Transport engineering for improving production and secretion of valuable alkaloids in Escherichia coli

Metabolic engineering of microorganisms to produce specialized plant metabolites has been established. However, these methods are limited by low productivity and the intracellular accumulation of metabolites. Here, we aimed to use transport engineering for producing reticuline, an important intermediate in the alkaloid biosynthetic pathway. We established a reticuline-producing Escherichia coli strain and introduced a multidrug and toxic compound extrusion transporter, Arabidopsis AtDTX1, into it. AtDTX1 was selected due to its suitable expression in E. coli and its reticuline-transport activity. Expression of AtDTX1 significantly enhanced reticuline production by 11-fold; produced reticuline was secreted into the medium. AtDTX1 expression conferred high plasmid stability, and up- or downregulated genes associated with biological processes including metabolic pathways for reticuline biosynthesis, leading to a high production and secretion of reticuline. The successful application of a transporter for alkaloid production suggests that the transport engineering approach may improve the biosynthesis of specialized metabolites via metabolic engineering.


Introduction
Specialized plant secondary metabolites perform diverse functions owing to the variety in their chemical structures. Many specialized metabolites have been used as medicinal resources 1 . However, meeting the commercial demand for these metabolites is difficult owing to their low concentration in plant cells, danger of extinction of plants, and costineffective methods of chemical synthesis. To circumvent these problems, biosynthetic enzymes for generating these useful metabolites have been studied, and the genes for the corresponding enzymes have been isolated. Further progress in this field has enabled the production of useful compounds by introducing biosynthetic genes into microorganisms such as Escherichia coli (E. coli) and yeast by metabolic engineering or synthetic biology 2, 3, 4 . For example, artemisinic acid, the precursor of anti-malarial drug artemisinin 5 ; thebaine 6, 7 , an important opiate; cannabinoid 8 , a potential medical compound; colchicine 9 , a medicine used for treating gout; and tropane alkaloids 10 , which act as neurotransmitter inhibitors, have been produced in microorganisms.
Although microbial production of central and specialized metabolites is now possible, growth retardation and low productivity have been reported in certain cases, Introduction of transporters in metabolite-producing microorganisms might increase production via removal of negative feedback inhibition and enable efficient recovery from the medium. Several examples of successful transport engineering have been reported for central metabolites such as alcohols 13 , alkanes 14 , and fatty alcohols 15 .
However, for specialized metabolites, most of which are valuable as medicinal resources and are toxic to microorganisms owing to their strong biological activities, reports regarding the use of efflux pumps are scarce. This is probably due to the dearth of knowledge regarding plant transporters for specialized metabolites. However, certain researchers have isolated transporters implicated in the transport of specialized metabolites 16,17,18,19,20,21 , which include ATP-binding cassette (ABC) transporters that transport substrates using the energy obtained from ATP hydrolysis 22 ; multidrug and toxic compound extrusion (MATE) transporters, which efflux substrates as proton antiporters 23,24 ; nitrate transporter 1/peptide transporter family (NPF) members, which import substrates as proton symporters 25 ; and purine permease (PUP) members, which import substrates as proton symporters ( Supplementary Fig. 1) 26 . We have studied the transport mechanisms of alkaloids and characterized the function of several transporters using microorganisms; for example, the ABCB-type ABC transporters are responsible for berberine translocation in Coptis japonica 27,28 and four MATE transporters and one PUP transporter are required for nicotine transport in Nicotiana tabacum 29,30,31,32 . In addition, other groups have reported several transporters that transport various alkaloids 19,20 . Furthermore, using metabolic engineering, we have established E. coli that can produce reticuline 33 , an important intermediate for various benzylisoquinoline alkaloids such as morphine and berberine, via three engineered pathways, 1) an L-tyrosineoverproducing pathway via glycolysis, pentose phosphate pathway, and shikimic acid pathway; 2) a pathway producing dopamine from L-tyrosine along with the tetrahydrobiopterin (BH4)-synthesis pathway; 3) a reticuline-producing pathway from dopamine ( Fig. 1). Therefore, we hypothesized that the production and secretion of specialized metabolites such as reticuline can be enhanced by combining transport engineering with metabolic engineering in E. coli using the information available regarding alkaloid transporters.
In this study, we aimed to investigate the effect of introducing a MATE transporter, AtDTX1 from Arabidopsis, selected due to its expression and transport activity in E. coli, on reticuline production and efflux in alkaloid-producing E. coli ( Fig.   1). Our observations suggested that the combination of transport engineering and metabolic engineering is a powerful tool that can be used for enhancing the productivity of specialized plant metabolites in microorganisms. This technology would lead to high production and a stable supply of useful pharmaceutical compounds in the future.

Results
Selection of appropriate transporters for transport engineering. For applying transport engineering to alkaloid production, appropriate transporters involved in transporting specialized metabolites must be selected 16,18,20 . First, we focused on the MATE family of plant transporters known to transport specialized metabolites 20, 23 , as members of this protein family efflux substrates from the cytosol (Supplementary Fig.   1) and their microbial expression is well studied. Among these, we focused on AtDTX1 from Arabidopsis thaliana and NtJAT1 from Nicotiana tabacum ( Supplementary Fig.   2). AtDTX1 transports plant-derived alkaloids such as berberine and palmatine, whose chemical structures are relatively similar to that of reticuline ( Supplementary Fig. 3).
AtDTX1 was isolated from a functional screen using E. coli 34 , indicating that it is well expressed in E. coli. NtJAT1 is expressed well in Saccharomyces cerevisiae and localizes to the plasma membrane, where it shows substrate specificity for alkaloids such as berberine ( Supplementary Fig. 2) 29 . Therefore, we investigated the expression and reticuline transport activities of AtDTX1 and NtJAT1 in E. coli BL21(DE3).
Expression and reticuline transport activity of AtDTX1 and NtJAT1 in E. coli. We investigated whether AtDTX1 and NtJAT1 are adequately expressed and are able to efflux reticuline in E. coli. Each cDNA was subcloned into pCOLADuet-1, a low copy plasmid, and introduced in E. coli BL21(DE3). The transformants and those harboring only the vector control were incubated in Luria Bertani (LB) medium, and isopropyl βd-1-thiogalactopyranoside (IPTG) (final concentration, 1 mM) was added at OD600 = 0.6 to induce the expression of each transporter. A 37 kDa immunoreactive band was detected in the membrane preparation of E. coli transformed with AtDTX1 cDNA using His-antibody and also in Coomassie Brilliant Blue (CBB) staining. This band was absent in the membranes of the vector control cells (Fig. 2a), indicating that AtDTX1 was expressed in E. coli. In contrast, a weak band of around 37 kDa was observed for NtJAT1 using the anti-NtJAT1 antibody (Fig. 2b). Next, BL21(DE3) cells expressing AtDTX1 or NtJAT1, or control cells were incubated in LB medium containing reticuline (final concentration, 250 µM), and the intracellular reticuline content was quantitatively analyzed using ultra performance liquid chromatography-mass 9 spectrometry (UPLC-MS). Cells expressing AtDTX1 or NtJAT1 accumulated significantly less content of reticuline than the control cells (Fig. 3). These data suggested that both MATE transporters were expressed at the plasma membrane of E.
Determination of the conditions for transporter expression and cell culture. During the reticuline transport assay, we observed that the growth of NtJAT1-expressing E. coli cells was significantly inhibited. As the expression of membrane proteins tends to reduce cell growth 13 , the growth rate of E. coli cells expressing AtDTX1 or NtJAT1 was determined. In the presence of IPTG (0.1 mM), AtDTX1 expression in BL21(DE3) cells retarded growth slightly, whereas NtJAT1 significantly inhibited growth ( Supplementary Fig. 4). As growth defect is undesirable for alkaloid production, only AtDTX1 was used for further analysis.
We introduced AtDTX1 or the control vector in reticuline-producing cells.
Reticuline-producing cells were established in this study from the BL21(DE3) strain via the introduction of four vectors with 14 genes related to reticuline biosynthesis as 10 described previously 33 . Although a different combination of vectors and genes was used (Supplementary Table 1), this compound was produced from glucose, a simple carbon source (Fig. 1). Reticuline-producing cells grew more slowly than the BL21(DE3) strain, probably because these cells harbored multiple vectors and were cultured in the presence of five antibiotics (tetracycline, ampicillin, spectinomycin, chloramphenicol, and kanamycin). AtDTX1 expression did not significantly affect the proliferation of E. coli in the presence of 0.1 mM IPTG ( Supplementary Fig. 5). A previous report 35 has shown that reticuline is produced efficiently using modified LB medium containing K2HPO4, KH2PO4, and glycerol (0.4%). Hence, we decided to investigate the reticuline production of these cells using modified LB medium in the presence of 0.1 mM IPTG.

AtDTX1 significantly improved reticuline production and secretion in the medium.
We determined the effect of AtDTX1 expression on cell growth and reticuline production under the conditions described above. At OD600 = 0.6, IPTG (0.1 mM) was added to the medium to induce the biosynthetic enzymes for reticuline synthesis and AtDTX1. AtDTX1-expressing cells grew almost similarly as the control cells during the exponential phase; however, they entered the stationary phase earlier than the control cells (Fig. 4). Reticuline levels in the cells and medium were monitored quantitatively using UPLC-MS analysis ( Supplementary Fig. 6). Time course analysis (from 0 to 72 h) showed that cellular reticuline content in AtDTX1-expressing E. coli cells was significantly higher than that of the vector control ( Fig. 5a). At 12 h, the reticuline content in AtDTX1-expressing cells reached 17.7 µg/g fresh weight, which was 3.5-fold higher than that of the control cells (5.1 µg/g fresh weight). Reticuline was detected in the medium of AtDTX1-expressing cells (0.23 mg/L at 8 h) earlier than in the medium of control cells (0.13 mg/L at 12 h). Reticuline content in AtDTX1-expressing cells sharply increased from 12 h to 24 h, and at 24 h, the difference in reticuline content between the two cell lines was 11-fold (14.4 mg/L in AtDTX1-expressing cells and 1.3 mg/L in control cells); this difference was maintained for 72 h (Fig. 5b). These results indicated that AtDTX1 expression significantly enhanced reticuline production and secretion in the medium.

Plasmid stability and transcriptomic analysis of AtDTX1-expressing E. coli cells. A
previous study regarding metabolic engineering of yeast cells for the biosynthesis of artemisinic acid reported that the end product, artemisinic acid, lowered plasmid stability and the productivity of this compound 12 . Therefore, to understand the mechanism underlying the enhancement of reticuline production due to AtDTX1 expression, we assessed plasmid stability in reticuline-producing cells. At 4 h, both cells  (Table 1, Supplementary Table 2, and Supplementary Fig. 9, 10). In 14 "Global and overview maps" (Table 1), the number of genes that were categorized into metabolic pathways (01100), biosynthesis of secondary metabolites (01110), microbial metabolism in diverse environments (01120), carbon metabolism (01200), biosynthesis of amino acids (01230), and biosynthesis of cofactors (01240) was high. In contrast, the number of genes related to 2-oxocarboxylic acid metabolism (01210), fatty acid metabolism (01212), and degradation of aromatic compounds (01220) was low. In detailed categories, many genes in carbohydrate metabolism, energy metabolism, nucleotide metabolism, amino acid metabolism, and membrane transport were altered (Supplementary Table 2). Interestingly, at 12 h, the time of maximum difference of cellular reticuline production was observed; several genes of pentose phosphate pathway (00030) were upregulated (Supplementary Table 2, and Supplementary Fig. 9).
Methionine is used for the biosynthesis of S-adenosyl-L-methionine (SAM), methyl donor for reticuline biosynthesis via three methyl transferases, that is, 4'OMT, CNMT, and 6OMT ( Fig. 1, 8). Some methyl transferases, such as 4'OMT and 6OMT, have been reported to be inhibited by end-product or related compounds-berberine or norreticuline 36,37 . Therefore, reticuline efflux from the cytosol via AtDTX1 may have relieved the negative feedback on methyltransferases and methionine biosynthesis was enhanced to supply sufficient amount of SAM.
These data suggested that AtDTX1 expression conferred high plasmid stability in the cells, significantly affected cellular processes, and induced the biosynthetic flow in the cell (Fig. 8). Overall, these changes led to the production of high amounts of reticuline and its efflux into the medium.

Discussion
Reports regarding the practical application of transporters for the production of specialized metabolites are scarce, as little has been known regarding plant transporters for specialized metabolites. In this study, we aimed to utilize transport engineering for the synthesis and efflux of plant specialized metabolites from E. coli. To the best of our knowledge, this is the first study in which a plant alkaloid transporter was used for microbial production, with the product subsequently effluxed into the medium.
Selection of an appropriate transporter is important for transport engineering. In this study, we focused on the MATE family. This is because NPF and PUP transport substrates inward as proton co-transport, which is not appropriate for studying product efflux from the cytosol to the medium ( Supplementary Fig. 1). Some plant ABC transporters show efflux activity. However, most plant ABC transporters, especially those involved in alkaloid transport (ABCB-type), function as a single polypeptide consisting of two transmembrane domains (TMDs) and two nucleotide binding domains (NBDs) (called full-size) 22  Although many plant transporters of specialized metabolites have recently been identified and characterized 16,18,19 , knowledge regarding their substrate specificities and expression in microorganisms is still limited. Hence, whether a selected transporter can transport a product in microorganisms should be determined, and novel transporters for specific products have to be identified in some cases. In addition, for utilizing plant transporters for the microbial production of specialized metabolites, transporter expression and activity in the host microorganism should be assessed. In this study, AtDTX1 was selected based on the results of a previous study 34 demonstrating the function and alkaloid transport profiles in E. coli. AtDTX1 was expressed well in E. coli and showed transport activity for reticuline ( Fig. 2 and 3). In contrast, NtJAT1 expression was weak (Fig. 2) and hindered cell growth significantly ( Supplementary   Fig. 4), and was hence not used further in this study. However, NtJAT1 is best suited for the transport engineering of yeast cells for producing alkaloids, as its expression and substrate specificity for alkaloids have been characterized in yeast cells 29 .
The expression level and time required for the induction of the transporter should be carefully determined. Overexpression of the transporter, similar to that of other membrane proteins, inhibits cell growth 13 . Thus, for n-butanol production using transport engineering, transporter expression was optimized to balance its toxicity while ensuring efficient production 13 . In this study, we suppressed the basal expression level of the transporter using a low copy plasmid, pCOLADuet-1. Nevertheless, significant growth inhibition was observed when expression was induced by high concentrations of IPTG (1 mM) at the lag phase of E. coli (OD600 = 0.1) (data not shown). Therefore, low concentration of IPTG (0.1 mM) was added after the cells grew to some extent (OD600 = 0.6). This experimental condition negligibly affected cell growth in reticuline-producing cells (Fig. 4) and resulted in high production and secretion of reticuline (Fig. 5).
Several studies have reported using transport engineering for the production of were also highly regulated, suggesting diverse effects of AtDTX1 expression in cellular function. It is noted that metabolic pathways relatively related to reticuline biosynthesis were altered. In addition to the pentose phosphate pathway, the metabolism of several amino acids other than methionine was affected; for instance, the genes for the biosynthesis of valine, leucine, and isoleucine, were upregulated at 12 h. The downregulation of the gene for tryptophan biosynthesis at 24 h might lead to higher production of tyrosine ( Supplementary Fig. 10). AtDTX1-dependent reticuline efflux might cause the decrease of biosynthetic intermediates like G3P, R5P, E4P, and tyrosine, and SAM, methyl donor for methyltransferase, which possibly resulted in the upregulation of related cellular metabolism to compensate for the supply of those compounds (Fig. 1). These alterations would have led to not only secretion to the medium but also high production in the cells (Fig. 5).
In contrast, AtDTX1 expression seems to induce stress in the cells. In the GO analysis, many genes categorized into response to stimulus (GO:0050896), detoxification (GO:0098754), antioxidant activity (GO:0016209) were upregulated (Fig.   7). Analysis of each gene revealed that genes, such as those encoding the universal stress protein UspG (locus tag: ECD_RS02910, 4.6-fold at 24 h) and stress-induced protein YchH (ECD_RS06300, 33-fold at 24 h) were highly upregulated (Supplementary Data 6). As reticuline did not significantly inhibit the growth of E. coli ( Supplementary Fig. 12), these stress-related genes might be induced by AtDTX1 expression. GO analysis also suggested that other genes, for example, those encoding Surprisingly, plasmid stability was high in AtDTX1-expressing cells (Fig. 6), suggesting that the percentage of reticuline-producing cells was high in the culture medium. These findings indicated that the expression of the transporter AtDTX1, in this case, induced diverse changes in the cells, and the integrated effect of these alterations increased the production of the desired compound.
Application of metabolic engineering to efflux transporters enables the high production and efficient recovery of valuable specialized metabolites. Some synthesized metabolites accumulate in vivo; therefore, procedures for the extraction and purification of metabolites are necessary, which renders some microbial productions commercially unfeasible. Combining transport engineering with metabolic engineering not only improves the production of valuable metabolites, but is also beneficial for the efficient 23 recovery of the products. Although the knowledge about transporters for transport engineering is still limited and the identification of the optimal transporters for the desired metabolites would be necessary, further progress in this field will enhance the production of medicinal resources from plants.
In conclusion, we successfully developed the first E. coli transport engineering platform that enables high production and secretion of a valuable alkaloid. The results of the present study are of considerable significance, as E. coli is a standard microorganism for industrial-scale production of plant medicinal resources 2, 3, 7 . The combined platform of metabolic engineering and transporter engineering described here will provide opportunities for low-cost production of various valuable alkaloids. The use of this technology for rapid mass production of useful plant metabolites will contribute to the health and welfare of society.

24
Chemicals. (S)-reticuline was purified as described previously 33 and was used as a standard. The other chemicals used in this study were purchased from FujiFilm Wako Pure Chemical Corporation (Osaka, Japan) or Nacalai Tesque (Kyoto, Japan). seconds of rest on ice were performed. The mixture was centrifuged at 3,000 ´ g for 15 min, and the supernatant was centrifuged at 20,000 ´ g for 20 min to pellet the membrane proteins. The membrane proteins were then denatured, subjected to electrophoresis on 10% sodium dodecyl sulfate-polyacrylamide gel, and transferred to a Immobilon polyvinylidene difluoride membrane (Millipore, Tokyo, Japan). The membrane was treated with BlockingOne (Nacalai Tesque) and incubated with anti-His monoclonal antibody (MBL, Japan) against His-AtDTX1 or anti-NtJAT1 29 against NtJAT1. The immunoreactive band was visualized using Chemi-Lumi One Super (Nacalai Tesque).

Reticuline transport of AtDTX1 and NtJAT1 in E. coli cells. BL21(DE3) cells
harboring pCOLADuet-1, pCOLADuet-1_AtDTX1, or pCOLADuet-1_NtJAT1 were cultured as described above. IPTG (1 mM) was added when OD600 of the cultures reached 0.6, followed by incubation for 3 h. The cells were harvested, suspended in LB medium containing 50 mg/L kanamycin, 1 mM IPTG, and 250 µM reticuline at OD600 = 0.7, and incubated at 30°C with shaking at 200 rpm for 6 h. Next, the cells were harvested and washed with LB medium, and the reticuline content in the cells was quantified as described below. Statistical analysis. Student's t-test (two-tailed) was used to determine significant differences compared to the control cells in reticuline production, and plasmid stability analyses. Multiple comparisons were conducted using repeated analysis of variance with Bonferroni test in reticuline transport.

Data availability statement
Data supporting the findings of this work are available within the paper and its Supplementary Information files. RNA-seq data are available in the DDBJ Sequenced Read Archive under the accession number DRA011247. All relevant data presented in this paper are available from the corresponding author upon request.  Reticuline is synthesized from simple carbon sources, i.e., glucose or sucrose, via the sequential action of enzymes dTH2, DODC, MAO, NCS, 6OMT, CNMT, and 4¢OMT.
As an efflux transporter of reticuline, AtDTX1 was introduced in this reticuline-    Yamada et al., Fig. 7