Recent advances in the biosynthesis strategies of nitrogen heterocyclic natural products

Contents

• 1. Introduction

• 2. Enzymatic biosynthesis of NHNPs
  • 2.1 Enzymatic biosynthesis of N-heterocyclic skeletons of NHNPs
  • 2.2 Tuning the tailing enzyme properties for NHNPs skeleton decoration by protein engineering
  • 2.3 Multi-enzyme cascade strategy for the biosynthesis of NHNPs
  • 3. Production of NHNPs in microbial cell factory
  • 3.1 Deciphering the biosynthesis pathway of NHNPs
  • 3.2 Heterologous biosynthesis of NHNPs in microbial cell factory

• 4. Conclusions and prospective

• 5. Conflicts of interest

• Acknowledgements

• References

---

Abstract

Covering: 2015 to 2020

Nitrogen heterocyclic natural products (NHNPs) are primary or secondary metabolites containing nitrogen heterocyclic (N-heterocyclic) skeletons. Due to the existence of the N-heterocyclic structure, NHNPs exhibit various bioactivities such as anticancer and antibacterial, which makes them widely used in medicines, pesticides, and food additives. However, the low content of these NHNPs in native organisms severely restricts their commercial application. Although a variety of NHNPs have been produced through extraction or chemical synthesis strategies, these methods suffer from several problems. The development of biotechnology provides new options for the production of NHNPs. This review introduces the recent progress of two strategies for the biosynthesis of NHNPs: enzymatic biosynthesis and microbial cell factory. In the enzymatic biosynthesis part, the recent progress in the mining of enzymes that synthesize N-heterocyclic skeletons (e.g., pyrrole, piperidine, diketopiperazine, and isoquinoline), the engineering of tailoring enzymes, and enzyme cascades constructed to synthesize NHNPs are discussed. In the microbial cell factory part, with tropane alkaloids (TAs) and tetrahydroisoquinoline (THIQ) alkaloids as the representative compounds, the strategies of unraveling unknown natural biosynthesis pathways of NHNPs in plants are summarized, and various metabolic engineering strategies to enhance their production in microbes are introduced. Ultimately, future perspectives for accelerating the biosynthesis of NHNPs are discussed.

---

Bo Gao

Bo Gao is a master's student at the Beijing Institute of Technology, mentored by Prof. Chun Li. He received his BS degree from the Beijing Institute of Technology in 2018. His research is focused on enzyme engineering for natural product biosynthesis.

Bo Yang

Bo Yang is currently a PhD student under the supervision of Prof. Chun Li at Tianjin University. Her research is focused on the efficient production of natural products in the microbial cell factory through metabolic engineering and synthetic biology.

Xudong Feng

Dr Xudong Feng is an assistant professor at the Beijing Institute of Technology. He obtained his PhD degree from the University of Auckland in 2014. His research interest is focused on the mining and engineering of key enzymes in natural products synthesis. He has published 50 peer-reviewed research papers in this field.

Chun Li

Dr Chun Li is a professor at the Tsinghua University and Distinguished Young Scholar of Natural Science Foundation of China. He received his PhD degree in Biochemical Engineering from the Tianjin University in 2001. He is a senior expert on synthetic biology, biocatalysis, and enzyme engineering, with more than 15 years of research experience. Since the past ten years, his research has been focused on the biosynthesis of plant natural products in microorganisms and the development of strategies and tools for increasing the productivity of plant natural products and the robustness of microbes. He has published 307 peer-reviewed research papers.

---

1. Introduction

Nitrogen heterocyclic natural products (NHNPs) are primary or secondary metabolites containing nitrogen heterocyclic (N-heterocyclic) skeletons, which have a wide range of biological activities. Among them, the primary metabolites of NHNPs include purines, pyrimidines, three N-heterocyclic amino acids (tryptophan, proline and histidine), and some members of the vitamin B family. They play important roles in storing genetic information, supplying energy, and participating in various important metabolic pathways of organisms.^1–3 The secondary metabolites of NHNPs include alkaloids, pseudoalkaloids, biogenic amine derivatives, and β-lactam antibiotics. Most alkaloids and pseudoalkaloids contain indole, isoquinoline, pyrrole, pyridine, or other nitrogen-containing heterocyclic skeletons, which belong to NHNPs.^4–12 A few alkaloids and pseudoalkaloids without nitrogen atoms in the heterocyclic ring do not belong to NHNPs. N-heterocyclic alkaloids and pseudoalkaloids have been widely used for developing new medicines and pesticides due to their remarkable anti-tumor, anti-bacterial, anti-fungal, insect toxicity, and other pharmacological activities. For example, morphine is widely used as an anesthesia due to its analgesic effects;¹³ scopolamine is used for treating gastric ulcers due to its effect of relieving gastrointestinal cramps;¹⁴ camptothecin is used as a chemotherapy drug for leukemia and gastric cancer.¹⁵

Nitrogen has a significant effect on the activity of the heterocyclic skeleton. On the one hand, nitrogen has a lone pair of electrons in the heterocyclic ring, which shows weak basicity. On the other hand, polar N–H bond exhibits weak acidity. These characteristics of the nitrogen element enable the N-heterocyclic skeleton to bind a variety of bioactive molecules, which improves the bioavailability of NHNPs and makes them exhibit excellent physiological activities. For example, five-membered heterocycles such as pyrrole and pyrazole usually form electron-rich compounds due to the lone pair of electrons of the nitrogen atom participating in the cyclic conjugated π bond and are prone to electrophilic reactions. The lone pair of electrons of six-membered heterocyclic nitrogen atoms such as pyridine and quinoline do not participate in conjugation and usually form electron-deficient compounds, and are prone to nucleophilic reactions. The combined action of the N-heterocyclic skeleton and the surrounding modification groups determine the biological activity of NHNPs. For example, indole is an electron-rich aromatic N-heterocycle. Many natural products containing indole rings are both weakly basic and weakly acidic, and can form π–π interactions with various active substances or receptors. Due to the unsaturated bond, the indole nucleus also has potential reducibility.¹⁶ A variety of natural products can be obtained by introducing substituents of different properties into the indole ring, which can exhibit various physiological activities such as anti-cancer, antibacterial, anti-oxidant, anti-viral, and anti-convulsant.¹⁷ Diketopiperazine is an N-heterocyclic compound formed by the dehydration condensation of two amino acids. The diketopiperazine skeleton contains two amide bonds; thus, there are two hydrogen bond acceptors and hydrogen bond donor sites, which allow them to bind to a variety of enzymes and acceptors. Adding modified groups to the diketopiperazine skeleton can form many natural products with higher receptor binding ability, inhibiting a variety of target proteins and produce a variety of pharmacological activities such as anticancer, anti-inflammatory, and anti-cardiovascular diseases.¹⁸

At present, more than 20 000 NHNPs have been reported.¹⁹ In the past two centuries, researchers have been continuously discovering new NHNPs and exploring their potential values, and remarkable progress has been achieved. However, the low native content of NHNPs severely restricts the industrial applications.^5,10 The traditional strategy is to directly extract NHNPs from plants. Several high-value NHNPs such as vincristine and morphine, are still obtained through large-scale plant cultivation, which exhibit many disadvantages. Firstly, most plants have a long growth cycle; thus, they are easily affected by environmental factors, such as insects, pests, and climate. Secondly, the extraction of active ingredients from the plants not only causes environmental pollution but also faces complex separation and purification problems. For microbe-derived NHNPs, indistinct biosynthetic pathways have hindered their industrial production.

Researchers are committed to the artificial synthesis of NHNPs to overcome the shortcomings of the traditional methods. Chemical synthesis is currently the most common method of obtaining NHNPs since it can provide different synthetic routes, obtain multiple analogs, and screen the best products.^20–24 However, for NHNPs with complicated structures (e.g., opioid alkaloids and terpenoid alkaloid), chemical synthesis has severe disadvantages, including multiple steps, low yield, by-product production, as well as poor regio- and stereo-specificity. In addition, similar to plant extraction, chemical synthesis can also cause serious environmental pollution.

For the development of biotechnology, biosynthetic strategies provide a promising option for obtaining NHNPs, especially those with complicated structures. The biosynthesis of NHNPs has the advantages of high efficiency, rigid specificity, and environment friendliness, which is attracting increasing attention from researchers. With the aid of sequencing technology, transcriptomics, and proteomics, researchers have elucidated the synthetic pathways of multiple NHNPs in native hosts.^9,25 Generally, N-heterocyclic structures are enzymatically synthesized by condensing amines or imine precursors, which were then modified to form NHNPs after oxidation, reduction, methylation, amination, acetylation, or glycosylation. Initially, plant cell engineering was used to produce NHNPs. However, the complex metabolic network of the plant cells and the interaction among the metabolic pathways severely restrict the synthesis of target NHNPs.²⁶ The emergence of DNA synthesis technology has provided a breakthrough in the biosynthesis of NHNPs. Researchers can conveniently reconstruct metabolic networks in easily-cultivable model microorganisms or use recombinant enzymes to synthesize NHNPs. Due to their high chemo-, regio-, and stereoselectivity toward natural structures, enzymatic biosynthesis and microbial cell factories are the most widely used biosynthetic strategies for obtaining NHNPs. Moreover, the rapid development of metabolic engineering, protein engineering, and synthetic biology have significantly improved the yield of NHNPs.^8,27

This review introduces the progress of biosynthesis of NHNPs, mainly in the recent five years. Secondary metabolites containing N-heterocyclic structures were selected as the object of review. For the enzymatic biosynthesis part, the important enzymes for producing typical NHNP skeletons are classified as three major groups including polyketide synthase/non-ribosomal peptide synthase (pyrrole/piperidine), cyclodipeptide synthases (diketopiperazine), and Pictet–Spenglerase (isoquinoline/β-carboline). Their mechanism and mining strategies are summarized. Subsequently, post-modification enzymes add various functional groups to the above skeletons to form valuable NHNPs. The strategies developed for engineering post-modification enzymes to improve their stability, activity, and specificity are also elucidated. The construction of a multi-enzyme cascade for the biosynthesis of precursors of NHNPs, including isoquinoline, pyrrole, and piperidine derivatives, has also been introduced. Tropane alkaloids (TAs) and tetrahydroisoquinoline (THIQ) alkaloids are selected as the representative compounds in the section of NHNPs production in microbial cell factories. The recent strategies developed for the decryption of the biosynthesis pathway of NHNPs are summarized. Then, the construction and regulation strategy of the NHNPs biosynthesis in microbial cell factories are introduced. This content can provide guidance for researchers who are engaged in deciphering the unknown synthetic pathways of natural products and optimizing the synthetic efficiency of the cell factories. Finally, the challenges in the biosynthesis of NHNPs have been discussed.

2. Enzymatic biosynthesis of NHNPs

The biosynthesis of NHNPs depends on the participation of multiple enzymes. Enzymatic biosynthesis has great potential for NHNPs production, especially for those with complex structures. Enzymatic biosynthesis in vitro allows the precise control of reaction conditions such as temperature and pH, which enables the effective synthesis of NHNPs without being affected by the inherent metabolic network of the host. This section introduces the progress of enzymatic biosynthesis of NHNPs in vitro. First, the key enzymes are summarized for the synthesis of N-heterocyclic skeletons, which provide the basic structure for NHNPs. Subsequently, the tailoring enzymes for introducing the functional groups into the N-heterocyclic skeleton to provide activity toward NHNPs are discussed. Recently, dramatic progress has been made to improve the properties of tailoring enzymes by protein engineering; thus, this part is especially emphasized. The key enzymes for forming skeletons can be combined with different tailoring enzymes to construct an in vitro cascade reaction system in order to produce diverse NHNPs and their derivatives, and this content is introduced at the end of this section (Fig. 1).

Fig. 1 The schematic diagram of the enzymatic biosynthesis of NHNPs.

2.1 Enzymatic biosynthesis of N-heterocyclic skeletons of NHNPs

The biosynthesis of NHNPs comprises two steps: the formation of N-heterocyclic skeletons and post-modification. In nature, the key enzymes for the synthesis of N-heterocyclic skeleton of secondary metabolites mainly include polyketide synthase/non-ribosomal peptide synthase (PKS/NRPS), cyclodipeptide synthase (CDPS), and Pictet–Spenglerase (PSase). These enzymes can synthesize common five-membered and six-membered N-heterocyclic skeletons, such as pyrrole, piperidine, and piperazine (Fig. 2). Then, these skeletons are decorated with various functional groups by post-modification enzymes, such as cytochrome P450, decarboxylase, methyltransferase, acyltransferase, and prenyltransferase to form NHNPs with diverse structures.^28–30 The discovery of enzymes involved in NHNPs biosynthesis provides availability for industrial applications. Recently, great progress has been made in terms of mining the enzymes that can synthesize N-heterocyclic skeletons. Thus, we reviewed the latest progress in the mining and in vitro characterization of these enzymes in this part.

Fig. 2 The reaction mechanism of key enzymes that synthesize N-heterocyclic skeletons. (A) PKS/NRPS. (B) CDPS. (C) PSase.

2.1.1 PKS/NRPS. PKS and NRPS are megacomplexes (>100 kDa) that play an important role in the formation of many NHNPs skeletons including pyrrole, pyridine, piperidine, acridine, isoquinoline, diazaheptane, and cyclic peptides. PKS/NRPS have been comprehensively introduced in some reviews previously.^31–35 Due to the limited space, this review mainly focuses on PKS/NRPS, which synthesize five-membered ring (pyrrole) and six-membered ring (piperidine) since these small and medium-sized skeletons can be easily modified to form many high-value NHNPs in multiple steps (Fig. 2A and 3).^36,37 Different domains of PKS, NRPS, and hybrid PKS-NRPS megacomplexes catalyze reactions to synthesize various NHNPs through three stages: substrate initiation, chain extension, and termination. The generation of N-heterocyclic skeletons is catalyzed by thioester reductase domains (TRs) on the C terminal region of PKS and NRPS through NAD(P)H-dependent reductive release of peptidyl-(S)-nonribosomal peptides or acyl-S-polyketides. Bioinformatics analysis is an efficient strategy for the mining of biosynthetic genes of PKS and NRPS. Through whole-genome sequencing and sequence identity analysis, PKS and NRPS with unknown function can be identified (Fig. 3).³⁸ For example, a recent study characterized five TRs that catalyzed the formation of pyrrole and piperidine skeletons in the biosynthesis of kirromycin, streptolydigin, tirandamycin B, α-lipomycin, and factumycin.³⁹

Fig. 3 Typical N-heterocyclic skeletons synthesized by TR.

2.1.2 CDPS. CDPSs are small enzymes (∼30 kDa) that assemble two aminoacyl-tRNAs (aa-tRNAs) into a 2,5-diketopiperazine skeleton. CDPSs mediate the amino nitrogen of the first aa-tRNA substrate to nucleophilically attack the carbonyl carbon of the second aa-tRNA substrate to generate the first peptide bond, and then catalyzes the formation of the intramolecular peptide bond to obtain the diketopiperazine skeletons (Fig. 2B). Most CDPSs are derived from microorganisms and exhibit substrate promiscuity. They can assemble multiple natural and unnatural amino acids into cyclic dipeptides. CDPS-encoding genes are usually located in the same cluster with the genes encoding modifying enzymes that modify diketopiperazine. The function of genes in the cluster would be easily illuminated if one of them was elucidated.⁴⁰ Gondry et al. proposed a method to mine CDPSs and analyze their substrate specificity.⁴¹ In their study, candidate sequences can be screened by aligning with the N-terminal amino acid sequence of previously characterized CDPSs. By analyzing the crystal structure of multiple CDPSs, they identified two conserved pockets related to the specificity of cyclic dipeptide synthesis, P1 and P2, which contain seven and eight residues respectively. CDPSs with similar P1 and P2 sequences may yield the same products. They finally mined 32 new CDPSs, which synthesized 16 new cyclic dipeptides. Meng et al.⁴² identified the biosynthetic pathway of bicyclomycin in Streptomyces sapporonensis by heterologous biosynthesis and in vitro biochemical assays. The CDPS encoded by bcmA in the biosynthetic gene clusters guided leucine and isoleucine to form a diketopiperazine skeleton (Fig. 4A). James et al.⁴³ identified two rare CDPSs in Nocardiopsis sp. CMB-M0232 genome, which specifically assembles two tryptophanyl-tRNAs to produce diketopiperazine (Fig. 4B). Through sequence alignment with the reported CDPSs specific to tryptophanyl-tRNA, several unique residues were considered to be related to the substrate specificity, including valine and asparagine residues, which recognize the first substrate, and lysine/arginine residues, which bind to the second substrate.

Fig. 4 Typical N-heterocyclic skeletons synthesized by CDPS. (A) BcmA. (B) NozA and NcdA.

2.1.3 PSase. PSases catalyze the intermolecular cyclization reaction of β-arylethylamine with aldehydes or ketones to form piperidine skeletons of isoquinoline or β-carboline natural products.⁴⁴ Several PSases participate in the formation of some important NHNP skeletons. For example, norcoclaurine synthase (NCS) catalyzes 4-hydroxyphenylacetaldehyde (4-HPAA) and dopamine to produce (S)-norcoclaurine, the key intermediate in the isoquinoline alkaloid biosynthesis (Fig. 2C). Strictosidine synthase catalyzes tryptamine and secologanin to yield 3-α(S)-strictosidine, the key intermediate in the monoterpenoid indole alkaloids biosynthesis.⁴⁵ In recent years, several new PSases have been identified and characterized in vitro. Lechner et al.⁴⁶ mined novel NCSs from Argemone mexicana and Corydalis saxicola, which showed high sequence identity with the reported NCS from Argemone mexicana. Mori et al.⁴⁷ characterized the in vitro activity of McbB, a PSase from Marinoctinospora thermotolerans, which catalyzed the formation of 1-acetyl-β-carboline and 1-acetyl-3-carboxy-β-carboline from L-tryptophan and oxaloacetaldehyde (Fig. 5A). Subsequently, several PSases with a high sequence identity to McbB were mined, such as NscB from Nocardiopsis synnemataformans and StnK2 from Streptomyces flocculu (Fig. 5A and B).^48,49

Fig. 5 Typical N-heterocyclic skeletons synthesized by PSase. (A) McbB and NscB. (B) StnK2.

2.1.4 Other key enzymes. In addition to the three kinds of enzymes mentioned above, some other enzymes can also catalyze the formation of the N-heterocyclic skeleton. Chen et al.⁵⁰ identified an FAD-binding oxidase, CnsA, which catalyzed the formation of seven-membered N-heterocycles in the Clavicipitaceae and Trichocomaceae families of fungi. Through in vitro functional verification, CnsA was proposed to catalyze the isomerization of the double bond, which subsequently caused the amino group to attack the electrophilic carbon to form a heterocyclic skeleton (Fig. 6A). Ryan et al.⁵¹ identified a heme-dependent N–N bond synthase, KtzT, in Kutzneria sp. 744. Du et al. identified a heme-dependent N–N bond synthetase, KtzT, in the kutzneride gene cluster of Kutzneria sp. 744, which catalyzed the formation of intramolecular N–N bonds between the N⁵-hydroxyl and amino groups of L-N⁵-hydroxyornithine (Fig. 6B). Transaminases and imine reductases (IREDs) can also synthesize the N-heterocyclic skeleton. ω-Transaminases catalyze amino transfer between the amino acids and keto acids. Five ω-transaminases were identified from different bacteria and fungi, which used a series of 1,5-diketones containing aliphatic chains and benzene ring substituents on C1 as the substrates to produce 2,6-disubstituted piperidine products (Fig. 6C).⁵² IREDs asymmetrically reduced cyclic imines to generate chiral amines. Borlinghaus et al.⁵³ used IRED from Myxococcus stipitatus to obtain a series of piperazine derivatives with different 1,2-diamine and 1,2-dicarbonyl substrates. Xu et al.⁵⁴ screened IREDs from Leishmania major and Micromonospora echinaurantiaca, which catalyzed asymmetric reductive amination to synthesize chiral 1,4-diazepanes with (R)- and (S)-enantioselectivity, respectively (Fig. 6D).

Fig. 6 Typical N-heterocyclic skeletons synthesized by other key enzymes. (A) CnsA. (B) KtzT. (C) ω-Transaminase. (D) IRED.

2.2 Tuning the tailing enzyme properties for NHNPs skeleton decoration by protein engineering

The multiple structures of NHNPs provide diverse bioactivities, which depend on the post-modification of the N-heterocyclic skeletons. However, post-modifications are usually the rate-limiting steps in NHNPs biosynthesis. To solve this problem, protein engineering strategies are widely used (Fig. 7). Enzymes with high activity, stability, substrate promiscuity, and selectivity can be obtained by mutating critical residues or fragments.^55–57 Directed evolution and rational design strategies are widely used in the construction of engineered enzymes. The mutant library was generated by random or rational mutation, and then the mutants in the library were screened under specific pressure. Mutants with improved properties were utilized as the templates for the next round of mutations until the desired mutants were obtained.^58,59 Protein engineering of post-modification enzymes for NHNP biosynthesis has made significant progress recently and this section mainly introduces the latest progress of this point.

Fig. 7 Construction of engineered enzymes to synthesize NHNPs. (A) Process of protein engineering. (B) Influence of protein engineering on the enzyme properties.

2.2.1 Engineering stability. Wild-type enzymes are generally poor in stability and heat resistance; thus, they cannot meet the requirements of industrial production. Improving the heat resistance through protein engineering directly improves the efficiency of enzymatic synthesis. One common strategy for improving the heat resistance is to introduce homologous enzymes as templates from thermophilic microorganisms.⁶⁰ However, this strategy is rarely used in the synthesis of NHNPs, possibly due to the lack of reports on corresponding enzymes from thermophilic microorganisms. Another common strategy is to analyze the B-factor of each amino acid residue and replace the residue with high B-factor value. Xu et al.⁶¹ analyzed the B-factor of all the residues of tryptophan synthase from E. coli K-12 and mutated the key residue G395 with the largest B-factor value (95.24). The optimal reaction temperature of mutant G395S was increased from 35 °C to 40 °C. Structural analysis showed that the enhanced thermal stability of mutant G395S was due to the increased rigidity of the enzyme.

2.2.2 Engineering activity. Recently, random mutation and saturation mutation have been widely used to increase the activity of enzymes that synthesize NHNPs (Fig. 8). Hyoscyamine 6β-hydroxylase (H6H) catalyzes the hydroxylation of (−)-hyoscyamine to form scopolamine. TA. Cao et al.⁶² performed error-prone PCR on H6H from Anisodus acutangulus to construct a mutant library containing ∼10 000 clones. They screened sixteen mutants with increased activity in the library, which contained five amino acid residue mutations. Single-site mutations were made to these five residues, and two key residues S14 and K97 that affected the enzyme activity were determined. Saturated mutations were then performed on these two residues, and a double mutant S14P/K97A with 4.4-fold increased hydroxylation activity was obtained. However, random mutation requires the screening of a large number of mutants, which reduces the efficiency. Some residues on the substrate binding pocket directly affect the activity; thus, saturation mutation on these residues can obtain mutants with increased activity efficiently. Fan et al.⁶³ performed saturation mutation on the substrate binding residue R224 of the cyclic dipeptide C4-prenyltransferase from Aspergillus fumigatus. R244 was proposed to bind to the side chain of substrate and has been verified to affect the activity. Thirteen mutants at R244 showed up to 76-fold higher turnover number toward seven cyclic dipeptides. Brockmeyer et al.⁶⁴ mutated two residues, T82 and Y196, in the substrate binding pocket P1 of a CDPS from Nocardiopsis prasina. The mutant T82V/Y196F showed 8-fold increased activity in the synthesis of cyclo-(L-Tyr-L-Tyr) and 10-fold increased activity in the synthesis of cyclo-(L-Tyr–L-Phe).

Fig. 8 Engineering the enzyme activity to synthesize NHNPs and their derivatives.

2.2.3 Engineering substrate promiscuity. Currently, enzymes for NHNP modification are still unknown. Recently, enzymes exhibiting catalytic promiscuity toward analogues of natural substrates have been successfully engineered for the modification of NHNP skeletons (Fig. 9). A representative example is the engineering of tryptophan synthase, which catalyzes indole and L-serine to form tryptophan. Although tryptophan synthase can accept several indole derivatives as nucleophiles, it cannot accept electrophiles except serine. Many high-value tryptophan derivatives can be obtained by improving the substrate promiscuity of tryptophan synthase. Arnold et al.^65–68 selected tryptophan synthase β subunits (TrpB, active) from different hosts for engineering. They constructed a TrpB mutant library by random mutations, and several TrpB variants showed excellent activity toward unnatural substrates, including L-threonine (electrophile) and more than 20 mono- or di-substituted indoles (nucleophiles) (Fig. 9A). The increased promiscuity was attributed to the improved stability of the intermediate. The mutation of residues on the conserved motifs and catalytic pocket is an effective way to change the substrate specificity. Aliphatic halogenases are highly selective biocatalysts to synthesize chlorine substituted organics. However, all the natural substrates of aliphatic halogenases contain isonitrile functionality, which limits their application in medicine. Duewel et al.⁶⁹ crystallized Wi-WelO15, an aliphatic halogenase from Westiella intricata HT-29-1, and docked Wi-WelO15 with its natural substrate derivative (without the isonitrile functionality) (Fig. 9B). First, they performed single-residue and multi-residue mutations on seven residues at the active site of Wi-WelO15. Then, they selected four residues around the active site and selected several best variants of the previous round of mutations as the templates for further mutations. They obtained a series of Wi-WelO15 variants, which catalyzed the chlorination of four unnatural substrates (without the isonitrile functionality). A widely used strategy to increase the substrate promiscuity is to replace the amino acid residues around the catalytic pocket to smaller ones, which can expand the catalytic pocket to gain tolerance toward larger substrates.⁷⁰ Scoulerine 9-O-methyltransferase (S9OMT) catalyzes the conversion of (S)-scoulerine to (S)-tetrahydrocolumbamine, which is a post-modification reaction in isoquinoline biosynthetic pathway. Valentic et al.⁷¹ obtained an S9OMT variant N191D/F205S with increased activity from Thalictrum flavum (TfS9OMT). They solved multiple crystal structures of this variants and identified several key residues affecting the substrate scope and methylation regiospecificity. The N191D mutation facilitated cofactor binding, while the F205S mutation improved the thermal stability. Then, they introduced the M111A mutation to expand the substrate pocket of TfS9OMT N191D/F205S, resulting in improved catalytic promiscuity. The TfS9OMT N191D/F205S/M111A mutant showed not only methylation activity toward three unnatural isoquinoline substrates, namely, (S)-reticuline, (S)-norcoclaurine, and (S)-norlaudanosoline, but also additional 2-O-methylation activity toward (S)-scoulerine, thus producing two new NHNPs, namely, (S)-tetrahydropalmatrubine and (S)-tetrahydropalmatine (Fig. 9C and 10).

Fig. 9 Engineering the substrate promiscuity to synthesize NHNPs and their derivatives. (A) TrpB. (B) wi-WelO15. (C) S9OMT. (D) RebH.

Fig. 10 Crystal structures of S9OMT wt (A) and S9OMT DS (B). Substrates and mutated residues are indicated by sticks. The images were reproduced from ref. 71.

The synthesis of NHNPs often involves the cooperation of multiple cascade enzymes, and the overall rate depends on the activity of the rate-limiting enzymes. Usually, rate-limiting enzymes with low activity reduce the supply of precursors, leading to the accumulation of intermediates and the reduction of the target NHNPs. Improving the substrate promiscuity can avoid the accumulation of the intermediate products to improve the overall rate. Halotryptamine is a common precursor for the biosynthesis of halogen-containing indole alkaloids. Tryptophan halogenase was introduced to form halotryptophan from L-Trp in plants, and then halotryptophan was transformed into halotryptamine by endogenous tryptophan decarboxylase (TDC). However, TDC showed low activity toward halotryptophan, resulting in reduced halotryptamine production. To solve this problem, Glenn et al.⁷² selected a 7-L-tryptophan halogenase (RebH) from actinomycetes and converted it into 7-tryptamine halogenase. Based on the crystal structure of RebH, they mutated several residues on the active site. Among them, RebH Y455W catalyzed the 7-chlorination of tryptamine instead of the natural substrate tryptophan with higher selectivity. The activity of RebH Y455W toward tryptophan was reduced by 10 times, while the activity toward tryptamine was increased by about 3 times. The change in the substrate specificity of RebH Y455W was interpreted as preventing tryptophan from binding to the active site (Fig. 9D). RebH Y455W was introduced to Catharanthus roseus to produce high value-added 12-chloro-19,20-dihydroakuammicine without accumulating 7-chlorotryptophan.

2.2.4 Engineering the reaction specificity. Generally, wild-type enzymes catalyze the specific reaction on the substrate. Through protein engineering, enzymes can exhibit new catalytic functions on natural substrates to obtain unnatural NHNP derivatives (Fig. 11). 3a-Hydroxyhexahydropyrrolo[2,3-b]indole-2-carboxylic (HPIC) is a common skeleton for many bioactive compounds, which is difficult to obtain by organic synthesis. Wei et al.⁷³ engineered tryptophan 2,3-dioxygenase from Xanthomonas campestris (xcTDO) to synthesize HPIC. Wild-type xcTDO catalyzes the stepwise oxygen insertion of tryptophan to form N′-formylkynurenine. They performed saturation mutation on residues near the substrate and the heme binding pocket of xcTDO. Due to the unstable transition state, the xcTDO mutant F51M/Q127Y lost the original activity of inserting the second oxygen atom into L-tryptophan. Instead, it catalyzed the stereoselective cyclization of L-tryptophan to generate HPIC with >3000 turnover numbers (Fig. 11A). Li et al.⁷⁴ performed two rounds of mutations (2–4 residues per round) at 12 active site residues of CYP102A1 (P450_BM3) from Bacillus megaterium, which constructed a library containing 48 mutants. In addition to the original C–H bond oxidation, some mutants exhibited other oxidase activities, including aromatization, C–C bond formation, and desaturation. With these mutants, 35 oxidized derivatives of tetrahydroquinoline were obtained using tetrahydroquinoline, 2-methyltetrahydroquinoline, and 8-methyltetrahydroquinoline as the substrates (Fig. 11B). Some NHNPs were synthesized from two or more precursors mediated by multiple enzymes, which restricts the substrate conversion. Protein engineering can convert monofunctional enzymes into bifunctional enzymes to balance the supply of different precursors. 3-Hydroxynorcoclaurine is a precursor of isoquinoline natural products, which is formed by the spontaneous Pictet–Spengler condensation of dopamine and 3,4-dihydroxyphenylacetaldehyde (DHPAA). These two compounds are formed from the common precursor L-3,4-dihydroxyphenylalanine (L-DOPA) catalyzed by L-3,4-dihydroxyphenylalanine decarboxylase (DDC) and 3,4-dihydroxyphenylacetaldehyde synthase (DHPAAS), respectively. The increase in 3-hydroxynorcoclaurine production depended on the balance of dopamine and DHAPP, which was hard to achieve with natural DDC and DHAPPS. Vavricka et al.⁷⁵ engineered DHAPPS from Bombyx mori into a bifunctional enzyme with both DDC and DHPAAS activities to control the balance between dopamine and DHAPP. They compared the structure of DHAPPS with multiple DDCs and determined that the key residue N192 in the active site affected the activity. They also found two conserved residues (Y79-F80 in DDC and F79-Y80 in DHAPPS), which may affect the activity. Thus, they designed a mutant F79Y/Y80F/N192H of DHPAAS, which showed bifunctional switching between DDC and DHAPPS (Fig. 11C), and the conversion rate of L-DOPA into 3-hydroxynorcoclaurine was increased from 16% to 23%. Protein engineering can also change the regioselectivity of the enzymes. Prenylated tryptophan-containing cyclic dipeptides are the precursors of many compounds with pharmacological activities. Only a few prenyltransferases can catalyze the C4, C6, and C7-geranylation of the indole ring with geranyl diphosphate as the only prenyl donor. Liao et al.⁷⁶ identified the gatekeeper residues that affected the size of the substrate binding pocket by structure-based alignments of multiple prenyltransferases. Then, they replaced these residues with glycine and generated a series of mutants that regioselectively geranylated multiple L-tryptophan-containing cyclic dipeptides. Notably, all seven possible positions of the indole ring can be attached by the geranyl moiety through those mutants (Fig. 11D).

Fig. 11 Engineering the reaction specificity to synthesize NHNPs and their derivatives. (A) xcTDO. (B) CYP102A1. (C) DHAPPS. (D) Prenyltransferase.

2.3 Multi-enzyme cascade strategy for the biosynthesis of NHNPs

Multi-enzyme cascade systems can convert commercially available raw materials into high-value NHNPs with ideal stereospecificity, which are difficult to achieve by chemical strategies. These multi-enzyme cascade systems usually contain three types of enzymes, namely, precursor synthesis enzymes, N-heterocycle synthesis enzymes, and post-modification enzymes. Multi-enzyme cascade systems can combine a variety of enzymes that cannot be combined by natural biosynthetic pathways. Multiple enzymatic reactions can be carried out simultaneously in the same reactor. By changing the recipe of enzymes and optimizing the reaction conditions, a large number of NHNPs have been efficiently synthesized. Recently, many researchers have combined norcoclaurine synthase (NCS), transaminase, and IRED with other precursor synthesis enzymes and post-modification enzymes to construct multi-enzyme cascades for the synthesis of NHNPs.

2.3.1 Enzyme cascade containing NCS. As described above, NCSs form isoquinoline skeletons in the natural pathway via Pictet–Spengler reaction. Generally, NCSs and post-modification enzymes are used to build cascade reactions. The former converts commercially available raw materials into isoquinoline skeletons, and the latter modifies these skeletons to form high-value compounds (Fig. 12). In 2015, Lichman et al.⁷⁷ designed a one-pot, two-enzyme, three-step chemoenzymatic strategy to synthesize (S)-tetrahydroprotoberine. First, a high-active transaminase Cv2025 from Chromobacterium violaceum was screened to convert dopamine to 3,4-dihydoxyphenylacetaldehyde with the addition of pyruvate. These two compounds were condensed by NCS to form (S)-norlaudanosoline. They optimized the concentration of dopamine and enzymes to increase the conversion from 74% to 87%. (S)-Norlaudanosoline was further Pictet–Spengler cyclized by formaldehyde to form 2,3,10,11-tetrahydroxyberbine with 47% conversion and >95% e.e (Fig. 12A). Subsequently, NCS from Thalictrum flavum (TfNCS) with good substrate promiscuity and stereoselectivity has become a common enzyme in cascade reactions. Bonamore et al.⁷⁸ designed a TfNCS-containing enzyme cascade to synthesize substituted THIQs. First, a variety of aliphatic and aromatic amines were transformed into the corresponding aldehydes by diamine oxidase from Lathyrus cicero (LCAO). Then, TfNCS stereoselectively converted aldehydes and dopamine to (S)-THIQs. In addition to optimizing the concentration of reactants and enzymes, they also determined the optimal pH (7.0). They added appropriate amounts of phenylhydrazine and ascorbic acid to prevent side reactions and finally obtained a series of (S)-THIQs with >95% conversion and >99% e.e (Fig. 12B). TfNCS was also used to synthesize multi-substituted THIQs. Erdmann et al.⁷⁹ designed a three-enzyme cascade to stereoselectively synthesize 1,3,4-trisubstituted THIQs, which have three chiral centers. Firstly, E. coli acetohydroxy acid synthase I catalyzed the carbonylation of 3-hydroxybenzaldehyde to form (R)-1-hydroxy-1-(3-hydroxyphenyl)propan-2-one. It was then converted by transaminase Cv2025. The transamination product and phenylacetaldehyde were condensed by TfNCS A79I, which showed the highest total conversion (88%) in the synthesis of (1S,3S,4R)-1-benzyl-3methyl-1,2,3,4-THIQ-4,6-diol (Fig. 12C). However, NCSs activity is strictly dependent on C6-hydroxyl and C8a-methylene; it is difficult to produce some high-value isoquinoline products containing other substituents on C6 or C8a. To solve this problem, Yang et al.⁸⁰ designed an IRED-containing cascade system to synthesize phenylisoquinoline derivatives that are difficult to obtain in NCS cascade systems. First, a series of chemically synthesized 1-phenyl and 1-benzyl-6,7-dimethoxy-dihydroisoquinoline precursors were reduced by imine reductase IR45 to form the corresponding (S)-THIQs, which were subsequently transmethylated by coclaurine N-methyltransferase to form high-value phenylisoquinoline derivatives. IR45 was structurally modeled, and F190 and W191 on the side of the active site were mutated to bind with bulkier substrates. Mutant F190L/W191F showed the highest conversion to bulkier substrates and two other dihydroisoquinoline medical precursors. This two-enzyme cascade can completely convert a series of dihydroisoquinoline precursors into N-methyl-(S)-THIQs (Fig. 12D).

Fig. 12 Enzyme cascade constructed to synthesize isoquinoline compounds. (A), (B), and (C) Construction of the NCS-containing enzyme cascade to synthesize isoquinoline compounds. (D) Construction of the IRED-containing enzyme cascade to modify isoquinoline compounds.

2.3.2 Enzyme cascade containing transaminase/IRED. Transaminase and IRED-containing biocatalytic cascades can enantioselectively produce chiral mono- and disubstituted piperidines and pyrrolidines. Transaminase and IRED can form cyclic imines, which are converted into high-value derivatives with different substituents by reductase or reducing reagents (Fig. 13). France et al.⁸¹ designed a cascade reaction involving carboxylic acid reductase (CAR), IRED and ω-transaminase to convert keto acids or keto aldehydes into a series of piperidine or pyrrolidine derivatives. First, Mycobacterium marinum CAR was selected to reduce many carboxylic acids to aldehydes, which were then converted to cyclic imines by the commercially available transaminase ATA-113 (Fig. 13A). To ensure the full conversion of aldehydes, DL-alanine and glucose dehydrogenase/lactate dehydrogenase system were introduced, which can control the transaminase equilibrium. Finally, the selected (R)-IRED and (S)-IRED were introduced in the cascade reaction to form the reduction products. They obtained pyrrole/piperidine derivatives containing four different substituents on C2 through this cascade reaction. Costa et al.⁸² designed a two-enzyme cascade reaction to synthesize 2-methyl-5-alkyl pyrrolidines containing one 3- to 9-carbons alkyl chain substituent. Same as that in the previous work, transaminase with the highest activity and stereoselectivity were selected to convert diketone substrates to cyclic imines. However, IREDs showed low conversion (<40%) to these alkyl chain substituted imines. Thus, they tested a class of novel reductases, reductive aminases (RedAms), among which RedAm from Ajellomyces dermatitidis showed the highest conversion rate (>95%). The optimized cascade reaction produced a series of 2-methyl-5-alkylpyrrolidines with >95% conversion and >99% e.e (Fig. 13B). They further replaced ω-transaminase with α,ω-diamine transaminase, which can directly cyclize biological diamines. A cascade reaction containing α,ω-diamine transaminase and lipase was designed to convert diamines and commercially available 3-ketoesters to 2-substituted pyrrolidine/piperidine derivatives.⁸³ YgjG, an α,ω-diamine transaminase from E. coli was selected to convert 1,4-diaminobutane or 1,5-diaminopentane into the corresponding cyclic imines. Simultaneously, a set of commercially available 3-keto acid esters were converted into unstable 3-keto acids by lipases with a wide substrate range. Subsequently, 2-substituted pyrrolidine/piperidine derivatives were formed by the decarboxylation reaction of 3-keto acids and cyclic imines with 50–99% conversion (Fig. 13C).

Fig. 13 Enzyme cascade constructed with transaminase/IRED. (A) Construction of transaminase and IRED-containing enzyme cascade to synthesize piperidine compounds. (B) Construction of transaminase and RedAm-containing enzyme cascade to synthesize pyrrole compounds. (C) Construction of α,ω-transaminase-containing enzyme cascade to synthesize pyrrole and piperidine compounds.

2.3.3 Enzyme cascade containing other key enzymes. Some other enzymes were also employed to synthesize NHNPs (Fig. 8). Fischereder et al.⁸⁴ designed a two-enzyme cascade reaction to synthesize C3-methyl-substituted strictosidine derivatives. Prochiral ketones were first converted to α-methyltryptamines by transaminases with different enantioselectivities, which were subsequently condensed with secologanin by strictosidine synthases-guided Pictet–Spengler reaction to form the piperidine nucleus products. Due to the highest activity, strictosidine synthase from Rauvolfia serpentina was screened and products with >98% diastereomeric excess were obtained (Fig. 14A). Wohlgemuth et al.⁸⁵ used two prenyltransferases from Aspergillus ruber, EchPT1 and EchPT2, to consecutively prenylate all the stereoisomers of cyclo-Trp-Pro and cyclo-Trp-Ala. EchPT1 first catalyzed the reverse C2-prenylation of the cyclic dipeptide substrates on the indole nucleus, which were subsequently accepted by EchPT2 for multiple prenylations and formed 2,5,7-triprenylated derivatives as the major products. Due to the substrate promiscuity of EchPT1 and EchPT2, multiple di-, tri-, and tetraprenylated cyclic dipeptides were produced by this cascade reaction (Fig. 14B). Wang et al.⁸⁶ designed a three-enzyme cascade reaction to synthesize 2-hydroxy-4H-quinolizin-4-one skeletons, including phenylacetate-CoA ligase, malonyl-CoA synthase, and type III PKS. Monosubstituted phenylacetate-CoA and malonyl-CoA were decarboxylatively condensed by HsPKS3 to form the quinolizin products (Fig. 14C).

Fig. 14 Enzyme cascade constructed with other key enzymes. (A) Enzyme cascade constructed with strictosidine synthases. (B) Enzyme cascade constructed with prenyltransferases. (C) Enzyme cascade constructed with type III PKS.

3. Production of NHNPs in microbial cell factory

There are some complex problems that remain to be solved in the process of enzymatic biosynthesis, for example, membrane proteins need to be attached to the membrane for displaying activity, and a number of enzymes rely on expensive cofactors. Microbial cell factories possess complete membrane structure and cofactor synthesis ability, which is an alternative or complementary strategy to synthesize high-value NHNPs.²⁶ One of the dominating obstacles in the heterologous synthesis of NHNPs is the indistinct metabolic pathways and unknown enzymes required for the catalytic reaction. Based on the comprehensive analysis and in-depth mining of multiple omics, including genome, transcriptome, proteome, and metabolome, the natural synthesis pathways of large number of NHNPs, such as morphine, noscapine, and sanguinarine, become gradually unambiguous, which has paved the way for heterologous synthesis in microorganisms. E. coli and Saccharomyces cerevisiae are the most widely adopted heterologous chassis due to the short growth cycle, clear genetic background, and simple culture conditions.⁸⁷ Various metabolic engineering strategies have been utilized to overcome the problem of inefficient production of NHNPs in microbes. Recently, researchers reconstructed the complete biosynthetic pathways in engineered strains by utilizing simple carbon and nitrogen sources for NHNPs production.^8,9,88 Huang et al.⁸⁹ summarized the entire research process of TAs biosynthesis in plants. The discovery process and related study of each enzyme in the pathway were also discussed. In contrast, this review mainly focused on the research progress of TAs and THIQ alkaloids in the past five years, which had made significant breakthroughs, and summarized the strategies of mining new enzymes and optimizing the production of NHNPs in microbial cell factories.

3.1 Deciphering the biosynthesis pathway of NHNPs

The main challenge in the biosynthesis of NHNPs is that the natural synthesis pathways or the enzymes required for NHNPs synthesis have not been completely elucidated; thus, it is impractical to achieve biosynthesis in heterologous hosts.⁹⁰ Focused on this scientific issue, many studies have been carried out. Several strategies are developed for pathway deciphering: (1) co-expression analysis based on the Pearson correlation coefficient of transcriptome data, coupled with sequence alignment and phylogenetic tree analysis, to identify new genes, (2) adjacent genes located in the same gene cluster, especially those with similar expression profiles, are likely to be related to the function and catalyze continuous reactions in the same pathway, (3) transient expression of candidates in Nicotiana benthamiana leaves and adding the substrate exogenously or harnessing the engineered microbe platform strains to verify the catalytic function of the target genes.

3.1.1 TAs. The biogenic routes of TAs that have become explicit along with the key pathway enzymes are gradually being elucidated since the recent years (Fig. 15). The roots of Anisodus acutangulus were dealt with deep RNA sequencing, and four genes, namely, ArAT, CYP80F1, ADH, and DAO encoding aromatic amino acid aminotransferase, littorine mutase/monooxygenase, alcohol dehydrogenase, and diamine oxidase, respectively, in TAs biosynthesis pathway were identified for the first time through BLAST search in multiple databases and path-based enrichment analysis, which laid the foundation for uncovering the TA synthesis pathways.⁹¹ Putrescine, the key precursor of TAs, was synthesized from ornithine catalyzed via ornithine decarboxylase (ODC) or arginine decarboxylase (ADC) in plants. In Atropa belladonna, it was verified that AbODC played a more important role than AbADC in the biosynthesis process of TAs.⁹² When ornithine was utilized as the substrate in vitro, AbODC exhibited higher catalytic efficiency than ODCs from other sources, such as human, Erythroxylum coca, and Nicotiana glutinosa. The overexpression of AbODC, the rate limiting enzyme of polyamine synthesis, significantly accelerated the putrescine accumulation in A. belladonna. Likewise, employing AbODC in the cell factories may greatly promote the production of putrescine and TAs.

Fig. 15 The biosynthesis pathway of TAs. The dashed arrow represents uncharacterized reaction steps. AR, L-arginase; ODC, Ornithine decarboxylase; PMT, putrescine N-methyltransferase; DAO, diamine oxidase; PYKS, pyrrolidine ketide synthase; CYP82M3, tropinone synthase; TRI, tropinone reductase I; TRII, pseudotropine reductase II; AT4, aromatic amino acid aminotransferase; PPAR, phenylpyruvic acid reductase; UGT1, phenyllactate UDP-glycosyltransferase; LS, littorine synthase; CYP80F1, littorine mutase/monooxygenase; H6H, hyoscyamine 6β-hydroxylase.

Tropinone is the simplest plant tropane in structure. The genes related to tropane biosynthesis in A. belladonna were preferentially expressed in the roots, and some candidates were obtained by dissecting the transcriptome of roots.⁹⁰ Combining the strategy of gene silencing and functional verification, two pivotal enzymes, pyrrolidine ketide synthase (AbPYKS), and tropinone synthase (AbCYP82M3), which catalyzed the successive conversion of N-methylpyrrolinium to tropinone, were identified. So far, the biosynthesis pathway of tropinone has been elucidated, paving the way for producing tropinone and derivatives in heterogenous microbes. However, the catalytic mechanism of PYKS is still elusive. RNA-Seq was performed on the hairy roots (where TAs was accumulated) of A. acutangulus, Datura stramonium, and A. belladonna, and 11 putative type III polyketide synthase (PKS) candidates were identified through phylogenetic tree analysis and amino acid sequence alignment.⁹³ By activity verification in vitro, AaPKS2, AbPKS3, and DsPKS1 were effective for the formation of 4-(1-methyl-2-pyrrolidinyl)-3-oxobutanoic acid from N-methylpyrrolinium. Among them, AbPKS3 was the previously reported AbPYKS.⁹⁰ According to the crystal structure of AaPYKS (formerly AaPKS2), it was found that malonyl-CoA produced 3-oxo-glutaric acid under the catalysis of PYKS, and then 3-oxo-glutaric acid reacted with N-methylpyrrolinium spontaneously to generate 4-(1-methyl-2-pyrrolidinyl)-3-oxobutanoic acid. The newly discovered mechanism provided a reference for the synthesis of other plant alkaloids.

Subsequently, tropinone, the molecular skeleton of TAs, was catalyzed by tropine-forming reductase (PtTRI) and pseudotropine-forming reductase (PtTRII) to form tropine and pseudotropine, respectively, in Przewalskia tangutica.⁹⁴ Compared with TRI from Brugmansia arborea,⁹⁵Hyoscyamus niger,⁹⁶ and Datura stramonium,⁹⁵ PtTRI possessed the highest affinity for tropinone. The same function of StTRI and StTRII from Solanum tuberosum was proved.⁹⁷ The reduction ability of TRII was stronger than that of TRI in potato leaves. One interesting phenomenon was that when TRI was disrupted or overexpressed, there was no obvious effect on the accumulation of calystegines, which are the products of the competitive pathway.

Multiple strategies including conserved domain analysis, tissue-specific expression profiles, and phylogenetic tree were adopted to parse the transcriptomes of A. belladonna, and three functional enzymes, namely, phenylpyruvic acid reductase (AbPPAR),⁹⁸ phenyllactate UDP-glycosyltransferase (UGT1), and littorine synthase (LS),⁹⁹ were identified, which characterized the biosynthetic pathway of littorine completely. AbPPAR was the first enzyme discovered in plants to reduce phenylpyruvic acid to phenyllactic acid.⁹⁸ The suppression of AbPPAR in transgenic root cultures disrupted phenyllactic acid production. Tropine and phenyllactylglucose, the glycosylation product of UGT1 from phenyllactate, were condensed via LS to generate littorine.

The chemodiversity of alkaloid was diverse in Mandragora spp. Three common alkaloids, namely, hyoscyamine, anisodamine, and scopolamine, are downstream metabolites of littorine. Scopolamine has the highest market demand due to its fewer side effects and stronger pharmacological activities. For example, it was used as the sedative or in the treatment of motion sickness. The successive hydroxylation and epoxidation of hyoscyamine to anisodamine and scopolamine were achieved by the catalysis of hyoscyamine 6β-hydroxylase (H6H). The functional inactivation of H6H by the substitution of a catalytic residue Gly220 to Cys in M. officinarum led to the accumulation of hyoscyamine and the lack of anisodamine and scopolamine.¹⁰⁰ The polymorphism of TAs content was conducive for understanding the catalytic mechanism of crucial enzymes and guiding further rational modification. In the hairy root cultures of Scopolia lurida, the accumulation of scopolamine and anisodamine were significantly improved by overexpressing H6H from Hyoscyamus niger (HnH6H) than that when introducing H6H from S. lurida (SlH6H).¹⁰¹ HnH6H may also be more suitable for scopolamine production in microbial cell factories.

3.1.2 THIQ alkaloids. Benzylisoquinoline alkaloids (BIAs) and phenethylisoquinoline alkaloids are the representative compounds of THIQ alkaloids (Fig. 16). Opium poppy is the sole commercial source of tyrosine-derived BIAs, such as thebaine, codeine, morphine, and noscapine. During the biosynthesis of thebaine, (7S)-salutaridinol-7-O-acetate (7-O-acetylsalutaridinol) can be spontaneously transformed into thebaine at pH 8–9, accompanied by a large number of labile hydrogenated byproducts.¹⁰² However, these byproducts were not detected in opium poppy, suggesting that there may be another pathway to produce thebaine in vivo. With (7S)-salutarinol as the substrate in vitro, alkaloid- and protein-abundant opium poppy latex together with SalAT could increase the yield of thebaine by 10 times and reduce the formation of byproducts, compared with SalAT alone.¹⁰³ Six candidate proteins were identified through the analysis of proteomics and opium poppy latex transcriptome. Among them, only Bet v1-1, termed as thebaine synthase (THS), belonging to the pathogenesis-related 10 protein (PR10) superfamily, could convert (7S)-salutaridinol-7-O-acetate into thebaine in vitro. The gene encoding THS is located in a novel gene cluster, which is responsible for transforming (S)-reticuline to thebaine. The expression of THS in genome and episomal plasmid increased the thebaine titer by 24-fold compared with spontaneous transformation in S. cerevisiae. The discovery of THS is of great significance to improve the production of thebaine in engineered yeast.

Fig. 16 The biosynthesis pathway of THIQs. CYP76AD5, tyrosine hydroxylase; Aro8, aromatic aminotransferase I; Aro9, aromatic aminotransferase II; DODC, DOPA decarboxylase; Aro10, phenylpyruvate decarboxylase; NCS, norcoclaurine synthase; 6OMT, norcoclaurine 6-O-methyltransferase; CNMT, coclaurine N-methyltransferase; CYP80B3, (S)-N-methylcoclaurine 3′-hydroxylase isozyme; 4′-OMT, 3′-hydroxy-N-methylcoclaurine 4′-omethyltransferase; REPI, reticuline epimerase; SalSyn, salutaridine synthase; SalR, salutaridine reductase; SalAT, salutaridine 7-O-acetyltransferase; THS, thebaine synthase; CODM, codeine 3-O-demethylase; T6ODM, thebaine 6-O-demethylase; COR, codeinone reductase; NISO, neopinone isomerase.

To promote the biosynthesis of thebaine derivatives, a new codeinone reductase (COR) isoform, COR-B, exhibited better catalytic performance than the previously characterized isoforms.¹⁰⁴ The reversible conversion between codeine and codeinone, morphine and morphinone, and the irreversible synthesis of neopine or neomorphine from neopinone or neomorphinone, respectively, could be achieved by COR with the dependence on NADP(H). It was found that the expression level of PR10-3 was the highest in the proteome and transcriptome of three opium poppy chemotypes and COR-coupled enzymatic reaction identified PR10-3 as a neopinone isomerase (NISO) to catalyze the isomerization of neopinone and codeinone, neomorphinone and morphinone, which were considered as spontaneous reactions previously.¹⁰⁵ The introduction of NISO into the in vitro system or engineering yeast containing T6ODM and COR-B promoted the biosynthesis of codeine and morphine from thebaine. Different from the spontaneous transformation, the recently identified THS¹⁰³ and NISO¹⁰⁵ significantly promoted the synthesis of products, which is beneficial to the heterologous production of thebaine derivatives.

In the synthesis branch of noscapine, four genes catalyzing the generation of noscapine from (S)-reticuline were located in a ten-gene cluster in the genome of opium poppy, and the transcription of the gene cluster was associated with the accumulation of noscapine.¹⁰⁶ The catalytic functions of CYP82X1, CYP82X2, AT1, and CXE1, whose encoding genes resided in the same gene cluster, were characterized by utilizing the substrates formed in the previous steps, thus clarifying the biosynthetic routes of noscapine.¹⁰⁷ Virus-induced gene silencing was further harnessed to confirm the roles of these four genes in opium poppy. Acetylation introduced by AT1 acted as a protective group to hinder hemiacetal generation, which was necessary for the oxidation of CYP82X1.

Biosynthetic gene cluster of noscapine and the genes responsible for converting (S)-reticuline to thebaine were located on chromosome 11 in the draft genome of opium poppy, which were co-expressed in stems.¹⁰⁸ Similar results were also discovered in which the expression patterns of the genes distributed in close proximity on the chromosomes tended to be similar,¹⁰⁹ and the function of genes in the same gene cluster might be related.¹⁰⁶ Mining key enzymes based on gene clusters is an effective strategy to parse the unknown metabolic pathways.

The biosynthesis pathways of numerous bioactive molecules from medicinal plants are still a mystery. Until recently, the nearly complete synthetic pathway of phenethylisoquinoline colchicine, a fascinating molecule used to treat gout and anti-inflammatory diseases, has been deciphered, which provided an effective strategy for pathway discovery.¹¹⁰ Each candidate methyltransferase or cytochrome P450 identified from Gloriosa superba transcriptome was transiently expressed in N. benthamiana leaves individually, and soaked in chemically synthesized substrate. The leaf extracts were analyzed to verify the function of the enzyme. The mined enzymes were gradually superimposed for the next round of screening, and the de novo synthesis of N-formyldemecoline in N. benthamiana was realized by combining the upstream module of the 1-phenethylisoquinoline scaffold. The residual three steps from N-formyldemecoline to colchicine remain to be illuminated.

3.2 Heterologous biosynthesis of NHNPs in microbial cell factory

Various strategies were combined to promote the efficient synthesis of NHNPs in microbial cell factory. Taking TAs and BIAs as examples, the techniques employed to boost the production mainly include: (A) enriching precursors through the overexpression of upstream enzymes or rate-limiting enzymes, introducing heterologous pathways or blocking the competition pathways; (B) enhancing the enzyme activities including codon optimization of heterologous enzymes, introduction of more efficient enzymes, enhancement of cofactors pool, or truncation of plant P450 signal peptide to modify the function; (C) identifying and optimizing the metabolic bottlenecks to regulate intracellular metabolic flux distribution; (D) modular stepwise fermentation and integrating transporters to promote the transport of intermediates in the co-culture (Fig. 17, Table 1).

Fig. 17 The strategies for optimizing the biosynthesis of NHNPs in the microbial cell factory. (A) Enriching the supply of precursors. (B) Enhancing the enzyme activities. (C) Identifying and optimizing the metabolic bottlenecks. (D) Modular stepwise fermentation.

Table 1 The summary of the production of TAs and BIAs in the microbial cell factory

Species	Products	Structures	Engineering strategy	Titer	References
E. coli	N-Methylpyrrolinium		Screening enzymes with high activities	3.02 mg L^−1	111
S. cerevisiae	N-Methylpyrrolinium		Screening enzymes with high activities, disruption of the competitive pathways, overexpression of the key genes	17.82 mg L^−1	111
S. cerevisiae	Tropine		Comparison of the catalytic performance of enzymes, enrichment of the precursor	0.13 mg L^−1	112
	Pseudotropine			0.08 mg L^−1
S. cerevisiae	Putrescine		Optimization of endogenous and exogenous pathways, disruption of the competitive pathways, recognition of the metabolic bottlenecks	34 mg L^−1	113
	Tropine			5.9 mg L^−1
	Cinnamoyltropine			6.0 μg L^−1
S. cerevisiae	(S)-Norcoclaurine		Optimization of the gRNAs and genomic loci, synthetic DNA landing pad (LP) system	130 μg L^−1	114
S. cerevisiae	(S)-Reticuline		Disruption of the side reactions, overexpression of the key enzymes, introduction of more efficient enzymes	4.6 g L^−1	115
E. coli	Thebaine		The N-terminus truncation of plant-derived P450s, selection of appropriate CPR, addition of heme precursor, stepwise fermentation	2.1 mg L^−1	116
S. cerevisiae	Noscapine		Optimization of N-terminal tags, promoters and culture temperatures, upregulation of the expression level of rate-limiting enzymes	1.64 μM	117
S. cerevisiae	Noscapine		Truncation of the N-terminal signal peptide, overexpression of the bottleneck enzymes, increasement of NADPH pool, optimization of fermentation conditions	2.21 mg L^−1	118

---

3.2.1 TAs. N-Methylpyrrolinium (NMP) was the critical intermediate metabolite in the synthesis of TAs, for instance, scopolamine, calystegine, cocaine, and nicotine. After comparing the kinetic parameters, three enzymes EcODC, AtPMT, and AaDAO3 from different plants were chosen to realize the synthesis of NMP in E. coli and S. cerevisiae, with the production of 3.02 and 2.07 mg L⁻¹, respectively.¹¹¹ Some efforts were made to further increase the titer in S. cerevisiae. Inactivating the aldehyde dehydrogenase ALD4, ALD5, and hexadecenal dehydrogenase in the competitive pathways and overexpressing SAM2 to enhance the supply of SAM, a necessary cofactor for PMT, resulted in a significant raise in NMP production, with the titer of 17.82 mg L⁻¹. Based on these results, three additional enzymes, AaPYKS, AaCYP82M3, AaTRI or AaTRII, from A. acutangulus were introduced into the NMP-producing strains to achieve the biosynthesis of tropine and pseudotropine.¹¹² AaTRI presented better catalytic performance than the outstanding PtTRI found in previous studies.⁹⁴ In order to enrich the precursor malonyl-CoA pool, an additional copy of endogenous acetyl-CoA carboxylase (ACC1) was integrated. Ultimately, the accumulation of tropine and pseudotropine reached 0.13 and 0.08 mg L⁻¹, respectively, in engineered S. cerevisiae. In the subsequent in-depth study, platform strains for the synthesis of TAs were reconstructed utilizing multi-target strategies.¹¹³ Combined with endogenous and exogenous metabolic pathways, putrescine titer was increased to 34 mg L⁻¹. The optimization of the polyamine regulation mechanism eliminated the competitive inhibition between ornithine decarboxylase (Spe1p) and antizyme-1 (Oaz1p), and weakened the alkylation of putrescine, which provided enough precursor putrescine and resulted in a 71-fold increase. Similar to previous studies, the disruption of all ALDHs boosted the NMP production by 76%. Identifying and engineering the metabolic bottlenecks catalyzed by PMT and PYKS in conjunction with fermentation optimization further improved the titer of tropine to 5.9 mg L⁻¹. The introduction of acyltransferase from E. coca achieved non-canonical cinnamoyltropine production with the output of 6.0 μg L⁻¹. These engineered strains offered a powerful platform to elucidate the downstream metabolic pathways of tropine alkaloids.

3.2.2 Benzylisoquinoline alkaloids. With the advance of metabolic engineering and synthetic biology, CRISPR/Cas9 has become a powerful tool for genome-editing¹¹⁹ and expression regulation.¹²⁰ Synthetic DNA landing pad (LP) system, mediated by CRISPR/Cas9, was developed to precisely control the copy numbers of the target genes from one to four in S. cerevisiae.¹¹⁴ After characterization and optimization, suitable gRNAs and genomic loci were chosen to constitute the LPs system. Pictet–Spengler condensation between dopamine and 4-HPAA was the major rate-limiting step in BIA biosynthesis.¹¹⁵ Ten orthologous enzymes of NCS, which performed inefficient catalysis during BIA production, were integrated into LP1–4 motifs, generating a 40-strain NCS library. NCS from Nandina domestica or Sanguinaria canadensis possessed prominent performance, and the yield of (S)-norcoclaurine showed an upward trend as the copy number of NCS increased from one to four. The production of (S)-norcoclaurine was improved to 130 μg L⁻¹ by increasing the NCS copy number. Multicopy gene integration completed by the LPs system enabled the rapid construction of the mutant libraries with different expression levels of the candidates to optimize the microbial synthesis of NHNPs.

(S)-Norcoclaurine undergoes 4-step continuous methylation and hydroxylation to form (S)-reticuline, the key precursor of BIAs. Pyne et al.¹¹⁵ established a high-yield yeast platform with a titer of 4.6 g L⁻¹ (S)-reticuline for producing natural or synthetic THIQ alkaloids. Several efforts have been made for the efficient production of (S)-reticuline. Five NADPH-dependent reductases and dehydrogenases were knocked out to prevent 4-HPAA from generating byproducts 4-hydroxyphenylacetic acid (4-HPAC) or tyrosol through oxidation or reduction reaction, respectively. However, there were still other unrecognized enzymes that catalyzed the aforementioned side reactions. Single gene knockout identified that Gre2 or Hfd1 could reduce or oxidize 4-HPAA to produce fused products. The simultaneous disruption of gre2 and hfd1 destroyed more than 80% of 4-HPAC and tyrosol synthesized intracellularly. The overexpression of enzymes in the upstream pathways or the introduction of more efficient enzymes, such as NCS from Coptis japonica (CjNCS) and CYP76AD5, provided sufficient precursors, 4-HPAA and dopamine, and further dramatically increased (S)-reticuline production. Utilizing the promiscuity of NCS, the condensation of endogenous amino acids or exogenous added amino acids with dopamine can realize the de novo synthesis of diverse non-canonical THIQs, enriching the structure of THIQs and broadening the application range of THIQ products.

Many types of BIAs were derived from (S)-reticuline, for instance, thebaine, noscapine, sanguinarine, and magnoflorine. During the biosynthesis of thebaine, which originated from L-tyrosine, E. coli with a high amino acid productivity was selected as the host. The N-terminus truncation of plant-derived P450s, including STORR and SalS, the selection of appropriate CPR, and the addition of the heme precursor, 5-aminolevulinic acid (5-ALA), were performed to improve the catalytic performance of P450s.¹¹⁶ Tyrosinase, responsible for transforming L-tyrosine to levodopa (L-DOPA), was able to degrade tetrahydropapaveroline (THP), which was the intermediate during thebaine production. The content of (R,S)-THP would influence the composition of R/S-reticuline, and the high concentration of (R,S)-THP reduced (R)-reticuline accumulation. In the process of thebaine production in E. coli, IPTG was found to possess an opposite effect on the synthesis of (R,S)-reticuline and thebaine. Given the interrelationship between the above pathways, the biosynthetic pathway of thebaine from glycerol was divided into four modules, with each module completed via different E. coli strains in a stepwise fermentation manner. Stepwise culture method divided the whole metabolic pathway into several parts and allowed each module to be optimized separately, increasing the thebaine titer to 2.1 mg L⁻¹. The introduction of corresponding downstream pathways could also achieve the efficient synthesis of (S)-reticuline derivatives.

Noscapine, a potential anticancer medicine, is another typical BIAs. The identification of a ten-gene cluster in charge of noscapine biosynthesis in the opium poppy genome provided a pivotal foundation for the de novo synthesis in S. cerevisiae.¹⁰⁷ The incorporation of heterogeneous pathways enabled the bioconversion of (S)-canadine to noscapine.¹¹⁷ Functional O-methyltransferases from plant sources were usually dimers. Consistent with the previous characterization results in vitro, monomer PsMT2 or PsMT3 was inactive when acting on narcotoline alone but their simultaneous expression might act as a heterodimer to methylate the 4′-O of narcotoline to form noscapine. The optimization of N-terminal tags, promoters, and culture temperatures, combined with the upregulating of the expression level of the rate-limiting enzymes, ultimately increased noscapine production to 1.64 μM from norlaudanosoline. In-depth studies were carried out sequentially to improve the yield. The introduction of 25 heterologous enzymes and 6 endogenous enzymes reconstituted the complete synthesis pathway of noscapine.¹¹⁸ In order to enhance the metabolic flux of the product, different modifications were implemented to different rate-limiting enzymes, such as truncating the N-terminal signal peptide of NCS, adopting codon-optimized tyrosine hydroxylase (TyrH) mutant, and overexpressing the bottleneck enzymes CYP82X2 and S9OMT. NADPH was the electron donor of plant cytochrome P450s, and increased the NADPH pool, contributing to enhanced catalytic efficiency. The overexpression of some endogenous genes for NADPH regeneration, either individually or in combination, significantly promoted noscapine production. The optimization of medium composition and fermentation conditions was a basic but effective strategy to boost the product synthesis.¹²¹ Under the optimal conditions, the yield of noscapine of up to 2.21 mg L⁻¹ was achieved, which was increased by 300 times. Feeding tyrosine derivatives and utilizing the promiscuity of enzymes realized the biosynthesis of halogenated BIAs, which extended the application of the platform strains and offered an alternative approach for the synthesis of synthetic compounds.

For products with long and complicated metabolic pathways, modular co-culture engineering strategy has been widely utilized to alleviate the heavy burden of multi-enzyme expression on the host or to avoid inhibition or competition among the pathways.¹²² Efficient transport of intermediates is critical for a multistrain co-culture. The transportation of berberine in Coptis japonica was mediated by the ATP-binding cassette (ABC) protein family including CjMDR1 (ref. 123) and CjABCB2,¹²⁴ and multiantimicrobial extrusion protein (MATE) family containing CjMATE1.¹²⁵ However, the transporter of opiate alkaloids in opium poppy is still elusive. In the gene cluster responsible for thebaine synthesis, there was a putative purine uptake permease (PUP), a transporter with affinity for alkaloids, which was termed as BUP1 pseudogene.¹²⁶ Eight additional BUPs (BUP2-BUP9) were identified in combination with transcriptome mining and sequence similarity analysis, among which, BUP1 localized on the plasma membrane and was verified to be capable of transporting a wide range of alkaloids in engineered yeast. In the three engineering yeast strains, the biosynthesis of codeine from DOPA was accomplished in a stepwise culture way, and the overexpression of BUP1 promoted the conversion significantly. The proposed transportation mechanism of BUP1 was to translocate the intermediates synthesized by the sieve element to the laticifer. In the stepwise fermentation mediated by multiple strains, introducing BUP1 would facilitate the transfer of the intermediates among different strains, thus promoting the synthesis of opiate alkaloids. There are plenty of transporters that remain to be parsed.

4. Conclusions and prospective

In summary, many biosynthetic pathways have been elucidated with the development of multi-omics and bioinformatics. A large number of new NHNPs are discovered annually, which are potential materials for manufacturing new food additives, pesticides, and drugs. Biosynthesis offers an alternative and efficient method for obtaining NHNPs, which can avoid large-scale planting, save land resources, and improve economic efficiency. Moreover, the biosynthesis of NHNPs from biomass resources can convert inorganic carbon sources such as CO₂ into high-value products and promote carbon balance. Despite the breakthroughs that have been made in enzyme engineering and microbial cell factory, the yield is still low and far from meeting the market demand. First, compared with chemical methods, only a few types of nitrogen-heterocyclic skeletons can be obtained by enzymatic methods. Enzymes that synthesize morpholine, pyrazole, tri/tetrazole, tri/tetrazine, and other high-value N-heterocyclic skeletons are still rarely reported. Second, most of the reported enzymes are not stable enough to meet industrial needs, which prevents them from showing higher benefits than chemical methods. Thirdly, the synthesis pathways of many NHNPs have not been fully elucidated, and researchers cannot find alternative enzymes to synthesize the target products. Today, researchers are committed to employing multiple strategies to optimize the synthesis, for example, utilizing retro-biosynthesis algorithm to predict the synthesis route of compounds, metabolic engineering coupled with a metabolic network model to optimize the metabolic flux, and biosensors for the dynamic regulation of metabolic processes. The biosynthesis of NHNPs will be further combined with big data and artificial intelligence technologies to meet the needs of higher data analysis and information integration. By designing high-throughput screening and detection methods, more enzymatic synthesis platforms and microbial cell factories will be constructed to achieve the efficient biosynthesis of NHNPs.

5. Conflicts of interest

There are no conflicts to declare.

6. Acknowledgements

This work was supported by the National Key Research and Development Program of China (2018YFA0901800), the Natural Science Foundation of China (No. 21736002, and 21878021), and the Beijing Nova Program of Science and Technology (No. Z191100001119099), Beijing Natural Science Foundation (No. M21010 and 2212017).

7. References

1. C. G. Acevedo-Rocha, L. S. Gronenberg, M. Mack, F. M. Commichau and H. J. Genee, Curr. Opin. Biotechnol., 2019, 56, 18–29.
2. T. Iyanagi, Biochim. Biophys. Acta Bioenerg., 2019, 1860, 233–258.
3. T. J. Titcomb and S. A. Tanumihardjo, Compr. Rev. Food Sci. Food Saf., 2019, 18, 1968–1984.
4. J. A. Joule, in Heterocyclic Chemistry In the 21st Century: A Tribute To Alan Katritzky, ed. E. F. V. Scriven and C. A. Ramsden, Elsevier Academic Press Inc, San Diego, 2016, vol. 119, pp. 81–106.
5. K. W. Wang and X. J. Gao, Mini Rev. Org. Chem., 2016, 13, 402–409.
6. S. S. Zhou and J. G. Jiang, Curr. Med. Chem., 2019, 26, 1833–1848.
7. V. Amirkia and M. Heinrich, Phytochem. Lett., 2014, 10, 48–53.
8. S. Schlager and B. Drager, Curr. Opin. Biotechnol., 2016, 37, 155–164.
9. A. Diamond and I. Desgagne-Penix, Plant Biotechnol. J., 2016, 14, 1319–1328.
10. L. Chen, Q. Y. Zhang, M. Jia, Q. L. Ming, W. Yue, K. Rahman, L. P. Qin and T. Han, Crit. Rev. Microbiol., 2014, 42, 454–473.
11. K. Gademann and C. Portmann, Curr. Org. Chem., 2008, 12, 326–341.
12. S. Asamizu, Biosci. Biotechnol. Biochem., 2017, 81, 871–881.
13. C. Hartley, F. Moultrie, A. Hoskin, G. Green, V. Monk, J. L. Bell, A. R. King, M. Buckle, M. van der Vaart, D. Gursul, S. Goksan, E. Juszczak, J. E. Norman, R. Rogers, C. Patel, E. Adams and R. Slater, Lancet, 2018, 392, 2595–2605.
14. B. Muller-Krampe, M. Oberbaum, P. K. Dipl-Math and M. Weiser, Pediatr. Int., 2007, 49, 328–334.
15. S. Gezici and N. Sekeroglu, Anticancer Agents Med. Chem., 2019, 19, 101–111.
16. T. V. Sravanthi and S. L. Manju, Eur. J. Pharm. Sci., 2016, 91, 1–10.
17. V. Sharma, P. Kumar and D. Pathak, J. Heterocycl. Chem., 2010, 47, 491–502.
18. A. D. Borthwick, Chem. Rev., 2012, 112, 3641–3716.
19. Z. X. Qing, P. Yang, Q. Tang, P. Cheng, X. B. Liu, Y. J. Zheng, Y. S. Liu and J. G. Zeng, Curr. Org. Chem., 2017, 21, 1920–1934.
20. B. K. Hong, H. H. Li, J. B. Wu, J. Zhang and X. G. Lei, Angew. Chem., Int. Ed., 2015, 54, 1011–1015.
21. P. Q. Huang, S. Y. Huang, L. H. Gao, Z. Y. Mao, Z. Chang and A. E. Wang, Chem. Commun., 2015, 51, 4576–4578.
22. Y. M. Huang, Y. Liu, C. W. Zheng, Q. W. Jin, L. Pan, R. M. Pan, J. Liu and G. Zhao, Chem.–Eur. J., 2016, 22, 18339–18342.
23. S. Y. Xu, J. Zhang, D. H. Ma, D. Y. Xu, X. G. Xie and X. G. She, Org. Lett., 2016, 18, 4682–4685.
24. Q. Zhang, F. M. Zhang, C. S. Zhang, S. Z. Liu, J. M. Tian, S. H. Wang, X. M. Zhang and Y. Q. Tu, J. Org. Chem., 2019, 84, 12664–12671.
25. N. Gerhards, L. Neubauer, P. Tudzynski and S. M. Li, Toxins, 2014, 6, 3281–3295.
26. C. Dziggel, H. Schafer and M. Wink, Biotechnol. J., 2017, 12, 1600145.
27. S. Kishimoto, M. Sato, Y. Tsunematsu and K. Watanabe, Molecules, 2016, 21, 1078.
28. T. Awakawa and I. Abe, Beilstein J. Org. Chem., 2019, 15, 1545–1551.
29. P. J. Facchini and J. S. Morris, Front. Plant Sci., 2019, 10, 1058.
30. X. Liu, X. Zhu, H. Wang, T. Liu, J. Cheng and H. Jiang, Synth. Syst. Biotechnol., 2020, 5, 187–199.
31. A. Miyanaga, F. Kudo and T. Eguchi, Nat. Prod. Rep., 2018, 35, 1185–1209.
32. A. Nivina, K. P. Yuet, J. Hsu and C. Khosla, Chem. Rev., 2019, 119, 12524–12547.
33. C. Beck, J. F. G. Garzon and T. Weber, Biotechnol. Bioprocess Eng., 2020, 25, 886–894.
34. M. J. Jaremko, T. D. Davis, J. C. Corpuz and M. D. Burkart, Nat. Prod. Rep., 2020, 37, 355–379.
35. P. D. Walker, A. N. M. Weir, C. L. Willis and M. P. Crump, Nat. Prod. Rep., 2021, 38, 723–756.
36. A. Zipperer, M. C. Konnerth, C. Laux, A. Berscheid, D. Janek, C. Weidenmaier, M. Burian, N. A. Schilling, C. Slavetinsky, M. Marschal, M. Willmann, H. Kalbacher, B. Schittek, H. Brotz-Oesterhelt, S. Grond, A. Peschel and B. Krismer, Nature, 2016, 535, 511–516.
37. F. Hemmerling and F. Hahn, Beilstein J. Org. Chem., 2016, 12, 1512–1550.
38. M. W. Mullowney, R. A. McClure, M. T. Robey, N. L. Kelleher and R. J. Thomson, Nat. Prod. Rep., 2018, 35, 847–878.
39. C. Gui, Q. L. Li, X. H. Mo, X. J. Qin, J. Y. Ma and J. H. Ju, Org. Lett., 2015, 17, 628–631.
40. P. Borgman, R. D. Lopez and A. L. Lane, Org. Biomol. Chem., 2019, 17, 2305–2314.
41. M. Gondry, I. B. Jacques, R. Thai, M. Babin, N. Canu, J. Seguin, P. Belin, J. L. Pernodet and M. Moutiez, Front. Microbiol., 2018, 9, 46.
42. S. Meng, W. Han, J. Zhao, X. H. Jian, H. X. Pan and G. L. Tang, Angew. Chem., Int. Ed., 2018, 57, 719–723.
43. E. D. James, B. Knuckley, N. Alqahtani, S. Porwal, J. S. Ban, J. A. Karty, R. Viswanathan and A. L. Lane, ACS Synth. Biol., 2016, 5, 547–553.
44. R. Roddan, J. M. Ward, N. H. Keep and H. C. Hailes, Curr. Opin. Chem. Biol., 2020, 55, 69–76.
45. E. Eger, A. Simon, M. Sharma, S. Yang, W. B. Breukelaar, G. Grogan, K. N. Houk and W. Kroutil, J. Am. Chem. Soc., 2020, 142, 792–800.
46. H. Lechner, P. Soriano, R. Poschner, H. C. Hailes, J. M. Ward and W. Kroutil, Biotechnol. J., 2018, 13, 1700542.
47. T. Mori, S. Hoshino, S. Sahashi, T. Wakimoto, T. Matsui, H. Morita and I. Abe, Chem. Biol., 2015, 22, 898–906.
48. Q. Chen, S. F. Zhang and Y. C. Xie, J. Biotechnol., 2018, 281, 137–143.
49. X. Z. Wang, D. K. Kong, T. T. Huang, Z. X. Deng and S. J. Lin, Org. Biomol. Chem., 2018, 16, 9124–9128.
50. K. L. Chen, C. Y. Lai, M. T. Pham, R. J. Chein, Y. Tang and H. C. Lin, Org. Lett., 2020, 22, 3302–3306.
51. Y. L. Du, H. Y. He, M. A. Higgins and K. S. Ryan, Nat. Chem. Biol., 2017, 13, 836–838.
52. R. C. Simon, B. Grischek, F. Zepeck, A. Steinreiber, F. Belaj and W. Kroutil, Angew. Chem., Int. Ed., 2012, 51, 6713–6716.
53. N. Borlinghaus, S. Gergel and B. M. Nestl, ACS Catal., 2018, 8, 3727–3732.
54. Z. Xu, P. Yao, X. Sheng, J. Li, J. Li, S. Yu, J. Feng, Q. Wu and D. Zhu, ACS Catal., 2020, 10, 8780–8787.
55. H. Renata, Z. J. Wang and F. H. Arnold, Angew. Chem., Int. Ed., 2015, 54, 3351–3367.
56. P. Molina-Espeja, J. Vina-Gonzalez, B. J. Gomez-Fernandez, J. Martin-Diaz, E. Garcia-Ruiz and M. Alcalde, Biotechnol. Adv., 2016, 34, 754–767.
57. F. H. Arnold, Angew. Chem., Int. Ed., 2018, 57, 4143–4148.
58. C. A. Tracewell and F. H. Arnold, Curr. Opin. Chem. Biol., 2009, 13, 3–9.
59. R. E. Cobb, R. Chao and H. M. Zhao, Biotechnol. Adv., 2013, 59, 1432–1440.
60. R. Parra-Cruz, P. L. Lau, H. S. Loh and A. Pordea, J. CO2 Util., 2020, 37, 1–8.
61. L. S. Xu, F. K. Han, Z. Dong and Z. J. Wei, Microorganisms, 2020, 8, 519.
62. Y. D. Cao, Y. C. He, H. Li, G. Y. Kai, J. H. Xu and H. L. Yu, J. Biotechnol., 2015, 211, 123–129.
63. A. Fan and S. M. Li, Appl. Microbiol. Biotechnol., 2016, 100, 5389–5399.
64. K. Brockmeyer and S. M. Li, J. Nat. Prod., 2017, 80, 2917–2922.
65. M. Herger, P. van Roye, D. K. Romney, S. Brinkmann-Chen, A. R. Buller and F. H. Arnold, J. Am. Chem. Soc., 2016, 138, 8388–8391.
66. J. Murciano-Calles, D. K. Romney, S. Brinkmann-Chen, A. R. Buller and F. H. Arnold, Angew. Chem., Int. Ed., 2016, 55, 11577–11581.
67. D. K. Romney, J. Murciano-Calles, J. E. Wehrmuller and F. H. Arnold, J. Am. Chem. Soc., 2017, 139, 10769–10776.
68. C. E. Boville, D. K. Romney, P. J. Almhjell, M. Sieben and F. H. Arnold, J. Org. Chem., 2018, 83, 7447–7452.
69. S. Duewel, L. Schmermund, T. Faber, K. Harms, V. Srinivasan, E. Meggers and S. Hoebenreich, ACS Catal., 2020, 10, 1272–1277.
70. A. Zqdlo-Dobrowolska, L. Hammerer, T. Paykov-Keller, K. Gruber and W. Kroutil, ACS Catal., 2020, 10, 1094–1101.
71. T. R. Valentic, J. T. Payne and C. D. Smolke, ACS Catal., 2020, 10, 4497–4509.
72. W. S. Glenn, E. Nims and S. E. O'Connor, J. Am. Chem. Soc., 2011, 133, 19346–19349.
73. Y. X. Wei, C. Lu, S. S. Jiang, Y. Y. Zhang, Q. C. Li, W. J. Bai and X. Q. Wang, Angew. Chem., Int. Ed., 2020, 59, 3043–3047.
74. Y. S. Li and L. L. Wong, Angew. Chem., Int. Ed., 2019, 58, 9551–9555.
75. C. J. Vavricka, T. Yoshida, Y. Kuriya, S. Takahashi, T. Ogawa, F. Ono, K. Agari, H. Kiyota, J. Li, J. Ishii, K. Tsuge, H. Minami, M. Araki, T. Hasunuma and A. Kondo, Nat. Commun., 2019, 10, 2015.
76. G. Liao, P. Mai, J. Fan, G. Zocher, T. Stehle and S. M. Li, Org. Lett., 2018, 20, 7201–7205.
77. B. R. Lichman, E. D. Lamming, T. Pesnot, J. M. Smith, H. C. Hailes and J. M. Ward, Green Chem., 2015, 17, 852–855.
78. A. Bonamore, L. Calisti, A. Calcaterra, O. H. Ismail, M. Gargano, I. D'Acquarica, B. Botta, A. Boffi and A. Macone, Chemistryselect, 2016, 1, 1525–1528.
79. V. Erdmann, B. R. Lichman, J. X. Zhao, R. C. Simon, W. Kroutil, J. M. Ward, H. C. Hailes and D. Rother, Angew. Chem., Int. Ed., 2017, 56, 12503–12507.
80. L. Yang, J. M. Zhu, C. H. Sun, Z. X. Deng and X. D. Qu, Chem. Sci., 2020, 11, 364–371.
81. S. P. France, S. Hussain, A. M. Hill, L. J. Hepworth, R. M. Howard, K. R. Mulholland, S. L. Flitsch and N. J. Turner, ACS Catal., 2016, 6, 3753–3759.
82. B. Z. Costa, J. L. Galman, I. Slabu, S. P. France, A. J. Marsaioli and N. J. Turner, ChemCatChem, 2018, 10, 4733–4738.
83. J. L. Galman, I. Slabu, F. Parmeggiani and N. J. Turner, Chem. Commun., 2018, 54, 11316–11319.
84. E. M. Fischereder, D. Pressnitz and W. Kroutil, ACS Catal., 2016, 6, 23–30.
85. V. Wohlgemuth, F. Kindinger and S. M. Li, Appl. Microbiol. Biotechnol., 2018, 102, 2671–2681.
86. J. Wang, N. Ding, Y. Wu, X. P. Shi, B. W. Qi, X. Liu, X. H. Wang, J. Li, P. F. Tu and S. P. Shi, RSC Adv., 2020, 10, 23566–23572.
87. A. Cravens, J. Payne and C. D. Smolke, Nat. Commun., 2019, 10, 3634.
88. A. Sharma, D. Amin, A. Sankaranarayanan, R. Arora and A. K. Mathur, Biotechnol. Lett., 2020, 42, 11–23.
89. J.-P. Huang, Y.-J. Wang, T. Tian, L. Wang, Y. Yan and S.-X. Huang, Nat. Prod. Rep., 2021.
90. M. A. Bedewitz, A. D. Jones, J. C. D'Auria and C. S. Barry, Nat. Commun., 2018, 9, 5281.
91. L. Cui, F. Huang, D. Zhang, Y. Lin, P. Liao, J. Zong and G. Kai, Mol. Genet. Genomics, 2015, 290, 1367–1377.
92. T. Zhao, S. Li, J. Wang, Q. Zhou and Z. Liao, ACS Synth. Biol., 2020, 9, 437–448.
93. J. P. Huang, C. Fang, X. Ma, L. Wang, J. Yang, J. Luo, Y. Yan, Y. Zhang and S. X. Huang, Nat. Commun., 2019, 10, 4036.
94. N. Wu, D. Jian, M. Xiang, M. Chen, X. Lan, Z. Liao and X. Liu, Biotechnol. Appl. Biochem., 2019, 66, 597–606.
95. W. Qiang, K. Xia, Q. Zhang, J. Zeng, Y. Huang, C. Yang, M. Chen, X. Liu, X. Lan and Z. Liao, Phytochemistry, 2016, 127, 12–22.
96. T. Hashimoto, K. Nakajima, G. Ongena and Y. Yamada, Plant Physiol., 1992, 100, 836–845.
97. N. Küster, S. Rosahl and B. Dräger, Planta, 2017, 245, 355–365.
98. F. Qiu, C. Yang, L. Yuan, D. Xiang, X. Lan, M. Chen and Z. Liao, Org. Lett., 2018, 20, 7807–7810.
99. F. Qiu, J. Zeng, J. Wang, J.-P. Huang, W. Zhou, C. Yang, X. Lan, M. Chen, S.-X. Huang, G. Kai and Z. Liao, New Phytol., 2020, 225, 1906–1914.
100. D. Schlesinger, R. Davidovich Rikanati, S. Volis, A. Faigenboim, V. Vendramin, F. Cattonaro, M. Hooper, E. Oren, M. Taylor, Y. Sitrit, M. Inbar and E. Lewinsohn, Plant Sci., 2019, 283, 301–310.
101. X. Lan, J. Zeng, K. Liu, F. Zhang, G. Bai, M. Chen, Z. Liao and L. Huang, Biochem. Biophys. Res. Commun., 2018, 497, 25–31.
102. R. Lenz and M. H. Zenk, J. Biol. Chem., 1995, 270, 31091–31096.
103. X. Chen, J. M. Hagel, L. Chang, J. E. Tucker, S. A. Shiigi, Y. Yelpaala, H. Y. Chen, R. Estrada, J. Colbeck, M. Enquist-Newman, A. B. Ibáñez, G. Cottarel, G. M. Vidanes and P. J. Facchini, Nat. Chem. Biol., 2018, 14, 738–743.
104. M. Dastmalchi, L. Chang, M. A. Torres, K. K. S. Ng and P. J. Facchini, Plant J., 2018, 95, 631–647.
105. M. Dastmalchi, X. Chen, J. M. Hagel, L. Chang, R. Chen, S. Ramasamy, S. Yeaman and P. J. Facchini, Nat. Chem. Biol., 2019, 15, 384–390.
106. T. Winzer, V. Gazda, Z. He, F. Kaminski, M. Kern, T. R. Larson, Y. Li, F. Meade, R. Teodor, F. E. Vaistij, C. Walker, T. A. Bowser and I. A. Graham, Science, 2012, 336, 1704–1708.
107. T.-T. T. Dang, X. Chen and P. J. Facchini, Nat. Chem. Biol., 2015, 11, 104.
108. L. Guo, T. Winzer, X. Yang, Y. Li, Z. Ning, Z. He, R. Teodor, Y. Lu, T. A. Bowser, I. A. Graham and K. Ye, Science, 2018, 362, 343–347.
109. G. Kustatscher, P. Grabowski and J. Rappsilber, Mol. Syst. Biol., 2017, 13, 937.
110. R. S. Nett, W. Lau and E. S. Sattely, Nature, 2020, 584, 148–153.
111. Y. Ping, X. Li, B. Xu, W. Wei, W. Wei, G. Kai, Z. Zhou and Y. Xiao, ACS Synth. Biol., 2019, 8, 257–263.
112. Y. Ping, X. Li, W. You, G. Li, M. Yang, W. Wei, Z. Zhou and Y. Xiao, ACS Synth. Biol., 2019, 8, 1257–1262.
113. P. Srinivasan and C. D. Smolke, Nat. Commun., 2019, 10, 3634.
114. L. Bourgeois, M. E. Pyne and V. J. J. Martin, ACS Synth. Biol., 2018, 7, 2675–2685.
115. M. E. Pyne, K. Kevvai, P. S. Grewal, L. Narcross, B. Choi, L. Bourgeois, J. E. Dueber and V. J. J. Martin, Nat. Commun., 2020, 11, 3337.
116. A. Nakagawa, E. Matsumura, T. Koyanagi, T. Katayama, N. Kawano, K. Yoshimatsu, K. Yamamoto, H. Kumagai, F. Sato and H. Minami, Nat. Commun., 2016, 7, 10390.
117. Y. Li and C. D. Smolke, Nat. Commun., 2016, 7, 12137.
118. Y. Li, S. Li, K. Thodey, I. Trenchard, A. Cravens and C. D. Smolke, Proc. Natl. Acad. Sci. U.S.A., 2018, 115, E3922–E3931.
119. C. Ronda, J. Maury, T. Jakociunas, S. A. B. Jacobsen, S. M. Germann, S. J. Harrison, I. Borodina, J. D. Keasling, M. K. Jensen and A. T. Nielsen, Microb. Cell Factories, 2015, 14, 97.
120. J. Lian, M. HamediRad, S. Hu and H. Zhao, Nat. Commun., 2017, 8, 1688–1696.
121. J. S. Morris, M. Dastmalchi, J. Li, L. Chang, X. Chen, J. M. Hagel and P. J. Facchini, in Synthetic Biology and Metabolic Engineering in Plants and Microbes, Pt A: Metabolism in Microbes, ed. S. E. O'connor, 2016, vol. 575, pp. 143–178.
122. S.-F. Yuan, X. Yi, T. G. Johnston and H. S. Alper, Microb. Cell Factories, 2020, 19, 143–154.
123. N. Shitan, I. Bazin, K. Dan, K. Obata, K. Kigawa, K. Ueda, F. Sato, C. Forestier and K. Yazaki, Proc. Natl. Acad. Sci. U.S.A., 2003, 100, 751–756.
124. N. Shitan, F. Dalmas, K. Dan, N. Kato, K. Ueda, F. Sato, C. Forestier and K. Yazaki, Phytochemistry, 2013, 91, 109–116.
125. K. Takanashi, Y. Yamada, T. Sasaki, Y. Yamamoto, F. Sato and K. Yazaki, Phytochemistry, 2017, 138, 76–82.
126. M. Dastmalchi, L. Chang, R. Chen, L. Yu, X. Chen, J. M. Hagel and P. J. Facchini, Plant Physiol., 2019, 181, 916–933.

---

Footnote

† With equal contribution to this work.

---