled to 22◦C, induced for gene expression as described above, and incubated an additional 5 days. Plasmid stability analysis was completed on the first and fifth days after cultures were induced. At these time points, dilutions from each culture were spread evenly on LB agar plates

for incubation overnight at 37◦C. Thirty colonies from each plate were selected and transferred to LB agar plates containing (1) combined antibiotics according to associated gene expression plasmids, (2) 1 mM IPTG, and (3) 1 mM IPTG + combined antibiotics as indicated in **Table 1**. Resulting colony development was then recorded, presented as a percentage of transferred colony growth on LB agar containing no antibiotics, and compared between strains.

## Salicylate 2-O-β-D-glucoside (SAG) Production Quantification

Post-culture, 1 mL of acetone was added per 50 mL culture and a 1 mL sample was centrifuged. Supernatant (50 µL) was analyzed by HPLC as described previously (Dean et al., 2005). Briefly, SAG and associated analogs were quantified using a ZORBAX Eclipse XDB-C18 column connected to an Agilent 1100 system equipped with a diode array detector. Solvent A was 0.1% acetic acid in water, solvent B was methanol, and a flow rate of 1 mL/min was used across the following gradient: 5–20% solvent B over 10 min; 20–80% solvent B over 5 min; 80% solvent B maintained for 5 min; reset to 5% solvent B. An absorbance wavelength of 274 nm was used for both SAG and associated analog quantification (Ahmadi et al., 2016). Peak area quantification was conducted compared to a standard calibration curve of pure SAG (Toronto Research Chemicals, Toronto, ON, Canada).

# LC-MS Analysis for SAG Analog Assessment

A 1 mL SAG analog culture sample was centrifuged and 50 µL of supernatant was used for analysis. LC-MS was performed using an API 3000 Triple Quad LC-MS with a Turbo Ion Spray source (PE Sciex) coupled with a Shimadzu Prominence LC system. Chromatography was performed through a Waters XTerra C18 column (5 mm, 2.1 mm × 250 mm) and MS analysis was conducted in positive ion mode. Following a 3 µL injection from the 50 µL sample, a linear gradient of 5–95% acetonitrile (balance water; both solutions containing 0.1% formic acid) was used for 20 min at a flow rate of 0.2 mL/min.

#### Statistical Evaluation

Data presented were generated from three independent experiments, and error bars represent standard deviation values.

# RESULTS

As outlined in **Figure 2**, cellular parameters were engineered to improve specific SAG production. First, the production system designed previously for SAG formation [represented by plasmids pMKA-41 and pGEX-UDP (Ahmadi et al., 2016)] was organized into one operon introduced to three different expression plasmids (**Figure 2A**). Plasmids pETcoco-1 and pET28a allowed operon expression from the T7 promoter system coupled to a bacterial strain containing the T7 RNA polymerase (encoded within DE3 cellular variants). The pBAD33 plasmid featured expression driven by an arabinose inducible promoter system (Guzman et al., 1995). Plasmids were then introduced to strains BW23(DE3) and BL21(DE3) to accommodate the T7 based plasmids or BW23 for the pBAD33 plasmid system (also used for pMKA-41 and pGEX-UDP).

A production comparison revealed that the BW23(DE3)/pRSQ2 strain generated the best relative SAG levels based upon volumetric and specific titer comparisons (**Figures 2B,C** and **Supplementary Figure S5**). When comparing performance by either SAG titer or production per cell density, the BW23(DE3)/pRSQ2 strain demonstrated a twofold to fivefold improvement relative to the original BW23/pMKA-41/pGEX-UDP system. Of the new expression plasmids tested, pRSQ1 showed the lowest levels of SAG production. Of the production strains tested, the BW23(DE3) background, engineered to support SAG metabolic channeling and to accommodate the strong T7 promoter, demonstrated the best overall titers.

The plasmids used in this study included a low-copy option (pRSQ2; pETcoco-1 [OriV/S, 1–2 copies per cell]) and two medium copy options: pRSQ1 (pET28a [pBR322, ∼40 copies per cell]) and pRSQ3 (pBAD33 [pACYC184/p15A]). **Table 1** presents plasmid stability data for the associated SAG production strains. From this analysis, the RSQ2 plasmid shows the best


TABLE 1 | Plasmid stability comparison with antibiotic selection, IPTG induction, and combined IPTG induction and antibiotic selection.

Numbers represent percentage of transferred colony growth. <sup>∗</sup>Cells carrying plasmid(s) are able to form colonies on plates containing corresponding antibiotics. ∗∗Those strains bearing no plasmid or mutants without the ability to express target genes can grow on plates with 1 mM IPTG. ∗∗∗Those strains bearing plasmids but having lost the ability to express the target genes will form colonies on plates with both antibiotics and 1 mM IPTG.

overall stability when tested for plasmid maintenance over time. The consolidated SAG biosynthetic pathways across plasmids pRSQ1-3 showed improved stability relative to the original dual expression plasmid system reliant upon pMKA-41 and pGEX-UDP. The same set of strains were also tested for stability when exposed to either IPTG induction or IPTG induction with antibiotic selection Novagen (1999). The plasmid stability data under these conditions indicate that, with the exception of pRSQ2, the newly constructed systems suffer from a combination of plasmid loss and mutant formation that results in lack of gene expression. From this perspective, the order of plasmid stability would be pRSQ2, pRSQ3, and pRSQ1.

**Figure 3** outlines a schematic to test the flexibility of the SAG pathway to generate analogs resulting from alternative glycosylation patterning. In particular, isomeric variants of the oliose and olivose deoxysugars were tested for glycodiversification of the incoming salicylate precursor (Aguirrezabalaga et al., 2000). Initial analog production efforts with the Ugt74F1 salicylate glucosyltransferase resulted in minimal product formation (data not shown). As an alternative, the urdamycin system glycotransferase (UrdGT) was used due to previously observed flexibility in glycosylation patterning (Hoffmeister et al., 2000, 2003). Using HPLC and LC-MS analysis, data supporting analog formation were generated for each deoxysugar variant (**Figures 4A,B** and **Supplementary Figures S6**, **S7**). However, production levels were significantly reduced compared to those from the original SAG production systems (**Figure 2**).

#### DISCUSSION

A combination of metabolic engineering and gene expression design resulted in plasmids pRSQ1, 2, and 3. Each plasmid design consolidated genes needed for SAG biosynthesis. Furthermore, the plasmids featured different copy numbers and promoter strengths and, when combined with E. coli strains engineered to support gene expression and SAG production, allowed for a systematic evaluation of final product values. Production levels were best for strain BW23(DE3)/pRSQ2, which featured

a metabolically engineered strain to streamline carbon flow to SAG production and support the strong T7 promoter driving expression from pRSQ2. Of note, this particular expression plasmid was the lowest copy version of those tested. pRSQ1 featured T7-based gene expression but from a higher copy plasmid; pRSQ3 utilized an arabinose-inducible system within pBAD33 (similar to use previously with the control system BW23/pMDA-41/pGEX-UDP). This particular study focused only on SAG production as a function of plasmid biosynthetic pathway design (**Figure 2**); however, future engineering approaches, such as tuning biosynthetic steps via expression variation, will likely be needed to maximally drive complete salicylate to SAG formation (Ahmadi et al., 2016).

One likely contributor to the heightened production observed for pRSQ2 was incorporation of the highly stable OriV/S replication system (Shizuya et al., 1992; Monaco and Larin, 1994; Tao and Zhang, 1998). As such, even though the copy number for this plasmid was reduced, stability was improved as indicated within the results presented in **Table 1**. The strong T7 expression system likely compensated for the reduced copy number level (Golomb and Chamberlin, 1974; Studier and Moffatt, 1986); whereas, the higher copy number T7 system represented by pRSQ1 showed lower relative SAG production and higher plasmid loss. The pRSQ2 system has the added advantage (not tested in this study) of plasmid copy-up capability (Wild et al., 1996, 2001, 2002; Wild and Szybalski, 2004). Thus, there is the potential for further increased production through the stable maintenance of the pRSQ2 plasmid during the growth phase of the host system followed by induction of both SAG biosynthesis and plasmid amplification to spur subsequent product generation.

Our group and others have studied natural product glycodiversification, which offers a directed means of natural product structural variation (Thibodeaux et al., 2007, 2008; Williams et al., 2008; Jiang et al., 2013; Zhang et al., 2015; Fang et al., 2018). Given the glycosylated nature of SAG, we were interested if this product could also accommodate alternative sugar moieties. In conducting this work, we relied on a series of deoxysugar pathways previously incorporated into polyketide biosynthesis to generate erythromycin analogs (Zhang et al., 2015). However, contingent upon this strategy working is the flexibility of a glycosylation enzymatic step capable of accepting new substrate groups. Use of the original glucosyltransferas (Ugt74F1) resulted in minimal analog production. As a result, we turned to an alternative glycotransferase from the urdamycin biosynthetic pathway [recognized for substrate flexibility (Hoffmeister et al., 2002; Luzhetskyy et al., 2005) and the same source as some of the deoxysugar pathways genes], which resulted in the production level of the analogs presented in **Figure 4**. The ability for the urdamycin glycotransferase to accommodate novel SAG analogs supports the general theme of glycodiversification applied to this compound. As in the case of several previous examples of analogs produced as a result of biosynthetic pathway modification (Jiang and Pfeifer, 2013; Zhang et al., 2015), titer levels of the SAG analogs were significantly reduced compared to the original product, likely to do the new substrates limiting the catalytic activity of the glycosyltransferase. We also note that additional analytical work is needed to fully chemically characterize these new analogs. However, indication of novel analogs provides a basis for future studies to test potential variation in bioactivity in applications that range from inflammatory relief [as we have tested previously with the original SAG compound (Ahmadi et al., 2016)] or plant stress protection.

#### AUTHOR CONTRIBUTIONS

GZ and BP designed and supervised the study. RQ executed the experimental plan.

#### FUNDING

The authors recognize funding provided by the NYS Pollution Prevention Institute through a grant from the NYS Department of Environmental Conservation. Any opinions, findings, conclusions, or recommendations expressed are those of the author(s) and do not necessarily reflect the views of the Department of Environmental Conservation. The authors also acknowledge funding support from the following agencies: SUNY 4E and NSF (1550378).

#### REFERENCES

fmicb-09-02241 September 18, 2018 Time: 16:50 # 7


#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.02241/full#supplementary-material

Novagen (1999). pET System Manual, 10th Edn. Birmingham: Novagen.


**Conflict of Interest Statement:** 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 © 2018 Qi, Pfeifer 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.

# Enhanced Biosynthesis of 2-Deoxy-scyllo-inosose in Metabolically Engineered Bacillus subtilis Recombinants

Joo Hyun Lim<sup>1</sup>† , Hyun Ha Hwang<sup>1</sup>† , Na Joon Lee<sup>1</sup> , Jae Woo Lee<sup>1</sup> , Eun Gyo Seo<sup>1</sup> , Hye Bin Son<sup>1</sup> , Hye Ji Kim<sup>1</sup> , Yeo Joon Yoon<sup>2</sup> and Je Won Park1,3 \*

<sup>1</sup> Department of Integrated Biomedical and Life Sciences, Graduate School, Korea University, Seoul, South Korea, <sup>2</sup> Department of Chemistry and Nanoscience, Ewha Womans University, Seoul, South Korea, <sup>3</sup> School of Biosystem and Biomedical Science, Korea University, Seoul, South Korea

#### Edited by:

Dipesh Dhakal, Sun Moon University, South Korea

#### Reviewed by:

Yuhui Sun, Wuhan University, China Kambiz Morabbi Heravi, University of Hohenheim, Germany

\*Correspondence: Je Won Park jewonpark@korea.ac.kr †These authors have contributed

#### Specialty section:

equally to this work

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

> Received: 04 July 2018 Accepted: 11 September 2018 Published: 27 September 2018

#### Citation:

Lim JH, Hwang HH, Lee NJ, Lee JW, Seo EG, Son HB, Kim HJ, Yoon YJ and Park JW (2018) Enhanced Biosynthesis of 2-Deoxy-scyllo-inosose in Metabolically Engineered Bacillus subtilis Recombinants. Front. Microbiol. 9:2333. doi: 10.3389/fmicb.2018.02333 2-Deoxy-scyllo-inosose (DOI) has been a valuable starting natural product for the manufacture of pharmaceuticals or chemical engineering resources such as pyranose catechol. DOI synthase, which uses glucose-6-phosphate (Glc6P) as a substrate for DOI biosynthesis, is indispensably involved in the initial stage of the biosynthesis of 2 deoxystreptamine-containing aminoglycoside antibiotics including butirosin, gentamicin, kanamycin, and tobramycin. A number of metabolically engineered recombinant strains of Bacillus subtilis were constructed here; either one or both genes pgi and pgcA that encode Glc6p isomerase and phosphoglucomutase, respectively, was (or were) disrupted in the sugar metabolic pathway of the host. After that, three different DOI synthase–encoding genes, which were artificially synthesized according to the codon preference of the B. subtilis host, were separately introduced into the engineered recombinants. The expression of a natural btrC gene, encoding DOI synthase in butirosin-producing B. circulans, in the heterologous host B. subtilis (BSDOI-2) generated approximately 2.3 g/L DOI, whereas expression of an artificially codonoptimized tobC gene, derived from tobramycin-producing Streptomyces tenebrarius, into the recombinant of B. subtilis (BSDOI-15) in which both genes pgi and pgcA are disrupted significantly enhanced the DOI titer: up to 37.2 g/L. Fed-batch fermentation by the BSDOI-15 recombinant using glycerol and glucose as a dual carbon source yielded the highest DOI titer (38.0 g/L). The development of engineered microbial cell factories empowered through convergence of metabolic engineering and synthetic biology should enable mass production of DOI. Thus, strain BSDOI-15 will surely be a useful contributor to the industrial manufacturing of various kinds of DOI-based pharmaceuticals and fine chemicals.

Keywords: 2-deoxy-scyllo-inosose, Bacillus subtilis, metabolic engineering, artificial gene, 2-deoxy-scylloinosose synthase

**Abbreviations:** DOI, 2-Deoxy-scyllo-inosose; Glc6P, glucose-6-phosphate.

# INTRODUCTION

fmicb-09-02333 September 25, 2018 Time: 18:0 # 2

Pyranose compounds have been produced in the traditional petrochemical sector from petroleum as a raw material. Most of these aromatic compounds including catechol and benzenoids are still being made from petroleum (Hansen and Frost, 2002; Baire et al., 2013), but because of limited petroleum reserves and worldwide regulations on carbon dioxide emissions, the development of environment-friendly and sustainable production processes using biomass (e.g., a fermentation process) is in demand.

2-Deoxy-scyllo-inosose (DOI) synthase, which uses Glc6P as a substrate when catalyzing the synthesis of pyranose compound DOI was first discovered as BtrC in Bacillus circulans that produces 2-deoxystreptamine-containing aminoglycoside butirosins (Kudo et al., 1999). This enzyme participates in the biosynthetic steps necessary for the core 2-deoxystreptamine scaffold (Llewellyn and Spencer, 2006; Park et al., 2013), and its catalytic product DOI has been broadly utilized as a starting material or a precursor of agrochemicals and pharmaceuticals (Hansen and Frost, 2002). Chemical synthesis of this DOI requires multistep reactions and hazardous and expensive metals, whereas the biosynthesis of DOI by DOI synthase allows for efficient synthesis in a single process. A method for producing DOI in a single enzymatic reaction was established (Takagi et al., 2006) in which a recombinant DOI synthase expressed in Escherichia coli is mixed with Glc6P. In addition, there was a report about a two-step enzymatic reaction that involves hexokinase and DOI synthase acting on D-glucose (Kakinuma et al., 2000). Furthermore, it was also reported that by the concentration of the enzymatic reactants followed by a reaction with hydrogen iodide under mild acidic conditions, DOI can be converted even to catechol (Suzuki et al., 2013).

Meanwhile, there was a publication concerning the biosynthesis of DOI via microbial fermentation of biomassderived glucose by a recombinant strain of E. coli, into which a heterologous DOI synthase-encoding gene was introduced (Kogure et al., 2007). Accordingly, along with glucose, the rare and expensive sugar alcohol mannitol was also required as an extra carbon source for the support of microbial proliferation and growth of E. coli. In other words, having DOI synthase expressed in a wild-type strain of E. coli alone resulted in low productivity in terms of DOI, but high production of DOI (29.5 g/L), as reported, was achieved by simultaneous disruption of three genes essential for the primary metabolic pathway of E. coli: phosphoglucose isomerase (pgi), glucose 6-phosphate-1-dehydrogenase (zwf), and phosphoglucomutase (pgm). With all the metabolic pathways via which glucose can enter glycolysis eventually being blocked, mannitol that can be used in alternate glycolysis should be requisite for E. coli growth. In addition, a method for producing DOI from plant-derived ingredients, including sucrose as the less expensive carbon source than glucose, was reported (Tokuda et al., 2011). That is, a recombinant strain of E. coli that harbors both a sucrose-6-phosphate hydrolase (CscA) gene and a DOI synthase (BtrC) gene was created, through which a cost-effective and scalable fermentation process that produces DOI from sucrose as the major ingredient of molasses, was next developed. Moreover, a novel DOI synthase that shows relatively higher heat resistance and pH stability than the existing DOI synthases was also discovered (Konishi and Imazu, 2010). DOI synthase derived from the Streptoalloteichus hindustanus JCM3268 strain was also discovered (Hirayama et al., 2006). Most recently, there was a report on overexpression of the above btrC gene in the photoautotrophic cyanobacterium Synechococcus elongatus, resulting in 400 mg/L DOI photosynthesis without any need for a carbon source for the recombinant (Watanabe et al., 2018).

As mentioned above, DOI synthase is a crucial enzyme that is involved in the beginning of the biosynthesis of 2 deoxystreptamine-containing aminoglycoside antibiotics. Therefore, the gene encoding DOI synthase must be present typically within a number of biosynthetic gene clusters essential for the biosynthesis of relevant antibiotics such as gentamicin, kanamycin, and tobramycin: genC originating from gentamicin-producing Micromonospora echinospora, kanC from kanamycin-producing Streptomyces kanamyceticus, and tobC from tobramycin-producing Streptomyces tenebrarius (Park et al., 2013). Herein, three codon-optimized artificial genes (genCopt, kanCopt, and tobCopt) were synthesized from the previously described genC, kanC, and tobC sequence templates (GenBank accession numbers AJ628149, AJ628422, and AJ810851, respectively) according to the codon usage preference of B. subtilis (**Figure 1**). Next, the glycolytic metabolic pathway in the heterologous host B. subtilis into which the abovementioned artificial genes were being separately introduced was engineered; either one or both genes pgi and pgcA, which encode Glc6P isomerase and phosphoglucomutase, respectively, was (or were) disrupted in the primary metabolic pathway of the host (**Figure 1**). On the other hand, a gene zwf encodes Glc6P-1 dehydrogenase which plays its role in the branched routes from Glc6P as an initiator of oxidative pentose phosphate pathway (**Figure 1**). Based on the previous report (Zamboni et al., 2004), the zwf knockout mutant showed about 1.5-folds reduced growth rate compared with the wild-type strain. However, in case of other two genes (pgi and pgcA), there has been no report for the negative effect on the host growth. Therefore, in this study, we constructed 1pgi and 1pgi1pgcA knockout mutants, in which the zwf gene is still intact. The present study involves the design and construction of a number of metabolically engineered recombinant strains of B. subtilis into which several natural and artificial DOI synthase–encoding genes were introduced. This approach should ultimately provide a microbial cell factory platform that can produce DOI with a high titer and productivity, as compared to the existing technology.

#### MATERIALS AND METHODS

#### Construction of Bacterial Strains and Plasmids

Strain B. subtilis 168 (genotype: trpC2) served as a negative control (Belda et al., 2013). A gene-targeting method that does not require selection markers (Fabret et al., 2002) was employed for the construction of recombinant strains of B. subtilis, in which either one or both pgi (Glc6P isomerase

gene) and pgcA (phosphoglucomutase gene) involved in the glycolysis metabolic pathway (KEGG pathway ID bsu00010) was (or were) disrupted in the genome of the host. After transformation of integration plasmid pCU-pgiFB (phenotype: Cm<sup>R</sup> [chloramphenicol resistance]) into B. subtilis strain 168, it was inserted into the B. subtilis genome via the first single-crossover recombination. The transformed recombinant 1pgi strain (genotype: trpC21pgi, phenotype: Cm<sup>R</sup> ) was selected on a Luria-Bertani (LB; BD Biosciences, Sparks, MD, United States) solid plate medium supplemented with 5 µg/mL chloramphenicol (Sigma-Aldrich, St. Louis, MO, United States). After cultivation in the liquid LB medium for 18 h, the recombinants were counterselected on the minimal medium (MM) supplemented with 10 µM 5-fluorouracil (Sigma-Aldrich). Similarly, another integration plasmid, pCU-pgcAFB (phenotype: Cm<sup>R</sup> ), was introduced into the recombinant 1pgi strain to obtain the recombinant 1pgi1pgcA strain (genotype: trpC21pgi1pgcA, phenotype: Cm<sup>R</sup> ) with both genes pgi and pgcA knocked out.

On the other hand, constitutive gene expression plasmid pHP13-P43 was constructed in the following manner. First, P43, one of the promoters originating from B. subtilis, was amplified by PCR using a forward primer, P43-F (50GGTA AAGCTTGCGGCTTCCTTGTAGAGCTCAG3<sup>0</sup> , underlined is a HindIII restriction enzyme cleavage site) and reverse primer P43- B (50CTCTCTGCAGCATGTGTACATTCCTCTC3<sup>0</sup> , underlined is a PstI site). P<sup>43</sup> promoter has been routinely utilized for the expression of heterologous gene in B. subtilis, as it is strong constitutive promoter (Song et al., 2016). After cleavage with both HindIII and PstI, the PCR products were ligated to B. subtilis expression plasmid pHP13 (Bacillus Genetic Stock Center, Columbus, OH, United States) that was digested with the same restriction enzymes, thus generating pHP13- P43 (genotype: P43, phenotype: Cm<sup>R</sup> , Em<sup>R</sup> [erythromycin resistance]).

The above-mentioned integration plasmid pCU-pgiFB was constructed in the following manner. The upper and lower DNA fragments of the pgi gene within the B. subtilis genome, pgi-F and pgi-B, respectively, were amplified from the B. subtilis 168 strain genome as a template with a pair of primers: D-pgi-FU/L (50CTAAACATGAACTGACAATTGAGGAAG3<sup>0</sup> ) and D-pgi-BU/L (50GAAGAAATATACAAGGTATCCAAAAGTATATG3<sup>0</sup> ). Then, after fusion of these two DNA fragments, fusion PCR was performed with the D-pgi-FsnU/L primer. The PCR products were cleaved with restriction enzymes SphI and KpnI, and pCU-pgiFB was generated by ligating the amplicons to pCU (phenotype: Cm<sup>R</sup> ) that was digested with the same restriction enzymes. The other integration plasmid pCU-pgcAFB was produced by fusion PCR with equivalent primers (D-pgcA-FU/L: 5 <sup>0</sup>TTAAGTTTATCGGTGAAAAGATTAAGGAATAC3<sup>0</sup> and D-pgcA-BU/L: 50AAAACCATATTCGTTAAAGAGATTGAT GAG3<sup>0</sup> ) and the same restriction enzyme sites.

The expression plasmid pHP13-P43-BtrC used for introducing btrC, a B. circulans-derived DOI synthase encoding gene, into B. subtilis strain 168 was assembled as follows. In other words, the genomic DNA of B. circulans NRRL B3312 was prepared as a template, and then subjected to PCR with forward primer BtrC-F (50GTGGGTACCGAGGTTAAA CATGACTAAAC3<sup>0</sup> , underlined is a KpnI site) and reverse primer BtrC-B (5<sup>0</sup> -CTCCTGCAGTTGTTATCGTGGATT AAATAATGG3<sup>0</sup> , underlined is a PstI site). After cleavage by both KpnI and PstI, the PCR products were ligated to expression plasmid pHP13-P43 that was digested with the same restriction enzymes, thereby yielding pHP13-P43-BtrC (genotype: P43-btrC, phenotype: Cm<sup>R</sup> , EM<sup>R</sup> ).

To amplify the expression of DOI synthase–encoding genes derived from microbes other than Bacillus (genC [Micromonospora echinospora DSM 43036, GenBank accession number AJ628149], kanC [Streptomyces kanamyceticus DSM 40500, GenBank AJ628422], and tobC [Streptomyces sp. DSM 40477, GenBank AJ810851]), the codon-optimized gene fragments (genCopt, kanCopt, and tobCopt) were designed and artificially synthesized by CosmoGenetech (Seoul, Korea) according to the codon usage preference of the heterologous host (B. subtilis). To all the codon-optimized gene fragments, we added a KpnI restriction site upstream of the ribosome-binding site and a PstI restriction site downstream of the termination codon, respectively. Three different artificial genes were treated with KpnI and PstI, and then ligated into plasmid pHP13-P43 that had been digested with the same restriction enzymes, thus generating pHP13-P43-GenCopt (genotype: P43-genCopt, phenotype: Cm<sup>R</sup> , EM<sup>R</sup> ), pHP13-P43-KanCopt (genotype: P43 kanCopt), and pHP13-P43-TobCopt (genotype: P43-tobCopt), respectively.

Gene manipulation was performed by standard techniques and transformation of both E. coli and B. subtilis was carried out by heat shock transformation and the natural competent transformation (Green and Sambrook, 2012). Antibiotics were added to the medium of recombinant strains at appropriate concentrations (kanamycin 5 µg/mL, ampicillin 100 µg/mL, and chloramphenicol 5 µg/mL; all from Sigma-Aldrich). All amplicons were routinely verified by sequencing.

# Shake-Flask Fermentation by the Recombinant Strains of B. subtilis

Unless specified otherwise, E. coli and recombinant B. subtilis strains were cultivated in the LB medium at 37◦C. To examine the growth of the recombinant strains and their DOI productivity, 20 mL of the 2YTG liquid medium (1.6% tryptone, 1% yeast extract, 0.5% NaCl, 3% glucose, all as w/v) was placed into a 250 mL baffled Erlenmeyer flask. After that, each recombinant strain was inoculated and cultivated for 12 h under the conditions mentioned above. A slightly modified 2YTG medium (1.6% tryptone, 1% yeast extract, 0.5% NaCl, 3% glucose, 2% glycerol; w/v) which contains, as another carbon source, 2% of glycerol in addition to glucose was prepared for a fed-batch fermentation process. After inoculation into 250 mL of the shake-flask fermentation medium in a 1 L baffled Erlenmeyer flask, each recombinant was cultivated for up to 60 h at 37◦C by reciprocal shaking at 200 rpm. Meanwhile, after 30 h of this shake-flask fermentation, 5 mL of a glucose solution (0.5 g/mL) that had been prepared under sterile conditions was added to the fed-batch culture to test whether the DOI productivity increases.

# Growth Curves of the Recombinant Strains and Analyses of Their DOI Titers

During the entire 60 h of fed-batch fermentation, an aliquot (2 mL) of fermentation broth was collected every 10 h to construct the growth curve in the following manner: 1 mL of serially diluted broth was placed into the UV-Vis spectrophotometer (Shimadzu, Japan) with absorbance measured at 680 nm. After refrigerated centrifugation of the same sample for 3 min at 5000 rpm (or ∼8000 × g), the precipitated cell debris were freeze-dried and then weighed. The linear curve of correlation between the dry cell weights and absorbance values was drawn, with which the dry cell weights of the recombinant strains of B. subtilis were determined during the fed-batch fermentation.

On the other hand, quantification of DOI produced via the fed-batch fermentation by the recombinants was performed by means of a modified version of the procedures described in other reports (Yamauchi and Kakinuma, 1992; Kharel et al., 2004) as follows: 100 µL of the supernatant obtained by the above-mentioned refrigerated centrifugation was mixed with 300 µL of an aqueous methanol solvent (methanol/water at 1:2, v/v) together with 40 µL of O-(4-nitrobenzyl) hydroxylamine hydrochloride (Sigma-Aldrich), 35 mg/mL. The oxime adducts of DOI were prepared via the reaction for 30 min in a water bath set to 60◦C (Kharel et al., 2004), and then the reactants were evaporated to dryness at room temperature by vacuum centrifugation. The DOI derivatives were reconstituted in 100 µL of methanol, and an aliquot (25 µL) was subjected to ultra highperformance liquid chromatography (UPLC) with electrospray ionization (ESI) and ion trap tandem mass spectrometry (MS/MS) analysis. The Spectra system P1000XR UPLC consists of a pump (Thermo Finnigan, San Jose, CA, United States) and a Spectra Series AS3000 autosampler (Thermo Finnigan, 20 µL loop). Isocratic chromatography was conducted on an Acquity CSH C<sup>18</sup> (Waters, Milford, MA, United States) reversedphase column at a flow rate of 120 µL/min of the mobile phase (acetonitrile/methanol/water/formic acid 1:1.5:8:0.002, v/v/v/v). The column effluent was directed to an LCQ ion trap mass spectrometer (Thermo Finnigan), operated in positive ion mode. The mass transition specific to DOI oxime adducts was m/z 315.3 > 164.3 (**Figure 2A**). Quantification of DOI produced during the fed-batch fermentation by each recombinant was carried out according to the calibration equation obtained from the correlation between the chromatographic peak areas and the concentration of the oxime adducts of an authentic DOI standard (GeneChem Inc., Daejeon, Korea). Each fermentation broth sample was also centrifuged, and then the supernatant of the sample was processed using two kinds of commercially available kits such as the D-glucose HK assay kit (Megazyme International Ltd., Wicklow, Ireland) and glycerol determination

kit (Sigma-Aldrich) to determine the residual concentration of glucose and glycerol that had been added as carbon sources during fed-batch fermentation.

Growth curves and the DOI titer of recombinant strains of B. subtilis, together with the remaining glucose and glycerol content in the fed-batch fermentation process, were presented by averaging the results in quadruplicate.

# RESULTS AND DISCUSSION

shake-flask fermentation broths of the recombinants.

### Comparison of the DOI Titer Produced by the Recombinants That Express a Heterologous DOI Synthase

Among the recombinant B. subtilis strains that carried out 60 h of shake-flask fermentation as described before, DOI productivity of four recombinant strains that expressed DOI synthase solely without further metabolic pathway engineering was determined and compared with that of strain BSDOI-01 as a negative control (**Table 1**). When a btrC gene that encodes the DOI synthase from B. circulans was expressed in the B. subtilis host (BSDOI-02), an average of 2.3 g/L DOI was produced. A recent publication revealed that DOI production of 1.5 g/L can be achieved through expression of the identical btrC gene in E. coli as a host (Kogure et al., 2007). Meanwhile, when an artificial gene—genCopt synthesized on the basis of a previously described genC sequence template—was introduced into the host (BSDOI-03), DOI titer in the shake-flask fermentation broth remained at 0.8 g/L on average, showing a failure of the enhancement of DOI productivity through genCopt expression. In contrast, heterologous expression of kanCopt and tobCopt, both of which originate from the Streptomyces genus, in B. subtilis hosts (BSDOI-04 and BSDOI-05) increased the DOI titer up to 3.4 and 3.6 g/L, respectively (**Figure 2B**). In particular, the GC contents of two codon-optimized artificial genes kanCopt and tobCopt (from Streptomyces) are 52 and 59% respectively, whereas that of genCopt (from Micromonospora) is 68% even being synthesized according to the codon-usage preference in B. subtilis host, suggesting the negative effect of high GC-content onto the gene expression in heterologous host. Therefore, we found that the expression of kanCopt or tobCopt improved the DOI titer as compared with btrC, which has been in common use according to the existing publications (Kogure et al., 2007; Tokuda et al., 2011; Watanabe et al., 2018). Thus, it was suggested that the expression of DOI synthases (orthologous to different species) that are involved in the same catalytic reaction could yield diverse DOI titers, whereas it was found that an improved DOI titer, as compared to the results from preceding studies (Kogure et al., 2007; Tokuda et al., 2011), could be achieved through the expression of artificially synthesized genes based on synthetic biology that takes into account codon usage preferences of the recombinant strains.

# Comparison of DOI Titers Produced by the Metabolically Engineered Recombinant Strains That Express btrC

DOI productivity of btrC-expressing recombinant strains (BSDOI-07 and BSDOI-12), in which either one or both pgi (Glc6P isomerase gene) and pgcA (phosphoglucomutase gene) involved in the glycolysis metabolic pathway was (or were) disrupted in the genome of B. subtilis as a host, was compared with that of the btrC-expressing BSDOI-02. BSDOI-07 represents

the btrC-expressing 1pgi strain, whereas BSDOI-12 denotes the btrC-expressing 1pgi1pgcA strain of B. subtilis. The resulting strains BSDOI-07 and BSDOI-12 showed average DOI productivity of 16.7 and 20.7 g/L, respectively, which was more than seven- and ninefold higher than the DOI productivity (2.3 g/L) of the control (BSDOI-02), which did not undergo a metabolic pathway modification (**Table 1**). In other words, when two kinds of enzymes (such as Glc6P isomerase and phosphoglucomutase that utilize Glc6P as the common substrate within the branched glycolytic metabolic pathway) were deleted in the host, intracellular accumulation of Glc6P as a typical substrate of DOI synthase was induced, thus leading to significantly elevated DOI productivity through catalysis by the BtrC enzyme. Moreover, the aforementioned three kinds of recombinant strains did not show noticeable differences in cell growth during the shake-flask fermentation process.

### Comparison of DOI Titers Among the Metabolically Engineered Recombinant Strains That Express Codon-Optimized Artificial Genes

A natural btrC gene introduced into the 1pgi strain (BSDOI-7) as described above was replaced by three different kinds of codon-optimized artificial genes such as genCopt, kanCopt, and tobCopt, yielding recombinant strains BSDOI-08 to BSDOI-10. Moreover, a similar gene replacement process was carried out in the 1pgi1pgcA strain (BSDOI-12), thereby generating recombinant strains BSDOI-11 to BSDOI-13. The DOI titer produced by the recombinants was compared with that of two recombinant strains: BSDOI-07 and BSDOI-12. At first, among the recombinant strains with single deletion of the pgi gene, additional expression of artificial gene genCopt showed a noticeably decreased DOI titer (5.5 g/L) relative to the btrCexpressing BSDOI-07 strain (**Table 1**). Nonetheless, replacement of btrC with other codon-optimized DOI synthases (kanCopt and

TABLE 1 | 2-Deoxy-scyllo-inosose titers shown by the recombinant B. subtilis strains during fed-batch fermentation.


tobCopt) showed an improved DOI titer (average of 22.5 and 24.2 g/L, respectively) as compared to the control (average of 16.7 g/L). Similarly, recombinant strains with double deletion of genes pgi and pgcA showed a similar pattern of DOI titer; in particular, expression of artificial gene tobCopt in the abovementioned recombinant host further increased DOI production to 37.2 g/L, on average (**Table 1** and **Figure 2**). Judging by the results above, engineering the glycolytic metabolic pathway in the host enables intracellular accumulation of Glc6P, and furthermore, the highest DOI titer and productivity were accomplished by means of artificial genes, as compared to other studies (Kakinuma et al., 2000; Kogure et al., 2007; Tokuda et al., 2011) that employed enzymatic reactions or an engineered recombinant strain of E. coli. Therefore, by combining the synthetic biology approach that spurs DOI biosynthesis in a metabolically engineered heterologous host E. coli strain, a maximum DOI production of 37.2 g/L was achieved. Hence these results highlight the above-mentioned 16 fold higher DOI titer compared with that of the btrC-expressing recombinant.

# Fed-Batch Fermentation by the BSDOI-15 Recombinant Strain

To examine time courses of cell growth and DOI production during the fermentation driven by a resultant engineered BSDOI-15 cell factory, the culture broth was collected every 10 h. The dry cell weight and the fermentation profiles of glucose and glycerol, together with the DOI titer were quantitatively determined (**Figure 3** and **Supplementary Figures 1**, **2**). Bacterial cell growth showed a typical fed-batch culture pattern, indicating that besides glucose, glycerol can serve as an extra carbon source for the growth of these DOI-producing cell factories. In fed-batch fermentation by the BSDOI-15 recombinant where glucose was added after 30 h fermentation, the profile of DOI production appears to closely correlate with that of glucose consumed. However, by the first 20 h of fed-batch fermentation, the concentration of glucose consumed was about 17.6 g/L, whereas DOI production was up to 27.9 g/L. Ideally, there should be one to one correspondence between the both concentrations. These discrepancies in reciprocal correlations between the glucose consumption and DOI production within 20 h fermentation might be due to the usage of complex media composition yeast extract. Further studies using chemically define medium, instead complex medium, could be a clue to the above question. Meanwhile, contrary to the sharp glucose consumption rate during the initial 20 h fermentation, the consumption rate of glycerol appears to be gradual slope; the intactness of Zwf on the pentose phosphate pathway during fed-batch fermentation will make extra flux to glyceraldehyde-3-P along with its precursors, thus causing catabolite repression of glycerol. A maximum DOI production of 38.0 g/L was reached at 50 h of fermentation, and then the DOI titer slightly decreased up to hour 60. Cell growth also seems to be in the stationary phase after 50 h fermentation, thus suggesting that there may be some relations between DOI production and cell stability in fed-batch fermentation. During the total 60 h of fermentation, the initial glucose level was set

to 30 g per liter of a culture medium. Considering that 10 g was additionally fed into the culture after 30 h fermentation, a total of 40 g of glucose as the main carbon source was converted to 38.0 g of DOI, meaning that a yield of approximately 95% was achieved. The yield of 95% seen in the present study is lower than the 99% yield achieved in a metabolically engineered recombinant strain of E. coli into which a natural DOI synthase–encoding btrC gene was introduced (Kogure et al., 2007), but the DOI titer was higher than what was reported in this previous publication (i.e., 38.0 vs. 29.5 g/L). Furthermore, considering the DOI productivity based on total fermentation time (i.e., 50 vs. 72 h), the DOI productivity obtained in this study was 0.76 g/(L.h) compared with 0.41 g/(L.h) in the previous study on the engineered E. coli recombinant. Our result represents ∼1.8-fold enhancement of DOI productivity. Meanwhile, when the titers obtained from shake-flask and fedbatch fermentations were compared, there was no significant difference (i.e., 37.2 vs. 38.0 g/L). But, considering with the timedependent productivity (i.e., 60-h vs. 50-h), the DOI productivity during shake-flask fermentation [0.62 g/(L.h)] was meaningfully lower than fed-batch fermentation [0.76 g/(L.h)]. Consequently, it was confirmed that the expression of a codon-optimized DOI synthase–encoding tobCopt gene in a metabolically engineered cell factory of B. subtilis—in which both the Glc6P isomerase pgi gene and phosphoglucomutase pgcA gene (involved in the glycolytic metabolic pathway) are disrupted—led to an approximately 38 g/L DOI titer within 50 h of fermentation that employs glycerol (besides glucose) as an extra carbon and energy source for growth. Namely, mass production of the desired DOI could be attained via the fed-batch fermentation by the engineered cell factory constructed through convergence of metabolic engineering and synthetic biology.

### CONCLUSION

Here, we report that the titer, yield, and productivity of DOI obtained via fed-batch fermentation by newly engineered B. subtilis cell factories are at least equivalent to those in another study [on engineered E. coli recombinants] (Kogure et al., 2007). For the production of DOI via fermentation by the recombinants constructed herein, we employed a dual carbon source: glucose and glycerol. Besides, in the case of the abovementioned study on engineered E. coli recombinants, glucose and mannitol (up to 4%, fed into the fermentation medium) were utilized for DOI production. Glycerol has been generated as the main byproduct of the manufacture of biodiesel, one of renewable energy sources for the substitution of petroleum, thus pointing to the potential utilization of glycerol as a carbon or energy source for industrial fermentation (da Silva et al., 2009). Hence, in comparison with cost-ineffective mannitol, glycerol is surely a favorable and cost-effective carbon source for industrial-scale fermentation. Although further optimization of the fermentation parameters (i.e., the ratio of carbon sources or the cell mass or density corresponding to the modification of carbon sources, and the time point of feeding) is required, it will be useful to determine whether other cost-effective and

sustainable carbon sources can be utilized to reach an equally high titer or productivity by means of these engineered recombinants. In this study, for the expression of natural and artificial DOI synthase-encoding genes, a strong constitutive P<sup>43</sup> promoter was employed as the control. However, the recent studies on promoter screening or its engineering for fine-tuned gene expression in B. subtilis host gave us a more details on the host expression patterns (Song et al., 2016; Liu et al., 2018), which are essential for synthetic biology and metabolic engineering approaches. Development of other inducible promoters and their application onto our host could be useful for industrial production of DOI. In addition, owing to the slight drop of the DOI titer after 50 h of fermentation, monitoring of DOI profiles during the fed-batch fermentation may be necessary for the maintenance of DOI stability. In particular, DOI synthase acts on Glc6P to biosynthesize DOI with the help of cofactor nicotinamide adenine dinucleotide (NAD) (Huang et al., 2005; Nango et al., 2008). Redox engineering has often been carried out and applied to various desired products (Qiao et al., 2017). Therefore, as a further study, redox engineering, or subsequent engineering of the NAD recycling (or regeneration) pathway in the B. subtilis recombinants constructed in this study, may be tested as a potential booster not only to enhance the DOI titer or productivity but also to develop cell factory platforms relevant to DOI-based pharmaceuticals and fine chemicals.

#### AUTHOR CONTRIBUTIONS

fmicb-09-02333 September 25, 2018 Time: 18:0 # 8

JP and YY conceived the project and wrote the manuscript. JHL, HH, NL, JWL, ES, HS, and HK designed and conducted all the experiments. JHL, HH, NL, YY, and JP analyzed the results.

#### FUNDING

This work was supported by the Cooperative Research Program for Agriculture Science & Technology Development [Grant No.

#### REFERENCES


PJ01317901] funded by the Rural Development Administration; the National Research Foundation of Korea [Grant Nos. 2015R1A2A2A01002524 and 2016 R1A2A1A05005078 (YY)] funded by the Ministry of Science, ICT and Future Planning, Republic of Korea.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmicb. 2018.02333/full#supplementary-material


**Conflict of Interest Statement:** 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 © 2018 Lim, Hwang, Lee, Lee, Seo, Son, Kim, Yoon and Park. 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.

# Single Amino Acid Substitution in Homogentisate Dioxygenase Affects Melanin Production in Bacillus thuringiensis

Wenjun Yang<sup>1</sup> , Lifang Ruan<sup>1</sup> , Jiangming Tao<sup>1</sup> , Donghai Peng<sup>1</sup> , Jinshui Zheng1,2 and Ming Sun<sup>1</sup> \*

<sup>1</sup> State Key Laboratory of Agricultural Microbiology, College of Life Science and Technology, Huazhong Agricultural University, Wuhan, China, <sup>2</sup> College of Informatics, Huazhong Agricultural University, Wuhan, China

#### Edited by:

Eung-Soo Kim, Inha University, South Korea

#### Reviewed by:

Gabriela Olmedo-Alvarez, Cinvestav Unidad Irapuato, Mexico Vania Aparecida Vicente, Universidade Federal do Paraná, Brazil

> \*Correspondence: Ming Sun m98sun@mail.hzau.edu.cn

#### Specialty section:

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

> Received: 24 May 2018 Accepted: 03 September 2018 Published: 11 October 2018

#### Citation:

Yang W, Ruan L, Tao J, Peng D, Zheng J and Sun M (2018) Single Amino Acid Substitution in Homogentisate Dioxygenase Affects Melanin Production in Bacillus thuringiensis. Front. Microbiol. 9:2242. doi: 10.3389/fmicb.2018.02242 Bacillus thuringiensis formulation losing its activity under field conditions due to UV radiation and photoprotection of B. thuringiensis based on melanin has attracted the attention of researchers for many years. Here, a single amino acid substitution (G272E) in homogentisate 1,2-dioxygenase was found to be responsible for pigment overproduction in B. thuringiensis BMB181, a derivative of BMB171. Disrupting the gene encoding homogentisate dioxygenase in BMB171 induced the accumulation of the homogentisic acid and provoked an increased pigment formation. To gain insights into homogentisate 1,2-dioxygenase in B. thuringiensis, we constructed a total of 14 mutations with a single amino acid substitution, and six of the mutant proteins were found to affect the melanin production when substituted by alanine. This study provides a new way to construct pigment-overproducing strains by impairing the homogentisate dioxygenase with a single mutation in B. thuringiensis, and the findings will facilitate a better understanding of this enzyme.

Keywords: Bacillus thuringiensis, tyrosine catabolism, homogentisate dioxygenase, site-directed mutagenesis, melanin

# INTRODUCTION

Bacillus thuringiensis, a gram-positive spore-forming soil bacterium, has been widely used in biological pest control due to the formation of parasporal crystal proteins that are toxic to the larvae of various insect pests (Sudakin, 2003). However, the insecticidal activity of the crystal proteins would be reduced or destroyed under field conditions because of UV damage in sunlight (Pusztai et al., 1991). To solve this problem, researchers have proposed a series of strategies to protect insecticidal crystal proteins from UV damage (Sanchis et al., 1999; Manasherob et al., 2002; Jallouli et al., 2014), and one of them is the use of melanin, a photoprotective agent, to reduce the damaging effect of UV radiation on B. thuringiensis toxins (Liu et al., 1993; Ruan et al., 2002; Zhang et al., 2008; Sansinenea and Ortiz, 2015).

Melanins are polymers of phenolic and/or indolic compounds and classified into three main categories: eumelanins, pheomelanins, and allomelanins (Plonka and Grabacka, 2006). These black pigments are widely distributed in nature and can be found in species of all biological kingdoms, including humans, fungi, and bacteria (Plonka and Grabacka, 2006). Melanins provide free-living species a survival advantage in the environment by protecting against different exogenous stresses, such as UV-irradiation, reactive oxygen species (ROS), metals, and defensins (Nosanchuk and Casadevall, 2003, 2006; Heinekamp et al., 2012). Both eumelanins and pheomelanins are produced from the oxidation of tyrosine or phenylalanine to o-dihydroxyphenylalanine (DOPA) and dopaquinone via tyrosinases or laccases. Allomelanins include a heterogeneous group of polymers formed through the oxidation and polymerization of the intermediates such as dihydroxynaphthalene, homogentisic acid (HGA), γ-glutaminyl-4-hydroxybenzene, catechols, and 4-hydroxyphenylacetic acid (Plonka and Grabacka, 2006).

Homogentisic acid is derived from the tyrosine or phenylalanine catabolism pathway (one branch of tyrosine metabolism) and further oxidized to acetoacetic acid and fumaric acid (Turick et al., 2010). Pyomelanin is formed from the autoxidation and self-polymerization of HGA with the deactivation of homogentisate dioxygenase (HmgA) or the disruption of the gene encoding HGA-oxidase (Rodriguez-Rojas et al., 2009; Schmaler-Ripcke et al., 2009; Valeru et al., 2009). A deficiency of this enzyme in humans causes the metabolic disease alkaptonuria (AKU), leading to the excretion of HGA in the urine in a large amount and its deposition in different tissues (Millucci et al., 2012). The synthesis of pyomelanin has been investigated in a broad range of bacteria, such as Aeromonas media, Burkholderia cepacia, Bacillus anthracis, Legionella pneumophila, Pseudoalteromonas, Pseudomonas aeruginosa, Pseudomonas putida, Ralstonia solanacearum, Shewanella colwelliana, and so on (Arias-Barrau et al., 2004; Valeru et al., 2009; Turick et al., 2010; Gonyar et al., 2015; Han et al., 2015; Wang et al., 2015; Ahmad et al., 2017; Zeng et al., 2017).

A number of studies have been performed to improve the photoprotection of the B. thuringiensis crystal proteins by producing melanin or increasing the melanin yield in B. thuringiensis cells. Some studies focused on screening B. thuringiensis mutants that can produce melanin in different conditions, and others attempted to construct recombinant B. thuringiensis strains with melanin production by genetic engineering (Hoti and Balaraman, 1993; Ruan et al., 2002; Saxena et al., 2002; Zhang et al., 2008; Sansinenea and Ortiz, 2015). It has been found that melanin could be produced by most B. thuringiensis strains in the presence of L-tyrosine at an elevated temperature (42◦C), but the insecticidal Cry proteins could not be synthesized at this temperature (Ruan et al., 2004). In our previous work, a mutant, B. thuringiensis strain BMB181, was identified to be able to produce the brownish black pigment as an alternative melanin without tyrosine supplementation in the growth medium, and this strain could achieve a high melanin yield in different media without additional L-tyrosine (Liu et al., 2013). However, the mechanism for melanin production by BMB181 remains unclear. Here, the pigment produced by the strain BMB181 was found to be derived from homogentisate acid. The inactivation of HmgA by a G272E amino acid substitution resulted in pigmentation in the strain BMB181. Six single-point mutations in HmgA resulted in observable changes of melanin production in B. thuringiensis. This study offers valuable information about HmgA and provides a new way of constructing the pigment production strain of B. thuringiensis.

# MATERIALS AND METHODS

# Bacterial Strains, Plasmids, and Growth Conditions

The bacterial strains and plasmids used in this study are shown in **Table 1**. The B. thuringiensis strain BMB171 and its derivative BMB181 have been reported in previous studies (He et al., 2010; Liu et al., 2013). Bacteria were grown in Luria-Bertani (LB) medium at 37◦C (E. coli) or 28◦C (B. thuringiensis) and 200 rpm. The antibiotics, including kanamycin (50 µg/mL), erythromycin (25 µg/mL), and ampicillin (100 µg/mL), were added when necessary.

# DNA Manipulation and Sequence Analysis

In this study, DNA was manipulated using the standard techniques as previously described (Green and Sambrook, 2012). The DNA of B. thuringiensis was extracted as previously reported (Andrup et al., 1993). The DNA fragments were amplified with the related primers, and the polymerase chain reaction (PCR) products were confirmed by DNA sequencing. The DNA sequences were analyzed using the DNASTAR software and the protein sequences were compared with those of other proteins using BLAST and CLUSTALX.

# Transformation Techniques

Escherichia coli was transformed using CaCl2-treated competent cells (Green and Sambrook, 2012), and B. thuringiensis was transformed by electroporation with the Bio-Rad Gene Pulser set (Bio-Rad, Hercules, CA, United States) (Peng et al., 2009).

# High Performance Liquid Chromatography (HPLC) Analysis of Culture Filtrates

The HPLC analysis of culture filtrates was performed using an HPLC apparatus equipped with a variable wavelength UVvisible detector (CapLC 2487, Waters) and a C-18 end-capped column (10 µm, 4.6 mm × 150 mm; Elite). Briefly, the bacteria were grown in LB medium at 28◦C and 200 rpm, and the culture supernatants were collected by centrifugation and filter sterilization. Next, 20 µL of the sample was injected into the column for HPLC. The mobile phase was 50 mM sodium phosphate buffer (pH 6.5)/methanol (80:20, v/v) at a flow rate of 0.5 mL/min as described by Fernandez-Canon and Penalva (1995). The chromatograms of standard solutions of HGA (from Sigma) were used as a reference to identify the corresponding HPLC peaks. The absorption maximum of HGA is 290 nm.

TABLE 1 | Bacterial strains and plasmids used in this study.

fmicb-09-02242 October 10, 2018 Time: 11:20 # 3


<sup>a</sup>Amp<sup>r</sup> , ampicillin resistance; Erm<sup>r</sup> , erythromycin resistance; Kan<sup>r</sup> , kanamycin resistance.

#### Isolation of the hmgA Gene

The hmgA gene (1,173 bp) encoding HmgA was amplified from B. thuringiensis BMB171 and BMB181 genomic DNA using the pair of primers, hmgA1 (5<sup>0</sup> -CGCGGATCCATGTTTTATCGT CACATGGGAG-3<sup>0</sup> , with the BamHI recognition sequence underlined) and hmgA2 (5<sup>0</sup> -CGCGTCGACTTATTTCACAG TATATGAACCT-3<sup>0</sup> , with the SalI recognition sequence underlined). The hmgA genes from BMB171 and BMB181 were designated as 171hmgA and 181hmgA, respectively.

## Insertional Inactivation of the hmgA Gene

To verify the function of the hmgA gene, the gene disruption strain was constructed for BMB171 via homologous recombination using a temperature-sensitive shuttle vector pHT304-ts containing the temperature-sensitive replication origin (Liu et al., 2010). Briefly, a 503-bp fragment and a 600-bp fragment, corresponding to the DNA regions upstream and downstream of the open reading frame of the hmgA gene in the strain BMB171, were generated by PCR using the primer pairs HmgA-up-1 (50 -CGCGGATCCTGGGAGAACTACCTCATAAAC-3<sup>0</sup> )/HmgA -up-2(5<sup>0</sup> -CCGGAATTCGCTATTCGCCTCTACAACA-3<sup>0</sup> ) and HmgA-down-1(5<sup>0</sup> -CCGGTCGACAATTGTTAGAGCATAGTCC G-3<sup>0</sup> )/HmgA-down-2(5<sup>0</sup> -CGGGGTACCATGAACCTTGTTCAA

TCCAG-3<sup>0</sup> ) and digested with BamHI- EcoRI and AccI-KpnI, respectively. A kanamycin resistance cassette (1,514 bp) was acquired by digesting plasmid pDG780 (Guerout-Fleury et al., 1995) with EcoRI-AccI. These three fragments were cloned into the plasmid pHT304-ts at the BamHI-KpnI site. The resulting plasmid, pBMB3143, was transformed into the strain BMB171 by electroporation.

The mutants were selected as follows. Specifically, the transformants were cultured in LB medium with kanamycin (50 µg/mL) for 8 h, followed by incubation at 42◦C for 4 days to eliminate unintegrated temperature-sensitive plasmids. Finally, the mutant strains that were resistant to kanamycin but sensitive to erythromycin colonies were confirmed by PCR using appropriate primers and sequencing.

## Genetic Complementation Analysis

For genetic complementation analysis, complementation plasmids were prepared by using the shuttle vector pHT304 (Arantes and Lereclus, 1991) and its derivative pBMBL (unpublished data) that contained a sporulationdependent promoter BtI-BtII and the cry1Ac transcription terminator. The amplified fragment 171hmgA was cloned into the vector pBMBL to yield the plasmid pBMB3141. To construct the plasmids pBMB3144 (carrying 171hmgA

operon) and pBMB3145 (carrying 181hmgA operon), the hmgA gene operon region (about 3.9 kb) containing the genes encoding 4-hydroxyphenylpyruvate dioxygenase (HppD), HmgA, and fumarylacetoacetate hydrolase (FahA) was amplified, respectively, from the strains BMB171 and BMB181 using special primers HMGAop-S (5<sup>0</sup> - CGCGGATCCAGATATATAAATACAATCATTC-3<sup>0</sup> , with the BamHI recognition sequence underlined) and HMGAop-A (50 -CGCGTCGACTCTTTCACTCCTCCAAGTTT-3<sup>0</sup> , with the SalI recognition sequence underlined) and subsequently cloned into pHT304. Finally, the two recombinant plasmids were transformed into the pigmented B. thuringiensis strains by electroporation separately.

# Measurement of the Bacterial Growth Curve and the Pyomelanin Production

The growth curve and the pigment production of the B. thuringiensis strains and their derivatives were evaluated according to optical density (OD). To monitor the growth curves of the strains, the bacteria were inoculated to 100 mL of LB medium (the flask volume is 500 mL) and incubated under shaking at 28◦C and 200 rpm, followed by the OD measurement of the cultures at 600 nm (OD600) at different time intervals. The melanin production of the strains was quantified by testing the absorbance of the centrifuged culture supernatant at 400 nm (OD400) at the indicated time points (Ruan et al., 2002; Liu et al., 2013).

# Alanine Scanning Site-Directed Mutagenesis

For performing alanine scanning site-directed mutagenesis, we constructed the plasmid pBMB3146 (a derivative of pHT304, containing the promoter region and the terminator region of the hmgA gene operon and the 171hmgA gene). Briefly, the promoter region of the hmgA gene operon (394 bp) was amplified from the strain BMB171 genomic DNA using the pair of primers, HMGAop-S (5<sup>0</sup> -CGCGGATCCAGATATATAAATACA ATCATTC-3<sup>0</sup> , with the BamHI recognition sequence under lined) and HMGAop-A2 (5<sup>0</sup> -CTCCCATGTGACGATAAAAC ATAATATCTTCATCTCCCTGTAA-3<sup>0</sup> ). Next, the 1,405-bp PCR product, containing the 171hmgA gene and the terminator region of the hmgA gene operon, was amplified using primers HMGA-S2 (5<sup>0</sup> -TTACAGGGAGATGAAGATATTATGTTTTAT CGTCACATGGGAG-3<sup>0</sup> ) and HMGAop-A (5<sup>0</sup> -CGCGTCGAC TCTTTCACTCCTCCAAGTTT-3<sup>0</sup> , with the SalI recognition sequence underlined). Finally, the vector pBMB3146 was generated by connecting the two fragments with overlapping PCR (Higuchi et al., 1988) using primers HMGAop-S and HMGAop-A, digesting the overlapping PCR fragment with BamHI and SalI, and inserting it into the shuttle vector pHT304.

A total of fourteen amino acid residues of HmgA were mutated to alanine residues by overlapping PCR using the primers shown in **Table 2**, with vector pBMB3146 as the template. The PCR fragments were cloned separately into the vector pHT304 between the BamHI and SalI restriction sites. The recombinant plasmids were transformed into E. coli DH5a and the positive clones were screened by restriction enzyme analysis. TABLE 2 | Primers used for site-directed mutagenesis.


<sup>a</sup>The restriction sites included in the oligonucleotide sequences are underlined. Nucleotide codons encoding the mutated amino acids are highlighted in bold types.

Before transformation into the pigmented B. thuringiensis strain BMB1711hmgA, all the resulting mutant plasmids were sequenced to ensure that the proper mutations were maintained.

#### Sequence Accession Number

The sequences reported in this paper have been submitted to the GenBank. The accession number for the complete genome of the strain BMB171 is CP001903.1. The accession number for the HmgA (locus tag is BMB171\_C0216) is ADH05034.1.

# RESULTS

#### Pigment Results From Polymerization of Homogentisate

harbors 171hmgA; BMB3142, a derivative of BMB181, contained pBMBL, as a negative control.

BMB181, a B. thuringiensis mutant with high melanin production, was obtained after subculturing the strain BMB171 for several generations at 42◦C (Liu et al., 2013). The red pigment produced by the strain BMB181 turned dark brown with the extension of incubation time (**Figure 1A**). Pigments can be formed from the oxidation and polymerization of compounds such as DOPA and HGA (Plonka and Grabacka, 2006). In our early work, we found that the pigment produced by the strain BMB181 has nothing to do with DOPA (data not shown). Therefore, we test whether HGA is the precursor of pigment produced by the strain BMB181. Here, ascorbic acid was added as an antioxidant to prevent HGA from oxidation. No pigment was observed when ascorbic acid (2 mM) was added to the cultures of the strain BMB181 (data not shown). Strains BMB171 and BMB181 were cultured in LB liquid media under shaking at 28◦C for 24 h and the culture samples were taken for HPLC analysis after centrifugation and filtration, using the commercially available HGA as the standard. A peak corresponding to HGA (with a retention time of 8.68 min) was identified in the culture supernatants of the strain BMB171 with HGA added to the culture during the logarithmic growth phase (**Figures 2A,B**). The peak with the same chromatography retention time as HGA was identified in the culture supernatants of the strain BMB181 (**Figure 2C**), suggesting that homogentisate could be secreted by the pigmented strain BMB181 and the pigment produced by the strain probably resulted from the accumulation and polymerization of homogentisate.

# Identification of Amino Acid Substitution in HmgA

Considering that the melanin production from homogentisate induced by a deficiency of HmgA in organisms is associated with the tyrosine metabolism pathway (**Figure 3A**), we analyzed the genome of BMB171 and found that BMB171 carried the genes encoding HppD, HmgA, and FahA (**Figure 3B**). The relationship between the biosynthesis of the pigment in the strain BMB181 and HmgA was tested by amplifying the hmgA gene from the BMB181 genomic DNA and sequencing. After aligning

the inferred amino acid sequence of HmgA from the publicly available BMB171 sequences (He et al., 2010), a glycine was found to be replaced by a glutamate at residue 272 in HmgA in the pigmented strain BMB181.

Whether this amino acid substitution is responsible for the observed pigmented phenotype was tested by performing a complementation analysis. Specifically, the plasmid pBMB3141 containing the 171hmgA gene was transformed into BMB181 to create the transformant strain BMB3141 (171hmgA+), and BMB3142, a BMB181 derivative containing the plasmid pBMBL was used as a control. It was found that BMB3141 (171hmgA+) was reverted to a non-pigmented phenotype, with no significant increase observed in the OD<sup>400</sup> of the supernatant relative to the control (**Figure 1B**). These results indicate that the G272E version of HmgA is responsible for melanin production in the B. thuringiensis pigmented strain BMB181.

# Disruption of HmgA Results in Pigment Formation Due to Inactivation of Homogentisate Dioxygenase

To verify that the lack of a functional HmgA is responsible for the pigment formation of B. thuringiensis, a hmgA insertion mutant strain was constructed by homologous recombination. Specifically, the plasmid pBMB3143 containing hmgA gene fragments flanking the kanamycin resistance cassette was constructed and transferred into the strain BMB171. The recombinants with double-crossover homologous recombination integration in the resident hmgA gene were selected and verified by PCR. The strain carrying the hmgA::kan disruption, named BMB1711hmgA, was able to produce the brownish black pigment in cultures (**Figure 4A**). HPLC analysis showed that HGA was secreted and accumulated in the culture supernatants of BMB1711hmgA (**Figure 2D**).

The complete hmgA gene operons from the strains BMB171 and BMB181 were amplified and used for the complementation analysis, and the plasmids containing different alleles of hmgA were named as pBMB3144 (carrying the operon containing the hmgA gene from BMB171) and pBMB3145 (carrying the operon containing the hmgA gene from BMB181). The plasmids expressing either allele of hmgA were transferred into the strain BMB1711hmgA (pigmented phenotype). It was found that the transformant harboring the 171hmgA gene operon (strain BMB3144) exhibited a non-pigmented phenotype, while the transformant harboring the 181hmgA variant operon (strain BMB3145) remained pigmented (**Figure 4B**). The same result was found when the two plasmids were introduced into the strain BMB181 (**Figure 4C**). These data suggest that the HmgA variant from the BMB181 mutant with residue Gly272 replaced by the residue Glu is not functional and ultimately results in pigment overproduction, while the functional HmgA enzyme from the strain BMB171 results in the absence of pigment from HGA.

## Mutational Analysis of HmgA in B. thuringiensis

To further test whether other single residue mutations have the same effect as G272E on melanin production, we designed 14

BMB1711hmgA. Homogentisate acid was identified by HPLC and is indicated by black arrows. (A) Supernatants of BMB171 (with ascorbic acid), as a negative control. (B) Supernatants of BMB171 (with ascorbic acid and Homogentisate), as positive control. (C) Supernatants of BMB181 (with ascorbic acid). (D) Supernatants of BMB1711hmgA (with ascorbic acid).

single-point mutations (G89A, N116A, G119A, D120A, G128A, F136A, K219A, G241A, P245A, H261A, N300A, G323A, H334A, and P336A) in 171HmgA based on the sequence alignment and the secondary structure of 171HmgA (**Figure 5**). A transcomplementation test was performed by introducing all the sequences under the control of the shuttle vector pHT304 separately into the pigmented strain BMB1711hmgA. Among the fourteen transformants, eight (containing G89A, N116A, G119A, D120A, K219A, P245A, N300A, and G323A, respectively) showed a non-pigmented phenotype, while the other six (containing G128A, F136A, G241A, H261A, H334A, and P336A, respectively) showed a pigmented phenotype (**Table 3**), implying that these six amino acid substitutions could result in the function loss of HmgA, making it unable to restore the non-pigment phenotype.

#### DISCUSSION

In aerobic organisms, L-Tyrosine degradation via homogentisate (HGA) is initiated by the conversion of tyrosine to 4 hydroxyphenylpyruvate by tyrosine aminotransferases, followed by the formation of HGA from 4-hydroxyphenylpyruvate

(Continued)

#### FIGURE 3 | Continued

fmicb-09-02242 October 10, 2018 Time: 11:20 # 8

HmgA, homogentisate 1,2-dioxygenase; BMB171\_C0217, amino acid permease; BMB171\_C0218, MFS transporter. The vector pBMB3141, a derivative of pBMBL, contained the hmgA gene (1,173 bp) amplified from the genomic DNA of BMB171 by using primers hmgA1 and hmgA2. The vector pBMB3144 carried the hmgA gene operon from BMB171, and the vector pBMB3145 carried the hmgA gene operon from BMB181. The hmgA gene operon region (about 3.9 kb) was amplified by using primers HMGAop-S and HMGAop-A. The vector pBMB3146, a derivative of pHT304, contained the transcription promoter region and the terminator region of the hmgA gene operon and the hmgA gene from BMB171. The promoter region (394 bp) was amplified from the genomic DNA of BMB171 by using primers HMGAop-S and HMGAop-A2. The hmgA gene (1,173 bp) and the terminator region (232 bp) of the hmgA gene operon was amplified from the genomic DNA of BMB171 by using primers HMGA-S2 and HMGAop-A.

by HppD and the oxidation of HGA to maleylacetoacetate (MAA) by HmgA. MAA is isomerized by maleylacetoacetate isomerase (MaiA) to fumarylacetoacetate, which is subsequently hydrolyzed by FahA to fumarate and acetoacetate (Schmaler-Ripcke et al., 2009; Valeru et al., 2009). The genes encoding these enzymes (HppD, HmgA, FahA, and MaiA) are adjacent on the chromosome of several organisms such as Aspergillus fumigatus, Pseudomonas chlororaphis, and Vibrio cholerae (Kang et al., 2008; Schmaler-Ripcke et al., 2009; Valeru et al., 2009). In the KEGG pathway database, three genes (BMB171\_C1377, BMB171\_C1351, and BMB171\_C2647) are predicted to encode the aminotransferases involved in the conversion of L-tyrosine to 4-hydroxyphenylpyruvate. The three enzymes (HppD, HmgA, and FahA) in B. thuringiensis BMB171 are probably transcribed as part of an operon as indicated by the genomic analysis (**Figure 3B**). Analyzing the genome sequences of B. thuringiensis strains of different H serotypes, we found that the gene operon for L-Tyrosine degradation via HGA is located on every genome (data not shown). However, the genes responsible for the isomerization of MAA cannot be directly identified in B. thuringiensis by analyzing the genomic data. This implies that the genes responsible for the MAA metabolism of B. thuringiensis might be different from that of the reported genes. Even so, we found that the inactivation of HmgA leads to pyomelanin hyperproduction in B. thuringiensis strains. We speculate that B. thuringiensis strains have the potential to produce pyomelanin via HGA through the tyrosine metabolism pathway by inactivating HmgA.

HmgA is involved in the catabolism of the phenylalanine and tyrosine pathway. The structure of the human HmgA shows that the enzyme forms a hexamer arrangement consisting of a dimer of trimers (Titus et al., 2000). HmgA has been predicted to belong to the cupin-like superfamily<sup>1</sup> , which refers to a β-barrel structural domain, on the basis of the primary sequence. Two conserved histidine-containing motifs provide the signature sequence for the cupin superfamily (Dunwell et al., 2001). Despite the considerable variability of the HmgA proteins from different organisms in their primary amino acid sequences (from 22 to 65%), the residues involved in coordinating Fe2<sup>+</sup> are highly conserved in the HmgA proteins from bacteria, fungi, plants, invertebrates, and humans (**Figure 5**). Three residues, including two residues His298 and Glu305 in motif 1, and one residue His334 in motif 2, are responsible for the metal ion binding of the 171HmgA protein (**Figure 5**). Additionally, the secondary structure of 171HmgA is composed of many β-strands (**Figure 5**). In this study, we found that a G272E mutation in HmgA resulted in pigment overproduction in the B. thuringiensis strain BMB181. Residues 272–274 of HmgA form a loop structure connecting two β-strands (β19 and β20). Gly272 is highly conserved in different HmgAs and the equivalent in the human HmgA is Gly309 (**Figure 5**). A G309V missense mutation has previously been found in AKU patients (Nemethova et al., 2016). These findings suggest that the residue Gly272 plays a very important role in the enzyme activity. To verify whether other residues in the loops or adjacent to the loops have the same effect as the residue Gly272, we designed 14 single-point mutants. Nine of the residues (Gly89, Gly119, Gly128, Phe136, Pro245, His261, Asn300, Gly323, and Pro336) are located in different loop regions, and the other five (Asn116, Asp120, Lys219, Gly241, and His334) are located at the start or end of different β-strands adjacent to the loop-forming residues. The result shows that only six mutants (G128A, F136A, G241A, H261A, H334A, and P336A) have an effect on melanin production (**Table 3**). The activity of the enzyme can be disrupted by mutations in many different ways. Some will directly or indirectly affect the active site, others will interfere with the folding of the subunit, and some will affect intersubunit interactions (Rodriguez et al., 2000). The relationship between these residues and the catalytic activity needs to be further confirmed by biochemical characterization and crystal structure.

Previous studies have shown that residues in many different sites of homogentisate 1,2 dioxygenase from humans are essential for enzyme function (Rodriguez et al., 2000; Vilboux et al., 2009; Zatkova, 2011). For instance, single amino acid

<sup>1</sup>https://www.ncbi.nlm.nih.gov/cdd/

(AAD29874.1); Ani, 2,5 dihydroxyphenylacetate oxidase from Aspergillus nidulans (AAC49071.1); and Vp, homogentisate 1,2-dioxygenase from Vibrio

parahaemolyticus 10329 (EGF43769.1). The sequences were aligned with ClustalW (Chenna et al., 2003) and ESPript (Robert and Gouet, 2014).

substitutions at different positions of HmgA can result in pyomelanin overproduction in AKU patients (Rodriguez et al., 2000; Vilboux et al., 2009; Zatkova, 2011). The deduced amino acid sequence of 171HmgA consists of 390 amino acids and shares 24% identity with the HmgA from humans. The singleresidue substitution in the 13 equivalents to the aforementioned 14 mutational residues in the human HmgA sequence would result in the loss of enzyme activity and pigment production (**Table 3**; Zatkova, 2011; Usher et al., 2015). These results indicate that five residues (G128, F136, G241, H334, and P336) are more conservative than the others (G89, N116, G119, D120, K219, P245, N300, and G323) in HmgA, and


TABLE 3 | Site-directed mutagenesis in HmgA.

fmicb-09-02242 October 10, 2018 Time: 11:20 # 10

<sup>a</sup>N: no pigment; P: pigment.

<sup>b</sup>?: no single amino acid substitution was found in the residue F290 in AKU patients in the present study.

mutations in residues G128, F136, G241, H334, and P336 could affect the production of pyomelanin in both humans and B. thuringiensis.

Pyomelanin production has been studied in different bacteria. The accumulation of pyomelanin does not affect the growth characteristics nor the expression of key virulence factors of B. anthracis (Han et al., 2015). Notably, the mutant P. aeruginosa strain PA141hmgA is significantly more virulent than the wild-type PA14 as PA141hmgA can kill the nematodes at a greatly accelerated rate compared with the wild-type PA14 (Harmer et al., 2015). Additionally, pyomelanin has been identified as the primary melanin produced by the A. media strain WS through the autoxidation and self-polymerization of HGA (Wang et al., 2015). The melanin produced by the strain WS serves as an excellent photoprotective agent for BTI against UV and sunlight radiation (Wan et al., 2007). A previous study has shown that the pigment produced by the B. thuringiensis strain BMB181 protects against UV radiation (Liu et al., 2013). To test the possible role of pyomelanin in H2O<sup>2</sup> resistance, we estimated the viability of the culture of the strain BMB171 under H2O<sup>2</sup> treatment and found that the supernatant from the mutant culture was able to protect the vegetative cells from the effect of H2O<sup>2</sup> based on the growth curves (data not shown). We speculate that the pyomelanin produced by B. thuringiensis strains may have a significant synergistic effect on the crystal proteins against nematodes and other insect pests by not affecting the growth and the expression of key virulence factors of B. thuringiensis and protecting B. thuringiensis cells from the stresses such as H2O<sup>2</sup> and UV during the life cycle.

In this study, we constructed several plasmid vectors for complementation analysis (**Figure 3B**). The plasmid pBMB3141 was constructed using the vector pBMBL (unpublished data) and the single hmgA gene fragment for the complementation analysis. However, we found that the strain BMB181 could not completely restore the non-pigment phenotype with the plasmid pBMB3141 transformed into it, and the pigment appeared after culturing the transformant BMB3141 (a derivative of BMB181 containing pBMB3141) in LB medium for over 40 h (**Figure 1B**), suggesting that the promoter of the vector pBMBL, a sporulation-dependent promoter (BtI-BtII), is not suitable for this restoration test. To solve this problem, we tried to construct a plasmid vector that contained its own promoter and terminator region of the hmgA gene for complementation analysis. As shown in **Figure 3B**, the genes hppD, fahA, and hmgA are organized as an operon as indicated by the genomic analysis, and the transcription of the operon would be influenced with the hmgA gene being interrupted by the kanamycin resistance cassette in the strain BMB1711hmgA. Thus, we constructed the plasmid vectors (**Figure 3B**) that contained the complete hmgA gene operon from strains BMB171 and BMB181 separately for the complementation analysis (**Figures 4B,C**). The functions of HmgA variants in melanin production were also tested by the complementation analysis. We firstly constructed the vector pBMB3146 that contained the promoter and terminator regions of the hmgA gene operon and the hmgA gene (**Figure 3B**). All of the point mutations in HmgA were carried out by gene splicing by overlap extension PCR (SOE PCR) with pBMB3146 as the template. By introducing the plasmids containing different hmgA gene variants into the pigmented strain BMB1711hmgA, the roles of different HmgA variants was preliminarily determined by the phenotypes (pigmented or non-pigmented) (**Table 3**). This indicates that some of these residues play important roles in enzyme activity. Nevertheless, the exact roles of these residues for the enzyme HmgA could not be

determined based only on the aforementioned data. The difference between HmgA variants should be distinguished by catalytic mechanisms in further related research.

In summary, B. thuringiensis strains could produce pyomelanin via HGA upon the deactivation of the HmgA or the disruption of the hmgA gene. We found that a G272E amino acid substitution in HmgA resulted in pigmentation. Several other residues in the loops or adjacent to the loops have the same effect as the residue Gly272 on the enzyme activity and melanin production. It is possible to generate more mutations in the hmgA gene and reintroduce them into B. thuringiensis and in many cases obtain a mutant phenotype (pigment overproducer). This approach has potential use to producing B. thuringiensis strains more resistant to UV. This is also an interesting model to study the human gene responsible for alkaptonuria.

# REFERENCES


#### AUTHOR CONTRIBUTIONS

WY, LR, and MS designed the research. WY and JT performed the research. WY, JZ, LR, DP, and MS analyzed the data. WY, LR, and MS wrote the manuscript.

# FUNDING

This work was supported by grants from the National Key R&D Program of China (2017YFD0201201), National Natural Science Foundation of China (31670085 and 31600005), China 948 Program of Ministry of Agriculture (2016-X21), and Fundamental Research Funds for the Central Universities (2662017PY094).

host endocytosis pathway and impact on virulence. Front. Microbiol. 3:440. doi: 10.3389/fmicb.2012.00440



**Conflict of Interest Statement:** 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 © 2018 Yang, Ruan, Tao, Peng, Zheng and Sun. 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.

# Microbial Platform for Terpenoid Production: Escherichia coli and Yeast

Chonglong Wang<sup>1</sup> \*, Mudanguli Liwei<sup>1</sup> , Ji-Bin Park<sup>2</sup> , Seong-Hee Jeong<sup>2</sup> , Gongyuan Wei<sup>1</sup> , Yujun Wang<sup>3</sup> and Seon-Won Kim<sup>2</sup> \*

<sup>1</sup> School of Biology and Basic Medical Sciences, Soochow University, Suzhou, China, <sup>2</sup> Division of Applied Life Science (BK21 Plus), PMBBRC, Gyeongsang National University, Jinju, South Korea, <sup>3</sup> Department of Marine Science, Qinzhou University, Qinzhou, China

#### Edited by:

Eung-Soo Kim, Inha University, South Korea

#### Reviewed by:

Sang Jun Lee, Chung-Ang University, South Korea Taek Soon Lee, Lawrence Berkeley National Laboratory (LBNL), United States

#### \*Correspondence:

Chonglong Wang clwang@suad.edu.cn Seon-Won Kim swkim@gnu.ac.kr

#### Specialty section:

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

> Received: 07 June 2018 Accepted: 25 September 2018 Published: 12 October 2018

#### Citation:

Wang C, Liwei M, Park J-B, Jeong S-H, Wei G, Wang Y and Kim S-W (2018) Microbial Platform for Terpenoid Production: Escherichia coli and Yeast. Front. Microbiol. 9:2460. doi: 10.3389/fmicb.2018.02460 Terpenoids, also called isoprenoids, are a large and highly diverse family of natural products with important medical and industrial properties. However, a limited production of terpenoids from natural resources constrains their use of either bulk commodity products or high valuable products. Microbial production of terpenoids from Escherichia coli and yeasts provides a promising alternative owing to available genetic tools in pathway engineering and genome editing, and a comprehensive understanding of their metabolisms. This review summarizes recent progresses in engineering of industrial model strains, E. coli and yeasts, for terpenoids production. With advances of synthetic biology and systems biology, both strains are expected to present the great potential as a platform of terpenoid synthesis.

Keywords: terpenoid, Escherichia coli, yeast, synthetic biology, MEP pathway, MVA pathway, strain engineering

#### INTRODUCTION

Terpenoids comprise a vast family of the most abundant (>55, 000 known members) natural products with diverse biological functions including cell integrity, hormones, electron transport, and photosynthetic machinery. The extracted products from some herbs or animal liver are realized as flavors, fragrances, colorants, commodity chemicals, vitamins, and pharmaceuticals from ancient time. Nevertheless, their low yields from natural sources limit mass production of terpenoid for industrial application. For instance, Artemisia annua (Qinghao) just yields artemisinin of <0.8% by dry biomass weight (Zyad et al., 2018), which severely restricts commercialization of this antimalarial drug. Humans have employed microbes to produce beverages (e.g., Saccharomyces cerevisiae) before civilization, and antibiotics (e.g., Penicillium chrysogenum) at an industrial scale since the Second World War (Nielsen and Keasling, 2016). Nowadays, advances in metabolic engineering enable us to build microbial cell factories to generate many interesting products, of which Escherichia coli and yeast are the well-characterized hosts for efficient and large-scale production. Since Amyris Inc., announced a record low manufacturing cost of \$1.75 per liter for farnesene from yeast in 2015, microbial farnesene has attracted a lot of interests from industry<sup>1</sup> .

<sup>1</sup>http://investors.amyris.com/press-releases

This review describes progress in biosynthesis of the diverse terpenoids in E. coli and yeast, where fantastic technologies are harnessed for development of microbial cell factories. It illuminated the great potential of E. coli and yeast on tackling a complexity of biosynthesis of the diverse terpenoids.

#### BIOSYNTHESIS PATHWAYS OF DIVERSE TERPENOIDS

Albeit structurally diverse, skeletons of terpenoids are composed of C<sup>5</sup> isoprene units which are successively assembled by biogenic isoprene rule. Precursors of the C<sup>5</sup> isoprene units are isopentenyl diphosphate (IPP) and dimethylallyl diphosphate (DMAPP), which are synthesized by either mevalonate (MVA) pathway in eukaryotes, archaea and a few bacteria or 2-C-methyl-Derythritol 4-phosphate (MEP) pathway in prokaryotes and plant plastids (Liao et al., 2016; Frank and Groll, 2017). MEP pathway starts from condensation of two glycolytic intermediates, pyruvate and glyceraldehyde-3-phosphate, and MVA pathway from condensation of acetyl-CoAs (**Figure 1A**). Both pathways require energies (ATP) and reductive powers [NAD(P)H] to proceed multi-enzymatic reactions to produce IPP and DMAPP. Given no consideration of energy balance, MEP pathway consumes 1 molecule of glucose per IPP with a carbon molar yield of 0.83, while the MVA pathway consumes 1.5 molecules of glucose per IPP with the lower yield of 0.56 (Eqs. 1, 2) (Schempp et al., 2018). In this scenario, MEP pathway requires 2 ATP and 2 NAD(P)H, while MVA pathway accompanies generation of 3 NAD(P)H. Thus, microorganisms need engineering to use both pathways to attain the carbon and energy balances, which results in a synergy between both pathways for production of terpenoids (Yang et al., 2016).

**MEP**: 1 Glucose + 2ATP + 2 NAD(P)H → 1 IPP + CO<sup>2</sup> (YIPP/Glucose = 0.83 C-mol/C-mol) (1) **MVA**: 1.5 Glucose → 1 IPP + 4 CO<sup>2</sup> + 3 NAD(P)H (YIPP/Glucose = 0.56 C-mol/C-mol) (2)

Next, prenyltransferases catalyze chain elongation of isoprenyl diphosphates including geranyl diphosphate (GPP, C10), farnesyl diphosphate (FPP, C15), geranylgeranyl diphosphate (GGPP, C20), and even longer chain isoprenyl diphosphates (LoPP, >C45), which determine the primary diversity of terpenoids in the chain lengths. Thus, terpenoids are classified to hemiterpenoids (C5), monoterpenoids (C10), sesquiterpenoids (C15), diterpenoids (C20), triterpenoids (C30), tetraterpenoids (C40), and long chain isoprenyl products (**Figure 1A**).

Terpene synthases convert isoprenyl diphosphates to numerous terpenes via carbocationic reaction cascades and/or hybridization of carbon atoms, which generate a myriad of carbon skeletons containing several stereocenters and present their diversity in the structure (Christianson, 2017). The skeletons are further regio- and stereo-selectively decorated by many tailoring enzymes such as cytochrome P450s (P450s), acyltransferase, and glycosyltransferase, which may act in a combinatory manner and finally render various functions to isoprenoids (e.g., artemisinin, taxol and ginkgolide) (King et al., 2016). Therefore, the microbes, for which a good genetic engineering toolbox is available, can be engineered to synthesize valuable terpenoids by assembling of the precursor synthesis, isoprenyl chain elongation, structural rearrangement and tailoring reactions.

#### MICROBIAL HOSTS FOR TERPENOID PRODUCTION

As most terpenoids have been discovered in plants, an extraction from plants is a major source of commercialized terpenoids. However, the plant extraction is compromised by its low productivity due to a slow growth of plants, a low product content in plants, and a limitation in cultivating some species. Even though there is a biotechnological progress in transgenic plants, it still remains difficult to engineer plants for improved isoprenoids contents. Microbes grow fast and do not heavily rely on land/water resources. Thus, they possess a great potential for sustainable mass production of terpenoids over plants. Microbial hosts in industry are required to meet the criteria: (i) a large metabolic potential supporting efficient synthesis of products of interest with robust and fast cell growth, (ii) a well-understood metabolism and well-developed genetic tools (e.g., expression vectors), and (iii) a great capacity to grow on cheap carbon sources. Terpenoids synthesis relies on MEP or MVA pathway for the generation of building units, IPP and DMAPP. Therefore, two representative microbial hosts E. coli and S. cerevisiae, which are based on MEP and MVA pathways, respectively, have been used for production of diverse terpenoids.

The terpenoids natively produced in E. coli are limited to small amounts (e.g., quinones). The native MEP pathway could be less efficient in IPP and DMAPP supply for terpenoid production. E. coli has been engineered to improve IPP and DMAPP synthesis by augmenting bottleneck enzymes of MEP pathway or introducing heterologous MVA pathway. S. cerevisiae accumulates a large amount of ergosterol, which presents a potential of its MVA pathway for terpenoid production. S. cerevisiae has redox systems that allow cytochrome P450 to modify terpenoids skeleton, whereas E. coli has none. S. cerevisiae is superior to E. coli in synthesis of value-added terpenoids with complicated structures. Yarrowia lipolytica is another good yeast for production of terpenoids based on MVA pathway because of its abundant acetyl-CoA, the initial substrate of MVA pathway (Zhu and Jackson, 2015). Besides, carotenogenic Rhodosporidium toruloides is developed to produce terpenoids from lignocellulosic hydrolysates in that it exerts ability to utilize multiple carbons and tolerates high osmotic stress (Yaegashi et al., 2017; Sundstrom et al., 2018). As many terpenoids have been isolated from Streptomyces species (Koksal et al., 2012; Rinkel et al., 2017), the bacteria could also be a potential host for terpenoids production in the future (Phelan et al., 2015; Khalid et al., 2017).

FIGURE 1 | Overview of terpenoids biosynthesis pathway (A), and pathway engineering strategies (B,C). Terpenoids biosynthesis is comprised of carbon assimilation, isoprene unit synthesis, terpenoids backbone synthesis, and terpenoids decoration. C<sup>6</sup> or C<sup>5</sup> sugar enters the central pathway of glycolysis. MEP and MVA pathways use central metabolites to initiate synthesis of IPP and DMAPP, the building blocks of terpenoids. Terpenoid synthesis pathways could be assembled and engineered in a tractable host (e.g., E. coli or yeasts) to create cell factories for their mass production. Two fashions of static and dynamic engineering have been used to optimize the synthesis pathways and coordinate them with host metabolic network. Static engineering approach generally constructs a matrix of genes, promoters, and regulatory elements (e.g., UTRs) to screen the best orchestra, while dynamic engineering approach relies on biosensors (metabolite response or transcription factor-based) to dynamically control the synthesis pathway. The red dots present the key intermediates of terpenoid biosynthesis, and the yellow and cyan dots present primary terpenenoids and decorated terpenoids, respectively (A,B). TF presents transcription factor (C).

# METABOLIC ENGINEERING OF E. coli AND YEASTS FOR TERPENOID PRODUCTION

Microbial production of terpenoids can be addressed by introduction of relevant genes for their synthesis into host strains (**Figure 1B**). Owing to a versatility of E. coli and S. cerevisiae, many strategies of metabolic engineering have been developed and examined to increase terpenoid production in the both model hosts. The engineering strategies can be classified to "static" and "dynamic" according to a regulatory means of synthetic pathway. Both approaches have been applied to terpenoid synthetic pathways for an enhancement of IPP/DMAPP flux, a minimization of byproducts formation, a toxicity reduction of pathway intermediates, etc., Static engineering employs variation of vectors, promoters, ribosome binding sites (RBSs), and genes of interest, which are assembled in a plethora of biological systems, to optimize a synthesis of target products (Ren et al., 2017; **Figure 1C**). Brewer's yeast has been engineered by combinatory regulation of pathway genes to biosynthesize aromatic monoterpenes that impart hoppy flavor to beer (Denby et al., 2018). Farnesol is a desirable biofuel molecule derived from FPP by phosphatases. By variation of phosphatases, 526 mg/L of farnesol was produced in an engineered E. coli overexpressing a membrane integral phosphatase, PgpB (Wang C. et al., 2016). Further phosphatase mining retrieved a cytosolic phosphatase NudB, whose overexpression led to farnesol of 1.42 g/L along with hemiterpenoids isopentenol (Zada et al., 2018). Optimization of plasmids carrying synthesis pathway of terpenoids brought an enhanced production of monoterpenes, 653 mg/L of 1,8-cineole of and 505 mg/L of linalool, from E. coli (Mendez-Perez et al., 2017). Oleaginous Y. lipolytica is able to synthesize a massive amount of acetyl-CoA, and lipid droplet is supposed to store lipophilic terpenoids. Tuning of promoter strength by promoters shuffling resulted in β-carotene production of 111.8 mg/L in Y. lipolytica (Larroude et al., 2018).

However, the static engineering is too laborious to gain a desired performance from an engineered strain due to intermediates toxicity or metabolic burdens often occurring from a mass production of terpenoids, whereas the dynamic engineering can address such adverse circumstances. Thus, there is also a great interest to develop tunable or inducible promoters and small regulatory RNAs (Ghodasara and Voigt, 2017; Marschall et al., 2017; Portela et al., 2017; Trassaert et al., 2017), which could benefit to a dynamic control of synthetic pathways of terpenoids. Biosensors have been developed to sense small molecules (metabolites), which are applied to a transcription control of the committed pathway in response to metabolite abundance (Dekker and Polizzi, 2017). They are built to a simple metabolite responsive promoter or a transcription factor (TF)-based binary system (**Figure 1C**), which are incorporated as a metabolic controller for dynamic flux control (Mannan et al., 2017). For example, FPP repressive promoters were developed to down-regulate FPP biosynthesis and FPP activated promoters could be used to up-regulate expression of amorphadiene synthase converting FPP to amorphadiene. These two promoters were combined in E. coli to implement the negative and positive feedback loops of the synthesis and conversion of FPP, which is a critical metabolic node in amorphadiene synthesis. Amorphadiene production of 1.2 g/L was obtained from the engineered E. coli, which was a two-fold increase in the production through the dynamic control (Dahl et al., 2013). In S. cerevisiae, a large amount of FPP is mainly used for synthesis of ergosterol via squalene. Thus, squalene synthase (Erg9) condensing two FPPs to a squalene could be a critical regulation point to divert FPP flux to synthesis of terpenoids of interests. A dynamic control of Erg9 expression using ergosterol-responsive promoter could alter FPP flux to amorphadiene synthesis (Yuan and Ching, 2015). Biosensors have been developed to provide an additional exquisite regulatory means for tuning of synthetic pathway, and improved in both their dynamic responsive range and substrates spectrum (Rogers et al., 2016; Liu et al., 2017). Natural TFs (AraC and Gal4) were evolved to respond to IPP accumulation (Chou and Keasling, 2013). Their application in controlling zeaxanthin biosynthesis pathway resulted in a successful production of the carotenoid (C40) at a titer of 722.5 mg/L (Shen et al., 2016). Synthetic biologists have also designed promoters responding to environmental signals such as pH and quorum sensing (QS) molecules (Tsao et al., 2010; Zargar et al., 2015; Rajkumar et al., 2016). It is capable control terpenoid biosynthesis with the environmental signals, not using expensive inducer molecules. A QS-based promoter system was successfully applied for an inducer free production of bisabolene from E. coli, where titer of bisabolene was increased to 1.1 g/L through optimization of LuxR and LuxI expressions (Kim et al., 2017).

## GENOME-LEVEL ENGINEERING OF TERPENOIDS PRODUCTION

Rewiring of metabolic network of host strain is required for a full performance of an engineered pathway for product synthesis, when potential of the engineered production pathway is restricted due to a metabolic characteristics of the host. Many genetic tools are available for genome engineering in E. coli and yeasts. Homologous recombination is the most popular method for deletion and replacement of genetic parts. The glyceraldehyde 3-phosphate and pyruvate supply was rebalanced by tuningdown of glyceraldehyde 3-phosphate synthase, which led to a two-fold increase in lycopene production via the MEP pathway in E. coli (Jung et al., 2016). The central pathway was rewired in S. cerevisiae to synthesize acetyl-CoA with reduced losses of carbon and ATP. As acetyl-CoA was the substrate of the MVA pathway, rewiring of the central carbon metabolism in S. cerevisiae resulted in an increase in farnesene production by 25% along with a reduced requirement of oxygen by 75% (Meadows et al., 2016). Genome engineering tools have been developed to simultaneously target multiple genomic loci like as multiplex automated genome engineering (MAGE), and to precisely edit a specific genomic locus without a scar based on the clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9 system (**Figure 2A**). MAGE was applied to

optimizing MEP pathway in E. coli for lycopene production (Wang et al., 2009). A modified method of MAGE was also successfully applied for β-carotene production in S. cerevisiae (Barbieri et al., 2017). As MAGE generates a vast of combinatorial variants in genome, it could be an effective genome engineering tool only when it is combined with a high-throughput screening (colorimetric or fluorescent) system enabling a selection of variants with desired traits. It is also difficult to incorporate large genetic parts (>1 kb) into genome by MAGE. CRISPR/Cas9 is currently powerful and popular genetic tool for precise genome editing (Jiang et al., 2013; Jessop-Fabre et al., 2016). The endonuclease Cas9 combined with guide RNA (gRNA) specifically recognizes a target site in genome, and facilitates sitespecific nucleotide base-pair mutations, gene deletions, or large DNA fragment (around 8 kb at least) insertions (**Figure 2A**). β-Carotene synthesis pathway was integrated into E. coli genome by CRISPR/Cas9-based method. Both MEP pathway and central metabolic pathways were optimized in the engineered strain, and the final strain was cultured in fed-batch, yielding 2.0 g/L of β-carotene (Li et al., 2015). A set of genomic loci of Y. lipolytica have been identified for targeted marker-less integration using CRISPR/Cas9 (Schwartz et al., 2017). An automated platform for multiplex genome-scale engineering in S. cerevisiae has been also developed based on CRISPR/Cas9 (Si et al., 2017). Plasmid-based expression of production pathway has a problem of segregational or structural instability of plasmids, which becomes worse in case of inhibitory products to host cells. Moreover, expensive antibiotics is required to exert a selection pressure on host cells for prevention of plasmid loss, which increases a production cost. However, genome-based expression of production pathway could deliver stable and reproducible production with no use of antibiotics, which is important for mass scale production of industry. The top portion of MVA pathway was integrated into genomic loci of adhE and ldhA in E. coli, and atpFH genes were deleted to increase glycolytic rate. The genome-engineered strain exhibited both high productivity (∼1.01 g/liter/h) and yield (86.1% of theoretical yield) after 48 h of shaking flask culture (Wang J. et al., 2016). However, engineering of genome integrated production pathway is a cumbersome task compared to plasmid-based engineering. Plasmid is more easy and convenient to modulate genes dosage and to construct various genetic expression cassettes. The genome level engineering for optimization of production pathway was facilitated by CRISPR/Cas9-based convenient chromosomal promoters change, which is successfully applied for bisabolene synthesis in E. coli (Alonso-Gutierrez et al., 2018). A heterologous MEP pathway was integrated into genome of S. cerevisiae with endogenous MVA pathway. As the MEP pathway does not produce extra NADH which need oxidizing to maintain redox status, the engineered yeast enabled to rely solely on the MEP pathway instead of the MVA pathway for terpenoids biosynthesis under low aeration conditions (Kirby et al., 2016). MVA pathway genes attached with mitochondrial-targeting signal sequences were integrated into genome of S. cerevisiae to enable utilization of both cytosolic and mitochondrial acetyl-CoAs via the native cytosolic and the engineered exogenous mitochondrial MVA pathways. The

engineered yeast produced 2.5 g/L of isoprene, an increase of 1.6-fold by the mitochondrial MVA pathway (Lv et al., 2016). A comprehensive understanding of whole metabolic networks is a prerequisite to rational editing of genomes or design of biosynthesis pathways (Campbell et al., 2017). Metabolic engineers paid efforts to interpret cellular phenomena at a systems level by measuring various cellular components including RNAs, proteins, and metabolites (**Figure 2B**). There are successful examples of terpenoid production driven by systems biology to debottleneck synthesis pathway (Alonso-Gutierrez et al., 2015; Li et al., 2017; Wada et al., 2017). An integrated approach of multi-level Omics data has been pursued to obtain a desirable phenotype of host strain for production, because a genetic manipulation may have positive impact at one metabolic layer but a negative or neutral impact at another (Goh and Wong, 2016; Lechner et al., 2016). IPP toxicity always challenges the engineered microbial systems for terpenoid production. By multi-level Omics-integrated analysis, formation of isoprenyl-ATP analog is recognized as a cause of the IPP toxicity, which suggests potential engineering strategies for terpenoids production from E. coli (George et al., 2018).

# CONSORTIA PROCESS FOR PRODUCTION OF TERPENOIDS

Complicated structures of terpenoids are often a hurdle of the production using a single microbe. As E. coli is generally not a tractable host for P450 chemistry, it is difficult to produce terpenoids decorated by P450 from E. coli. Although an oxygenated taxanes is produced in E. coli with optimization of P450 expression and its reductase partner interaction (Biggs et al., 2016), it requires lots of elaborate endeavors and cannot be copied into an engineering of other P450s. It would be beneficial to engineer multiple organisms to carry out a complicated task together rather than an engineering of a single organism alone (Hays et al., 2015). With greater understanding of microbial traits, coordinated microbial consortia can be designed and built to address complex tasks. Expansion of metabolic capacities and improvement of production yields could be attained in the synthetic microbial communities along with simplification of engineering of a complicated long pathway to a few simple shorter pathways distributed in the community microbes. The synthetic microbial consortia can be classified an intraspecies homo-consortia and an interspecies heteroconsortia, of which each strain is assigned an individual role toward product synthesis (**Figure 2C**). A key issue of synthetic microbial consortia engineering is to balance each microbial population for a successful collaboration to attain the goal (Johns et al., 2016). In order to produce acetylated taxanes, the entire pathway was divided into taxadiene synthesis expressed in E. coli and P450 modification in S. cerevisiae (Zhou et al., 2015). E. coli was engineered to use xylose, but produce a wasteful acetate, while auxotrophic S. cerevisiae assimilated solely acetate for cell growth in presence of xylose as a sole carbon source. The mutualistic consortium could maintain a population balance of the co-cultured strains and doubled the

can be built in the same species (homo-consortia) or the different species (hetero-consortia).

production in comparison with a co-culture using glucose. Deeper understanding of autotrophy, interspecies cross-feed, and QS machinery would provide more exquisite approaches for production of complicated terpenoids using synthetic microbial consortia.

# CONCLUSION

Currently, many enabling technologies of synthetic biology allow us to tailor microbes for terpenoids production. Among the microbial species, E. coli, and yeasts have proved as the most attractive microbial platform, which can be conveniently developed to generate either bulk or value-added terpenoids owing to versatile enabling technologies for the microbes. A great potential of E. coli and yeasts as platform strains has been demonstrated with many successes of synthetic pathway rewiring, genome editing, and microbial consortia building for improvement of production. With deeper understanding of their metabolism, more promising approaches are expected to boost production of terpenoids in E. coli and yeasts.

# AUTHOR CONTRIBUTIONS

CW and S-WK developed the ideas and drafted the manuscript. ML, J-BP, and S-HJ collected the literatures and drew the figures. GW, YW, and S-WK professionally edited the manuscript.

# FUNDING

This work was supported by C1 Gas Refinery Program through the NRF funded by the MSIP (NRF-2016M3D3A1A01913246) and a grant from the Next-Generation BioGreen 21 Program (SSAC, grant#: PJ01326501), RDA, Korea. J-BP and S-HJ were supported by scholarships from the BK21 Plus Program, MEST, Korea.

# ACKNOWLEDGMENTS

CW would like to thank the support of the China Postdoctoral Science Foundation (2017M610350). CW and GW also thanked a project funded by the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD).

# REFERENCES

fmicb-09-02460 October 12, 2018 Time: 12:40 # 7


reveromyceticus SN-593. ACS Synth. Biol. 6, 2339–2349. doi: 10.1021/acssynbio. 7b00249



fermentation through chromosomal integration. Appl. Environ. Microbiol. 82, 7176–7184. doi: 10.1128/AEM.02178-16


**Conflict of Interest Statement:** 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 © 2018 Wang, Liwei, Park, Jeong, Wei, Wang and Kim. This is an openaccess 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.

# Recombinant Protein Expression System in *Corynebacterium glutamicum* and Its Application

#### Min Ju Lee and Pil Kim\*

*Department of Biotechnology, The Catholirc University of Korea, Bucheon, South Korea*

*Corynebacterium glutamicum*, a soil-derived gram-positive actinobacterium, has been widely used for the production of biochemical molecules such as amino acids (i.e., L-glutamate and L-lysine), nucleic acids, alcohols, and organic acids. The metabolism of the bacterium has been engineered to increase the production of the target biochemical molecule, which requires a cytosolic enzyme expression. As recent demand for new proteinaceous biologics (such as antibodies, growth factors, and hormones) increase, *C. glutamicum* is attracting industrial interest as a recombinant protein expression host for therapeutic protein production due to the advantages such as low protease activity without endotoxin activity. In this review, we have summarized the recent studies on the heterologous expression of the recombinant protein in *C. glutamicum* for metabolic engineering, expansion of substrate availability, and recombinant protein secretion. We have also outlined the advances in genetic components such as promoters, surface anchoring systems, and secretory signal sequences in *C. glutamicum* for effective recombinant protein expression.

#### *Edited by:*

*Eung-Soo Kim, Inha University, South Korea*

#### *Reviewed by:*

*Yutaka Kawarabayasi, Kyushu University, Japan Sergio Adrian Guerrero, Universidad Nacional del Litoral, Argentina*

> *\*Correspondence: Pil Kim kimp@catholic.ac.kr*

#### *Specialty section:*

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

> *Received: 02 July 2018 Accepted: 03 October 2018 Published: 26 October 2018*

#### *Citation:*

*Lee MJ and Kim P (2018) Recombinant Protein Expression System in Corynebacterium glutamicum and Its Application. Front. Microbiol. 9:2523. doi: 10.3389/fmicb.2018.02523* Keywords: *Corynebacterium glutamicium*, expression host systems, cytosolic expression, secretory expression, recombinant protein, surface displayed expression

#### INTRODUCTION

Recombinant proteins, including biologics and enzymes, are useful in the biopharmaceutical, food, and chemical industries (Butenas, 2013). To date, more than 400 recombinant biologics have been approved by the US Food and Drug Administration (FDA), and more recombinant biologics are in the clinical development stage (Sanchez-Garcia et al., 2016). The demand for new biologics (such as antibodies, growth factors, and hormones) for the treatment of severe chronic diseases (such as cancer, anemia, and multiple sclerosis) has increased, and the market for recombinant proteins is expected to grow over the next few decades (https:// www.grandviewresearch.com/press-release/global-protein-expression-market). Corynebacterium glutamicum, which is a non-lethal and non-emulsifying gram-positive bacterium, exhibits a low protease activity in the culture supernatant and can secrete protease-sensitive proteins into the culture supernatant (Liu et al., 2013). Absence of lipopolysaccharide (endotoxin) in C. glutamicum, which is the gram-negative bacterial surface component that should be removed for the production of therapeutic proteins (Srivastava and Deb, 2005), may increase the heterologous protein yield by minimizing the purification steps. C. glutamicum has generally been used as a generally recognized as safe (GRAS) host for the industrial production of biochemicals including L-glutamate and L-lysine (Lee et al., 2016). As a result, C. glutamicum is favorable for producing high yields of proteins that are difficult to secrete in the host and the proteins that must remain active in a non-pathogenic environment.

The industrial production of biochemicals including nutraceuticals has been established using C. glutamicum as a host (Nakayama et al., 1961). Since C. glutamicum was first isolated as an L-glutamate producer by Kinoshita and Udaka in 1956 (Kyowa Hakko Bio Ltd. Co., Japan) (Kinoshita et al., 1957), many L-amino acids have been produced using this soil bacterium. In addition, many biochemicals (biopolymers, organic acids, rare sugars, etc.) have been commercially produced from metabolically engineered C. glutamicum strains. The metabolic processes of C. glutamicum may be rationally modified for the production of various biochemicals using three approaches: (1) amplification of biosynthetic pathway enzymes to increase target products, (2) reduction of by-product formation, and (3) introduction of important enzyme feedback controls to optimize target biomaterials. All these approaches involve the use of recombinant protein expression in the cytosol to produce beneficial biochemicals.

This review summarizes the recent studies on the heterologous expression of the recombinant protein in C. glutamicum for various applications including metabolic engineering, expansion of substrate availability, and recombinant protein secretion. It also lists the advancements of genetic components for effective recombinant protein expression.

# CYTOSOLIC PROTEIN EXPRESSION IN *C. glutamicum* FOR METABOLIC ENGINEERING

A common method for producing biochemicals from C. glutamicum is the overexpression of enzymes involved in the biosynthetic pathway of the target product in cytosol (**Table 1**), which involves recombinant protein expression. Jensen and Wendisch overexpressed the ornithine cyclodeaminase (OCD) gene from Pseudomonas putida for the production of L-proline, which is a biochemical that is typically used as a commodity chemical or feed additive; this overexpression resulted in an increased product yield of 0.36 g proline/substrate (Jensen and Wendisch, 2013). Another foreign protein (D-lactate dehydrogenase) from Lactobacillus delbrueckii was expressed to address the limitations of using lactic acid bacteria, which require a relatively expensive complex medium for D-lactate production, and Okino et al. reported a high level of D-lactate production in C. glutamicum (Okino et al., 2008).

Jojima et al. designed protein expression systems as a way to reduce by-product formation in L-alanine production (Jojima et al., 2015). In a C. glutamicum strain, genes involved in the organic acid biosynthetic pathway (1ldhA: lactate dehydrogenase; 1ppc: phosphoenolpyruvate carboxylase; 1alr: alanine racemase) were inactivated; however, the alaD of Lysinibacillus sphaericus (encoding L-alanine dehydrogenase) along with the gapA of L. sphaericus (encoding glyceraldehyde 3-phosphate dehydrogenase promoting glucose consumption) were overexpressed, leading to a metabolic flux from organic acids to L-alanine. As a result, a high product (L-alanine) concentration (98 g/Lmedium) was obtained.

As a large amount of oxygen and energy is required in the production of L-amino acids in C. glutamicum (Kwong and Rao, 1991), Liu et al. reported a novel approach for improving the intracellular oxygen supply by expressing hemoglobin (Liu et al., 2008). They modulated the metabolism to increase the productivity of L-glutamate by inducing metabolic flux into the tricarboxylic acid (TCA) cycle and additionally expressed the hemoglobin protein of Vitreoscilla sp. (VHb) in C. glutamicum to increase the oxygen and energy supply, resulting in the increased production of L-glutamine.

In addition, cytosolic protein expression in C. glutamicum has contributed to the production of biochemicals such as polyhydroxyalkanoate (PHA) (Matsumoto et al., 2011), ethanol (Jojima et al., 2015), and γ-aminobutyric acid (GABA) (Choi et al., 2015). The industrial techniques for the production of rare saccharides such as D-tagatose (Shin et al., 2016), D-sorbose and D-psicose (Yang et al., 2015), D-allose (https://patents.google. com/patent/WO2017111339A1/en), and GDP-L-fucose (Chin et al., 2013) are also the methods of nutraceutical production that involve cytosolic protein overexpression in C. glutamicum.

#### SURFACE-DISPLAYED ENZYME EXPRESSION IN *C. glutamicum*

### Present Applications of Surface-Displayed Systems

The cell surface display systems can be used for a wide range of biotechnological and industrial applications: (1) a live vaccine that induces antigen-specific antibody responses by exposing heterologous epitopes to pathogenic bacterial cells (Lee et al., 2000); (2) screening of displayed peptides by sequential binding and elution (Boder and Wittrup, 1997); (3) expression of surface antigens to produce polyclonal antibodies in animals (Martineau et al., 1991); (4) using biological adsorbents for the removal of heavy metals (Bae et al., 2000); (5) using biological adsorbents for the removal of herbicides and environmental pollutants (Dhillon et al., 1999); (6) detecting single amino acid changes in target peptides after random mutagenesis; and (7) using biosensors with immobilized enzymes, receptors, or other signal-sensitive components (Aoki et al., 2002).

Purified enzymes have been used in many industrial bioconversion processes as immobilized enzyme catalysts. The immobilization of enzymes is time-consuming and costly because it involves several steps: growth of culture, disruption of cells, purification of enzymes, and immobilization of enzymes. When an enzyme is expressed on the cell surface of a microorganism as a whole-cell catalyst, additional purification steps are unnecessary and the whole-cell catalyst can be used repeatedly.

#### Development of Surface-Display Systems

The first surface expression system was developed in the mid-1980s to attach small peptides to proteins fused to bacteriophage surfaces (Smith, 1985). Thereafter, various phage display systems have been developed to express heterologous proteins on the surface of phages; however, the size of exogenous proteins that TABLE 1 | Examples of cytosolic protein expressions in *Corynebacterium glutamicum* for productions of biochemicals.


*(Continued)*

#### TABLE 1 | Continued


can be expressed on the surface is limited (Li, 2000). For example, foreign proteins that can be expressed in M13 phages have a length of 6.6 nm (Lee et al., 2012), while most enzymes are more than 10 nm in diameter (http://book.bionumbers.org/how-bigis-the-average-protein).

To address this problem, microbial cell surface display systems have been developed (i.e., C. glutamicum is 2,000–6,000 nm in length and 500 nm in diameter). Microbial cell surface display is generally accomplished by expressing a passenger protein on the cell surface fused with preexisting microbial surface proteins or with anchoring motifs of the membrane protein (**Figure 1**). A C-terminal fusion, N-terminal fusion, or sandwich fusion strategy can be considered, depending on the characteristics of the fixed motif and the target protein. A good anchoring motif should meet the following requirements: (1) the premature fusion protein must have an efficient signal peptide or transport signal to pass through the inner membrane, (2) an anchoring motif must have a strong immobilization structure to retain the fusion protein on the cell surface, (3) an anchoring motif must be compatible with the inserted or fused foreign sequences (i.e., the anchoring motif should not become unstable following the insertion or fusion of heterologous sequences), and (4)

an anchoring motif must be resistant to attack by proteases present in the periplasmic space or media. In cell display systems, the properties of the target protein are known to significantly affect transport to the cell surface. In particular, the folded structure of the target protein (e.g., a disulfide bridge) in the outer membrane (the periplasmic side) can affect the movement of the target protein (Maurer et al., 1997). In addition, the insertion of amino acid sequences that contain multiple charged residues or hydrophobic residues within the target protein can result in ineffective sequence secretion in bacterial hosts.

Good hosts for surface display should be compatible with the expressed protein and should be easily culturable without cell lysis. In addition, the host cells must have a low extracellular protease activity. Gram-negative bacteria, including Escherichia coli, have a complex surface structure, which consists of the cytoplasmic membrane, periplasm, outer membrane, and many anchoring motifs that have been developed based on the outer membrane proteins. Therefore, the anchoring motif fused with the target protein in gram-negative bacterial hosts must be transferred to the outer membrane through the cytoplasmic membrane and periplasm. On the other hand, gram-positive bacteria are considered to be more suitable for whole-cell catalysts and whole-cell adsorbents because of the robust structure of their cell walls. Many surface proteins can be covalently immobilized on the cell walls of Bacillus spp., Staphylococcus spp., and C. glutamicum. A eukaryotic GRAS host, Saccharomyces cerevisiae, has protein folding and secretion systems that are similar to those in mammalian cells; it has been reported that mammalian proteins could be linked to the cell wall via a glycosylphosphatidylinositol (GPI) anchor or disulfide bonds (Lee et al., 2003).

# Examples of Surface-Displayed System in *C. glutamicum*

The use of C. glutamicum, an important industrial biochemical producer and as a Gram-positive bacterial host, is advantageous in a cell surface display system because of the presence of various enzymes on the surface of C. glutamicum cells that extend the range of carbon sources during the production of biochemicals (**Table 2**). Starch is used as an industrial carbon source for microorganisms; however, C. glutamicum cannot consume starch directly. Starch should be provided in a hydrolyzed form using αamylase or glucoamylase. Tateno et al. have described the use of starch as a carbon source directly by a surface-displayed enzyme on C. glutamicum (Tateno et al., 2007). PgsA, a transmembrane protein derived from Bacillus subtilis, is a part of the polyγ-glutamate synthetase complex. It was used to anchor the α-amylase from Streptococcus bovis 148 on the cell surface. The resulting display system was able to produce L-lysine (yield of 6.04 g/Lmedium) from starch. In addition, a system displaying α-amylase fused with PgsA as an anchor produced 6.4% of poly-β-hydroxybutyrate (PHB) (Song et al., 2013) in a metabolically engineered C. glutamicum, using starch as raw material.

In addition to PgsA, porin has been used as an anchor protein in C. glutamicum. Porin is a cell wall-related protein of C. glutamicum that is present in the mycolic acid layer. Adachi et al. have produced 1.08 g/Lmedium of L-lysine from a cellobiose carbon source using a system in which β-glucosidase was displayed using PorC as an anchor protein (Adachi et al., 2013). In addition, Imao et al. have reported the display of β-xylosidase on the cell surface using PorH as an anchor protein (Imao et al., 2017). In this expression system, xylooligosaccharides were used as a carbon source to produce 0.12 g/Lmedium of 1,5-diaminopentane (cadaverine).

The use of other anchor proteins has also been reported by Yao et al. They displayed S. bovis 148 α-amylase on the cell surfaces using the C-terminally truncated NCgl1221 anchor protein. In this system, 19.3 g/Lmedium of L-glutamate was produced from starch (Yao et al., 2009).

Recent reports of Choi et al. suggested that the proteins from 19 known mycolic acid layers in the extracellular membrane of C. glutamicum can be used as anchoring motifs in surface display systems (Choi et al., 2018). The α-amylase of S. bovis was screened using a portion of NCgl1337 as an anchoring motif; this portion has a signal peptide and a predicted Omycoloylation site. As a result, 10.8 g/Lmedium of L-lysine was obtained from starch; this result demonstrates the potential of whole-cell biotransformation using the cell membrane proteins of C. glutamicum.

# SECRETIONS OF PROTEINS FROM *C. glutamicum*

# Characteristics of Natural Secretion Systems in *C. glutamicum*

The direct secretion of proteins into the culture medium by C. glutamicum over protein expression in the cytosol has several advantages. First, it is easy to obtain the target protein by purification because it does not require cell disruption and there are fewer proteins in the culture medium than in the cytoplasm (Nguyen et al., 2007). In addition, the oxidative environment of the extracellular culture fluid is suitable for the formation of disulfide bonds, which leads to protein folding and the expression of the active protein form (Makrides, 1996). Furthermore, the low extracellular protease activity of C. glutamicum contributes to the stability of the target proteins (Suzuki et al., 2009).

Two major translocation pathways have been known identified in C. glutamicum: the secretory (Sec)-pathway and the twin-arginine translocation (Tat)-pathway. The Secpathway transports unfolded proteins, whereas the Tat-pathway transports folded proteins (Kudva et al., 2013). These two pathways have the signal peptides necessary for the protein to pass through the cell membrane (von Heijne, 1985). The difference between the signal peptides of the two pathways is that the N-region of the Tat-type signal peptide is longer than that of the Sec-type signal peptide because the Tat-type signal peptide contains a conserved twin-arginine residue (RR) at the end of the N-region (Berks et al., 2000; **Figure 2A**).

The Sec-pathway is a system that secretes proteins in an unfolded state (**Figure 2B**). Sec-dependent protein secretion systems have a co-translational targeting system and a posttranslational targeting system (Fröderberg et al., 2004). In the co-translational targeting system, the signal recognition particle (SRP) binds to the nascent peptide and leads the complex (nascent peptide + ribosome) to a membrane protein FtsY along with ribosomes; then, this SPR subsequently leads the nascent peptide to the channel complex (SecYEG). In the posttranslational targeting system, the translation-finished peptide binds to SecB and SecA to reach the SecYEG channel (Singh et al., 2014). Once the linear peptide passes through the SecYEG channel, the signal peptide is cleaved by Type I signal peptidase and the protein is released from the membrane (Schallenberger et al., 2012).

The Tat-pathway is a twin-arginine translocation pathway with a conserved twin-arginine motif (RR) in the signal peptide (**Figure 2C**). The basic structure of the Tat system is divided into two complexes: a docking complex and a pore complex. The docking complex (TatB and TatC) recognizes the RR motif of the Tat signal peptide in the folded protein. Then, the folded protein is translocated across the active pore complex (TatA), with a structural change of the docking complex (Goosens et al., 2015).


Frontiers in Microbiology | www.frontiersin.org

Lee and Kim Protein Expressions in *Corynebacterium glutamicum*

carboxy-terminal region (C-region). The difference between the two pathways is that the N-region of the twin-arginine translocation (Tat)-type signal peptide is longer than the secretory (Sec)-type signal peptide because the Tat-type signal peptide contains a conserved twin-arginine residue (RR) at the end of the N-region [CspAC.a, surface (S)-layer protein from *Corynebacterium (Continued)* FIGURE 2 | *ammoniagenes*; PorBC.g, porinB from *C. glutamicum*; TorAE.c, TMAO reductase from *Escherichia coli*; PhoDC.g, alkaline phosphatase from *C. glutamicum*] (Berks et al., 2000). (B) Protein translocation by the Sec pathway. Sec translocase consists of the following components: SecYEG, a core protein in Sec translocase that forms the transmembrane protein-conducting channel (PCC), and SecDF, interacts with YajC to improve protein transport efficiency driven by the proton motive force (Scotti et al., 1999). In the co-translational targeting Sec pathway, signal recognition particles (SRPs) bind to the signal peptide at the beginning of translation where proteins are still bound to ribosomes. Then, the SRPs and the initial ribosomal protein (nascent protein) migrate to the SRP receptor and membrane protein FtsY and subsequently come in contact with SecYEG. The nascent protein passes through SecYEG while the ribosome is attached. In the post-translational targeting Sec pathway, a translation-finished protein binds to SecB without ribosome and then migrates to SecA-SecYEG complex. The delivered protein then passes through SecYEG while SecA is attached. (C) Protein translocation by the Tat pathway. The Tat system consists of TatA-like proteins (TatA, TatB, and TatE) and TatC (TatE seems to have the same function as TatA, though the difference is not clear yet). Translocation begins when the folded cargo proteins interact with the docking complex. The twin-arginine (RR) motif of the Tat signal peptide attaches to the signal peptide-binding loop of TatBC. The docking complex recognizes the cargo protein and inserts it into the membrane. TatA receives the cargo protein from the docking complex, and the cargo protein is translocated across the active pore complex. The signal peptide is then cleaved by type I signal peptidase, and the mature protein is separated from the cell membrane (Tuteja, 2005).

# Examples of Recombinant Protein Secretion in *C. glutamicum*

The production of a recombinant protein in C. glutamicum by a protein secretion system (the secretion of α-amylase from Bacillus amyloliquefaciens using the Sec system) was first reported by Smith et al. (Smith et al., 1986). Subsequently, protease, transglutaminase, green fluorescent protein (GFP), subtilisin, and endoglucanase have been produced in the C. glutamicum secretion system (**Table 3**).

Some studies have shown that eukaryotic proteins (such as human or camelid proteins) and microbial proteins could be successfully expressed in C. glutamicum. Yim et al. produced 68 mg/Lmedium of a single-chain variable fragment (scFv) with anthrax toxin as an antigen in a Sec system by codon optimization using a strong promoter (Yim et al., 2014). When the same M18 scFv was expressed in E. coli, a slightly higher level of the protein was obtained (89.8 mg/Lmedium). Nevertheless, the use of C. glutamicum may be safer for drugs such as antibodies because endotoxins are not produced by a GRAS host, unlike the case of E. coli, and the secreted proteins are stable because there is no extracellular protease activity. Grampositive bacteria such as C. glutamicum have no outer membrane; thus, target proteins need to pass through only one membrane to move out of the cell (van Wely et al., 2001). In addition, yeast cells that can glycosylate proteins when producing fulllength antibodies are mainly used. However, in contrast to the glycosylation system of mammalian cells, yeast cells have a mannose-rich glycosylation system; thus, they are often not suitable for use in medicine. In particular, post-translational modifications such as glycosylation in Pichia pastoris often lead to unexpected protein structure and function (Dai et al.,

#### TABLE 3 | Examples of protein secretions in *Corynebacterium glutamicum*.


\**B.s, Bacillus subtilis; C.a, Corynebacterium ammoniagenes; C.g, Corynebacterium glutamicum; E.c, Escherichia coli.*

2015). Nevertheless, C. glutamicum may be advantageous as a host for the expression of antibody fragments such as the scFv and Fab (antigen-binding fragment), which do not require glycosylation (Yim et al., 2014). Matsuda et al. produced 57 mg/Lmedium of an Fab fragment of anti-human epidermal growth factor receptor 2 (anti-HER2) using the Sec-secretion system with a cell wall protein-deficient C. glutamicum strain (Matsuda et al., 2014). This was based on the formation of an intermolecular disulfide bond when the heavy and light subunits of anti-HER2 Fab fragments were present at the same time. In another study, Date et al. reported the production of 156 mg/Lmedium of an active human epidermal growth factor (hEGF) with six cysteine residues that form three disulfide bonds, using the Sec-secretion system in C. glutamicum (Date et al., 2006). Therefore, C. glutamicum is an attractive secretory expression host for the production of medicinal proteins containing disulfide bonds as well as heterologous enzymes.

Efforts have also been made to introduce new signal peptides in C. glutamicum. An analysis of the secretion of C. glutamicum at high cell densities showed that the most abundant protein (51% of extracellular proteins) in the culture supernatant was a hypothetical protein encoded by cg1514. Using the promoter and signal peptide of the Cg1514 protein, three target proteins [endocolanicol A, 1.07 g/Lmedium; S. bovis α-amylase, 0.78 g/Lmedium; camelid antibody fragment (VHH) for human lysozyme, 1.58 g/Lmedium] were produced (Yim et al., 2016). These results suggest that Cg1514-derived expression and secretion signals may be particularly effective in the production of secretory proteins from C. glutamicum.

Although not as common as the Sec system, there have been attempts to secrete proteins such as GFP or α-amylase using the Tat system in C. glutamicum. In particular, the Tat system is sometimes necessary because protein folding and the insertion of some cofactors into the proteins must occur in the cytoplasm. As the Tat system can transduce the substrates in a

#### TABLE 4 | Examples of inducible and constitutive promoters in *Corynebacterium glutamicum*.


#### TABLE 5 | Examples of expression vectors in *C. glutamicum*.


fully collapsed state through the cytoplasmic membrane, the use of the Tat-pathway for enzyme secretion has been investigated. To this end, FAD cofactor-containing sorbitol-xylitol oxidase (SoXy), which is a cytosolic enzyme of Streptomyces coelicolor, was expressed in C. glutamicum using the Tat secretion system (Scheele et al., 2013). This study demonstrated that heterologous proteins containing cofactors can also be produced using the C. glutamicum secretion system.

# GENETIC TOOLS FOR PROTEIN EXPRESSION IN *C. glutamicum*

To express recombinant proteins efficiently for the amino acid, food, and pharmaceutical industries, it is necessary to precisely control the expression of genes and to optimally control the metabolic flow toward the target protein or amino acid. Therefore, to produce the target protein efficiently, it is important to do the following: (1) optimize the promoter to increase expression efficiency, (2) construct a plasmid vector for various kinds of proteins, (3) construct an efficient protein-secretion pathway, and (4) design a C. glutamicum bioreactor culture system for high-yield production.

There have been attempts to increase the yield of expression systems, by using promoters (**Table 4**), which are mainly used for the production of amino acids and industrial enzymes with C. glutamicum as a host. The selection of optimal promoter and regulatory sequences is essential for producing useful products in living organisms. Promoters that are mainly used in C. glutamicum include several inducible promoters such as PlacUV<sup>5</sup> , Ptac, Ptrp, ParaBAD, Ptrc, , and the phage PR/P<sup>L</sup> promoter from E. coli (Rytter et al., 2014). However, due to the low isopropyl-β-D-thiogalactopyranoside (IPTG) permeability of C. glutamicum, an IPTG-inducible expression system would have a lower expression level in C. glutamicum than in E. coli. Therefore, studies have been carried out to improve the promoter core sequence and membrane permeability of C. glutamicum and to increase gene expression by the site-directed mutagenesis. For example, a single-site mutation of the wild-type lac promoter has been used to enhance its protein expression level (Brabetz et al., 1991). In addition, the tac-M primer for constructing the tac promoter was found to increase the promoter activity following a mutation at the −10 region (Xu et al., 2010).

An auto-inducible promoter is a promoter that expresses proteins according to variables such as the nutrient type/concentration, oxygen level, pH level, and cell growth stage (Chou et al., 1995); this is advantageous for producing recombinant proteins on an industrial scale. Kim et al. (2016) engineered the SigB-dependent cg3141 promoter in C. glutamicum to develop an auto-inducible promoter system that is capable of expressing recombinant proteins during the transition phase between the log phase and the stationary phase of the cells. As a result, the model protein, glutathione S-transferase, was successfully produced on a lab-scale bioreactor (5 L) by introducing the P4−N14 promoter (Kim et al., 2016).

The use of constitutive promoters is advantageous because they do not require expensive reagents for induction or optimized induction conditions (Yim et al., 2013). Constitutive promoters derived from the genome of C. glutamicum, such as PsodA, PgapA, Peftu, and PcspB, are known to have high expression levels. However, the strength of the promoters cannot be directly compared and the use of strong promoters can also be affected by other genetic elements such as the 5′ -untranslated region (5′ -UTR) (Teramoto et al., 2011) and transcription initiation region (TIR) (Yim et al., 2013). Therefore, the selection of an optimal promoter is required because a strong promoter alone does not guarantee high protein expression. Yim et al. have developed the first synthetic promoter in C. glutamicum (Yim et al., 2013). Sequences including PL10, PL26, PI16, PI51, PH30, and PH36 were selected from the promoter library, which consisted of 70 randomly chosen nucleotide sequences. Among them, PH36 was the strongest promoter, and it successfully induced the expression of antibody fragments and endoxylanase (746 mg/Lmedium), as model proteins. Appropriate expression vectors and promoters are also important for increasing the yield of recombinant proteins. Currently, several C. glutamicum–E. coli shuttle expression vectors are being widely used (**Table 5**).

# CONCLUSION

C. glutamicum can be used as an industrial L-glutamate and L-lysine producer. In addition, various types of recombinant proteins can be expressed in C. glutamicum, which has been used for several decades for the production of microbial enzyme. Furthermore, C. glutamicum has been used to increase yields, develop new anchoring systems, and signal peptides (for the efficient production of biochemicals and nutraceuticals, enzymes, medicinal proteins, and biopolymers), and screen synthetic promoters of various strengths. However, using C. glutamicum as an expression host has several disadvantages when compared with using E. coli as an expression host: (1) a much lower transformation efficiency, (2) fewer available expression systems, and (3) lower yields for some proteins, especially antibodies. Therefore, further studies are necessary to develop various tools to enhance protein yields and reduce manufacturing costs. Recent advances in bioinformatics, such as next-generation sequencing (NGS), RNA-seq, and proteomics, would provide more information on the protein production pathways in C. glutamicum.

# AUTHOR CONTRIBUTIONS

ML wrote the manuscript and PK supervised. All authors have made intellectual contributions to the work, and approved it for publication.

# ACKNOWLEDGMENTS

This work was financially supported by the grants from the National Research Foundation of Korea (2016R1E1A1A01943552). The authors extend their appreciation to the A3 Foresight Program (2016K2A9A2A10005545) for supporting their travel.

# REFERENCES


deletion of insertion sequences (IS elements). Microb. Cell Fact. 14:207. doi: 10.1186/s12934-015-0401-7


from Rhodococcus rhodochrous for acrylamide production. Appl. Microbiol. Biotechnol. 98, 4379–4387. doi: 10.1007/s00253-014-5544-7


Corynebacterium glutamicum in anaerobic conditions. AMB Express 3:72. doi: 10.1186/2191-0855-3-72


**Conflict of Interest Statement:** 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 © 2018 Lee 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.

# Engineering of the Filamentous Fungus Penicillium chrysogenum as Cell Factory for Natural Products

Fernando Guzmán-Chávez1,2† , Reto D. Zwahlen1,2, Roel A. L. Bovenberg2,3 and Arnold J. M. Driessen1,2 \*

<sup>1</sup> Molecular Microbiology, Groningen Biomolecular Sciences and Biotechnology Institute, University of Groningen, Groningen, Netherlands, <sup>2</sup> Synthetic Biology and Cell Engineering, Groningen Biomolecular Sciences and Biotechnology Institute, University of Groningen, Groningen, Netherlands, <sup>3</sup> DSM Biotechnology Centre, Delft, Netherlands

#### Edited by:

Eung-Soo Kim, Inha University, South Korea

#### Reviewed by:

Gang Liu, Institute of Microbiology (CAS), China Juan F. Martin, Universidad de León, Spain

> \*Correspondence: Arnold J. M. Driessen a.j.m.driessen@rug.nl

#### †Present address:

Fernando Guzmán-Chávez, Synthetic Biology and Reprogramming of Plant Systems, Department of Plant Sciences, University of Cambridge, Cambridge, United Kingdom

#### Specialty section:

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

> Received: 19 June 2018 Accepted: 29 October 2018 Published: 15 November 2018

#### Citation:

Guzmán-Chávez F, Zwahlen RD, Bovenberg RAL and Driessen AJM (2018) Engineering of the Filamentous Fungus Penicillium chrysogenum as Cell Factory for Natural Products. Front. Microbiol. 9:2768. doi: 10.3389/fmicb.2018.02768 Penicillium chrysogenum (renamed P. rubens) is the most studied member of a family of more than 350 Penicillium species that constitute the genus. Since the discovery of penicillin by Alexander Fleming, this filamentous fungus is used as a commercial β-lactam antibiotic producer. For several decades, P. chrysogenum was subjected to a classical strain improvement (CSI) program to increase penicillin titers. This resulted in a massive increase in the penicillin production capacity, paralleled by the silencing of several other biosynthetic gene clusters (BGCs), causing a reduction in the production of a broad range of BGC encoded natural products (NPs). Several approaches have been used to restore the ability of the penicillin production strains to synthetize the NPs lost during the CSI. Here, we summarize various re-activation mechanisms of BGCs, and how interference with regulation can be used as a strategy to activate or silence BGCs in filamentous fungi. To further emphasize the versatility of P. chrysogenum as a fungal production platform for NPs with potential commercial value, protein engineering of biosynthetic enzymes is discussed as a tool to develop de novo BGC pathways for new NPs.

Keywords: Penicillium chrysogenum, natural products, nonribosomal peptides, polyketides, gene activation, biosynthetic gene clusters, cell factory

#### INTRODUCTION

Since the discovery of penicillin by Alexander Fleming produced by the filamentous fungus Penicillium notatum, the genus Penicillium has been deeply studied for its capacity to produce a wide range of natural products (NPs) (secondary metabolites), many of them with biotechnological and pharmaceutical applications. P. chrysogenum (recently renamed as P. rubens) is the most relevant member of more than 354 Penicillium species that constitute the genus (Nielsen et al., 2017). Penicillium is usually found in indoor environments and associated with food spoilage. It is known as an industrial producer of β-lactam antibiotic in particularly penicillin, and current production strains result from several decades of classical strain improvement (CSI) (Gombert et al., 2011; Houbraken et al., 2011). The CSI program began in 1943 with the isolation of P. chrysogenum NRRL 1951 capable of growing in submerged cultures. This strain was subjected to a long serial process of mutations induced by 275 nm ultraviolet and X-ray irradiation, nitrogen mustard gas and nitroso-methyl guanidine exposure, single spore selection and selection for loss

of pigments, improved growth in large scale industrial fermenters and enhanced levels of penicillin production. CSI programs were developed in several companies (Barreiro et al., 2012), and this has resulted in an increase of penicillin titers by at least three orders of magnitude (van den Berg, 2010). As consequence, numerous genetic modifications were introduced in P. chrysogenum. Some have been studied in detail, most notably the amplification of the penicillin biosynthetic clusters and DNA inversions in this region (Fierro et al., 1995, 2006; Barreiro et al., 2012). Although the CSI had a major impact on the production of β-lactams by P. chrysogenum, it also affected secondary metabolism in general. Indeed, a proteome analysis performed between P. chrysogenum NRRL 1951 and two derived strains (Wisconsin 54-1255 and AS-P-78) showed reduced levels of proteins related to secondary metabolism in the higher penicillin producer strains (Jami et al., 2010). Genome sequencing of P. chrysogenum Wisconsin 54-1255 revealed the presence of several secondary metabolite encoding biosynthetic gene clusters (BGCs) in addition to the penicillin cluster, most of which have only be poorly studied and remain to be characterized (**Figure 1**). The products of the BGCs are either nonribosomal peptides (NRPs), polyketides (PKs) or hybrid molecules.

Penicillium chrysogenum produces a broad range of secondary metabolites such as roquefortines, fungisporin (a cyclic hydrophobic tetrapeptide), siderophores, penitric acid, ω-hydroxyemodin, chrysogenin, chrysogine, sesquiterpene PR-toxin and sorbicillinoids, but likely also possesses the ability to produce other compounds not detected before. For most of the identified compounds, the responsible BGCs are unknown. The development of new bioinformatics tools (SMURF, AntiSMASH) (Khaldi et al., 2010; Weber et al., 2015; Blin et al., 2017) and the increase in the number of fungal genomes sequenced to date has opened the possibility to discover new NPs with novel properties (genome mining). The genes involved in the biosynthesis, regulation and transport of secondary metabolites tend to be arranged in the genome in clusters. Importantly, these gene clusters include the core biosynthetic genes which either encode polyketide synthases (PKSs), nonribosomal peptide synthetases (NRPSs) or terpene synthases genes (Smanski et al., 2016). Recently, a global analysis was performed on 24 genomes of Penicillium species and this identified 1,317 putative BGCs predominated by two classes based on PKS (467) and NRPS (260) (Nielsen et al., 2017). In P. chrysogenum there are 33 core genes in the secondary metabolism that encode 10 NRPS, 20 PKS, 2 hybrid NRPS–PKS, and 1 dimethyl-allyl-tryptophan synthase (van den Berg et al., 2008; Khaldi et al., 2010; Medema et al., 2011; Samol et al., 2016) (**Figure 1**). A large number of PKS and NRPS enzymes are found also in other Penicillium species but only part of these gene clusters are shared, which suggests an unexplored potential of the secondary metabolome even in a single genus.

Here, we summarize the most recent strategies for engineering filamentous fungi with particular attention to P. chrysogenum, a promising cell factory of novel products with new application spectra. A brief description of the key biosynthetic enzymes involved in biosynthesis of secondary metabolites in fungi is provided.

# THE BUILDING ENZYMES OF THE NATURAL PRODUCTS

Nonribosomal peptide synthetases are large, highly structured and complex enzymatic machineries, closely related to other modular enzymes such as PKSs, NRPS–PKS hybrid synthetases and fatty acid synthetases (FASs). They have certain distinct properties in common, the most striking one being their structural division in domains and modules, which is manifested in their shared evolutionary history (Smith and Sherman, 2008). Every enzyme minimally consists of one module, a functionally distinct unit, which allows for the recruitment and subsequent incorporation of a precursor into a growing product. Domains as well as modules are clearly defined and evolutionary exchangeable structures amongst multi-modular enzymes. In the case of PKS and NRPS, this led to the occurrence of a variety of NRPS–PKS hybrids (Du and Shen, 2001; Shen et al., 2005; Li et al., 2010; Nielsen et al., 2016).

# Nonribosomal Peptides (NRPs) and Nonribosomal Peptide Synthetases (NRPSs)

In comparison to most ribosomally derived peptides, NRPs are low molecular weight products. The structural diversity of NRPs is tremendous, mostly due to their chemical complexity. Significantly contributing to this diversity is the fact that NRPS are not only reliant on proteinogenic amino acids, but up until now more than 500 substrates were identified, which serve as NRPS building blocks (Caboche et al., 2008). These molecules are predominantly amino acids, but not exclusively, since fatty acids, carboxylic acids and others substrates have been reported in NRPs (Marahiel et al., 1997). NRP thus represent a diverse group of natural compounds and occur as linear, branched, circular or macrocircular structures (Dang and Sussmuth, 2017; Sussmuth and Mainz, 2017). The natural functions of NRPs are as diverse as their structures. Signaling, communication, metal-ion chelation, host protection are important functions performed by NRPs, though many compounds are not yet fully characterized in this respect. Nevertheless, the characterization of NPs for applied purposes is well developed and led to a vast collection of ground-breaking pharmaceuticals, including antibiotics, antifungal agents, immunosuppressants as well as cytostatic drugs (Frisvad et al., 2004; Watanabe et al., 2009; Dang and Sussmuth, 2017).

Structurally, every NRPS module, initiation (1), elongation (n) or termination (1), requires a minimal set of domains (**Figure 2A**) (Stachelhaus and Marahiel, 1995). The two domains essential to every module are the adenylation domain (A) and the non-catalytic thiolation domain (T). This tandem di-domain enables the specific selection and activation of a given substrate. However, the T-domain must first go through 4 0 -phosphopantetheinyl transferase (PPTase) and coenzyme A (CoA) dependent activation after expression, by transferring the phosphopantetheine moiety of CoA onto a conserved serine residue, in order to enter the holo state. Also, adenylation domains (A) have accompanying factors, or proteins, called

MbtH-like proteins (MLPs) (Quadri et al., 1998; Baltz, 2011). In contrast to PPTases, MLPs are merely interacting with the A-domain, however, they do not have an intrinsic enzymatic activity, but rather a chaperoning function upon binding a distinct part of the A domain (Felnagle et al., 2010; Miller et al., 2016; Schomer and Thomas, 2017). In addition to these domains, any elongation module will require a condensation domain (C), which connects two modules and links upand downstream activated substrates via a peptide bond. C-domains are stereospecific for both, up- and downstream activated substrates and render the resulting intermediate compound attached to the downstream T-domain. Lastly, the C-terminal termination module essentially requires a thioesterase domain (Te), to catalytically release the covalently bound compound of the NRPS, returning the NRPS complex to the ground state for another reaction cycle. In addition to these essential domains, we can distinguish a series of additional domains, performing epimerization, halogenation, cyclization, macrocyclization, multimerization or methylation (Ansari et al., 2008; Horsman et al., 2016; Bloudoff and Schmeing, 2017).

Theses enzymatic machineries can be classified as type I NRPS when the modules are arranged on a single protein, while the type II NRPS are independent proteins in an transient manner during the NRP synthesis (Sattely et al., 2008; Hur et al., 2012). A NRPS can be as simple as a single modular unit containing three domains, although the most complex and largest structure known contains 15 modules with 46 domains (Wang H. et al., 2014; Bode et al., 2015) yielding a 1.8 MDa protein complex (type I NRPS). Although the size of a NRPS, as well as the modular sequence, limits the setup of the resulting NRP, it is common that NRPS cluster and interact with tailoring enzymes in order to produce products of a higher complexity (Yin and Zabriskie, 2006). To enable such specific interactions, NRPS can contain small stretches of up to 30 amino acids at the Cor N-terminus, which form a rather specific recognition point, thus enabling communication (COM-domain) between multiple NRPS of one cluster (type II NRPS) (Hahn and Stachelhaus, 2006; Dowling et al., 2016). To date three types of NRPS system have been described according to their synthesis mode (or strategy of biosynthesis): Type A (linear), type B (iterative) and type C (non-linear) (**Figure 3**). The **type A system** harbors the typical domain organization A-T(C-A-T)n−1-Te/C, where n represents the number of amino acids in the peptide. In this linear NRPS, the order and number of modules correlates with the amino acid sequence in the NRP and thus it is possible to predict the product that will be formed. Usually, in fungal NRPSs the cyclisation reaction is performed by a specialized C domain instead of Te domain. Since each module catalyzes one cycle during the chain elongation of the nascent NRP due to its specific activity, this system is considered analogous to type I PKS. In fungi, the most prominent examples of this type of NRPS are ACV synthetases (β-lactams), cyclosporin synthetases (Cyclosporin A) and peptaibol synthetases (peptaboils, a class of antibiotics with a high content of α-aminoisobutyric acid) (Wiest et al., 2002; Tang et al., 2007; Felnagle et al., 2008; Eisfeld, 2009). The **type B system** is characterized to employ all their modules or domains more than once during the synthesis of a single NRP, which enables the assembling of peptide chains that contain repeated sequences along the structure (Mootz et al., 2002). An example of this mode of synthesis occurs in Fusarium scirpi during the biosynthesis of enniatin (antibiotic), which is achieved through the repeated use of two modules. Other examples of type B NRPSs are the siderophore synthetases, which only contain three A domains that catalyze the biosynthesis of ferrichrome (Mootz et al., 2002; Eisfeld, 2009). In **type C system**, the non-linear NRPSs have at least one domain conformation that deviate from (C-A-T)n−<sup>1</sup> organization contained in linear NRPSs. Likewise, in these synthetases the module arrangement does not correspond to the amino acid sequence in the NRP. Unlike type A NRPS, in type C NRPSs the non-linear peptide is

produced by a branch-point synthase and contains uncommon cyclization patterns. Another important difference is that nonlinear NRPSs can incorporate small soluble structures, such as amines into the rising NRP through specialized C domains (Tang et al., 2007; Felnagle et al., 2008; Hur et al., 2012). Capreomycin, bleomycin and vibriobactin are examples of NRPs produced by this type of synthetases (Felnagle et al., 2008). In continuation, a brief description of the main NRPS domains features is provided.

#### Adenylation (A) and Thiolation (T) Domains

Any NRPS module minimally consists of an A- and T-domain (or peptidyl carrier protein, PCP), enabling single module functionality and multi-modular functionality upon addition of C domains (Linne and Marahiel, 2000; Bergendahl et al., 2002; Felnagle et al., 2008; Kittilä et al., 2016; Bloudoff and Schmeing, 2017). They are often referred to as "gatekeeper" domains, as there is no subsequent product formation without prior adenylation and thioesterification of a substrate (Sun et al., 2014). The two core functions of the A-domain are characterized first, through the hydrolysis of ATP or adenylation, allowing an AMP-substrate conjugate to be formed, which is subsequently transferred to the free thiol group of the 4 0 -phosphopantetheinyl-moiety (Ppant), which is anchored to a conserved serine residue in the downstream T-domain (Ku et al., 1997; Weber and Marahiel, 2001; Neville et al., 2005).

#### Condensation Domains (C)

C-domains are approximately 450 residue NRPS domains, representing a highly versatile class of NRPS domains. Any NRPS composed of more than one module must consequently contain at least one C-domain. However, also single modular NRPS may contain C-domains, especially if they cooperate with other NRPS. Essentially, the primary target of a C-domain is the condensation of the up- and downstream activated substrates through a nucleophilic attack, mainly leading to the formation of an n-peptide linked via a peptide bond. Nonetheless, several residues of the C-domain may have the intrinsic potential to fulfill multiple functions (Balibar et al., 2005; Teruya et al., 2012; Haslinger et al., 2015).

#### Epimerization Domains (E)

The E-domains are among the most abundant modification domains intrinsic to NRPS. In contrast to the structurally similar C domains they are responsible for the site specific epimerization of a substrate, predominantly performing this function after peptide bond formation has occurred (Bloudoff and Schmeing, 2017).

#### Thioesterase Domains (Te)

fmicb-09-02768 November 13, 2018 Time: 17:47 # 6

The thiotemplate based enzymatic systems rely on a catalytic activity in order to remove a product or product-scaffold of the primary enzyme. Therefore, most NRPS contain a domain on their C-terminus responsible for precisely this purpose, the thioesterase domain (Te). Te-domains are a common commodity in single and multi-modular NRPS, although, in multi-NRPS systems only the terminal NRPS contains this domain (Horsman et al., 2016). Additionally, this domain harbors a quality control activity (proofreading) to verify the correct configuration of the nascent peptide (Martín and Liras, 2017).

#### Intrinsic Product Modifying Domains

In addition to the C-domain related epimerization domain, discussed previously, there are cyclization (Cy), oxygenation (Oxy) as well as methyl-transferase (MT). These domains have been characterized to the extent of classifying their functions, although, especially Cy- and Oxy-domains may occur as a singular bi-functional unit or in a serial manner, respectively (Walsh, 2016). Cy- and Oxy-domains, specifically replace the classic function of C-domains, omitting amino acid condensation through peptide bond formation, resulting in thiazoline, oxazoline or methyloxazoline structures (Sundaram and Hertweck, 2016; Walsh, 2016). Those reactions predominantly occur in siderophore producing NRPS and rely on the presence of serine, threonine and cysteine residues (Patel et al., 2003; Kelly et al., 2005). Also MT-domains follow the common di-sub-domain structural patterning, which is also seen in A-, C-, E-, and Cy-domains. Fundamentally, MT-domains, however, are more restricted in their function, which covers the transfer of methyl-groups from S-adenosylmethionine to N (N-MT), C (C-MT), O (O-MT) or for certain residues S (S-MT) atoms resolving around the amino acids Cα carbon (Miller et al., 2003) and in case of S-MT Cβ, respectively (Al-Mestarihi et al., 2014).

In P. chrysogenum 10 NRPS have been identified (**Table 1**) (van den Berg et al., 2008; Medema et al., 2011; Samol et al., 2016), of which only two have no attributed function. In this fungus, next to fungisporin (**Figure 1**) which is a cyclic hydrophobic tetrapeptide generated by a singular NRPS, three biosynthetic pathways involving a NRPS have been described in detail: penicillins, roquefortine/meleagrin, fungisporin and chrysogine (**Figure 4**).

#### Polyketides and Polyketide Synthase

Polyketides were already discovered in 1883 by James Collie, but the interest in these compounds (enzymes) was revived only as late as the 1950s by the work of Arthur Birch on the aromatic polyketide-6-methyl salicylic acid from P. patulum. These molecules are a class of NPs, that display different types of biological activities such as antibiotic (erythromycin A), antifungal (amphotericin B), immunosuppressant (rapamycin), antitumor (geldanmycin) and hypolipidemic (lovastatin) (Nair et al., 2012; Jenner, 2016; Weissman, 2016). Their assembly process is similar to that in the fatty acid biosynthesis, the main difference is the optional full reduction of the β-carbon in the PK biosynthesis. The group of enzymes that catalyzes the

biosynthesis of PKs is referred to as PKSs (Keller et al., 2005; Caffrey, 2012).

In addition to the NRPSs, PKSs are the main enzymes that build the structural scaffold of a wide range of secondary metabolites and NPs in plants, bacteria, insects and fungi (Brakhage, 2012; Nair et al., 2012). Usually, these enzymes are encoded by genes that are grouped into clusters, that also specify genes encoding tailoring enzymes (oxygenases, oxidoreductases, reductases, dehydrogenases,and transferases), that further modify the scaffold produced by the PKS into a final product (Brakhage, 2012; Lim et al., 2012). PKSs are multimodular and multidomain enzymes that use a specific acyl-coenzyme A (acyl-CoA; usually malonyl-CoA or methylmalonyl-CoA) as building block, and subsequently catalyze a decarboxylative Claisen-type condensation of ketide units. The basic structural architecture consists of an acyl carrier protein (ACP), a ketosynthase (KS) and an acyltransferase (AT) domain. These combined domains extent a linear intermediate by two carbon atoms. An optional set of domains (dehydratase (DH), ketoreductase (KR), enoyl reductase (ER) and thioesterase (TE) may provide further modifications of the linear intermediate (Staunton and Weissman, 2001; Brakhage, 2012; Nair et al., 2012; Dutta et al., 2014).

According to their protein architecture and mode of action, PKS enzymes are classified into types I, II, and III (**Figure 5**). **Type I PKSs** are mainly found in bacteria and fungi. These multidomain proteins can be further subdivided in two categories: modular and iterative (Nair et al., 2012) Modular type I PKSs or non-iterative PKSs are unique for bacteria and are characterized by presenting a sequence (or set) of modules, each constituted with a set of specific catalytic domains. In consequence, the number of precursors fused in the PK is equivalent to the number of modules which are present (Chan et al., 2009). In contrast, iterative type I PKSs use the same catalytic core domains as modular type I PKSs, but the catalytic reaction is repeated to yield the complete PK backbone. A representative example of this type is LovB, that together with LovC (a enoyl reductase) catalyzes around 35 reactions to produce dihydromonacolin L, an intermediate in the lovastatin biosynthesis (Chan et al., 2009; Campbell and Vederas, 2010). Like iterative type I PKS enzymes, fungal PKSs (**Figure 2B**) are restricted to a single module and the consecutive domains act in sequential order during the synthesis of the complete PK. They are equipped with basic structural domains typically found in PKS enzymes (ACP-KS-AT domains) but may also contain optional units (KR, DH, ER, and Te domains). Depending on the presence or absence of reducing domains, these enzymes can be divided into highly reducing (HR), nonreducing (NR) and partially reducing (PR) PKS (**Figure 2B**) (Keller et al., 2005; Crawford and Townsend, 2010; Jenner, 2016).

#### Highly Reducing PKS (HR-PKS)

Highly reducing PKS (HR-PKS) produce the linear or cyclic scaffold of some compounds such as fumonisins, T-toxins, solanapyrone E, squalestatin or/and lovastatin (Chiang et al.,

2014; Roberts et al., 2017). Usually, they start with a KS domain, followed by an AT, DH and C-Met domain, although the latter does not always follow the DH domain. The ER domain is an optional unit in HR-PKS enzymes, but when the ER is missing, the corresponding region is filled with a polypeptide domain with an unknown function. Furthermore, these enzymes do not contain a product template domain (PT) or N-terminal SAT domain, whereas these special domains are present in NR-PKS enzymes (Cox and Simpson, 2010).

#### Partially Reducing PKS (PR-PKS)

Structurally, these enzymes have a domain architecture that is similar to the mammalian FAS: a N-terminal KS-domain followed by an AT-, DH-, and "core"-KR-ACP domain. These


A, adenylation; T, thiolation; E, epimerization; te, thioesterase; C, condensation. <sup>∗</sup>Point mutations present in nrps genes of industrial P. chrysogenum strains subject to CSI program. Modified from Salo et al. (2015), Samol et al. (2016), and Guzmán-Chávez et al. (2018).

enzymes lack an ER domain (Wang L. et al., 2015), and also do not have a Te domain, which suggests an alternative mechanism of product release than hydrolysis. PR-PKS enzymes produce small aromatic molecules such as 6-methylsalicylic acid (MSA), but in most cases the chemical product is unknown (Cox and Simpson, 2009, 2010; Kage et al., 2015).

#### Non-reducing PKS (NR-PKS; Aromatic PKs)

Non-reducing PKS (NR-PKS; aromatic PKs) typically, consist of six catalytic domains that are covalently associated and arranged in four components: loading (SAT), chain extension (KS-MAT-PT-ACP), cyclisation and processing components (TE-CLC) (Bruegger et al., 2013).

**Type II PKSs** are unique for bacteria and use a similar iterative mechanism as observed in iterative type I PKSs. However, the different catalytic domains are encoded by independent genes. In general, they often constitute a "minimal PKS," which comprises of two KS units (KSα and KSβ) and an ACP protein that holds the growing PKS chain. The KSβ domain defines the length of the PK chain. The folding pattern of the poly-β-keto intermediates is determined by optional PKS units such as aromatases, ketoreductases, and cyclases. Other tailoring modifications are performed by oxygenases, methyl and glycosyl transferases. Known metabolites synthetized by type II PKSs are tetracyclines, anthracyclines and aureolic acids (Hertweck et al., 2007; Jenner, 2016). **Type III PKSs** have originally been discovered in plants but are also present in bacteria and fungi. They consist of a single KS domain that catalyzes a defined number of elongations, usually generating small phenols or naphtol rings. The enzyme transfers the acyl

group from the CoA to the active site histidine, which is a highly conserved residue. However, the amino acid sequence of the his motif is not similar to those found in KS domains of type I and II PKS enzymes (Shen, 2003; Chan et al., 2009; Bruegger et al., 2014; Jenner, 2016). Importantly, independent of the mechanistic or structural differences, all the PKs synthetized by PKS enzymes follow the same decarboxylative condensation mechanism of the acyl-CoA precursors. However, these precursors should be activated in prior by the ACP domain, in the case of the type I and II PKS enzymes, whereas type III PKS enzymes act independently of ACP domains (Shen, 2003; Hopwood, 2009). Acridones, pyrenes as well as (and) chalcones are some examples of the compounds produced by type III PKS enzymes (Yu et al., 2012). Below, a brief description is provided on the main catalytic features of PKS domains.

#### Acyltransferase Domains (AT)

A main unit during PK biosynthesis is the AT domain that selects the starter unit (malonyl-CoA or methylmalonyl-CoA) before it is transferred to the ACP domain for the chain elongation cycle (Dunn et al., 2013). This process involves two steps, i.e., the acylation and the transfer to the ACP (Jenner, 2016).

#### Acyl Carrier Protein (ACP)

The ACP is an essential cofactor that participates in PK biosynthesis. This protein belongs to a highly conserved carrier family, and consists of 70–100 amino acid residues (Byers and Gong, 2007). To perform the PK biosynthesis, the holo-ACP (active) form is generated by the phosphopantetheinyl transferase enzyme (PPTase) through a post-translational modification of ACP whereby a 4<sup>0</sup> -phosphopantetheine (4<sup>0</sup> -PP) moiety from CoA is transferred to the conserved serine (Evans et al., 2008; Kapur et al., 2010; Jenner, 2016) resulting in the formation of the Ppant arm. ACP modulates three important events during PK biosynthesis. First, it allows the condensation during chain elongations since it transfers the starter unit from the AT domain to the KS domain. Second, it shuttles the growing chain between the up and downstream domains, as well as to optional PKS domains, probably involving protein–protein recognition between domains. Third, it prevents premature cyclization and enolization of the PK chain (Yadav et al., 2013).

#### Ketosynthase Domains (KS)

The KS is a homodimeric condensing domain that catalyzes the extension of the β-ketoacyl intermediate by a decarboxylative Claisen condensation. This domain contains two active sites which are accessible to the ACP through its flexible Ppant arm, which receives the β-carboxyacyl-CoA extender unit from the AT. At that stage, a thioester bond is formed between the activesite cysteines' thiol group of the KS and the growing PK. Only when both units are covalently attached onto the module, a decarboxylative Claisen condensation occurs, which involves two conserved his residues. Therefore, mechanistically the KS domain acts at three stages: acylation, decarboxylation and condensation (Chen et al., 2006; Caffrey, 2012; Yadav et al., 2013; Jenner, 2016; Robbins et al., 2016).

#### Ketoreductase Domains (KR)

The KR domain functions as a β-carbon processing unit that belongs to the family of short-chain dehydrogenase/reductases. This domain reduces the β-keto group, that is formed during the condensation process, into a hydroxyl group (a β-hydroxyl intermediate) using NADPH (Keatinge-Clay and Stroud, 2006; Caffrey, 2012). Additionally, some KR domains are equipped with epimerase activity. The epimerizing module has a more open architecture, enabling the catalytic epimerization of methyl groups in acyl-ACP substrates, a reaction that involves the conserved serine and tyrosine residues which are also employed during ketoreduction (Ostrowski et al., 2016c; Bayly and Yadav, 2017).

#### Dehydratase Domains (DH)

The DH domain is usually coupled to B-type KR domains (Btype). This domain catalyzes water elimination (via syn or anti) at the β-hydroxy acyl chain position thereby producing trans double bonds (α,β-unsaturated moieties) (Caffrey, 2012; Bruegger et al., 2014; Jenner, 2016; Bayly and Yadav, 2017).

#### Enoyl Reductase Domains (ER)

The ER domain is an optional tailoring unit involved in the final oxidation state of the growing PK. It reduces α,β-enoyl groups and thereby generates saturated α–β bonds. This reaction involves NAD(P)H as hydride donor in a Michael addition type of mechanism. In the enoyl reduction, the products formed during this reaction have a specific stereochemistry (3R,2R) or (3R,2S) due to the β-carbon attack performed by the pro-4R hydride of NADPH, contrasting the KR domain that utilizes the pro-4S hydride (Chen et al., 2006; Bruegger et al., 2014).

#### Thioesterase Domain (Te)

Termination of PK biosynthesis involves the Te domain, which produces macrolactones via intramolecular cyclization or linear PKs by hydrolysis (Keatinge-Clay, 2012). In both events, an acyl-Te intermediate is formed through the transfer of the PK chain from the last ACP to the active serine on Te domain (Jenner, 2016).

#### Special Domains

In non-reducing PKS, the **ACP transacylase (SAT)** domain acts as starter unit that loads the ACP whereupon chain extension is mediated for KS and AT domain. During this process, the malonyl-CoA:ACP transacylase **(MAT) domain** transfers the extension units from malonyl-CoA to the ACP, while the product template **(PT) domain** stabilizes the reactive poly-β-keto intermediates. The processing component acts after the initial assembly when the cyclized or PK intermediate is still attached to the ACP. Final cyclization and release is catalyzed by the Te/Claisen cyclase **(CLC) domain** (Cox and Simpson, 2010; Crawford and Townsend, 2010; Bruegger et al., 2013; Chiang et al., 2014).

In P. chrysogenum, 20 PKS genes have been identified (**Table 2**) (van den Berg et al., 2008; Medema et al., 2011; Samol et al., 2016), but for only six the products are known. To date, in P. chrysogenum only four PK-related pathways have been

(2013), Heinekamp et al. (2013), Matsuda et al. (2013), Salo et al. (2015), Salo et al. (2016), Druzhinina et al. (2016), Meng et al. (2016), Salo (2016), Samol et al.

(2016), Guzmán-Chávez et al. (2017), Guzmán-Chávez et al. (2018), Nielsen et al. (2017), and Rojas-Aedo et al. (2017).

described in detail: sorbicillinoids, MSA-6/yanuthones, DHNmelanin and andrastin A (**Figure 6**).

#### Terpenoids Biosynthesis

In addition to NRPs and PKs, terpenoids are another class of NPs that are synthetized by filamentous fungi (Ascomycota) although less abundant as compared to Basidiomycota (Schmidt-Dannert, 2014). Fungal terpenoids or isoprenoids are structurally diverse molecules derived from isoprene units (C5 carbon skeleton): isopentenyl pyrophosphate (IPP) and dimethylallyl pyrophosphate (DMAPP), which are synthetized in the mevalonate pathway from acetyl-CoA (Chiang et al., 2014; Soltani, 2016). The head-to-tail condensation of these C5 units is catalyzed by isoprenyl diphosphate synthases (IDSs), producing isoprenyl diphosphates with 10 (geranyl, GPP), 15 (farnesyl, FPP), and 20 (geranylgeranyl, GGPP) carbons. Eventually, these linear chains of different length are further modified by cyclases, terpene synthases (TPs) and prenyl transferases (PTs), yielding different subclasses of terpenoids (Schmidt-Dannert, 2014; Chen et al., 2016). For instance, monoterpenoids, sesquiterpenoids, diterpenoids, sesterterpenoids, and triterpenoids, which harbor two to six isoprene units, respectively (Soltani, 2016). Terpenoids are oxygenated derivatives of terpenes, which are also derived of isoprene (Stashenko and Martinez, 2017).

In filamentous fungi such as Aspergillus, Penicillium, Claviceps, and Neosartorya, ABBA-type PTs are involved in the biosynthesis of a range of toxins (Schmidt-Dannert, 2014). For the synthesis of indole-diterpenoids, IPPS-type PTs transfer GGPP to a indole group, while UbiA-type PTs are involved in the biosynthesis of meroterpenoids, which are hybrid NPs (terpenoids and PKs) (Itoh et al., 2010; Schmidt-Dannert, 2014). In A. nidulans, AusN (UbiA-type TPs) converts the product of a NR-PKS (3,5-dimethylorsellinic acid) as part of an earlier step in the dehydroaustinol/austinol biosynthesis pathway (Lo et al., 2012).

Terpene synthases catalyze cyclization reactions forming the carbocation by substrate ionization (class I) or substrate protonation (class II) (Zhou et al., 2012; Meguro et al., 2014).

A relevant example of class I TPs are sesquiterpene synthases, which cyclize the FPP to obtain a sesquiterpene scaffold (C15 backbone) (Quin et al., 2014). Recently, the prx1 to prx4 gene cluster involved in the biosynthesis of PR-toxin in P. roqueforti was cloned and sequenced. This cluster contains the gene prx2 (ari1) that encodes for a aristolochene synthase which forms a sesquiterpene aristolochene derivative (precursor of PR-toxin). Interestingly, an orthologous gene cluster was identified in P. chrysogenum (Pc12g06300 to Pc12g06330), as part of BGC of eleven genes, which is also involved in the biosynthesis of PR-toxin (Hidalgo et al., 2014).

#### STRATEGIES FOR ACTIVATION OF BGCS

Natural products represent a broad range of molecules produced by animals, plants and microorganisms. These molecules may display different biological activities (e.g., antiviral, antimicrobial, anti-tumor, immunosuppressive agents) and it is estimated that the majority of these compounds are derived from filamentous fungal sources and from filamentous bacteria belonging to the genus Streptomyces. With respect to antibiotics, most of the chemical scaffolds used today were discovered during the golden age of antibiotics discovery (1940–1960s). This was followed by four decades during which hardly any new scaffolds from a natural source were developed (Reen et al., 2015; Smanski et al., 2016; Okada and Seyedsayamdost, 2017). However, there is also a current understanding that only a small fraction of the potential possible molecules has been discovered to this date. This follows from genomic studies revealing large numbers of uncharacterized BGCs, while many of these gene clusters are not expressed (silent or sleeping gene clusters) under laboratory conditions (Brakhage and Schroeckh, 2011). Furthermore, metagenomics studies indicate that the majority of microbes present in the environment have not been cultured nor characterized. Thus, there are many challenges that need to be overcome in order to harness the natural diversity of NPs, to cultivate potential strains under laboratory conditions and to activate the BGCs for expression. To achieve the synthesis of new NPs, three main approaches (**Figure 7**) were used in recent years, which may be successfully applied in P. chrysogenum: manipulation of cultivation conditions, engineering of NRPS and PKS and genetic interference.

#### Manipulation of Cultivation Conditions

Under natural conditions, fungi face a variety of biotic and abiotic conditions to survive. The cellular response to the environment involves complex regulatory networks that respond to stimuli such as light, pH, availability of carbon and nitrogen sources, reactive oxygen species, thermal stress, and interspecies-crosstalk (Brakhage, 2012; Reen et al., 2015).

#### OSMAC (One Strain Many Compounds) Approach

This strategy is derived from the observation that changes in the metabolic output of microorganisms can be achieved by alternating the medium composition and other cultivation parameters. It is well known that glucose, ammonium, or


ks, ketosynthase; at, acyltransferase; dh, dehydratase; mt, methyltransferase; er, enoyl reductase; kr, ketoreductase; acp, acyl carrier protein; te/red, thioester reductase. <sup>∗</sup>Point mutations present in pks genes of industrial P. chrysogenum strains subject to CSI program. Modified from Salo et al. (2015), Samol et al. (2016), and Guzmán-Chávez et al. (2018).

phosphate at high concentrations act as repressors of secondary metabolism, whereas iron starvation and nitrogen limitation can stimulate secondary metabolite production. The latter is for instance exploited for the production of terrain by A. terreus (Bode et al., 2002; Brakhage and Schroeckh, 2011; Gressler et al., 2015). This strategy can readily be implemented using highthroughput methods, where an array of culture conditions can be screened for new metabolite profiles (Spraker and Keller, 2014). In combination with bioinformatics tools, this strategy can be a powerful tool to investigate the production of new molecules, as exemplified by the discovery of aspoquinolones A–D in A. nidulans (Scherlach and Hertweck, 2006). However, despite the fact that the OSMAC approach has led to the discovery of increased numbers of new molecules with antimicrobial activity, some chemical and physical conditions are still missing under the laboratory tested conditions as the activation often concerns a limited number of BGCs (Chiang et al., 2009).

#### Interspecies-Crosstalk

The production of secondary metabolites is a natural strategy that microorganisms have developed to cope with specific environmental conditions and challenges. They serve as intermediary agents to establish a symbiotic association between species or as a weapon against other organism to compete for nutrients and space. These conditions, that are not present in axenic cultures, boost the production of molecules that are constitutively present and/or that are cryptic and normally are not synthetized due to silencing of the respective BGCs (Demain and Fang, 2000; Marmann et al., 2014). The strategy in which different organisms are cultivated together is called "co-culture," which has been successful in several cases yielding new metabolites. A. fumigatus produces fumiformamide when co-cultivated with Streptomyces peucetius, while co-cultivation of this fungi with S. rapamycinicus results in the production of fumicyclines A and B, two novel PKs with antibacterial activity, are examples of the use of this strategy (Netzker et al., 2015; Adnani et al., 2017). Interestingly, the association of two marine organisms, Emericella sp. and Salinispora arenicola, results in the biosynthesis of emericellamides A and B which are equipped with antibacterial activity (Oh et al., 2007). Also, the interactions between fungi and insects result in the production of volatile secondary metabolites (Rohlfs and Churchill, 2011).

#### Engineering of NRPS and PKS

Nonribosomal peptide synthetase and PKS are highly structured and multi-facetted enzymes, containing a tremendous potential for the exploitation of their product scaffold structure for the generation of novel, bioactive compounds. However, due to the complexity of all interactions within these mega enzymes, the elucidation and implementation of engineering strategies is an extremely challenging task. Several strategies have been developed and applied with different degrees of success, though the overall approaches can be grouped as module, domain, sub-domain or site directed, respectively. Owing to their large size, utilization of a random mutagenesis approach proved to be difficult, but other more directed strategies are met with a great success. Nevertheless, all of these strategies have their inherent difficulties, advantages and disadvantages in respect to the complexity and success rate of NRPS/PKS engineering efforts.

#### Subunit, Module, and Domain Swapping

Extensive efforts targeting the active site of A-domains has been a major focus in NRPS engineering. Multiple studies confirmed that the substrate specificity of a NRPS A-domain can be successfully altered, however, at the cost of substantially lowered catalytic velocity (Thirlway et al., 2012; Zhang et al., 2013). Similar successes and limitations were observed when domains were swapped or replaced by synthetic versions (Beer et al., 2014). The most challenging way of obtaining novel NRPS, however, is the swapping or combining of entire modules (Kries, 2016). Domain swapping overall created not only functional parts or domains, but also complete NRPS though with limited success (Beer et al., 2014).

Due to the strict arrangement of NRPS in domains and modules, the possibility of exchanging a unit appears to be the most straight forward approach for altering its intrinsic properties. A series of studies targeting the enzymes linked to the production of daptomycin (Nguyen et al., 2006; Baltz, 2014) elucidated the possibilities and borders of a combinatorial swapping strategy in context of novel compound production. The daptomycin biosynthetic cluster comprises three NRPS containing a total of 13 modules for the incorporation of an equal number of substrates. Different levels of domain and module swap approaches were followed, starting with the exchange of modules 8 and 11 (C-A-T), representing an internal module exchange. The resulting NRPS exhibited the production of novel daptomycin compounds with an inverted amino acid composition at the predicted sites at a near native rate (Nguyen et al., 2006). A similar combinatorial approach has been chosen for altering the PK stereochemistry. The exchange of a R domain with a TE domain in a NR-PKS from A. niger produced two alternative NR-PKS that harbor carboxylic acids instead of the aldehydes present in the original products (Yeh et al., 2013; Weissman, 2016). In Aspergillus, this rational domain swap has also been used to diversify the native substrates that NR-PKS takes as starter unit to produce new products. This involved exchanging the starter unit ACP transacylase domain in the PKS (Liu et al., 2014). Likewise, an analogous approach was used to produce new hybrids (PK–NRPs) in A. nidulans via module swapping of the two PKS–NRPS natural hybrids involved in the syn2 and cytochalasin E pathways from Magnaporthe oryzae and A. clavatus respectively (Nielsen et al., 2016). Despite the successful use of this strategy in some filamentous fungi, the engineering of NRPS and PKS in P. chrysogenum remains unexplored.

#### Genetic Interference

Another mechanism to stimulate the expression of silent BGCs in P. chrysogenum is by genetic interference, for instance by direct manipulation of the regulatory network related to BGCs expression. The regulation of BGCs is effected at many levels, through specific (or local) and global regulators up to epigenetic regulation involving the modification of the chromatin landscape (Lim et al., 2012; Spraker and Keller, 2014).

#### Global and Specific Regulators

fmicb-09-02768 November 13, 2018 Time: 17:47 # 13

#### **Global regulator-based regulation**

Pleiotropic transcriptional regulators or global regulators are proteins that respond to environmental signals such as pH, temperature, and N- and C-sources. They provide the link between the production of secondary metabolites and external cues. In fungi, these proteins control the regulation of BGCs that do not contain other regulatory factors. Up to 40% of the known clusters do not encode a local and specific regulator (or obvious regulatory genes). Additionally, global regulators also act on genes that do not belong to secondary metabolism (Brakhage, 2012; Rutledge and Challis, 2015; Fischer et al., 2016). Global regulators that have been reported as key players in the biosynthesis of secondary metabolites are featured below.

Velvet complex. This heterotrimeric complex is a conserved regulator present in most of the fungi, except yeast. It consists of at least three proteins: VeA, VelB, and LaeA. Likewise, this complex provides a link between sexual development and secondary metabolism through light regulation (Yin and Keller, 2011; Deepika et al., 2016), since light has an inhibitory effect on VeA expression. The formation of the velvet complex takes place in the nucleus, where the complex VeA–VelB via the α-importin KapA meets the methyltransferase LaeA. It has been hypothesized that the velvet complex acts as a transcriptional factor as it contains a DNA binding fold that resembles the corresponding region of the NF-κB transcription factor of mammals (Sarikaya-Bayram et al., 2015). The role of the velvet complex in secondary metabolism mostly follows from the control that the LaeA protein executes on several BGCs in filamentous fungi. LaeA (loss of aflR expression-A) was identified in 2004 as a global regulator in Aspergillus. Deletion of this gene results in the repression of many BGC, such as the one responsible for the production of penicillin, lovastatin, and sterigmatocystin. Overexpression of LaeA causes an opposite phenotype. Interestingly, LaeA is negatively regulated by AflR (Zn2Cy<sup>6</sup> transcriptional factor) in a feed loop mechanism (Bok and Keller, 2004). It has been hypothesized that LaeA acts at different levels, i.e., as a methyltransferase, epigenetically and as a direct member of the velvex complex. Structurally, LaeA has a S-adenosyl methionine (SAM)-binding site with a novel S-methylmethionine auto-methylation activity, although this activity does not seem to be essential for its function. LaeA is not a DNA-binding protein, but it does affect chromatin modifications. In an A. nidulans 1laeA strain, high levels of the heterochromatin protein 1 (HepA) are detected and an increase in trimethylation of the H3K9 in the sterigmatocystin cluster. When LaeA is present, the levels of HepA, ClrD (H3K9 methyltransferase) and H3K9me3 decrease while the sterigmatocystin levels are raised. The heterochromatic marks stay until the sterigmatocystin cluster is activated, and apparently LaeA influences the offset of these marks in this particular cluster (Reyes-Dominguez et al., 2010; Brakhage, 2012; Jain and Keller, 2013; Sarikaya-Bayram et al., 2015; Bok and Keller, 2016). Orthologs of LaeA have been discovered in many other filamentous fungi as Penicillium, Fusarium, Trichoderma, Monascus spp. and LaeA exhibits positive and negative effects on the synthesis of NPs. For instance, LaeA1 of F. fujikuroi positively regulates the production of fusarin C, fumonisins and gibberellins, and represses bikaverin biosynthesis. In P. chrysogenum, LaeA controls the biosynthesis of penicillin, pigmentation and sporulation (Keller et al., 2005; Kosalková et al., 2009; Jain and Keller, 2013). In Trichoderma reesei, Lae1 positively modulates the expression of cellulases, xylanases, β-glucosidases. Interestingly the stimulation of these genes was not directly influenced by the methylation of H3K4 or H3K9 (Wiemann et al., 2010; Yin and Keller, 2011; Lim et al., 2012; Seiboth et al., 2012; Jain and Keller, 2013).

LaeA is not the only member of the velvet complex that has influence on the regulation of secondary metabolite production. VeA of A. parasiticus is necessary for the expression of two transcriptional factors of the aflatoxin cluster (AflR and AflJ), which regulate the pathway. In A. fumigatus, veA regulates 12e BGCs (Dhingra et al., 2013). This study also revealed that veA modulates the biosynthesis of fumagillin via the regulation of FumR, a transcriptional factor of the fumagillin cluster, which in turn is also regulated by LaeA. Similarly, a transcriptome analysis in A. flavus revealed that 28 of 56 BGCs are dependent on veA, in particular the aflavarin cluster which is differentially expressed. Likewise, orthologs of veA are also present in other fungi such as in P. chrysogenum, F. oxysporum, Botrytis cinerea, F. verticillioides (Yin and Keller, 2011; Dhingra et al., 2013; Jain and Keller, 2013; Cary et al., 2015). Despite the clear interaction between veA and LaeA in the velvet complex and its influence on secondary metabolism, it is thought that veA may be acting as molecular scaffold of the velvet complex, since it interacts also with three other methyl transferases [LaeA-like methyltransferase F (LlmF), velvet interacting protein C (VipC), and VipC associated protein B (VapB)]. This suggests that veA functions in a supercomplex or in dynamic network control. Taken together, modulation of the velvet complex is a useful tool to activate BGCs (Sarikaya-Bayram et al., 2015) but results are difficult to predict.

bZIP transcription factors. Basic leucine zipper (bZIP) transcription factors are highly conserved in the eukaryotes. The dimeric bZIP transcriptional factors play an important role in the cellular responses to the environment. Regarding their structure, they contain a conserved leucine zipper domain and a basic region, which controls the dimerization of the protein and establishes sequence-specific DNA-binding, respectively. Once dimeric, bZIPs target palindromic DNA sequences by two mechanisms: redox and phosphorylation (Amoutzias et al., 2006; Knox and Keller, 2015). In fungi, bZIP proteins have been implicated in multiple metabolic processes, such as in the regulation of development, morphology and in stress responses. Several orthologs of the Yap family bZIPs, which were first described in yeast, have been characterized in Aspergillus spp. (AtfA, NapA, Afyap1, Aoyap1, and Apyap1) and these regulators have recently been associated with the production of secondary metabolites in filamentous fungi. In A. nidulans, overexpression of RsmA (restorer of the secondary metabolism A, Yap-like bZIP) has a compensatory effect on secondary metabolism in a strain in

which LaeA and veA are missing. However, these transcription factors also display negative regulation. For instance, an increase in the biosynthesis of aflatoxin and chratoxin has been observed when yap1 is deleted in A. parasiticus and A. ochraceus (Yin et al., 2013; Knox and Keller, 2015; Wang X. et al., 2015). MeaB is another bZIP transcriptional factor which was discovered in A. nidulans. Its function is associated in nitrogen regulation and has a negative effect on the biosynthesis of aflatoxin in A. flavus and bikaverin production in F. fujikuroi (Wagner et al., 2010; Amaike et al., 2013).

Other global regulators. AreA is a highly conserved transcriptional factor in fungi that belongs to the GATA family and it is characterized by Cys2His<sup>2</sup> zinc finger DNA binding domains. Likewise, it is involved in the repression of nitrogen metabolism when ammonium or glutamine are present. Recently, this transcription factor and its orthologs have been shown to influence secondary metabolism. For instance, areA deletion strains of F. verticillioides are not able to produce fumonisins on mature maize kernels. In Acremonium chrysogenum, the deletion of areA resulted in the reduction of cephalosporin because of a reduced expression of the enzymes involved in cephalosporin biosynthesis. Additionally, AreA is a positive regulator of the production of gibberellins, trichothecene deoxynivalenol (DON), fusarielin H, beauvericin and zearalenone (Li et al., 2013; Tudzynski, 2014; Knox and Keller, 2015 and Keller, 2015). The carbon catalytic repressor CreA also influences secondary metabolism. CreA is a Cys2His<sup>2</sup> zinc finger transcription factor that is involved in the repression of genes associated with the use of carbon sources other than glucose (Knox and Keller, 2015). This transcription factor acts by direct competition with activator proteins for specific binding sites (5<sup>0</sup> -SYGGRG-3<sup>0</sup> ) and by direct interaction with activators (Janus et al., 2008). In P. chrysogenum CreA represses penicillin biosynthesis and causes a reduced expression of the pcbAB gene that encodes a NRPS involved in this pathway. Mutations in the putative CreA binding site in the pcbAB promoter result in enhanced enzyme expression when cells are grown in the presence of glucose (Cepeda-García et al., 2014). In contrast, mutations in the CreA binding sites of the ipnA promoter (pcbC in other species) of A. nidulans revealed that in this organism repression of penicillin biosynthesis by glucose is independent of CreA (Knox and Keller, 2015). CreA has been implicated in the variable metabolite profiles when fungi are grown in the presence of different carbon sources (Yu and Keller, 2005). Recently, the xylanase promoter binding protein (Xpp1) of Trichoderma reesei was used as a reporter to fulfill a dual role in the regulation of primary and secondary metabolism. Xpp1 is an activator of primary metabolism, while its deletion boosts the production of secondary metabolites, including sorbicillinoids (Derntl et al., 2017b). Another Cys2His<sup>2</sup> zinc finger transcription factor conserved in fungi is PacC, which is involved in pH dependent regulation. Deletion of the ortholog of this gene (BbpacC) in Beauveria bassiana resulted in a loss of dipicolinic acid (insecticide compound) and oxalic acid production, compounds that reduce the pH of the medium. However, also production of a yellow pigment was noted. When A. nidulans is grown

at alkaline pH, PacC modulates the expression of the acvA (pcbAB) and ipnA of the penicillin BGC, while it acts negatively on the expression of the sterigmatocystin BGC (Deepika et al., 2016; Luo et al., 2017). In filamentous fungi, another global regulatory element is the CCAAT-binding complex (CBC). This complex consists of three proteins (HapB, HapC, and HapE) that respond to redox stimuli and an additional unit HapX, a bZIP protein that interacts with the complex for modulating the iron levels. In A. nidulans this complex binds to CAATT motifs, which are present in the penicillin BGC stimulating the expression of the ipnA and aatA (penDE) genes (Bayram and Braus, 2012; Brakhage, 2012). Whereas in F. verticillioides the ortholog core of this complex (FvHAP2, FvHAP3, and FvHAP5) is deleted, cells show an altered hyphal morphology, reduction of growth, reduced pathogenesis and a deregulation of secondary metabolism (Ridenour and Bluhm, 2014).

#### **Specific regulator-based regulation**

In addition to the global regulators, the expression of BGCs can be also modulated by specific regulatory elements, which most of the times are encoded by genes that are part of the same cluster that they regulate. In some cases, such regulators also influence the expression of other BGCs. It is estimated that about 60% of the fungal BGCs contain a gene encoding a potential regulator amidst the gene cluster. With PKS containing BGCs mostly containing a regulator that belongs to the Zn2Cys<sup>6</sup> binuclear cluster domain family (around 90%). With NRPS containing BGCs, the putative transcription factors are more diverse. The Zn2Cys<sup>6</sup> family of transcription factors contain a DNA binding domain (DBD) that has two zinc atoms coordinated by six cysteines. There are three sub regions: a linker, a zinc finger and a dimerization domain. Additional to a DBD, these proteins contain two further functional domains, the acidic region and the regulatory domain. These transcription factors can act as monomers, hetero- and homodimers. They recognize single or multiple trinucleotide sequences, commonly CCG triples, in a symmetric or asymmetrical format. The affinity of the DBD for a given DNA stretch is also determined by the nucleotides surrounding this triplet. The transcriptional activity of these proteins is regulated by phosphorylation, exposing the activation and DNA binding domains for DNA binding (MacPherson et al., 2006; Brakhage, 2012). Some of these regulators have been shown to control the expression of BGCs. For instance, in F. verticillioides the disruption of FUM21 gene, that encodes a Zn2Cys<sup>6</sup> protein, reduces fumonisin production as a result of a downregulation of the BGC (Brown et al., 2007). Interestingly, fumonisin production is also regulated by another Zn2Cys<sup>6</sup> protein that is encoded by a gene located outside of the fumonisin cluster (Flaherty and Woloshuk, 2004). Mlcr is another example of a positive regulator that controls compactin production in P. citrum (Abe et al., 2002). AflR is a Zn2Cys<sup>6</sup> protein that regulates the biosynthesis of aflatoxin/sterigmatocystin through binding to a palindromic sequence (5<sup>0</sup> -TCG(N5)GCA) that is found in most of the promoters of this BGC, albeit a second binding sequence has been reported that is associated with the autoregulation mechanism of the expression of AflR. The disruption of AlfR abolishes

the production of aflatoxin/sterigmatocystin. Likewise, some BGCs encode multiple regulatory proteins. Next to the aflR gene in the aflatoxin cluster resides the aflS (formerly aflJ) gene. The corresponding transcription factor binds to AflR to enhance the transcription of early and mid-biosynthetic genes in the aflatoxin pathway (Georgianna and Payne, 2009; Yin and Keller, 2011). In P. chrysogenum and Trichoderma reesei, the sorbicillin BGC is regulated by two transcriptional factors through a coordinated action (Derntl et al., 2016, 2017a; Guzmán-Chávez et al., 2017). Also, regulation of BGCs via crosstalk has been observed in filamentous fungi. For instance, the alcohol dehydrogenase promoter has been used to induce the expression of putative pathway-specific regulatory gene (scpR) in A. nidulans, which controls the expression of two pathway associated NRPS genes (inpA and inpB). Surprisingly, two PKS genes (afoE and afoG) and one transcriptional activator (afoA) belonging to the asperfuranone BGC are also upregulated by ScpR, allowing the production of asperfuranone (Bergmann et al., 2010). For some regulators, no clear phenotype is observed. For instance deletion of the chyR gene of the chrysogine BGC in P. chrysogenum, has no effect on the expression of the corresponding BGC (Viggiano et al., 2017).

#### **Manipulation of regulatory elements as strategies for the activation of BGCs**

Gene deletion. It is a classical strategy that consists of the abolishment of the expression of a certain gene by its elimination whereupon the impact on the metabolite profile is examined by HPLC or LC-MS. A major limitation of this approach is that it can only be used in BGCs that are not totally silenced under laboratory conditions. Using this strategy, it was possible to elucidate the highly branched biosynthetic pathway for the synthesis of roquefortine as well as the biosynthetic pathways of sorbicillinoids and chrysogine in P. chrysogenum (García-Estrada et al., 2011; Ali et al., 2013; Ries et al., 2013; Deepika et al., 2016; Guzmán-Chávez et al., 2017; Viggiano et al., 2017). Likewise, this approach can be used to remove transcriptional repressor genes, as in the case of TetR-like pathway-specific repressor proteins, whose deletion induced the production of gaburedins in Streptomyces venezuelae (Rutledge and Challis, 2015). Global regulators, such as LaeA have also been targeted using this strategy (Chiang et al., 2009).

Promoter replacement. Another method concerns the replacement of the endogenous promoter of the gene(s) in a BGC by a strong constitutive or inducible promoter. For instance in A. nidulans replacement of the native promotor of the scpR gene (secondary metabolism cross-pathway regulator) for the inducible promoter of alcohol dehydrogenase AlcA induced the expression of a silent cluster that contained two NRPS genes (inpA and inpB) and scpR itself. Additionally, it also led to the expression of the asperfuranone BGC, which is normally silent (Bergmann et al., 2010; Yin and Keller, 2011; Lim et al., 2012). Recently in P. chrysogenum a promising promoter toolbox for bioengineering purposes was developed. This included the analysis of four constitutive promoters from P. chrysogenum and six from A. niger, which were evaluated using a reporter system and assorted by promoter strength (Polli et al., 2016).

Overexpression of a specific or global regulator. This approach is one of the most used strategies to turn on cryptic BGCs, since a change in expression level of a regulator may boost the expression of a whole cluster. Usually, this strategy is applied in combination with the promoter replacement approach. Using this strategy, i.e., overexpression of the transcription activator ApdR under control of the alcohol dehydrogenase promoter alcAp, it has been possible to induce the expression of a hybrid PKS-NRPS BGC in A. nidulans. This resulted in the production of aspyridones A and B (Bergmann et al., 2007). Similarly when the global regulator FfSge, which is associated with vegetative growth of F. fujikuroi, is overexpressed, some BGCs are forced to express under these unfavorable conditions (low nitrogen concentrations) leading to the identification of the corresponding products (Michielse et al., 2015).

#### Chromatin-Mediated Regulation

In fungal cells, chromosomal DNA is wrapped in a complex of DNA, histone proteins and RNA called chromatin. This chromatin structure consists of a basic unit called nucleosome, which consists of superhelical DNA (147 base pairs) that binds an octamer of four different core histone proteins (two each of H2A, H2B, H3, and H4) in 1.75 turns. It has been shown that modifications of the chromatin structure (boosts or alters) changes gene expression, amongst other genes involved in the biosynthesis of secondary metabolites. Structurally, chromatin represents an obstacle that complicates access of DNA-binding factors to their corresponding binding regions. According to the compaction level, chromatin can be in a dense (heterochromatin) or relaxed (euchromatin) state. These compaction levels are regulated by post-translational modification of the histone proteins by acetylation, methylation, ubiquitination, ethylation, propylation, butylation, and phosphorylation events. Regions that display low transcriptional activity have been associated with the heterochromatic conformation. In contrast, the euchromatic conformation is present in regions with abundant coding sequences and is usually highly active during transcription. Such regions are also linked with hyper-acetylated nucleosomal histones. Likewise, it has been reported that methylation of H3K9, H3K27, and H4K20 are typical markers of the heterochromatin, while in euchromatin methylation occurs at H3K4 (Brosch et al., 2008; Strauss and Reyes-Dominguez, 2011; Gacek and Strauss, 2012; Spraker and Keller, 2014; Rutledge and Challis, 2015).

#### **Histone methylation, acetylation, and sumoylation**

As mentioned above, LaeA influences secondary metabolite production through chromatin modification. The methylation state of H3K9 has been correlated with the heterochromatin protein A (HepA), since this protein needs the di- and trimethylation of H3K9 for binding to chromatin and to form heterochromatin. Deletion of LaeA allows the unobstructed binding of HepA to the AlfR promoter, thereby affecting the expression of the sterigmatocystin pathway. The deletion of the methyltransferase encoding clrD and ezhB genes in Epichloe festucae, that act on H3K9 and H3K27, respectively (in axenic culture), results in the activation of the ergot alkaloids and lolitrem BGCs. These compounds are necessary to establish a

symbiotic association with the plant Lolium perenne. Compass (complex of proteins associated with Set1) which methylates H3K4 in yeast, also impacts secondary metabolism in filamentous fungi. The deletion of one of its components (cclA) in A. nidulans allowed the activation of a cryptic BGC and the production of emodin (Palmer and Keller, 2010; Gacek and Strauss, 2012; Chujo and Scott, 2014; Netzker et al., 2015; Deepika et al., 2016). Likewise, in F. fujikuroi and F. graminearum, the deletion of cclA caused the overproduction of secondary metabolites derived from BGCs close to the telomeres, but this seems to relate to a H3K4 methylation independent mechanism (Studt et al., 2017). Other types of histone modification may alter the chromatin landscape, such as acetylation which is a reversible process governed by two antagonist enzymes: histone acetyltransferases (HATs) and deacetylases (HDACs). Active transcription is usually associated with histone acetylation, although recently the deacetylation of histones has been shown to cause activation of genes (Brosch et al., 2008). Usually, histones are acetylated by several complexes with acyltransferase activity, such as Saga/Ada and NuA4. In A. nidulans a chromatin immunoprecipitation (ChIP) analysis revealed that GcnE and AdaB, the catalytic subunits of the complex Saga/Ada, are needed for acetylation of histone H3 (Deepika et al., 2016). Indeed, the interaction between A. nidulans and Streptomyces rapamycinicus can be linked to a GcnE dependent increase in the acetylation of H3K14 that shields the promoters of the orsellinic acid BGC. The Saga/Ada complex is a key player in the induction of the penicillin, terrequinone and sterigmatocystin BGCs (Nutzmann et al., 2011; Brakhage, 2012). In contrast, deletion of hdaA (encoding a HDAC) in A. nidulans resulted in major changes in the metabolite profile (Rutledge and Challis, 2015). HdaA is a class 2 histone deacetylase involved in the regulation of BGCs that are located near the telomeres, such as the penicillin and sterigmatocystin clusters in A. nidulans. Indeed, deletion of the hdaA gene results in the increased and early gene expression of these two BGCs, and the production of the corresponding secondary metabolites. In A. fumigatus, the hdaA gene is involved in growth and production of secondary metabolites, and the deletion of this gene increases the production of many secondary metabolites while it causes a reduction of gliotoxin production. In contrast, HdaA overexpression shows the opposite effect (Shwab et al., 2007; Lee et al., 2009). In P. chrysogenum was demonstrated that HdaA (histone deacetylase) mediates the transcriptional crosstalk among sorbicillinoids biosynthesis and other BGCs, since a new compound as detected only under conditions of sorbicillinoids production (Guzmán-Chávez et al., 2018).

Histone deacetylases are ubiquitously distributed in filamentous fungi, and therefore HDAC inhibitors can be used to improve the synthesis of NPs by epigenome manipulation (Shwab et al., 2007; Lee et al., 2009). For instance, the metabolite profile of Cladosporium cladosporioides and A. niger underwent a significant change when these strains were exposed to suberoylanilide hydroxamic acid (SAHA), a HDAC inhibitor, allowing the detection of two new compounds, cladochrome and nygerone A, respectively (Rutledge and Challis, 2015). An exploratory analysis performed in 12 fungi treated with different types of DNA methyltransferase and histone deacetylase inhibitors, revealed the production of new secondary metabolites but also the elevated amounts of known compounds (Williams et al., 2008). In this respect, the chromatin state can directly influence the binding of transcription factors, and thereby modulate expression (Palmer and Keller, 2010; Macheleidt et al., 2016). It has been hypothesized that histone sumoylation may modulate secondary metabolite production. This process is mediated by a small protein termed SUMO (small ubiquitin-like modifier) that shares structural similarity to the ubiquitin protein. In A. nidulans, deletion of the sumO gene enhanced the production of asperthecin, whereas synthesis of austinol, dehydroaustinol, and sterigmatocystin was reduced. Although the molecular mechanism still needs to be elucidated, it is thought that sumoylation acts at several levels, such as on epigenetic regulators (COMPASS, Clr4, SAGA/ADA and HDACs) or at the level of transcriptional regulators (Brakhage and Schroeckh, 2011; Spraker and Keller, 2014; Wu and Yu, 2015).

#### **Modification of the chromatin landscape to activate BGCs**

Many fungal BGCs are located in distal regions of the chromosomes. In these heterochromatin rich regions, transcription of the BGCs can be activated by epigenetic regulation. Therefore, the encoding genes of proteins that influence histone modification are prime targets, although these modifications can also be achieved by chemical treatment (Williams et al., 2008; Brakhage, 2012). A recent study in P. chrysogenum showed that the expression of a set of PKS and NRPS encoding genes is induced when an ortholog of a class 2 histone deacetylase (HdaA) is deleted. This allowed for the overproduction of sorbicillinoids, the reduction of chrysogine related metabolites and the detection of a new compound whose origin still unknown (Guzmán-Chávez et al., 2018).

#### Other Targets for Regulation

Secondary metabolites produced by fungi can be toxic to the producer organisms, and often fungi are equipped with detoxification mechanisms. One of these mechanisms is toxin excretion by transporters, which are membrane proteins whose genes often localize to the BGCs. Transporters may belong to different protein families but the major facilitator superfamily (MFS) and ABC superfamily are most commonly encoded by BGCs (Keller, 2015). Since biosynthesis of secondary metabolites may take place in different cell compartments, also intracellular transport may be evident (Kistler and Broz, 2015). Despite their assumed biological importance, the deletion of transporter genes from the BGCs often does not impact secondary metabolite production. For instance, deletion of the A. parasiticus aflT gene, that encodes a MFS transporter, does not result in reduced aflatoxin excretion, despite the fact that aflT belongs to the aflatoxin BGC and its expression is regulated by a specific transcription factor, AflR, of the pathway. Probably, this protein is redundant, and other transporters may participate in excretion, detoxification or self-defense. In A. fumigatus, GliA facilitates the excretion of gliotoxin. Similarly, the tri12 gene contained in the trichothecene BGC encodes for a membrane protein required for the biosynthesis of trichothecene and virulence of F. graminearum on wheat crops (Chang et al., 2004; Menke

et al., 2012; Wang D.N. et al., 2014; Keller, 2015). Often, however, the deletion of the transporter gene in BGCs has no effect on production. Possibly, these metabolites are also recognized by other promiscuous transporters, or transporters that are not part of the BGC (Keller, 2015). For example, ZRA1 of Gibberella zae, whose gene is not localized to the zearalenona BGC, impacts zearalenone production. However, the expression of the zra1 gene is regulated by the transcriptional factor ZEB2, whose gene localizes to the corresponding BGC (Lee et al., 2011). Also, the penicillin BGC of P. chrysogenum lacks a transporter gene whereas export of penicillin occurs against the concentration gradient, probably through the activity of multiple transporter proteins (van den Berg et al., 2008; Kistler and Broz, 2015). Furthermore, compartmentalization of the biosynthesis of penicillin is well documented requiring transport of penicillin precursors across the membrane of intracellular organelles (Weber et al., 2012a,b).

#### Other Genetic Engineering Strategies for the Activation of BGCs

Several approaches have been used to activate the expression of cryptic BGCs in a targeted manner. Usually, this is achieved by manipulation of pathway-specific regulatory genes, or by replacing endogenous promoters for inducible systems or strong promoters (Rutledge and Challis, 2015). The various approaches are summarized in **Figure 7**.

#### **Manipulating biosynthetic pathways by genome editing**

Due to the increasing number of sequenced filamentous fungi, it is necessary to make use of efficient genome editing tools to explore new potential sources of secondary metabolites. For many years, the unique strategy available for the genome edition of P. chrysogenum was based on the use of ku70/80 disrupted strains to improve the homologous recombination instead of the Non-Homologous End Joining (NHEJ) pathway (Weber et al., 2012a). This strategy allowed for the generation P. chrysogenum strains with high copy numbers of the penicillin cluster, the identification of a biosynthetic branch of the roquefortine cluster and the reactivation of the sorbicillinoid gene cluster (Nijland et al., 2010; García-Estrada et al., 2011; Ali et al., 2013; Ries et al., 2013; Salo et al., 2016; Guzmán-Chávez et al., 2017). Recently, a CRISPR/Cas9 based system was developed for genome modifications in P. chrysogenum (Pohl et al., 2016, 2018). This study demonstrated that the deletion of full gene clusters is feasible with minimal cloning efforts, which opens the possibilities to engineer new synthetic pathways and the re-factoring P. chrysogenum as platform organism.

#### **Ribosome engineering**

This approach has been applied for activating silent or poorly expressed BGCs (Ochi and Hosaka, 2013). Basically, this concept is derived from the activation of the actinorhodin BGC in S. lividans due to a point mutation in the rpsL gene, which encodes for the ribosomal S12 protein (Shima et al., 1996). Another successful examples in the BGCs activation have been reported by modifying the transcription and translation pathways via targeting different ribosomal proteins, RNA polymerases (RNAP) and translation factors (Ochi and Hosaka, 2013). In P. purpurogenum G59, a marine derived strain, the insertion of gentamicin resistance after treatment with high concentrations of this antibiotic, altered ribosomal functions of this fungus which allowed for the activation of dormant secondary metabolite gene clusters (Chai et al., 2012).

#### **Heterologous expression and refactoring**

Due to the broad range of molecular tools available to express heterologous pathways in yeast, several attempts have been undertaken to express NRPS and PKS genes with the remainder of the pathway in yeast (Rutledge and Challis, 2015). A recent study demonstrated that the baker's yeast Saccharomyces cerevisiae can be used as a platform to produce and secrete penicillin when the biosynthetic genes are expressed in this organism (Awan et al., 2016). Although the first step was performed when the acetyl-CoA:isopenicillin N acyltransferase (IAT), which catalyzes the last step in the penicillin biosynthesis was amplified from the P. chrysogenum penDE gene and expressed in Hansenula polymorpha (Lutz et al., 2005). However, most of the times the main obstacle is the large size (>40 kbp) of the DNA fragment that needs to be cloned, the effective activation/maturation of the expressed enzymes, and the toxicity of the produced compounds (Rutledge and Challis, 2015). Alternatively, fungi may be used as platform organism, as it was for instance demonstrated with the reconstruction of the citrinin gene cluster of Monascus purpurea in A. oryzae (Spraker and Keller, 2014). Likewise, the in vivo assembly of genetic elements has been successfully applied in P. chrysogenum through the overlapping of bi-partite fragments that reconstituted a functional amdS gene (marker), which eventually is integrated in the genome of this fungus proving the uncharacterized potential of P. chrysogenum as heterologous host (Polli et al., 2016). The potential of this approach follows a recent study employing A. nidulans as a host for the plasmid based expression of a diverse range of BGCs from other filamentous fungi (Clevenger et al., 2017).

The introduction of revolutionary new genetic tools, such as CRISPR/Cas9 offers more effective solutions to express specific BGCs. Such methods can contribute to product identification but also to the production of unique compounds by introduction of specific tailoring enzymes. These are the main strategies that are used for the activation of silent BGCs or for the modification/redirection of known biosynthetic pathways in order to increase NP diversification (Smanski et al., 2016). Specifically, this involves the expression of pathways from a plasmid in a suitable production host and a screen for product formation.

#### CONCLUDING REMARKS

For many years P. chrysogenum has been used as one of the main industrial strains to produce penicillins (β-lactams). Its genome sequence revealed an unexplored potential of P. chrysogenum as a source of NPs (van den Berg et al., 2008). Despite the development of bioinformatics tools for genome mining of BGCs to identify novel molecules (Blin et al., 2017), the experimental validation of product structure and identity is still

necessary. However, most of the secondary metabolite associated genes in P. chrysogenum are silent or poorly expressed. Given the urgent need for new molecules based on novel chemical scaffolds for the use in the medical and biotechnological fields (e.g., antibiotics, anti-cancer agents, antivirals, nutraceuticals, pigments, surfactants and many more), the use of organisms that have been genetically domesticated offers a promising target solution for NP discovery due to the availability of molecular tools for their genetic modification. Here, we have summarized the main approaches that have been applied for P. chrysogenum and other filamentous fungi to bioengineer secondary metabolite BGC pathways which have led to a greater understanding of the main obstacles to be overcome to use this fungus as a generic cell factory. We discussed the main characteristics of the building enzymes (PKS and NRPS) in filamentous fungi. Despite the apparent modular organization, the complexity of these mega enzymes and the inherent interactions between the various domains within their structures has not allowed a straight forward approach for the PKS/NRPS engineering (Thirlway et al., 2012; Zhang et al., 2013; Weissman, 2016). However, the combinatorial swapping strategy of structural elements such as recognition regions has increased the perspectives for designing de novo biosynthetic pathways. Further research needs to be focused around PKS/NRPS engineering in filamentous fungi to facilitate the rational design of biosynthetic enzymes to produce another generation of novel compounds. To mine the secondary metabolome of filamentous fungi, general methods such as manipulating cultivation conditions have been used that can also be implemented as a high-throughput strategy. Another avenue is interfering with the genetic regulatory systems, either through the manipulation of specific or global regulators. This strategy also revealed crosstalk between certain BGCs and an important role of chromatin remodeling in BGC expression. Because of its pleiotropic effect that leads to the activation or silencing of biosynthetic pathways, chromatin remodeling might be used as a more general strategy to explore the production of new metabolites in filamentous fungi (Guzmán-Chávez et al., 2018).

A further major advance is the development of genomeediting tools that allow for efficient genetic engineering of complex fungal cell factories (Jakoèiunas et al., 2016). In P. chrysogenum, improved methods for homologous recombination and CRISPR/Cas9 as genome editing tool now facilitates more advanced engineering of this fungus (Pohl et al., 2016, 2018). This is further stimulated by the development of a synthetic biology toolbox using promoters and terminators as building blocks and more complex regulatory devices to control the expression of genes. Importantly, the in vivo assembly of genetic elements in P. chrysogenum offers a promising tool to build entire pathways from scratch with reducing cloning efforts at minimal costs (Polli et al., 2016). In particular the low cost of DNA synthesis will allow rapid progress using such approaches as exemplified by a study using A. nidulans as a host where a diverse set of BGCs was expressed from an

extra-chromosomal vector, the AMA plasmid (Clevenger et al., 2017). A further challenge is the generation of a platform strain in which endogenous BGCs have been removed to allow for more optimal carbon and nitrogen flow toward the production of the compounds of interest. Herein, the CRISPR/Cas9 based methods are instrumental (Pohl et al., 2016, 2018). Such a Penicillium platform might be used as a heterologous host to express a vast arsenal of BGCs from others filamentous fungi and represents a good alternative to yeast as expression host, an organism that does not naturally produce NRP and PK. The use of a such industrial strains to rapidly achieve the high level production of a novel metabolite has proven to be successful for pravastatin production (McLean et al., 2015) but a further step is to make use of secondary metabolite deficient industrial strains.

Despite the progress in genetic engineering and bioinformatics tools to identify BGCs, the main bottleneck to identify potentially interesting compounds has not yet been solved. Bioinformatic tools perform poorly in the prediction of the structures formed, and therefore future discovery programs will mostly dependent on high throughput methods to express

#### REFERENCES


foreign pathways and then use advanced metabolomics to identify the novel products. Since such approaches depend on high throughput, further efforts are needed to implement high throughput cloning methods to P. chrysogenum which will enable further studies to harness the enormous untapped source for NPs hidden in fungal (meta-)genomes.

#### AUTHOR CONTRIBUTIONS

FG-C and RZ wrote the manuscript. AD supervised, conceived, and designed the manuscript. RB co-supervised the manuscript.

#### FUNDING

FG-C was supported by Consejo Nacional de Ciencia y Tecnología (CONACyT, Mexico) and Becas Complemento SEP (Mexico). RZ was supported by the BE-Basic Foundation, an international public–private partnership.





introduction of a new metabolic pathway. FEMS Yeast Res. 5, 1063–1067. doi: 10.1016/j.femsyr.2005.07.002


synthetase activation in Aspergillus fumigatus. Chembiochem 6, 679–685. doi: 10.1002/cbic.200400147




#### **Conflict of Interest Statement:** RB is an employee of DSM Biotechnology, Delft.

The remaining 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 © 2018 Guzmán-Chávez, Zwahlen, Bovenberg and Driessen. 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.

# Engineered Streptomyces lividans Strains for Optimal Identification and Expression of Cryptic Biosynthetic Gene Clusters

Qinying Peng, Guixi Gao, Jin Lü, Qingshan Long, Xuefei Chen, Fei Zhang, Min Xu, Kai Liu, Yemin Wang, Zixin Deng, Zhiyong Li\* and Meifeng Tao\*

State Key Laboratory of Microbial Metabolism, Joint International Research Laboratory of Metabolic and Developmental Sciences, School of Sciences and Biotechnology, Shanghai Jiao Tong University, Shanghai, China

#### Edited by:

Dipesh Dhakal, Sun Moon University, South Korea

#### Reviewed by:

Hiroyuki Morita, University of Toyama, Japan Sergey B. Zotchev, Universität Wien, Austria

\*Correspondence:

Zhiyong Li zyli@sjtu.edu.cn Meifeng Tao tao\_meifeng@sjtu.edu.cn

#### Specialty section:

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

> Received: 07 October 2018 Accepted: 26 November 2018 Published: 10 December 2018

#### Citation:

Peng Q, Gao G, Lü J, Long Q, Chen X, Zhang F, Xu M, Liu K, Wang Y, Deng Z, Li Z and Tao M (2018) Engineered Streptomyces lividans Strains for Optimal Identification and Expression of Cryptic Biosynthetic Gene Clusters. Front. Microbiol. 9:3042. doi: 10.3389/fmicb.2018.03042 Streptomyces lividans is a suitable host for the heterologous expression of biosynthetic gene clusters (BGCs) from actinomycetes to discover "cryptic" secondary metabolites. To improve the heterologous expression of BGCs, herein we optimized S. lividans strain SBT5 via the stepwise integration of three global regulatory genes and two codonoptimized multi-drug efflux pump genes and deletion of a negative regulatory gene, yielding four engineered strains. All optimization steps were observed to promote the heterologous production of polyketides, non-ribosomal peptides, and hybrid antibiotics. The production increments of these optimization steps were additional, so that the antibiotic yields were several times or even dozens of times higher than the parent strain SBT5 when the final optimized strain, S. lividans LJ1018, was used as the heterologous expression host. The heterologous production of these antibiotics in S. lividans LJ1018 and GX28 was also much higher than in the strains from which the BGCs were isolated. S. lividans LJ1018 and GX28 markedly promoted the heterologous production of secondary metabolites, without requiring manipulation of gene expression components such as promoters on individual gene clusters. Therefore, these strains are well-suited as heterologous expression hosts for secondary metabolic BGCs. In addition, we successfully conducted high-throughput library expression and functional screening (LEXAS) of one bacterial artificial chromosome library and two cosmid libraries of three Streptomyces genomes using S. lividans GX28 as the library-expression host. The LEXAS experiments identified clones carrying intact BGCs sufficient for the heterologous production of piericidin A1, murayaquinone, actinomycin D, and dehydrorabelomycin. Notably, due to lower antibiotic production, the piericidin A1 BGC had been overlooked in a previous LEXAS screening using S. lividans SBT5 as the expression host. These results demonstrate the feasibility and superiority of S. lividans GX28 as a host for high-throughput screening of genomic libraries to mine cryptic BGCs and bioactive compounds.

Keywords: optimal hosts, global regulatory genes, heterologous expression, biosynthetic gene clusters (BGCs), secondary metabolites, library expression and function-directed screening system (LEXAS)

# INTRODUCTION

fmicb-09-03042 December 7, 2018 Time: 17:16 # 2

Microbial secondary metabolites display tremendous diversity in chemical structure and bioactivity and play an important role in drug discovery and development (Cragg and Newman, 2013). In recent years, the exploitation of the potential of "cryptic" biosynthesis, in the form of biosynthetic gene clusters (BGCs), in microbial genomes has become the focus of natural product research (Scherlach and Hertweck, 2009). Heterologous expression of BGCs is now an important technology for genome mining, biosynthetic study, and metabolic engineering. In addition, genome mining and biosynthetic studies on slowgrowing or uncultured microorganisms can only be carried out through heterologous expression (Ongley et al., 2013). Many Streptomyces strains have the advantages of rapid growth and simple genetic manipulation. Moreover, as Streptomyces spp. and closely related actinomycetes are rich in secondary metabolite resources and therefore have the ability to provide precursors and cofactors required for efficient biosynthesis, engineered Streptomyces strains are highly suitable hosts for the heterologous expression of BGCs (Martinez et al., 2004; Wenzel and Müller, 2005). Excellent hosts must also be able to express all of the enzymes of the candidate biosynthetic pathway efficiently, including gene transcription, translation, and posttranslational modifications, in order to successfully produce the corresponding compounds (Baltz, 2010). Ideally, endogenous secondary metabolic pathways should also be deleted to make a clean metabolic background and avoid substrate competition between endogenous and heterologous pathways (Komatsu et al., 2010). Such optimized hosts include Streptomyces coelicolor M1152, M1154, Streptomyces avermitilis, and Streptomyces albus, which has a naturally minimized genome (Gomez-Escribano and Bibb, 2011; Komatsu et al., 2013; Kallifidas et al., 2018).

Streptomyces lividans, a species closely related to S. coelicolor, has additional advantages as an expression host since it does not restrict (cleave) exogenous methylated DNA, whereas most actinomycetes such as S. coelicolor and S. avermitilis cleave methylated plasmid DNA from most Escherichia coli strains (MacNeil, 1988). Furthermore, when used as a recipient for E. coli–Streptomyces intergeneric conjugation, S. lividans exhibits a high efficiency of conjugative transfer, and this advantage is particularly important for experiments that require high-throughput transfer of arrayed library clones for screening genomic or metagenomic libraries by the function of unknown compounds (Wang et al., 2000). Indeed, an S. lividans TK24-derived strain was chosen as the expression host for the expression and screening of a metagenomic BAC library (Martinez et al., 2004). On the other hand, the wild-type strains of S. lividans, such as strain 1326, have disadvantages as they contain endogenous BGCs for secondary metabolites, such as act (for actinorhodin), red (for streptorubin or undecylprodigiosin) (Liu et al., 2017), and cda [for calcium-dependent antibiotic (CDA)]. Of more concern, such wild-type strains do not produce corresponding antibiotics under most culture conditions, implying that at least these BGCs are silent in the wild-type host (Hu et al., 2002).

Streptomyces lividans TK24 is a spontaneous rpsL[K88E] mutant that increases the production of actinorhodin; rpsL[K88E] has been shown to induce global upregulation of secondary metabolite biosynthesis (Shima et al., 1996; Okamoto-Hosoya, 2003). Additionally, the global regulatory genes afsRScla from S. clavuligerus significantly promote the synthesis of actinorhodin, streptorubin, and CDA in S. lividans TK24 (Chen et al., 2012). On the basis of the above findings, we knocked out the act, red, and cda BGCs from S. lividans TK24 and inserted 1–2 copies of afsRScla, obtaining S. lividans SBT5 and SBT18 (Bai et al., 2014; Xu et al., 2016). Using these strains as expression hosts, Xu et al. (2016) optimized the previous functional genomic screening protocol (Martinez et al., 2004) and developed a library heterologous expression and function-directed screening system (LEXAS) for the screening of BGCs. LEXAS has facilitated the activation of cryptic BGCs and helped in mining compounds and new BGCs using the genomic libraries of Streptomyces spp. (Gao et al., 2017; Zheng et al., 2017; Chen et al., 2018).

In this study, in order to further improve the expression efficiency of heterologous BGCs and improve the screening efficiency of biologically active natural products by LEXAS technology, we optimized S. lividans SBT5 using a number of global positive and negative regulatory genes and genes encoding drug efflux pumps. We also demonstrated the superiority of the new strains in expression of heterologous BGCs and for LEXAS screening of cosmid and BAC libraries.

# MATERIALS AND METHODS

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

Streptomyces lividans SBT5 [S. lividans TK24 1act1redKL 1cdaPS3-SLI3600::afsRScla] (Bai et al., 2014) was used as the parent strain to construct optimized hosts for the heterologous expression of BGCs. Streptomyces griseoruber Sgr29 and Streptomyces galtieri Sag48 were isolated from Shennongjia (Eastern Hubei, China) forest soil (China Center for Type CultureCollection, CCTCC). Streptomyces parvulus 10 was isolated from the marine sponge Phyllospongia foliascens collected from Yongxing Island (South China Sea). S. lividans TK24 was used to construct the wblAsl knockout plasmid, and S. coelicolor M1154 (Gomez-Escribano and Bibb, 2011) was used as the cloning template for nusGsc. Mannitol soy flour agar (MS) was used for sporulation of Streptomyces spp. Liquid culture was performed in TSBY medium containing 3% tryptone soy broth medium, 0.5% yeast extract, and 10.3% sucrose (Kieser et al., 2000). Agar media YBP (Ou et al., 2009), R3 (Shima et al., 1996), GYM (Ochi, 1987), No18 and No24 (Farnet et al., 2008) were used for the fermentation of Streptomyces. Streptomyces cultures were grown and fermented at 30◦C. To determine growth curves for recombinant strains of S. lividans, strains were cultured in baffled flasks at 30◦C with TSBY liquid medium (30 mL/flask).

Escherichia coli strains, Staphylococcus aureus CICC 10201, and Bacillus mycoides were cultured in Luria-Bertani (LB) medium at 37◦C. E. coli XL1-Blue (Stratagene) was used as

the host for cosmid library construction, and E. coli DH10B (Invitrogen) was used for general cloning, plasmid maintenance, and as host for a BAC library. E. coli ET12567/pUB307 was used as a helper strain mediating tri-parental E. coli–Streptomyces intergeneric conjugation (MacNeil et al., 1992). S. aureus CICC 10201, B. mycoides, and Saccharomyces sake were used as indicator strains in the bioassay experiments.

pJTU2554 (Li et al., 2008) is a pSET152-derived, triplet COS site-bearing vector used to construct genomic cosmid libraries of S. galtieri Sag48 and S. parvulus 10. SuperCos 1 (Stratagene) was used to construct the genomic cosmid library of S. lividans. pHL921 (Xu et al., 2016) was the vector for the genomic BAC library of S. griseoruber Sgr29. pJTU2554-, pHL931-, and pHL921-derived clones carry the attP and int loci of the Streptomyces temperate phage 8C31, and therefore can integrate into Streptomyces genomes at the attB site (Combes et al., 2002). pMS82, which bears the integration site attP and int loci of the Streptomyces temperate phage 8BT1, was used as an integrative vector to carry genes of interest into the chromosome of S. lividans (Gregory et al., 2003). pUB307 is an RK2 derived, self-mobilizable plasmid that facilitates the intergeneric conjugation of oriTRK2-plasmids from E. coli to Streptomyces (Bennett et al., 1977). pIJ773 and pIJ778 were used as templates for PCR amplification of aac(3)IV and aadA resistance markers, respectively (Gust et al., 2003). pHL851 (Chen et al., 2012) was the source of afsRScla.

# Chemicals

The actinomycin D standard was purchased from Aladdin Bio-Chem Technology, Co. Standard compounds of murayaquinone, hybrubin A, dehydrorabelomycin, and piericidin A1 were prepared as described (Liu et al., 2012, 2018; Zhao et al., 2016; Gao et al., 2017).

# Construction of Integrative Plasmids Carrying nusGsc, Efflux Pump Genes, and afsRScla

A 1.4 kb fragment containing nusGsc was amplified by PCR using S. coelicolor M1154 genomic DNA as template and primers nusG-R (5<sup>0</sup> -CTAGTTCTTCTGGATCTGGTGCTTG-3<sup>0</sup> ) and nusG-F (50 -GTGACGGACGCCGTGGGCTCCA-3<sup>0</sup> ), and then cut with XbaI and ligated with pMS82 to yield pJTU6725. Two codonoptimized genes, mdfAco and lmrAco, were synthesized based on the protein sequences of MdfA of Enterobacteriaceae (AFH35853) and LmrA of Lactococcus lactis subsp. cremoris MG1363 (CAL98427.1), respectively, with the codon usage table of S. coelicolor, and following the Codon Adaptation Tool<sup>1</sup> . The mdfAco gene has an overall GC content of 67% and a GC content of 99.5% for the third codon position. The lmrAco gene has an overall GC content of 65% and a GC content of 99.5% for the third codon position. The optimized lmrAco-mdfAco DNA fragment was synthesized at HongXun Biotechnology, Co., Ltd., and ligated into pJTU6725 to construct pJTU6727. The paired genes afsRScla (SCLAV\_3382, SCLAV\_3383) and the upstream ermE<sup>∗</sup> promoter in pHL851 were cloned into pJTU6727 to construct pJTU6728. The plasmids pJTU6725, pJTU6727, and pJTU6728 retain the hygromycin resistance gene (hyg), origin of transfer (oriTRK2), and the attP and int loci of Streptomyces phage 8BT1 from the integrative vector pMS82.

# Construction of Genomic Cosmid Libraries of Streptomyces spp.

The genomic cosmid library of S. lividans TK24 was constructed using XL1-Blue as a host and SuperCos 1 as a vector according to the standard protocol (Kieser et al., 2000). The S. lividans TK24 genomic DNA was extracted and partially digested with Sau3AI. The 40–60 kb fragments were isolated by pulsed-field gel electrophoresis (PFGE), dephosphorylated, and ligated into the SuperCos 1 vector. The ligation product was packaged with λ phage packaging protein, transferred into E. coli XL1-Blue, and transformants were selected by kanamycin. The cosmid clones were extracted and digested with EcoRI and BamHI to verify that the average inserted exogenous fragment size was 39 kb. The genomic cosmid libraries of S. galtieri Sag48 and S. parvulus 10 were constructed using E. coli XL1-blue MR/pUZ8002 as a host and pJTU2554 as a vector as described (Chen et al., 2012).

# wblAsl Knockout of S. lividans SBT5

The genomic cosmid library of S. lividans TK24 was screened by PCR amplification using primers wblA-F (5<sup>0</sup> -CGTCC TCAACTGGCGGCGGTGAAT-3<sup>0</sup> ) and wblA-R (5<sup>0</sup> -GGCCCC TGATCCGGCCTCGGGGCT-3<sup>0</sup> ), and the cosmid clone 10H1 containing wblAsl was obtained. The wblAsl gene knockout plasmid was then constructed using 10H1 according to a PCR-targeting protocol (Gust et al., 2003). Firstly, the wblAsl in 10H1 was replaced by an aac(3)IV cassette amplified from pSET152 by λ-Red recombination. The resulting plasmid 10H1-1wblA::aac(3)IV was then transformed into E. coli DH5α containing the recombinant plasmid BT340, which expresses the FLP recombinase gene, to remove the aac(3)IV cassette by FLP recombination (Gust et al., 2003), yielding 10H1- 1wblA. The bla (ampicillin resistance gene) on the backbone of 10H1-1wblA was then replaced by the aac(3)IV-oriT cassette amplified from pIJ773 by λ-Red recombination, resulting in the wblAsl knockout plasmid pHLJ42. The wblAsl gene in S. lividans SBT5 was deleted by homologous recombination between pHLJ42 and the chromosomal DNA. pHLJ42 contains 25.7 and 14.7 kb regions of S. lividans chromosomal DNA flanking either side of the mutated wblAsl locus. When pHLJ42, which does not contain an autonomous replication region or integration locus, was introduced into S. lividans SBT5 by conjugation, the apramycin-resistant (Apr<sup>R</sup> ) exconjugant should be a single-crossover mutant. To identify double-crossover mutants, the offspring colonies from the single-crossover mutant were screened for the loss of apramycin resistance, indicating the loss of aac(3)IV. A doublecrossover mutant strain S. lividans SBT51wblA, i.e., a wblAsl mutant, was confirmed by PCR (herein renamed S. lividans LJ101).

<sup>1</sup>http://www.jcat.de

### Conjugation Using Mycelia as Recipient

Conjugation using mycelia was conducted following literature (Du et al., 2012). S. lividans LJ1018 was grown in 30 mL TSB liquid medium in a baffled flask, shaking at 180 rpm, 28◦C for 48 h. The mycelia was collected by centrifugation at 5,000 rpm and washed with equal volume of 10% of glycerol once and 2× YT twice. Then 0.6 mL of washed mycelia was resuspended in 0.3 mL of 2× YT in an Eppendorf tube, mixed with 0.3 mL exponential phase donor E. coli cells. The mixture was spun at 5,000 rpm for 10 s, and the precipitate was spread on MS agar plate.

## High-Throughput Screening (LEXAS) of Streptomyces Antibiotic BGCs Using S. lividans GX28 as the Library Expression Host

The high-throughput, tri-parental E. coli–Streptomyces conjugation of an arrayed genomic library, high-throughput fermentation, and bioactivity assay were carried out according to the LEXAS procedure (Xu et al., 2016). Cosmid or BAC libraries in E. coli DH10B in the format of 96-well plates were used as the arrayed donors for conjugation. Spores of S. lividans GX28 were used as recipients, and E. coli ET12567/pUB307 was used as the helper strain. The E. coli strains containing cosmid or BAC clones were cultured in LB liquid medium (150 µL/well), supplemented with apramycin, at 37◦C overnight, and then transferred to antibiotic-free LB, cultured for 4–6 h until the optical density at 600 nm (OD600) reached 0.4 to 0.6. E. coli ET12567/pUB307 (helper strain) was cultured in LB (120 mL/library) containing a final concentration of 50 µg/mL chloramphenicol at 37◦C until the OD<sup>600</sup> was between 0.4 and 0.6. The cells were then collected by centrifugation and resuspended in 20 mL LB medium. Next, 20 µL of ET12567/pUB307 was pipetted into each well of the 96-well plates in which the BAC/cosmid library was inoculated, and the plates were shaken on a rotary shaker at 200 rpm for 5 min to allow thorough mixing. S. lividans GX28 (recipient) was grown on MS sporulation medium for 5–6 days at 30◦C. The fresh spores were collected and resuspended in 4 mL of 2× YT medium, heat-shocked at 50◦C for 10 min, and then spread on MS plates supplemented with Mg2<sup>+</sup> (20 mM). The donor-helper E. coli mixtures were replicated from 96-well plates onto spore-coated MS plates using a 48-pin replicator. After incubation at 30◦C for 12 to 16 h, the MS plates were covered with apramycin and trimethoprim to final concentrations of 50 µg/mL to inhibit the E. coli strains. The exconjugants were cultured for another 4–6 days and then replicated to MS plates containing final concentrations of 50 µg/mL apramycin and 25 µg/mL nalidixic acid to remove the residual E. coli. The S. lividans GX28 exconjugants were fermented and subjected to high-throughput screening based on antibacterial activity. The S. lividans GX28 exconjugants of libraries were replicated to the agar fermentation media YBP, R3, GYM, No18, and No24 by replicator and cultured at 30◦C for 7 days. The surface of the fermentation media were covered with soft agar premixed with indicator bacteria, and the inhibition zones produced by heterologous expression of the active compounds were observed after 1–2 days of incubation.

# Sequence Analysis

The sequences of both ends of the inserts in BAC clones were determined with primers pHL921F (5<sup>0</sup> -ATGTTTTT CGTCTCAGCC-3<sup>0</sup> ) and pHL921R (5<sup>0</sup> -CCTTTAGTTG TTCCTTTC-3<sup>0</sup> ). The end sequences of cosmid clones were determined with primers pJTU2554F (TGTAA AACGACGGCCAGT) and pJTU2554F (GGCACCTG TCCTACGAGTTG). The DNA end sequences were then mapped to the genomic sequences. The DNA sequences of BAC or cosmid inserts were submitted to antiSMASH<sup>2</sup> for the analysis of secondary metabolic BGCs.

# Isolation and Analysis of Compounds

Actinorhodin was isolated and measured using a published method (Bystrykh et al., 1996). The S. lividans strains were cultured on solid YBP medium for 84 h, 500 mg agar culture was taken from each plates, 500 µL of 1 M NaOH was added, followed by crushing using a homogenizer (5,000 rpm, 15 s; twice). The samples were centrifuged at 12,000 × g, 5 min and the absorbance of the supernatants was measured at 633 nm. The isolation and analysis of piericidin A1, murayaquinone, dehydrorabelomycin, and actinomycin D, followed a similar approach as the following. Fermented culture (40 mL) was extracted three times with ethylacetate (150 mL). The combined extracts were concentrated on a rotary evaporator (Buchi R210) at 37◦C and then dissolved in 1 mL methanol. The crude extract (20 µL) was filtered and injected onto a C18 reversed-phase column (Agilent Zorbax ODS C18, 5 µm, 4.6 by 250 mm) and analyzed by high performance liquid chromatography (HPLC) in the Agilent 1260 HPLC system using mobile phase A (H2O supplemented with 0.1% formic acid) and mobile phase B (acetonitrile) at a flow rate of 0.6 mL/min. The elution procedure was: 0–2 min, 5% B (and 95%A); 2–25 min, 5–40% B; 25–35 min, 40–100% B; 35–40 min, 100% B; 40–45 min, 100–5% B; 45–55 min, 5% B.

The isolation and identification of hybrubin A was carried out as described (Zhao et al., 2016). Hybrubin A was eluted using the following HPLC conditions: mobile phase A was H2O (supplemented with 0.1% formic acid), mobile phase B was methanol; flow rate of 0.6 mL/min; 0 min, 40% B; 5–15 min, 65–80% B; 15–20 min, 80–100% B; 20–25 min, 100% B; 25– 26 min, 100–40% B; 26–35 min, 40% B.

Agilent G6530 HR ESI-QTOF mass spectrometry equipped with Agilent 1260 HPLC system was used to identify piericidin A1, dehydrorabelomycin, murayaquinone, actinomycin D, and hybrubins.

# RESULTS

#### Engineering of S. lividans SBT5-Derived Strains Using Global Regulatory Genes

The nusGsc gene of S. coelicolor A3(2) encodes an anti-terminator that is functionally conserved in prokaryotes, eukaryotes, and archaea (Mason and Greenblatt, 1991; Burmann et al., 2010).

<sup>2</sup>http://antismash.secondarymetabolites.org/

For cloning this gene with its native promoter, we amplified a 1.4 kb fragment containing the nusGSC coding region and the 536 bp upstream region by PCR, and then the fragment was ligated into pMS82 to yield the integrative plasmid pJTU6725 (**Figure 1A**). pJTU6725 was conjugated to S. lividans SBT5 to generate S. lividans GX25, in which the plasmid is integrated into the genome at the attB8BT1 site.

The lmrA gene in Lactococcus lactis subsp. cremoris MG1363 encodes a multidrug resistance ABC transporter ATP-binding and permease protein (van Veen et al., 1996), and the mdfA gene in Escherichia coli K-12 encodes a multidrug efflux transporter protein. Both lmrA and mdfA confer hosts with resistance to a variety of antibiotics by heterologous expression (Edgar and Bibi, 1997). The G+C contents of the original lmrA and mdfA genes were 39.0 and 52.4%, respectively. For the expression of these two multidrug resistance genes in the high G+C content genome of S. lividans, we synthesized the codon-optimized twin gene cassette lmrAco-mdfAco based on the protein sequences and the degenerate codon usage table of the S. coelicolor genome. The promoter of the non-ribosomal peptide synthase (NRPS) gene cdaPS1 from the CDA BGC was placed upstream of lmrAco to control the expression of lmrAco and mdfAco. The previously reported production of CDA in afsRSclacarrying S. lividans strains suggested that the PcdaPS<sup>1</sup> promoter has been activated (Chen et al., 2012; Bai et al., 2014). The synthetic operon PcdaPS1-lmrAco-mdfAco was ligated to pJTU6725 to construct pJTU6727 (**Figure 1A**). The integrative plasmid pJTU6727 was conjugated to S. lividans SBT5 to yield S. lividans GX27.

The global transcriptional regulator AfsR/Scla from S. clavuligerus ATCC 27064 (NRRL3585) increased the production of actinorhodin and CDA in S. lividans TK24 (Chen et al., 2012). We cloned afsR/Scla and the ermE<sup>∗</sup> promoter from pHL851 into pJTU6727 to construct pJTU6728 (**Figure 1A**), which was conjugated to S. lividans SBT5 to construct S. lividans GX28 (**Figure 1B**).

The gene wblAsl (SLIV\_20395) of S. lividans TK24 encodes a global transcriptional regulator of the WhiB family (Yu et al., 2014) and has 99% similarity to S. coelicolor wblAsc (SCO3579). To knock out wblAsl in S. lividans SBT5, the cosmid clone 10H1 containing wblAsl was obtained from a genomic cosmid library of S. lividans TK24. An in-frame deletion was made in wblAsl on 10H1 to construct the gene knockout vector pHLJ42, which contains wblAsl flanking sequences of 25.7 and 14.7 kb for homologous recombination. The wblAsl-knockout strain LJ101 was constructed using pHL42 via homologous recombination and confirmed by PCR (**Figure 1C**). Compared with the parental strain SBT5, S. lividans LJ101 exhibited a "white" phenotype (**Figure 1D**): no spore pigment was produced and the white aerial hyphae did not develop into spores. This indicated that wblAsl plays an important role in aerial hyphae development similar to the wblA from S. coelicolor A3(2) (Fowler-Goldsworthy et al., 2011). pJTU6728 was conjugated to S. lividans LJ101 to yield S. lividans LJ1018, which displayed a "bald" phenotype with only sparse white mycelium (**Figure 1D**).

Growth curves indicated that the introduction of pJTU6725, pJTU6727, and pJTU6728 into S. lividans SBT5 did not significantly affect the growth and biomass accumulation of the host strain. However, the biomass of the wblAsl deletion strains S. lividans LJ101 and S. lividans LJ1018 was significantly improved. The dry weight of the two strains was 1.6 times higher than that of S. lividans GX28 after 72 h of culture (**Figure 1E**; p < 0.0001).

#### Heterologous Expression of the Pigmented Polyketide Antibiotic Actinorhodin in Engineered S. lividans Strains

To test the ability of the engineered S. lividans strains to express polyketide BGCs, we expressed the actinorhodin BGC using S. lividans GX25, S. lividans GX27, S. lividans GX28, S. lividans LJ1018, and the parent strain S. lividans SBT5 as the expression hosts. The act BGC is a 22 kb type II polyketide synthase (PKS) BGC, and actinorhodin (**Figure 2A**) is a pH-sensitive, pigmented aromatic polyketide antibiotic that is red at acidic pH and blue at alkali pH. Plasmid pMM1 (45 kb) carrying the complete act BGC (Zhou et al., 2012) was introduced into the S. lividans series of hosts by conjugation. High conjugation frequencies, ca. 10−<sup>2</sup> /cfu, were observed when S. lividans GX25, S. lividans GX27, S. lividans GX28, and S. lividans SBT5 were used. Because S. lividans LJ1018 is deficient in sporulation, mycelium was used as the recipient for conjugation, and 100s of exconjugants were obtained on each conjugation plate, with a conjugation frequency of around 10−<sup>6</sup> /cfu. After fermentation in YBP medium for 72 h, the blue color of actinorhodin was observed due to the heterologous expression of the act BGC in the exconjugants. Observation of the color of the YBP fermentation medium revealed that the heterologous expression of actinorhodin in the optimized hosts S. lividans GX25, GX27, GX28, and LJ1018 progressively increased compared to levels in SBT5 (**Figure 2B**). The yield of actinorhodin of GX25/pMM1 was 1.3 times higher than that of SBT5/pMM1 (p < 0.001), and the yields of actinorhodin in S. lividans GX27/pMM1, GX28/pMM1, and LJ1018/pMM1 were 12.8, 21.6, and 23.3 times higher than that of SBT5/pMM1, respectively (p < 0.0001; **Figure 2C**), indicating that the addition of nusGsc, the drug efflux pump genes, and afsRScla and the knockout of wblAsl in SBT5 up-regulated the production of actinorhodin.

#### Heterologous Expression of the Aromatic Polyketide Antibiotic Murayaquinone in the Engineered S. lividans Strains

Murayaquinone is a tricyclic, angular aromatic polyketide 9,10 phenanthraquinone antibiotic produced by a type II PKS

shown. pMM1, plasmid carrying the S. coelicolor actinorhodin BGC. (C) Quantification of actinorhodin production by various expression hosts carrying pMM1 on

YBP medium. The productivity related to S. lividans SBT5/pMM1 was present. Data are from three biological replicates.

pathway (**Figure 3A**, Gao et al., 2017). The murayaquinone BGC is about 56 kb and was cloned into BAC clone 3B4. 3B4 was obtained by screening the genomic BAC library of S. griseoruber Sgr29 using S. lividans SBT5 as the high-throughput heterologous expression host, conferring the exconjugants with antibacterial activity against S. aureus (Gao et al., 2017). The exconjugants carrying 3B4 were fermented on solidified media No18 and No24, and the crude extracts were analyzed by HPLC. Three new peaks were observed on the HPLC trace of S. lividans GX28/3B4 fermented on No18 medium, with the same retention time as the standard samples of murayaquinone and murayalactone 1, 2 (**Figure 3B**). Murayalactone 1 and 2 were the main products on No18 medium, while murayaquinone was the main product on No24 medium. The identity of murayaquinone and murayalactones isolated from S. lividans GX28/3B4 was further confirmed by high-resolution mass spectrometry (HR-MS) (**Figure 3C**). However, murayaquinone and murayalactones were not detectable by HPLC from the S. griseoruber Sgr29 fermentation culture (**Figure 3B**), which is consistent with the literature. To compare the production of murayaquinone, all exconjugants were fermented on No24, and the areas of murayaquinone peaks in the HPLC traces were measured. The yield of murayaquinone was extremely low (about 0.11 mg/L) in the parent host SBT5/3B4 but was significantly increased in all engineered hosts (p < 0.05). The yields of murayaquinone in S. lividans GX28/3B4 and S. lividans LJ1018/3B4 were

much higher than that of the original host SBT5/3B4 (74 and 96 times higher, respectively, p < 0.0001), and the yield of murayaquinone in S. lividans LJ1018/3B4 was 10.6 mg/L (**Figure 3D**).

#### Heterologous Expression of Hybrubins in the Engineered S. lividans Strains

The hbn BGC from Streptomyces variabilis Snt24 is a small PKS-NRPS hybrid BGC responsible for the biosynthesis of 5-ethylidenetetramic acid (ETA); the truncated red pathway in S. lividans SBT5 synthesizes 4-methoxy-2,2<sup>0</sup> -bipyrrole-5 carbaldehyde (MBC), and condensation of ETA with MBC produces the "non-natural" red compounds named hybrubins (Zhao et al., 2016). pZZL3 is an integrative plasmid containing a 13 kb hbn BGC cloned from the S. variabilis Snt24 genome. The heterologous expression of hbn BGC in S. lividans SBT5 led to the production of the red-pigmented secondary metabolites hybrubin A-C (**Figure 4A**, Zhao et al., 2016). When the pZZL3 carrying exconjugants of S. lividans SBT5, GX25, GX27, GX28, and LJ1018 were fermented with R3 medium, red pigment was observed in the crude extract whereas the vector control did not produce red pigment (**Figure 4B**). HPLC analysis indicated that hybrubins A-C were produced and that hybrubin A was the main component (**Figure 4C**). The identity of hybrubin A was confirmed by HR-ESI-MS (**Figure 4D**). The relative yield of hybrubin A was evaluated based on the HPLC peak area.

FIGURE 3 | Heterologous expression of the murayaquinone BGC in engineered S. lividans strains. (A) Structure of murayaquinone and murayalactones. (B) HPLC analysis of the exconjugants carrying the murayaquinone BGC from BAC 3B4 and of the original strain Streptomyces griseoruber Sgr29. The absorbance was measured at 350 nm. No18 and No24 agar media were used for fermentation. 3B4, a BAC clone containing the murayaquinone BGC. No murayaquinone or murayalactones were detected from S. griseoruber Sgr29. (C) HR-MS spectrum of murayaquinone and murayalactone 1 isolated from S. lividans GX28/3B4. The spectra of murayalactone 1 and 2 are identical. (D) Quantification of murayaquinone production on No24 medium by engineered S. lividans strains carrying 3B4. Data are from three biological replicates.

The yield of hybrubin A in GX25/pZZL3 and GX27/pZZL3 was slightly higher than in SBT5/pZZL3 (2.2 and 2.5 times, respectively, p < 0.05), whereas the yield in GX28/pZZL3 and LJ1018/pZZL3 was greatly increased, reaching 13 times and 29 times the yield in SBT5/pZZL3, respectively (**Figure 4E**).

# Discovery of a Piericidin A1 BGC Using LEXAS and S. lividans GX28

We tested the ability of the engineered strain S. lividans GX28 to serve as a host for LEXAS screening of antibiotics and their corresponding BGCs, using the S. griseoruber Sgr29 genomic BAC library, which contains 912 arrayed clones with an average insertion size of about 100 kb (Gao et al., 2017). S. lividans SBT5 had been used as a host for the high-throughput heterologous expression in a previous screening, and seven positive BAC clones with S. aureus resistance were obtained from this genomic BAC library, three of which contained the murayaquinone BGC (Gao et al., 2017). Using S. lividans GX28 as the expression host, nine new S. aureus-resistant positive BAC clones were obtained, five of which (4E6, 4F9, 1H5, 2D3, and 4G10) shared overlapping DNA regions (**Figure 5A**). The termini of these five BACs were sequenced with primers pHL921-F/R, and then the sequences were aligned with the S. griseoruber Sgr29 genomic sequence. The five BAC plasmids were found to have a 98 kb overlapping region, and analysis of this region by AntiSMASH revealed that it contains a 50 kb piericidin A1 BGC, which included six type I polyketide synthase (PKSI) genes and five postmodification genes highly homologous to piericidin A1 BGC genes in S. piomogeues. Piericidin A1 is an α-pyridone antibiotic (**Figure 5B**) that inhibits the mitochondrial respiratory chain and NADH-ubiquinone oxidase and exhibits weak antimicrobial and antitumor activities (Liu et al., 2012; Chen et al., 2014). To verify the function of the piericidin A1 BGC, one of the BAC clones, 1H5, was transferred to the expression host S. lividans GX28, and the exconjugants was fermented with R3 medium. The fermented culture of GX28/1H5 had inhibitory activity against B. mycoides, whereas the empty vector control (S. lividans GX28/pHL921) did not produce an inhibition zone (**Figure 5C**). HPLC and HR-MS analysis indicated that piericidin A1 was produced by GX28/1H5 and S. griseoruber Sgr29 (**Figures 5D,E**).

To detect the yield of piericidin A1 in different expression hosts, 1H5 was transferred into the five S. lividans hosts for heterologous expression, and the resulting exconjugants and

the natural strain S. griseoruber Sgr29 were fermented. HPLC analysis showed that, although S. griseoruber Sgr29 produced high levels of piericidin A1 (156.6 mg/L), the yields resulting from heterologous expression in the S. lividans hosts were significantly higher (p < 0.001). The yield of piericidin A1 from S. lividans GX28/1H5 was 2.4 times that of S. griseoruber Sgr29, and the yield from S. lividans LJ1018/1H5 was even higher, at 3.1 times the yield from S. griseoruber Sgr29 and reaching 478 mg/L (**Figure 5F**).

## Discovery of a Dehydrorabelomycin BGC Using LEXAS and S. lividans GX28

We constructed a genomic cosmid library of S. galtieri Sag48, a species isolated from forest soil by CCTCC, and performed LEXAS screening using S. lividans GX28 as the highthroughput heterologous expression host. The LEXAS screening identified an exconjugant displaying weak inhibition activity against B. mycoides and which contained cosmid plasmid 8F5. Sequencing analysis revealed that 8F5 has a 32 kb insertion sequence containing 27 genes having high level of similarity (81–95%) to the alpA-alpW genes in the type II polyketide BGC of kinamycin from S. ambofaciens (**Figure 6A**). The complete kinamycin BGC is 63 kb and cannot be packaged into a single cosmid clone (Wang et al., 2015; Liu et al., 2018). Although cosmid 8F5 contains PKS genes (alpABC) and early modification genes for the synthesis of kinamycin intermediates, it does not contain other genes required for the synthesis of the final product (i.e., kinamycin). To analyze the metabolites produced via this cosmid, 8F5 and the vector pJTU2554 were introduced into the five S. lividans hosts by conjugation, and the resulting exconjugants and the natural strain S. galtieri Sag48 were fermented on No18 agar plates. Extracts of the fermented cultures were analyzed by HPLC. The crude extract of S. lividans LJ1018/8F5 produced an absorption peak at 39 min, which was not produced by S. galtieri Sag48 and the vector control strain S. lividans LJ1018/pJTU2554 (**Figure 6B**). The compound was detected by LC-MS, and its molecular weight, with an m/z value of 321.0710, was consistent with that of dehydrorabelomycin (m/z of [M+H]<sup>+</sup> calcd. 321.0763) (**Figures 6C,D**), which is an intermediate of the kinamycin biosynthetic pathway.

Quantitative comparison of dehydrorabelomycin production indicated that S. lividans GX28/8F5 and LJ1018/8F5 yielded levels 6.7 times and 12.7 times, respectively, the amount

produced by the original host SBT5/8F5. The highest yield of dehydrorabelomycin was produced by S. lividans LJ1018/8F5, reaching a level of 118.0 mg/L (**Figure 6E**).

# Discovery of an Actinomycin D BGC Using LEXAS and S. lividans GX28

Streptomyces parvulus 10 was isolated from the marine sponge Carteriospongia foliascens collected from the South China Sea by the Zhiyong Li Group. We constructed a cosmid library of the S. parvulus 10 genome and used S. lividans GX28 as the heterologous expression host for high-throughput library screening. An exconjugant exhibiting S. aureus inhibitory activity was observed. Sequencing analysis of the corresponding cosmid, 5H11, revealed that it contains an NRPS BGC (**Figure 7A**) with 16 genes highly similar (78–95% similarity) to genes of the actinomycin C of Streptomyces anulatus (Keller et al., 2010). The compound produced by S. lividans GX28/5H11 was determined to be actinomycin D (**Figure 7B**) by HPLC and LC-MS (m/z of [M+H]<sup>+</sup> obsd. 1255.6359, calcd. 1255.6363) (**Figures 7C,D**).

To compare the heterologous expression of actinomycin BGC in different hosts, 5H11 was transferred to the five S. lividans hosts, and the exconjugants were fermented in YBP medium. After 4 days of fermentation, the bioactivity test indicated that the inhibition zones produced by the exconjugants of the newly engineered S. lividans hosts were larger than for the parental strain. The S. lividans LJ1018/5H11 fermented culture displayed the largest inhibition zone against B. mycoides (**Figure 7E**). HPLC quantitative determination confirmed that the production of actinomycin D in S. lividans strains increased in turn (p < 0.01 or p < 0.001), i.e., the production in LJ1018/5H11 > GX28/5H11 > GX27/5H11 > GX25/5H11 > SBT5/5H11 (**Figure 7F**). LJ1018/5H11 was capable of producing 52.1 mg/L actinomycin D, which was 3.5 times the level produced by SBT5/5H11. The native strain S. parvulus 10 also synthesized actinomycin D (18.0 mg/L). The actinomycin D yields of strains GX27/5H11, GX28/5H11, and LJ1018/5H11 were 32% (p < 0.0001), 95% (p < 0.0001), and 190% (p < 0.0001) higher than that of S. parvulus 10, respectively.

# DISCUSSION

Whether a host can effectively express all the essential genes of a given heterologous synthetic pathway is key to the success of the heterologous production of secondary metabolites. When

high-throughput heterologous expression methods are used to screen metagenomic or genomic libraries (Baltz, 2008), ideally overall gene expression should be improved by manipulating global regulatory genes in the expression host, rather than by attempting to modify all of the individual promoters within BGCs, a potentially complex and cumbersome task (Chen et al., 2010) and one not possible with previously unknown BGCs. Although previously engineered hosts, such as S. coelicolor, have altered global regulatory genes to promote the expression of BGCs, due to the restriction of methylated DNA and the slightly lower frequency of conjugative transfer (MacNeil, 1988), these strains are not well-suited to be LEXAS high-throughput screening hosts (Chen et al., 2012).

Streptomyces lividans has the advantage of high frequency of conjugative transfer and no restriction on exogenous methylated DNA (Martinez et al., 2004), and S. lividans TK24 strain itself contains an rpsL[K88E] mutation that promotes gene expression (Ochi, 2007). We previously added 1–2 copies of the global regulatory gene afsRScla to the TK24 genome, and the resultant strains indeed contributed significantly to the establishment of a high-throughput library expression and screening system (LEXAS) (Xu et al., 2016). To further optimize the S. lividans host and the LEXAS system, in this study we continued to optimize S. lividans with global regulatory genes, including nusGsc and wblAsl, as well as drug efflux pump genes, in addition to afsRScla.

Many antibiotic BGCs carry export genes, such as actII-ORF2 in the actinorhodin BGC (Fernándezmoreno et al., 1991) and rifP in the rifamycin BGC (August et al., 1998). These efflux pumps secrete the antibiotics out of the cell, thereby reducing the feedback inhibition of the end-products on the biosynthetic enzymes, while increasing the self-tolerance to the antibiotics. Therefore, overexpression of antibiotic efflux pumps is helpful when engineering strains to increase the production of antibiotics of interest (Qiu et al., 2011). MdfA of E. coli is a multi-drug transporter of the major facilitator superfamily (Sigal et al., 2006); it has a broad-spectrum recognition and efflux function for toxic compounds and enhances the tolerance of the host strains to natural or synthetic antibiotics such as daunomycin, rifampin, puromycin, aminoglycoside antibiotics, and quinolones (Edgar and Bibi, 1997). LmrA of Lactococcus lactis subsp. cremoris MG136362 belongs to a family of multidrug resistance ABC (ATP-binding cassette) transporters driven

by ATP hydrolysis (van Veen et al., 1996; Wilkens, 2015), and its sequence is highly similar to that of the multi-drug resistance export pump P-glycoprotein (MDR) in mammals (Margolles et al., 1999). LmrA and MDR1 increased the tolerance of bacterial cells to compounds such as daunomycin, ethidium, rhodamine 6G, and tetraphenylphosphonium (van Veen et al., 1996). We added two codon-optimized efflux pumpencoding genes, mdfAco and lrmAco, into S. lividans GX25 to construct GX27. Our quantitative data on heterologous expression suggested that the introduction of these two efflux pump genes significantly increased the yield of four antibiotics, including actinorhodin, dehydrorabelomycin, piericidin A1, and actinomycin D, demonstrating that it is applicable to use multidrug transporters for the general improvement of antibiotics production in heterologous hosts.

The second group of ideal engineering targets are global regulators. NusG, the regulator of the NusG-like family, functions as an RNAP processivity clamp and is the only antiterminator factor conserved among the kingdoms of prokaryotes, eukaryotes, and archaea (Burmann et al., 2010). Behnken et al. (2012) used the constitutive strong promoter Pthl to increase the expression level of nusG, thereby successfully activating the originally silenced polythioamides BGC in the genome of the anaerobic bacterium Clostridium cellulolyticum and unveiling seven new compounds. In this study, we inserted S. coelicolor nusGsc into the S. lividans SBT5 and used the resultant GX25 as a heterologous host to express six different types of antibiotic BGCs, the production of five out of six antibiotics increased significantly (p < 0.05). Similarly, additional copy of the positive regulatory gene afsRcla and sigma factor-like gene afsScla increased the production of six antibiotics significantly. These results suggest that global positive regulatory genes like nusGsc and afsRScla are applicable for improving the heterologous expression of PKS, NRPS, and NRPS-PKS BGCs.

WblA is a global negative regulator unique to actinomycetes (Kang et al., 2007) and has obvious sequence similarity to the developmental differentiation factor WhiB (Chater et al., 2000). In many actinomycetes, knocking out wblA significantly improved antibiotic biosynthesis in the mutant strains (Kang et al., 2007; Noh et al., 2010; Rabyk et al., 2011; Nah et al., 2012; Yu et al., 2014). The molecular mechanism by which WblA negatively regulates antibiotic synthesis remains unclear. We knocked out wblAsl in S. lividans GX28 to obtain LJ1018, which led to significant increases in the production of hybrubins, dehydrorabelomycins, and actinomycin D (p < 0.05), and slight increases of the three antibiotics actinorhodin, murayaquinone, and piericidin A. However, the mutant strains S. lividans LJ1018 and LJ101 that we constructed do not produce spores. This is not surprised since WblA plays an important role in the formation of aerial hyphae in Streptomyces (Fowler-Goldsworthy et al., 2011). After knocking out wblA in S. coelicolor and S. chattanoogensis L10, no spores were formed on the aerial hyphae (Yu et al., 2014). As a consequence, we had to use mycelium instead of spores as the recipient during conjugation transfer, which reduced the frequency of conjugation sharply. Nevertheless, this characteristic did not affect the introduction of target BGCs into the host for heterologous expression, since dozens to 100s of exconjugants could be obtained for each mycelium conjugation in our laboratory. However, when we attempted to use S. lividans LJ1018 mycelium as LEXAS host for high-throughput expression of arrayed cosmid libraries and BAC libraries, only sporadic exconjugants emerged, so LJ1018 is not suitable as a host for high-throughput heterologous expression of arrayed libraries.

In contrast, strains GX25, GX27, and GX28 still produce abundant spores. Both high-throughput and conventional conjugation transfer worked as efficiently as with the parental strain SBT5. When screening the two cosmid libraries (from S. galtieri Sag48 and S. parvulus 10) using GX28 as the high-throughput expression host, 3948 out of 4032 cosmid clones (98%) yielded exconjugants, and when we screened a BAC library using GX28, 818 out of the 912 clones (93%) produced exconjugants. BGCs producing dehydrorabelomycin and actinomycin were identified from the cosmid libraries. Five clones containing the complete piericidin A1 BGC, in addition to clones carrying the murayaquinone BGC, were identified from the S. griseoruber Sgr29 genomic BAC library. Notably, these piericidin BGC clones had been overlooked during the previous screening using SBT5 as a host (Gao et al., 2017). Indeed, the corresponding SBT5 exconjugants did not show significant antibacterial activity, since no zone of inhibition was produced. Our genomic screening results demonstrate that the GX28 strain is an excellent expression host for the LEXAS procedure to screen for functional BGCs in arrayed cosmid libraries and BAC libraries, and also demonstrates the superiority of GX28 as a heterologous expression host.

In summary, S. lividans provides excellent host strains for high-throughput screening of genomes (such as LEXAS screening) due to its rapid growth, abundant sporulation, high frequency of conjugation transfer, no methylation restriction on methylated DNA, and efficient expression of heterologous BGCs after rational engineering. By sequentially engineering global regulatory genes and multi-drug transporters, we obtained four engineered strains of S. lividans, which in turn increased the yield of multiple synthesized antibiotics that involve PKS, NRPS, and PKS-NRPS hybrid pathways. Since cryptic BGCs in microorganisms usually encode new or unknown biosynthetic pathways, it is very difficult to specifically engineer pathwayspecific regulatory factors or to appropriately modify promoters. Our optimized host GX28 produces high yields of antibiotics and does not require one-by-one modification of the promoters in a BGC of interest, so it is an excellent heterologous expression host for LEXAS high-throughput screening of cosmid or BAC libraries for the discovery of new, previously silenced BGCs and corresponding compounds. In addition, our study has revealed that the positive regulatory genes nusGsc and afsRScla, the negative regulatory gene wblAsl, and efflux pump genes, which are not regulatory genes by definition, have synergistic effects on the synthesis of antibiotics when they are combined in one host. The yields of the tested antibiotics were increased several times, even dozens of times in the case of hybrubins, over yields from the parent strain, SBT5. Furthermore, since the strain engineering conducted here mainly utilizes plasmid integration, these plasmids, especially pJTU6728, which integrates nusG, afsRScla, and efflux pump genes, can be used for the engineering of other strains for high-yield production of antibiotics in future. Therefore, the strains we have generated and the approaches we have used should aid in the identification of new BGCs and in optimizing the production of secondary metabolites of clinical and industrial value.

# AUTHOR CONTRIBUTIONS

fmicb-09-03042 December 7, 2018 Time: 17:16 # 13

MT, ZD, and ZL were responsible for the original concept and designed the experiments. MT and YW analyzed the data. QP, GG, JL, QL, XC, FZ, MX, KL, and YW performed the experimental work. QP and MT wrote the manuscript. All authors read and approved the final manuscript.

### REFERENCES


# FUNDING

This work was supported by the Chinese Ministry of Science and Technology through a China-Australia Joint Grant (Grant No. 2016YFE0101000), the National Natural Science Foundation of China (Grant No. 31770036), the Science and Technology Commission of Shanghai Municipality (Grant No. 15JC1400401), and the National Key Research and Development Program of China (Grant No. 2018YFC030900).

# ACKNOWLEDGMENTS

We thank Dr. Songwang Hou and Prof. Tianshen Tao for gifts of bacterial strains.



biosynthetic genes. Org. Lett. 18, 572–575. doi: 10.1021/acs.orglett.5b0 3609


**Conflict of Interest Statement:** 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 © 2018 Peng, Gao, Lü, Long, Chen, Zhang, Xu, Liu, Wang, Deng, Li and Tao. 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.

# Exploring Structural Diversity of Microbe Secondary Metabolites Using OSMAC Strategy: A Literature Review

#### Rui Pan<sup>1</sup> , Xuelian Bai<sup>2</sup> , Jianwei Chen<sup>1</sup> , Huawei Zhang<sup>1</sup> \* and Hong Wang<sup>1</sup> \*

<sup>1</sup> School of Pharmaceutical Sciences, Zhejiang University of Technology, Hangzhou, China, <sup>2</sup> College of Life and Environmental Sciences, Hangzhou Normal University, Hangzhou, China

#### Edited by:

Dipesh Dhakal, Sun Moon University, South Korea

#### Reviewed by:

Juan Carlos Aon, GlaxoSmithKline, United States Anil Shrestha, Ewha Womans University, South Korea

\*Correspondence:

Huawei Zhang hwzhang@zjut.edu.cn Hong Wang hongw@zjut.edu.cn

#### Specialty section:

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

> Received: 11 August 2018 Accepted: 04 February 2019 Published: 26 February 2019

#### Citation:

Pan R, Bai X, Chen J, Zhang H and Wang H (2019) Exploring Structural Diversity of Microbe Secondary Metabolites Using OSMAC Strategy: A Literature Review. Front. Microbiol. 10:294. doi: 10.3389/fmicb.2019.00294 Microbial secondary metabolites (MSMs) have played and continue to play a highly significant role in the drug discovery and development process. Genetically, MSM chemical structures are biologically synthesized by microbial gene clusters. Recently, however, the speed of new bioactive MSM discovery has been slowing down due to consistent employment of conventional cultivation and isolation procedure. In order to alleviate this challenge, a number of new approaches have been developed. The strategy of one strain many compounds (OSMAC) has been shown as a simple and powerful tool that can activate many silent biogenetic gene clusters in microorganisms to make more natural products. This review highlights important and successful examples using OSMAC approaches, which covers changing medium composition and cultivation status, co-cultivation with other strain(s), adding enzyme inhibitor(s) and MSM biosynthetic precursor(s). Available evidences had shown that variation of cultivation condition is the most effective way to produce more MSMs and facilitate the discovery of new therapeutic agents.

Keywords: OSMAC strategy, microbe secondary metabolite, structural diversity, medium composition, co-cultivation, epigenetic modification

#### INTRODUCTION

Microbial secondary metabolites (MSMs) have been recognized as the primary source of new compounds for drug discovery and development (Gunatilaka, 2006; Rateb et al., 2011b; Deng et al., 2013). Traditional chemical investigation of microorganism mainly focuses on extraction and isolation of structurally and highly active compounds from fermentation broth and mycelium. However, these processes are becoming inefficient due to high rate of the re-discovery of known MSMs. It is commonly believed that a large portion of microbial gene clusters are silenced under standard fermentation conditions (Scherlach and Hertweck, 2009; Wasil et al., 2013). By mining microbial genome and targeting biosynthetic gene clusters of MSM, researchers can exploit the potential of microbes in a more objective way, such as knocking down, introduction or heterologous expression of microbial genes, regulation of promoters, induction of mutations, or changing cultivation conditions to stimulate MSM genes expression (Schneider et al., 2008). Variation of cultivation condition has been deemed to be the simplest and most effective strategy, which is termed as "one strain many compounds (OSMAC)" by professor Zeeck and coworkers (Bode et al., 2002). On basis of extensive literature search, important and successful examples using OSMAC strategy are summarized in this review, which consists of variation of medium, changing cultivation condition, cocultivation with other strain(s), adding epigenetic modifier(s) or biosynthetic precursor(s).

#### VARIATION OF MEDIUM

fmicb-10-00294 February 23, 2019 Time: 16:42 # 2

Culture medium has a greater effect not only on microbe growth but also on metabolism. It has been reported that C/N ratio, salinity, and metal ion can regulate the degree and pattern of MSM gene expression and result in production of various secondary metabolites.

#### Medium Composition

Generally, carbon and nitrogen sources are major components in the culture medium. The carbon source not only provides the basis for building biomass and represents the source of energy for all heterotrophs but also delivers carbon units for secondary metabolites. The nitrogen source is required for the synthesis of essential proteins and nucleic acids, and likewise N-containing units for secondary metabolites. The type of used carbon and nitrogen sources is known to have a significant influence on microbial secondary metabolism (Ruiz et al., 2009; Singh et al., 2017). Furthermore, the C/N ratio is one of important factors that affect fermentation products (Karakoç and Aksöz, 2004; Brzonkalik et al., 2012; Dinarvand et al., 2013). Notably, the consumption of carbon and nitrogen-based medium components can greatly affects the pH of the cultivation broth, e.g., by formation of organic acids or the accumulation of basic ammonium. Thus, microorganisms cultured in medium containing different components may exhibit differently adapted metabolism and express specific sets of biosynthetic genes, which produced a differential biosynthesis of specialized metabolites (Ma et al., 2009).

One marine-derived strain Asteromyces cruciatus 763 was shown to produce a new pentapeptide lajollamide A (**1**), when cultivated in the Czapek-Dox broth contained arginine solely as nitrogen source rather than NaNO3, which was missed in the normal Czapek-Dox medium (Gulder et al., 2012). One sediment-derived Aspergillus niger BRF-074 produced a novel furan ester derivative (**2**), a compound has toxicity acidity against HCT-116 cancer cell line (Uchoa et al., 2017), when cultivated in MPDB (malt peptone dextrose broth) medium. But this compound failed to appear in PDB (potato dextrose broth) or PDYB (potato dextrose yeast broth) media. A fungus Aspergillus sp. from Waikiki Beach (Honolulu, HI), generated six isotopically labeled metabolites (**3**–**8**) when grown on the deuterium-enriched Czapek broth (Wang et al., 2015a), whereas this strain was found to metabolite a novel prenylated indole alkaloid, waikialoid A (**9**) when cultivated in PDB medium. Bioassay results indicated that compound **9** possessed potent

Five new polyketides (**10**–**14**) were detected in the crude extract of rice-based medium of a marine-derived Cladosporium sphaerospermum 2005-01-E3 (Wu et al., 2014). Another two new hybrid polyketides (**15**–**16**) were accessed when the same strain was fermented on the soybean flour (Yu et al., 2015). The organic extract of Dothideomycete sp. CRI7 was elaborated by four comparative medium. The strain growing in PDB made with potato tubers led to the isolation of azaphilone derivatives (**17**–**18**) and a novel tricyclic polyketide (**19**). Only compound **19** exhibited a broad spectrum of cytotoxic activities (Senadeera et al., 2012). It is interesting that MSM production by strain CR17 was sensitive to sources of potato and malt extract used for the preparation of PDB and Czapek malt media, respectively. Three new polyketides (**20**–**22**) were produced when strain CR17 was grown in PDB broth prepared from a commercial potato powder instead of fresh tubers of potato, while this strain produced several other compounds (**20**–**21** and **23**–**25)** in Czapek malt medium. Compound **24** exhibited cytotoxic activity against cancer cell lines MOLT-3, HuCCA-1and A549 with IC<sup>50</sup> values of 17.4, 48.1, 46.5 µg/mL, respectively (Hewage et al., 2014). One fungus strain of Fusarium tricinctum isolated in Beni-Mellal, which can colonize the rhizomes of Aristolochia paucinervis, could afforded three new fusarielins (**26**–**28**). But these metabolites were not detected when cultivated in normal rice medium supplemented with fruit and vegetable juice. Bioassay results suggested that compound **26** possessed cytotoxic effect on human ovarian cancer cell line A2780 with an IC<sup>50</sup> value of 12.5 µM (Hemphill et al., 2017). A new diketopiperazine (**29**) was isolated from Eurotium rubrum MPUC136 cultured by

wheat medium, which displayed more powerful bioactivity than the Czapek-Dox agar medium, and shown to have cytotoxicity against B<sup>16</sup> melanoma cell line with an IC<sup>50</sup> value of 60 µM (Kamauchi et al., 2016).

HPLC analysis of crude extracts of an actinomycete strain Lentzea violacea AS08 indicated different composition in three media including CYPS (casein yeast peptone), SCP-1 (starch casein peptone), and SC (starch casein) (Hussain et al., 2017). Only one new eudesmane sesquiterpenoid (**30**) and a new analog of virginiae butanolide E (**31**) were detected in SC medium, and compound **30** exhibited moderate cytotoxic effect on HCT-116 and A549 tumor cell lines with IC<sup>50</sup> values of 19.2 and 22.3 µM, respectively. One rhizosphere fungus Paraphaeosphaeria quadriseptata produced a known C18 polyketide monocillin I together with several analogs when incubated in PDA medium constituted with tap water (Wijeratne et al., 2004). However, the same fungal strain could make six new trihydroxybenzene lactones, cytosporones F-I (**32**–**37**), when the tap water was changed as distilled water (Paranagama et al., 2007). Similarly, one new naphthalopyran compound (**38**), which possesses an unusual oxygenated aromatic structure with a lactone bridge, could be metabolized by the fungus P. hordei grown on plant tissue agar such as macerated tulip and yellow onion, oatmeal and red onion, while it was not detected in CYA (caffeic acid agar), MEA (malt extract agar), and YES (yeast extract with supplements) media (Overy et al., 2005). When cultivated in rice medium, a hard coral-derived fungus Scopulariopsis sp. from the coastline of Red Sea was shown to afford six secondary metabolites including xanthone derivatives (**39**– **40**), phenolic bisabolane-type sesquiterpenes (**41**–**42**), one new alkaloid (**43**) and one new α-pyrone derivative (**44**) (Elnaggar et al., 2016). Interestingly, this strain could biosynthesize a new naphthoquinone derivative (**45**) and two new triterpenoids (**46**– **47**) in the protein-rich white bean medium (Elnaggar et al., 2017).

Chemical investigation of one marine-derived strain Streptomyces sp. C34 grown on ISP2 (yeast malt extract agar) medium led to the isolation of four new ansamycin-type polyketides (**48**–**49**). But only compounds **48**, **50,** and **51** could be extracted from modified ISP2 medium, which contained glycerol rather than glucose. Bioassay results indicated that

compound **51** had a selective inhibitory effect on S. aureus ATCC 25923 with a MIC value of 0.05 µg/mL (Rateb et al., 2011a). The utilization of a defined medium to cultivate strain C34 resulted in the observation of three novel 22-membered lactone polyketides (**52**–**54**) (Reid et al., 1995). Compounds **50**–**52** possessed strong antibacterial activities against L. monocytogenes and B. subtilis with MIC values range from 3 to 6 µg/mL and against S. aureus with MIC values of <1 µg/mL (Rateb et al., 2011a). Four media applied to strain Streptomyces sp. CS resulted in production of various natural products including three new macrolides (**55–57**) from YMG agar medium, five new 16-membered macrolides (**58–62**) from ISP2 broth, five novel polyketides (**63**–**67**) from sterilized Waksman Synthetic medium and three new naphthomycins (**68–70**) from oatmeal medium. Compounds **55** was shown to have inhibitory effect on Fusarium moniliforme with a MIC value of 300 µg/mL and compounds **58–62** exhibited cytotoxicity toward the MDA-MB-435 human cancer cell line with IC<sup>50</sup> values of 4.2, 4.5, 5.5, 3.8, and 11.4 mM, respectively (Lu and Shen, 2003, 2004; Li et al., 2008, 2010; Yang et al., 2012). Streptomyces sp. ML55 in a medium consisting of glycerin, molasses, casein, polypeptone led to the isolation of three novel antimycins, JBIR-02 (**71**), JBIR-06 (**72**), and JBIR-52 (**73**), while this strain had capacity to produce two novel depsipeptides (**74–75**) in GYM medium (Ueda et al., 2007, 2008; Kozone et al., 2009; Li X. et al., 2013). An ant-derived actinomycete Streptomyces sp. 1H-GS5 was found to produce one new spectinabilin derivative (**76**) when cultivated in the medium consisting of corn starch 10%, soybean powder 1%, cotton flour 1%, α-amylase 0.02%, NaCl 0.1%, K2HPO<sup>4</sup> 0.2%, MgSO4 · 7H2O 0.1%, CaCO<sup>3</sup> 0.7%, cyclohexanecarboxylic acid 0.1%, pH 7.0, while this stain made another new cytotoxic spectinabilin (**77**) when reducing the proportion of nutrients (Liu S. et al., 2015; Liu C. X. et al., 2016).

When cultured in an oat bran medium, one strain Streptomyces sp. A1 was found to produce rubromycin derivatives, while other three known compounds were biosynthesized in a mannitol/soybean meal medium and three new congeners (**78**–**80**) and streptenol E (**81**) in medium (degreased soybean meal 2%, mannitol 2%, agar 2%) with soil as an addition, provide. Compound **81** had significant cytostatic effect on four tumor cell lines including HMO2, HEP G2, MCF7 and Kato III with GI<sup>50</sup> values (the concentration that causes 50% growth inhibition) of 0.15, 0.3, 10, and 0.7 µM, respectively (Puder et al., 2001). Phytochemical study of one filamentous soil fungus, Talaromyces wortmannii, cultivated in maize culture medium, led to the separation of three new polyketones (**82**–**84**), which were absent in rice or dextrose agar media. Compounds **82**–**84** displayed inhibitory activities against NFRD (fumarate reductase) with IC<sup>50</sup> values of 8.8, 11, and 13 µM, respectively (Liu W. C. et al., 2016). Interestingly, this stain was found to produce four novel 22-membered macrolides (**85**–**88**) (Dong et al., 2006) and four novel tetraene lactones (**89**–**92**) (Dong et al., 2009) when grown in the still-cultured medium (2.5% soybean meal and 97.5% rice). Compounds **85**–**88** exhibited in vitro moderate cytotoxic activities against human cancer cell lines (HCT-5, HCT115, A549, MDA-MB-231, and K562) with IC<sup>50</sup> values range from 28.7 to 130.5 µM, while compounds **89**–**92** showed potent inhibitory effects on cathepsin B.

#### Salinity

Salinity is an important factor in determining many aspects of the chemistry of natural water and of biochemical process within cultivation system, and is a thermodynamic state variable that, along with temperature and pressure, governs physical characteristics like the osmotic pressure and enzymes involved in microbial growth and metabolism (Blunt et al., 2015). Suitable salinity is needed for normal microbial growth and high osmotic pressure makes cells dehydrated and affects microbial biochemical reactions (Poolman and Glaasker, 1998; Wang Y. et al., 2011).

Microorganisms exposed to different types of media supplemented with various halogens maybe trigger their synthesis pathway to restore osmotic imbalance, thus activating different hidden MSM biosynthetic gene clusters. Compare to that grown in seawater, one marine-derived fungus Aspergillus unguis CRI282-03 was shown to produce new brominated depsidones (**93**–**95**) and two new orcinol derivatives (**96**–**97**) in KBr medium and a new depsidone (**98**) in KI broth (Sureram et al., 2013). Bioassay results indicated that compounds **95** and **96** possesses aromatase inhibitory effects (Sureram et al., 2012). Nine new polyketides (**99**–**107**), which were absent in the broth contained KI or deionized water, were produced by the fungus Dothideomycete sp. CRI7 isolated from Tiliacora triandra when cultivated in the medium supplemented with KBr and seawater (Wijesekera et al., 2017).

Chemical investigation of one symbiotic stain Aspergillus sp. D from Edgeworthia chrysantha led to isolation of five known heterocyclic alkaloids from normal Czapek medium, while a new meroterpenoid (**108**) and four known analogs were obtained from Czapek medium with 3% salty (Zhang et al., 2018a,b). One mangrove-derived endophyte Wallemia sebi PXP-89 cultivated in 10% NaCl broth produced a new cyclopentanol pyridine alkaloid (**109**), which was not detected in normal medium (Peng et al., 2011). When cultivated in medium containing 10% sea salt, strain Spicaria elegans KLA-03 was shown to biosynthesize a new antimicrobial diacrylic acid (**110**) (Wang F. Z. et al., 2011). Strain Streptomyces sp. DSM 14386 could metabolize five new compounds (**111**–**115**) in 1.5% NaCl medium, while this strain produced two brominated congeners (**116–117**) in 1.5% NaBr medium. Antimicrobial tests showed that compounds **113** and **117** displayed potent antibiotics against MRSA (methicillin-resistant Staphylococcus aureus) with the same MIC values of 16 µg/mL, and compound **117** also had strong activity toward Mycobacterium smegmatis (IC<sup>80</sup> = 2 µg/mL) (Onaka, 2017). Two rare epidithiodiketopiperazines, gliovirin and pretrichodermamide A, were detected in 1.5% NaCl

broth of a marine-derived Trichoderma sp. TPU199, while this strain produced a new iodo derivative (**118**) from freshwater medium with 3.0% NaI and 3.0% NaBr as well as 5-bromo-5 deoxy derivative (Yamazaki et al., 2015a).

#### Metal Ion

Metal ion affects physiological structure and function of microorganism. The interaction between metal ion and microbe is usually assumed in three pathways, including causing reactions in cells, conserving energy in the process of dissimilation, and assimilating reactions (Thorneley, 1990).

One marine-derived strain Ascotricha sp. ZJ-M-5 was shown to produce a new 3,4-split ring lanolin alkyl triterpene (**119**) and a new cyclonerols derivative (**120**), when cultivated in eutrophic medium made up with sea salt (Xie et al., 2013a,b). However, three new caryophyllene derivatives (**121**–**123**) were detected in modified Czapek Dox medium, while compound **122** was absent in the fermentation broth without Mg2<sup>+</sup> (Wang W.J. et al., 2014). Strain Aspergillus sclerotiorum C10WU derived from hydrothermal vent sediment in Taiwan (China) could produce three new alkaloids (**124**–**126**) under normal medium. However, this strain metabolized one unelucidated compound due to the low amount available together with aspochracin when grown in the stressed culture medium with Cu2<sup>+</sup> as a supplement. Likewise, two compounds, namely deoxytryptoquivaline and tryptoquivaline A (**127**–**128**), were purified from the normal extract of A. clavatus C2WU, while only metabolite **129** was found in normal medium containing Cu2<sup>+</sup> and Cd3<sup>+</sup> (Jiang et al., 2014). A novel antibacterial cyclodepsipeptide, named NC-1 (**130**), was produced by a red soil-derived strain Streptomyces sp. FXJ1.172 when cultured in GYM (glucoseyeast extract-malt extract) medium added with ferric ion (Liu M. et al., 2016).

#### CULTIVATION CONDITION

Suitable cultivation conditions, such as appropriate temperature, pH, oxygen concentration, and cultivation status, are essential for the growth and biochemical reactions of microorganisms. However, many biosynthetic genes of MSMs are not expressed under normal culture conditions, thus it is essential to change the cultivation condition to activate these silent gene clusters to diversify their MSMs.

#### Temperature

Chemical diversity of MSM is directly influenced by microbe enzyme activity, which is susceptible to cultivation temperature. The normal function of microbial enzyme is dependent on appropriate temperature. Generally, the higher the cultivation temperature is, the faster the enzyme deactivation rate will be (Feller et al., 1994). For example, when the temperature was lower than 30◦C, secondary metabolites of an uncoded strain Streptomyces sp. were composed of chlortetracycline, while only tetracycline was synthesized when cultivation temperature went up to 35◦C (Cui et al., 1996).

#### pH

During microbe fermentation process, the decomposition and utilization of nutrients as well as the accumulation of secondary metabolites usually causes the variation of medium pH (Gibson et al., 1988; Tan et al., 1998). It affects not only the activity of each enzyme, but also the surface charge of the membrane. The nature and permeability of cell membrane could change the rate of utilization of substrate, thus affecting the growth of microorganisms and biosynthesis of final products. Chemical study of one desert-derived strain Nocardiopsis alkaliphila nov. YIM-80379 led to isolation of two new pyran-2-one derivatives (**131**–**132**) when cultivated on Gause's synthetic agar slants with pH = 10. However, the neutral medium was unsuitable for its growth (Hozzein et al., 2004; Wang et al., 2013c). Acidic medium (pH = 5) dramatically increased the production of bioactive compounds of a mangrovederived fungus Rhytidhysteron rufulum AS21B, including two new antitumor spirobisnaphthalenes (**133**–**134**). However, these compounds were not detected in neutral medium (Siridechakorn et al., 2017).

#### Oxygen Concentration

fmicb-10-00294 February 23, 2019 Time: 16:42 # 6

Changes in oxygen supply can affect the biochemical reactions and activate different set of functional gene clusters for different secondary metabolites production (Sato, 1990). For example, [ <sup>13</sup>C]-labeled acetates and a small amount of [18O2] were used to investigate the biosynthetic pathway of aspinonene (**135**) in the culture broth of Aspergillus ochraceus DSM-7428. It is interesting that aspyrone (**136**) was produced by increasing dissolved oxygen concentration during fermentation, accompanied by reduced amounts of compound **135** under an oxygen enriched atmosphere (Fuchser et al., 1995).

### Cultivation Status

A growing body of evidence has indicated that cultivation status can directly affect microbe metabolic process, including solid or liquid, static or dynamic. Compared with solid and static cultivation, liquid and dynamic modes not only ensure the full contact of microorganisms and nutrients, but also affect their biochemical reactions by changing oxygen supply and activating functional gene clusters. Till now, MSMs from 12 genera had been investigated under different fermentation status, including Arthrinium, Aspergillus, Myxotrichum, Nodulisporium, Lentinus, Paraphaeosphaeria, Penicillium, Pestalotiopsis, Phomopsis, Spicaria, Streptomyces, Ulocladium.

#### Arthrinium

One marine sponge-derived fungus Arthrinium arundinis ZSDS1-F was shown to metabolize a novel naphthalene glycoside (**137**) (Wang J.F. et al., 2014), five cytochalasins (**138**–**142**) (Wang et al., 2015b), and three alkaloids (**143**–**145**) when cultivated in a rotary liquid medium (Wang et al., 2015c). However, only phenethyl 5-hydroxy-4-oxohexanoate (**146**) was traced in rice medium (Li Y. L. et al., 2017). Bioassay suggested that compounds **143**–**146** possessed in vitro cytotoxicity against cancer cell lines A549, BGC823, Huh-7, K562, H1975, MCF-7, HL60, U937, Hela, and MOLT-4 with IC<sup>50</sup> values in range of 0.24–45 µM. In addition, compounds **143** and **145** displayed significant AchE (acetylcholine esterase) inhibitory activity with IC<sup>50</sup> values of 47 and 0.81 µM, respectively.

#### Aspergillus

By comparison of solid and liquid fermentation products of an endophytic strain A. fumigates LN-4 from stem bark of Melia azedarach L., their HPLC profiles were obviously different (Zhang et al., 2013). Strain A. versicolor ZLN-60 could produce two new cyclic pentapeptides (**147**–**148**) and four new prenylated diphenyl ethers (**149**–**152**) in static liquid condition (Zhou et al., 2011; Gao et al., 2013). Biological tests indicated that compound **151** displayed moderate cytotoxicity against Hela and K562 cancer cell lines with IC<sup>50</sup> values of 31.5, 48.9 µM, respectively. However, further purification of its crude extract of solid medium resulted in the detection of four other novel cyclic peptides (**153**–**157**) (Peng et al., 2014). Chemical study of one marine-derived fungus A. terreus cultivated in 11 different culture conditions indicated that static agar was ideal for the production of antifungal lovastatins (**158**–**159**) and 7-desmethylcitreoviridin (**160**), which were absent in the shaking fermentation (Adpressa and Loesgen, 2016).

#### Lentinus

Two new prenyl phenols (**161–162**), one indole alkaloid echinuline (**163**) and one anthraquinone fiscione (**164**), were biosynthesized by Lentinus strigellus under static condition. While in shaking fermentation broth, this strain produced benzopyrans (**165**–**168)** together with panepoxydone (**169**) and isopanepoxydone (**170**). Bioassay indicated that striguellone A (**171**) displayed moderate cytotoxicity against HeLa cancer cells (Zheng et al., 2009; Barros et al., 2012).

#### Myxotrichum

One fungal strain Myxotrichum sp. isolated from lichen Cetraria islandica (L.) Ach in Laojun Mountain (China), was shown to

make one novel austdiol analog (**172**), three new fulvic acid derivatives (**173**–**175**) and a new citromycetin analog (**176**) in rotary PDB medium (Yuan et al., 2013), while four new polyketides (**177**–**180**) were acquired from rice medium under static fermentation status. And compound **179** was shown to restrain Arabidopsis seeds root markedly with the inhibition rate of 75.9% at 8 µg/mL (Yuan et al., 2016).

#### Nodulisporium

Chemical investigation of one symbiotic strain Nodulisporium sp. (No. 65-12-7-1) from the lichen Everniastrum sp. resulted in the isolation of two rarely 4-methyl-progesteroids (**181**–**182**) when grown in rice medium (Zheng et al., 2013). Whereas this strain could biosynthesize ten novel nodulisporisteroids (**183**–**192**) in shaking PDB medium (Zhao et al., 2015).

#### Paraphaeosphaeria

A fungal strain Paraphaeosphaeria photiniae, inhabiting Roystonea regia collected from Jianfeng Mountain (China), was shown to yield six new unique benzofuranone-derived γ-lactones (**193**–**198**) when cultivated in shaking liquid medium (Ding et al., 2009), while only two different δ-lactone derivatives (**199**–**200**) were detected in its rice medium (Ding et al., 2012).

#### Penicillium

When grown on solid PDA medium, one mangrove-derived fungus Penicillium brocae MA-231 could produce six new disulfide-bridged diketopiperazine derivatives (**201**–**206**). Bioassay results showed that compounds **201**, **202**, **205,** and **206** had cytotoxic activities against Du145, Hela, HepG2, MCF-7, NCI-H460, SGC-7901, SW1990, SW480, and U251 tumor cell lines with IC<sup>50</sup> values ranging from 0.89 to 9.0 µM (Meng et al., 2014). When cultivated in liquid media (PDB or Czapek), however, five new penicibrocazines (**207**–**211**), four new thiodiketopiperazine alkaloids (**212**–**215**) and two new N-containing p-hydroxyphenopyrrozin derivatives (**216**–**217**) were detected in its fresh mycelia, which compounds **207**–**209** displayed antimicrobial activities against Staphylococcus aureus with MIC values of 32.0, 0.25, 8.0 µg/mL, respectively. In addition, **209**–**211** exhibited potent antimicrobial effect on Gaeumannomyces graminis with MIC values of 0.25, 8.0 and 0.25 µg/mL, respectively. And compound **216** showed powerful inhibitory effect on Fusarium oxysporum and S. aureus (Meng et al., 2015b, 2017).

Chemical study of one marine sponge-derived strain P. adametzioides AS-53 led to isolation of two new bisthiodiketopiperazine derivatives (**218**–**219**) from shaking PDB broth, whereas two new acorane sesquiterpenes (**220**–**221**) were found in its static rice medium. Compound **218** showed strong lethality against brine shrimp (Artemia salina) with an LD<sup>50</sup> value of 4.8 µM and a broad spectrum of antimicrobial effect on Aeromonas hydrophilia, S. aureus, Vibrio spp. V. harveyi, Gaeumannomyces graminis and V. parahaemolyticus (Liu Y. et al., 2015). Six novel azaphilone derivatives (**222**–**227**) as major secondary metabolites were obtained from rotary PDB medium of one marine-derived strain P. commune QSD-17 (Gao et al., 2011), whereas other new compounds isophomenone (**228**) and 3-deacetylcitreohybridonol (**229**) were detected in its static rice medium (Gao et al., 2012).

Three novel penipanoids (**230**–**232**) were characterized from one marine-derived strain P. paneum SD-44 grown in rice medium (Li et al., 2011). The exploration of changing fermentation conditions of P. paneum SD-44 to a seawater-based culture broth under dynamic fermentation condition gave five

new anthranilic acid derivatives (**233**–**237**). Metabolites **233** and **237** exhibited inhibitory activity toward human colon cancer RKO cell lines with IC<sup>50</sup> values of 8.4, 9.7 µM, respectively (Li C. S. et al., 2013). One deep sea-derived fungus Penicillium sp. F23-2 biosynthesized terpenoids, diketopiperazines, and meleagrin alkaloids when incubated in sea-water-based culture medium under static condition (Du et al., 2009, 2010), whereas five new nitrogen-containing sorbicillinoids (**238**–**242**) were metabolized by this stain when cultivated in PYG (peptone yeast glucose) medium under shaking status (Guo et al., 2013).

#### Pestalotiopsis

When grown in rice medium, one endophytic strain of Pestalotiopsis fici from Camellia sinensis was found to be a prolific producer of bioactive secondary metabolites, including pupukeanane chloride (**243**) (Liu et al., 2008a), chloropestolide A (**244**) (Liu et al., 2009a), seven isoprenylated chromones (**245**–**251**) (Liu et al., 2010), three highly functionalized compounds (**252**–**254**) (Liu et al., 2009b), and three cytotoxic pupukeanane chlorides (**255**–**257**) (Liu et al., 2011). In vitro cytotoxic assays suggested that compound **244** possessed potent inhibitory effects on HeLa and HT29 with GI<sup>50</sup> values of 0.7, 4.2 µM, respectively. However, this strain produced new cyclopropane derivatives (**258**–**262**) when cultivated in shaking liquid medium (Liu et al., 2008b). An endophytic fungus P. foedan, residing in Bruguiera sexangul, synthesized a new reduced spiro azaphilone derivative (**263**) together with two new isobenzofuranones (**264**–**265**) in solid GYM (glucose, yeast extract, malt) medium (Ding et al., 2008). But, in liquid modified PDB medium, a pair of novel spiro-γ-lactone enantiomers (**266**–**267**) were identified (Yang and Li, 2013).

#### Phomopsis

An endophytic fungus Phomopsis sp. sh917 isolated from fresh stems of Isodon eriocalyx var. laxiflora collected in Kunming Botanical Garden of China, was shown to produce six new polyketides (**268**–**273**) on solid rice medium but metabolize a new polyketide (**274**) in shaking liquid FM4 medium (Tang et al., 2017).

#### Spicaria

Nine new cytochalasins Z7-Z15 (**275**–**283**), one novel spicochalasin (**284**), five new aspochalasins (**285**–**289**), and three new aspochalasin derivatives (**290**–**292**) were synthesized by a marine-derived fungus Spicaria elegans KLA03 in the seawater-based medium under static fermentation status. Compounds **235** and **276** displayed strong cytotoxicity against P388 and A-549 cancer cell lines with IC<sup>50</sup> values in range of 8.4–99 µM (Liu et al., 2005, 2006, 2008c; Lin et al., 2009a, 2010). However, new aromatic polyketide (**293**) was obtained from shaking seawater medium (Luan et al., 2014).

#### Streptomyces

One marine-derived stain Streptomyces sp. CHQ-64 was found to produce six new antifungal polyene-polyols (**294–299**) and two new cytotoxic hybrid isoprenoid alkaloids (**300–301**) in liquid medium under shaking condition, while this strain made only one new hybrid isoprenoid alkaloid (**302**) under static condition (Che et al., 2012, 2013, 2015, 2016). When cultivated in liquid Gause's No. 1 medium, strain Streptomyces sp. DT-A37 could produce a new ring-opened lactam (**303**), while in rice medium one unknown holomycin (**304**) and two new cyclopropaneacetic acids (**305**–**306)** were detected (Ding et al., 2017). Strain Streptomyces sp. HZP-2216E cultured in 2216E solid medium, GYM solid medium and GMSS (Gause's medium with sea salt) liquid medium resulted in isolation of two new compounds of 23-O-butyrylbafilomycin D (**307**), streptoarylpyrazinone A (**308**) a unique indolizinium alkaloid streptopertusacin A (**309**). It was noted that compound **307** showed potent activity in suppressing the proliferation of the four tested glioma cell lines with IC<sup>50</sup> values in a range from 0.35 to 2.95 µM and antibacterial activity with MIC value of 7.4 µM for MRSA and IC<sup>50</sup> values of 0.44 to 0.98 µM for glioma cells (Zhang et al., 2017c,d).

#### Ulocladium

fmicb-10-00294 February 23, 2019 Time: 16:42 # 9

Two antifungal polyketides (**310**–**311**) were characterized from rice medium of Ulocladium sp. that was isolated from the lichen Everniastrum sp. (Wang X.E. et al., 2012), whereas three new tricycloalternarenes F-H (**312**–**314**) and five ophiobolane sesterterpenes (**315**–**319**) were detected in liquid Czapek or PDB medium (Wang et al., 2013a,b). Compounds **315** and **319** exhibited moderate antibacterial activity against meticillinresistant S. aureus and Bacillus subtilis and displayed strong in vitro cytotoxicity against cancer cell lines KB and HepG2.

## CO-CULTIVATION WITH OTHER STRAIN(S)

In one culture medium, the relationship between one strain and other(s) may be competitive, antagonistic or friendly. Co-cultivation of two or more strains usually has positive effect of an enhanced production of known compounds or an accumulation of cryptic compounds that are not detected in axenic culture (Bohni et al., 2013; Marmann et al., 2014). This effect maybe results from the production of enzymes that activate metabolite precursors or that other strain(s) may induce epigenetic modifications of the producer strain.

# Fungus and Other Fungal Strain

An endophytic strain Acremonium sp. Tbp-5 from the European yew (Taxus baccata L.) could produce new lipoaminopeptides (**320**–**322**) when co-cultivated with Mycogone rosea DSM 12973 (Degenkolb et al., 2002). Chemical investigation of the mixed fermentation broth of two epiphytic strains Aspergillus sp. FSY-01 and FSW-02 from marine mangrove Avicennia marina led to the isolation of a novel alkaloid (**323**), which had antibacterial activity against Bacillus dysenteriae, B. proteus, and E. coli (Zhu et al., 2011). The production of 2-alkenyl-tetrahydropyran analogs (**224**–**326**) was provoked by Chaunopycnis sp. CMB-MF028 in the mixed culture with a partner strain Trichoderma hamatum CMB-MF030, which were isolated from the inner tissue of marine pulmonate false limpet (Shang et al., 2017). Co-cultivation of Monascus sp. J101, used as the producer of Monascus pigment, with Saccharomyces cerevisiae KCCM 11371 or A. oryzae KCCM 11773 on the solid sucrose medium could result in two folds of accelerated cell growth and 30–40 folds of increased pigment production (Shin et al., 1998). Strain J101 was shown to stimulate cell growth and reproduction by interacting with S. cerevisiae, which resulted in production of more hydrophobic pigments compared to the mono-culture (Suh and Shin, 2000a,b). When co-cultivated with Beauveria felina, one marine-derived P. citrinum could biosynthesize two new compounds (**327**–**328**) featuring in a unique tetracyclic framework, whereas neither strain could produce these compounds in axenic medium. Antimicrobial assay showed that compounds **327** and **328** had strong inhibitory effects on human pathogens S. aureus and E. coli (Meng et al., 2015a). Penicillium sp. IO1 derived from mediterranean sponge Ircinia oros could produce a new fusarielin analog (**329**). However, co-cultivation of Penicillium strains IO1 and IO2 resulted in the accumulation of two known compounds norlichexanthone and monocerin, which were not detected in axenic controls (Chen et al., 2015a). Four new polyketides (**330**–**333**) were detected in a dual culture of the deep-sea-derived fungus Talaromyces aculeatus and a mangrove-derived fungus

P. variabile, while these compounds were not identified in single culture. Compounds **333** displayed strong cytotoxicity against A549, K562, HCT-116, HeLa, MCF-7 and HL-60 human cancer cell lines with IC<sup>50</sup> values ranging from 1.2 to 9.8 µM (Zhang et al., 2017b).

One novel 1-isoquinolone analog (**334**) and its methyl ester (**335**) were detected in mycelia and culture filtrate of mixed fermentation of two endophytic fungi Nos. 1924# and 3893# , whereas these compounds were not traced in axenic medium under the same conditions (Zhu and Lin, 2006). The formation of new antibiotics (**336**–**337**) was emerged during co-cultivation of a multi-antibiotic stable mutant strain of Rhodococcus fascians and a strain Streptomyces padanus, neither of which was capable of yielding an antibiotic (Kurosawa et al., 2008). An terrestrial bacterium Tsukamurella pulmonis TP-B0596 co-cultured with strain Streptomyces sp. NZ-6, coincided with stimulation of three new metabolites (**338**–**340**) of unprecedented di-andtricyclic macrolactams (Hoshino et al., 2015). The yield of red pigment was detected in the dual induction of T. pulmonis TP-B0596 and S. lividans TK23. Co-cultivation of T. pulmonis and S. endus S-522 resulted in the production of one new antibiotics called alchivemycin A (**341**) (Onaka et al., 2011). A soil-dwelling actinomycetes S. coelicolor was shown to significantly improve the yield of red compound undecylprodigiosin, when co-cultured with Corallococcus coralloides (Schäberle et al., 2014).

#### Bacterium and Other Bacterial Strain

Only two new macrolactams (**342**–**343**) were detected in the co-culture broth of a rare actinomycete Micromonospora wenchangensis HEK-797 and Tsukamurella pulmonis TPB0596, whereas the axenic medium of strain HEK-797 produced a polyene macrolactam (**344**), which was possibly the precursor of compounds **342** and **343** (Hoshino et al., 2017). Investigation of the interaction of the portable predator Myxococcus Xanthus and Streptomyces coelicolor showed that actinorhodin production of S. coelicolor was raised up to 20-fold and stimulated aerial mycelium production (Pérez et al., 2011). Co-cultivation of two sponge-derived actinomycetes, Nocardiopsis sp. RV163 and Actinokineospora sp. EG49, induced ten reported compounds, including diketopiperazine, angucycline, and β-carboline derivatives, while only three natural products were isolated in mono-culture (Dashti et al., 2014). Mixed culture of Pseudomonas maltophilia 1928 and S. griseorubiginosus 43708 resulted in the production of one peptide antibiotic, biphenomycin A (**345**) (Ezaki et al., 1992). However, the accumulation of biphenomycin A, which could be obtained from the transformation of biphenomycin C (**346**), was inhibited in single culture of strain 1928 (Uchida et al., 1985; Ezaki et al., 1993). Interspecies interactions between Streptomyces coelicolor M145 with other actinomycete stains (Amycolatopsis sp. AA4, Streptomyces sp. E14, Streptomyces sp., SPB74 and S. viridochromogenes DSM 40736) resulted in the production of at least 12 different versions of a molecule called desferrioxamine (Traxler et al., 2013).

#### Fungus and Bacterium

Co-cultivation of one fungal strain A. terreus with B. cereus and B. subtilis resulted in the yield of two novel butyrolactones (**347**–**348**), which were absent in single culture medium (Chen et al., 2015b). An endophyte Chaetomium sp. from the Cameroonian plant Sapium Ellipticum (Euphorbiaceae) was shown to produce two novel shikimic acid analogs (**349–350**) and four new butenolide derivatives (**351–354**) when co-cultivated with Pseudomonas aeruginosa, while none of these chemicals was traced in axenic medium (Ancheeva et al., 2017). Strain Bacillus subtilis 168 trpC2 was shown to greatly activate the biosynthesis of three novel chemicals (**355–357**) of fungal endophyte Fusarium tricinctum during co-culture process. And these compounds were not duplicated in axenic fungal culture (Ola et al., 2013).

Co-cultivation of one marine fungus Libertella sp. CNL-523 symbiotic on an ascidian collected from the Bahamas and a fellow strain Thalassospira sp. CNJ-328 resulted in the production of four new diterpenoids (**358**–**361**). Compound **360** exhibited remarkable cytotoxicity against HCT-116 human adenocarcinoma cell line with an IC<sup>50</sup> value of 0.76 µM (Oh et al., 2005). A new pyridone alkaloid (**362**) was isolated from the mixed culture extract of Paecilomyces lilacinus and Salmonella typhimurium, which had 57.5 ± 5.50% of AChE inhibition (Teles and Takahashi, 2013). Co-culture of an endophyte Pestalotiopsis sp. from Drepanocarpus lunatus with B. subtilis was found to biosynthesize two novel sesquiterpenoids (**363**–**364**) while new compounds **365** and **366** emerged in axenic culture (Liu et al., 2017). The mixed cultivation of Trichoderma sp. 307 colonizing in Clerodendrum inerme and one bacterium Acinetobacter johnsonii B2 led to the production of two new sesquiterpenes (**367**–**368**) and three novel de-O-methyl lasiodiplodins (**369**–**371**). Compounds **369** and **370** displayed potent α-glucosidase inhibitory effect with IC<sup>50</sup> values of 25.8 and 54.6 µM, respectively (Zhang et al., 2017a).

Chemical study of an endophytic stain Aspergillus austroafricanus from Eichhornia crassipes led to the isolation of a highly oxygenated heterodimeric xanthone (**372)** and a new sesquiterpene (**373**) in axenic culture. Mixed fermentation of A. austroafricanus with B. subtilis or S. lividans afforded several

diphenyl ethers, including one new austramide (**374**) (Ebrahim et al., 2016). Two novel N-formyl alkaloids (**375–376**) were characterized from a mixed fermentation of A. fumigatus and S. peucetius. Compound **376** displayed in vitro cytotoxic effect on cancer cell line NCI-60 with an IC<sup>50</sup> value of 1.12 µM (Zuck et al., 2011). Seven known diketopiperazine alkaloids associated with ergosterol and 11-O-methylpseurotin A were traced in response to the supplement of A. fumigatus MBC-F1-10 to an established culture of S. bullii, whereas neither strain metabolized these compounds when cultivated alone (Rateb et al., 2013). Co-cultivation of A. fumigatus MR2012 with S. leeuwenhoekii C34 in ISP2 medium resulted in the yield of a new luteoride derivative (**377**) and a new pseurotin derivative (**378**). None of these compounds could be detected in axenic culture. When strain MR2012 was co-cultivated with strain C58, a lasso peptide chaxapeptin (**379**) was made, which displayed significant inhibitory effect on human lung cancer cell line A549 (Elsayed et al., 2015; Wakefield et al., 2017).

Physical interaction of A. nidulans RMS011 with S. hygroscopicus was found to trigger biosynthesis of four new aromatic polyketides (**380**–**383**), which were absent in the axenic medium (Schroeckh et al., 2009). A new polyketide glycoside (**384**) was formed in the dual induction of two Gram-positive bacteria, S. tendae KMC006 and Gordonia sp. KMC005, which were obtained from an acidic mine drainage sample (Park et al., 2017). In response to S. coelicolor A3(2) M145, strain A. niger N402 was shown to be apt to produce 2-hydroxyphenylacetic acid and cyclic dipeptide cyclo(Phe–Phe). Biotransformation of a new hexadienedioxic acid (**385)** and a new phenol derivative (**386**) was achieved by co-culture of these strains (Wu et al., 2015). More interestingly, co-cultivation of one marinederived fungus Emericella sp. CNL-878 with Salinispora arenicola CNH-665 resulted in the higher yields of two novel antimicrobial cyclic depsipeptides (**387**–**388**) than axenic culture (Oh et al., 2007).

#### EPIGENETIC MODIFIER

Epigenetic modifiers are those chemicals that are able to change microbial characteristics in correspondence to alteration of their epigenetic status, such as DNA methyltransferase (DNMT) inhibitor and histone deacetylase (HDAC) inhibitor. The addition of these modifiers usually suppresses the activity of related enzymes in the biosynthetic pathway and promotes the progress of other metabolic pathways (Seyedsayamdost, 2014).

#### DNA Methyltransferase Inhibitor

DNA methylation is a process by which methyl groups are added to DNA. When located in a gene promoter, DNA methylation typically acts to repress gene transcription and causes chromatin structure changes in the corresponding regions, preventing the binding of specific transcription factors and suppressing gene expression (Araujo et al., 2001). 5-Azacytidine (5-AC) is the most common DNMT inhibitor used to modify the function of microbe DNA followed by repressing gene transcription. Chemical investigation of a marine-derived fungus Aspergillus sydowii afforded three novel bisabolane-type sesquiterpenoids (**389**–**391**) when its culture medium was supplemented with 5-AC (Chung et al., 2013b). An entomopathogenic fungus Cordyceps indigotica yielded a novel aromatic polyketide

glycoside (**392**) when cultivated in PDB media, while the strain produced another unusual glycoside (**393**) when supplement with 5-AC (Asai et al., 2012e). Several other examples that adding 5-AC as epigenetic modifier in culture medium could lead to the production of new metabolites were also reported, such as novel diethylene glycol phthalate esters (**394**–**400**) from a marine-derived strain Cochliobolus lunatus TA26-46 (Chen et al., 2016), a new benzoic acid (**401**) from Pestalotiopsis microspora (Yang et al., 2017), one new coumarin (**402**) from P. crassiuscula NBRC 31055 associated with Fragaria chiloensis (Yang et al., 2014), and novel meroterpenes (**403**–**404**) from P. citreonigrum (Wang et al., 2010).

#### Histone Deacetylase Inhibitor

The acetylation or deacetylation of histone affects its binding to DNA in microbe. There are many chemical modifications in the tail of histone that regulate the gene expression. The introduction of hydrophobic acetyl group into the N-terminal lysine residues of histone could increase the electrostatic attraction and steric hindrance between histone and DNA, which is conducive to facilitate the depolymerization of DNA and the binding of transcription factors (Fukuda et al., 2006; Cole, 2008). Suberoyl bishydroxamic acid (SBHA), suberoylanilide hydroxamic acid (SAHA), and nicotinamide are the most common HDAC chemicals used to inhibit the deacetylation and facilitate gene transcription and expression in microbes (Moore et al., 2012).

Many reports suggested the presence of SAHA in culture medium could result in production of new natural compounds, such as a novel metabolite nygerone A (**405**) from a soildwelling fungus A. niger ATCC 1015 (Henrikson et al., 2009), two new aromatic norditerpenes (**406**–**407**) tied with an oxygenated derivative (**408**) from a marine-derived A. wentii na-3 residing in the brown alga Sargassum fusiforme (Miao et al., 2014), three novel cyclodepsipeptides (**409**–**411**) from Beauveria feline (Chung et al., 2013a), one novel chlorinated polyketide (**412**) from Daldinia sp. (Du et al., 2014), a new cyclodepsipeptide of hybrid EGM-556 (**413**) from one marine sediment-derived fungus Microascus sp. (Vervoort et al., 2011).

In SBHA-treated culture medium, Chaetomium indicum could produce two novel spironolactone polyketides (**414**–**415**) and six novel prenylated aromatic polyketides (**416**–**421**) (Asai et al., 2013b,c). Similarly, when exposed to SBHA, four new 2,3-dihydrobenzofurans (**422–425**) and a new aromatic polyketide (**426**) were characterized from an entomopathogenic fungus Cordyceps annullata (Asai et al., 2012c), six new aromatic polyketides (**427**–**432**) were synthesized by C. indigotica (Asai et al., 2012f), two new fusaric acid derivatives (**433**–**434**) were produced by Fusarium oxysporum associated with medicinal plant Datura stramonium L. (Chen et al., 2013), and a series of novel prenylated tryptophan analogs (**435**–**437**) were metabolized by an entomopathogenic fungus Torrubiella luteorostrata (Asai et al., 2011). Supplement of nicotinamide [a Zn(II)-type HDAC inhibitor] in culture medium of C. cancroideum could generate three novel polyketides (**438**–**440**) (Asai et al., 2016). The use of this inhibitor to strains Eupenicillium sp. LG41 and Graphiopsis chlorocephala had the similar effect, which the former supplied two new decalin-containing compounds (**441**–**442**) (Li G. et al., 2017) and the later afforded a serious of new benzophenones (**443**–**444**) and diverse new C13-polyketides (**445**–**453**) (Asai et al., 2012d, 2013a).

#### Multiple Chemical Epigenetic Modifiers

Interactions between epigenetic features play an important role in regulation of gene expressing or silencing in microorganisms, such as DNA methylation and histone modification. Many references that looked into the combined effect of epigenetic processes suggested that these chemicals could regulate the activity of genomic regions of varying sizes, from single genes to entire domains and chromosomes. Epigenetic markers could also interact with other nuclear proteins to work together to form chromatin structures and to create genomic functional discrete

regions that induce the production of new secondary metabolites (Tammen et al., 2013).

One symbiotic strain Alternaria sp. from medicinal plant Datura stramonium Linn. was shown to produce four new aromatic polyketides (**454**–**457**) and a new tenuazonic acid (**458**) when incubated in medium containing 5-AC and/or SBHA. While these compounds were absent in normal culture medium. Interestingly, the yield of these secondary metabolites was higher in the medium of adding HDAC and DNMT inhibitors than that of addition of any other inhibitors (Sun et al., 2012). Chemical investigation of one marine-derived fungus Aspergillus sp. SCSIOW2 or SCSIOW3, exposed with an integration of SHBA and 5-AC, led to production of three new eremophilane-type sesquiterpenes (**459**–**461**) together with a new diphenylether-Oglycoside (**462**) (Wang L. et al., 2016; Li X. et al., 2017). Bioactivity tests indicated that the glycosylated compound **462** exhibited a protective activity toward free radicals with an EC<sup>50</sup> value of 20.8 µM. One strain Cladosporium cladosporioides from a tidal pool was found to display different responses to the treatment with 5- AC and SHBA. Exposure of C. cladosporioides to 5-AC resulted in substantially increased biosynthesis of three oxylipins (**463**–**465**), whereas SHBA induced the yield of two new perylenequinones (**466**–**467**) (Williams et al., 2008).

Concomitant supplement of SHBA and N-phthalyl-Ltryptophan (DNMT inhibitor) to the fermentation medium of an entomopathogenic fungus Gibellula formosana induced the formation of two new highly oxidized ergosterols (**468**–**469**) and five new isariotin analogs (**470**–**474**) (Asai et al., 2012a). The same method was applied to expand MSM profile of Isaria tenuipes, which resulted in the yield of one new polyketide (**475**) (Asai et al., 2012b). An endophytic strain Leucostoma persoonii from red mangrove was subject to large-scale cultivation with sodium butyrate (HDAC inhibitor) and 5-AC, which resulted in the increased yield of known cytosporones and the production

of one new cytosporone (**476**) (Beau et al., 2012). Three novel aromatics (**477**–**479**) were produced by Pestalotiopsis acacia from Taxus brevifolia when its culture medium was supplemented with SHBA and 5-AC (Yang and Li, 2013). Application of this approach also led to production of a new glycolipid ustilagic acid C (**480**) by Ustilago maydis (Yang et al., 2013).

#### OTHER FACTORS

#### Enzyme Inhibitor

Beside DNMT and HDAC, other microbial enzymes also played important role in the regulating the biosynthesis of secondary metabolites, such as monooxygenase and hydroxylase. Some chemicals can selectively inhibit the activity of these enzymes in the biosynthetic pathway and promote the progress of other metabolic pathways, such as metyrapone, tricyclazole, jasplakinolide, and DMSO.

Chemical study of Chaetomium subaffine in the presence of metyrapone (an inhibitor of cytochrome P-450) led to purification of five new polyketides (**481**–**485**) and two new less oxidized analogs (**486**–**487**) (Oikawa et al., 1992). A soilderived strain Phoma sp. SNF-1778 was shown to yield a new cytochalasin (**488**) when inoculated with metyrapone (Kakeya et al., 1997). When added with the F-actin inhibitor jasplakinolide in culture medium, one marine sponge-derived fungus Phomopsis asparagi could afford three unusual cytotoxic compounds, chaetoglobosin-510 (**489**), chaetoglobosin-540 (**490**), and chaetoglobosin-542 (**491**) (Christian et al., 2005). Two novel bisnaphthalene compounds (**492**–**493**) were characterized from Sphaeropsidales sp. F-24'707 cultured with tricyclazole, which was shown to inhibited the regular biosynthesis of 1,8 dihydroxynaphthalene (Bode and Zeeck, 2000). Continuous study showed that metyrapone supplementation in the culture of Spicaria elegans led to the isolation of two novel 7-deoxy-cytochalasins (**494**–**495**). Compound **494** had weak

cytotoxicity against human lung cancer cell line A-549 at 15.0 mM (Lin et al., 2009b). One marine-derived strain Trichoderma cf. brevicompactum elicited an unprecedented epidiketopiperazine (**496**), which has a trisulfide bond between the α–β positions of two amino acid residues, by adding DMSO to its natural seawater medium (Yamazaki et al., 2015b).

#### Biosynthetic Precursor

fmicb-10-00294 February 23, 2019 Time: 16:42 # 14

Biosynthetic precursor refers to one chemical that is apt to be directly incorporated into the final product. Adding various biosynthetic precursors in the fermentation medium may change biosynthesis pathways of secondary metabolites and result in the production of novel compounds (Ramm et al., 2017).

An endophytic strain Penicillium crustosum from the ripe berry of Coffea arabica L., treated with ferulic acid and quinic acid or cinnamic acid and 3,4-(methylenedioxy) cinnamic acid, was shown to produce mycophenolic acid (**497**) and 5-hydroxy-7-methoxy-4-methylphthalide (**498**) (Valente et al., 2013). Three novel cytochalasins Z21–Z<sup>23</sup> (**499**–**501**) were characterized from one marine-derived fungus Chaetomium indicum KLA03 when cultivated in medium supplied by L- and D-tryptophan. Compound **498** exhibited potent cytotoxic effect on A549 cell lines with an IC<sup>50</sup> value of 8.2 µM (Wang F. Z. et al., 2011). Strain S. griseoviridis Tü 3634 could afford a wide variety of acyl α-L-rhamnopyranosides (pyrrolyl, indolyl, thienyl, furanyl, and pyridyl derivatives) if its culture medium, respectively, added corresponding precursors, heteroaromatic carboxylic acid, benzoic acid, cinnamic acid, aminobenzoic acid, and salicylic acid (Grond et al., 2000, 2002).

# CONCLUSION

Microorganisms are susceptible to culture conditions, such as medium composition, temperature, pH, salinity, culture status, axenic or mixed culture, epigenetic modifier, biosynthetic precursor, and so on. Variation of these factors may result

#### REFERENCES


in changing chemical diversity of secondary metabolites. Traditional culture method of microbe is limited to the expression of a large number of metabolic pathways that many MSMs could not be biologically synthesized. A growing body of evidence has suggested that OSMAC strategy can provide a simple, quick and effective approach for enhancing chemodiversity of MSM to obtain new drug leads through activating silent gene clusters. Moreover, employment of this strategy could avoids the waste of time and resources caused by multiple inoculation, screening, culturing and separation in comparison with mutation strategy (Fang et al., 2014) and ribosome engineering (Ochi et al., 2004). Nowadays, the rate of discovery of new MSM is getting lower and the possibility of the rediscovery of known compounds is higher than before. Therefore, OSMAC strategy would be an important alternative way to alleviate this challenge. There is a great need for new method to assist in isolating and identifying novel bioactive MSMs, such as bioassay-guided isolation, microbe genomes mining (Hug et al., 2018) and LC-MS/MS based molecular networking analysis (Wang M. et al., 2016).

# AUTHOR CONTRIBUTIONS

RP made a draft of the review. XB and JC searched and collected the references. HZ and HW conceived and revised this review.

# FUNDING

Financial supports from the National Key Research and Development Program of China (2018YFC0311004 and 2017YFE0103100), the National Natural Science Foundation of China (41776139 and 81773628), the Zhejiang Provincial Natural Science Foundation of China (LY16H300007 and LY16H300008), and the Hangzhou Science and Technology Development Program of China (20170432B02) were greatly appreciated.


Paeonia lactiflora cultivated in the presence of an NAD(+)-dependent HDAC inhibitor. Org. Lett. 15, 2058–2061. doi: 10.1021/ol400781b


fmicb-10-00294 February 23, 2019 Time: 16:42 # 15

marine-derived fungus Scopulariopsis sp. Tetrahedron 72, 2411–2419. doi: 10. 1016/j.tet.2016.03.073


Aspergillus sp. SCSIOW3, cultivated in the presence of epigenetic modifying agents. Molecules 22:E1302. doi: 10.3390/molecules22081302




**Conflict of Interest Statement:** 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 Pan, Bai, Chen, Zhang and Wang. 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.

# A Review of the Microbial Production of Bioactive Natural Products and Biologics

Janette V. Pham1,2, Mariamawit A. Yilma1,2, Adriana Feliz1,2, Murtadha T. Majid1,2 , Nicholas Maffetone1,2, Jorge R. Walker1,2, Eunji Kim<sup>3</sup> , Hyo Je Cho<sup>4</sup> , Jared M. Reynolds1,2 , Myoung Chong Song<sup>3</sup> , Sung Ryeol Park1,2,5 \* and Yeo Joon Yoon<sup>3</sup> \*

<sup>1</sup> Geisinger Commonwealth School of Medicine, Scranton, PA, United States, <sup>2</sup> Baruch S. Blumberg Institute, Doylestown, PA, United States, <sup>3</sup> Department of Chemistry and Nanoscience, Ewha Womans University, Seoul, South Korea, <sup>4</sup> School of Life Sciences and Biotechnology, Kyungpook National University, Daegu, South Korea, <sup>5</sup> Natural Products Discovery Institute, Doylestown, PA, United States

#### Edited by:

Dipesh Dhakal, Sun Moon University, South Korea

#### Reviewed by:

Alejandra Prieto-Davó, Universidad Nacional Autónoma de México, Mexico Yousong Ding, University of Florida, United States

#### \*Correspondence:

Sung Ryeol Park sung.park@bblumberg.org; sungryeolpark0906@gmail.com Yeo Joon Yoon joonyoon@ewha.ac.kr

#### Specialty section:

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

> Received: 18 January 2019 Accepted: 04 June 2019 Published: 20 June 2019

#### Citation:

Pham JV, Yilma MA, Feliz A, Majid MT, Maffetone N, Walker JR, Kim E, Cho HJ, Reynolds JM, Song MC, Park SR and Yoon YJ (2019) A Review of the Microbial Production of Bioactive Natural Products and Biologics. Front. Microbiol. 10:1404. doi: 10.3389/fmicb.2019.01404 A variety of organisms, such as bacteria, fungi, and plants, produce secondary metabolites, also known as natural products. Natural products have been a prolific source and an inspiration for numerous medical agents with widely divergent chemical structures and biological activities, including antimicrobial, immunosuppressive, anticancer, and anti-inflammatory activities, many of which have been developed as treatments and have potential therapeutic applications for human diseases. Aside from natural products, the recent development of recombinant DNA technology has sparked the development of a wide array of biopharmaceutical products, such as recombinant proteins, offering significant advances in treating a broad spectrum of medical illnesses and conditions. Herein, we will introduce the structures and diverse biological activities of natural products and recombinant proteins that have been exploited as valuable molecules in medicine, agriculture and insect control. In addition, we will explore past and ongoing efforts along with achievements in the development of robust and promising microorganisms as cell factories to produce biologically active molecules. Furthermore, we will review multi-disciplinary and comprehensive engineering approaches directed at improving yields of microbial production of natural products and proteins and generating novel molecules. Throughout this article, we will suggest ways in which microbial-derived biologically active molecular entities and their analogs could continue to inspire the development of new therapeutic agents in academia and industry.

Keywords: natural products, biologics, biological activity, microbial cell factories, genetic engineering, combinatorial biosynthesis, production improvement

# INTRODUCTION

Natural products originate as secondary metabolites from a myriad of sources, including terrestrial plants, animals, marine organisms, microorganisms, terrestrial vertebrates and invertebrates (Chin et al., 2006). These structurally and chemically diverse molecules act as a remarkable class of therapeutics to heal various ailments. The earliest documentation of the application of natural

products to improve human health dates back to the ancient Mesopotamia's sophisticated medicinal system from 2900 to 2600 BCE (Borchardt, 2002; Siddiqui et al., 2014). By the early 1900's, approximately 80% of all medicines were obtained from plant sources (Sneader, 1997; Siddiqui et al., 2014). The discovery of penicillin from Penicillium notatum by Alexander Fleming in 1928 marked a significant shift from plants to microorganisms as a source of natural products (Fleming, 1944). Since then, microorganism-derived compounds have been utilized in medicine, agriculture, food industry and scientific research (Sanchez et al., 2012). The early years of antibiotic research discovered streptomycin from Streptomyces griseus (Waksman et al., 1946), chloramphenicol from Streptomyces venezuelae (Duggar, 1948), chlortetracycline from Streptomyces aureofaciens (Ehrlich et al., 1947), cephalosporin C from Cephalosporium acremonium (Newton and Abraham, 1955), erythromycin from Saccharopolyspora erythraea and vancomycin from Amycolatopsis orientalis (Geraci et al., 1956). Given these historical successes, large pharmaceutical companies have continued to invest in this traditional domain (Dias et al., 2012). Currently, approximately 60% of approved small molecule medicines are related to natural products, and 69% of all antibacterial agents originate from natural products (Patridge et al., 2016; Matsumura et al., 2018). However, many natural compounds with potential as novel drug candidates occur in low concentrations in nature, often making drug discovery and development burdensome and economically impractical. Therefore, an emerging alternative solution is to express biosynthetic genes from the original producers in microbial hosts, notably bacteria and fungi (Song et al., 2014). Engineered microbes can produce appreciable amounts of scarce natural compounds, thereby facilitating the synthesis of the target novel compound and potent derivatives, as well as the validation of their activities (Matsumura et al., 2018).

The natural product sector is not the only area that has undergone substantial growth or utilizes therapeutic products generated in/from living organisms. Prokaryotic and eukaryotic microbial cells, in combination with the advancement of recombinant DNA techniques, have been responsible for an explosion of biologics. Biologics are a set of molecules whose active pharmaceutical ingredients are derived from living organisms such as animals, plants, microorganisms, human blood products, and tissue transplants that are too complex to be produced through organic synthesis (Revers and Furczon, 2010). They can be categorized into five main classes: (1) monoclonal antibodies, like trastuzumab (Herceptin <sup>R</sup> ) and rituximab (Rituxan <sup>R</sup> ); (2) blood factor derivatives, like coagulation factor VIIa (NovoSeven RT <sup>R</sup> ) and epoiten alfa (Epogen <sup>R</sup> ); (3) vaccines; (4) enzymes; and (5) recombinant proteins, such as immunomodulatory cytokines, and thrombolytic agents (Lacana et al., 2007). Since the approval of recombinant human insulin and recombinant human growth hormone as some of the first modern biopharmaceuticals, large numbers of additional biopharmaceuticals have been developed, approved, and marketed using different microbial expression systems; many more are currently in the development pipeline (Graumann and Premstaller, 2006). After the successful production of the recombinant human insulin Humulin <sup>R</sup> , Escherichia coli quickly became the prevalent expression platform in the 1980s when the biopharmaceutical sector emerged and was followed by yeast Saccharomyces cerevisiae (Sanchez-Garcia et al., 2016). Microbial cells constitute the majority of hosts employed in the production of currently approved recombinant pharmaceuticals for human treatment, mainly because of their lack of unconventional post-translational modifications, proteolytic instability, poor solubility and activation of cell stress responses (Graumann and Premstaller, 2006). This demonstrates that microbial hosts represent convenient and robust platforms for the efficient production of recombinant proteins despite some bottlenecks and obstacles.

Herein, we will summarize the biological activities and applications of a variety of natural products and biologics and review the microbial systems used to produce these pharmaceutical compounds. We will also cover past and current attempts at improving the microbial production of these biological molecules and generating new molecules using diverse engineering approaches. In addition, we will discuss the challenges of the production of natural products and biologics in microbial systems and advances that can help overcome them for drug discovery and development. Future prospects for cutting-edge developments and technological advances in microbial production of bioactive natural products and recombinant proteins as the most valuable sources of therapeutics are also discussed.

# BIOLOGICAL ACTIVITIES OF NATURAL PRODUCTS AND BIOLOGICS

Natural products have diverse biological activities relevant to human health, including antibiotic, antifungal, anticancer, immunosuppressive, anti-inflammatory, biofilm inhibitory activities, etc. In this section, we will focus on the biological activities of natural products, which can be grouped into several categories. The biological activities of microbial recombinant proteins will be also reviewed.

#### Antibiotics

Natural products are a rich source for antibiotic drug development, but the most clinically useful of these scaffolds can be classified as polyketides, non-ribosomal peptides, and aminoglycosides (Wright, 2014). Polyketides, assembled by polyketide synthases (PKS), make up one of the largest classes of chemically diverse natural products and are among the most important secondary metabolites for their applications in medicine, agriculture, and industry (Tae et al., 2007). For example, pikromycin was the first known polyketide antibiotic produced from S. venezuelae in 1950 (Vazquez, 1967; Jung et al., 2006). It has been reported that pikromycin is very potent against multi-drug resistant respiratory pathogens (Woo et al., 2014). Another remarkable polyketide antibiotic with significant clinical applications is erythromycin A (1; **Figure 1** and **Table 1**), which was first discovered in 1952 as a broad-spectrum antibiotic produced by S. erythraea (McGuire et al., 1952). This antibiotic

is prescribed to a treat wide variety of bacterial infections, such as respiratory and gastrointestinal infections, whooping cough, syphilis, and acne, especially in patients who have adverse reactions against penicillin (Cobb et al., 2013). While many natural antibiotics fail to inhibit Gram-negative organisms, tetracyclines (2; **Figure 1** and **Table 1**) are active against both Gram-positive and Gram-negative bacteria (Chopra and Roberts, 2001; Demain, 2009).

As previously mentioned, penicillin is a well-known antibiotic secondary metabolite from P. notatum and is effective against Gram-positive bacteria, which are responsible for diseases such as scarlet fever, pneumonia, gonorrhea, meningitis, and diphtheria (Fleming, 2001; Tan and Tatsumura, 2015). Penicillin belongs to non-ribosomal peptide antibiotics along with vancomycin (Fischbach and Walsh, 2006). Non-ribosomal peptides, assembled by non-ribosomal peptide synthetase (NRPS), possess bioactivity that can be exploited for therapeutic applications and are amongst the most widespread and structurally diverse secondary metabolites. Vancomycin (3; **Figure 1** and **Table 1**) is another potent non-ribosomal peptide against pathogenic bacteria, including Clostridium difficile, Listeria monocytogenes, Streptococcus pneumoniae, Staphylococcus epidermidis, and methicillin-resistant Staphylococcus aureus (MRSA) (Dasgupta, 2012).

Aminoglycosides are another class of antibiotics that act by binding to the rRNA subunit of the 30S bacterial ribosome and inhibiting protein synthesis (Moazed and Noller, 1987). Streptomycin (4; **Figure 1** and **Table 1**) produced by S. griseus is the first aminoglycoside discovered in 1944 and effective against pulmonary tuberculosis (Schatz et al., 1944). Since the discovery of streptomycin, aminoglycoside antibiotics such as kanamycin, gentamicin, sisomicin, and lividomycin have been discovered and widely used to treat infectious organisms that have developed resistance against streptomycin after prolonged use (Park et al., 2013). Despite their excellent antibacterial activity, aminoglycosides have met with resistant organisms. In order to combat antibiotic resistance to aminoglycoside antibiotics, semisynthetic aminoglycoside antibiotics were specifically tailored to shield against these enzymes (Van Bambeke et al., 2017). Semi-synthetic aminoglycoside antibiotics such as amikacin, netilmicin, dibekacin, and isepamicin are developed as a result of semi-synthetic derivatives of the natural product (Miller et al., 1976; Leggett, 2015).

Natural antimicrobials have also been important to the food industry in terms of food safety against foodborne pathogens. Microbes such as lactic acid bacteria, produce a wide range of chemicals that have been shown to inhibit the growth and development of other microbial species. Nisin A (5; **Figure 1**

TABLE 1 | Biological activities of microbial-derived natural products and biologics.


and **Table 1**), a bacteriocin produced by Lactococcus lactis, is approved to preserve food in over 50 countries and is very active against Gram-positive bacteria resistant to conventional antibiotics (Li and Vederas, 2009; Gyawali and Ibrahim, 2014). Reuterin (6; **Figure 1** and **Table 1**) from Lactobacillus reuteri has been shown to have antimicrobial activities against foodborne pathogens and spoilage organisms when evaluated in milk, dairy, and meat products (Talarico et al., 1988; Gyawali and Ibrahim, 2014).

#### Antifungal Agents

Nystatin, one of the first effective polyene antifungal agent, was obtained from Streptomyces noursei in 1950 and effective against Aspergillus species (Stanley and English, 1965). Clinically, nystatin plays a significant role as a topical antifungal agent in treating oral, gastro-intestinal, and genital candidosis (Fjærvik and Zotchev, 2005). Amphotericin B (7; **Figure 2** and **Table 1**) is also a traditional polyene antifungal product of Streptomyces nodosus utilized against life-threatening fungal infections caused by Aspergillus species, and especially effective in patients who have undergone organ transplantation, received aggressive chemotherapy or with acquired immunodeficiency syndrome (Abu-Salah, 1996; Tevyashova et al., 2013).

Recently, in a review of natural products with anti-Candida albicans activity, 71 substances of the 142 evaluated were determined to have antifungal activity under the criteria of

having minimum inhibitory concentration (MIC) values below 8 mg/mL (Zida et al., 2017). The glycolipids ieodoglucomide C (8; **Figure 2** and **Table 1**) and ieodoglycolipid were isolated from the marine bacterium Bacillus licheniformis and exhibited antifungal activities with a 21 µg/L MIC against Aspergillus niger, Rhizoctonia solani, Botrytis cinerea, Colletotrichum acutatum, and C. albicans (Tareq et al., 2015). Both ieodoglucomide C and ieodoglycolipid also exhibit good antibiotic properties against S. aureus, Bacillus subtilis, Bacillus cereus, Salmonella typhi, E. coli and Pseudomonas aeruginosa with MICs ranging from 0.01 to 0.05 µM, establishing these compounds as strong potential candidates for the development of new fungicides (Tareq et al., 2015).

#### Anticancer Agents

There are many microbe-derived anticancer agents that have been evaluated through clinical trials. For instance, the polyketide actinomycin was isolated from Streptomyces parvulus in 1940 and was the first antibiotic shown to have anticancer activity (Waksman and Woodruff, 1940; Hollstein, 1974). In particular, actinomycin D, also known as dactinomycin, is approved by FDA and has been widely used in clinical practice as an anticancer drug for treating many tumors, such as Wilms' tumor, childhood rhabdomyosarcoma, Ewing's sarcoma, and metastatic, non-seminomatous testicular cancer.

Another notable example is the therapeutic combination of the microbial product bleomycin (9; **Figure 2** and **Table 1**), the plant compound etoposide, and the synthetic agent cisplatin, which has played a significant role in increasing the cure rate for disseminated testicular cancer from 5% in 1974 to 90% in 2011 (Einhorn and Donohue, 2002; Demain and Vaishnav, 2011). Bleomycin is a glycopeptide produced by Streptoalloteichus hindustanus and has been used for squamous cell carcinomas, melanomas, sarcomas, testicular, and ovarian cancer, Hodgkin's and non-Hodgkin's lymphomas, and testis tumors as an anticancer agent (Demain and Vaishnav, 2011). Its derivative, blenoxane is also used clinically with other compounds against lymphomas, skin carcinomas, and tumors of the head, neck, and testicles (Demain and Vaishnav, 2011). The anthracyclines are also an important family of polyketides produced by Streptomyces species by iterative PKS pathways and include daunorubicin (10; **Figure 2** and **Table 1**) (Di Marco et al., 1981), epirubicin (Cersosimo and Hong, 1986), and doxorubicin (Metsä-Ketelä et al., 2008). The FDA approved the use of daunorubicin and doxorubicin for cancer therapy in the 1960s. Daunorubicin is used in the treatment of acute lymphoblastic or myeloblastic lymphoma. Meanwhile, doxorubicin is used in the treatment of breast cancer, solid tumors in children, soft tissue sarcomas, and aggressive lymphomas (Giddings and Newman, 2013).

Among numerous recent examples, rapamycin, wortmannin, and geldanamycin have been found to have antiproliferative activities during clinical use as novel chemotherapeutic agents (da Rocha et al., 2001). Rapamycin, a natural product derived from Streptomyces rapamycinicus has anticancer activity in addition to its immunosuppressive, anti-inflammatory, and antifungal activities (Kim et al., 2014). It performs antitumor activity on a tumor cell by hindering its proliferation, triggering apoptosis, and inhibiting angiogenesis (Law, 2005). Wortmannin is a fungal furanosteroid derivative of Penicillium funiculosum (Davidson et al., 2013). It has shown as an effective selective inhibitor of phosphoinositide 3-kinases (PI3Ks) and PI3K- related enzymes which are play a key role in intracellular signaling pathways (Sieber et al., 2010). A study on the proliferation and apoptosis of human breast MCF-7 cells treated with wortmannin uncovered that wortmannin shows antitumor activity by triggering apoptosis and impeding proliferation of cancer cells by suppressing PI3K/Akt signaling and NF-κB protein expression (Yun et al., 2012). Geldanamycin is a benzoquinone ansamycin antitumor compound derived from Streptomyces hygroscopicus var. geldanus (Singh et al., 2010). Geldanamycin prevents ATPase activity by binding to the heat shock protein and hindering the stability and function of oncogenic protein kinases involved in signal amplification cascade that controls proliferation and apoptosis (Singh et al., 2010). Geldanamycin and its analogs play

a key role as anticancer agent in multiple myeloma, breast, and prostate cancer (Gorska et al., 2012). Another example is epothilone, an anticancer agent produced from mycobacterium Sorangium cellulosum. It obstructs microtubule depolymerization thereby causing G2-M interphase arrest of the cell cycle (Molnar et al., 2000). There are also marine microbial natural products that have anticancer activities, such as dolastatin, which is originated from cyanobacteria of the genera Symploca and Lyngbya (Simmons et al., 2008).

#### Immunosuppressive Agents

Rapamycin (also known as sirolimus) (11; **Figure 3** and **Table 1**) and FK506 (tacrolimus) (12; **Figure 3** and **Table 1**) are microbial natural products with immunosuppressive properties. Rapamycin blocks the proliferation of most cell types in response to activation by IL-2, IL-3, platelet-derived growth factor, epidermal growth factor, and insulin (Vezina et al., 1975). Rapamycin also exhibits synergism with other immunosuppressants, such as cyclosporin, to significantly reduce kidney toxicity and acute renal allograft rejection (Yoo et al., 2017). This compound has been developed to coat coronary stents and prevent organ transplant rejection and lymphangioleiomyomatosis; it was approved by the FDA for wider use in 1999 (Mann, 2001). In addition to its immunosuppressive activity, rapamycin possesses several other biological activities, including antitumor, neuroprotective/neuroregenerative, antineoplastic, and lifespan extension activities (Law, 2005; Pan et al., 2008; Yoo et al., 2017).

FK506 is also an immunosuppressive drug and was first discovered in soil samples containing Streptomyces tsukubaensis and several other Streptomyces species (Tanaka et al., 1987). FK506 is used to minimize organ rejection and to induce immunosuppression via calcineurin inhibition and interruption of T cell activation pathway (Migita and Eguchi, 2003). It has been demonstrated to be more effective than cyclosporin and non-toxic in low doses (Demain, 2014). The discovery of its immunosuppressive activity led to its use in heart, liver, and kidney transplants with overwhelming success (Demain, 2014). Like rapamycin, FK506 possesses various biological activities, including antifungal, anti-inflammatory, neuroprotective, and neuroregenerative activities (Ban et al., 2016).

#### Anti-inflammatory Agents

Some natural products also have anti-inflammatory activities. FK506 has shown efficacy in the treatment of refractory rheumatoid arthritis, a chronic inflammatory disease (Migita and Eguchi, 2003). Rapamycin also inhibits the inflammatory response after spinal cord injury by diminishing the activation and proliferation of inflammatory cells and the expression of inflammatory cytokines, thereby reducing secondary injury in the spinal cord and providing a neuroprotective effect (Song et al., 2015). Recently, strepsesquitriol, isolated from Streptomyces sp. SCSIO 10355, has been found to have anti-inflammatory activity through the inhibition of tumor necrosis factor-α production in lipopolysaccharide-activated macrophages (Yang et al., 2013). Salinamides A and B from marine Streptomyces sp. CNB-091 also displayed potent topical anti-inflammatory activity through a phorbol ester-induced mouse ear edema assay (Trischman et al., 1994). One study evaluated 7 peptides found in the Faecalibacterium prausnitzii supernatant, all belonging to a protein named microbial anti-inflammatory molecule (Breyner et al., 2017). These peptides were able to inhibit the NF-κB pathway in vitro and showed anti-inflammatory properties in vivo in a dinitrobenzene sulfate-induced colitis model (Breyner et al., 2017).

# Biofilm Inhibitory Agents

Parasitic microorganisms adhere to solid surfaces and form layers of a complex polysaccharide matrix called a biofilm that confers resistance against antibiotics as wells as inflicts significant chronic bacterial infections (Singh et al., 2017). Analogs of 5-benzylidene-4-oxazolidinones are small molecules derived from marine natural products. These molecules inhibit 89% of biofilm formed by MRSA at 0.78 µM and disperses pre-formed biofilms at 4.7 µM (Edwards et al., 2017). A synthetic library of 2-aminoimidazole triazoles was able to successfully inhibit 94% of biofilm formation in Acinetobacter baumannii and MRSA at 100 µM (Rabin et al., 2015). Another recent example is cahuitamycins A-C (13; **Figure 3** and **Table 1**) derived from the marine bacterium Streptomyces gandocaensis. Cahuitamycins have been evaluated as inhibitors of A. baumannii biofilms and it has been found that cahuitamycin C shows half maximal inhibitory concentration (IC50) at 14.5 µM. Modifications of cahuitamycins through selective mutasynthesis have yielded cahuitamycins D and E with an increased the potency of antibiofilm activity against A. baumannii (Park et al., 2016). The FDA-approved antitumor agent actinomycin D has also significant biofilm inhibitory activity against methicillin resistant- and sensitive-strains of S. aureus (Fracchia et al., 2010; Lee et al., 2016). In addition to bacterial biofilm, fungal biofilm associated with Candida pathogens is responsible for serious C. albicans infections linked to biofilm formation on medical devices. One study showed that Lactobacillus biosurfactants displayed high anti-adhesive biofilm formation properties against C. albicans and also prevented biofilm formation of L. monocytogenes, Salmonella arizonae, E. coli, and S. aureus (Fracchia et al., 2010).

#### Others

Natural products can also act as antiparasitic agents. The avermectins (14; **Figure 3** and **Table 1**) and the derivative ivermectin have shown antiparasitic activity by significantly lowering the incidence of onchocerciasis and lymphatic filariasis (Shen, 2015). Spinosad and milbemycin also have insecticidal activity. Spinosad is a combination of spinosyn A and D, which are both produced by Saccharopolyspora spinosa and have insecticidal activity against livestock external parasites via the disruption of nicotinic acetylcholine receptors (Sanchez et al., 2012). Milbemycin is an isolated fermentation product of S. hygroscopicus subsp. aureolacrimosus that acts as an insecticide and acaricide with GABAergic activity on the post-synaptic membranes of the inhibitory motor neurons of mites and arthropods through hyperpolarization and impeding neuronal

signal transduction mechanisms (Copping and Duke, 2007). Mollemycin A 20 (15; **Figure 3** and **Table 1**) is a first-inclass glycol-hexadepsipeptide-polyketide from a Streptomyces sp. and has antibacterial properties against certain Grampositive and Gram-negative bacteria, as well as extremely potent antimalarial activity against drug sensitive and MDR Plasmodium falciparum clones (Blunt et al., 2016). Microbial natural products also function as enzyme inhibitors. Lipstatin (16; **Figure 3** and **Table 1**) is a pancreatic lipase inhibitor produced by Streptomyces toxytricini that is used to combat obesity and diabetes by interfering with the gastrointestinal absorption of fat (Weibel et al., 1987). Lipstatin contains a beta-lactone structure that is likely responsible for irreversibly binding to the active site of lipase (Sanchez et al., 2012).

#### Biological Activity of Microbial Biologics

Since Humulin <sup>R</sup> (**Figure 4A**) became the first recombinant biopharmaceutical as a treatment for diabetes (Johnson, 1983), additional FDA-approved microbial biologics have been produced by E. coli. Somatrem (Protropin <sup>R</sup> ) and somatropin (Humatrope <sup>R</sup> ) are used to treat children with growth hormone deficiency (Baeshen et al., 2015; Sanchez-Garcia et al., 2016). Another biopharmaceutical produced from E. coli is pegloticase (Krystexxa <sup>R</sup> ) for the treatment of chronic gout and interferon (IFN) α-2b (Intron <sup>R</sup> A; **Figure 4B**) for the treatment of certain types of genital warts, malignant melanoma, hairy cell leukemia, follicular lymphoma, Kaposi's sarcoma, and chronic Hepatitis B or C (Baeshen et al., 2015; Sanchez-Garcia et al., 2016). Top selling biopharmaceuticals of 2015 from microorganisms include insulin glargine (Lantus <sup>R</sup> ) derived from E. coli, which functions as an insulin analog, and the pneumococcal vaccines (Prevnar <sup>R</sup> family) derived from S. pneumoniae and Corynebacterium diphtheriae (Jozala et al., 2016; Sanchez-Garcia et al., 2016). Biopharmaceuticals are also utilized for their antitumoral properties, such as the cytokines filgrastim (Neupogen <sup>R</sup> ) and granulocyte colony stimulating factor pegfilgrastim (Neupeg <sup>R</sup> ; **Figure 4C**), which are both derived from E. coli. Flgrastim stimulates hematopoiesis for bone marrow transplantation and cancer chemotherapy-induced neutropenia, whereas pegfilgrastim stimulates the differentiation, proliferation and activation of neutrophilic granulocytes for cancer chemotherapy-induced neutropenia (Sanchez-Garcia et al., 2016). Recombinant human interleukin-3 (hIL-3; **Figure 4D**) protein is a cytokine that regulates the differentiation and proliferation of the various cells of the immune system (Hercus et al., 2013). The hIL-3 protein is derived from B. subtilis, B. licheniformis, and E. coli and has utility as a laboratory reagent in hematology for cell cultures, differentiation studies and

functional assays. It has shown that hIL-3 has potential in treating bone marrow failure, hematological malignancies, and can support engraftment after bone marrow transplantation (Westers et al., 2006). In addition, recombinant Pfs48/45 is a disulfide-rich malaria transmission-blocking vaccine produced by L. lactis that provides immunization against malaria from P. falciparum (Song et al., 2017).

Recombivax is produced by S. cerevisiae and can prevent infection of all known subtypes of the Hepatitis B virus (Sanchez-Garcia et al., 2016). Some examples of currently approved protein therapeutics derived from yeast include human serum albumin (Recombumin <sup>R</sup> and Albucult <sup>R</sup> ; **Figure 4E**), human insulin (Actrapid <sup>R</sup> ) and primary immunization for infants born of Hepatitis B virus (HBV) surface antigen (Pediarix <sup>R</sup> ), all of which are obtained exclusively from S. cerevisiae (McAleer et al., 1984; Ballance, 1999; Nielsen, 2013; Nandy and Srivastava, 2018). Recombinant human serum albumin is utilized to increase the shelf life of protein drugs by preventing physical and chemical degradation. Actrapid <sup>R</sup> is used to treat diabetes, and subcutaneous injections of Pediarix is designed for immunization against diphtheria, tetanus, pertussis, poliomyelitis, and infection caused by all known subtypes of HBV (Nandy and Srivastava, 2018). Ecallantide (Kalbitor <sup>R</sup> ) is an FDA-approved recombinant peptide produced by Pichia pastoris for the treatment of hereditary angioedema (Sheffer et al., 2011). Additionally, anakinra (Kineret <sup>R</sup> ) was approved in 2001 in the United States for rheumatoid arthritis (Baeshen et al., 2015). Anakinra is expressed in E. coli and functions as an IL-1 receptor antagonist that is effective and safe for patients with systemic-onset juvenile idiopathic arthritis, adult-onset Still's disease, hereditary autoinflammatory syndromes, and Schnitzler's syndrome (Kalliolias and Liossis, 2008; Jozala et al., 2016).

# MICROBIAL CELL FACTORIES

Selecting a suitable host strain is one of the most important aspects in the design of natural product and recombinant protein bioprocesses. We will review the characteristics of the microbial strains used to produce natural products and biologics in this section. We will also present the tools and strategies that facilitate engineering of the hosts as microbial cell factories for the production of biopharmaceutical compounds (**Table 2**).

#### Gram-Negative Bacteria Escherichia coli

Escherichia coli has been seen as one of the optimal systems for the production of natural products because it is easily manipulated, highly productive, there is an availability of genetic tools to use with it and there is a deep knowledge of its physiology. Artemisinin, a sesquiterpene lactone endoperoxide from Artemisia annua L. plants, has strong antimalarial activity against the multi-drug resistant parasite P. falciparum (Abdin et al., 2003). Yet the synthesis of artemisinin is costly and low yields are isolated from the natural plant source. Researchers reported the production of approximately 24 mg/L of amorpha-4,11-diene (amorphadiene), an artemisinin

TABLE 2 | Comparison between different microbial host systems for production of recombinant proteins and natural products.


• Tetracycline • Daptomycin • Chloramphenicol

#### TABLE 2 | Continued

fmicb-10-01404 June 19, 2019 Time: 15:53 # 10


• Thermo-tolerant

precursor, by the expression of a codon-optimized synthetic amorphadiene synthase gene and the mevalonate pathway from S. cerevisiae in E. coli. Additionally, after further processing modifications and optimal conditions, they were able to produce 105 mg/L of artemisinic acid (Chang et al., 2007). However, there are some obstacles and limitations with E. coli as a dominant host in natural product biosynthesis. E. coli requires extensive genetic manipulation and lacks native natural product biosynthetic machinery and/or precursors. An example is phosphopantetheinyl transferase, which is responsible for the activation of the carrier protein domains of the PKSs and NRPSs. This enzyme must be introduced into E. coli to support of natural product biosynthesis (Zhang M.M. et al., 2016). There have been efforts to overcome these hurdles, such as the production of erythromycin A and its derivatives in the engineered E. coli strain (Zhang et al., 2010). The study generated two analogs through directed manipulation of polyketide biosynthesis in which variations were made to the deoxyerythronolide B synthase (DEBS) 1 and DEBS3 enzymes in order to utilize the multi-catalytic capability of the modular polyketide synthase (Zhang et al., 2010).

Escherichia coli has also been the pioneering host for recombinant protein production. To date, E. coli continues to be the first-choice microorganism for manufacturing recombinant proteins at laboratory and industrial scales. Its success is mostly due to its fast growth, simple culture procedures, cost-effectiveness, unusually high versatility, and the associated systems that make it adaptable to varying production demands (Ferrer-Miralles and Villaverde, 2013; Sanchez-Garcia et al., 2016). From 2004 to 2013, 24% of the biopharmaceuticals approved by the FDA and the European Medicines Agency were derived from E. coli (Baeshen et al., 2014). Currently, biopharmaceuticals produced from E. coli are used in the treatment of diabetes, growth hormone-deficiency in children, leukemia, gout, and many other therapeutic indications as previously discussed in Section "Biological Activity of Microbial Biologics" (Baeshen et al., 2015). A major concern when using E. coli as a production platform is the lack of post-translational modifications (PTMs) present in most eukaryotic proteins; lacking PTMs can lead to protein products being insoluble, unstable, or inactive (Ferrer-Miralles et al., 2009). However, it is possible to add synthetic PTMs to generate versions of the protein that are more stable than the original naked product (Ferrer-Miralles et al., 2009). Examples of this include pegylated products, like human growth hormone, stimulating factor, IFNs α-2a and α-2b, (Ferrer-Miralles et al., 2009). Additionally, there is a risk of translational errors due to the presence of a large number of rare codons that appear in human genes that are different from those occurring in E. coli genes. Even at low levels, these errors may cause an impact on the tertiary structure, premature termination of protein synthesis or amino acid misincorporation which results in low protein expression (Gustafsson et al., 2004). One strategy to bypass the issue with codon bias is to synthesize the whole human gene based on codon usage in E. coli through site-directed mutagenesis, which is currently a preferred method; however, it is limited by the high cost of production and time consumption (Sørensen and Mortensen, 2005). An alternative method that is less time consuming utilizes the co-transformation of E. coli strains with a plasmid(s) containing a gene encoding the tRNA cognate to the rare codons (Dieci et al., 2000). Increasing the copy number allows for E. coli to be manipulated to match the codon usage frequency in heterologous genes (Dieci et al., 2000). Currently, there are numerous commercial E. colistrains available that harbor plasmids containing gene sequences encoding the tRNA for rare codons, such as BL21(DE3) CodonPlus-RIL, BL21(DE3) CodonPlus-RP and Rosetta (DE3) (Baeshen et al., 2015). Another common problem associated with recombinant protein expression in E. coli involve inclusion body formation, which refers to insoluble and inactive protein aggregates (Hartley and Kane, 1988). E. coli producing recombinant proteins have the ability to assemble in cells and form conglomerates of inclusion bodies as well as result in erroneous protein folding which hinder the extraction of proteins directly from the cell leading to costly purification of proteins (Zweers et al., 2008). Inclusion bodies formed from lack of proportion between protein solubilization and aggregation can be resolved by combining the desired protein with a solubility enhancer fusion partner acting as an intrinsic chaperone in order to ensure the production of soluble recombinant proteins (Rosano and Ceccarelli, 2014). The fusion of maltose-binding protein to polypeptides such as human tissue inhibitor of metalloproteinase and p16 improved their solubility significantly in E. coli (Kapust and Waugh, 1999).

# Gram-Positive Bacteria

#### Lactococcus lactis

Lactococcus lactis is becoming an attractive alternative in genetic engineering for the production of various recombinant proteins. Unlike E. coli, which uses intracellular production strategies that involve expensive and often challenging purification processes, L. lactis utilizes extracellular secretion system. This is because L. lactis has a monolayer cell wall that allows direct secretion into the extracellular environment (Schneewind and Missiakas, 2012). The presence of exported proteases such as HtrA in L. lactis contributes to recombinant protein production by minimizing the destruction of heterologous proteins in the medium (Morello et al., 2008; Song et al., 2017). Additionally, L. lactis does not generate undesired glycosylation of protein, is generally recognized as safe (GRAS), does not produce endotoxins, and has probiotic properties (Singh et al., 2018). Another advantage of L. lactis includes a lack of inclusion body formation (Theisen et al., 2017). There is a diverse selection of cloning and inducible expression vectors available for use with this host that are compatible with large-scale upstream and downstream processes (Song et al., 2017).

Lactococcus lactis has been used for centuries in the fermentation of food, especially in cheese, yogurt, and sauerkraut because of its production of nisin (Song et al., 2017; Singh et al., 2018). Beyond the food industry, lactic acid is used as an emulsifier and moisturizing agent in the cosmetic industry and as an important raw material in the pharmaceutical industry (Papagianni, 2012). The L. lactis host has also been chosen after researchers had unsuccessfully attempted to obtain correct conformation using a variety of other

prokaryotic and eukaryotic recombinant protein expression systems. L. lactis has multiple advantages over E. coli for expression of 5<sup>0</sup> -methylthioadenosine/S-adenosylhomocysteine nucleosidase (Pfs) recombinant proteins, including the following: (1) codon-optimization of the recombinant gene is not necessary to achieve successful expression in L. lactis; (2) the recombinant protein is secreted into the L. lactis culture supernatant, which results in easier and less expensive down-stream processing, and (3) there is no lipopolysaccharide contamination in L. lactis expression (Singh et al., 2018). L. lactis has been used in the successful production of recombinant Pfs48/45, a vaccine candidate against P. falciparum (Song et al., 2017). GMZ2, a recombinant fusion protein expressed in L. lactis, is also a malaria vaccine candidate that has been shown to elicit high levels of active IgG antibodies with inhibitory activity against a broad range of P. falciparum strains (Jepsen et al., 2013). A recent study concluded phase 2 trial of GMZ2 adjuvanted with aluminum hydroxide in a cohort of 1,849 children revealed GMZ2 as well tolerated with modest efficacy (Sirima et al., 2016). Not only is L. lactis being utilized for the production of recombinant proteins for vaccines, but the host is also being tested as a factory for antigen production, allowing the bacteria to function as live vaccines. Using L. lactis as a vaccine carrier is beneficial because it can induce both mucosal and systemic immune responses, has adjuvant properties and is free from the risks associated with the use of conventional attenuated live pathogens, such as Salmonella species and Mycobacterium species (Song et al., 2017). However, while L. lactis has been studied against an array of antigens from various pathogens, there is no current live vaccines under clinical trial which may be due to a lack of containment strategies (Bahey-El-Din, 2012). Without a plan for containment, studies on live L. lactis vectors risk the chance of proliferation and dispersion. An additional limitation of AT rich L. lactis as a cell factory is due to codon usage as well as distribution of rare codons to express heterologous genes (Mierau and Kleerebezem, 2005).

#### Streptomyces Species

Another major species that has shown promise as a cell factory through its wide production of natural products and biologics is Streptomyces. This Gram-positive bacterium has been studied for more than 70 years and has proven to be of great use in biotechnology due to its ability to produce natural products acting as antibiotics, anticancer agents, and immunosuppressants (Yoo et al., 2015). Some examples include tetracycline, daptomycin, and chloramphenicol (De Lima Procópio et al., 2012). There are many species of Streptomyces currently to produce various natural products and biologics. Among the Streptomyces species, Streptomyces coelicolor, Streptomyces lividans, Streptomyces albus, and S. venezuelae are favored heterologous hosts to produce specialized metabolites due to the relative ease of their genetic manipulation, the availability of their genome sequences, and the abundant supply of their natural substrates (Park et al., 2010). Streptomyces has also been used to produce a wide array of heterologous proteins of both eukaryotic and prokaryotic origin (Gilbert et al., 1995) because Streptomyces has well-developed genetic manipulation and fermentation technologies as well as efficient protein secretion systems such as the secretory (Sec) pathway and the twin-arginine-translocation (Tat) pathway (Hamed et al., 2018). The Sec-pathway catalyzes the translocation of unfolded proteins while the Tat pathway allows for the export of folded proteins across the cytoplasmic membrane (Natale et al., 2008). Tumor necrosis factor (TNF) α and human interleukin (hIL) 10 are able to be expressed in both the Sec- and Tat-pathways (Schaerlaekens et al., 2004). In particular, S. lividans could be the ideal Streptomyces host due to limited restriction-modification systems and low endogenous protease activity (Nakashima et al., 2005). Streptokinase (Pimienta et al., 2007), transforming growth factor-α (Taguchi et al., 1995), IL-2 (Bender et al., 1990) and many other proteins have been successfully expressed and secreted from S. lividans. However, aside from its efficient secretory pathways, when in culture, Streptomyces grows as mycelial networks, leading to the formation of pellets or clumps (Van Dissel et al., 2015). These pellets are unappealing from an industrial standpoint because of mass-transfer problems, slow growth, and culture heterogeneity which leads to lower product yield (Van Dissel et al., 2015).

#### Bacillus Species

Bacillus species are some of the most popular species used in enzyme production. It accounts for roughly 50% of enzymes market within the industrial sphere (Barros et al., 2013). Certain species, like B. subtilis, Bacillus amyloliquefaciens, and B. licheniformis are favored because of their outstanding fermentation properties, high protein production yield, and their completely toxin-free production (Van Dissel et al., 2015). The fermentation capacity of Bacillus species in acid, neutral, and alkaline pH ranges in addition to thermophiles accounts for the prolific production of enzymes that have desirable temperature, pH, and stability, which makes them appealing for specific use in various industries (Schallmey et al., 2004). Bacillus species are known for their production of iturins and fengycin which have antifungal activity as well as surfactin for its function as a surfactant (Wang T. et al., 2015).

Among these species, B. subtilis is the most widely studied due to (1) its flexibility during genetic engineering, (2) its naturally high secretory capacity, (3) its ability to produce valuable antibiotics, such as bacillomycins D-L and bacitracin, and (4) its ability to produce enzymes, such as stable alkaline cellulase, alkaline protease and alkaline α-amylase. It may also elicit better folding conditions, leading to the prevention of inclusion bodies (Peypoux et al., 1984; Van Dijl and Hecker, 2013). In addition, its ability to adapt to varying environmental conditions as well as its classification as toxic free GRAS has contributed tremendously to its success in the industrial platform (Baysal and Y*ı*ld*ı*z, 2017). B. subtilis as an endotoxin free host amplified its utilization in the production of sterile recombinant and therapeutic proteins expression as compared to E. coli which could have potential contamination due to the lipopolysaccharide endotoxins (Wang et al., 2013). For instance, B. subtilis and Bacillus megaterium were the preferred hosts over E. coli in the production of bioengineered heparin in order to diminish toxin contamination (Wang et al., 2013). Moreover, B. subtilis is able to produce high yield in enzyme as it secretes the enzymes

straight into the fermentation medium due to the absence of outer membrane which allows easy recovery of purified proteins from the medium into their active form (Zweers et al., 2008). It has the capacity to secrete about 20–25 g/L of enzymes into the medium (Schallmey et al., 2004). Enzymes produced by B. subtilis has a wide variety of applications ranging from pharmaceutical, leather, detergent, food, and waste management industries (Singh et al., 2016).

Aside from enzyme production, cytokines like hIL-3 have been produced by B. subtilis and B. licheniformis. The production of hIL-3 has been tested in various host organisms, including E. coli. However, the production of IL-3 within other organisms has often exhibited problems, such as insolubility or the degradation of produced hIL-3. This led to the use of B. licheniformis and B. subtilis to minimize such complications. The production of hIL-3 in B. licheniformis was engineered to lack four C-terminal residues, resulting in a fully active hIL-3 protein. However, residual proteolytic degradation of some hIL-3 still occurred, leading to use B. subtilis to achieve complete folding and full biological activity of hIL-3 (Westers et al., 2006).

Among the Bacillus species, Bacillus thuringiensis is best known for being widely used within the agricultural industry due to its insecticidal properties through its production of parasporal crystals during the stationary phase of its growth cycle (Höfte and Whiteley, 1989; Schnepf et al., 1998). Upon ingestion, the parasporal crystals are solubilized in the midgut of insects, resulting in the release of protoxin proteins known as δ-endotoxins, leading to the formation of pores throughout the cell membrane (Höfte and Whiteley, 1989; Gill et al., 1992). Parasporal proteins also have selective cytotoxicity against liver and colon cancer cells while leaving normal cells unharmed (Ito et al., 2004).

However, the use of Bacillus has been restricted to mainly enzyme production and non-recombinant protein therapeutics, which may be due to the lack of associated expression vectors, plasmid instability and the presence of native proteases (Westers et al., 2004). Despite B. subtilis success as the industrial workhorse, it has its drawbacks in the production of heterologous proteins. Heterologous protein yield could diminish when using the Bacillus as a host due to the proteolytic destruction of foreign protein by host secreted extracellular proteases (Nijland and Kuipers, 2008). Efforts have been made to improve the production of heterologous protein by manipulating the expression of proteins involved in the post translocation phase resulting in amplified levels of heterologous protein secretion (Vitikainen et al., 2005). In contrast to E. coli, the absence of distinguished and controllable promoters in B. subtilis interferes with successful expression of heterologous genes resulting in inefficient production of heterologous proteins (Schallmey et al., 2004).

#### Yeast/Fungi

#### Saccharomyces cerevisiae

As with E. coli, S. cerevisiae has been extensively used for the production of recombinant human insulin since the early 1980s, and it currently produces half of the world's supply of insulin (Meehl and Stadheim, 2014). S. cerevisiae is preferred because it is also cost-effective, fast growing, technically practical, and is amenable to large-scale fermentation in bioreactors. Yeast is often utilized as a cell factory when the target protein is not produced in a soluble form in prokaryotic systems or when a specific PTM cannot be produced or added to the naked product. S. cerevisiae, as with other yeast species, can perform many PTMs such as O-linked glycosylation, phosphorylation, acetylation, and acylation, which allows recombinant proteins to be expressed in a soluble, correctly folded, and functionally active form (Ferrer-Miralles et al., 2009; Baeshen et al., 2014). Some examples of currently approved protein therapeutics derived from yeast include human serum albumin, insulin, and primary immunization for infants born of HBV surface antigen, all which are obtained in S. cerevisiae (McAleer et al., 1984; Ballance, 1999; Nielsen, 2013; Nandy and Srivastava, 2018). However, the significant drawback to producing protein therapeutics from S. cerevisiae is that higher eukaryotes have a different pattern of N-linked glycosylation, which can decrease the half-life and cause hyper-immunogenicity, leading to less effective therapeutics (Ferrer-Miralles et al., 2009). In recent years, there have been some advances in limiting S. cerevisiae hypermannosylation. These yeast glycoengineering techniques involve two main stages, (1) the removal of yeast hypermannosylation and (2) the conversion to complex glycoforms containing terminal sugars, such as N-acetylglucosamine, galactose, or sialic acid. These recent reports on yeast N-glycan humanization indicate a move from the proof of concept phase to implementation (Meehl and Stadheim, 2014).

Another current area of study is the production of plant and microbe-derived secondary metabolites. Due to the structural complexity of secondary metabolites, chemical synthesis is an inefficient route for large-scale production, and fermentation is the main mode for economic commercial production of pharmaceutically useful natural products (McDaniel et al., 2001). S. cerevisiae could be an ideal candidate as a microbial host as it boasts relatively rapid growth, and it is accompanied by highly developed genetic tools and advanced fermentation science. Like E. coli, S. cerevisiae has been shown to be an outstanding production host for artemisinic acid, a precursor to the antimalarial agent artemisinin, with a high productivity (25 g/L) that led to the industrial production of semi-synthetic artemisinin beginning in 2013 (Paddon et al., 2013; Kung et al., 2018). Research has also produced the paclitaxel (Taxol <sup>R</sup> ) precursor taxadiene (∼73 mg/L) by engineering the taxol biosynthetic genes in S. cerevisiae (Ding et al., 2014). Besides plant secondary metabolites, S. cerevisiae has generated a remarkable titer (1.7 g/L) of microbial polyketide 6-methylsalicylic acid in un-optimized shake-flask fermentations. In addition, S. cerevisiae has been developed as a heterologous host to express fungal cryptic BGCs and their respective metabolites. In this study, 30 ADH2-like promoters in Saccharomyces species were developed as tools for expression of 41 heterologous BGCs, 22 of which were capable of producing heterologous compounds, including novel compounds. For example, BGCs derived from basidiomycete were found to produce N-, S-bis-acylated amino acids and a leucine O-methyl ester with an additional polyketide chain amidated to the amino ester (Harvey et al., 2018). However, barriers still exist to the heterologous production of complex molecules. This includes the production of polyketides in S. cerevisiae, as the host lacks some required polyketide precursor pathways, its codon usage is biased toward A + T (most microbial polyketide producers have high G + C genome content) and it lacks the appropriate endogenous phosphopantetheinyl transferase capable of the necessary PTMs (Mutka et al., 2006).

#### Aspergillus Species

fmicb-10-01404 June 19, 2019 Time: 15:53 # 14

Multicellular filamentous fungi, such as A. niger and Aspergillus oryzae, can also offer great potential in the production of a desired substance by fermentation due to the following reasons: (1) they are well-characterized GRAS organisms, (2) are amenable to scaled-up fermentation, (3) can be genetically engineered, (4) they are capable of secreting a high level of proteins and (5) can withstand adjustable cultivation conditions, including temperature (5–45◦ ), pH (2–11), salinity (as much as 34%), water activity (0.92–0.98), and both nutrient rich and poor environments (Maheshwari, 2006; Meyer et al., 2011).

Aspergillus niger has been predominantly used for industriallevel production of citric acid through anaerobic fermentation process. As a weak acid, citric acid can serve as a natural preservative, flavoring agent in food and beverages, antioxidant, acidulant, pH-regulator, chelating agent or vegetable rinse, as well as comparable applications in the pharmaceutical and cosmetics industries (Cairns et al., 2018). Due to its wide variety applications, its ease of production, and economical potential of citric acid, the market of citric acid has become one of the fastestgrowing regions of the food additives market due to the rising demand: according to estimations, in 2007, the market value of citric acid exceeded \$2 billion in 2014 and is predicted to rise to \$3.6 billion by 2020 (Show et al., 2015; Cairns et al., 2018). Phytase is another example that have produced by A. niger fermentation (Papagianni et al., 1999). The significance of phytase enzymes lie in its ability to interact with the nutrient rich compounds known as phytate. Phytate, or phytic acid, is a common energy source found in oilseeds, cereals, and legumes, which are used in providing nutrition to animal feeds (Schlemmer et al., 2009). Combining citric acid with phytase has also been shown to enhance phytase activity on phytate, producing greater nutrient outcomes in tested animals (Boling et al., 2000). In addition, two different humanized immunoglobulin G1(κ) antibodies and an Fab<sup>0</sup> fragment were produced by A. niger, and the antibodies were successfully secreted into the culture supernatant (Ward et al., 2004). Aspergillus strains have been also used to produce the human iron-binding glycoprotein lactoferrin and hIL-2 with the yields of 25 and 150 mg/L, respectively (Maheshwari, 2006). Bicoumanigrin with cytotoxic activity against human cancer cell lines and aspernigrin B with neuroprotective effects have both been isolated from A. niger (Hiort et al., 2004). The product spectrum of Aspergillus species is not restricted to biologic molecules. As an example, a novel cyclic peptide compound, KK-1, with potent antifungal activity was produced in A. oryzae by introducing gene clusters spanning approximately 40 kb from the plant-pathogenic fungus Curvularia clavata into the genome of A. oryzae. Although the amount of KK-1 produced by the host was lower than that produced by the original producer C. clavata, this result indicated that a gene twice as large as the largest native gene in A. oryzae could be successfully expressed (Yoshimi et al., 2018). Furthermore, when the industrial fungus A. niger was engineered as a heterologous host, it produced high titers (up to 4,500 mg/L) of enniatin belonging to non-ribosomal peptides with antibacterial, antiviral, and anticancer activities (Richter et al., 2014).

#### Hansenula polymorpha

Another industrially important yeast species that has shown promise in the production of peptides is Hansenula polymorpha (Gellissen et al., 1992; Boer et al., 2007). H. polymorpha is a methylotrophic yeast species with the ability to use and grow on methanol, glucose, or glycerol as its primary carbon source (Gellissen et al., 1992). Like S. cerevisiae and Aspergillus species, H. polymorpha, classified as GRAS organism, does not harbor pyrogens, toxins, pathogens, or viral inclusions (Ubiyvovk et al., 2011). It is distinguished by very high cell densities in bioreactors and characterized by simple cultivation mode in inexpensive growth media. For example, H. polymorpha has allowed for cost-effective production of phytase through cheap carbon sources (Mayer et al., 1999). It possesses well-established genetic tools such as strong regulatory and constitutive promoters, which consequently give high product yield (Van Dijk et al., 2000). It also has thermotolerance properties, making H. polymorpha successful in crystallographic studies and in the production of recombinant proteins like IFN-2α, IL-6, recombinant human serum albumin, glucose oxidase, and catalase (Kunze et al., 2009; Celik and Calik, 2012). A notable feature of H. polymorpha is the significant growth of peroxisomes when grown on methanol which allows for high storage capacity of soluble proteins. The lack of protein modifying enzymes in the matrix of peroxisomes also provides an advantage for the development of heterologous proteins that are susceptible to proteolytic degradation (Van Dijk et al., 2000). Furthermore, the host has been used to produce L antigens found on the HBV viral envelope in attempt to produce the HBV vaccine. The L protein produced by H. polymorpha has increased stability in comparison to other yeast species, such as S. cerevisiae and P. pastoris (Janowicz et al., 1991). In addition to its use in vaccine production, H. polymorpha is also used in the production of human hemoglobin through the use of a single expression vector (Hollenberg and Gellissen, 1997). However, hyperglycosylation has been observed as a main drawback of H. polymorpha to produce heterologous products (Müller et al., 1998).

#### EFFORTS IN PRODUCT IMPROVEMENTS AND GENERATION OF NEW ANALOGS

There are multiple approaches that have been taken to advance product improvement for microbial natural products and biologics. This section will discuss efforts to combat the

challenges of production of natural products and its analogs, including strain improvement, increasing precursor supply, pathway engineering, combinatorial biosynthesis, and genome mining (**Figure 5**).

#### Strain Improvement

Whole-genome shuffling is a process that utilizes the advantages of the multi-parental crossing allowed by DNA shuffling with the genome recombination normally associated with conventional breeding (Zhang et al., 2002). Genome shuffling has been successfully improved the titers of variety of microorganisms. For example, two strains of Streptomyces fradiae generated from two rounds of genome shuffling were able to produce up to a ninefold increase in antibacterial tylosin production in comparison to the initial strain (Zhang et al., 2002). Using genome shuffling in a combination of protoplast fusion, mutant strain of S. cellulosum GSUV3−<sup>205</sup> generated a 130 fold increase (104 mg/L) in production of epothilone when compared to starting strain S. cellulosum So0157-2 (0.8 mg/L) (Gong et al., 2007). Ribosome engineering is also a method useful in increasing secondary metabolite production titer and productivity (Sun and Alper, 2015). Studies demonstrate

that rpoB mutations are effective in activating silent and poorly expressed secondary metabolite biosynthetic gene clusters (BGCs) at the transcriptional level in S. griseus, S. coelicolor, and S. erythraea (Ochi and Hosaka, 2013). For example, the H437R mutant of rpoB from S. erythraea was screened for drug resistance and was found to have an increased production of erythromycin (Tanaka et al., 2013; Ochi et al., 2014). Another study found a 37-fold increased production of avilamycin in a recombinant Streptomyces viridochromogenes strain due to a mutation in ribosome protein S12 (rps12) acquired through a combination of gene shuffling and ribosome engineering (Lv et al., 2013).

#### Engineering Precursor Supply

Precursor supply is defined as the enhancement of the availability of primary metabolites or molecules derived from primary metabolism involved in the biosynthesis of natural products (Shiba et al., 2007). Precursor supply engineering can be achieved by manipulating either the pathways or enzymes involved with the precursor supply. Malonyl-CoA and methylmalonyl-CoA are the most commonly used and metabolically available precursors for the biosynthesis of polyketides. One study found that supplying methyl oleate enhanced the internal concentration of methylmalonyl-CoA, which is a biosynthetic precursor for FK506, and led to a 2.5-fold increase in FK506 production in Streptomyces clavuligerus CKD1119 (Mo et al., 2009). In another study, propionyl-CoA carboxylase with supplementation of propionate was found to effectively increase methylmalonyl-CoA and rapamycin titers in the mutant strain S. rapamycinicus UV2−<sup>2</sup> induced by ultraviolet mutagenesis in comparison to wild-type strain (7 ∼ 8 mg/L) (Jung et al., 2011). The mutant strain was found to have a 3.2-fold improvement (23.6 mg/L) in comparison to wild-type strain S. rapamycinicus ATCC 29253 (7 ∼ 8 mg/L) (Jung et al., 2011). Further, Méndez and coworkers improved precursor metabolite pools for the production of the antitumor polyketide mithramycin in Streptomyces argillaceus by increasing the precursor supply of malonyl-CoA and glucose-1-phosphate (Zabala et al., 2013). Several classes of natural products utilize aromatic amino acids or other metabolites derived from the shikimate pathway as precursors, including flavonoids, alkaloids, polyketides, and non-ribosomal peptides (Knaggs, 2001). The production of the vancomycin analog balhimycin was increased 2.5-fold in Amycolatopsis sp. Y-89,21022. This was achieved by increasing the non-ribosomal peptide precursor 3-deoxy-D-arabino-heptulosonate7-phosphate synthase, the first enzyme in the shikimate pathway (Thykaer et al., 2010). In addition, manipulating key enzymes that direct carbon flux through core biochemical pathways involved in glucose, fatty acid, and amino acid metabolism can increase biosynthetic precursor pools. A study on the modulation of carbon flux between the pentose phosphate pathway and the glycolysis pathway found that a deletion of phosphofructokinase isoenzymes led to the enhanced production of antibiotics actinorhodin and undecylprodigiosin in S. coelicolor by increasing carbon flux through the pentose phosphate pathway (Borodina et al., 2008).

Precursor supply engineering has been successfully used to produce most of the major classes of natural products, with application to heterologous producing strains as well as native producers. When the native host is slow-growing or cannot be easily manipulated genetically, this process could be performed effectively in an appropriate heterologous host. A study bypassed the native deoxyxylulose 5-phosphate pathway and instead introduced the mevalonate pathway from S. cerevisiae to E. coli, which allowed for an increased production of amorpha-4,11-diene which is a precursor to antimalarial artemisinin (Martin et al., 2003). Combined approach of exogenous supplementation and engineering of intracellular pathway responsible for precursors can be also performed in a heterologous host. The biosynthetic process of the hybrid non-ribosomal peptide-polyketide yersiniabactin was known to rely on the supply of salicylate, L-cysteine, S-adenosyl-L-methionine, malonyl-CoA, and NADPH. When exogenous cysteine was fed to the culture of E. coli harboring yersiniabactin BGC and an additional set of genes (hmwp1-2) responsible for yersiniabactin precursor biosynthesis was introduced, the yersiniabactin production in E. coli was boosted to approximately 175 mg/L (Ahmadi and Pfeifer, 2016).

Precursor engineering strategy can be also employed to increase recombinant protein production by reducing unwanted by-products. One of the primary obstacles observed in high cell density cultivations of E. coli for the production of recombinant proteins is the formation of acetate, which is a byproduct caused by an excess influx of carbon during aerobic fermentation. This acetate accumulation hampers cell growth and recombinant protein formation, even at low concentrations (Waegeman and Soetaert, 2011). A number of engineering approaches have focused on minimizing acetate formation in order to enhance recombinant protein production in E. coli. When a heterologous anaplerotic pyruvate carboxylase from Rhizobium etli is overexpressed in E. coli, the resulting strain had a 57% reduction in acetate formation and a 68% increase in β-galactosidase production (March et al., 2002). It is also possible to combine different strategies to reduce the formation of undesired by-products, including acetate. For example, one study found that a mutant E. coli strain containing a defective acetate pathway and an overexpressed phosphoenolpyruvate carboxylase-encoding ppC reduced acetate and other by-product formation and produced five time more β-galactosidase activity when compared the wild type strain (De Mey et al., 2010).

#### Pathway Engineering

Metabolic pathway engineering can be performed in the native host through repetitive gene expression, gene deletion, and introduction of new genes to enhance production of natural products (Pickens et al., 2011). For example, overexpression of the 4–12 tandem copies of the actinorhodin cluster resulted in a 20-fold increase in actinorhodin production in S. coelicolor (Murakami et al., 2011). Additionally, a S. hygroscopicus strain with 3–5 tandem copies of the 40 kb validomycin A cluster showed a 34% increase in production and a maximum titer of approximately 20 g/L (Zhou et al., 2014). Deletion of genes may be useful to eliminate competing pathways that may siphon off

important precursors or intermediates, or simply contribute to an unnecessary use of cellular resources which result to improve yields of products of interest. During in vivo bioconversion of lovastatin intermediate monacolin J to simvastatin using E. coli expressing heterologous acyltransferase LovD, it was found that E. coli could unexpectedly hydrolyze the synthetic thioester substrate. The responsible hydrolase BioH was knocked out to improve simvastatin production (Xie et al., 2007). The regulatory component of the pathway can be manipulated to enhance production of the resulting natural product. Negative regulation by pathway specific repressors can help regulate secondary metabolite pathways. For example, one study improved the titer by 100-fold of antibiotics platensimycin (323 mg/L) and platencin (255 mg/L) through the inactivation of a gene encoding protein PtmR1 belonging to GntR family of transcriptional repressors (Smanski et al., 2009). On the other side of the spectrum, Streptomyces antibiotic regulatory protein (SARP) is a positive regulator of antibiotic production (Chen et al., 2010). Overexpression of SARPs and/or increasing SARP gene dosage using multi-copy plasmids has been demonstrated to increase production titers. Overexpression of mgsA or chxA, SARP family members that are positive regulators for the iso-migrastatin and cycloheximide biosynthetic machinery, respectively, in Streptomyces amphibiosporus ATCC 53964 led to a fivefold increased production of antibiotic lactimidomycin (Zhang et al., 2016). Members of the large ATP-binding regulators of the LuxR (LAL) family also generally function as transcriptional activators, and constitutive overexpression of these LAL-type activators was found to increase production of rapamycin in S. rapamycinicus and FK506 in S. tsukubaensis (Kušèer et al., 2007; Mo et al., 2012).

Competing pathways can also be deleted to ensure the production of important precursors or intermediates and to save useful cellular resources. When deleting pathways, the idea is to create a host with a minimized genome to ensure the efficient production of necessary secondary metabolites. Deleting nonessential genes and directing cellular resources toward pathways that are essential for the survival and product biosynthesis can improve cellular efficiency and streamline biochemical production. For example, the genome of Streptomyces avermitilis was effectively minimized to 83% of its original size. When heterologous streptomycin gene cluster was introduced into the genome-minimized S. avermitilis, the resulting strain produced a higher titer of streptomycin than both the parent S. avermitilis carrying the same heterologous gene cluster and the native streptomycin producer S. griseus (Komatsu et al., 2010). However, large scale deletions may result in unintended effects as the complete workings of the cell are not yet entirely understood.

Similar approaches have been employed to improve the secretion capability and productivity of biologics. Engineering the protein trafficking pathway represents one successful approach to improve the secretion of heterologous recombinant proteins. For example, the secretion of the heterologous proteins human insulin precursor and α-amylase from A. oryzae in S. cerevisiae was improved by the over-expression of Sec1/Munc18 proteins, which are involved in the protein secretory pathway (Hou et al., 2012). Increasing the copy number of genes that are associated with protein secretion can also enhance protein secretion. This is seen with the Necator americanus secretory protein (Na-ASP1), which shows potential as a vaccine protein for hookworm infections. Increasing the Na-ASP1 gene copy number caused saturation of secretory capacity in P. pastoris, a species of methylotrophic yeast, led to a decreased amount of secreted protein. This was remedied by the overexpression of the protein disulfide isomerase, which allowed for the increased secretion of Na-ASP1 protein in high copy clones (Inan et al., 2006). Another study showed that deletion of obstructive protease genes involved in fission could lead to the enhanced secretion of protease-sensitive human growth hormones (hGH) in Schizosaccharomyces pombe. The production of hGH was hampered by the intracellular retention of secretory hGH, and it was determined that the multi-protease deletant strain plays a role in hGH retention. Deletion of vps10, which encodes a carboxypeptidase Y sorting receptor and is involved in the traffic between the late-Golgi and prevacuolar compartments, resulted in an approximate twofold increase in hGH secretion (Idiris et al., 2006).

#### Combinatorial Biosynthesis

Combinatorial biosynthesis is one genetic engineering application that can modify biosynthetic pathways in order to yield new and altered natural product structures (Hopwood et al., 1985). This approach exploits indiscriminate substrates and uses engineered enzymes and pathways for the production of new natural product analogs.

Modular megasynthases, such as PKS and NRPS enzymes, constitute a class of multifunctional proteins that govern complex enzymatic mechanisms and catalyze multiple reactions useful for combinatorial biosynthesis. Type 1 PKSs consist of multiple modules which are responsible for incorporating acyl-CoAs into a polyketide backbone for elongation. Meanwhile, NRPSs are composed of a modular set of repeating enzyme domains for the activation and incorporation of amino acids (Park and Yoon, 2015). The modular NRPSs typically consist of a condensation domain, adenylation domain, and a thiolation domain, while type I PKSs generally contain a ketosynthase domain, acyltransferase domain, and an acyl carrier protein (Komaki et al., 2015; Skiba et al., 2018). Natural product structures can be modified by mixing and matching the megasynthases at the subunit, module, and domain levels. Genetic manipulation of PKS and NRPS encoding genes can result in predictable changes in structure that is difficult to achieve with standard chemical derivatization or total synthesis methods (Park et al., 2010). This approach to manipulating substrate incorporation and biosynthetic PKS and NRPS machinery has allowed for the generation of a great number of natural product analogs. Examples include erythromycin (McDaniel et al., 1999), pikromycin from type I modular PKS (Yoon et al., 2002) and daptomycin from NRPS (Robbel and Marahiel, 2010).

Post-assembly modifications, such as glycosylation, oxidation, and halogenation are performed by diverse enzymes and can lead to structurally and biologically diverse natural compounds (Park et al., 2010). Sugar moieties attached to the core structure of polyketides or non-ribosomal peptides by glycosyltransferases can also contribute to an extension of combinatorial biosynthesis.

Since several glycosyltransferases have been known to be flexible toward sugar donors and sugar accepters, arrays of analogs differing in glycosylation patterns via tailoring enzymes can also be generated by combinatorial engineering of glycosyltransferases from different pathways. For example, one study found that A. orientalis-derived glycosyltransferases accepted the unnatural sugar 4-epi-vancosamine in the presence of vancomycin pseudoaglycone or the glucosylated teicoplanin scaffold to generate novel hybrid glycopeptide compounds such as 4-epi-vancosaminyl form of vancomycin (Losey et al., 2001). Besides sugar biosynthesis, combinatorial biosynthesis can be applied for other modifications such as oxidation and halogenation. Oxidase genes from polyketide pathways have been used to induce structural alterations of important functional groups that are essential for biological activities. It has been reported that 5-O-desosaminyl erythronolide A, a potent precursor of ketolides and the latest generation of antibiotic compounds derived from erythromycin A, was produced by expressing the monooxidase gene pikC from the pikromycin pathway in a mutant strain of S. erythraea lacking of a EryBV glycosyltransferase (Basnet et al., 2008). In addition, a recent study obtained nine analogs of the antitumor antibiotic xantholipin through the individual in-frame mutagenesis of five tailoring enzymes (Zhang et al., 2012). In another study, fluorosalinosporamide, a derivative of the potent anticancer agent salinosporamide A, was produced by replacing the chlorinase gene salL from Salinispora tropica with the fluorinase gene flA from Salinispora cattleya (Eustaquio et al., 2010).

However, a common concern with this approach regards limited tolerance of downstream enzymes or domains to the new substance introduced by combinatorial biosynthesis and metabolic engineering. Rational design or directed evolution is one solution to this concern. Rational design is the strategy of creating new molecules with a certain functionality based on predicting how the molecule's structure will affect its behavior, while directed evolution refers to methods to alter enzyme function using mutagenesis and selection (Nannemann et al., 2011). In a recent study, the reactivity of PikC was modified through protein engineering driven by molecular dynamics and quantum mechanical calculations. The computation-driven PikC engineering yielded a PikCD50N mutant that showed improved catalytic efficiency compared to the wild-type PikC (Narayan et al., 2015). This study demonstrated that a rationally designed protein using a crystal structure of protein and/or a computational analysis can develop a predictive model for substrate scope and selectivity of natural product biosynthesismediated reactions. Directed evolution is also a powerful tool to modify the activity of key enzymes responsible for the biosynthesis of natural products and can lead a higher diversity of natural products by generating novel and more potent analogs (Williams et al., 2013). As an example, a few rounds of directed evolution restored and enhanced the activity of an impaired chimerical enterobactin NRPS that has been swapped with a non-cognate aryl-carrier protein (Zhou et al., 2007). In order to reduce the risk of limited tolerance and reduce concerns of efficiency, directed evolution requires a large, high-quality library and an efficient screening strategy. The swapping of functional domains often results in non-functional or heavily impaired chimerical enzymes, and this remains an existing problem when manipulating modular PKS and NRPS systems.

#### Mutasynthesis

Novel natural product analogs can also be generated through gene disruption and mutasynthesis. Disruption of a gene, such as a tailoring enzyme acting downstream in a pathway, can serve to introduce a structural change. Two FK506 analogs, 9-deoxo-31- O-demethylFK506 and 31-O-demethylFK506, were produced by targeting gene disruption in Streptomyces sp. MA6548 (Shafiee et al., 1997; Ban et al., 2013). These two recombinant mutants were genetically engineered via disruption of fkbD and fkbM genes that code for 31-O-demethylFK506 methyltransferase and 9-deoxo-31-demethylFK506 hydroxylase/oxidase (Shafiee et al., 1997; Ban et al., 2013). Inactivation of individual domains within the multidomain modular PKSs and NRPSs serves as an alternative to the deletion of a whole gene. Mutasynthesis involves the coupling of a gene inactivation strategy with precursor feeding to generate new structural analogs. Precursor feeding is useful due to the substrate-promiscuity of the biosynthetic enzyme. Precursor feeding may lead to the acceptance of similar substrates or mutasynthons, a natural substrate substitute that can replace the natural substrate of a disrupted gene after being added to the growth medium, to ultimately generate new analogs. Mutasynthesis can generate new analogs for many classes of compounds. For example, the analog cahuitamycin D was produced through mutasynthetic generation with twofold-enhanced biofilm inhibitory activity in comparison to its natural product counterpart (Park et al., 2016). Recently, this approach was applied to generate nonbenzoquinone analogs of the Hsp90 inhibitor geldanamycin, which has anti-proliferative activity on tumor cells (Shin et al., 2008; Wu et al., 2011). By removing the biosynthetic genes for the 3-amino-5-hydroxybenzoic acid starter unit and feeding the culture with various 3-aminobenzoic acids and related heterocycles, a chloro-substituted nonbenzoquinone analog with significantly improved therapeutic properties was produced along with other geldanamycin analogs (Kim et al., 2007, 2009). This has been also seen in the generation of new analogs of rapamycin (Khaw et al., 1998), balhimycin (Weist et al., 2002), and novobiocin/chlorobiocin (Li and Heide, 2005).

# FUTURE PROSPECTS

An increasing number of natural products and natural product-derived compounds have been launched over the years (Butler et al., 2014). Since 2000, 77% of FDAapproved antibiotics are natural products, all of which were derived from microbes (Patridge et al., 2016). There have been extensive reviews of natural products, semisynthetic natural products, and nature-inspired molecules currently approved by the FDA that show the continued importance of natural products for medicine and health (Sanchez et al., 2012; Newman and Cragg, 2016). Microbial

biologics are expected to remain prominent in the global biologics market, which was valued at 277 billion USD in 2015 and was recently estimated to reach 400 billion USD by 2025 (Grand View Research, 2017). While many of the biological activities of microbial natural products and biologics are well known, new advances and insights continue to be discovered. Chemical diversity from microbial natural products continue to be relevant to future drug discovery, with a continuing need for novel drugs with antibiotic, anticancer, and immunosuppressant effects, along with other pharmacological activities (Sanchez et al., 2012).

At the same time, there are a multitude of challenges facing microbial production of natural products and biologics. Some challenges to natural products-based drug discovery involve low production titers, difficulty in product isolation or structural identification. Similarly, there is much room for improvement in terms of the expression of recombinant proteins in microbial platforms. Accumulation of the end product in the microbial cell can cause global stress responses that result in cell growth inhibition. Also, the formation of misfolded and biologically inactive proteins can lower the yield of recombinant proteins. In particular, membrane proteins, high-molecular weight proteins, and multi-domain proteins are often expressed in inclusion bodies. Additionally, expressing eukaryotic proteins in a prokaryotic-based heterologous system can result in a product that is not correctly modified by posttranslational enzymes, which are often required for functionality (Rosano and Ceccarelli, 2014). However, a wide variety of engineering strategies can be used with the conventional recombinant DNA technologies, including genome editing, ribosome engineering, precursor engineering, mutagenesis, and overexpression of structural genes, making it possible to facilitate the efficient production of natural products and pharmaceuticals in microbial systems.

Current technologies, such as CRISPR/Cas (Clustered Regularly Interspaced Short Palindromic Repeats/CRISPR associated protein), should also be considered as tools for genome editing for additional improvements and to increase production (Gaj et al., 2013). For example, in vitro CRISPR-Cas9 cloning with Gibson assembly has provided an alternate strategy for heterologous expression of cryptic BGCs from genetically recalcitrant actinomycetes strains (Wang J.W. et al., 2015; Jiang and Zhu, 2016). Additionally, CRISPR-Cas9 has many prospects in the analyses of biochemical pathways in Streptomyces strains due to the development of a pCRISPomyces expression system (Cobb et al., 2015). With genome editing tools, it is also possible that nonmodel native hosts can be engineered to be heterologous production hosts to become platforms for combinatorial biosynthesis to create synthetic natural products and natural product derivatives. Some intrinsic limitations for nonmodel microbial hosts can be also improved by this genetic modification. Corynebacterium glutamicum, a GRAS organism, has been used for the industrial production of various amino acids for over five decades (Becker et al., 2011). Recently, it also showed promising potential for use as a protein expression system (Kallscheuer et al., 2016). However, this

bacterium has intrinsic disadvantages, including a much lower transformation efficiency and lower yields of protein production. The CRISPR/Cas9 system was successfully employed to disrupt four different genes in C. glutamicum, opening new possibilities to use non-model strains as improved cell factories for the production of recombinant proteins and natural products (Peng et al., 2017).

Genome mining is another alternative process to discover secondary metabolites and is done by extracting information from genome sequencing (Bachmann et al., 2014; Zhongyue et al., 2018). Evaluating silent cryptic BGCs through genome mining has provided valuable avenues to generate novel molecules. For example, a genome mining strategy combined with bioinformatics predictions was used to isolate the novel natural product orfamide A by feeding a predicted precursor to a culture of Pseudomonas fluorescens (Zerikly and Challis, 2009). In a recent study, genome miningbased combinatorial biosynthesis approach also led to the discovery of new members of the leinamycin family of natural products. Leinamycin has been considered a promising anticancer drug lead due to its potent anticancer activities, unique molecular architecture and interesting modes of action. However, no leinamycin analog had been isolated in the past three decades until this study (Pan et al., 2017).

A combination of approaches can also lead to improvements in the field of microbial natural products, such as gene shuffling and ribosome engineering for increased secondary metabolite production. Additionally, the integration of 'omics' information has great potential in natural product drug discovery, such as with metabolomics to accurately quantify biochemical changes and metabolic pathways. Advancements in metagenomics has allowed for further understanding of diverse and complex microbial sources, including lakes, rivers, marine environments and extreme conditions, such as sub-seafloor sites and ice cores (Chandra Mohana et al., 2018).

In terms of structural characterization, X-ray crystallography and cryo-electron microscopy are advanced techniques that can allow for structure solving with high precision. Cryoelectron microscopy has been a leading method for evaluating macromolecular structures at near-atom resolution (Shoemaker and Ando, 2018). For example, single particle electron cryomicroscopy has been used to visualize pikromycin PKS module 5 from S. venezuelae, which allowed for 3D map construction with resolutions of 7.3–9.5 Å to reveal secondary structures (Dutta et al., 2014). These techniques are just a few among many that should be considered for structural studies on natural product biosynthesis. Advancements in computational strategies have led to the identification of BGCs in genome sequences and predictions of product chemical structures. Sequencing campaigns for natural product discovery should be directed toward samples likely to yield novel natural products along with well-characterized clades, such as actinomycetes, as they are still a resource for natural product discovery and have yet to be fully exhausted. The development of algorithms to mine the ever-increasing amounts of metagenomic data will allow for

the potential of genome mining to be realized (Medema and Fischbach, 2015). Finally, developing additional host platforms for high-throughput refactoring and functional expression of pathways has the potential to overcome current limitations in precursor supply, product toxicity, the ability to express very large gene clusters and more (Schmidt-Dannert, 2015).

Overall, microbial natural products and biologics will continue to broaden their diverse and integral role in human life. The potential for recombinant drugs is expanding through the utilization of new protein production platforms and efforts in product improvement. Microbial cells will remain as potent protein factories because of their versatility and cost-effectiveness. Engineering strategies and recombinant DNA technologies will also allow for the increased production of microbial natural products and recombinant proteins despite the many challenges faced. Continued efforts in natural product analog development will provide an avenue for the discovery of compounds with improved biological activities in comparison to their natural counterparts. Current advanced technologies can be utilized to further advance the field of microbial natural products, which remain a steadfast resource for novel compounds in drug discovery.

#### REFERENCES


## AUTHOR CONTRIBUTIONS

SP and YY designed, directed, and coordinated this project. JP, MY, AF, MM, NM, JW, EK, HC, JR, MCS, SP, and YY made substantial contributions in providing critical feedback and drafting the manuscript. SP and JP reviewed the final manuscript.

#### FUNDING

This work was supported by a National Research Foundation of Korea (NRF) grant (2019R1A2B5B03069338) funded by the Ministry of Science and ICT (MSIT), the Bio & Medical Technology Development Program of the NRF funded by the MSIT (2018M3A9F3079662), the Collaborative Genome Program of the Korea Institute of Marine Science and Technology Promotion (KIMST) funded by the Ministry of Oceans and Fisheries (MOF) (No. 20180430), and Cooperative Research Program for Agriculture Science & Technology Development (PJ01316001) funded by the Rural Development Administration, South Korea. MCS was supported by the RP-Grant 2019 funded by the Ewha Womans University.


system to simultaneous comparative assessment. FEMS Yeast Res. 7, 1181–1187. doi: 10.1111/j.1567-1364.2007.00254.x





lipolytica. Cloning of two novel promoters from Yarrowia lipolytica. Yeast 14, 1267–1283. doi: 10.1002/(sici)1097-0061(1998100)14:14<1267::aid-yea327>3. 3.co;2-u


and future prospects. Appl. Microbiol. Biotechnol. 85, 1227–1239. doi: 10.1007/ s00253-009-2326-8


platencin. Antimicrob. Agents Chemother. 53, 1299–1304. doi: 10.1128/aac. 01358-08



**Conflict of Interest Statement:** 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 Pham, Yilma, Feliz, Majid, Maffetone, Walker, Kim, Cho, Reynolds, Song, Park and Yoon. 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.