Pathway engineering for high-yield production of lutein in Escherichia coli

Abstract Lutein is an industrially important carotenoid pigment, which is essential for photoprotection and photosynthesis in plants. Lutein is crucial for maintaining human health due to its protective ability from ocular diseases. However, its pathway engineering research has scarcely been performed for microbial production using heterologous hosts, such as Escherichia coli, since the engineering of multiple genes is required. These genes, which include tricky key carotenoid biosynthesis genes typically derived from plants, encode two sorts of cyclases (lycopene ε- and β-cyclase) and cytochrome P450 CYP97C. In this study, upstream genes effective for the increase in carotenoid amounts, such as isopentenyl diphosphate isomerase (IDI) gene, were integrated into the E. coli JM101 (DE3) genome. The most efficient set of the key genes (MpLCYe, MpLCYb and MpCYP97C) was selected from among the corresponding genes derived from various plant (or bacterial) species using E. coli that had accumulated carotenoid substrates. Furthermore, to optimize the production of lutein in E. coli, we introduced several sorts of plasmids that contained some of the multiple genes into the genome-inserted strain and compared lutein productivity. Finally, we achieved 11 mg/l as lutein yield using a mini jar. Here, the high-yield production of lutein was successfully performed using E. coli through approaches of pathway engineering. The findings obtained here should be a base reference for substantial lutein production with microorganisms in the future.


Introduction
Carotenoids are isoprenoids with a large number of conjugated double bonds that exhibit yellow to red coloration. The pigments are biosynthesized by all photosynthetic organisms composed of photosynthetic bacteria, cyanobacteria, algae and plants (1)(2)(3). Photosynthetic organisms must utilize carotenoids to fulfill light harvest in photosynthesis and to protect themselves from oxidative damage generated by excessive light (4,5). The photosynthetic organs of land plant chloroplasts that include higher plants and liverworts retain lutein [(3R,3 ′ R,6 ′ R)-β, ε-carotene-3,3 ′ -diol] as the predominant carotenoid (ordinarily ∼45% and ∼58% of the total carotenoids in higher plants and liverwort, respectively (6,7), which is a crucial component of light-harvesting complex II) (8).
In humans, lutein exists in the macula lutea along with zeaxanthin owing to their dietary intake. Lutein is considered to protect the human eyes from photooxidative damage by filtering high-energy blue light and preventing ocular diseases such as agerelated macular degeneration and cataract (9,10). This pigment is also present in other human tissues and is suggested to protect the skin from ultraviolet-induced damage and to reduce the risk of cardiovascular diseases such as atherosclerosis (9,11). Thus, lutein is now commercially produced for functional supplements and foods, typically by cultivating marigold and extracting the pigment from flowers. According to a BCC research report (12), lutein's global market value is about 300 million dollars, which corresponds to the maximal amount among carotenoid species for sale.
Pathway engineering, which is a metabolic engineering with heterologous microbial hosts (13), has been arduously performed to efficiently produce several commercialized carotenoids, i.e. lycopene, β-carotene, zeaxanthin and astaxanthin, using Escherichia coli, representatively (e.g. [14][15][16][17][18]. In such a pathway engineering research of these carotenoids, bacterial carotenoid biosynthesis (crt) genes have been utilized for their functional expression, e.g. crtE, crtB, crtI, crtY, crtZ and crtW genes have been used for astaxanthin production (15,19,20). Alternatively, few reports have described lutein biosynthesis in the E. coli host by the functional expression of carotenogenic genes that were derived from a plant (liverwort) (genes of lutein biosynthesis from lycopene) and bacteria (the crtE, crtB and crtI genes for lycopene biosynthesis) (21,22). This study shows the construction of a recombinant E. coli that produced lutein efficiently by engineering multiple genes derived from bacteria and a plant (liverwort) (Figure 1).

Plasmid construction
Plasmids and E. coli strains used in this study are summarized in Supplementary Figure S1 and Supplementary Tables and S2. The sequences of the plasmids constructed in this study are provided in the Supplementary information. We constructed the plasmid pAC-HIEBI that contained the Haematococcus pluvialis IDI (IPP isomerase), Pantoea ananatis crtE (GGPP syntthase), crtB (phytoene synthase) and crtI (phytoene desaturase) genes (20; accession no. ERWCRT), flanked by the tac promoter (Ptac) and the rrnB terminator (TrrnB), as described below. Ptac and TrrnB fragments were amplified by the polymerase chain reaction (PCR) and inserted into the NcoI and EcoRII sites of an E. coli vector pACYC184. Next, IDI, crtE, crtB and crtI gene fragments were amplified by PCR, sequentially ligated and inserted into the XhoI and KpnI sites between Ptac and TrrnB of the above-constructed plasmid to construct the plasmid pAC-HIEBI. Similar to pAC-HIEBI, we constructed the plasmid pRK-HIEBI. Ptac and TrrnB fragments were amplified by PCR and inserted into the HindIII and BglII sites of the E. coli vector pRK404, after which IDI, crtE, crtB and crtI gene fragments were amplified by PCR and inserted into the XhoI and KpnI sites between Ptac and TrrnB of the above-constructed plasmid to construct the plasmid pRK-HIEBI.
The coding region of the LCYe gene from Marchantia polymorpha (MpLCYe) was artificially synthesized with the codons optimized to those of E. coli K12 strain and then inserted into the plasmid pAC-HIEBI. This plasmid was named as pAC-HIEBI-MpLCYe. The accession number of the synthetic MpLCYe gene is LC603184.
The coding region of the MpLCYb gene was amplified with the plasmid pETD-MpLCYb/MpLCYe (7; accession no. AB794089) as a template and cloned into the pAC-HIEBI and pRK-HIEBI. The resulting plasmids were named as pAC-HIEBI-MpLCYb and pRK-HIEBI-MpLCYb. The transit peptide of the MpLCYb gene product was predicted by ChloroP (http://www.cbs.dtu.dk/services/ ChloroP/). The truncated form of MpLCYb gene (MpLCYb∆TP) was then amplified by PCR and cloned into plasmids similar to MpLCYb. The codon-optimized MpLCYb (MpLCYbop) (accession no. LC603183) was chemically synthesized and used to construct the plasmid pAC-HIEBI-MpLCYbop.
Artificial to E. coli K12, were chemically synthesized. These CYP97C fragments were cloned into the pUC18 between the EcoRI and BamHI sites.
To construct the plasmid CDF-MpLCYe, PCR was performed using the synthetic MpLCYe gene described above as a template. The amplified fragment was inserted into the pCDF (Merck, MO, USA), after which the MpCYP97C gene was also amplified and inserted into the CDF-MpLCYe, resulting in the plasmid CDF-MpCYP97C-MpLCYe.
The sequences of the primers used in this study are shown in Supplementary Table S3.

Construction of the genome modified E. coli
The neomycin phosphotransferase (NPT) gene was amplified by PCR with pKD13 (25), which contained flippase recognition target (FRT) sequences as a template. Ptac and TrrnB fragments were then amplified by PCR and were inserted into the SalI and BamHI sites of the NPT fragments mentioned above. This NPT fragment including Ptac and TrrnB was named as KDPT.
The manX and manZ genes' fragments were then amplified by PCR with E. coli DNA as a template and ligated. The two distinct fragments of yjfP gene were also amplified and ligated. These ligated manXZ and yjfP fragments were then cloned into the pACYC184 between the EcoRI and NcoI sites. The resulting plasmids were named pAC-manXYZ and pAC-yjfP.
The KDPT fragment was digested with EcoRV and SacI and inserted into the SmaI and SacI sites of pAC-manXYZ and pAC-yjfP, resulting in the plasmids pAC-manXYZ-KDPT and pAC-yjfP-KDPT, respectively. The IDI gene fragment was amplified by PCR and inserted into the XhoI and KpnI sites between Ptac and TrrnB of the pAC-manXYZ-KDPT, yielding the plasmid pAC-manXYZ-KDPT-HI. The Aacl-pnbA fragment was amplified by PCR with pAC-Mev/Scidi/Aacl/pnbA as a template and inserted into the plasmid pAC-yjfP-KDPT, resulting in the plasmid pAC-yjfP-KDPT-Aacl-pnbA. The DNA fragments for the recombination were also obtained by PCR using the primers shown in Supplementary  Table S2.
Either E. coli JM101(DE3) or JM101 (DE3) ∆(manXYZ)[IDI] carrying the Red helper plasmid pKD46 (25) was grown in SOB medium with ampicillin and 1 mM L-arabinose. The electroporationcompetent cells were prepared as described previously (26). The cells were mixed with PCR fragments in an ice-cold 0.1cm cuvette and electroporated at 1.8 kV (25 µF, 200 Ω; Gene Pulser Xcell, Bio-Rad, USA). After selection with kanamycin, the transformants were cultured overnight at 42 • C and tested for ampicillin sensitivity to check for loss of the helper plasmid. Colony PCR was then performed to confirm the genome recombination.
The FLP helper plasmid pCP20 (27) was introduced into the transformants to eliminate the NPT gene between FRT sequences at 30 • C. After selection with ampicillin resistance, the transformants were cultured at 42 • C overnight and tested for kanamycin and ampicillin sensitivity to check for the loss of the NPT gene and helper plasmid, respectively.

Extraction and HPLC analysis of carotenoids from E. coli cells
Each transformed E. coli was grown in 2YT medium at 37 • C until an optical density of 0.8-1.0 was achieved, induced with 0.05 mM of isopropyl β-D-thiogalactopyranoside (IPTG), and further cultured at 21 • C for 2 days. Basically, we used test tubes for the bacterial cultures. When the genome-modified E. coli was grown, 0.1% (v/v) ethyl acetoacetate (EAA) was added together with IPTG and cultured at 21 • C for 3-5 days. In the case of EAA addition, we used shake flasks for an appropriate agitation.
Extraction of carotenoids from E. coli was performed using the method described previously (7). E. coli cultures were centrifuged and cell pellets were extracted in methanol using a mixer for 5 min. Tris-HCl (50 mM, pH 7.5) (containing 1 M NaCl) was added and mixed. Then, chloroform was added to the mixture and incubated for 5 min. After centrifugation, the chloroform phase was removed and dried by centrifugal evaporation. Dried residues were resuspended with ethyl acetate and applied to the High Performance Liquid Chromatography -Photodiode Array (HPLC-PDA).
Chromatography was carried out on a Waters Alliance 2695-2996 system (Waters, Milford, MA, USA) with a column, TSKgel ODS-80Ts (4.6 × 150 mm, 5 µm; Tosoh, Tokyo, Japan), according to the method described previously (28). Briefly, the extract was eluted at a flow rate of 1.0 ml/min at 25 • C with solvent A (water-methanol, 5:95, v/v) for 5 min, followed by a linear gradient from solvent A to solvent B (tetrahydrofuranmethanol, 3:7, v/v) for 5 min, solvent B alone for 8 min and then back to solvent A. Carotenoids were identified by comparing both their retention times and absorption spectra monitored using PDA relative to those of the authentic standards. For quantitative analysis, we performed the experiments three times.

Fermentation conditions
Cells were cultured overnight in liquid Luria Broth (LB) medium at 30 • C, and then, 10 ml of the cell culture was inoculated into 1 l of modified Terrific Broth (TB) medium (per liter: 12 g Bacto Tryptone; Gibco), 24 g Bacto yeast extract, 9.4 g K 2 HPO 4 , 2.2 g KH 2 PO 4 and appropriate antibiotics (100 mg spectinomycin, 10 mg tetracycline and 30 mg chloramphenicol). Cultures were grown at 25 • C in a 3-l jar fermenter (BMJ-03P, ABLE). The pH was maintained at 7.0 by automatic addition of 28% NH 4 OH and 25% H 3 PO 4 . The agitation speed was 100 rpm. At the time of inoculation, dissolved oxygen levels were allowed to fall to 10% of O 2 saturation with a continuous air supply of 1 volume per minute. The glucose concentration was maintained at 0.4 g/l by the addition of 15% (w/v) glucose solution. 0.1 mM IPTG and 0.1% (v/v) ethyl 3-oxobutanate were then added to the culture when Optical Density at 600 nm (OD 600 ) reached 10.

Detection and quantification of chemical compounds
Glucose in the culture medium was analyzed by the mutarotaseglucose method using a Glucose CII Test Wako (Wako, Japan). To analyze carotenoid compounds in the culture medium, cultures were collected every 12 h by an autosampler (LA-11, ABLE). Cells were corrected by centrifugation at 5000 × g for 5 min and stored at −20 • C.
Cells from 0.2 ml culture medium were homogenized with 0.5 ml acetone. About 1 ml hexane/diethyl ether (1:1) was added to acetone extract and vortexed well. Also, 1 ml water was added and

Statistical analysis
For statistical analysis, we performed Student's t test.

Results and discussion
Our previous study succeeded in synthesizing lutein in the transgenic E. coli, which has three plasmids, pACHP-Lyc, pETD-MpLCYb/MpLCYe and pRSF-MpBHY/MpCYP97C (22). However, the productivity of lutein was low (∼0.1 mg/l). To increase lutein productivity, we first tried rearranging these plasmids and made the plasmids pAC-HIEBI-MpLCYb-MpLCYe-Z and pUC -MpCYP97C. The E. coli with these plasmids could produce lutein, as shown below. Here, using these plasmids, we investigated the factors that affected the productivity of lutein.

Examination of the lycopene β-cyclase
Plant lycopene β-cyclases (LCYbs) are typically bicyclases, which immediately convert lycopene to β-carotene with two β-rings via γ-carotene with one β-ring. Alternatively, a few bacteria have lycopene β-monocyclase genes, for example, cruA from C. phaeobacteroides (29) and crtYm from a unique marine bacterium P99-3 (23). To inhibit the production of zeaxanthin (both β-rings), we investigated these lycopene β-monocyclase genes in place of the plant LCYb. When we introduced the plasmid pAC-HIEBIYm into E. coli, only the β-carotene was found (Figure 2A), which indicated that CrtYm functioned as a bicyclase like the plant LCYbs in our system. When the plasmid pAC-HIEBIA was introduced, however, the γ-carotene was detected in addition to lycopene ( Figure 2B). This result showed that CruA functioned as a monocyclase as expected. Then, we constructed and transformed the plasmid either pAC-HIEBIA-MpLCYe or pAC-HIEBI-MpLCYe-A into E. coli. In both cases, no peak corresponding to α-carotene was found except for lycopene and δ-carotene with one ε-ring, suggesting that CruA cannot convert δ-carotene into α-carotene ( Figure 2C and D). From these experiments, we found that neither CrtYm nor CruA produce α-carotene with the plant lycopene ε-cyclase (LCYe). In the future, it may be required to find the β-monocyclase, which shows high substrate specificity to δ-carotene.
Next, to balance between LCYb and LCYe activities, we tried modulating the MpLCYb activity. Plant LCYb genes have the chloroplast transit peptides (cTPs). The existence of TPs affects their activities in E. coli. So, we deleted the predicted TPs from the coding region of MpLCYb (MpLCYb∆TP) and constructed the plasmid pAC-HIEBI-MpLCYb∆TP-MpLCYe-Z. In our system, it was difficult to distinguish α-carotene from β-carotene, so we compared zeinoxanthin (an α-carotene derivative) and zeaxanthin (a β-carotene derivative). In plants, the BHY and CYP97A genes function as the β-ring hydroxylase for β-carotene and zeinoxanthin, respectively. However, in E. coli, the bacterial CrtZ can hydroxylate both compounds with a higher activity than the plant genes BHY and CYP97A. Thus, we used P. ananatis crtZ for the hydroxylation of β-carotene and zeinoxanthin. In the E. coli having the plasmid pAC-HIEBI-MpLCYb∆TP-MpLCYe-Z, the ratio of zeinoxanthin to zeaxanthin (2.2 ± 0.1) was higher than that (1.5 ± 0.1) in the E. coli carrying pAC-HIEBI-MpLCYb-MpLCYe-Z ( Figure 3A and B), suggesting that the deletion of TP decreased the activity of MpLCYb. Since the lycopene was not detected in the pAC-HIEBI-MpLCYb∆TP-MpLCYe-Z carrying E. coli, it was suggested that the activity of MpLCYb∆TP was not too weak. In contrast, when we tested the codon-optimized MpLCYb (MpLCYbop), the ratio of zeinoxanthin to zeaxanthin was 0.5 ± 0.1, indicating that the activity of MpLCYbop was higher than that of MpLCYe ( Figure 3C). These results suggested that MpLCYb∆TP was most suitable to produce zeinoxanthin, the precursor of lutein.

Selection of the LCYe (lycopene ε-cyclase)
Our previous studies showed that the activity of the MpLCYb was stronger than that of the MpLCYe (7). Therefore, we tested several LCYes to find the stronger LCYe. We selected two LCYe genes from L. sativa (LsLCYe) and T. erecta (TeLCYe) in addition to MpLCYe. Most of the higher plants do not accumulate εcarotene or ε-carotene derivatives such as lactucaxanthin, probably because the activities of their LCYes are not strong compared with their LCYbs. However, lettuce (L. sativa) accumulates lactucaxanthin with two ε-rings, and the activity of LsLCYe is considered quite strong (30). Marigold (T. erecta) flower is known to be rich in lutein, suggesting that the activity of TeLCYe was relatively stronger (31).
For this purpose, we constructed the plasmids pAC-HIEBI-MpLCYb∆TP-LCYe-Z containing each LCYe gene. As a result, the peaks of zeaxanthin were predominantly detected in both cases of LsLCYe and TeLCYe ( Figure 4B and C). These results indicated that both LsLCYe and TeLCYe genes did not function in E. coli. In contrast, the peak of zeinoxanthin was dominantly detected in the case of MpLCYe ( Figure 4A). These results suggested that MpLCYe showed the highest activity among the three LCYes tested in E. coli. Therefore, we used the MpLCYe gene for further experiments.
In this study, the lettuce LCYe (LsLCYe) could synthesize εcarotene in E. coli, showing its high activity (data not shown). In contrast, the MpLCYe could synthesize only δ-carotene but not εcarotene. However, when the LsLCYe combined with MpLCYb, it did not exhibit its ability. One of the reasons is that the combination of LsLCYe and MpLCYb was not good to function together. We tried to express LsLCYb in E. coli, but its activity was significantly weaker than that of MpLCYb (data not shown). From these results, it is observed that the combination of MpLCYb and MpLCYe was useful in our system.

Screening of the CYP97C gene
Next, we tried to find the most suitable CYP97C for the effective lutein production in E. coli. We selected eight CYP97C genes in addition to MpCYP97C, which we had initially used. The eight genes are from C. reinhardtii (CrCYP97C), H. pluvialis (HpCYP97C), B. napus (BnCYP97C), C. quinoa (CqCYP97C), O. sativa (OsCYP97C), L. sativa (LsCYP97C), N. tabacum (NtCYP97C) and H. annuus (HaCYP97C). We constructed each plasmid pUC-CYP97C and transformed it with pAC-HIEBI-MpLCYb∆TP-MpLCYe-Z into E. coli. As a result, in all cases, the peaks of zeinoxanthin, which did not convert to lutein, were found ( Figure 5). This result suggested that the activities of these CYP97Cs were not sufficient to produce lutein exclusively. Nevertheless, the lowest peak of zeinoxanthin was observed in E. coli carrying pUC-MpCYP97C (Supplementary Figure S3). These results suggested that MpCYP97C was the most active CYP97C in E. coli among the nine CYP97Cs tested. Therefore, we used MpCYP97C for further experiments.

Addition of the MpLCYe gene
As described above, MpLCYe was suitable for the production of lutein along with MpLCYb. However, the activity of MpLCYe was weaker than that of MpLCYb. To balance the activities of β-cyclase and ε-cyclase, we added more than one copy of the MpLCYe gene as the plasmid CDF-MpLCYe. For the following experiments, we used the plasmid pRK-HIEBI-MpLCYb∆TP-MpLCYe-Z instead of pAC-HIEBI-MpLCYb∆TP-MpLCYe-Z. The E. coli with the latter plasmid showed almost the same carotenoid profile as that carrying the former one ( Figure 6A). The addition of MpLCYe also decreased zeaxanthin and increased zeinoxanthin, suggesting the effectiveness of improving lutein production ( Figure 6B). Then, to add the MpLCYe gene and reduce the number of plasmids, we constructed the plasmid CDF-MpCYP97C-MpLCYe. The E. coli carrying this plasmid and pRK-HIEBI-MpLCYb∆TP-MpLCYe-Z thus accumulated mainly lutein as expected ( Figure 6C). At this point, the lutein productivity was ∼1.0 mg/l.

Enhancement of the upper pathway
To increase the lutein production, we tried to enhance the upper pathway with three strategies. First, we tried integrating the IDI gene into the manXYZ region of E. coli chromosome (Supplementary Figure S2A). Second, we added crtE gene (crtE Pg ) from P. agglomerans, which showed a higher activity than P. ananatis. Studies suggested that the rate limitation of carotenoid production in E. coli depends on the activity of crtE (18). Third is the insertion of Mavalonic Acid (MVA) pathway via chromosomal integration and plasmid. Many studies have proven that the addition of MVA pathway was effective for enhancing carotenoid  or sesquiterpene production in E. coli (16,(32)(33)(34)(35)(36)(37)(38)(39)(40). Moreover, we can use EAA as a substrate for the MVA pathway by using the Aacl and pnbA genes to convert EAA to acetoacetyl-CoA ( Figure 7) (41). The Aacl and pnbA genes were integrated into the yjfP region of the chromosome of E. coli ∆(manXYZ)[IDI] (Supplementary Figure S2B). Moreover, we introduced the plasmid pAC-Mev/Scidi/Aacl/pnbA with pRK-HIEBIMpLCYb∆TP-MpLCYe-Z-E Pg and CDF-MpCYP97C-MpLCYe into E. coli. As a result of these strategies, the lutein productivity was improved to 2.6 mg/l.

Optimization of fermentation conditions for the biosynthesis of lutein
Finally, to improve the yield of lutein, the fed-batch fermentation technique was applied. Figure 8A shows the chromatogram of carotenoids extracted from E. coli cells. Many carotenoids, especially lutein and zeaxanthin, were separated by Ultra Performance Liquid Chromatography (UPLC). The results of aerobic batch and continuous cultivations of E. coli strains indicated that less acetate was accumulated (data not shown) with a higher lutein yield at 25 • C as compared to the case at 30 • C ( Figure 8B). As a result of comparing the IPTG concentrations between 0.1 mM and 0.2 mM, the ratio of zeaxanthin was extremely high in 0.2 mM IPTG (data not shown), which was not preferable for lutein synthesis. Therefore, 0.1 mM IPTG was used as an induction condition for gene expression.  Figure 8C and D). Strain 2 showed the highest carotenoid productivity and the highest lutein yield of 6.5 mg/l. Since it is known that CYP97C, a key enzyme of lutein synthesis, contains heme (42), we investigated whether the addition of FeCl 3 to the fermentation medium contributed to the increase in lutein yield. Results showed that the addition of FeCl 3 maximized the yield of lutein, and in particular, when 0.5 mM FeCl 3 was added, the productivity of lutein was 11.0 mg/l ( Figure 8E and F).

Conclusion
So far, we have produced lutein in E. coli by metabolic engineering (22); however, its productivity was low (∼0.1 mg/l; our unpublished data). Indeed, no reports have been published describing the yield of lutein biosynthesized in the metabolically engineered E. coli. In this study, we applied several strategies such as the rearrangement of the plasmid, selecting the proper genes and enhancing the upper pathway. As a result, we achieved a yield of 11 mg/l. However, the intermediates (zeinoxanthin and α-carotene) or by-products (zeaxanthin) remained. For the higher lutein productivity, it will be required to employ enzymes with higher catalytic activities and to enhance the introduced genes' expressions. In our experiments, the lutein produced by E. coli is almost present as free form, whereas most lutein in the marigold flowers or other plants are esterified, which suggests that lutein obtained from the genetically engineered E. coli could be advantageous for purification and commercialization.

Supplementary data
Supplementary data are available at SYNBIO online.

Data availability
Data and materials used in this study are available on request to the corresponding author.

Funding
'Smart Cell Project' organized by the New Energy and Industrial Technology Development Organization (NEDO) (16100920-0).